EPA 635/R-03/007
www.epa.gov/iris
?/EPA
TOXICOLOGICAL REVIEW
OF
DICHLOROACETIC ACID
(CAS No. 79-43-6)
In Support of Summary Information on the
Integrated Risk Information System (IRIS)
August 2003
U.S. Environmental Protection Agency
Washington, DC
-------
DISCLAIMER
This document has been reviewed in accordance with U.S. Environmental Protection
Agency policy and approved for publication. Mention of trade names or commercial products
does not constitute endorsement or recommendation for use. Note: This document may undergo
revisions in the future. The most up-to-date version will be made available electronically via the
IRIS Home Page at http://www.epa.gov/iris.
11
-------
TABLE OF CONTENTS
LIST OF TABLES v
LIST OF FIGURES vi
FOREWORD vii
AUTHORS, CONTRIBUTORS, AND REVIEWERS viii
1. INTRODUCTION 1
2. CHEMICAL AND PHYSICAL INFORMATION RELEVANT TO ASSESSMENTS 3
3. TOXICOKINETICS 4
3.1. ABSORPTION 4
3.2. DISTRIBUTION 4
3.3. METABOLISM 4
3.3.1. Mechanistic Metabolic Considerations 11
3.4. ELIMINATION 14
3.5. PHYSIOLOGICALLY-BASED TOXICOKINETIC MODELS 15
4. HAZARD IDENTIFICATION 17
4.1. STUDIES IN HUMANS 17
4.2. STUDIES IN ANIMALS 21
4.2.1. Acute and Subchronic Studies 21
4.2.2. Chronic Studies and Cancer Bioassays 30
4.3. REPRODUCTIVE/DEVELOPMENTAL STUDIES 37
4.4. OTHER STUDIES 42
4.4.1. Mechanistic Studies 42
4.4.2. Genotoxicity Studies 51
4.5. SYNTHESIS AND EVALUATION OF MAJORNONCANCER EFFECTS .... 58
4.5.1. Metabolic Alterations 58
4.5.2. Hepatic Toxicity 59
4.5.3. Reproductive/Developmental Toxicity 61
4.5.4. Neurotoxicity 61
4.6. SYNTHESIS AND EVALUATION OF CANCER EFFECTS AND MODE OF
ACTION 62
4.6.1. Data Summary 62
4.6.2. Potential Mode of Carcinogenicity 65
4.6.3. Cancer Characterization 71
4.7. SUSCEPTIBLE POPULATIONS AND LIFE STAGES 72
5. DOSE-RESPONSE ASSESSMENTS 74
5.1. ORAL REFERENCE DOSE (RfD) 74
5.1.1. Methods of Analysis 74
5.1.2. NOAEL/LOAEL Approach 74
iii
-------
5.1.3. Benchmark Dose Approach 80
5.1.4. Summary of Oral RfD Derivation 88
5.2. INHALATION REFERENCE CONCENTRATION (RFC) 88
5.3. CANCER ASSESSMENT 89
5.3.1. Choice of Principal Studies and Cancer Endpoints 89
5.3.2. Dose-Response Data 89
5.3.3. Dose Conversion 90
5.3.4. Dose-Response Characterization in the Range of Observation 90
5.3.5. Selection of a Dose-Response Model 91
5.3.6. Extrapolation to Doses Below the Range of Observation 92
5.3.7. Confidence in the Cancer Assessment 93
6. MAJOR CONCLUSIONS IN THE CHARACTERIZATION
OF HAZARD AND DOSE-RESPONSE 95
6.1. HUMAN HAZARD POTENTIAL 95
6.2. DOSE-RESPONSE 100
7. REFERENCES 102
APPENDIX A: RESPONSE TO PEER REVIEW SUMMARY DOCUMENT
APPENDIX B: SUMMARY OF STUDIES ON DC A TOXICITY AND APPLICABILITY FOR
BMD ANALYSIS
APPENDIX C: BENCHMARK DOSE-RESPONSE FOR NONCANCER ENDPOINTS
APPENDIX D: BENCHMARK DOSE-RESPONSE FOR CANCER ENDPOINTS
DEANGELO ET AL, 1999 (5 DOSES)
APPENDIX E: BENCHMARK DOSE-RESPONSE FOR CANCER ENDPOINTS
DEANGELO ET AL., 1999 (4 AND 6 DOSES)
IV
-------
LIST OF TABLES
Table 3-1. Enzyme kinetics for GSTZ: kast/Km ratios (M^sec"1) 13
Table 4-1. Biomarkers of tissue DCA exposure:
incidence (%) of altered hepatic history 44
Table 4-2. Frequency of spontaneous and DCA-induced mutations of codon 61
in exon 2 of the H-ras oncogene mutations in E6C3Fl mice 48
Table 4-3. Summary of in vitro genotoxicity tests 53
Table 4-4. Summary of in vivo genotoxicity tests 55
Table 4-5. Drinking water exposures, cancer response and simulated internal dose metrics 64
Table 5-1. Summary of noncancer studies considered for benchmark modeling 76
Table 5-2. Criteria for selecting studies appropriate for BMD modeling 82
Table 5-3. Cardiovascular defects induced by DCA 83
Table 5-4. Effects of DCA on fetal body weight 84
Table 5-5. Liver weight data set 85
Table 5-6. Male reproductive data sets used for BMD modeling 85
Table 5-7. Summary of noncancer BMD modeling results 87
Table 5-8. Cancer dose-response data evaluated using BMD modeling: male mice 89
Table 5-9. Summary of cancer BMD modeling results 92
-------
LIST OF FIGURES
Figure 3-1. Metabolism of DCA 7
Figure 5-1. Summary of noncancer effects of DCA 77
Figure 5-2. Multistage dose-response model fit for combined hepatocarcinoma
and adenoma incidence in male mice 93
VI
-------
FOREWORD
The purpose of this Toxicological Review is to provide scientific support and rationale
for the hazard and dose-response assessment in IRIS pertaining to chronic exposure to
dichloroacetic acid (DCA). It is not intended to be a comprehensive treatise on the chemical or
toxicological nature of DCA.
In Section 6, EPA has characterized its overall confidence in the quantitative and
qualitative aspects of hazard and dose response. Matters considered in this characterization
include knowledge gaps, uncertainties, quality of data, and scientific controversies. This
characterization is presented in an effort to make apparent the limitations of the assessment and
to aid and guide the risk assessor in the ensuing steps of the risk assessment process.
For other general information about this assessment or other questions relating to IRIS,
the reader is referred to EPA's IRIS Hotline at 202-566-1676.
vn
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS
CHEMICAL MANAGER
Joyce Morrissey Donohue, Ph.D.
Health and Ecological Criteria Division
Office of Science and Technology
Office of Water
U.S. Environmental Protection Agency
Washington, DC
AUTHORS
Joyce Morrissey Donohue, Ph.D.
Health and Ecological Criteria Division
Office of Science and Technology
Office of Water
U.S. Environmental Protection Agency
Washington, DC
Hend Galal-Gorchev, Ph.D.
Health and Ecological Criteria Division
Office of Science and Technology
Office of Water
U.S. Environmental Protection Agency
Washington, DC
William Brattin, Ph.D.
IS SI Consulting
Denver, Colorado
John J. Liccione, Ph.D.
Sciences International, Inc.
Alexandria, VA
KaraB. Altshuler, Ph.D.
ICF Consulting
Fairfax, VA
REVIEWERS
This document and summary information on IRIS have received peer review both by
EPA scientists and by independent scientists external to EPA. Subsequent to external review
and incorporation of comments, this assessment has undergone an Agency-wide review process
viii
-------
whereby the IRIS Program Director has achieved a consensus approval among the Office of
Research and Development; Office of Air and Radiation; Office of Prevention, Pesticides, and
Toxic Substances; Office of Solid Waste and Emergency Response; Office of Water; Office of
Policy, Economics, and Innovation; Office of Children's Health Protection; Office of
Environmental Information; and the Regional Offices.
INTERNAL EPA REVIEWERS
Jeff Gift
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Karen Hogan
National Center for Environmental Assessment
U.S. Environmental Protection Agency
John Lipscomb, Ph.D.
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Carolyn Smallwood, Ph.D.
National Center for Environmental Assessment
U.S. Environmental Protection Agency
EXTERNAL PEER REVIEWERS
Richard Bull, Ph.D.
Department of Pharmacology/Toxicology
Washington State University Tricities
Denise Robinson, Ph.D.
International Life Sciences Institute
Lauren Zeise, Ph.D.
California Environmental Protection Agency
Summaries of the external peer reviewers' comments and the disposition of their
recommendations are in Appendix A.
IX
-------
1. INTRODUCTION
This document presents background and justification for the hazard and dose-response
assessment summaries in EPA's Integrated Risk Information System (IRIS). IRIS Summaries
may include an oral reference dose (RfD), inhalation reference concentration (RfC) and a
carcinogenicity assessment.
The RfD and RfC provide quantitative information for noncancer dose-response
assessments. The RfD is based on the assumption that thresholds exist for certain toxic effects
such as cellular necrosis, but may not exist for other toxic effects such as some carcinogenic
responses. It is expressed in units of mg/kg-day. In general, the RfD is an estimate (with
uncertainty spanning perhaps an order of magnitude) of a daily exposure to the human
population (including sensitive subgroups) that is likely to be without an appreciable risk of
deleterious noncancer effects during a lifetime. The inhalation RfC is analogous to the oral RfD,
but provides a continuous inhalation exposure estimate. The inhalation RfC considers toxic
effects for both the respiratory system (portal-of-entry) and for effects peripheral to the
respiratory system (extrarespiratory or systemic effects). It is generally expressed in units of
mg/m3.
The carcinogenicity assessment provides information on the carcinogenic hazard
potential of the substance in question and quantitative estimates of risk from oral exposure and
inhalation exposure. The information includes a weight-of-evidence judgment of the likelihood
that the agent is a human carcinogen and the conditions under which the carcinogenic effects
may be expressed. Quantitative risk estimates are presented in three ways. The slope factor is
the result of application of a low-dose extrapolation procedure and is presented as the risk per
mg/kg-day. The unit risk is the quantitative estimate in terms of either risk per |ig/L drinking
water or risk per |ig/m3 air breathed. Another form in which risk is presented is a drinking water
or air concentration providing cancer risks of 1 in 10,000; 1 in 100,000; or 1 in 1,000,000.
Development of these hazard identification and dose-response assessments for
dichloroacetic acid (DCA) has followed the general guidelines for risk assessment as set forth by
the National Research Council (1983). EPA guidelines that were used in the development of this
assessment may include the following: Guidelines for Carcinogen Risk Assessment (U.S. EPA,
1986a), Guidelines for the Health Risk Assessment of Chemical Mixtures (U.S. EPA, 1986b),
Guidelines for Mutagenicity Risk Assessment (U.S. EPA, 1986c), Guidelines for Developmental
1
-------
Toxicity Risk Assessment (U.S. EPA, 199 la), Guidelines for Reproductive Toxicity Risk
Assessment (U.S. EPA, 1996b), Guidelines for Neurotoxicity Risk Assessment (U.S. EPA,
1998a), Draft Revised and Draft Final Guidelines for Carcinogen Assessment (U.S. EPA, 1999,
2003), Recommendations for and Documentation of Biological Values for Use in Risk
Assessment (U.S. EPA, 1988), (proposed) Interim Policy for Particle Size and Limit
Concentration Issues in Inhalation Toxicity (U.S. EPA, 1994b), Methods for Derivation of
Inhalation Reference Concentrations and Application of Inhalation Dosimetry (U.S. EPA,
1994c), Use of the Benchmark Dose Approach in Health Risk Assessment (U.S. EPA, 1995),
Science Policy Council Handbook: Peer Review (U.S. EPA, 1998b, 2000a), Science Policy
Council Handbook: Risk Characterization (U.S. EPA, 2000b), Benchmark Dose Technical
Guidance Document (U.S. EPA, 2000c) and Supplementary Guidance for Conducting Health
Risk Assessment of Chemical Mixtures (U.S. EPA, 2000d).
The literature search strategy employed for this compound was based on the CASRN and
at least one common name. At a minimum, the following databases were searched: RTECS,
HSDB, TSCATS, CCRIS, GENE-TOX, DART/ETIC, EMIC, TOXLINE, CANCERLIT, and
MEDLINE. Any pertinent scientific information submitted by the public to the IRIS Submission
Desk was also considered in the development of this document. The relevant literature was
reviewed through January 2003.
-------
2. CHEMICAL AND PHYSICAL INFORMATION RELEVANT TO
ASSESSMENTS
Dichloroacetic acid (DCA) is a colorless to slightly yellow liquid, with a pungent acid-
like odor. In aqueous solutions, DCA and its conjugate base, dichloroacetate, exist as an
equilibrium mixture, the proportions of each depending primarily on the pH of the solution.
With a pKa of 1.48 at 25°C, DCA occurs almost exclusively in the ionized form in normal
drinking water (pH range 6-9). Other selected chemical and physical properties for this chemical
are listed below (Merck Index, 1996; Lewis, 1997).
CASRN 79-43-6
Empirical Formula CHC12COOH
Molecular Weight 128.94
Melting Point 13.5°C
Boiling Point 193-194°C
Density 1.5724 g/mL at 13°C
Physical State Liquid
Solubility Soluble in water, alcohol and ether
Specific Gravity 1.563 at 20/4 °C
Vapor Pressure 0.19 mbar (19 Pa) at 20°C
DCA has a very low vapor pressure and is not expected to volatilize from drinking water
or contaminated environmental media to any appreciable extent. Therefore, inhalation exposure
from volatilized DCA is negligible and is not considered in this document.
-------
3. TOXICOKINETICS
3.1. ABSORPTION
Studies in humans and animals indicate that DCA is readily absorbed by the
gastrointestinal tract (Lin et al., 1993; Larson and Bull, 1992; Stacpoole et al., 1998a).
Following oral administration of radiolabeled DCA to rats and mice, only about 1-2% of the
label was found in the feces, indicating almost complete gastrointestinal absorption (Lin et al.,
1993; Larson and Bull, 1992). In fasted human subjects, peak plasma DCA concentration occurs
within 15 to 30 minutes of oral dosing (Stacpoole et al., 1998a).
3.2. DISTRIBUTION
Lin et al. (1993) and James et al. (1998) investigated the distribution of absorbed DCA to
internal tissues in rats. Lin et al. (1993) reported that 48 hours after oral DCA administration
(gavage), 21 to 36% of the tracer [14C] was recovered from tissues, with the precise amount
dependant on the dose as well as the form of labeled DCA that was administered. The majority
of tracer was found in the liver, muscle, skin, blood and intestines. At 24 hours, all of the other
tissues combined (kidney, adipose, stomach, testis, lung, spleen, heart, brain, and bladder)
contained 10 to 15% of the label (James et al., 1998) while at 48 hours these tissues contained
-1-2% of the original dose given (Lin et al., 1993).
3.3. METABOLISM
The primary metabolic pathway for DCA involves oxidative dechlorination to form
glyoxylate (Larson and Bull, 1992). This reaction, once thought to be microsomal Cytochrome
P-450 mediated, has now been shown to be NADPH- and GSH-dependent and occurs
predominantly in the cytosol (Lipscomb et al., 1995; Cornett et al., 1997; Stacpoole et al., 1998a;
Board et al., 1997). Recent work by Tong et al. (1998 a; 1998 b) has identified a rat liver
cytosolic enzyme, glutathione-S-transferase Zeta (GST Zeta), that catalyzes the conversion of
DCA to glyoxylate. This enzyme is considered the rat ortholog of human GST zeta.
Data on DCA metabolism in humans are available because DCA has been used
experimentally in the therapeutic treatment of several metabolic disorders. The data obtained
support the hypothesis that DCA metabolism is similar in both humans and rodents (Stacpoole et
-------
al., 1998a). The occurrence of oxalic acid in the urine of DCA-treated patients indicates that
DCA is oxidatively dechlorinated to glyoxylate, which is then converted to oxalate. In one child
with congenital lactic acidosis, monochloroacetic acid was present in plasma in addition to
oxalate and glyoxylate during the first four hours after the initial dose. Monochloroacetic acid
concentrations were then below detection for the remainder of the observation period. Initially,
the concentration of monochloroacetic acid in plasma exceeded that for glyoxylate, but not
oxalate (Stacpoole et al., 1998a). These data indicate that in at least some individuals, the
reductive dechlorination pathway can occur initially after DCA administration, but continued
DCA metabolism occurs through the oxidative dechlorination pathway.
GSTZ appears to be identical to maleylacetoacetate isomerase (MAAI), the enzyme in
the pathway for tyrosine catabolism that converts the cis double bond in maleylacetoacetate
(MAA) to the trans double bond in fumarylacetoacetate, using GSH as a cofactor (Fernandez-
Canon and Penalva, 1998). GSTZ/MAAI appears to have an active site geometry that is highly-
conserved across species and is sufficiently-plastic that it can participate in cis/trans
isomerization reactions and dehalogenation of molecules as diverse as DCA and
pentachlorophenol. It also has moderate GSH peroxidase activity (Anandarajah et al., 2000;
Polekhina et al., 2001; Sheehan et al., 2001). In humans GSTZ is expressed mostly in the liver
followed by kidney, and skeletal muscle. It is also expressed in the placenta, heart, pancreas,
mammary tissues, seminal glands, and fetal liver (Fernandez-Canon et al., 1999; Polekhina et al.,
2000). The complete pathway for tyrosine catabolism is found only in the liver and kidney. The
presence of the enzyme in other tissues suggests that it has functions other than the isomerization
of MAA (Fernandez-Canon et al., 1999).
There are species and age-related differences in the activity of GSTZ. The relative rate
of DCA transformation in mouse hepatic cytosol was greater than in rat hepatic cytosol which in
turn was greater than in human hepatic cytosol (Tong et al., 1998a). The Km and Vmax/Km values
for DCA in mice were 81.9 ± 5.6 jiM and 52.9 ± 2.46 (xlO"3) respectively, those in rats were 70.1
± 5.3 |iM and 32.4 ± 4.87 (xlO'3), and those in humans were 47.3 ± 6.7 |iM and 8.25 ± 1.37
(xlO"3) (Tong et al., 1998a). Km and Vmax/Km values for DCA transformation in naive young
mice were 108 jiM ± 16 and 6.72 ml/hr/mg, while those for aged mice were 56.1 jiM ± 14.2 and
8.92 ml/hr/mg, demonstrating a difference in the response of the enzyme in the young versus the
older mice (Schultz et al., 2002).
-------
Among humans there are known polymorphisms in GSTZ which may account for
differences in the ability to metabolize DC A and other halogenated compounds (Sheehan et al.,
2001). The polymorphisms result from A/G transitions at nucleotides 94 and 124 of the coding
region and T/C transitions at positions 23 and 245 (Blackburn et al., 2000, 2001). The GSTZ
variants are the products of the different combinations of the bases at the variant positions and
were designated GSTZla-la, GSTZlb-lb, GSTZlc-lc, GSTZld-ld, and GSTZle-le (Blackburn
et al., 2000, 2001; Tzeng et al., 2000). Analysis of a Caucasian (unselected, European Australian
blood donors) population (141 subjects: 68 females and 73 males, ages 16 to 69) by Blackburn et
al. (2000) showed that the first three allele variants were present with frequencies of 0.09, 0.28,
and 0.63, respectively. Blackburn et al. (2001) reported the results of an analysis for five
variants in a similar population of 128 subjects where the variant distribution was 0.086, 0.285,
0.473, 0.156 and 0 for GSTZla-la, GSTZlb-lb, GSTZlc-lc, GSTZld-ld, and GSTZle-le,
respectively. GSTZla-la has been demonstrated to have different catalytic properties toward
DCA than the other variants, including a 4-5-fold higher activity. However, excluding the GSTZ
le-le variant, the most active human GSTZ variants toward the catabolism of DCA appeared at
the lowest frequency in the populations studied by Blackburn et al. (2000, 2001). The most
common variant, GSTZlc-lc, had the highest activity toward the isomerization of MAA (MAA
is chemically too unstable to be used in the enzyme studies) using maleylacetone as a surrogate.
Glyoxylate formed from the metabolism of DCA may be routed though several different
pathways (Figure 3-1). Transamination by peroxisomal alanine-glyoxylate transaminase forms
glycine, which can be incorporated into proteins, used in the synthesis of serine, or degraded
releasing carbon dioxide. Conversion to oxalate occurs via a (S)-2-hydroxyacid dehydrogenase
such as lactate dehydrogenase. Glyoxylate can also be converted to glycolate by glyoxylate
reductase (Michal, 1999).
-------
of Dicloroacetic Acid
H o CI
1 # \
C/"* .uttfii.^.-.—i-ffi* .rfC. ..niHimi /*! if* i ii if"*
o ^ ^ ^ oi v v
I \ i
Ho- H
o
^
\
o-
Monochloroacetate Dichloroacetate
j ,
1
1
^
GSH
?
i '
o 7
P j%
v **#
/ \
NADH+H*
——Xgsa^
°v
NAD N>
p,
H OH
Glyoxylate
s*
X
HO O-
Glycolate
O,
Alanine
Pyruvate
Thiodiacetate
f
C02+NH4+
5,10-Methylene THF
H ...
HO—C— C— C
I I \
H H O"
Serine
Figure 3-1. Metabolism of DCA
Source: Adapted from Michal, 1999.
7
-------
There may be other metabolic pathways for DCA. Oxalate, glycine, CO2, glycolate,
monochloroacetic acid and thiodiacetic acid have been shown to be metabolites of DCA in
rodents, although the relative amount of each seems to be species-specific (Larson and Bull,
1992; Lin et al., 1993; Gonzalez-Leon and Bull, 1996; Xu et al., 1995). While urinary
metabolites (glyoxylate, glycolate, oxalate, monochloracetic acid, and thiodiacetic acid) account
for -12-30% of the administered dose in rats and mice, CO2 excretion may differ between these
two species. Larson and Bull (1992) report that exhaled CO2 generated from radiolabeled DCA
was approximately 24 to 30% of a single administered dose in rats, but represented only 2% of
the same dose in mice. However, a later study (Xu et al., 1995) indicates that approximately
45% of a single administered dose of DCA in mice was exhaled as CO2 in the first 24 hours after
dosing. A problem associated with the recovery of label in the Larson and Bull (1992) study
may have resulted in the lower value in mice. In both species the nonchlorinated acids were the
primary metabolites detected in urine. Thiodiacetic acid concentrations were much greater than
monochloroacetic acid, which was present in only trace quantities (Larson and Bull, 1992).
To account for the production of metabolites that are not metabolically linked to
glyoxylate, Stacpoole et al. (1998a) and Larson and Bull (1992) proposed reductive
dechlorination of DC A yielding monochloroacetic acid as an alternate metabolic pathway. The
monochloroacetic acid is converted to thiodiacetic acid via glutathione conjugation. While it has
been speculated that this pathway might involve the formation of free radicals (Larson and Bull,
1992), it has not been investigated (Stacpoole et al., 1998a).
DCA metabolites (e.g., glyoxylate) can enter intermediary metabolism, and the carbon
atoms originally present in DCA can become incorporated into endogenous proteins and other
biomolecules. Stevens et al. (1992) investigated this possibility and reported that a substantial
portion of a radiolabeled dose of DCA was not excreted, but was oxidatively metabolized into
glycine and incorporated into serum albumin. These data are consistent with the results from a
study by Larson and Bull (1992), who reported that within 3 hours of dosing mice and rats, high
concentrations of radiolabel were incorporated into various plasma proteins.
One of the unique features of DCA toxicokinetics is the ability of the compound to
inhibit its own metabolism. Lin et al. (1993) administered single oral doses of 100 and 282
mg/kg to rats, and measured urinary output of the parent compound. At the low-dose, only 1-2%
of the administered DCA appeared in the urine. However, with the 282 mg/kg dose, 20% of the
parent compound was excreted, suggesting that the metabolic capacity of rats was exceeded in
-------
the high-dose range. Similar effects have been reported in healthy human volunteers treated
with DCA, indicating that inhibition also occurs in humans (Stacpoole et al., 1998a).
Plasma clearance rates for DCA can vary substantially among species. Following a
single oral dose of approximately 50 mg/kg in humans, the plasma half-life of DCA was 0.5-2
hours, with less than 1% of the parent compound excreted in the urine (Lukas et al., 1980; Curry
et al., 1991). This was similar to results reported in mice and rats (Lin et al., 1993; Larson and
Bull, 1992; James et al., 1998). Although Lukas et al. (1980) found that rats cleared DCA from
their blood with a half-life of 2.0-4.4 hours following intravenous injection of 100 mg/kg,
clearance of the radiolabel from the ingested DCA metabolites took a much longer time (21-36
hours). Dogs clear DCA from their blood at a slower rate. Following injection of 100 mg/kg
DCA, Lukas et al. (1980) found that initial blood levels of DCA in two dogs were approximately
double those in three treated rats. DCA concentrations fell from their peak levels to half the
concentration at some point between 6 and 24 hours leading to an estimated half-life of 17.1 to
24.6 hours. Sampling from one dog was discontinued at 24 hours because of collapsing veins.
However, at 48 hours, the rats and the remaining dog had approximately the same percent of the
initial DCA concentration as residual in plasma (10-20% for the three rats and 10% for the dog).
Prior exposure to DCA significantly inhibits its metabolism (Curry et al., 1991;
Gonzalez-Leon et al., 1997a, b, 1999; Lukas et al., 1980; Schultz et al., 2002). Studying the
plasma half-life of DCA in human volunteers, Curry et al. (1985) found that the mean half-life of
DCA increased from 63.3 minutes to an average of 374 minutes following the fifth in a series of
50 mg/kg doses administered intravenously at 2-hour intervals. In another study (Stacpoole et
al., 1998a) of healthy adults in which 25 mg/kg DCA was administered daily for five days, the
half-life increased about eightfold on the fifth day when compared to the first (1.09 ± 0.45 hr vs.
8.03 ± 5.62 hr). The most likely basis for the decrease in DCA clearance observed with repeated-
or high-dose exposure is the inactivation of one or more of the enzymes involved in its
metabolism.
Recent work by Tong et al. (1998a) demonstrates that prior DCA exposure in rats
substantially reduces the cytosolic conversion of DCA to glyoxylate from the inhibition of
GSTZ. The rate constants for the DCA-dependent inactivation of the four polymorphic variants
of recombinant human GSTZ were in the following order: variant la-la < Ib-lb ~lc-lc ~Id-Id
(Tzeng et al., 2000). Thus, the most frequent human GST variant (GSTZlc-lc) observed by
Blackburn et al. (2000) has a low activity toward DCA and is impacted by DCA inhibition to a
9
-------
greater extent than the most active enzyme variant (GSTZla-la). The observations of Tzeng et
al. (2000) were confirmed by Lantum et al. (2002) where each of the four enzyme variants were
tested in vitro with chlorofluoroacetate as the substrate. Residual enzyme activities of the Ib-lb,
Ic-lc, and Id-Id variants were 3, 4.5, and 4% of the original activities while la-la retained 12%
of its original activity. Accordingly, one might expect poor clearance of DC A from human
plasma via oxidative dechlorination when exposure is continuous even in individuals that carry
the la-la GSTZ variant.
Work by Anderson et al. (1999), long et al. (1998a), and Wempe (1999) suggests that
the inhibition of GSTZ is due to the formation of a covalent bond between GSH and DCA
forming an S-(a-chlorocarboxymethyl) glutathione intermediate. This intermediate can then
undergo hydrolysis releasing the remaining chloride and forming S-(a-hydroxycarboxymethyl)
glutathione liberates glyoxylate and regenerates the GSH. Alternately, S-(a-
hydroxycarboxymethyl) glutathione reacts with a nucleophilic residue on the enzyme (i.e.,
histidine or tyrosine) and modifies and inhibits the enzyme. Anderson et al. (1999) demonstrated
that the inactivation of GSTZ by DCA is irreversible and is accompanied by a loss of
immunoreactive GSTZ protein. Additional support for irreversible enzyme inhibition is provided
by the work of Schultz et al. (2002), which demonstrates that the recovery from enzyme
inhibition requires protein synthesis.
Schultz et al. (2002) found that the metabolic clearance of DCA in mice depends on age,
dose, and the presence or absence of pretreatment. For example, pretreatment of the animals
with increasing DCA doses decreased metabolic clearance based on in vitro kinetic
measurements of hepatic enzymes. However, the metabolism of DCA by pre-exposed older rats
was comparable to the older naive rats. In hepatic tissues of young mice, decreased metabolic
clearance was accompanied by a decrease in immunoreactive GST zeta, while in aged mice the
amount of immunoreactive protein remained constant (Schultz et al., 2002). The authors
hypothesized that this was the result of decreased turnover of the inhibited enzyme with
increasing age. Aged rats (16-month) showed a decreased capacity to metabolize the second of
two doses of DCA when compared to rats that were three to four months old (James et al., 1998).
The aged rats also had peak plasma concentrations that were 5-fold higher than the young rats,
while elimination half-life was approximately doubled.
Additional evidence to support enzyme inhibition comes from studies in which rodents
were predosed with DCA in their drinking water for 2 weeks, followed by a single intravenous
10
-------
dose of DC A. This predosing regimen significantly lowered DCA-derived CO2 excretion in rats,
but not in mice (Cornett et al., 1997). The predosing regimen also resulted in a significant
increase in plasma area under the curve (AUC) values for the parent compound relative to
controls in rats and mice (Cornett et al., 1997). A similar study by Gonzalez-Leon et al. (1999)
found no significant change in the amount of labeled CO2 formed in mice treated with 2 g/L
DCA in drinking water for two weeks and a subsequent 100 mg/kg dose of DC A by gavage. The
plasma AUC for DCA was about three times that for the untreated controls. These results in
mice support the findings of Cornett et al. (1997).
Schultz et al. (2002) conducted an experiment in young (8-week-old) mice by exposing
them to 2.0 g/L DCA for 14 days and an intravenous dose of 20 mg/kg DCA at 6-, 16-, 36-, or
48-hours after cessation of the drinking water exposure. Metabolic clearance of DCA was
greatly inhibited 6 and 16 hours after the end of the drinking water exposure periods, slightly
inhibited 36 hours after the end of the drinking water exposure and essentially the same as
untreated controls 48 hours after the end of the DCA drinking water exposure period. A similar
inhibition of metabolism of sub sequent DC A doses after the administration of a 50 mg/kg dose
has been shown in rats (James et al., 1998).
The toxicological relevance of the inhibition of glutathione-S-transferase
biotransformation by DCA in different species is not entirely clear. For instance, DeAngelo et
al. (1996) determined that Fischer (F344) rats were more sensitive than B6C3F1 mice with
regard to DCA-induced hepatocarcinogenicity based on the mean daily doses at which 50% of
the animals exhibited liver neoplasia (Tong et al., 1998a). However, the rates of DCA
biotransformation were much greater in mice than rats. Accordingly, Tong et al. (1998a)
concluded that the carcinogenicity of DCA does not appear to be directly related to its
glutathione-S-transferase-dependent biotransformation.
Cornett et al. (1999) suggested that differences in carcinogenicity may be related to
tyrosine metabolites that accumulate when GSTZ is inhibited rather than DCA metabolites. The
study authors proposed that DCA concentrations that inhibit GSTZ also increase the
concentration of MAA and its decarboxylated end product, maleylacetone, both of which are
postulated to be alkylating agents and are linked to the mechanism for carcinogenesis for those
that suffer from hereditary tyrosinemia I (Schultz et al., 2002).
3.3.1. Mechanistic Metabolic Considerations
11
-------
There are a number of unanswered questions about the metabolism of DCA and its relevance
to toxicity in laboratory species and humans. The question of whether or not there is more than one
metabolic pathway for DCA remains unanswered. Recent work by Schultz et al. (2002) comparing
DCA clearance in pre-exposed young (10-week-old) versus aged (60-week-old) mice suggests that
there may be more than one metabolic option in mature animals. In the aged mice, DCA clearance was
minimally affected by pretreatment with DCA, but it was significantly suppressed (>80%) in the young
mice with recovery times of less than 16 hours. This is in contrast to the in vitro work that had
demonstrated that the hepatic GSTZ was inhibited in the aged mice to almost the same extent as in the
young mice and suggested the possibility of extrahepatic metabolism in the older mice.
The relative affinities of the GSTZ/MAAI active site for DCA versus maleylacetoacetate, when
DCA concentrations are low, is another consideration that needs investigation. Maleylacetoacetate is a
degradate of phenylalanine and tyrosine, both essential amino acids in mammals. Because of this, the
activity of GSTZ with MAA as a substrate may be favored over that of GSTZ with DCA at low
concentrations such as those present in chlorinated water. This could favor DCA metabolism by
another pathway. Lantum et al. (2002) examined GSTZ activity using compounds similar to MAA and
DCA that provide information on the relative variant activities. Maleylacetone was used as a surrogate
for maleylacetoacetate while chlorofluoroacetate was considered as a surrogate for DCA.
The studies by Lantum et al. (2002) were conducted in vitro with the la-la, Ib-lb, Ic-lc,
and Id-Id enzyme variants. Reactions with the substrates (maleylacetone — 0 tol mM or
chlorofluoroacetate - 0 to 2 mM) were carried out at pH 7.4 and 25°C for 30 seconds with
measurement of product by HPLC. Triplicate samples were analyzed and the values for Km, k^, Vmax
and kcat/Km were determined. There was considerable variability in the Vmax and Km values with
maleylacetone as the substrate as reflected in the standard estimates of the means. Enzyme activity
seemed to be driven by the k^ differences to a greater extent than the Km values. In other words, the
catalytic activity in the active site (k^) appeared to impact Vmax to a greater extent than the affinity of
the active site for the substrate (Km).
Despite the variability of results, the lower catalytic efficiency of the la-la variant with
maleylacetone as a substrate was apparent in the k^ values as was its greater efficiency with
chlorofluoroacetate. The Ic-lc variant had the highest k^ with maleylacetone as a substrate. The
kcat/Km ratios (Table 3-1) also reflect the lower effectiveness of la-la with maleylacetone as
12
-------
a substrate but the ratios are fairly consistent for the chlorofluoroacetate, suggesting comparable
effectiveness of the enzyme variants when substrate concentrations are below saturation.
Table 3-1. Enzyme kinetics for GSTZ: krat/Km ratios (M *sec *)
Enzyme Variant
GSTZ la- la
GSTZ Ib-lb
GSTZ Ic-lc
GSTZ Id- Id
Substrate
maleylacetone
7.6 xlO5
20xl05
14.5 xlO5
20.6 xlO5
chlorofluoroacetate
4.3 x 103
4.7 x 103
5. Ox 103
4.1 x 103
Source: Adapted from Lantum et al. (2002).
Lantum et al. (2002) also evaluated the inhibitory effect of maleylacetone and its product,
fumarylacetone, on the reaction of chlorofluoroacetate. Lineweaver-Burke plots of the inhibition
indicate that it was neither purely competitive nor noncompetitive. Accordingly, the authors
described the effects of maleylacetone and fumarylacetone as mixed inhibition.
The work of Lantum et al. (2002) suggests that GSTZ would preferentially react with
MAA under conditions where DCA and MAA were competing for the enzyme's active site in
individuals possessing the Ib-lb, Ic-lc, and Id-Id variants and that there would be a greater
opportunity for haloacetic acid to be favored with the la-la variant. However, it is important to
recognize that the reactions were carried out under conditions where there had been no
preexposure of the enzymes to a halo-acid and, thus, no prior inhibition. Results with an
inhibited enzyme might be quite different.
The identity of the DCA toxic intermediate(s) for cancer and noncancer effects is also
unknown. As mentioned previously, Cornett et al. (1999) suggested that MAA and
maleylacetone, the tyrosine metabolites that could accumulate when GSTZ is inhibited, might be
involved with DCA toxicity because they are alkylating agents. Fernandez-Canon et al. (2002)
demonstrated that there is a metabolic bypass to this reaction in MAAI/GSTZ-deficient mice.
Homozygous-MAAI null mice were monitored for up to 22 months and displayed normal
growth and reproductive success when compared to the controls. No adverse effects on tissue
histopathology were seen at two and six months in the organs examined (including liver and
13
-------
testes). Some biochemical abnormalities were observed. For example, the authors determined
that fumarylacetoacetate and succinylacetone were found in the urine of the enzyme-deficient
mice, but not in the controls and that there was an induction of NMO-1 mRNA in the liver in the
MAAI deficient mice. NMO-1 has the ability to reduce quinones via a mechanism that prevents
the generation of free radical oxygen thereby protecting cells from oxidative stress.
Another important difference between the enzyme-deficient mice and the controls was
their response to a diet enriched in phenylananine, tyrosine ,or protein. Increased intake of
protein, phenylalanine, or tyrosine caused a rapid loss of weight and death in the enzyme-
deficient mice. This study serves to reduce but not remove the concern that the toxicologically-
active metabolite in DCA-exposed mice is MAAI or maleylacetone, rather than DCA or a DCA
metabolite. The presence of the bypass reaction, however, may not be completely protective for
species such as humans that consume a high-protein diet.
Carcinogenic and genotoxic effects of DCA have been most strongly associated with
high doses where DCA metabolism is inhibited. This observation may indicate that DCA or a
metabolite produced when the availability of the GSTZ pathway becomes limiting is the most
actively toxic compound. Tzeng et al. (2000) reported that the relative rate of DCA-induced
inactivation of liver GSTZ was greater in rats than in mice or humans. GSTZ activity was
greater in mouse liver than human liver. This could mean that humans are more sensitive to
DCA toxicity than other species if toxicity is due to unmetabolized DCA.
Dose is another factor to consider in evaluating the toxicity of DCA in acutely- and
chronically-exposed subjects. Saghir and Schultz (2002) examined the oral bioavailability of
DCA in rats at doses of 0.25 to 100 mg/kg. Previously unexposed animals were given 1, 5 or 20
mg/kg DCA; blood samples were collected and analyzed for DCA at intervals over a 24-hour
postdosing period. DCA was rapidly metabolized for the 1 mg/kg dose and plasma
concentrations were less than 6 ng/mL (the limit of detection) within 15 minutes of dosing. With
the 5, 20, and 100 mg/kg doses, the amounts of DCA in the plasma (oral bioavailability) were
10, 13 and 81% of the dose, respectively. In rats that had been pretreated with DCA in drinking
water (0.2 mg/L) for seven days to inhibit GSTZ, the estimates of the oral bioavailability were
14, 28, 31, 75 and 100%, respectively, for oral doses of 0.25, 1, 5, 20, and 100 mg/kg.
3.4. ELIMINATION
14
-------
Only a small fraction of DCA (-1-2%) is found in the feces in animal studies (Lin et al.,
1993; Larson and Bull, 1992). There is also minimal (~ 1%) excretion of unmetabolized DCA in
the urine at low doses (Lukas et al., 1980; Curry et al., 1991; Lin et al., 1993), but as the DCA
dose increases the amount of parent compound in the urine also increases (Lin et al., 1993).
Kim et al. (1999) collected morning urine samples from 25 women who were part of a
study on neural tube defects and whose drinking water was chlorinated. Exposure to DCA was
estimated based on analysis of a single tap water sample from the subjects home. Subjects spent
most of their time in their household for the 48-hour study period. While DCA was detected in
the urine of all subjects, there was no relationship between estimated exposure and urinary DCA
excretion or creatinine-normalized DCA excretion. As part of the same study, two women
ingested water containing 4.0 or 6.3 \ig/L DCA. DCA appeared in the urine immediately after
exposure and accounted for 2 to 5% of the ingested dose. This amount is a slightly higher
fraction of the ingested dose than has been reported for animals.
Oxalate is the primary urinary metabolite of DCA; it is formed by the oxidation of
glyoxylate (Stacpoole et al., 1998a). In humans and animals, variable quantities of glyoxylate,
glycolate, monochloracetic acid, and thiodiacetic acid are found in the urine (Larson and Bull,
1992; Lin et al., 1993; Gonzalez-Leon and Bull, 1996; Stacpoole et al., 1998a). A fraction of the
glyoxylate produced from DCA is oxidized to carbon dioxide and is exhaled. Carbon dioxide is
also produced by the degradation of glycine formed from glycoxylate.
3.5. PHYSIOLOGICALLY-BASED TOXICOKINETIC MODELS
A pharmacokinetic model for DCA used Advanced Continuous Simulation Language and
data from B6C3F1 male mice exposed to DCA by intravenous injection and oral gavage (Barton
et al., 1999). Some of the tested animals had no prior exposure to DCA, while others had been
pretreated with drinking water containing 2 g/L for two weeks prior to the administration of 20
or 100 mg/kg test doses. A two-compartment model was developed to project expected blood
concentrations and area under the curve in the liver (AUCL) after DCA exposure; the model
included compartments for the lumen of the small intestine, the liver, and the body (with its
volume of distribution corrected for the liver volume). DCA uptake by way of intravenous,
injection, gavage, and drinking water exposures plus elimination rates via hepatic metabolism
and excretion were included in the model. Experimental data were used to determine the volume
15
-------
of distribution, metabolic rate, uptake kinetics, and elimination constant that would fit the data in
the model.
The control animals were found to clear DCA from blood rapidly (metabolic rate: 40
mg/hr/kg body weight). The rate of clearance for the preexposed mice was considerably lower
(metabolic rate: 3-8 mg/hr/kg body weight) resulting in higher blood concentrations of DCA in
these animals (Barton et al., 1999). While the model seemed to fit the data, it under-predicted
blood concentrations for mice intravenously exposed to 100 mg/kg at about 1 hour post exposure
for naive and pretreated mice. The projected clearance time from blood for the naive mice was 1
hour while it was 3.5 hours in the pretreated mice, reflecting the inhibition of DCA metabolism.
For mice exposed via drinking water, the model predicted a nonlinear AUCL for both
naive and pretreated animals (Barton et al., 1999). This reflects saturation of metabolism in both
instances. The projected AUCL was higher in the pretreated animals than in the naive animals
with drinking water concentrations of 0.01 to about 100 g/L. The model projected an AUCL for
the pretreated animals that was about 8-fold higher than the untreated animals at concentrations
between 0.01 and about 0.8 g/L. With drinking water concentrations between 1 and 10 g/L, the
modeled difference between the naive animals and the pretreated animals increased dramatically
to a greater than 200-fold difference and then narrowed until it was the same for both groups at
concentrations of about 100 g/L.
The objective for the development of the DCA pharmacokinetic model was to provide a
mechanism for estimating liver concentrations of DCA that would be useful in refining the tissue
dose-response for liver tumors. The model has some utility in projecting liver concentrations
under conditions where the metabolism of DCA is not inhibited and again under conditions of
maximum inhibition. However, it cannot provide estimates under conditions of partial metabolic
inhibition or project how liver concentrations might vary with differences in the activity of
GSTZ isozymes.
16
-------
4. HAZARD IDENTIFICATION
4.1. STUDIES IN HUMANS
The following studies in humans are comprised of clinical reports and case studies.
There are currently no epidemiological studies regarding the chronic effects of DC A exposure in
humans with numbers adequate to provide information regarding carcinogenicity or toxicity at
doses lower than those discussed in the following sections. Studies in humans cannot predict
carcinogenicity in the exposed individuals because of several limitations: in most studies too few
individuals were studied, the exposure period or observation period was limited and a minimum
number of endpoints were monitored. The reader is therefore advised to consider the following
information as indicative only of potential toxicity in humans exposed to DC A at therapeutic
levels. Current studies are inadequate to support predictions regarding potential adverse effects
in humans exposed to DCA at concentrations approximating those currently detectable in
finished drinking water.
For over 25 years, DCA has been used clinically as an investigational drug to treat
several metabolic disorders (congenital lactic acidosis, familial hypercholesterolemia, and
diabetes). At the present time the most active pharmaceutical use of DCA is its application in
the treatment of congenital lactic acidosis; applications in the treatment of diabetes and
hypercholesteremia do not appear to have continued. Congenital lactic acidosis includes a group
of inborn metabolic disorders that result in increased blood lactate concentrations. In most cases
the metabolic defect is located in the pyruvate dehydrogenase complex, but it can also involve
enzymes in the citric acid cycle, enzymes in the respiratory chain, pyruvate carboxylase or
phosphoenolpyruvate kinase (Stacpoole et al., 1998b). Each of these enzymes is
involved either in bridging the end products of glycolysis to the citric acid cycle or in
mitochondrial oxidative metabolism. Affected children exhibit accumulation of lactate and
hydrogen ions in blood, urine or cerebrospinal fluid, failure to thrive, and neuromuscular
degeneration. Approximately 250 new cases are identified per year and there is about a 20%
annual mortality rate for the affected population (Stacpoole et al., 1998b). Some cases of
congenital lactic acidosis do not respond to DCA treatment.
Effects of DCA treatment have been limited to transient central neuropathy (sedation),
peripheral neuropathy (tingling in fingers and toes and nerve conduction changes), and metabolic
17
-------
changes such as decreases in fasting glucose, plasma lactate and cholesterol, and alanine. For
example, Stacpoole et al. (1978) studied diabetic or hyperlipoproteinemic patients, ranging in
age from 42 to 71 years. They each received a daily oral dose of 3 to 4 g DCA (43 to 57 mg/kg-
day, assuming a 70-kg body weight) for 6 or 7 days. Seven female patients were studied over
the subsequent 7 days, while four patients (three female, one male) were studied in more detail
over a 15-day period after treatment. Some patients experienced mild sedation, but no other
laboratory or clinical evidence of adverse effects were noted either during or immediately after
the treatment phase.
Dichloroacetate treatment significantly reduced fasting blood glucose levels an average
of 24% and produced marked, concomitant decreases in plasma lactate (73%) and alanine (82%)
(Stacpoole et al., 1978). Plasma cholesterol levels significantly decreased (22%) and triglyceride
levels decreased by 71%. Plasma insulin, free fatty acid, and glycerol levels were not altered.
The treatment also depressed uric acid excretion, resulting in elevated serum uric acid levels.
Maximum effects were generally noted at the end of the 6- to 7-day treatment period and
returned to pretreatment levels during the post-treatment observation period. Plasma cholesterol
levels were not altered by treatment in one patient, and the depression of cholesterol levels in the
others returned to the pretreatment levels during the recovery period.
The effects of DCA on intermediary metabolites appear to be the result of its activation
of pyruvate dehydrogenase, a key enzyme controlling the flow of three carbon metabolites into
the citric acid cycle. Pyruvate dehydrogenase exists in active and inactive forms, and is
deactivated by phosphorylation through the action of pyruvate dehydrogenase kinase. It is
activated through the removal of the phosphate via pyruvate dehydrogenase phosphatase. DCA
is an inhibitor of the kinase, thus maintaining the enzyme in its active form (Stacpoole et al.,
1998a).
Moore et al. (1979) evaluated clinical effects in two individuals treated with
dichloroacetate for radically elevated serum cholesterol. An 8-year-old boy with severe familial
hypercholesteremia was given 50 mg/kg-day DCA orally. Total serum cholesterol levels
decreased from >1,000 to 849 mg/dL within 7 days. Continued treatment for 5 weeks resulted in
a further decrease to 727 mg/dL. No adverse clinical or laboratory signs were detected in this
individual.
18
-------
In a case study of a 21-year-old man reported by Moore et al. (1979), dichloroacetate
treatment (50 mg/kg-day) decreased total serum cholesterol levels from 578 to 372 mg/dL in 1
week. At this point, the patient was switched to therapy with nicotinic acid and cholestyramine,
but treatment was ineffective and cholesterol levels rose to more than 500 mg/dL. Therapy was
reinstated and serum cholesterol levels decreased to 363 mg/dL after 2 weeks and to 325 mg/dL
after 10 weeks. After 16 weeks of treatment, the patient complained of tingling in his fingers
and toes. Physical examination revealed slight decreases in the strength of facial and finger
muscles, diminished to absent deep tendon reflexes, and decreased strength in all muscle groups
of the lower extremities (distal muscle groups being most severely affected). Electromyographic
studies revealed denervation changes in foot and distal leg muscles. Mild slowing of conduction
velocity was noted in both posterior tibial nerves, and no measurable response was obtained in
the peroneal or sural nerves. Treatment was immediately discontinued. Eight weeks after
treatment stopped, the patient stated that the tingling sensation had subsided. The strength of his
facial muscles was normal, and strength in his legs and feet was slightly improved. Six months
after treatment was stopped, the patient exhibited normal motor strength, increased deep tendon
reflexes and marked improvement in electromyographic and nerve conduction examinations.
Serum cholesterol returned to its former high level following the cessation of treatment.
Stacpoole et al. (1998a, b) reviewed observations in humans that have accrued from
nearly 25 years of experimental DCA clinical use, primarily in the treatment of congenital lactic
acidosis. Therapeutic doses of DCA are usually in the range of 25-50 mg/kg-day (either oral or
intravenous). In several cases, treatments at 25 mg/kg-day have occurred for as long as 5 years.
Evidence of clinically-significant DCA toxicity in humans is primarily limited to the central and
peripheral nervous system. Approximately 50% of patients receiving 25-50 mg/kg-day
experience sedative effects. This effect is observed following oral, intravenous, or repeated
dosing regimens. There have been three reported cases of peripheral neuropathy following DCA
treatment, but all were completely reversible within 6 months of cessation of treatment. In one
case, following the reversal of neurological symptoms, reinstitution of DCA at 10 to 25 mg/kg-
day was maintained for 2 years without further evidence of neuropathy. Two children that were
treated for congenital lactic acidosis with 25-75 mg/kg-day DCA orally for several months had a
two-fold increase in serum transaminases, suggesting preclinical hepatic toxicity. This increase
was also reversible after the treatment ended. One child received oral doses of <25 mg/kg-day
for five years before death from pneumonia.
19
-------
Nerve conduction velocities and amplitudes were studied for one year in 27 patients with
congenital lactic acidemia who received sodium dichloroacetate treatment (Spruijt et al., 2001).
The patients (16 male, 11 female) whose age ranged from 9 months to 37.4 years (mean 9.8 ± 9.4
years), were started on 50 mg/kg-day DCA and were coadministered 100 mg/day thiamine.
Lactate and plasma DCA concentrations were measured at 3, 6, and 12 months, and
pharmacokinetics of DCA was measured at 3 and 12 months (data were not reported for these
time intervals). All but two of the patients had normal baseline nerve conduction tests prior to
DCA administration. Twelve of the patients (9 male, 3 female) who had prior normal baseline
electrophysiology showed evidence of neuropathy (decreased nerve conduction velocity and
response amplitude) by the end of treatment. Three patients showed neuropathy early, within 3
months of treatment. Neuropathy increased during treatment in the two patients who exhibited
neuropathy prior to the start of therapy. Patients with neuropathy were notably older than those
with normal electrophysiology; while age was significantly correlated with the deterioration in
conduction of some nerves at certain time periods, there was an insufficient number of
individuals in the study to provide statistical power for testing age and the deterioration of most
nerves.
Data on DCA in humans are scarce, and the fact that available studies in humans have
predominantly focused on individuals who were being treated for a disease complicates the
assessment of DCA-mediated toxicity. Many of these individuals were extremely ill and the fact
that they were being dosed with other medications in addition to DCA presents the possibility
that any adverse effects of DCA treatment might not be observed by a clinician. For example,
effects might have been masked or developed over a longer period than the treatment period
used. To date, there have been no reports of DCA-induced neoplasia in any tissue or gonadal
toxicity in humans.
20
-------
4.2. STUDIES IN ANIMALS
4.2.1. Acute and Subchronic Studies
Mice
Male mice were administered varying levels of DCA (0.1 to 3 g/L) in their drinking
water for up to 8 weeks and were subsequently examined for accumulation of glycogen in their
liver (Kato-Weinstein et al., 1998). Significant increases in the glycogen content of the liver
were observed after two weeks with concentrations as low as 0.5 g/L (100 mg/kg-day).
Glycogen concentrations reached maximum levels within 1 week of treatment at concentrations
of DCA in drinking water of 1 g/L (200 mg/kg-day) and above. The glycogen that accumulated
at this early stage was subject to mobilization by fasting. However, with continued treatment,
the deposited glycogen became increasingly resistant to mobilization. After approximately 8
weeks, the glycogen content of the livers of DCA-treated mice were the same for animals tested
in fasted and nonfasted states.
Male B6C3FJ mice (12/dose level) were exposed to dichloroacetate concentrations of 0,
0.3, 1, or 2 g/L in drinking water for 14 days (Sanchez and Bull, 1990). This corresponded to
doses of approximately 0, 57, 190 or 380 mg/kg-day. Male and female Swiss-Webster mice
(4/sex/dose) were exposed to 0, 1 or 2 g/L (0, 190 or 380 mg/kg-day) for 14 days. In male
B6C3FJ mice, exposure to 190 and 380 mg/kg-day increased the liver weight and hepatocyte
diameters. Increased hepatocyte size was attributed to increased glycogen deposition. At these
dosage levels, livers had pale streaks running on the surface and, occasionally, discrete round
white areas. In Swiss-Webster mice, liver weight increased at the high dose level in both sexes
and the relative liver-to-body weight ratio increased in both sexes in a dose-related manner.
Localized areas of necrosis were observed at both doses. A significant increase in the labeling
index of hepatocytes was observed in male B6C3FJ mice treated with 2 g/L at day 14, but not at
lower doses. These observations led the authors to speculate that the carcinogenic effects of
DCA seen in other studies (Bull et al., 1990) may be related to DNA damage and increased
repair activities, and abnormal glycogen deposition may be an underlying mechanism of DCA-
induced hepatotoxicity. Based on histological evidence of liver toxicity the study identifies a
NOAEL of 57 mg/kg-day for E6C3Fl mice and a LOAEL of 190 mg/kg-day for E6C3Fl and
Swiss-Webster mice.
21
-------
Rats
In an acute study investigating DCA-induced metabolic changes, male Sprague-Dawley
rats were administered a single 100 mg/kg dose of DC A by gavage (Evans and Stacpoole, 1982).
Animals were sacrificed 0.5, 1, 3, 6, 12, or 24 hours later (three animals at each time period).
Blood glucose, pyruvate, and lactate were significantly decreased at 3 hours after dosing with a
return to basal levels at 6 hours. No significant change in pyruvate dehydrogenase complex
activation was detected in the liver. The authors did not evaluate other health effects.
The effect of multiple doses of DC A on pyruvate dehydrogenase complex activity was
assessed by the administration of three successive 100 mg/kg doses at 6-hour intervals to male
Sprague-Dawley rats. Groups of three animals were sacrificed 3, 6, 12, 24, 48, and 72 hours
after dosing. Multiple dosing resulted in a progressive rise in pyruvate dehydrogenase complex
activity with each dose. Activity was determined as the ratio of active to total (CaCl2- and
MgCl2-activated) pyruvate dehydrogenase complex. Activity returned to basal levels 24 hours
after the second and third dose (Evans and Stacpoole, 1982).
In a third experiment, adult Sprague-Dawley rats were exposed to 100 mg/kg-day DC A
by gastric intubation for 7 days. Blood lactate was decreased, and the reduced level was
maintained until 48 hours after the final dose. The activity of pyruvate dehydrogenase
significantly increased in muscle and liver tissue, but returned to basal levels within 24 hours
after cessation of dosing.
Davis (1990) performed a similar study of DCA-induced metabolic changes in rats, but
used slightly larger dose groups and two dose levels of DC A. Groups of Sprague-Dawley rats
(5/sex/dosage group) were administered a single dose of DC A by gavage (three times in one day)
for a total dose of 0, 120, or 316 mg/kg-day. Animals were subsequently examined for
alterations in glucose and lactate levels in the plasma, liver and kidney. Decreased plasma
lactate levels were observed in both sexes in both dosage groups. Plasma glucose levels were
not decreased and, although tissue lactate levels were reduced, the differences were not
significantly different from the controls.
In an earlier study, Davis (1986) evaluated the administration of DC A in drinking water
for a two-week period on metabolism in the rat. Sprague-Dawley rats (5/sex/group) were given
22
-------
water containing 0, 30, 125, 500, or 1,875 mg/L of DCA for 14 days. These concentrations
correspond to target dose levels of 0, 10, 40, 150, or 600 mg/kg-day. Rats in the high-dose
group lost weight during the first week and then began gaining weight normally during the
second week. This effect was not statistically significant and was correlated with decreased
water and food consumption in the high-dose group. Urine volume and osmolarity were not
significantly affected in any groups; however, there was a trend toward decreased volume and
increased osmolarity with increased DCA exposure (consistent with decreased water
consumption). Ammonia excretion and renal phosphate-dependent glutaminase activity tended
to increase with increasing exposure. These effects were considered by the author to be normal
adaptation to an acid load. Lactate and pyruvate levels in females were not significantly affected
in either the liver or kidney, although a trend toward decreased liver lactate was observed.
Blood glucose levels were not significantly affected in either males or females. The study
identified a NOAEL in rats of 150 mg/kg-day.
In a subchronic study, the metabolic and toxic effects of DCA were investigated in rats
following dietary administration of the compound (Yount et al., 1982). Doses varied from 4
mmol/kg-day at the beginning of the study to 2.5 mmol/kg-day (516 to 323 mg/kg-day) during
the 12-week study period. Dichloroacetic acid did not affect plasma glucose levels, but led to
decreased plasma triacylglycerol and increased plasma ketone bodies. Hind limb weakness and
abnormal gait were observed in exposed animals within 2 to 4 weeks, while decreased nerve
conduction velocities was observed in sural, tibial, and motor nerves. Decreased food
consumption and decreased weight gain occurred in exposed animals, and organ-to-body weight
ratios were increased for the adrenal glands, brain, and kidney. Dichloroacetic acid also caused
hepatomegaly and there was evidence of testicular degeneration.
Groups of five male Sprague-Dawley rats were administered DCA (0 or 1,100 mg/kg-
day) in drinking water for 90 days (Bhat et al., 1991). Body weights were monitored throughout
the study. Following the 90-day treatment regimen, the animals were sacrificed and selected
organs were isolated for evaluation. The following organs were weighed and examined for
histopathological alterations: liver, lung, heart, spleen, thymus, kidney, testes, and pancreas.
The brain and liver were also examined for collagen deposition. At sacrifice, the average body
weight of the DCA-treated group was 66% that of the control group. The DCA-treated animals
also had increased liver weight (p<0.01), increased liver-to-body weight ratios (p<0.01), and
increased liver collagen deposition compared to control animals. Perivascular inflammation was
23
-------
noted in the lungs and focal vacuolation and gliosis were present in the forebrain and brain stem
of the DCA-treated group. The authors reported a progressive DCA-related decrease in water
intake (presumably taste aversion) with a concomitant drop in food consumption. This potential
confounder could be the underlying cause for the observed weight loss in the treated animals and
could have contributed to other reported effects of DC A including liver alterations. The
examination of a single high-dose limits the study.
Mather et al. (1990) administered Sprague-Dawley rats (10 males/dosage group) DCA in
their drinking water at 0, 0.05, 0.5, or 5 g/L (equivalent to dosage levels of approximately 0, 3.9,
35.5, or 345 mg/kg-day, respectively) for 90 days. Water consumption was significantly
(p<0.05) reduced in the 0.5 and 5 g/L treatment groups when measured at two months of
exposure. Terminal body weights were significantly reduced (p<0.05), and there were increases
(p<0.05) in liver- and kidney-to-body weight ratios at dose levels of 35.5 mg/kg-day or greater.
At the highest dose, the spleen-to-body weight ratio increased. Total serum protein levels were
significantly depressed at all doses. Significant increases in alkaline phosphatase were seen at
the two highest doses, while alanine aminotransferase levels were increased at the highest dose.
Hepatic peroxisomal beta-oxidation activity was significantly increased at the highest dose (as
measured by [14C] palmitoyl-CoA oxidation), but no effects were seen on hepatic microsomal
enzyme activity. Liver effects were also observed at the high dose, including focal
hepatocellular enlargement, intracellular swelling, and glycogen accumulation. Kidney effects at
the highest dose used were characterized by diffuse degeneration of the tubular epithelium and
glomeruli. Although spleen weights increased, histopathological changes in the spleen were not
observed. No consistent effects were observed for immunological parameters, such as antibody
production, delayed hypersensitivity, natural killer cell cytotoxicity, or production of PGE2 or
IL-2. Based on hepatic and renal effects in male rats, this study defined a NOAEL and a
LOAEL of 3.9 and 35.5 mg/kg-day, respectively.
Katz et al. (1981) evaluated the effect of DCA in rats following a 3-month exposure and a
postexposure recovery period. Sodium dichloroacetate was administered to Sprague-Dawley
rats (10 to 15/sex/group) by gavage at dose levels of 0, 125, 500, or 2,000 mg/kg-day for 3
months. Five more rats per sex were added to the control group and the high-dose group. They
were monitored for an additional 4 weeks after the 3-month feeding period was discontinued.
Two rats of each sex in the 2,000 mg/kg-day group died during the study. The major signs of
intoxication were hind limb paralysis and frequent urination. Two rats (one of each sex)
24
-------
exhibiting these signs appeared to recover completely during the 4-week recovery period. Body
weight gain was significantly depressed in a dose-dependent manner at all dose levels during the
dosing period. Minimal effects on hematological parameters were observed at the two highest
doses. All groups exhibited significant depressions in blood glucose and lactate, while creatinine
levels increased. Male rats exhibited significantly depressed blood levels of total protein,
triglycerides, iron, and calcium, as well as elevated levels of total and direct bilirubin, sodium,
and potassium. Cessation of treatment was followed by a return to baseline levels in all
parameters. The mean relative weights of the liver, kidneys, and adrenal glands were
significantly increased in a dose-dependent fashion, but both absolute and relative organ weights
tended to approach those of the controls during the 4-week recovery period. The brain and testes
were the target organs of DC A intoxication. Brain lesions (characterized by vacuolization of the
myelinated white tracts) were observed in the cerebrum and cerebellum of treated rats of both
sexes in all dose groups (combined incidence rates of 60% at 125 mg/kg-day and 100% at 500
and 2,000 mg/kg-day). In 3/8 rats, the brain lesions persisted after cessation of treatment. Based
on these effects on organ weights and brain lesions, this study identified a LOAEL of 125
mg/kg-day, the lowest dose tested.
Moser et al. (1999) extended the evaluation of the neurotoxic effects of DC A exposure in
a series of experiments in weanling and adult rats. The study used a neurobehavioral screening
battery under varying exposure durations (acute, subchronic, and chronic) and routes of
administration (oral gavage and drinking water). The following is a description of the
subchronic study which consisted of several experiments (designated by the authors as
experiments 2, 4, 5a, 5b, 6a, 6b, and 7a and 7b). None of the experiments employed a control
group. Experiments 2, 7a and 7b examined adult rats, while experiments 4, 5a, 5b, 6a, and 6b
were conducted on weanling rats.
In experiment 2, Long-Evans (LE) rats (80-days-old) were treated with 30, 100, 300, or
1,000 mg/kg-day by oral gavage for 5 d/wk for 10 weeks, with a 1-week recovery period. The
results revealed alterations in the gait of adult rats in the 300 and 1,000 mg/kg-day dose groups.
In addition, mild tremor, hypotonia, and decreased forelimb grip strength was observed at the
high dose. The 100 mg/kg-day dose was a NOAEL.
In experiments 7a and 7b, LE and F344 rats (68-69-days-old) were administered DCA
(via drinking water) at doses of 23, 122, or 220 mg/kg-day (LE rats) or 18, 91, or 167 mg/kg-day
(F344 rats) for 8 weeks, plus 2-week recovery period. Some of the F344 rats in the low-dose
25
-------
group showed gait abnormalities. Gait abnormalities and decreased forelimb and hind limb grip
strength were noted in the mid- and high-dose LE and F344 rats. In addition, increased foot
splay was noted in the F344 rats. A chest-clasping response was seen in the high-dose F344 rats.
With the exception of gait deficit and decreased hind limb grip strength, both strains showed
recovery 2 weeks after exposure was discontinued. In F344 rats, the low dose of 18 mg/kg-day
was a LOAEL for gait abnormalities. In LE rats, 23 mg/kg-day was a NOAEL and the LOAEL
was 122 mg/kg-day.
Experiment 4 involved exposure of weanling F344 rats (30-days-old) to 162 or 308
mg/kg-day DCA in the drinking water for 12 weeks, plus a 5-week recovery period. Exposure to
the high dose was discontinued at 3 weeks due to severe toxicity; the time-weighted intake for
this dose group was 308 mg/kg-day. The low-dose group was exposed to DCA for 12 weeks as
originally planned. The high-dose animals exhibited gait abnormalities that were still evident 14
weeks after exposure ended. The high-dose animals also displayed decreased hind limb grip
strength, decreased forelimb grip strength, altered righting reflex, and lowered motor activity.
Although more pronounced in the high-dose animals, these effects were also observed in the
low-dose animals, with peak effects evident during the 9th and 12th week of exposure. A dose
of 162 mg/kg-day was a LOAEL for neurotoxic effects in weanling F344 rats.
Experiment 5 included two segments (5 a and 5b) and was intended to compare the
potency of DCA in drinking water and by oral gavage. In experiment 5a, weanling F344 rats
(28-29-days-old) were exposed to drinking water containing 16, 66, or 172 mg/kg-day DCA for
12 weeks, plus a 15-week recovery period. Another group of weanling F344 rats (experiment
5b) were treated by gavage to 176 mg/kg-day DCA for 12 weeks, plus a 15-week recovery
period. Clear signs of neurotoxicity were observed in high-dose (172 mg/kg-day) weanling rats
of experiment 5a (drinking water route). Neurotoxic signs consisted of gait abnormalities,
righting reflex deficits, decreased motor activity, decreased grip strength, and tremors.
Progressive gait changes and decreased motor activity were evident in the mid-dose animals (66
mg/kg-day). Low-dose (16 mg/kg-day) animals exhibited moderate effects on gait. In gavage-
dosed rats (176 mg/kg-day; experiment 5b), gait abnormalities developed within 3 weeks and
became progressively worse during dosing. In contrast to the drinking water route, hind limb
grip strength and other neuromuscular endpoints were not affected by gavage treatment.
Experiment 6 involved the exposure of weanling LE rats (experiment 6a) or weanling
F344 rats (experiment 6b) to drinking water containing DCA at dose levels of 17, 88 or 192
26
-------
mg/kg-day (experiment 6a) or 16, 89, 173 mg/kg-day (experiment 6b) for 13 weeks. The
purpose of experiment 6 was to investigate potential strain differences in the response to DC A.
Changes were assessed using a functional observation test battery and monitoring of motor
activity. The results of experiment 6 revealed that both rat strains showed progressive changes
in gait in all treated groups (LOAEL, 17 mg/kg-day in LE rats and 16 mg/kg-day in F344 rats).
The effect was most pronounced in the high-dose F344 rats. Hind limb grip strength was
decreased throughout exposure in the mid- and high-dose LE rats (no dose-response relationship
was apparent) and in the high-dose F344 rats. The effect was more pronounced in the high-dose
F344 rats. Other effects at the high-dose in both strains included tremor, hypotonia and
inhibition of pupil reflex. The study authors indicated that F344 rats, but not the LE strain,
showed a progressive decrease in motor activity, righting deficits, and forelimb grip strength,
and an increase in foot splay. Data were presented only for forelimb grip strength which was
slightly decreased (<5%) at the mid-dose; the decrease was more pronounced (approximately
20%) at the high dose.
Results of the study indicated that DCA is a more potent neurotoxicant when
administered to adult rats via drinking water rather than by gavage. The results also revealed
that gait abnormality is a critical effect for DCA. The effect was observed at doses as low as 16
mg/kg-day (in the absence of other neuromuscular changes) and was persistent in adult rats of
both strains at doses >91 mg/kg-day (F344) even following a 2-week recovery period. The data
are consistent with the persistent histological effects in the rat cerebrum observed by Katz et al.
(1981) at doses > 125 mg/kg-day. The data also revealed that hind limbs may be preferentially
affected by DCA.
Data from experiment 5 demonstrated partial recovery of neurotoxic effects, e.g.,
following a 13-week intake of 172 mg/kg-day. Experiments 6 and 7 illustrate that F344 rats are
more sensitive than the LE rat strain to DCA. In regard to age differences, limited results show
that the severity of neuromuscular toxicity was somewhat greater in rats when exposures began
shortly after weaning.
27
-------
Doss
Three studies have been performed in dogs. The first (Ribes et al., 1979) reported
decreased lactate and pyruvate levels persisting for 48 hours (35 and 27% of basal values,
respectively) following a single oral dose of 150 mg/kg of sodium dichloroacetate. Blood
glucose levels were unchanged for the first 4 hours postdosing, but declined significantly
(p<0.01) at 24-28 hours and then returned to their initial levels at the end of the 48-hour study.
Longer DC A administration (150 mg/kg-day for 7 days) caused decreases in blood glucose,
lactate, pyruvate, cholesterol, and oxaloacetate concentrations. All serum values returned to
their initial values within 2-6 days following treatment. Ketone bodies were not reported in the
urine.
Katz et al. (1981) studied the subchronic administration of sodium dichloroacetate (0, 50,
75, or 100 mg/kg-day by capsule for 13 weeks) to four beagle dogs/sex in the control and high-
dose groups and three dogs/sex in the other groups. Female dogs at all doses showed markedly
reduced appetites and both sexes exhibited dose-dependent weight losses, which were reversed
after the treatment ended. One female at 75 mg/kg-day died on day 40, and one at 100 mg/kg-
day died on day 88. The animals exhibited anorexia, ataxia, hind limb weakness, and reduced
activity. Bloody stools, vomiting, and paralysis were also observed at the highest dose level.
Dose-related decreases in erythrocyte counts, hematocrits, and hemoglobin levels were reported.
Mean blood glucose, lactate, and pyruvate levels were significantly decreased in all treated
animals. The parameters returned to normal in those animals monitored following treatment.
Treated dogs also exhibited lung consolidation. Histopathology showed neurological effects
(slight to moderate vacuolization of white myelinated tracts in the cerebrum and cerebellum),
and liver and gall bladder effects (an increased incidence of hemosiderin-laden Kupffer cells in
the liver and cystic mucosal hyperplasia in the gall bladder); these effects were persistent
through the 5-week recovery period. Indirect effects including increased incidence and/or
severity of pulmonary inflammatory lesions were also attributed to DC A treatment. Based on
the study results, the lowest dose of 50 mg/kg-day was identified as a LOAEL.
Slightly lower doses were used in a separate subchronic study (Cicmanec et al., 1991), in
which juvenile beagle dogs (4-months-old; 5/sex/dose) received daily oral doses of 0, 12.5, 39.5,
or 72 mg/kg-day DCA in gelatin capsules for 90 days. At study termination, organ weights were
determined and tissues were examined microscopically. Overt clinical signs were evident in the
high-dose animals throughout the duration of the experiment. Dyspnea (shortness of breath or
28
-------
difficulty in breathing) was observed in high-dose animals starting at day 45, and worsened with
time. Partial paralysis of the hind limbs was observed in three animals in the high-dose group
during the latter half of the exposure period. Conjunctivitis was observed in 24/30 treated
animals and a few controls during the first month, and became more severe later in the study.
The occurrence of ocular effects appeared to be dose-related, with 8/10 high-dose dogs affected.
Reduction of food and water intake was noted in DCA-treated dogs, although the effect did not
appear to be dose-related. High-dose males exhibited a 16% reduction in body weight, while
high-dose females and mid-dose males experienced a 9% reduction in weight gain over the
duration of the study. Mid-dose females exhibited an 11% reduction in weight gain. Dogs in the
mid- and high-dose groups experienced sporadic diarrhea. The most severely affected dogs
required fluid therapy to prevent severe dehydration. One female and two males treated at 72
mg/kg-day died during the study. These deaths were attributed to pneumonia and dehydration.
Statistically significant decreases in erythrocyte count and hemoglobin levels were
observed in high-dose dogs at day 30. Trend analysis of serum biochemistry data indicated
apparent increases in lactate dehydrogenase, alanine aminotransferase, and aspartate
aminotransferase activity in the high-dose groups at some time points. These findings were
consistent with microscopic tissue observations. Relative liver weight was significantly
increased in all dose groups, and absolute liver weight was increased in all but high-dose males.
Pathological examination revealed multiple changes in the organs of animals treated with DC A,
including: mild vacuolar change (most prevalent at the low dose), inflammation, and
hemosiderosis in the liver; and chronic inflammation and acinar degeneration in the pancreas.
While the primary lesions included pale and discolored kidneys, the severity of these lesions was
ranked as mild or moderate. Microscopic examination of the brain revealed mild vacuolization
of white myelinated tracts in the cerebrum and/or cerebellum of some animals in low-, mid-, and
high-dose DCA treatment groups. Mild vacuolar change was noted in the medulla and spinal
cord of some males, while mild meningoencephalitis was present in one high-dose female.
Microscopic testicular lesions were also noted in treated dogs, and are further discussed in
Section 4.3. A LOAEL of 12.5 mg/kg-day can be identified, based on visual organ effects
(neurological changes, hepatic vacuolization, and testicular effects) and increased liver weights.
29
-------
4.2.2. Chronic Studies and Cancer Bioassays
Mice
In one of the earliest studies of DCA tumorigenesis in mice, Herren-Freund et al.(1987)
gave male B6C3FJ mice (28-days-old) drinking water containing 0, 2, or 5 g/L of DCA
(corresponding to about 0, 400, or 1,000 mg/kg-day). The 400 and 1000 mg/kg-day groups were
also pretreated with ethylnitrosourea (ENU). An additional high-dose group (1,000 mg/kg-day)
did not receive ENU. Animals were sacrificed after 61 weeks of exposure and examined for
tumors. In the control group (not exposed to either ENU or DCA), there were no hepatocellular
carcinomas. In mice pretreated with ENU and subsequently administered 400 or 1,000 mg/kg-
day DCA, the incidence of hepatocellular carcinomas was 66 and 78%, respectively. In the
1,000 mg/kg-day dose group treated only with DCA, hepatocellular carcinoma incidence was
81%; this result prompted the authors to conclude that DCA was carcinogenic at this dose in the
absence of initiation.
Bull et al. (1990) used a stop-dosing regimen to evaluate the time required to onset liver
tumors in dosed mice. Groups of B6C3FJ mice were provided drinking water containing DCA at
concentrations of 0 mg/L (35 males, 10 females), or 1 g/L (11 males) for 52 weeks, or 2 g/L (11
males) for 37 weeks with a 15-week recovery period, or 2 g/L (24 males, 10 females) for 52
weeks. Based on the authors' graphical data for total dose, mean intake rates were
approximately 140 mg/kg-day (52 weeks) at the low dose and 280 mg/kg-day (37 weeks) or 300
mg/kg-day (52 weeks) at the high dose. Although treatment did not affect survival or body
weight, increased hepatic lesions were observed in all low- and high-dose groups, including:
increased absolute and relative liver weights, cytomegaly, massive accumulation of glycogen in
hepatocytes, and foci of necrosis or basophilic cellular alteration. The LOAEL for chronic non-
neoplastic effects established by this study was approximately 140 mg/kg-day for 52 weeks.
At sacrifice, no liver tumors were reported in the female mouse group, but hyperplastic
nodules were observed microscopically in the livers of 3/10 treated animals (Bull et al., 1990).
In the male mouse group exposed to 140 mg/kg-day for 52 weeks, a total of 3 hepatic lesions
were noted in 2 of 11 mice; the single lesion examined histologically was a hyperplastic nodule.
Of the 24 male mice exposed to 300 mg/kg-day for 52 weeks, 92 hepatic lesions were scored in
23 mice; of the 23 lesions in 10 mice that were examined histologically, 15 lesions in 9 mice
were hyperplastic nodules, 2 lesions in 2 mice were adenomas, and 6 lesions in 5 mice were
30
-------
hepatocellular carcinomas. Finally, of the 11 male mice exposed to 280 mg/kg-day for 37
weeks, 23 hepatic lesions were found in 7 mice; of the 19 lesions in 7 mice that were examined
histologically, 15 lesions in 6 mice were hyperplastic nodules, 2 mice had an adenoma, and no
hepatocarcinomas were observed. The authors concluded that tumorigenesis by DCA may
depend largely on stimulation of cell division secondary to hepatotoxic damage.
The U.S. EPA (1991b) evaluated the carcinogenicity of DCA in female B6C3FJ mice.
Following exposures to 0, 0.5, or 3.5 g/L DCA (approximately 0, 80 or 400 mg/kg-day) in
drinking water for 104 weeks, the high-dose group had a 100% hepatocellular tumor incidence
and a tumor multiplicity of 8.36 tumors/animal. Mice receiving 0.5 g/L DCA had a tumor
incidence of 20% and a tumor multiplicity of 0.2 tumors/animal. The untreated control group
had an incidence of 7.7% and a multiplicity of 0.1 tumors/animal.
DeAngelo et al. (1991) evaluated differential exposure doses and exposure durations on
the development of tumors in male mice. Dichloroacetic acid was administered to B6C3FJ mice
(50 males/dosage group) in their drinking water at concentrations of 0, 0.05, 0.5, 3.5, or 5.0 g/L
for 60 weeks. These doses correspond to levels of 0, 7.6, 77, 410, and 486 mg/kg-day. Other
groups of mice were administered DCA at 7.6 or 77 mg/kg-day for 75 weeks. In high-dose
treated mice, water consumption was reduced to 60% of controls. Body weight was decreased at
the two highest dose levels, and relative liver weight was increased at the three highest dose
levels. An increase in kidney weight was seen only at 410 mg/kg-day. No effects were seen on
testes or spleen weight. Therefore, the LOAEL for increased relative liver weight was 77
mg/kg-day for the 60-week study, and the NOAEL was 7.6 mg/kg-day. At 75 weeks, the
relative liver weight for the 77 mg/kg-day dose was increased, but the difference from controls
was not statistically significant. In mice receiving 410 mg/kg-day, 58% had hyperplastic
nodules, 100% had hepatocellular adenomas and 67% had hepatocellular carcinomas. At the
higher dose level of 486 mg/kg-day, 83% of the mice had hyperplastic nodules, 80% had
hepatocellular adenomas, and 83% had hepatocellular carcinomas. Incidences in other groups
(7.6 and 77 mg/kg-day) were similar to controls.
In a limited-dose cancer study, Daniel et al. (1992) exposed B6C3Fj male mice (33/dose
level) to DCA in drinking water at concentrations of 0 or 0.5 g/L (0 or 88 mg/kg-day, mean
weighted average) for 104 weeks. At terminal sacrifice, absolute and relative liver weights were
significantly increased (p<0.01) when compared to untreated controls. The mean daily water
31
-------
consumption was not significantly reduced (6.1 vs. 6.2 mL/mouse/day). There was an increased
incidence of hepatocellular necrosis, chronic inflammation, and cytomegaly in the treated group
when compared to controls. No significant changes were found in other organ weights (kidney,
testes, and spleen), body weights, or survival in the treated groups when compared to untreated
controls. A LOAEL for nonneoplastic hepatic effects in this study was 88 mg/kg-day.
Hepatocellular carcinomas (15/24 or 63% versus 2/20 or 10% in controls) and
hepatocellular adenomas (10/24 or 42% versus 1/20 or 5%) increased in animals that survived
104 weeks to terminal sacrifice (Daniel et al., 1992). The increase in the number of hyperplastic
nodules observed in treated animals (2/24 or 8%, versus 0 in controls) was not statistically
significant. No adenomas or nodules were found at the 30-week interim necropsies and there
was no interim sacrifice at 60 weeks which might have provided data on whether or not
hyperplastic nodules had started to form.
Two recent studies evaluated the carcinogenic response of DC A exposure in the female
mouse. In the first study, female B6C3FJ mice were administered 2.0, 6.67, or 20.0 mmol/L
DCA in drinking water (40, 115, or 330 mg/kg-day) from 7 to 8 weeks of age to sacrifice at 360
or 576 days (-51 or 82 weeks) of exposure (Pereira and Phelps, 1996). Significant increases in
the percentage of animals with altered hepatocyte foci and liver adenomas were seen in the 115
and 330 mg/kg-day groups, including: after 51 weeks, 40.0% with foci and 35% with adenomas
at 330 mg/kg-day; after 82 weeks, 39.3% with foci and 25% with adenomas at 115 mg/kg-day,
and 89.5% with foci and 84.2% with adenomas at 330 mg/kg-day. A significant increase in the
percentage (26.3%) of animals with liver carcinomas was only seen in the 330 mg/kg-day group
after 82 weeks of exposure. The authors concluded that the relationship of altered hepatocyte
foci frequency, hepatocellular adenoma occurrence, and hepatocellular carcinoma occurrence to
DCA concentration were best described by second-order regression.
In the second study, liver tumors were initiated in female B6C3FJ mice with 25 mg/kg
methylnitrosourea (MNU); the mice were then administered 2.0, 6.67, or 20.0 mmol/L DCA in
their drinking water (50, 167, or 468 mg/kg-day) from age 7 weeks to sacrifice 31 or 52 weeks
later to characterize tumor promotion by DCA (Pereira and Phelps, 1996). A 4 mL/kg sterile
saline vehicle control was included in the study. Significant increases in the percentage of
animals with liver adenomas were seen in the 468 mg/kg-day group after 31 weeks of exposure
(50.0% versus 0% in control) and 52 weeks of exposure (73.1% versus 17% in control). A
significant increase in the percentage of animals with altered hepatocyte foci was also seen after
32
-------
31 weeks of exposure (80.0% versus 20.0% in control) and 52 weeks of exposure (50.0% versus
10.0% in control). When the exposure to 468 mg/kg-day DCA was terminated after 31 weeks it
was followed by a 21-week recovery period. The authors observed decreased yield of altered
hepatocytes and tumors, indicating that continued existence of these lesions was dependent on
continuous exposure to DCA. The tumor-promoting activity of DCA exhibited a second-order
relationship to drinking water concentration, so that a sharp rise in potency was seen between
167 and 468 mg/kg-day.
To better understand the mechanisms of dichloroacetate carcinogenicity, Stauber and
Bull (1997) investigated changes in the replication and phenotype of cells from hepatic tumors.
Male B6C3FJ mice were pretreated with 2.0 g/L of dichloroacetate in drinking water for 38 or 50
weeks, respectively. The mice (12 animals/dose) were then administered drinking water
containing 0, 0.02, 0.1, 0.5, 1.0, or 2.0 g/L dichloroacetate for two additional weeks. At three
days prior to sacrifice, 5-bromo-2-deoxyuridine (BrdU) was administered via subcutaneously
implanted pumps to label the DNA in vivo. The animals were sacrificed and the liver tissue was
stained and examined. A transient, but significant increase in hepatocyte division rates as
compared to controls was evident for the first 14 days of treatment with 2 g/L, but was not
apparent at 28, 280 and 350 days of treatment. DCA-induced tumors were stained with anti-c-
Jun and anti-c-Fos antibodies. Dichloroacetate-induced altered hepatic foci (AHF) and tumors
were largely basophilic and reacted uniformly to antibodies against c-Jun and c-Fos (nuclear
transcription factors). The c-Jun protein was localized in the cytoplasm and the c-Fos protein
was found in the nucleus. The AHF and tumors that were c-Jun positive displayed a dose-
dependent increase in cell replication during the labeling period. The cell replication rate in
dichloroacetate-induced AHF and tumors were dependent on dichloroacetate treatment, but this
effect was observed only in the c-Jun positive regions of the lesions (see Section 4.4 for
additional data on the effects of DCA on transcription factors).
DeAngelo et al. (1999) reported on the carcinogenesis of DCA in male B6C3Fj mice.
The mice were exposed to 0, 0.05, 0.5, 1, 2, or 3.5 g/L of DCA in drinking water for 90-100
weeks. The exposures corresponded to mean daily doses of 0, 8, 84, 168, 315, or 429 mg/kg-
day, respectively. The cumulative incidence of hepatocellular carcinomas was significantly
increased in animals exposed to 1 g/L (71%), 2 g/L (95%), and 3.5 g/L (100%) when compared
to control (26%) (see Table 5-8). Hepatocellular carcinoma multiplicity (tumor/animal)
significantly increased in all treatment groups as follows: 0.05 g/L (0.58), 0.5 g/L (0.68), 1 g/L
(1.29), 2 g/L (2.47) and 3.5 g/L (2.90) when compared to the control group (0.28). The
33
-------
cumulative incidence of hepatocellular adenomas significantly increased in animals exposed to 1
g/L (51.4%), 2 g/L (42.9%), and 3.5 g/L (45%) when compared to controls (10%) and the 0.5
g/L group (20%). Hepatocellular adenoma multiplicity (tumor/animal) significantly increased in
the following dose groups: 0.5 g/L (0.32), 1 g/L (0.80), 2 g/L (0.57), and 3.5 g/L (0.64) as
compared to controls (0.12). By the end of the study, body weights decreased 18% in mice
treated with 2 and 3.5 g/L compared to the controls. All DCA doses, except the lowest dose
(0.05 g/L), resulted in an increase in the severity of hepatic necrosis compared to the controls,
when measured at 26 weeks. Necrosis was mild (between 25 and 50% of the liver sections were
affected) and transient at 1 and 2 g/L (severity did not increase at later time points with these
doses). Hepatic peroxisome proliferation increased in the high-dose group, but did not correlate
with liver tumor response. The severity of hepatotoxicity increased with DCA concentration.
Below 1 g/L, hepatotoxicity was mild and transitory (as evidenced by histopathological
examination and serum enzyme levels) and there was no significant increase in labeling index
outside of proliferative lesions. Based on these observations, the authors concluded that DCA-
induced liver cancer does not appear to be dependent upon peroxisome induction or chemically-
sustained cell proliferation. Hepatotoxicity, especially at the higher doses, may exert an
important influence on the carcinogenic process.
Rats
DeAngelo et al. (1996) reported the results of two studies of male Fischer 344 rats
exposed to DCA in drinking water. Cancer as well as noncancer toxicity endpoints were
assessed. The two studies are independent of each other (they were conducted in different
laboratories with different animals) and are described separately below.
In the first study, 28-day-old male F344 rats were given drinking water containing DCA
at concentrations of 0 (78/group), 0.05 (60/group), 0.5 (60/group) or 5.0 g/L (78/group). A
second control group (50/group) was provided water containing 2.0 g/L NaCl. Animals were
observed daily for physiological and behavioral responses and for overt signs of toxicity. Body
weights and water consumption were measured throughout the study. All animals were treated
for 100 weeks, except for animals in the 5.0 g/L group, which exhibited signs of peripheral
neuropathy. In response to this overt toxic effect, the concentration was sequentially lowered to
2.5 g/L at 9 weeks, then 2.0 g/L at 23 weeks and finally to 1.0 g/L at 52 weeks. When the
neuropathy did not reverse or diminish, the animals were sacrificed at 60 weeks and excluded
from the report. Based on measured water intake in the 0, 0.05 and 0.5 g/L groups, the time-
34
-------
weighted average doses were 0, 3.6, and 40.2 mg/kg-day, respectively. Interim sacrifices for
each dose group were performed at 15, 30, 45, 60, and 100 weeks while the NaCl control group
was sacrificed at 104 weeks. The body, liver, kidneys, testes, and spleen were weighed and
examined for gross lesions at the interim sacrifices, while at the final sacrifice, a complete
necropsy was performed on all animals. No differences were observed in water consumption,
final body weight, absolute or relative liver weight, and kidneys or spleen weight, at dosages of
3.6 or 40.2 mg/kg-day at any time point. However, absolute and relative testicular weights were
mildly, but significantly, increased at the 40.2 mg/kg-day dose at final sacrifice. Increased
hepatocellular vacuolization was detected, but there was no increase in hepatocyte proliferation
at any dose group. There was also a noted lack of necrosis observed in doses carried out to final
sacrifice at 100 weeks.
Hepatic neoplastic lesions were examined at sacrifice (DeAngelo et al., 1996). At a dose
of 40.2 mg/kg-day DCA, there was a statistically significant increase in the cumulative incidence
of combined hepatocellular neoplasia (21.4% vs. 4.4%; p<0.05) and total proliferative lesions in
the liver (34.9% vs. 8.7%; p<0.05) compared to controls. This was not observed at the lower
dose of 3.6 mg/kg-day. Tumor multiplicity was significantly increased in the 40.2 mg/kg-day
group as compared to controls. There was also a significant increase in combined hepatocellular
neoplasia (0.04 vs. 0.3) and total proliferative lesions (0.41 vs. 0.09). Other tumors were not
increased over control values.
In the second study by DeAngelo et al. (1996), male F344 rats were exposed to DCA
concentrations of 2.5 g/L DCA in their drinking water (78/group) or to deionized water
(78/group). The concentration of DCA was lowered to 2 g/L at 5 weeks, to 1.5 g/L at 8 weeks,
and to 1.0 g/L at 26 weeks. This corresponded to a time-weighted average concentration of 1.6
g/L and a time-weighted average dose of 139 mg/kg-day over the 103-week exposure period.
Interim sacrifices for each dose group were performed at 14, 26, 52, 78 and 103 weeks. The
liver, kidneys, testes, thyroid, stomach, rectum, duodenum, ileum, jejunum, colon, urinary
bladder, and spleen were examined for gross lesions at all time periods. In this study the mean
final body weight of DCA-exposed animals was significantly reduced to 73% of the deionized
water control group. Absolute testes weight decreased, but relative testes weight was not
significantly lower than the control group. Signs of liver pathology were also minimal in this
study, and this dose of DCA suppressed hepatocyte proliferation. Consistent with the first study,
there was a lack of liver necrosis observed at final sacrifice.
35
-------
Hepatic tumor incidence significantly increased in exposed animals compared to
controls, as follows: carcinoma (21.4% vs. 3.0%, p<0.05), combined hepatocellular neoplasia
(28.6% vs. 3.0%; p<0.01) and total proliferative lesions (32.1% vs. 6.1%; p<0.01, DeAngelo et
al., 1996). Tumor multiplicity was also significantly increased in the exposed group compared
to controls: combined hepatocellular neoplasia (0.36 vs. 0.03), total proliferative lesions in the
liver (0.39 vs. 0.06), and carcinomas (0.25 vs. 0.03). Other tumors examined were not increased
over control values.
Male Fischer 344 rats were administered time-weighted average concentrations of 0,
0.05, 0.5, or 2.4 g/L (0, 4, 40, or 296 mg/kg-day) DCA in drinking water, followed by sacrifice at
intervals for up to 104 weeks by Richmond et al. (1995). No hepatoproliferative lesions were
seen in the 4 mg/kg-day group, and the negative control group had only 4% hepatic adenomas.
The 40 mg/kg-day group had 10% hyperplastic nodules, 21% hepatic adenomas, and 10%
hepatocarcinomas after 104 weeks, while the 296 mg/kg-day group had 70% hyperplastic
nodules, 26% hepatic adenomas, and 4% hepatocarcinomas after terminal sacrifice at 60 weeks.
Increased numbers of altered hepatocyte foci were also seen in the 4 and 40 mg/kg-day groups,
but the differences were significant only in animals from the 45-week sacrifice.
Tumor marker expression was examined in the DCA-induced hyperplastic nodules
(Richmond et al., 1995). The expression of six histochemical markers of neoplastic cells (p21
ras, p39 c-jun, p55 c-fos, aldehyde dehydrogenase, glutathione S-transferase, and alpha
fetoprotein) were examined by immunohistochemical and image analysis methods. The
hyperplastic nodules were identified as having preneoplastic characteristics, while altered
hepatic foci did not have preneoplastic characteristics. These observations were reported to be
consistent with results obtained for DCA-induced hepatocarcinogenesis in B6C3FJ mice (Daniel
etal., 1992).
36
-------
4.3. REPRODUCTIVE/DEVELOPMENTAL STUDIES
Mice
Unlike those studies in rats (discussed below), studies regarding potential reproductive
effects of DC A exposure in mice have been limited to in vitro methodology. Dichloroacetic acid
was found to inhibit in vitro fertilization of B6D2FJ mouse gametes by Cosby and Dukelow
(1992). The percent of gametes fertilized dropped from 87.0% (controls) to 67.3% or 71.8%
with exposure to 100 or 1,000 mg/L DCA, respectively. A study by Hunter et al. (1996) exposed
CD-I mouse whole-embryo cultures to 0 to 14.7 mM DCA for 24 hours. The study authors
found significant increases in neural-tube defects at treatment concentrations of 5.9 mM and
above, heart and pharyngeal arch defects were seen at concentrations of 7.3 mM and above and
eye defects, rotational defects and somite dysmorphology at concentrations of 11 mM and above.
In a follow-up study to the whole embryo culture study performed by Hunter et al.
(1996), Ward et al. (2002) investigated cell-cycle disruptions in mice neurulation-stage (gd8)
embryos exposed to 11 mM DCA for 6, 12, 18, or 24 hours. Dichloroacetic acid caused a slight,
but not statistically-significant, increase in the number of heart cells in S phase and a slight
decrease in those cells in Gl phase (measured by flow cytometry), compared to controls.
Dichloroacetic acid induced a statistically-significant increase in sub-Gl events (defined as
hypodiploid peaks and cells or cellular debris with less than 2n copies of DNA) in embryos
incubated > 12 hours, which was interpreted to be an increase in the induction of apoptosis. This
effect was consistent in the head, heart, midpiece, and hindpiece regions of embryos exposed to
DCA for 24 hours.
Bis I, an inhibitor of protein kinase C, did not induce sub-Gl events while staurosporine,
a nonspecific protein kinase inhibitor did. When the alteration in apoptosis was analyzed using
fluorescence microscopy, DCA treatment increased signals in the primordial optic tissue, the
prosencephalon brain vesicle in the embryo, and the branchial arches (fetal gill-like tissue), but
not the heart region. This is consistent with the cell-cycle data and the neural tube defects
observed by Hunter et al. (1996). Bis I did not increase the signal in the brain region. The data
suggest that DCA's inhibition of protein kinase may be a mechanism for apoptosis. However,
the inhibition of protein kinase C is unlikely to be the predominant mediator of DCA-induced
embryotoxicity.
37
-------
Rats
Several studies have been performed to determine the potential reproductive and/or
developmental toxicity of DCA exposure in rats. However, there are no single or multiple-
generation studies of DCA reproductive toxicity.
The testicular toxicity of DCA was evaluated in adult male rats administered both single
and multiple (up to 14 days) oral doses of 0, 18, 54, 160, 480, or 1440 mg/kg-day (Linder et al.,
1997). Delayed spermiation and altered resorption of residual bodies were observed in rats
given single doses of 1,500 and 3,000 mg/kg body weight; these effects persisted to varying
degrees on posttreatment days 2, 14, and 28. Delayed spermiation and formation of atypical
residual bodies also were observed on days 2, 5, 9, and 14 in rats dosed daily with 1,440, 480,
160 or 54 mg/kg-day, respectively. Distorted sperm heads and acrosomes were observed in step
15 spermatids after doses of 480 and 1,440 mg/kg-day for 14 days. Decreases in the percentage
of motile sperm occurred after 9 days at doses of 480 and 1440 mg/kg-day, and after 14 days at
160 mg/kg-day. Increased numbers of fused epididymal sperm were observed on days 5, 9, and
14 in rats dosed with 1440, 480 and 160 mg/kg-day, respectively; other morphologic
abnormalities occurred at 160 mg/kg-day and higher. On day 14, a significant decrease in
epididymal weight was observed at 480 and 1,440 mg/kg-day, and epididymal sperm count was
decreased at 160 mg/kg-day and higher (see also Table 5-6).
Limited, but significant, reproductive toxicity was reported by Bhat et al. (1991)
following the subchronic oral dosing of DCA in male rats. Groups of male Sprague-Dawley rats
(5/group) were administered 0 or 1,100 mg/kg-day DCA in drinking water for 90 days. Body
weights were monitored throughout the study. The animals were sacrificed at 90 days, and
selected organs, including the testes, were isolated for evaluation. Dichloroacetic acid exposure
decreased testis weight (p<0.01) and was associated with signs of tissue atrophy. In addition, the
seminiferous tubules contained very few spermatocytes, and no mature spermatozoa.
In a subchronic toxicity study, Katz et al. (1981) dosed rats (10 to 15/sex/dosage group)
with 0, 125, 500, or 2,000 mg/kg-day of sodium dichloroacetate by gavage daily for 3 months.
Mammary glands, prostate glands, testes with epididymis, ovaries, and uterine horns were
among the large number of tissues examined for histopathological changes. Testicular germinal
epithelial degeneration was seen in 40% of males at 500 mg/kg-day and in all males at 2,000
38
-------
mg/kg-day. In all males at 2,000 mg/kg-day, the testes appeared aspermatogenic and contained
syncytial giant cells in the germinal epithelium, while the epididymis ducts were devoid of
spermatozoa. Syncytial giant cells in the germinal epithelial were seen in 20% of the male rats
dosed at 500 mg/kg-day. No other effects were noted at the 125 or 500 mg/kg-day dose levels.
No effects were noted in the reproductive tissues of female rats. Five rats of each sex that had
received the highest dose were maintained on a normal control diet for 5 weeks after
dichloroacetate treatment had been discontinued. In some of the male rats, there was evidence of
germinal epithelium regeneration (50%) and spermatogenesis (25%).
Toth et al. (1992) also studied the potential reproductive effects in male rats following
subchronic oral exposure to DCA using lower doses than the earlier studies. Male Long-Evans
rats (18 to 19/dose) were administered 0, 31.25, 62.5, or 125 mg/kg-day sodium dichloroacetate
for 10 weeks by oral gavage. Reduced final animal weights relative to controls were observed at
the mid- and high-dose groups. At 31.25 mg/kg-day NaDCA and higher, relative liver weights
increased, while relative kidney and spleen weights and absolute liver weights were increased at
62.5 and 125 mg/kg-day NaDCA. Significant (p<0.05) reductions in the absolute weight of the
preputial gland and epididymis were noted at all dose levels, but the absolute weight of the testis
was not affected at any dose. At the two higher doses (62.5 and 125 mg/kg-day), there were
significant (p<0.05) reductions in the percentage of motile sperm, effects on sperm motion (i.e.,
velocity, linearity, amplitude of lateral head displacement) and reduced epididymis sperm head
counts. At 125 mg/kg-day, animals also had reduced accessory organ (prostate and seminal
vesicle) weights and increased relative testis weights. Histological examination of testis cross
sections did not reveal any gross lesions at any dose, and cellular structures in the epididymis
epithelium appeared normal. Impaired spermiation was noted in 4 of the 10 mid-dose (62.5
mg/kg-day) animals and 9 of the 10 high-dose (125 mg/kg-day) animals, and was attributed to
the retention of late-step spermatids in the seminiferous tubules, as observed histologically. This
finding corroborated the observed reductions in epididymal, but not testicular late-step spermatid
head counts. The fertility of treated males, although reduced in the high-dose group, did not
differ significantly from controls at any dose level. Based on the organ weight changes reported
for the preputial gland and epididymis, as well as impaired sperm formation, a LOAEL of 31.25
mg/kg-day was identified (see also Table 5-6).
Epstein et al. (1992) investigated the time-sensitivity of DCA dosing on the development
of the fetal rat. Pregnant Long-Evans rats were exposed, via oral intubation, to DCA as follows:
1,900 mg/kg-day on consecutive gestation days 6 to 8, 9 to 11, or 12 to 15; single doses of 2,400
39
-------
mg/kg-day on gestation day 10, 11, 12, or 13; or single doses of 3,500 mg/kg-day on gestation
day 9, 10, 11, 12, or 13. No treatment effects on maternal body or organ weights were observed.
Within the 1,900 mg/kg-day exposure group, reduced mean fetal body weight was observed for
days 6 to 8, and increased cardiac malformations for days 9 to 11 and 12 to 15. Single gestation
day exposures increased the incidence of cardiac defects (2.5 to 3.3% and 2.9 to 3.6% at the
2,400 and 3,500 mg/kg-day doses, respectively). Collectively, these studies indicate a
developmental LOAEL of 1,900 mg/kg-day.
Smith et al. (1992) performed a similar investigation into the developmental toxicity of
ingested DC A when administered during organogenesis in the pregnant rat. Pregnant Long-
Evans rats (19-2I/group) were treated by oral intubation with 0, 900, 1,400, 1,900, or 2,400
mg/kg-day DC A on gestational days 6 to 15. Eight dams in the three high-dose groups died
during treatment, appearing anorexic and sluggish prior to death; one death was determined to be
accidental. Maternal weight gain, adjusted for gravid uterine weight, was significantly decreased
to approximately 60% of the control value in all treatment groups. The absolute liver, spleen,
and kidney weights significantly (p<0.05) increased (approximately 13 to 19%, 16 to 28%, and
12 to 18%, respectively, compared to the control) in all dose groups, with corresponding
hypertrophy in these organs. The mean percentage of resorbed implants per litter was
significantly elevated in all treated dose-groups. The number of live fetuses/litter was
significantly reduced by 27% at 2,400 mg/kg-day. All dose groups exhibited significant, but
relatively small, dose-dependent reductions in fetal weight (approximately 89% of control group)
and fetal crown-rump length (75 to 86% of the control group). There was a significant increased
incidence (dose-dependent) of soft tissue and cardiovascular anomalies in all treatment groups,
and of external malformations beginning at the 1,400 mg/kg-day group. No skeletal
malformations were observed.
In a second experiment by Smith et al. (1992), pregnant Long-Evans rats (19-20/group)
were administered 0, 14, 140, or 400 mg/kg-day DC A by gavage on gestational days 6 to 15. A
significant decrease in maternal weight gain, adjusted for gravid uterine weight, was found in the
mid- and high-dose dams (63 and 77% of control, respectively), as well as an increase in spleen
and kidney weights at the highest dose. Absolute liver weight was significantly elevated for all
dose groups compared to the control group, with 3, 8, and 14% increases observed, respectively.
Dose-related hypertrophy in the liver, spleen and kidneys was reported in the two high-dose
groups (no incidence data). Reduced fetal crown-rump length (5% decrease) and fetal body
weight (7% decrease) were significant in the high-dose group. A dose-related increase in soft
40
-------
tissue anomalies, primarily cardiovascular, was reported in the 140 and 400 mg/kg-day groups.
The increase in soft tissue abnormalities was significant for the two highest dose groups and the
cardiac abnormalities for the highest dose. An intraventricular septal defect between the
ascending aorta and the right ventricle was most commonly observed with less frequent
urogenital defects (bilateral hydronephrosis and renal papilla) and defects of the orbit also
reported. Collectively, these studies determined a NOAEL of 14 mg/kg-day and a LOAEL of
140 mg/kg-day DCA for developmental effects (soft tissue anomalies) and maternal effects
(reduced body weight and organ hypertrophy).
Moser et al. (1999) investigated the chronic-duration neurotoxic effects of DCA in
weanling rats. Rats were exposed via drinking water to 2.5 or 3.5 g/L DCA for 24 months.
However, exposures to the high dose were discontinued before the study ended because of
excessive toxicity. In addition, the low dose was decreased at 6 weeks and at 10 weeks.
Estimated intake levels over the exposure period were 235 mg/kg-day (for 6 months) and 137
mg/kg-day (for 24 months) for the high- and low-dose groups, respectively. Severe gait
abnormalities, decreased hind limb grip strength, righting deficits, and tremors (>50% incidence)
were evident in both dose groups throughout the 2-year period. Recovery was not evident even
18 months after exposure ended in the high-dose group. Treated rats also showed decreased
forelimb grip strength, chest clasp, and an inhibited pupil response.
The potential developmental toxicity of DCA was studied in vitro using a rat whole
embryo culture system (Saillenfait et al., 1995). Groups of 10 to 20 explanted embryos from
Sprague-Dawley rats were cultured for 46 hours in 0, 1.0, 2.5, 3.5, 5.0, 7.5, or 10 mM DCA. A
significant, dose-dependent decrease in crown rump length was seen at 3.5 mM and above, while
significant, dose-related decreases in yolk sac diameter, head length, somite (embryonic
segment) number, protein content, and DNA content were seen at 2.5 mM and above. In
addition, several defects which were nonexistent in the 0 and 1.0 mM groups were present to a
substantial degree in the higher dose groups. At 2.5 mM, 30% of the embryos had brain defects,
45% had eye defects, and 10% had reduced embryonic axis. At 3.5 mM, 95% had brain defects,
75% had eye defects, 80% had reduced embryonic axis, 15% had reduced first branchial arch,
40% had otic system defects, and 15% had defective flexion. The results indicated a teratogenic
effect from DCA in this system.
41
-------
Doss
Two subchronic studies in the beagle dog examined endpoints relevant to reproductive
toxicity. In the first study (Katz et al., 1981), male and female beagle dogs (3 to 4
animals/sex/dosage group) were administered oral doses of 0, 50, 75, or 100 mg/kg-day of
sodium dichloroacetate by gavage for 13 weeks. Prostate gland atrophy and testicular changes
(degeneration of germinal epithelium, vacuolation of Ley dig cells, formation of syncytial giant
cells) were observed in all treated males. These effects were qualitatively judged by the authors
to be dose-dependent (no data provided). After a 5-week recovery period in one male, the
prostate appeared normal and there was evidence of germinal epithelium regeneration with
spermatogenesis.
In the second study (Cicmanec et al., 1991), four-month-old male and female beagle dogs
(5 animals/sex/dose) were administered 0, 12.5, 39.5, or 72 mg/kg-day of dichloroacetate in
gelatin capsules for 90 days. Testicular changes were reported in the males at all dose levels
(except for control), including syncytial giant cell formation and degeneration of testicular
germinal epithelium. Severity of the lesions increased in the mid- and high-dose animals (see
also Table 5-5). Prostate glandular atrophy characterized by a significant reduction of glandular
alveoli was also noted in mid-and high-dose groups. The testes of affected males did not show
lesions upon gross necropsy. Absolute and relative testicular weights were unaffected by DCA
treatment. A reproductive LOAEL of 12.5 mg/kg-day, the lowest dose tested, was established in
this study. Data on the nonreproductive endpoints examined in this study are provided in
Section 4.2.1.
4.4. OTHER STUDIES
4.4.1. Mechanistic Studies
A number of studies have evaluated the mechanism of action for DCA toxicity. Most,
however, have concentrated on possible mechanisms for carcinogenicity rather than noncancer
effects. Studies performed to elucidate the mechanism of toxicity of DCA have included in vitro
and in vivo analyses with endpoints such as cell death, cell communication, response to growth
factors, and the formation of tissue or DNA lesions.
42
-------
In an effort to shed light on the cellular events preceding the development of malignant
liver tumors in male B6C3FJ mice, Carter et al. (2003) examined 1,355 slides from liver samples
from 327 animals used by DeAngelo et al. (1999). Tissues collected from mice sacrificed
throughout the DeAngelo et al. (1999) study were used to evaluate the effects of dose (0, 0.05
0.5, 1.0, 2.0 and 3.5 g/L DCA) and time (26, 52, 78 and 100 weeks) on liver lesions. Slides were
processed for standard histological examination and were evaluated for the occurrence of altered
hepatic foci (AHF), large foci of cellular alteration (LFCA, formerly called hyperplastic
nodules), adenomas (AD) and carcinomas (CA). In order to minimize interhuman variability in
the classification of tissue abnormalities, all of the slides were read by two observers who were
blinded to treatment group and time-of-sacrifice.
In addition to the four main categories described above, lesions were subcharacterized
into three groupings as follows: eosinophilic, dysplastic, and basophilic and/or clear cell.
Eosinophilic cells showed increases in smooth endoplasmic reticulum and mitochondria. The
dysplastic cells displayed atypical or enlarged nuclei. Tissue lesions from all four major
categories and all three subcategories were identified in liver tissues from control and exposed
animals. The basophilic cells had increased rough endoplasmic reticulum and/or ribosomes.
The clear cells had accumulation of glycogen and/or lipids (steatosis).
When the histological examination of the tissues was completed, the data were arrayed
by dose and time-to-sacrifice and reexamined to determine if there was a pattern of lesion
progression with either dose or duration of exposure. The observed patterns of lesion frequency
and their progression across the time- and dose-range gave rise to the hypothesis that there were
three possible routes to the development of malignant tumors. In one case, eosinophilic cells
seemed to progress from eosinophilic AHF to eosinophilic AD and CA. The basophilic cells and
clear cells showed two patterns of progression. They either progressed from AHF to LFCA and
then to CA or from LFCA to AD and then to CA. The dysplastic cells seemed to progress
directly from AHF to CA. All three patterns of lesion progression were observed in the livers of
mice treated with DCA and were significantly different from controls at some time or dose
points. The majority of the cancers arose from the basophilic/clear cell progression.
The researchers also examined the relationship of necrosis, glycogen accumulation,
cytomegaly, accumulation of lipid droplets, atypical nuclei, and enlarged nuclei to malignancies
(Table 4-1). The strongest correlation was observed for cytomegaly. A correlation with
glycogen accumulation and necrosis was observed for some doses but there was no consistent
43
-------
dose-response pattern. The lack of dose-response at the high doses may be due to a decrease in
the amount of liver tissue (non-involved liver) that had not been impacted by the tumors (AD or
CA). While clear cells (lipid containing) were negatively correlated to the length of DC A
exposure, this finding is consistent with the hypolipidemic effects of DCA.
Table 4-1. Biomarkers of tissue DCA exposure:
incidence (%) of altered hepatic histology
Dose/Tissue Abnormality
Cytomegaly
Glycogen
Steatosis
Necrosis
Atypical nuclei
Enlarged nuclei
CA
Control
1.2
3.8
26.3
2.5
22.5
41.2
7.5
0.05 g/L
0
0
66.7
6.1
33.3
39.4
15.2
0.5 g/L
0
20
34.5
1.8
32.7
54.5
10.9
1.0 g/L
30.8
10.8
21.5
20
55.4
55.4
20
2.0 g/L
41.2
11.8
0
11.8
58.8
45.1
39.2
3.5 g/L
34.9
27.9
7
30.2
46.5
41.9
37.2
Source: Adapted from Carter et al. (2003).
Bruschi and Bull (1993) used hepatocyte suspensions from male B6C3FJ mice and
Sprague-Dawley rats to investigate the possible role of cytotoxic effects in DCA-induced
hepatocarcinogenicity. Cytotoxicity was measured by the release of lactate dehydrogenase,
trypan blue exclusion by the exposed cells,and depletion of intracellular reduced glutathione. No
effects were seen in DCA-treated cells of either species using concentrations up to 5.0 mM and
exposure times up to 240 minutes, suggesting little cytotoxicity from exposure to DCA as
measured by the biomarkers employed.
Cellular changes that might indicate the potential mechanism of DC A-induced
hepatotoxicity and carcinogen!city were studied in two parallel sets of experiments using the
same strain of male mouse and an identical dosing regimen. In the first set of analyses, Carter et
al. (1995) dosed male E6C3Fl mice with 0, 0.5, or 5 g/L (0, 95, or 440 mg/kg-day, respectively)
of DCA in drinking water for up to 30 days in two phases: Phase I was 5-15 days of treatment
and Phase II was 20-30 days of treatment. Thymidine incorporation in hepatic DNA was
measured by administering [3H]-thymidine by a mini osmotic pump for 5 days prior to sacrifice.
Groups of five animals were sacrificed at 5-day intervals. Significant, dose-related increases in
absolute and relative (to total body weight) liver weights were seen at each 5-day interval. These
44
-------
trends increased with the length of exposure. Hepatocytes from Phase I animals in the high-dose
group exhibited reduced thymidine incorporation (labeling index) and inhibition of mitosis. In
Phase II, a decrease in the labeling index was observed among the low- and high-dose groups.
Differences from the control group were significant at 20 and 25 days, but not at 30 days. Both
treatment groups had enlarged nuclei, which may suggest polyploidy. The hepatocytes also
exhibited glycogen accumulation, suggesting alterations in cellular metabolism. The authors
concluded that DCA exposure initially inhibits rather than stimulates cell proliferation
(hyperplasia), and that the increased liver weight is due to hepatocyte enlargement rather than
regenerative hyperplasia following cell death.
A second segment to the Carter et al. (1995) project examined the role of apoptosis
(programmed cell death) suppression as a contributing factor to DCA-induced
hepatocarcinogenicity. The results were published as Snyder et al. (1995). Apoptotic cells were
visualized by in situ nick-end labeling of DNA from the livers of animals sacrificed at 5-day
intervals. Regression analysis revealed a significant trend toward decreased apoptosis as the
dose and length of exposure increased. The lowest dose, 0.5 g/L, was shown to significantly
(p<0.05) decrease apoptosis at the earliest time point (5 days) and also at days 15, 25, and 30.
For the high-dose group, apoptosis was significantly depressed as compared to controls for all
time points except the 20-day point. The authors suggested that DCA may suppress the
apoptotic mechanism by which initiated tumor cells would otherwise be removed.
Benane et al. (1996) examined the effects of 1-, 4-, 6-, 24-, 48-, and 168-hour exposures
to DCA (0, 5, 10, or 50 mM) on gap junction intercellular communication in Clone 9 cell
cultures (normal rat hepatocytes). No differences in intercellular communication were seen
between the 5 mM groups and controls, as measured by a dye transfer protocol, but there was a
difference between all 50 mM groups and controls. The shortest exposure time and lowest
exposure concentration which significantly reduced dye transfer was for the 6-hour, 10 mM
group. A 41 mM DCA concentration produced a 50% reduction in dye transfer over a 24-hour
period. The significance of the disruption in intercellular communication has not been
elucidated, but DCA's ability to disrupt communication was much weaker (>5.8-fold) than other
chlorinated compounds tested, including: perchloroethylene, trichloroacetic acid,
trichloroethanol, and chloral hydrate.
Tsai and DeAngelo (1996) examined the effects of DCA administered to male B6C3FJ
mice on the subsequent responsiveness to growth factors of isolated hepatocytes in culture.
45
-------
Mice were administered drinking water with 0 or 3.5 g/L DCA for up to 90 days. Incorporation
of [3H]-thymidine in the presence of epidermal, hepatocyte, or acidic fibroblast growth factors
was then measured in the isolated hepatocytes, with or without the mito-inhibitory transforming
growth factor Px. Inhibition of basal DNA synthesis was noted in cells isolated from animals
exposed to DCA for 30, 60, or 90 days. However, cells from DCA-treated mice that were treated
in culture with growth factors exhibited enhanced DNA synthesis similar to that seen in cultured
cells from control mice. The authors suggested that the early depression of cell proliferation
seen in other studies of DCA-induced tumorigenesis is due to some mechanism other than an
impaired ability to respond to growth factors.
As knowledge of the complex sequence of cytosolic and nuclear events that influence
neoplasia increases, it is important to consider data on genetically-linked events including
changes in the genetic messages (proto-oncogenes) for transcription factors and signal
transduction proteins. The data base for DCA includes the results of several studies that
examined the ras signal transduction genes, plus the juntos and myc transcription factors.
Anna et al. (1994) exposed male B6C3FJ mice to drinking water containing 0 (50
animals) or 5 g/L DCA (110 animals, about 900 mg/kg-day), 5 days/week for 76 weeks. Mice
treated with DCA had an increased incidence of both hepatic adenomas (93% of treated mice had
at least one adenoma vs. 8% positive for control animals), as well as hepatocarcinomas (74% of
the treated mice had at least one carcinoma vs. 8% for control animals). There were no
significant differences in H-ras codon 61 mutation frequency among DCA-induced and
spontaneous hepatocellular tumors. However, significant changes were seen in the mutation
spectra of H-ras codon 61 in the DCA-treated mice as compared to the control animals. In the
spontaneous tumors from the controls (study controls plus historical controls) the CAA of codon
61 became AAA in 59 % of the tumors, CGA in 28% and CTA in 14%. In the DCA-treated
mice, the H-ras codon 61 changes were 28% AAA, 35% CGA and 38% CTA. The authors
suggest that these differences were due to nonspecific secondary DNA damage by DCA. The
authors further suggest that DCA exposure, while not necessarily causing mutations in the H-ras
gene, may nevertheless provide a selective growth advantage to mutations that arise
spontaneously. Ras proteins are GTPases that are involved in the activation of a series of protein
kinases that control cell growth and differentiation. Ras is activated by binding of a ligand to a
cell surface receptor.
46
-------
The findings of Anna et al. (1994) are partially supported by those of Velazquez (1995)
in which DNA was examined from normal liver and tumor tissues obtained from male B6C3FJ
mice that were administered 0.5 g/L (90 mg/kg-day) DCA in drinking water for 2 years.
Sequences of the H-ras gene were amplified using PCR (polymerase chain reaction); it was
observed that H-ras codon 61 mutations were present in three out of five (60%) of the DCA-
induced tumors. In this case, the spectrum of mutations associated with DCA was the same as
that of spontaneous tumors from untreated animals. The significance of this observation is
limited by the fact that there were only three tumors with codon 61 mutations.
In another study, male B6C3FJ mice were administered 1.0 or 3.5 g/L (180 or 630 mg/kg-
day) DCA in drinking water for 104 weeks, and then sacrificed (Ferreira-Gonzalez et al., 1995).
The incidence of liver carcinomas was 19% in the untreated mice, 70.6% in the 180 mg/kg-day
group, and 100% in the 630 mg/kg-day group. DNA samples were examined from 32
spontaneous liver tumors from the control group, 13 tumors from the 180 mg/kg-day group, and
33 tumors from the 630 mg/kg-day group. The DNA was analyzed for K- and H-ras proto-
oncogene mutations in the DCA-induced and spontaneous tumors. Point mutations in exons 1, 2,
and 3 of the K- and H-ras genes were quantified by single-stranded conformation polymorphism.
Similar frequencies of H-ras proto-oncogene exon 2 mutation were found in all three groups
(58% in spontaneous tumors, 46% in 180 mg/kg-day group, and 50% in the 630 mg/kg-day
group). Mutation frequencies in other exons were minimal.
Comparative sequence analysis of exon 2 mutations from spontaneous and DCA-induced
tumors revealed a substantial shift in the spectrum of base changes in codon 61. Sequence
analysis of spontaneous tumors revealed changes in codon 61 from CAA to AAA in 80% and
CAA to CGA in 20% of the examined tumors (Table 4-2). No CAA to CTA conversion was
observed in spontaneous tumors. In contrast, the frequency of CAA to AAA conversion was
16% and 21% at DCA doses of 180 and 630 mg/kg-day, respectively. CAA to CGA conversion
was noted in 50% of the tumors from mice treated with either 180 or 630 mg/kg-day, and CAA
to CTA conversion was observed in 34% and 29% of the two dosage groups, respectively. Thus,
although DCA-induced and spontaneous tumors involved similar levels of H-ras mutation, the
mechanisms of tumor induction may be different. Differences in codon 61 mutation spectra
between spontaneous and DCA-induced tumors in this study are similar to those reported in the
Anna et al. (1994) study, where there was also a lower number of CAA to AAA conversions and
a higher number of CAA to CTA conversions in the DCA-induced tumors as opposed to the
spontaneous tumors.
47
-------
Table 4-2. Frequency of spontaneous and DCA-induced mutations of codon 61 in
exon 2 of the H-ras oncogene mutations in B6C3F, mice
Dose (mg/kg-day)
Untreated (spontaneous mutations)
180
630
Mutation Frequency (%)
AAA
80
16
21
CGA
20
50
50
CTA
~
34
29
Source: Adapted from Ferreira-Gonzalez et al. (1995).
Schroeder et al. (1997) examined DCA-induced tumors in female B6C3FJ mice for H-ras
codon 61 mutations. There was a H-ras mutation in only one of 22 tumors. However, this one
mutation was consistent with the observations of other researchers in that it involved a CAA to
CTA conversion.
Stauber et al. (1998) demonstrated that DCA increases cell proliferation ofc-Jun positive
hepatocytes in vitro. As mentioned previously, c-Jun is a nuclear transcription factor that is
associated with apoptosis and cell transformation. Its expression is linked to the H-ras signal
transduction cascade (Johnson et al., 1996). The investigators treated isolated hepatocytes from
neonatal mice with DCA and plated the cells to allow them to form colonies. Exposure of the
cells to 0.5 mM DCA significantly increased colony formation (no cytotoxicity) over controls.
Interestingly, the colonies that were induced by DCA were c-Jun positive. This is noteworthy
because this is the same phenotype observed in DCA-induced liver tumors in whole mice
exposed to DCA (Stauber and Bull, 1997). When mice were pretreated for 2 weeks with DCA in
their drinking water prior to preparation of hepatocytes, DCA again induced c-Jun positive
colony formation, but only required 0.02 mM DCA for the same degree of induction.
While Pereira et al. (2001) investigated the effect of DCA treatment on proto-oncogene
gene expression in the liver, the study considered the effect of DCA treatment on the
hypomethylation and expression of the c-myc gene and the promotion of liver and kidney
tumors. The c-Myc gene is a nuclear protein that is involved in transcriptional response and
proliferation of liver cells. Hypomethylation of the c-myc gene seems to enhance its expression
and thus cell division.
48
-------
In the first of two experiments by Pereira et al. (2001), 7- to 8-week-old female B6C3FJ
mice were administered 400, 800, or 1600 mg/L chloroform in drinking water for 17 days. On
the last 5 days of treatment, the mice were also administered 500 mg/kg-day of DC A via gavage.
Methylation of the c-myc gene was determined by enzymatic DNA hydrolysis using a Hpall
restriction endonuclease to digest unmethylated CCGG sites combined with Southern blot
analysis. Gene expression was evaluated using Northern blot analysis for c-myc mRNA.
Dichloroacetic acid decreased c-myc methylation and increased expression of the gene more than
chloroform. Doses >800 mg/kg-day chloroform, coadministered with DC A, significantly
reduced the ability of DC A to increase gene expression.
In the second experiment, five-week-old male and female B6C3FJ mice were
administered 3.2 g/L DCA in drinking water, either alone, or in conjunction with 800 or 1600
mg/L chloroform (Pereira et al., 2001). Prior to DCA exposure, the mice had been initiated with
a single (300 mg/kg) intraperitoneal dose of MNU at 15 days of age. The mice were sacrificed at
36 weeks of age. Greater numbers of hepatic foci were observed in DCA-treated animals
(females more than in males). The tumor response was greater in males than in females.
Chloroform in conjunction with DCA at both doses drastically reduced the adenomas and
adenocarcinomas. One interesting effect of combining DCA exposure with chloroform in the
MNU-treated mice was the occurrence of kidney tumors. While treatment alone produced few
tumors in the kidney, coexposure with chloroform increased the tumor multiplicity.
Thai et al. (2001) investigated changes in early gene expression in mice liver following
DCA exposure. Four-week-old mice were administered 2 g/L DCA in drinking water for 4
weeks. Differential display of mRNA levels revealed that 381 genes showed differences in
intensity of the display between the exposed mice and the controls. Upon further refinement of
the data, six genes were identified that were expressed differently in control and exposed mice
(one gene induced, the other five suppressed). Four genes were identified: stearoyl-CoA
desaturase was induced, while alpha-1 protease inhibitor, cytochrome b5, and carboxylesterase
were suppressed. All but alpha-1 protease inhibitor are endoplasmic reticular enzymes involved
in fatty acid metabolism. Four of the six genes were found to be similar in hepatocellular
carcinomas (from additional mice treated with 3.5 g/L DCA for 93 weeks) and in the livers of
mice treated with DCA for 4 weeks. The identified genes that were similar in the tumors and the
DCA-treated mice were those for alpha-1 protease inhibitor, cytochrome b5, carboxylesterase,
and an unnamed gene. The expression of stearoyl-CoA desaturase and one other identified gene
were the same in the control mice and the tumors. The significance of these findings is
49
-------
unknown, relative to the carcinogenic properties of DC A. Changes in the expression of some of
the genes may merely reflect metabolic perturbations induced by DCA rather than cancer-linked
events.
A second study by the same researchers (Thai et al., 2003) utilized the same mouse liver
tissue samples and gene microarrays. The first array was the Atlas Mouse Cancer 1.2 Array,
which contains 1,176 unique complementary cDNA fragments from genes known to be
implicated in cancer development (Clontech, 2000). The second was an array of 140 genes
representing mouse stress/toxicity response elements. There were approximately 50 genes that
were common to the two microarrays. In the mouse stress/toxicity array there were 13 genes that
were differentially expressed: five showed increased expression and 8 decreased expression.
From the mouse 1.2 cDNA array, 11 genes were differentially expressed; the expression of two
was increased while expression of nine was decreased. Thai et al. (2003) considered the results
of their 2001 and 2003 analyses and concluded that the affected genes were related to three cell-
response groupings: tissue remodeling and/or angiogenesis, xenobiotic metabolism, and damage
response. Most of the genes in each of these groupings were suppressed. The authors
hypothesized that the suppressed gene expression in the tissue remodeling and angiogenesis
group plus the tissue repair grouping facilitated tumor growth. The GSTZ gene and other genes
involved in glycogen and lipid metabolism were not present in the microarrays that Thai et al.
(2003) employed. The PPARa gene was present but not activated in the microarray.
On the other hand, the finding that DCA induced peroxisomal enzymes in some studies,
suggested that the PPARa gene can be activated. Since peroxisomes generate hydrogen
peroxide through some of their metabolic reactions, they are often associated with oxidative
change to cellular DNA. Austin et al. (1996) investigated the potential for DCA to increase
intercellular lipid peroxidation and the oxidation of DNA. Male B6C3FJ mice were treated with
a single oral dose of DCA (0, 30, 100, or 300 mg/kg). Nuclear DNA was extracted at various
times in order to assess increases in relative guanosine hydroxylation. A significant increase was
seen in the 300 mg/kg group from 4 to 6 hours postdosing, but returned to near control levels at 8
hours postdosing. The authors suggest that DNA hydroxylation indicates oxidative stress
induced by DCA in mouse livers. The level of hydroxylation appeared to be related to the ability
to induce thiobarbituric acid-relative substances (TEARS), which is an indicator of lipid
peroxidation. Significant increases in lipid peroxidation have also been shown in cultured
primary rat and mouse hepatocytes following exposure to DCA concentrations as low as 0.5 mM
(mouse) and 1.0 mM (rat; Everhart et al., 1998).
50
-------
4.4.2. Genotoxicity Studies
Observations on DCA
There have been multiple studies investigating the hypothesis that DCA is a genotoxic
agent. Results from in vitro studies are summarized in Table 4-3. The majority of these studies
indicate that DCA is only genotoxic at high doses or after long durations. Most of the in vitro
tests are negative or equivocal, either in the presence or absence of metabolic activation.
While one report indicated that DCA may increase prophage X induction in E. coli
(DeMarini et al., 1994), this finding has not been confirmed by other laboratories and required
DCA concentrations in the mM range to achieve significance. In the Ames assay, DCA has been
evaluated using strains TA98, 100, 1535, 1537, 1538, 1950, 2322, and TS24 in the presence and
absence of S9 activation (Fox et al., 1996; Herbert et al., 1980; Waskell, 1978). The only clear
positive results were reported by DeMarini et al. (1994) using DCA in the vapor phase and strain
TA100. However, the results have not been replicated in other reversion assays and it is possible
that the form of DCA affected the study results. The increased revertants may be the result of
low pH resulting from the use of the free acid in the vapor phase or differences in the membrane
transport of the non-ionized acid.
Herbert et al. (1980) reported an equivocal increase in revertants in strains TA98 and
TA1538 when exposed to 1-10 |_ig/plate DCA (salt); all other strains gave negative results. The
revertant numbers were similar in both the presence and absence of metabolic activation, ranging
from -64 to 102 revertants/plate (compared to negative control values of 59-61 revertants/plate).
The response in TA98 was considered by the study authors to be evidence of a weak mutagenic
effect because the response in TA1538 was not unequivocally dose-related. The results should
not be considered strong evidence for the mutagenic capacity of DCA. Only the results in TA98
were presented in the published paper and they show a slight increase above the spontaneous
reversion rate. The increase (1.4- to 1.7-fold) did not reach the limit (2- to 3-fold) that most
laboratories would typically require for the compound to be identified as mutagenic. The dose-
response trend reached statistical significance, however, in both the absence and presence of S9.
Nevertheless, the authors could not exclude the possibility that the mutagenicity observed was
the result of a contaminant in the DCA.
51
-------
Data from genotoxicity assays using mammalian cells in culture provided negative
results. Harrington-Brock et al. (1998) reported that DCA induces mutations at the thymidine
kinase locus, as well as gross chromosomal aberrations in L5178Y mouse lymphoma cells in
vitro, but the concentrations required to induce these effects were in the mM range.
Additionally, DCA did not induce micronuclei formation in the cells. The authors compared the
dose-response curve of DCA mutagenicity to that of ethylmethane sulfonate, noting that the
mutagenic potency was similar to, but less than, that of the classic mutagen. Recognizing that
their data provides evidence for the mutagenic capacity of DCA, the authors noted that the
compound is unlikely to be a mutagen at concentrations found in finished drinking water
(Harrington-Brock et al., 1998). The results of studies of DNA strand breaks in several cell lines
(Chang et al., 1992), Chinese hamster ovary cell chromosomal aberrations (Fox et al., 1996) and
DNA repair (Waskell, 1978) were negative in the absence of S9 activation. The chromosomal
aberration assay was the only one conducted in the presence of the microsomal S9 factor and
those results were also negative.
52
-------
Table 4-3. Summary of in vitro genotoxicity tests
Assay
A Prophage induction in
Escherichia coli WP2
Bacterial reverse mutation
assay
TS24
TA 2322
TA 1950
TA100
TA 1535
TA 1537
TA 1538
TA98
E. coli WP2uvrA
DNA strand breaks
mouse hepatocytes
rat hepatocytes
human CCRF-
CEM cells
Result
Without S9
-
-
-
-
-
-
+
-
-
-
-
+(a)
-
+
-
-
-Kb)
With S9
+
-
-
-
-
-
+
-
-
-
-
+(a)
-
+
-
-
Concentration
2,500 |ig/mL
31,OOO^ig/mL
31,OOO^ig/mL
31,OOO^ig/mL
5,000 |ig/plate
1-10 |ig/plate
l^g/mL
NA
100-1,500 |ig/mL
(-S9); 1,500-7,500
l^g/mL (+S9)
1-10 |jg/plate
5,000 |jg/plate
1-10 |jg/plate
5,000 |jg/plate
1-10 |ig/plate
5,000 |jg/plate
1-10 |ig/plate
5,000 |jg/plate
5,000 |^g/ plate
2,580 |ig/mL
1,290 |ig/mL
1,290 i^g/mL
Reference
DeMarini et al., 1994
Waskell, 1978
Waskell, 1978
Waskell, 1978
Foxetal., 1996
Herbert etal., 1980
DeMarini et al., 1994
Matsudaetal., 1991
Ciller etal., 1997
Herbert etal., 1980
Foxetal., 1996
Herbert etal., 1980
Foxetal., 1996
Herbert etal., 1980
Foxetal., 1996
Herbert etal., 1980
Foxetal., 1996
Foxetal., 1996
Chang etal., 1992
Chang etal., 1992
Chang etal., 1992
53
-------
Assay
L5178Y/TK+/- mouse
lymphoma mutation assay
Chromosome aberration
assay (Chinese hamster
ovary cells)
Newt micronucleus test
DNA repair
Repair deficient strains
TA1535 (umu operon)
E.coli PQ37
Result
Without S9
+(c)
-(d)
-
NA
+
With S9
NA
NA
-
-(e)
NA
+
Concentration
5,000 |ig/mL
600-800 |ig/mL
5,000 |ig/mL
20, 40, 80 |ig/mL
31 mg/plate
58.5 [ig/mL
500 |ig/mL
Reference
Foxetal., 1996
Harrington-Brock et
al., 1998
Foxetal., 1996
Ciller etal., 1997
Waskell, 1978
Onoetal., 1991
Ciller etal., 1997
NA=not applicable/not available
(a) The results in TA1538 were positive but did not "exhibit an unequivocal dose-response relationship" (Herbert et
al., 1980).
(b) Small increase in strand breakage (7%) seen after 4-hour exposure, but not at 1 hour; response deemed
negligible.
(c) Mutations/chromosome aberrations.
(d) Micronuclei induction.
(e)Test is performed using stage 53 newt larvae in the absence of exogenous S9; any metabolic activation is from the
test animal.
Results from in vivo studies are shown in Table 4-4. In this case, results are mixed with
no consistent pattern of positive or negative results for mouse micronucleus assay, DNA strand
breaks in mouse and rat cells, or DNA adduct formation. In particular, DCA has been
investigated in vivo for its ability to induce single-strand breaks in DNA. Chang et al. (1992)
exposed B6C3FJ mice to drinking water containing 0.05, 0.5, or 5.0 g/L DCA for 7 and 14 days
while F344 rats were exposed to drinking water containing 0.05, 0.5 or 2 g/L DCA for 30 weeks.
Analysis of damaged DNA was conducted in hepatocytes as well as in epithelial cells taken from
the spleen, stomach and duodenum. Consistent with their in vitro results, the authors reported no
evidence of increased DNA strand breakage at any dose tested in mice or rats. While the authors
reported a 7% increase in strand breaks in mice dosed in vivo, they considered this result to be
insignificant.
The findings by Chang et al. (1992) are in direct opposition with earlier work published
by Nelson et al. (1989) and Nelson and Bull (1988). In these two studies, DCA exposure
significantly increased DNA single-strand breaks in the livers of mice and rats. It is interesting
54
-------
to note that the reported DNA damage occurred at an oral dose of 10-13 mg/kg-day, almost 2
orders of magnitude lower than the doses used in the Chang et al. (1992) study. The basis of the
differences in results between laboratories is not clear but may be the result of differences in
methodology. The results of Nelson et al. (1989) and Nelson and Bull (1988) may reflect the
rapid repair of hydroxylated guanines (Austin et al., 1996), which require the formation of
single-strand breaks. Fuscoe et al. (1996) reported no significant increase in DNA migration
(evidence of DNA strand breaks) in mice exposed in vivo up to the highest concentration of 3.5
g/L. At the highest dose, however, there was a reduction in migration rates that was interpreted
to be evidence of DNA cross-linking. This result is in contrast with the negative findings of
other assays, which can also measure DNA cross-linking (Chang et al., 1992; Fox et al., 1996).
Table 4-4. Summary of in vivo g
Assay
Micronuclei assay
(mouse)
DNA Strand Breaks
Mouse leukocytes
Mouse hepatocytes
Mouse splenocytes
Mouse epithelial cells (a)
Fischer rat hepatocytes
Sprague-Dawley rat
hepatocytes
8-OH DNA adducts
Lac I operon
transgenic mutations
Result
+
-
+
+
-
-
-
+
+
+
Concentration or
Dose
3.5 g/L
3.5 g/L
1,000 mg/kg
3.5 g/L
13 mg/kg
10 mg/kg
0.05, 0.5, 5 g/L
(1,290 mg/kg)
5 g/L (1,290 mg/kg)
5 g/L (1,290 mg/kg)
0.05, 0.5, 2 g/L
30 mg/kg
oral 300 mg/kg
up to 2.0 g/L
land 3.5 g/L
enotoxicity tests
Duration
9 days
28 days
3 days
28 days
1 dose
1 dose
7 & 14 days
14 days
14 days
30 weeks
1 dose
1 dose (gavage)
3 & 10 weeks
60 weeks
Reference
Fuscoe etal., 1996
Fuscoe etal., 1996
Fox etal., 1996
Fuscoe etal., 1996
Nelson and Bull, 1988
Nelson etal., 1989
Chang etal., 1992
Chang etal., 1992
Chang etal., 1992
Chang etal., 1992
Nelson and Bull, 1988
Austin etal., 1996
Parrishetal., 1996
Leavittetal., 1997
(a) Epithelial cells from the stomach and duodenum.
In an in vivo micronucleus assay, Fox et al. (1996) exposed Sprague-Dawley rats
intravenously to 275, 550, and 1,100 mg/kg DC A, and did not detect an effect. Fuscoe et al.
(1996) evaluated micronuclei induction in poly- and normochromatic erythrocytes (PCEs and
55
-------
NCEs, respectively) in male B6C3FJ mice following in vivo exposure to drinking water
containing DCA at approximate doses of 95, 190, 380, or 665 mg/kg-day for up to 28 days in
one experiment and 31 weeks in another experiment. In the first study, an increase in
micronucleated PCEs was noted in the high-dose group, but only at day 9. This was apparently a
transient effect and by day 28, the increased incidence of micronuclei was no longer evident.
There was also no increase in micronucleated NCEs. It should be noted, however, that the
control frequency was twice as high in the 28-day study as in the 9-day study, which might have
affected the ability of the assay to detect slight increases in micronuclei.
Fuscoe et al. (1996) also measured the effect of tocopherol (vitamin E) administration on
the induction of micronuclei at the high dose (665 mg/kg-day) at both 9 and 28 days (to
determine if increased intracellular oxygen radicals were causing the DNA damage). Vitamin E
treatment had no effect on micronuclei formation. Interestingly, doses of 665 mg/kg-day DCA
plus vitamin E significantly increased micronuclei at 9 and 28 days, when compared to the
vitamin E controls.
In the second experiment, mice were administered 665 mg/kg-day DCA for 10, 26, or 31
weeks (with water administered alone following the exposure period up to sacrifice at 31 weeks).
At each time point, slight but significant increases in NCEs were observed while micronucleated
PCEs slightly increased in a dose-dependent manner, but did not reach statistical significance.
The response was greater for PCEs than for NCEs. Data reflected the much higher control
micronuclei frequency for PCEs than for NCEs (Fuscoe et al., 1996).
Austin et al. (1996) reported increases in the DNA adduct 8-hydroxy-2-deoxyguanosine
in DCA-treated mice. The increase was noted in animals at all doses tested (single doses of 30,
100 or 300 mg/kg DCA), but was statistically significant only in the high-dose group (300
mg/kg) and only at 4 and 6 hours postdosing. This finding was interpreted to indicate the
potential for DCA to oxidatively damage hepatic DNA. In contrast, Parrish et al. (1996), treated
B6C3Fj male mice for 10 weeks with 540 mg/kg-day DCA and saw no evidence of increased 8-
hydroxy-2-deoxyguanosine.
Leavitt et al. (1997) exposed transgenic mice (Big Blue) to 1 or 3.5 g/L DCA
(approximate doses of 190 or 665 mg/kg-day) in their drinking water for 60 weeks. The
concentrations were comparable to those used in chronic bioassays. At interim time points (4 and
10 weeks), neither concentration of DCA induced an increased frequency of mutations in the Lac
56
-------
I loci. However, at 60 weeks, both concentrations of DC A induced a significantly elevated
mutational frequency at this loci. This time-response pattern suggests that the mutational events
might be secondary to toxicological changes in the liver rather than a direct genotoxic effect,
since a direct effect would be expected to be time-independent. The results indicate that a large
cumulative dose (due to the 60-week exposure period) is necessary to increase mutations in this
in vivo system. A second complicating issue regarding this study is the clonal expansion of
preneoplastic and neoplastic cells within the tissue, which may account for the apparent increase
in mutation rate at 60 weeks (WHO, 2000). Although, the investigators accounted for this
potential confounder by analyzing the type of mutation (i.e., base substitutions) and subtracting
duplicate identical mutations recovered in the same animal, the proportion of mutation types
recovered from control and treated mice were still statistically different after the adjustment.
The study authors, however, did not provide any justification for this correction.
Observations on DCA Metabolites
In contrast to the findings reported for DCA, glyoxylate has been shown to be mutagenic
in four independent studies (Marnett et al., 1985; Sasaki and Endo, 1978; Yamaguchi and
Nakagawa, 1993; Sayato et al., 1987). However, the concentrations of glyoxylate required to
produce positive results are very high (in the mM range) and it is not known whether these can
be reasonably achieved in vivo from the metabolism of DCA. Consequently, it is uncertain
whether the results are likely to be relevant to the issue of DCA genotoxicity. Haworth et al.
(1983) reported that oxalate was not mutagenic in Salmonella.
Summary
The genotoxicity/mutagenicity of DCA has been investigated in a number of studies.
The preponderance of in vitro studies are negative, with only a few equivocal or positive results.
Studies in vivo are mixed, with internally inconsistent results between studies and between
endpoints. The difference in results does not appear to be clearly related to differences in
exposure levels. While Leavitt et al. (1997) found an increased frequency of mutations in the
Lac I loci after exposure for 60 weeks, this was not observed at interim time points (4 and 10
weeks). The findings suggest that duration of exposure may be an important variable. The
importance of these findings and the potential relevance to the issue of DCA carcinogenesis is
further discussed in Section 4.6.
57
-------
4.5. SYNTHESIS AND EVALUATION OF MAJOR NONCANCER EFFECTS
The noncancer effects of DC A can be grouped into one of four major categories:
metabolic alterations, hepatic toxicity, reproductive/developmental toxicity, and neurotoxicity.
Important observations on each of these effect categories are briefly summarized below.
4.5.1. Metabolic Alterations
Multiple independent studies demonstrate that DCA has the ability to alter normal
carbohydrate metabolism. Dichloroacetic acid treatment results in a significant reduction in
plasma levels of glucose, pyruvate, and lactate. This finding has been consistently reported in
DCA-treated rats, dogs and humans (Ribes et al., 1979; Katz et al., 1981; Evans and Stacpoole,
1982; Davis, 1986, 1990; Stacpoole et al., 1978). The primary mechanism of action associated
with the decrease in blood glucose and lactic acid appears to be enhancement of pyruvate
dehydrogenase activity (Crabb et al., 1981; Stacpoole, 1989). This is the enzyme that
decarboxylates pyruvate and provides acetyl-CoA for the citric acid cycle. Activation of
pyruvate dehydrogenase occurs indirectly by DCA inhibition of the protein kinase that maintains
it in its inactive form (Whitehouse et al., 1974; Stacpoole, 1989). By stimulating the pyruvate
dehydrogenase, DCA accelerates pyruvate, lactate, and alanine oxidation, and results in plasma
level decreases of these metabolites. As lactate is oxidized first to pyruvate and then to acetyl-
CoA by pyruvate dehydrogenase, there is a corresponding decrease in the hydrogen ions that
exist in a 1:1 stoichiometry with lactate, and the subsequent generation of bicarbarbonate ions.
This serves as the basis for using DCA in the treatment of severe cases of lactic acidosis
(Stacpoole et al., 1998a). Similarly, by the removal of lactate and alanine, two major
gluconeogenic substrates, DCA inhibits hepatic glucose output and induces the resulting
decrease in circulating glucose levels (Stacpoole et al., 1998a).
A potential confounder in the interpretation of some studies of the effect of DCA on
metabolism is an exposure-related decrease in water and food consumption, especially in the
high-dose groups (Bhat et al., 1991; Katz et al., 1981; Davis, 1986; Mather et al., 1990; Yount et
al., 1982). A drop in water and food consumption could contribute to the reported decreases in
body weight observed in some high-dose group animals and could potentially impact glucose
metabolism as well. The relationship between nutritional status and aerobic glycolysis is
58
-------
complex, but glucose metabolism could clearly be modified by a significant decrease in the
caloric intake of treated animals. However, changes in plasma levels of glucose or lactic acid
have been seen in DCA-treated humans with no associated weight loss. Additionally, metabolic
changes have been consistently observed in DCA-treated animals at doses below those resulting
in body weight changes. Therefore, metabolic effects are not artifacts of altered food or water
intake.
Dichloroacetic acid exposure also results in a decrease in plasma cholesterol levels. This
has been observed in experimental animals and humans, and was even briefly exploited
therapeutically in the treatment of a few individuals with hypercholesterolemia (Moore et al.,
1979; Stacpoole et al., 1978). Dichloroacetic acid has been shown to be a noncompetitive
inhibitor of the rate-limiting microsomal enzyme in cholesterol biosynthesis,
hydroxymethylglutaryl (HMG) CoA reductase (Stacpoole, 1989). It also inhibits hepatic
synthesis of triglycerides by an unknown mechanism (Stacpoole and Greene, 1992). The net
effect of these inhibitory activities is a decrease in serum lipids and lipoproteins in vivo
following DCA dosing. All DCA-induced metabolic alterations appear to be transient, with full
recovery to basal/control levels observed following cessation of DCA administration.
Cornett et al. (1999) demonstrated that DCA can significantly alter tyrosine metabolism
as a consequence of its inhibitory effect on GSTZ. Inhibition of tyrosine metabolism can result
in increased levels of reactive tyrosine metabolites such as maleylacetoacetate and
maleylacetone, metabolites that may adversely affect the heart, liver and nerves, targets of DCA
toxicity. In humans, hereditary tyrosinemia II (a disease involving a deficit in tyrosine
metabolism) is often associated with the development of polyneuropathy, and/or hypertrophic
cardiac myopathy in young patients (Tanguay et al., 1996; LaBerge et al., 1986).
4.5.2. Hepatic Toxicity
Another consistent finding in DCA ingestion studies is a dose-related increase in liver
size (DeAngelo et al., 1999; Sanchez and Bull, 1990; Yount et al., 1982; Mather et al., 1990;
Smith et al., 1992), generally accompanied (or caused) by an increase in glycogen deposition in
the liver (Kato-Weinstein et al., 1998; Bhat et al., 1991). The enzymatic basis for increased
hepatic glycogen accumulation remains unclear, although it has been shown that DCA treatment
does not alter glycogen synthetase or the amount of active hepatic phosphorylase (Kato-
59
-------
Weinstein et al., 1998). The increase in liver size and glycogen accumulation resemble changes
occurring in glycogen storage disease, suggesting that failure of glycogenolysis, through either
glycogen phosphorylase or a debranching enzyme, may play a role in the observed accumulation.
The dose-response for glycogen deposition in the liver is in the same range that is required for
inducing hepatocarcinogenesis (Bull, 2000).
The glycogen accumulation and hepatomegaly observed in DCA-treated rats are similar
to changes observed in humans with glycogen storage disease VI. This human genetic disorder
is believed to be the result of a deficiency in liver, rather than muscle, glycogen phosphorylase b
kinase. This kinase is responsible for the conversion of inactive glycogen phosphorylase b to
active glycogen phosphorylase a. The symptoms of glycogen storage disease VI include
accumulation of liver glycogen, liver enlargement and a tendency for development of liver
adenomas and carcinomas (Hers et al., 1989). The disorder is also marked by increased levels of
plasma cholesterol and triglycerides in some subjects (Hers et al., 1989). The increase in plasma
lipids is different from the typical decrease in lipids observed following DCA exposure. It can be
noted that some, but not all, of the cancers observed by DeAngelo et al. (1999) in mice appear to
have originated from the clear cells involved in glycogen storage (Carter at al., 2003).
Liver toxicity, as evidenced by increases in serum levels of liver enzymes, has been seen
in DCA-treated mice, rats, dogs and humans (DeAngelo et al., 1991, 1999; Mather et al., 1990;
Cicmanec et al., 1991; Stacpoole et al., 1998a; Katz et al., 1981). Frank hepatic cytotoxicity in
the form of necrosis has been consistently reported in DCA-treated mice, with exposure levels of
0.5 g/L (-77 mg/kg-day) associated with necrosis of scattered individual hepatocytes, and
exposures of 1-5 g/L (-150 to -1000 mg/kg-day) resulting in larger areas of coagulative necrosis
(DeAngelo et al., 1991; Bull et al., 1990; Daniel et al., 1992; Sanchez and Bull, 1990; ILSI,
1997). Interestingly, frank liver necrosis has not been seen in rats, even at the highest
concentration used (5 g/L) (DeAngelo et al., 1996), nor has it been reported in dogs or humans.
The reason for the preferential severity of the hepatotoxic response in mice is not known.
Sanchez and Bull (1990) suggested that liver necrosis observed in DCA-treated mice was not the
result of DCA-induced hepatocytotoxicity per se, but occurred in infarcted areas caused by
extensive liver hypertrophy.
Another hypothesis is that liver necrosis is secondary to lipid peroxidation. This is
supported by evidence of oxidative damage to hepatic DNA in DCA-treated mice (Austin et al.,
1996). Though direct evidence of lipid peroxidation in the mouse is limited, this observation
suggests the potential for DCA to oxidatively damage the liver. Evidence of lipoperoxidation
60
-------
has also been reported in Fischer 344 rats treated with doses of 300-1000 mg/kg-day (Larson and
Bull, 1992). However, this finding was not confirmed by Mather et al. (1990) who reported no
evidence of lipoperoxidation in Sprague-Dawley rats treated with doses of 50-250 mg/kg-day.
Nevertheless, the fact that necrosis has been observed at concentrations that have not been
shown to induce lipid peroxidation (-77-159 mg/kg-day) argues against this being a significant
mechanism of cell death.
4.5.3. Reproductive/Developmental Toxicity
There is an extensive and consistent data base demonstrating the reproductive toxicity of
DCA in males and females (Katz et al., 1981; Yount et al., 1982; Bhat et al., 1991; Cicmanec et
al., 1991; Toth et al., 1992; DeAngelo et al., 1996; Linder et al., 1997; Smith et al., 1992; Epstein
et al., 1992). Section 4.3 presents further details of these studies. In male rats, DCA may cause
decreases in testicular weight at 1,100 mg/kg and viable sperm production at 62.5 mg/kg-day.
While testicular degeneration was observed in rats at 500 mg/kg-day and dogs at 12.5 mg/kg-
day, it has not been reported in exposed humans. However, testicular effects in humans have not
been specifically examined, because they cannot be readily assessed by noninvasive techniques.
In female rats, DCA exposure to dose levels of 140 mg/kg-day during gestation can lead to
impaired fetal maturation and result in soft tissue anomalies (primarily of cardiac origin) in the
offspring (Smith et al., 1992; Epstein et al., 1992).
To date, no specific cellular or molecular mechanism of action has been proposed to
explain the testicular or developmental toxicity associated with DCA administration.
Dichloroacetic acid can freely and rapidly cross the placenta (Smith et al., 1992); however,
during early organogenesis, the embryo relies almost exclusively on glycolysis for energy, a
process stimulated by DCA (Smith et al., 1992). It is not known if the DCA-mediated effect on
glycolysis may be relevant to the mechanism of action of developmental toxicity. Although the
mechanism by which DCA targets the embryonic heart is not clear, there is evidence that DCA
concentrates in rat myocardial mitochondria (Smith et al., 1992; Kerbey et al., 1976).
4.5.4. Neurotoxicity
Neurologic symptoms and morphologic changes in the nervous system have been
reported in humans, dogs, and rats at comparable doses (when expressed as mg/kg). Reversible
61
-------
peripheral neuropathy has been noted in humans after several months of daily, oral doses of 50 to
100 mg/kg-day (Moore et al., 1979; Spruijt et al., 2001; Stacpoole et al., 1998a). Beagle dogs
treated orally with DCA developed partial paralysis of the hind limbs at doses of 72 mg/kg and
above (Cicmanec et al., 1991; Katz et al., 1981). Morphologic alterations in the CNS, including
vacuolization of the myelinated white tracts in the cerebellum, cerebrum, and spinal cord, were
observed in dogs at doses of 12.5 to 72 mg/kg-day. In rats, dose-limiting toxicity is associated
with hind limb paralysis and peripheral neuropathy (Katz et al., 1981; DeAngelo et al., 1996).
Brain lesions characterized by the vacuolization of white tracts have been noted at doses of 125
to 2000 mg/kg-day (Katz et al., 1981). Focal vacuolation and gliosis were present in the
forebrain and brain stem of rats treated with 1,100 mg/kg-day (Bhat et al., 1991). Progressive
changes in gait were observed in 2 strains of weanling rats exposed, via drinking water, to 16, 89
or 173 mg/kg-day DCA (F344 strain) and 17, 88 or 192 mg/kg-day DCA (LE strain) for 13
weeks (Moser et al., 1999). In addition, hind limb grip strength was decreased throughout
exposure in LE rats treated with 88 or 192 mg/kg-day (but no dose-response relationship was
evident) and in F344 rats treated with 173 mg/kg-day DCA. Also, tremor, hypotonia, and
inhibition of pupil reflex were observed in both high-dose strains. Moser et al. (1999) provided
information pertaining to the potency of DCA in drinking water vs. oral gavage, reversibility of
effects, and strain and age differences. To date, no signs of neurologic toxicity or morphologic
changes of the nervous system have been reported in DCA-treated mice.
4.6. SYNTHESIS AND EVALUATION OF CANCER EFFECTS AND MODE OF
ACTION
4.6.1. Data Summary
No epidemiological investigations of the carcinogenicity of DCA in humans have been
performed. However, there have been a number of studies on cancer risk in humans who ingest
chlorinated drinking water (which may contain DCA as a disinfection by-product). A number of
these studies show a weak correlation between exposure to chlorinated drinking water and risk of
bladder cancer. However, available data are not sufficient to establish a causal relation between
the ingestion of chlorinated water and the risk of developing cancer (U.S. EPA, 1998d). Further,
even if data ultimately establish an increase in cancer risk that is attributable to the ingestion of
chlorinated water, it cannot be concluded from these studies that DCA per se is carcinogenic in
62
-------
humans, since chlorinated water contains a wide spectrum of potentially carcinogenic
disinfection by-products.
In animals, there have been a number of independent studies investigating aspects of the
carcinogenicity of DC A. Among these studies, statistically significant increases in hepatic
carcinomas alone and/or hepatic carcinomas plus adenomas was seen in: (1) all male B6C3FJ
mouse studies (Herren-Freund et al., 1987; Bull et al., 1990; DeAngelo et al., 1991, 1999; Daniel
et al., 1992; Ferreira-Gonzalez et al., 1995); (2) all but the last cited female B6C3FJ mouse study
(Pereira and Phelps, 1996; Pereira, 1996; U.S. EPA, 1991b; Bull et al., 1990); and (3) in three
F344 rat studies (DeAngelo et al., 1996; Richmond et al., 1995). Based on these findings, it is
recognized that DCA is hepatocarcinogenic in male mice and rats, and that exposure to high
concentrations of DCA in drinking water can significantly increase the incidence of liver
adenomas and/or carcinomas. Exposure levels causing increased incidence tumors in animals
range from 0.5 to 5 g/L. However, concentrations as low as 0.05 g/L (8 mg/kg-day) increase the
multiplicity of tumors in male mice (DeAngelo et al., 1999).
The induction of liver tumors in mice is a widely debated endpoint in cancer bioassays.
This is particularly true for male B6C3Fj mice, which are especially susceptible to developing
liver tumors from a variety of chemical insults. However, the positive findings in female mice
and rats indicate that the carcinogenicity of DCA is not restricted to male mice, and that the
tumorigenic response is likely to be relevant across different species. The fact that DCA induces
liver tumors in the rat at lower doses than in the mouse also strengthens the overall weight of
evidence for DCA's tumorigenicity.
Some support for the relevance of the hepatic tumors observed in rodents to humans is
provided by the fact that liver tumors are sometimes a consequence of glycogen storage disease
(VI) and hereditary tyrosinemia I. As mentioned previously, several of the hepatic
manifestations of DCA-exposure in rodents (liver enlargement and glycogen accumulation) are
similar to the consequences of untreated glycogen storage disease (VI) while the DCA inhibition
of GSTZ produces increased concentrations of the same intermediary tyrosine metabolites that
are increased in tyrosinemia I. In addition, the work of Carter et al. (2003) seems to support a
multifactorial origin for DCA-induced liver tumors in animals. Examination of the tissues from
the mice used in the DeAngelo et al. (1999) study suggests that cancerous liver tumors can
originate from eosinophilic, dysplastic, and basophilic or clear cells of exposed animals. One or
more of these origins may be relevant to humans.
63
-------
As with many carcinogens, the duration of exposure is an important determinant of the
magnitude of the tumorigenic response. Several studies in male and female B6C3FJ mice found
multiple tumors per animal with treatment concentrations of 2 g/L and above within one year
(Herren-Freund et al., 1987; Bull et al., 1990; DeAngelo et al., 1991; Pereira, 1996). At this time
point, the dose-response is very steep with no response observed at concentrations of 1 g/L or
lower. However, lower concentrations (0.5 g/L) resulted in a hepatic tumor incidence of
approximately 80% in a full two-year study in male mice (Daniel et al., 1992). This same
temporal relationship occurred in the rat at doses as high as 2.4 g/L required to induce tumors at
60 weeks. Doses as low as 0.5 g/L induced liver tumors (incidence, 41%) when exposure was
104 weeks (Richmond et al., 1995).
There is a considerable increase in the internal DCA dose with drinking water
concentrations between 1 and 10 g/L in cases where GSTZ is inhibited according to the Barton et
al. (1999) pharmacokinetic model. The increase is coincident with the concentration associated
with a statistically significant increase in cancer prevalence in the study by DeAngelo et al.
(1999) and supports the hypothesis that either DCA or the alkylating tyrosine metabolites that
accumulate when GSTZ is inhibited may be the causative agent (Table 4-5). However, this
hypothesis does not explain the significant increase in tumor multiplicity at lower doses.
Additional support for classifying DCA as a carcinogen comes from the data base of
other carcinogenic compounds such as perchloroethylene, trichloroethylene, trichloroacetic acid,
and chloral hydrate. Each of these compounds produce DCA as a metabolite (IARC, 1979; Lash
and Parker, 2001). However, DCA production alone is unable to account for the carcinogenic
properties of the more highly halogenated two-carbon precursor compounds.
Table 4-5. Drinking water exposures, cancer response
and simulated internal dose metrics
Drinking Water
Concentration
(g/L)
0
0.05
0.5
1
Daily Dose
mg/kg-day
0
8
84
168
Modeled AUCL
mg-hr/L
0
0.041
0.72
15.8
Carcinoma
Prevalence (%)
26
33
48
71*
Carcinoma
Multiplicity
0.28
0.56*
0.68*
1.29*
64
-------
2
2.5
315
429
417
1064
95*
100*
2.47*
2.90*
* Statistically different from control.
Source: Adapted fromDeAngelo et al. (1999).
4.6.2. Potential Mode of Carcinogenicity
While a number of studies provide some information on the mode of action by which
DC A may increase cancer incidence in animals, none of them provide a satisfactory mode of
action for the carcinogenicity of DCA. Note that it is not necessary to assume that only one
mode is operative, and the possibility exists that different modes may be acting in different
species, or even in the same species at different doses. The most likely modes of action for the
carcinogenic activity of DCA are briefly summarized below.
Mutagenicity and Genotoxicity
The genotoxicity database on DCA has been extensively reviewed by several scientific
organizations including IARC (1995), ILSI (1997), WHO (2000), and EPA (1998c). Based on
an evaluation of data available at the time, IARC (1995) and ILSI (1997) reached independent
conclusions that DCA was not genotoxic. A more recent review by WHO (2000) concluded that,
although there is some evidence that DCA is genotoxic at high concentrations, these effects are
not likely to be involved in the mechanism of DCA tumorigenesis. In another recent review,
Moore and Harrington-Brock (2000) concluded that the available genotoxicity data indicate that
DCA is very weakly mutagenic. In contrast, the National Center for Environmental Assessment
(U.S. EPA, 1998c) concluded that available data indicate that DCA is a direct-acting genotoxic
agent. This conclusion is based on recent studies conducted at the National Health and
Environmental Effects Research Laboratory (DeMarini et al., 1994; Fuscoe et al., 1996; Leavitt
et al., 1997; Harrington-Brock et al., 1998) that reveal DCA's ability to cause mutational
damage, induce point mutations in DNA and cause chromosomal aberrations. Note that several
of these newer studies were published after the IARC and ILSI evaluations.
The majority of evidence indicates that DCA is a weak mutagen, inducing mutations and
chromosome damage in vitro and in vivo assays predominantly at high concentrations.
Nevertheless, in the absence of causal data, EPA considers it prudent to assume that DCA might
be genotoxic, at least under in vivo exposure levels that are associated with detectable increases
65
-------
in tumor incidence (particularly at the higher doses). Whether DCA is genotoxic at lower doses
(which would suggest a linear dose-response curve for cancer risk) is not known.
Cytotoxicity and Compensatory Hyperplasia
One plausible hypothesis for DCA-induced tumorigenesis is that it occurs secondary to
increased cell proliferation (Ames and Gold, 199la, b). Increasing the rate at which a cell
divides increases the probability that some critical genetic error will occur or a quiescent error
will be clonally expanded, potentially resulting in a transformed cell. Regenerative hyperplasia
in response to cytotoxicity is one way that this can occur. As previously discussed in Section 4.5,
there is evidence of liver toxicity in DCA-treated mice at doses shown to cause tumor
development. However, data in the rat do not fit this paradigm. Although frank liver toxicity
(necrosis) has not been reported in DCA-treated rats, there is explicit evidence in the literature to
the contrary. DeAngelo et al. (1996), for example, examined the livers of the rats in all dose
groups including the highest dose (2 g/L) and reported no evidence of increased hepatocellular
necrosis and no increase in labeling index. These findings indicate that cell killing and
regenerative hyperplasia were minimal. Subsequently, the authors reported to ILSI (1997) that
there were small increases in serum enzymes associated with DCA exposure, suggesting that a
low level of cell death and lysis may have been occurring. Whether this could account for the
strong tumorigenic response is not clear. Subsequent studies by DeAngelo et al. (1999) detected
an increased incidence and multiplicity of hepatic tumors in male mice that displayed no
apparent cytotoxic or regenerative response to DCA. It also seems reasonable that if
necrosis/cytotoxicity were an important precursor to DCA hepatocarcinogenesis, then the rat,
which demonstrates far less liver toxicity, should by extension be less susceptible to DCA
tumorigenicity. As previously described, this is not the case (DeAngelo et al., 1991, 1996).
Utilizing magnetic resonance imaging techniques, Miller et al. (2000) demonstrated that
DCA can affect growth rates of liver tumors. In particular, the results suggest that the primary
effect of DCA in tumor induction is mediated through accelerated growth of spontaneously-
initiated cells. To some extent this conclusion is supported by studies suggesting that DCA
influences the structure of H-ras protein (Anna et al., 1994; Velazquez, 1995; Ferreira-Gonzalez
et al., 1995; Schroeder et al., 1997) and the concentrations of several nuclear transcription
factors (Stauber et al., 1998; Stauber and Bull, 1997; Pereira et al., 2001).
66
-------
Several chemical carcinogens have been shown to induce mouse liver tumors with
specific point mutations at codon 61 of the H-ras proto-oncogene. This mutation also occurs
with very high frequency in spontaneously-derived liver tumors (-50%). Interestingly, in DCA-
induced liver tumors, the frequency of ras mutations was the same as the frequency of ras
mutations in the spontaneously-derived tumors of male mice (Anna et al., 1994). However, the
spectrum of mutations observed in hepatic tumors was different between control and DCA-
treated animals. Apparently, DCA treatment resulted in a decrease in the frequency of AAA
mutations in H-ras with a corresponding increase in CTA mutations. The toxicological
significance of this finding is not clear. Anna et al. (1994) interpreted these observations to
suggest that oncogene activation was not the primary mechanism of DCA action, and that DCA
may act by providing a selective growth advantage to hepatocytes bearing this type of mutation
in the H-ras proto-oncogene.
It is clear that sufficiently high doses of DCA can cause cytolethality and regenerative
hyperplasia in the liver of exposed mice, and that this response occurs in some dose groups in
which a tumorigenic response is also observed. However, data from mice exposed to lower
doses (DeAngelo et al., 1999) and data from rats (DeAngelo et al., 1996) indicate that this
response is not required for tumorogenesis. On this basis, it is concluded that the mechanism is
unlikely to completely account for the tumorigenic response, at least at doses <0.5 g/L that do
not produce clear hepatotoxicity.
Peroxisome Proliferation
Increased number and/or size of hepatic peroxisomes (peroxisome proliferation) is a
common finding in the livers of rodents treated with some types of hepatocarcinogens. The
proliferation is regulated by a class of nuclear receptors known as peroxisome proliferator-
activated receptors (PPARs) which are believed to mediate at least some of the effects reported
for hepatocarcinogens, including the initiation of certain cellular events leading to
transformation (U.S. EPA, 1998c). At this time, however, the precise role that increased
peroxisome proliferation or PPARs plays in the actual induction of tumor formation is not clear
(U.S. EPA, 1998c). Interestingly, significant species differences exist in the expression of
various PPARs, and are especially prevalent between rodents and humans (U.S. EPA, 1998c).
Available data suggest that humans are less responsive to a variety of peroxisome proliferators
than are rats and mice. This has generated some controversy over whether peroxisome
proliferators are carcinogenic in humans.
67
-------
DCA has been shown to be a weak peroxisome proliferator in mice and rats (DeAngelo et
al., 1998, 1999; Daniel et al., 1992; Mather et al.,1990). Transient transfection studies
demonstrate that DCA activates PPARs (Zhou and Waxman, 1998). Dichloroacetic acid has also
been shown to activate mouse and human PPARs with similar receptor sensitivity (Maloney and
Waxman, 1999). However, the relevance of this finding to DCA tumorigenesis is not fully
understood. Recently, Thai et al. (2003) found no activation of the PPARa gene in mouse liver
tissues from the DeAngelo et al. (1999) animals using a microarray containing this gene.
Accordingly, there is some inconsistency in the data base.
While one of the effects attributed to PPARs is the suppression of c-Jun activity and
expression (Sakai et al., 1995), several recent studies demonstrate that DCA-induced tumors are
c-Jun positive (Stauber and Bull, 1997; Stauber et al., 1998). This finding is inconsistent with a
role for PPARs in DCA-induced tumorigenesis. It has also become apparent that DCA induces
hepatic tumors in rodents at doses that are significantly below those required to induce
significant peroxisome proliferation (DeAngelo et al., 1999). Collectively, these observations
suggest that peroxisome proliferation is not likely to be important in the tumorigenicity of DCA.
Tumor promotion and alterations in cell replication and death
Several sets of observations suggest that DCA may be acting through a tumor promotion
mechanism. For example, Stauber et al. (1998) demonstrated that DCA increased cell
proliferation of c-Jun positive hepatocytes in vitro. The investigators treated isolated
hepatocytes from neonatal mice with DCA and plated the cells to allow them to form colonies.
While exposure of the cells to 0.5 mM DCA did not cause any cytotoxicity, it significantly
increased colony formation over controls. Interestingly, the colonies that were induced by DCA
were c-Jun positive. This is noteworthy because this is the same phenotype observed in DCA-
induced liver tumors in mice exposed to DCA in vivo by Stauber and Bull (1997). The authors
then pretreated the animals for 2 weeks with DCA in their drinking water and repeated the
experiments. Dichloroacetic acid again induced c-Jun positive colony formation, but it only
required 0.02 mM DCA for the same degree of induction. This observation indicates that DCA
provided a selective growth advantage for (promoted) hepatocytes with a specific phenotype.
Dichloroacetic acid has also been reported to alter cell replication rates in other assay systems,
but in a complex manner. In vivo, exposure of cells to DCA stimulates cell proliferation at low
doses in the short term, but increasing doses and chronic exposure appears to sharply inhibit
68
-------
hepatocyte replication (U.S. EPA, 1998c). In normal hepatocytes (c-Jun negative), in vitro DC A
administration consistently inhibits hepatocyte replication (Pereira, 1996; Carter et al., 1995). In
contrast to normal cells, hepatocytes from DCA-induced liver tumors (c-Jun positive) are
resistant to the inhibitory effects of DCA on cell proliferation (Stauber and Bull, 1997).
Data from Pereira and Phelps (1996) indicate that MNU-initiated female mice exposed to
DCA in drinking water at a concentration of 20 mM exhibited a statistically significant increase
in adenomas (compared to controls) when measured after 31 and 52 weeks of exposure. Herren-
Freund et al. (1987) also examined DCA promotion using ENU as an initiator and doses of 400
or 1,000 mg/kg-day DCA. In mice treated with ENU and DCA, the incidence of hepatocellular
carcinomas was 66% (400 mg/kg-day) or 78% (1,000 mg/kg-day), while in mice treated only
with DCA at 1,000 mg/kg-day the carcinoma incidence was 81%. Thus, at high doses DCA is
able to act as a complete carcinogen. This conclusion is supported by the results of DCA
bioassays (Daniel et al., 1992; DeAngelo, 1991, 1996, 1999).
Chen (2000) proposed a biologically-based dose-response model for liver tumors induced
by TCE and DCA. The model incorporated parameters pertaining to initiation rate, proliferation
rate, conversion rate, probability of tumor progression, and death rate. A stochastic model was
used to predict tumor response in TCE bioassays on the basis of its metabolite, DCA alone. The
modeling results suggest that DCA may be responsible for most, if not all, TCE-induced
carcinomas. Adenomas, hyperplastic nodules and other tumors were not considered in the dose-
response modeling. Dosimetry was based on an unpublished PBPK model. However, in
modeling liver DCA concentrations generated from trichloroethylene metabolism, Barton et al.
(1999) concluded that the doses of DCA formed from trichloroethylene could not account for the
tumorigenic properties of this compound.
69
-------
Other Potential Mechanisms
Hypomethylation
Mammalian DNA contains the methylated base 5-methylcytosine. While the extent of
DNA methylation is known to play a role in gene expression (Stroger et al., 1993), decreases in
DNA methylation levels is a frequent finding in tumors and is considered to be a key factor in
expanding clones and precancerous cells during neoplastic progression (Counts and Goodman,
1994, 1995). Additionally, the level of methylated DNA is lower in chemically-induced liver
tumors than in normal liver tissue (Lapeyre and Becker, 1981).
Recent studies by Tao et al. (1996) demonstrate that DCA treatment decreases 5-
methylcytosine levels in hepatic DNA from treated compared to control mice. Methylation of
cytosine in the promoter region of genes regulates mRNA expression including that of the proto-
oncogenes, c-Jun and c-myc (Tao et al., 1998, 2000). These proto-oncogenes participate in the
control of cell proliferation. Cells from DCA-induced tumors have been identified as c-Jun
positive (Stauber and Bull, 1997). Increased levels of mRNA and protein for c-jun and c-myc
genes have been reported in liver and liver tumors from mice treated with DCA. Decreased
methylation in the promoter regions of the c-jun and c-myc genes and increased levels of
associated mRNAs and proteins were reported in the livers of mice exposed to DCA. While this
observation is consistent with the hypothesis that DCA might increase tumor risk by inhibiting
DNA methylation, the actual importance of this event in mediating the tumorigenic response to
DCA is not known.
Conclusions Regarding Cancer Mode of Action
There are numerous questions that remain unanswered about the toxicity of DCA; many
of which relate to carcinogenicity and prevent identification of a single mode of action as the
only or most important pathway leading to cancer. The number of metabolic pathways affected
by DCA and species differences in metabolism are still not known, nor has the ultimate toxic
substance been identified. Examination of the liver tissues from animals with carcinogenic
tumors suggest that the tumors can originate from several different cell lines and through more
than one pathway. The impact of DCA inhibition of GSTZ and other enzymes is incompletely
characterized and may be important based on the observed tendency for hepatic tumor
development in humans with hereditary tyrosinemia I or glycogen storage disease VI. The
70
-------
genotoxicity data for DCA are internally inconsistent, and there is little basis for judging whether
genotoxic effects - including alterations in the genetic messages for various proto-oncogenes -
are important in the carcinogenic response, and if so, whether the dose-response curve for
genotoxic effects is linear or nonlinear. If DCA is acting as a promoter, it is possible that the
dose-response curve might be linear. However, although Pereira and Phelps (1996) found some
evidence for promotion, the mechanism for DCA-induced promotion is not known and so the
shape of the dose-response curve is uncertain.
The importance of these issues regarding the mechanism and shape of the dose-response
curves for genotoxicity and carcinogenicity are highlighted by comparing the concentrations of
DCA in water that are carcinogenic in animals (0.05 to 5 g/L) with those that are commonly
observed in chlorinated drinking water (10 to 100 jig/L) (U.S. EPA, 1994a; IARC, 1995). Thus,
concentration values are about 4-5 orders of magnitude lower in drinking water than were used
in experimental studies in animals. This difference is further magnified by the lower water
intake per unit body weight of humans (approximately 0.03 L/kg-day) compared to rodents
(about 0.1-0.2 L/kg-day).
4.6.3. Cancer Characterization
Previous Classification
EPA performed a cancer weight-of-evidence review for DCA in 1994 (U.S. EPA, 1994d),
that was updated in 1996 (IRIS, 1996). The reviews classified DCA as a Group B2 (probable
human carcinogen) in accordance with the 1986 EPA Guidelines for Carcinogen Risk
Assessment (U.S. EPA, 1986a).
In 1995, IARC concluded that, based on the data available at that time, "DCA is not
classifiable as to its carcinogenicity to humans," and placed DCA in the IARC Group 3 category.
In 2002, IARC classified DCA in Group 2B (possibly carcinogenic to humans) based on
sufficient evidence of carcinogenicity in experimental animals and inadequate evidence of
carcinogenicity in humans. It is important to keep in mind that at the time of the first IARC
evaluation in 1995, there was no information regarding positive carcinogenicity responses in the
rat.
Current Characterization of DCA Carcinogenesis
71
-------
Based on current data and the lack of conclusive data regarding the mode of action of
DCA at environmentally relevant doses, DCA is considered likely to be carcinogenic in humans
(U.S. EPA, 1999, 2003). This assessment is based on the strength of the evidence in animal
bioassays. In particular, there are a number of independent studies reporting: consistently
positive results at roughly comparable doses, site concordance for tumor formation between two
species, consistent observations in different species and sexes, clear evidence of a dose-response
relationship, and no clear data supporting a cohesive mode of action.
4.7. SUSCEPTIBLE POPULATIONS AND LIFE STAGES
While no data were located to establish that any particular human subpopulation is likely
to be especially susceptible to the toxic effects of DCA, the toxicity of DCA appears to be
related to the ability of the body to clear parent DCA by metabolism (Lukas et al., 1980;
Cicmanec et al., 1991). Thus, individuals who have low activity GSTZ isozymes, or isozymes
that are particularly vulnerable to inhibition by DCA might be more susceptible than the general
population. Available data from a population with a Caucasian-European lineage suggest that
individuals with GSTZ Ib-lb, Ic-lc, and Id-Id isozymes might have a different response to
DCA than those with la-la isozyme (Blackburn et al., 2000, 2001). Individuals with glycogen
storage disease (an inherited deficiency or alteration in any one of the enzymes involved in
glycogen degradation) represent another group that may be more susceptible to DCA toxicity.
There is some evidence that alterations in glycogenolysis precede the development of many
tumor types (Bannasch, 1986). The dose-response for DCA-induced effects on hepatic glycogen
is in the same range as that required for inducing liver tumors (Bull, 2000). In addition, DCA is
thought to be metabolized by at least one free radical generating pathway, and peroxidation has
been proposed as a mechanism for DCA toxicity. Thus, it is possible that catalase-deficient
individuals may also experience increased risk.
Individuals with hyperoxaluria Type 1, a rare genetic disorder, would be susceptible to
elevated glyoxalate originating from DCA. While data are unavailable regarding the prevalence
of this rare disorder in the United States, data from France indicate that the prevalence is 1.05
per 1 million individuals (Cochat et al., 1995). In this condition, the inability to convert
glyoxyalate to glycine leads to the formation and excretion of oxalate (Montgomery et al., 1990).
72
-------
Specific information on whether children are more susceptible than adults to the effects
of DC A are not available. At this time there are no indications that there are unique GST Zeta
isoforms expressed in the fetus or neonate. GST zeta is expected to be active in neonates,
infants, and children because of its role in catabolism of tyrosine, an amino acid, which is
present in milk other protein foods and needed for growth and development. Accordingly, GST
zeta alone is unlikely to play a direct role in childhood susceptibility.
In female rats, DCA exposure during gestation resulted in the impairment of fetal
maturation and soft tissue anomalies (primarily of cardiac origin) indicating that the developing
fetus is susceptible to DCA-induced toxicity (Smith et al., 1992). Data collected by Moser et al.
(1999) provide limited evidence for increased susceptibility of rats to DCA-induced
neurotoxicity when exposures begin shortly after weaning.
73
-------
5. DOSE-RESPONSE ASSESSMENTS
5.1. ORAL REFERENCE DOSE (RfD)
The oral reference dose (RfD) is based on the assumption that thresholds exist for certain
toxic effects (e.g., cellular necrosis) and is expressed in units of mg/kg-day. The RfD is an
estimate (with uncertainty spanning perhaps one order of magnitude) of a daily exposure to
humans (including sensitive individuals) that is likely to be without an appreciable risk of
deleterious effects during a lifetime.
5.1.1. Methods of Analysis
Data on the noncancer effects of DC A were used to estimate RfD values using two
different approaches: (1) the traditional NOAEL-LOAEL approach (Section 5.1.2.), and (2) the
Benchmark Dose (BMD) modeling approach (Section 5.1.3).
5.1.2. NOAEL/LOAEL Approach
Data Summary
Figure 5-1 graphically presents the NOAELs and LOAELs from studies that examined
the noncancer effects of DCA. NOAEL values are shown by open symbols and LOAEL values
are shown by closed symbols. A table that summarizes the data displayed in Figure 5-1 is
included as Appendix B. Some of the studies from Figure 5-1 were not considered suitable for
quantitative risk assessment because there was no LOAEL or there were too few dose groups to
permit an assessment of dose-response. The key studies that were considered for quantitative
risk assessment are summarized in Table 5-1.
Oral exposure levels of 12.5-200 mg/kg-day have been demonstrated to cause all of the
characteristic noncancer effects in animals, and most of these effects (impacts on metabolism,
neurotoxicity, liver effects) have also been observed in humans at similar doses. Based on the
general similarity in the effect levels reported for each response category, it is not apparent that
any one effect occurs at a clearly lower dose than the others, and that one type of effect should
be considered critical.
74
-------
Selection of Principal Study and Critical Endpoint
The study of Cicmanec et al. (1991) identifies the lowest LOAEL (12.5 mg/kg-day) that
has been determined to date (Table 5-1). In this study, beagle dogs were administered oral doses
(12.5 to 72 mg/kg-day) of DCA in capsules for 90 days. Adverse effects noted in the low-dose
group (12.5 mg/kg-day) included mild to severe testicular degeneration in four of five males,
along with mild to moderate hepatic vacuolization and mild vacuolization of the myelinated
white tracts of the cerebrum and cerebellum in males and females. This study is supported by
the findings of Katz et al. (1981), who noted marked testicular degeneration and myelin
vacuolization in dogs administered oral doses of 50 mg/kg-day or higher for 13 weeks. While
testicular effects have not been noted in humans, this effect has not been monitored in the human
population. Data from humans administered DCA as a pharmaceutical indicate that doses of 25-
50 mg/kg-day produce neurological effects, including sedation and peripheral neuropathy
(Moore et al., 1979; Spruijt et al., 2001; Stacpoole et al., 1998a). Based on these considerations,
the study of Cicmanec et al. (1991) is judged to identify a LOAEL that is likely to be appropriate
for humans, and is selected as the principal study. The effects of concern are testicular
degeneration accompanied by mild histopathological alterations in the liver and brain.
75
-------
Table 5-1. Summary of noncancer studies considered for benchmark modeling
Endpoint
Metabolism
Sepatic
STeurologic
Reproductive
Species
Rat
Dog
Mouse
Rat
Dog
Rat
Dog
Rat
Dog
Reference
Kate etal., 1981
Kate etal., 1981
Sanchez and Bull, 1990
tCato-Weinstein et al.,
1998
Bull etal., 1990
DeAngelo et al., 1991
Mather etal., 1990
Kate etal., 1981
Smith etal., 1992
loth etal., 1992
Cicmanec etal., 1991
Kate etal., 1981
Cicmanec etal., 1991
Kate etal., 1981
Smith etal., 1992
Linder etal., 1997
loth etal., 1992
DeAngelo etal., 1996
Kate etal., 1981
Cicmanec et al., 1991
Duration
3 months
13 weeks
2 weeks
2 weeks
1 year
60-75
weeks
3 months
3 months
gestation
10 weeks
3 months
3 months
3 months
13 weeks
gestation
days 6- 15
2 weeks
10 weeks
100 weeks
13 weeks
3 months
Dose Grps.a
4
4
4
6
3
3
4
4
4
4
4
4
4
4
4
5
4
3
4
4
NOAEL
57
20
7.6
3.9
14
14
3.6
LOAEL
125
50
190
100
140
77
35.5
125
140
31
12.5
125
12.5
50
140
54
31
40.2
50
12.5
Effect
Decreased serum lactate and
glucose
Decreased serum metabolites
increased glycogen, focal necrosis
Increased glycogen
increased glycogen
increased liver weight
increased liver weight
increased liver weight
increased liver weight
increased liver weight
Increased liver weight,
inflammation
Sistological brain lesions
Vacuolar changes in brain
Vacuolar changes in brain
Decreased fetal wt, increased
resorptions
impaired sperm formation
impaired sperm formation
increased testicular weight
r'rostate atrophy, testicular
shanges
lesticular degeneration
Number of dose groups, including control
NCV = Nerve conduction velocity
76
-------
1000
100 -
S^
ro
T3
i5
a
o
Q
10 -
"I _
1"
^
A
A
0 I
A
O
•
: •
$
•
0
Metabolic Hepatic Neurological Reproductive
ENDPOINT
• Human-LOAEL
AMouse-NOAEL
AMouse-LOAEL
O Rat-NOAEL
• Rat-LOAEL
• Dog-LOAEL
Figure 5-1. Summary of noncancer effects of DCA.
77
-------
Calculation of the NOAEL-LOAEL Based RfD
Based on the LOAEL of 12.5 mg/kg-day identified by Cicmanec et al. (1991), the oral
RfD for DCA is calculated as follows:
RfD = (12.5 mg/kg-day)/(3000) = 0.0042 mg/kg-day (Rounded to 4E-03 mg/kg-day)
where:
12.5 mg/kg-day = LOAEL, based on lesions observed in the testes, cerebrum,
cerebellum, and the liver of dogs exposed orally (via gel capsules)
to dichloroacetic acid for 90 days.
3,000 = Uncertainty factor. This uncertainty factor includes a factor of 10
to account for potential inter-human variability in susceptibility to
DCA, a factor of 3 to account for extrapolation from animal data to
humans, a factor of 10 to account for use of a LOAEL, a factor of
3 to account for the use of a less-than-lifetime study in which frank
effects were noted, and a factor of 3 to account for deficiencies in
the database.
A factor of 10 was applied for intrahuman variability because of the observation that the
most frequent human GSTZ variant (GSTZ Ic-lc) is one that has a low activity toward DCA and
is also impacted by DCA inhibition to a greater extent than the most active, but less frequent
human variant (GSTZ la-la). Accordingly, one might expect poor clearance of DCA from
human plasma via oxidative dechlorination when exposure is continuous.
A threefold factor was applied for interspecies variability. There are several reasons for
this choice and the resulting partial reduction of the UF from the default of 10. First, death
occurred at a dose of about 75 mg/kg-day DCA (90 day study) in 3/10 and 1/3 dogs after 51 and
74 days of dosing and 50 days of dosing, in the principle study by Cicmanec et al. (1991) and the
study by Katz et al. (1981) respectively. Conversely, Stacpoole et al. (1998b) reported on cases
of five children with lactic acidosis who received 25-60 mg/kg-day orally for two months to four
years without clinical signs of DCA toxicity (elevation of liver enzymes and neuropathy).
Although two of the children died during treatment, death was the result of infection and not
from the lactic acidosis or DCA treatment. Annual mortality in patients with congenital lactic
acidosis, even with treatment, is 20%.
78
-------
Additional support for this conclusion is provided by the fact that metabolic effects of
DCA on serum lactate and glucose in dogs (Katz et al., 1981; Ribes et al., 1979) parallel those in
humans (Stacpoole et al., 1998a, b). Stacpoole et al. (1998b) reported that lactate concentrations
decreased by at least 20% within 24 hours after oral doses of 25 to 100 mg/kg in humans. In the
study by Katz et al. (1981), there was an approximate 40% reduction in serum lactate
concentrations of dogs (male and female) after 13 weeks of exposure to 50 mg/kg-day DCA.
Limited toxicokinetic data suggest that dogs metabolize DCA at a slower rate than
humans and rodents supporting the concept of their increased sensitivity (Section 3.3; Lukas et
al, 1980; Curry et al, 1991; Lin et al, 1993; Larson and Bull, 1992; James et al, 1998). A single
intravenous dose of 100 mg/kg in two dogs lead to peak plasma levels that were twice as high as
the same levels in rats (Lukas et al., 1980).
Lastly, the structure of GST Zeta appears to be highly conserved across species making
it unlikely that the metabolic differences in humans will differ from dogs by a full order of
magnitude, also taking into consideration that the full UF=10 has been applied for intrahuman
variability. Under these circumstances an interspecies uncertainty factor of 3 rather than the
default 10-fold value is justified.
The factor of 10 for the use of a LOAEL is justified by the observed effects of DCA on
the nervous system in sensitive humans (those under treatment for lactic acidosis and other
disorders) at doses of 25-50 mg/kg-day. These doses are within the same order of magnitude as
the LOAEL in the Cicmanec et al. (1991) study and the LOAEL for neutotoxicity in F-344 and
LE rats in adult and weanling rats in the Moser et al. (1999) study. There are no human data on
testicular effects from DCA.
Threefold factors were applied for both the use of data from a less-than-lifetime study
and database inadequacies. The database for DCA lacks a multi-generation study of
reproductive toxicity and a developmental neurotoxicity study, thus, meriting an uncertainty
factor of 3 for database insufficiency. Otherwise the database is comprehensive with
information from subchronic and lifetime animal studies, studies in three animal species, and
over 25 years of experience with the use of DCA as a experimental pharmaceutical in the
treatment of several human disorders.
79
-------
The richness of the data base does not abrogate all concern associated with using a
subchronic study as the basis of the RfD, but is sufficient to reduce the uncertainty factor from a
10 to a 3. The neurological effects of DCA in the principal study are a concern as is the fact that
there are no data on the mechanism for the observed neurological or testicular effects.
Neurological effects were seen in humans and other animal species (rats, dogs) at doses
comparable to the LOAEL in the Cicmanec et al. (1991) study. They were severe enough in
human patients to alter the clinical treatment regime. About 20 to 50% of patients with lactic
acidosis experience sedative effects with single and repeated oral or intravenous doses of 25 to
50 mg/kg-day. The effects were reversed with the withdrawal of DCA, but in some patients
reversal was slow (Stacpoole et al., 1998a). The effects on the nervous system seen in dogs
involved vacuolization of myelin. This observation can be mechanistically linked to the
decreased nerve conduction velocity observed in human subjects (Spruijt et al., 2001) since
nerve impulses travel faster in myelinated nerves. Therefore, the use of an uncertainty factor of
3 to extrapolate from subchronic to chronic exposures is appropriate. There are no data that
permit an assessment of the clinical progression of the neurological effects. The data on
testicular effects could be more robust, but are mitigated by the lack of testicular histopathology
in the DeAngelo et al. (1996) rat cancer study. Rats are susceptible to testicular effects as a
result of DCA exposure (Linder et al., 1997; Toth et al., 1992), but the data indicate they are less
sensitive to this effect than dogs.
5.1.3. Benchmark Dose Approach
Selection of Data Sets for Modeling
It is apparent from Figure 5-1 that DCA produces effects on metabolism, the liver, and
the nervous system at doses between 10 and 100 mg/kg-day in rats, mice, dogs and humans.
Effects on the reproductive system are seen in rats, mice, and dogs at the same doses. Some
observed effects such as the metabolic changes are biomarkers of exposure and others are
unequivocally adverse at doses between 10 and 100 mg/kg-day. The data from these studies
were evaluated for their suitability in establishing a protective dose-response curve using
benchmark dose analysis following the criteria outlined in Table 5-2. Studies were eliminated
from consideration for benchmark modeling because of the following reasons:
• there were only one or two doses;
• there was no LOAEL;
80
-------
• the LOAEL exceeded 200 mg/kg-day (a LOAEL of 200 mg/kg-day is more than
10 times the lowest LOAEL of 12.5 mg/kg-day seen in the Cicmanec et al. (1991)
study); and
• the effects were not definitively adverse.
All of the studies that were considered as potential candidates for modeling are included
in Table 5.1. Among the studies that were included by the quantitative criteria outlined above,
several were excluded from modeling on the basis of semi-quantitative or qualitative criteria, as
follows:
The study by Katz et al. (1981) on the effect of DCA on serum metabolite levels in dogs
was not selected because the doses administered resulted in markedly reduced appetites
at all doses, and both sexes exhibited dose-dependent weight losses. Thus, effects on
serum metabolite levels might simply be secondary to decreased food intake. In addition,
one of three mid-dose female dogs and one high-dose male dog died during the study.
Thus, the number of animals surviving at the end of the study is too small to allow
reliable BMD modeling.
Data from the study by DeAngelo et al. (1991) demonstrating increased liver weight was
not used because the increase was due to hepatic tumor growth. Increased liver weight,
in this case, is not an appropriate indicator of noncancer effects.
The data set on liver histopathological lesions reported by Cicmanec et al. (1991) was not
retained because most of the lesions were ranked as mild and are not likely to be of
significant toxicological concern.
Data on the effects of DCA on testicular weight in rats (DeAngelo et al., 1996) were not
used. Although DCA caused a slight, but significant increase in absolute and relative
testes weight at 40.2 mg/kg-day, there were no accompanying histopathological effects in
these tissues. In a second study in male rats performed by the same investigators, a
significant decrease in absolute (but not relative) testes weights was observed at the
single dose used (139 mg/kg-day). When considered together, the data indicate that
increases in testes weight observed at a lower dose range might reverse at higher doses.
The data were considered inappropriate for modeling. In addition, the endpoint was not
deemed to be the most sensitive because no histopathological effects were noted.
Data on vacuolar changes in brain reported by Cicmanec et al. (1991) were not utilized
because there was no consistent dose-response trend in the data.
81
-------
Table 5-2. Criteria for selecting studies appropriate for BMP modeling51
Category
Criterion
Rationale
Possible Exceptions
Quantitative
The chosen study should have at
least 2 dose groups, plus a control
group.
A study with only 1 dose group and a control
does not provide enough data to define the
shape of the dose-response curve.
If the one exposure group yields a response
near the BMR, a suitable estimate of the
BMD may be possible.
The LOAEL from the chosen study
should be within a factor of 10 of
the lowest LOAEL from other
studies.
Studies that identify LOAEL values more than
10-times the lowest LOAEL are very unlikely to
be based on the most sensitive endpoint and are
unlikely to yield the lowest BMD.
The NOAEL from the chosen study
should not exceed the lowest
LOAEL from other studies.
Studies that identify NOAEL values that exceed
the lowest LOAEL are very unlikely to be based
on the most sensitive endpoint and are unlikely
to yield the lowest BMD.
If the study defines a reasonable dose-
response trend below the NOAEL, but the
NOAEL is elevated because of lack of
statistical power, the study might be worthy
of evaluation.
Semi-
Quantitative
The LOAEL should not be a near-
maximal adverse response.
If the response in the lowest dose group is at the
high end of the dose-response curve, the data
will not provide information on the shape of the
curve at doses that produce responses near the
BMR, and BMD estimates will be unreliable.
If the shape of the dose-response curve is
very steep, then the dose-response curve will
be reasonably constrained even if the
response at the low-dose group is well above
the BMR.
The data should have a clear dose-
response trend, preferably
smoothly graded (monotonic).
If no clear dose-response trend is apparent, the
data are not suitable for establishing a dose-
response curve.
Qualitative
The endpoint for which there is
dose-response data should have
clear lexicological relevance.
There is little basis for setting a BMR (and,
hence, estimating a BMD) for endpoints which
are not easily interpretable in terms of their
lexicological significance. Endpoints which are
known to be early indicators of Ihe adverse
effecls of Ihe chemical are preferred.
U.S. EPA, 2000C
The findings of Katz et al. (1981) on histopathological changes in brain and testes were not used because no quantitative data
on the severity or incidence of these effects were provided.
82
-------
Based on these evaluations, the following data sets were judged to warrant BMD
modeling:
• low birth weight and cardiac malformations in rats (Smith et al., 1992);
increased relative liver weight (Mather et al., 1990);
impaired sperm formation in rats (Linder et al., 1997);
impaired sperm formation in rats (Toth et al., 1992); and
testicular lesions in dogs (Cicmanec et al., 1991).
In the case of the Smith et al. (1992) study, EPA had already modeled the data using the
THRESH Benchmark Dose program (U.S. EPA, 1998e). Because the original data are no longer
available from the study (based on oral communication with the author), model inputs were
limited to the data provided in the published paper.
The Smith et al. (1992) study was conducted in two segments: one segment utilized a
DC A dose range of 900-2,400 mg/kg-day and the second utilized a dose range of 14-400 mg/kg-
day. The cardiac malformation data were presented as the percent of fetuses affected per litter.
The data were converted to affected fetuses per litter using the average number of fetuses per
litter (calculated from the number of fetuses and the number of litters [Table 5-3]). Data from all
dose groups were modeled to determine the BMD and BMDL. According to the best-fit model
for the cardiovascular defects, the BMDL10 (10% response level) was 567 mg/kg-day.
Table 5-3. Cardiovascular defects induced by DCA
Dose Group (mg/kg-
day)
0
14
140
400
900
1,400
1,900
2,400
Mean Percent Fetuses
Affected per Litter
0*
0.69
1.02
8.07
8.15
23.91
43.67
68.75
Average Number of
Fetuses per Litter
13
8.4
9.2
8.6
8.6
8.6
8.5
6.8
Estimated Incidence
0
1
1.8
13.2
11.2
37.1
63.3
74.0
* The control data from both segments of the study have been added together.
Source: Adapted from Smith et al. (1992).
83
-------
For the fetal body weights, it was necessary to use a combined mean weighted average
for males and females in each dose group because the fetal body weights were reported as a
continuous variable. In addition, it was necessary to estimate the mean responses of the animals
and the corrected sum of squares based on the published information in order to model the data
(Table 5-4). The parameters were not reported in the published version of the study. Initial
modeling for the fetal body weights incorporated data from all dose groups; however, use of the
control and the three lower dose groups from part two of the study provided the best fit in the
low-dose range. The BMDL values for 10, 5 and 1% reductions in mean fetal body weight were
458, 259 and 60 mg/kg-day, respectively. All of these values exceed the NOAEL identified by
the Cicmanec et al. (1991) study and, thus, were not considered to be suitable for derivation of
theRfD.
Table 5-4. Effects of DCA on fetal body weight
Dose Group (mg/kg-
day)
Combined Mean Fetal
Body Weight (g)
Combined Standard
Deviation
Estimated Corrected
Sum of Squares
Low-Dose Segment of the Study
0
14
140
400
3.58
3.68
3.54
3.36
0.22
0.31
0.21
0.31
11.81
20.84
11.27
22.75
High-Dose Segment of the Study
0
900
1400
1900
2400
3.57
3.06
2.90
2.77
2.68
0.22
0.22
0.31
0.31
0.24
13.24
9.39
20.81
19.32
8.93
Source: Adapted from Smith et al. (1992).
The software employed for benchmark dose modeling of the remaining studies was
BMDS Version 1.2 or 1.3.1, downloaded from EPA's NCEA web site. The data for
dichotomous endpoints were fit to each of the dichotomous models provided in the software,
including gamma, logistic, multi-stage, probit, quantal-linear, quantal-quadratic, and Weibull.
The data for continuous endpoints were fit to each of the continuous models offered in the
BMDS software (linear, polynomial, power, and Hill).
84
-------
Mather et al. (1990) identified a NOAEL for the liver weight effects from a study that
included four dose groups and was considered suitable for modeling (Table 5-5). Several other
studies (Cicmanec et al., 1991; Kato-Weinstein et al., 1998; Toth et al., 1992) provide
quantitative dose-response data on increases in absolute or relative liver weight which justifies
consideration of this endpoint for modeling. The findings were attributed to increased glycogen
accumulation. Glycogen accumulation may play a role in the toxicity of DC A. Significant
increases in relative liver weight were also observed in a chronic-duration study in which rats
were administered a time-weighted average dose of 1.6 g/L DCA (DeAngelo et al., 1996).
Table 5-5. Liver weight data set
Endpoint
(rats)
Increased Relative
Liver Weight
Parameter
Mean±S.E.
Dose (mg/kg-day)
0
3.8 ±0.10
3.9
4.08 ±0.01
35.5
4.73 ±0.09
345
6.42 ±0.10
Source: Adapted from Mather et al. (1990).
Studies by Cicmanec et al. (1991), Katz et al. (1981), Linder et al. (1997), and Toth et al.
(1992) have all identified effects of DCA on testicular histopathology and/or sperm parameters.
Several of the data sets were suitable for benchmark dose modeling. Table 5-6 summarizes the
dose-response data sets that were evaluated for the male reproductive system effects.
Table 5-6. Male reproductive data sets used for BMP modeling
Endpoint
(rats)
Epididymal Sperm
Count (106/g)
Sperm Morphology
(percent normal)
Sperm Motility
(percent motile)
Parameter
N
Mean
Stdev
N
Mean
Stdev
N
Mean
Stdev
Dose (mg/kg-day)
0
19
630.3
204.8
20
85.1
19.2
15
54.6
10.2
31.25
18
582.5
137.0
19
86.7
16.9
14
54.1
11.2
62.5
18
502.6
163.5
17
80.4
14.1
17
39.5
12.0
125
19
367.8
91.6
19
58.9
16.2
19
27.1
9.8
Source: Toth etal. (1992).
85
-------
Endpoint
(rats)
Cauda Sperm Count
(106)
Sperm Motility
(percent motile)
Parameter
N
Mean
Stdev
N
Mean
Stdev
Dose (mg/kg-day)
0
8
224
40
8
72
12
18
8
248
32
8
74
8
54
8
208
25
8
72
9
160
8
165
21
8
41
13
480
8
106
35
8
20
17
1440
8
86
11
8
6
7
Source: Linder et al. (1997).
Endpoint
(dogs)
Testicular
Degeneration
(Incidence)
Parameter
N
Affected
Dose (mg/kg-day)
0
5
0
12.5
5
4
39.5
5
5
72
5
5
Source: Cicmanec etal. (1991).
Results of the BMD model fitting (with the exception of the Smith et al. [1992] data
discussed above) are detailed in Appendix C. Within each data set, the preferred model was
selected based mainly on the quality of the model fit to the data, judged in part by the/? value
and in part by visual inspection of the fit. Models that yielded ap value less than 0.100 were not
considered further. When more than one model gave similar quality fits, the model that yielded
the lowest BMD was preferred. The results are summarized in Table 5-7.
86
-------
Table 5-7. Summary of noncancer BMP modeling results
Reference
Cicmanec et
al. (1991)
Linder et al.
(1997)
Toth et al.
(1992)
Mather et al.
(1990)
Data Set
Incidence of testicular
degeneration
Cauda Sperm count
Sperm motility
Epididymal Sperm
count
Sperm motility
Sperm morphology
Relative Liver Weight
(Liver weight/body
weight ratio)
BMR
10% Extra
risk
1 Stdev
1 Stdev
Point Risk
1 Stdev
1 Stdev
1 Stdev
Preferred
Model (a)
Quanta!
Quadratic
Hill
Hill
Linear
Hill
Power
Hill
P value
1.000
0.180
0.173
0.891
-(c)
0.619
-(c)
BMD
mg/kg-day
3.2
73.8
74.4
87.4
55.8
101.5
3.3
BMDL
mg/kg-day
2.1
-(b)
52.0
-(b)
40.4
74.7
3.0
(a) The preferred model is the one that fits the data best. If more than one model gave comparable fits, the preferred
model is the one yielding the lowest BMD.
(b) The software could not calculate a BMDL, including an attempt to define BMR as a point risk.
(c) Chi-Square Test for fit not valid (degrees of freedom are less than or equal to 0).
As seen in Table 5-7, two of the data sets could not be adequately described by any of the
continuous models: Toth et al. (1992) [sperm motility data] and Mather et al. (1990) [liver
weight]. Of the remaining data sets, four (based on the data of Linder et al. [1997] and Toth et
al. [1992]) yielded BMD values ranging from 74-102 mg/kg-day. In two of these cases, the
software was not able to calculate a BMDL (the lower confidence bound on the BMD). In the
other two cases, the BMDL values ranged from 52-75 mg/kg-day.
In contrast to these results, the data set reporting incidence of testicular lesions in dogs
(Cicmanec et al., 1991) yielded a very high quality fit (p = 1.000), and much lower BMD and
BMDL values (3.2 and 2.1 mg/kg-day, respectively) than the data from rats. However, this
apparent goodness-of-fit is an artifact because none of the dose groups in this study yielded a
response near the BMR, so the shape of the curve is essentially unconstrained in the low-dose
range. Thus, even though the fit appears to be of high quality, both the BMD and the BMDL are
judged to be unreliable.
87
-------
Calculation of the BMD-BasedRfD
BMD modeling of effects on sperm formation in rats (Toth et al., 1992; Linder et al.,
1997) yield BMD values of 74-102 mg/kg-day. These values (based on a BMR of one standard
deviation decrease from control) are substantially higher than the dose of 12.5 mg/kg-day, which
yielded a high incidence (4/5) of testicular degeneration in dogs (Cicmanec et al., 1991). This
suggests that rats are not as sensitive to the testicular effects of DC A as dogs, so the results from
BMD modeling in rats are not considered appropriate for deriving a RfD. As noted above, even
though BMD modeling of the data from dogs yields a model fit of an apparently high quality
model, the numeric values of the BMD and the BMDL derived from this data set are not
considered to be reliable, since the lowest dose tested yielded a high response, and there are no
dose groups yielding a response near the BMR. It is recommended for BMD analyses that
LOAEL values not be a near-maximal adverse response, which is the case for this endpoint
(Cicmanec et al., 1991).
On the basis of these considerations, it is concluded that none of the available noncancer
data sets provide a suitable basis for deriving an RfD for DCA via the benchmark dose modeling
approach.
5.1.4. Summary of Oral RfD Derivation
Although BMD modeling often offers a number of advantages over the traditional
NOAEL-LOAEL approach for deriving a reliable RfD (U.S. EPA, 1995), in this case none of the
available noncancer dose-response data sets provided a suitable basis for deriving a RfD via the
BMD approach. Therefore, the RfD of 4E-03 mg/kg-day derived using the NOAEL-LOAEL
approach is judged to be the most appropriate assessment of chronic noncancer risk based on the
current data for DCA.
5.2. INHALATION REFERENCE CONCENTRATION (RFC)
There are no data from toxicity studies of DCA that employed the inhalation route of
exposure. As noted in Section 2, both the acid and salt forms of DCA have low volatility,
therefore, inhalation exposure is not considered to be of concern. On this basis, an inhalation
RfC for DCA is not considered necessary.
88
-------
5.3. CANCER ASSESSMENT
5.3.1. Choice of Principal Studies and Cancer Endpoints
As discussed above, there are multiple studies in B6C3FJ mice that establish that
ingestion of DC A results in increased incidence of hepatic tumors (both hepatocellular adenomas
and adenocarcinomas). Of the available studies in mice, the best one for cancer dose-response
modeling is that reported by DeAngelo et al. (1999), since this study was specifically designed to
establish a multi-point dose-response curve, and data are available for five dose groups plus a
control group. In addition, the duration of this study spans the expected lifetime of a mouse.
It appears that the highest dose administered in this study approaches the MTD, based on
the significantly elevated hepatic necrosis indices throughout the study, as well as significant
decreases in body weight gain in mice from 52 weeks onward. For this reason, the highest dose
group was excluded from the analysis.
Cancer dose-response data are also available from studies in F344 rats (DeAngelo et al.,
1996; Richmond et al., 1995). This data set is less robust than that for mice, having only three
dose groups, the highest of which apparently exceeded the MTD, based on a marked decrease in
body weight. In addition, the data were collected in two separate studies in two separate
laboratories. Therefore, these data were not used in the cancer benchmark dose modeling.
5.3.2. Dose-Response Data
Table 5-8 summarizes the tumor incidence data for liver carcinomas in B6C3Fj mice in
the study by DeAngelo et al. (1999). The numbers of animals with either adenomas or
carcinomas at 100 weeks were used to model the cancer dose-response relationships.
Table 5-8. Cancer dose-res
Cone.
in water
(g/L)
No. of
animals
enterin
g study
Mean
BW(g)
at 100
weeks
ponse data evaluated using BMD modeling: male mice3
Dose (mg/kg-day)
Animals with
hepato-
carcinomas at
100 weeks
Animals with
hepato-
adenomas at
100 weeks
Animals with
either hepato-
carcinomas or
adenomas at
100 weeks
89
-------
0
0.05
0.5
1
2
3.5
50
33
25
35
21
11
43.9
43.3
42.1
43.6
36.1
36.0
Animal
0
8.0
84
168
315
429
HEDb
0
1.3
13.2
26.5
47.5
64.6
%
26%
33%
48%
71%
95%
100%
N
13
11
12
25
20
11
%
10%
3%
20%
51%
43%
45%
N
5
1
5
18
9
5
%
36
33
56
86
100
100
N
18
11
14
30
21
11
(a) High-dose group excluded from benchmark modeling (see text)
(b) HED calculated using a dose scaling factor of BW°75
Source: DeAngelo et al. (1999)
5.3.3. Dose Conversion
Because the exposure of mice to DC A in drinking water was continuous for the
approximate full life span of the animals, no adjustment is needed to account for duration of
exposure or duration of study. Doses in animals are converted to human equivalent doses (HED)
by assuming that doses (mg/day) in animals and humans are lexicologically equivalent when
scaled by body weight raised to the 3/4 power (U.S. EPA, 1992):
Dose (mg / day)"
BW
3/4
animal
Dose (mg / day)^
BW3/4
human
When doses are expressed as mg/kg-day, this yields the following:
HED (mg/kg-day) = Dose in animals (mg/kg-day) • (BWa / BWh)°'25
The group mean body weights for animals in each exposure group are shown in Table 5-8 above.
The body weight of humans was assumed to be 70 kg (U.S. EPA, 1988). The resulting HED
values are also shown in Table 5-8.
5.3.4. Dose-Response Characterization in the Range of Observation
The dose-response data sets presented in Table 5-8 were modeled using the BMDS
software system (Version 1.3.1) developed by the U.S. EPA National Center for Environmental
Assessment (NCEA). The benchmark dose was estimated using the numbers of animals with
90
-------
adenomas or carcinomas in the five lowest dose groups. As noted above, the highest dose group
was excluded because the highest dose (429 mg/kg-day) approached the MTD. The results of
the model fitting are detailed in Appendix D, and the findings are summarized in Table 5-9.
Judging by the chi-squared/?-values, all the models except the quantal linear fit the data
reasonably well (p-values > 0.7). The multistage and quantal quadratic models had the highest
/•-values (0.981) and the lowest AIC values (174.62), and thus appear to provide the best fit to
the data. For a benchmark risk (BMR) level of 0.10, the estimated benchmark dose values for
the best-fitting models (p-values > 0.7) range from 3.1 to 9.4 mg/kg-day, and the BMDL values
range from 2.1 to 5.7 mg/kg-day.
5.3.5. Selection of a Dose-Response Model
The multistage and quantal quadratic models provide essentially identical fits to the data.
The multistage model estimate was selected for dose-response extrapolation because the quantal
quadratic model has no first-order term and therefore may predict zero slope at zero dose. Given
the uncertainty surrounding the carcinogenic mechanism of DCA (see Section 4.6), it was
decided that the zero slope assumption was not justified. The multistage model gives a BMDL
estimate of 2.1 mg/kg-day (2.05, rounded to two significant figures). The fit of the multistage
model to the DeAngelo et al. (1999) data is shown in Figure 5-2.
91
-------
Table 5-9. Summary of cancer BMP modeling resultsa'b
Model
Multi-Stage (2)
Quantal-quadr.
Probit
Logistic
Weibull
Gamma
Log-Probit
Log-Logistic
Quantal-linear
BMD
6.86
6.86
3.16
3.10
7.53
8.45
9.36
9.17
1.88
BMDL
2.05
5.69
2.54
2.43
2.50
2.55
4.27
4.07
1.37
p-value
0.981
0.981
0.816
0.728
0.916
0.858
0.779
0.703
0.370
AIC
174.62
174.62
175.54
176.11
176.59
176.81
177.10
177.45
178.41
(a) Data = DeAngelo et al. (1999), animals with hepatocarcinoma or adenoma,
excluding high-dose group
(b) BMD estimated using BMR = 0.10, BMDL estimated as 95% LCL
5.3.6. Extrapolation to Doses Below the Range of Observation
Selection of the Point of Departure
Based on the data summarized above, the point of departure (POD) selected for the
quantification of cancer risk from DCA is the BMDL of 2.1 mg/kg-day, derived from the fit of
the multistage model to the cancer incidence data in male mice, with the high-dose group
excluded.
Extrapolation to Low Dose
In 1996, the U.S. EPA published its Proposed Guidelines for Carcinogen Risk
Assessment (U.S. EPA, 1996a). The guidelines were recently updated by the Agency (U.S.
EPA, 1999, 2003). Under the revised guidelines, two alternative approaches may be used to
quantify cancer risk, depending on what is known about the mode of carcinogenicity and the
shape of the dose-response curve. A linear default approach is used for a chemical when
available evidence indicates that the chemical is mutagenic or DNA-reactive, or supports another
mode of action that is anticipated to be linear. An inference of linearity may also be supported if
existing human exposure is thought to be on the linear part of a dose-response curve, even
though the overall dose-response is sub-linear. The linear approach is used as a matter of policy
if the mode of carcinogenicity is not understood. Non-linear models may be used when the
mode of carcinogenicity is reasonably understood, and the weight of evidence supports the
conclusion that the dose-response curve is likely to be nonlinear.
92
-------
Multistage Model with 0.95 Confidence Level
1
0.9
0.8
-------
humans who consume chlorinated drinking water) are also likely to increase cancer risk.
Assuming that low doses of DC A are carcinogenic, then the estimate of cancer potency derived
from the dose-response study in male mice is considered to be reliable.
The BMDL is estimated based on four dose groups plus a control, from a high-quality
study. The BMDL estimate is moderately sensitive to the model selected to estimate it, but the
best-fitting models, which do not assume zero slope at zero dose, all predict BMDLs that are
approximately within a factor of two of the multistage model, with the majority of models
predicting BMDLs between 2.0 and 3.0 mg/kg-day.
Based on these considerations (strong dose-response data and good model fits in mice,
but lack of understanding of the mode of action), confidence in the quantification of the cancer
risk for DC A is rated as medium.
94
-------
6. MAJOR CONCLUSIONS IN THE CHARACTERIZATION
OF HAZARD AND DOSE-RESPONSE
6.1. HUMAN HAZARD POTENTIAL
Exposure Pathways
Dichloroacetic acid occurs at low levels in most drinking water systems that are
disinfected with chlorine. Humans are exposed mainly by ingestion of chlorinated drinking
water. Dermal contact may also occur during showering or bathing. Because DCA has low
volatility, inhalation exposure is not expected to be significant.
Toxicokinetics
Dichloroacetic acid is well absorbed by the gastrointestinal tract and is distributed to
multiple tissues throughout the body. Dichloroacetic acid is metabolized in the liver by
oxidative dechlorination to yield glyoxylate, which can enter intermediary metabolism and either
be oxidized to oxalate and excreted, converted to carbon dioxide, and/or incorporated into amino
acids or other cellular molecules. In most species, including humans, clearance of DCA from the
plasma is relatively rapid, with a half-time of 2-3 hours for single doses. In contrast, limited data
suggest that dogs clear single doses of DCA more slowly. Repeated exposure to DCA results in
a decreased ability to metabolize DCA in humans and animals, most likely because DCA inhibits
GSTZ, the cytosolic enzyme needed to carry out the metabolism of the parent compound via
oxidative dechlorination. Inhibition of DCA metabolism can result in blood levels of DCA that
are 8 to 10 times higher than after single doses. DCA can also be metabolized to
monochloroacetic acid and thiodiacetic acid.
95
-------
Characteristic Non-cancer Effects
Dichloroacetic acid causes a wide spectrum of adverse effects in animals and humans,
including:
Effects on Metabolism
Dichloroacetic acid treatment results in a significant reduction in plasma levels of
glucose, pyruvate and lactate. This finding has been consistently reported in DCA-treated rats,
dogs, and humans. Metabolic effects are seen in humans at oral doses as low as 10 mg/kg-day.
DC A exposure also decreases plasma cholesterol levels. The metabolic effects of DC A have led
to its experimental use in the treatment of lactic acidosis, diabetes, and familial
hypercholesteremia. As noted in the section on hepatotoxicity below, DCA exposure induces a
marked accumulation of glycogen in the liver and thus bears some similarity to glycogen storage
disease VI in humans. The metabolic basis for this accumulation has not been elucidated. Most
DCA-induced metabolic alterations appear to be transient, with full recovery to basal/control
levels following cessation of DCA administration, although in at least one study, liver glycogen
levels in mice became resistant to change after 8-weeks of DCA administration (Kato-Weinstein
etal., 1998).
Hepatic Toxicity
DCA causes a dose-related increase in liver size, generally accompanied (or caused) by
an increase in glycogen deposition in the liver. Liver toxicity, as evidenced by increases in
serum levels of liver enzymes, has been seen in DCA-treated mice, rats, dogs, and humans.
Hepatic necrosis has been consistently reported in mice exposed to high doses of DCA. Frank
liver necrosis has not been seen in rats, even at the highest concentration used, nor has it been
reported in dogs or humans.
Reproductive/Developmental Toxicity
In males, DCA causes decreases in testicular weight and viable sperm production.
Testicular effects were observed in rats and dogs. Dogs are apparently the most sensitive
species, displaying testicular toxicity at a dose substantially lower than for other test species. In
96
-------
female rats, DCA exposure during gestation can lead to impaired fetal maturation and result in
soft tissue anomalies (primarily of cardiac origin) in the offspring.
Neurotoxicity
In humans, exposure to DCA causes sedation in many individuals, and occasionally
results in reversible peripheral neuropathy. Neurological effects have also been reported in rats
and dogs, including hind limb paralysis and morphologic alterations in the CNS. Gait
abnormalities have been observed in weanling and adult rats exposed to DCA either by gavage
or via the drinking water route. To date, signs of neurologic toxicity or morphologic changes of
the nervous system in DCA-treated mice have not been reported.
Characteristic Cancer Effects
There are no apparent studies which have been conducted to explore whether DCA
exposure is associated with increased cancer risk in humans. Multiple studies in rodents,
however, have revealed that DCA exposure causes increased incidence and multiplicity of
hepatic adenomas and adenocarcinomas. This effect has been observed in male and female mice
and in male rats. Livers of DCA-treated mice displayed adenomas and carcinomas that
developed from eosinophilic, dysplastic, basophilic or clear cells indicating several origins for
the tumorous growths. Increased tumors have not been observed in other rodent tissues.
Mode of Action
Non-cancer Effects
Dichloroacetic acid is known to inhibit the protein kinase that maintains pyruvate
dehydrogenase in its inactive form. By inhibiting this protein kinase, the activity of pyruvate
dehydrogenase is increased, which in turn results in a spectrum of changes in intermediary
metabolism, including a decrease in plasma glucose and lactate. Dichloroacetic acid has also
been shown to be a noncompetitive inhibitor of the rate limiting microsomal enzyme in
cholesterol biosynthesis (HMG CoA reductase), which likely accounts for its effect on plasma
cholesterol levels. It may inhibit glycogen phosphorylase b kinase or a debranching enzyme
leading to hepatic glycogen accumulation and GSTZ, possibly leading to the accumulation of
alkylating tyrosine metabolites.
97
-------
The detailed mode of DCA-induced hepatotoxicity is not known. Potential modes that
have been suggested include peroxidative damage secondary to DCA metabolism, abnormal
glycogen storage, infarction caused by extensive liver hypertrophy, and disruption of cell cycle
control through DNA/RNA-centered changes related to signal transduction and nuclear
transcription factors. No specific cellular or molecular hypotheses have been advanced to
explain the neurotoxicity and reproductive toxicity of DCA, but inhibition of key cellular
enzymes in the affected tissues is likely to be involved. The potential relevance of GSTZ
inhibition to the toxic mechanism of DCA is not known.
Cancer Effects
A number of potential modes of DC A-induced hepatocarcinogenicity have been
proposed, including the following:
Direct-acting Genotoxicity
The genotoxicity data base on DCA has been extensively reviewed by several scientific
organizations including IARC (1995), ILSI (1997), WHO (2000), and U.S. EPA (1998c). Based
on an evaluation of data available at the time, IARC (1995) and ILSI (1997) each reached
independent conclusions that DCA was not genotoxic. More recently, WHO (2000) concluded
that, although there is some evidence of DCA being genotoxic, these effects occur at such high
concentrations that they are not likely to be involved in the mode of DCA tumorigenesis. In
contrast, NCEA (U.S. EPA, 1998c) concluded that available data indicate that DCA is a direct
acting genotoxic agent. NCEA (U.S. EPA, 1998c) considered new data that were not available
to IARC and ILSI during their respective reviews. NCEA (U.S. EPA, 1998c) stated that the test
results reveal the ability of DCA to cause mutational damage, both point mutations and
chromosomal aberrations, although generally at relatively high exposure levels. NCEA (U.S.
EPA, 1998c) also indicated that mutations are viewed as exhibiting linear low-dose responses
according to EPA's Proposed Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2003).
Because the data on DCA genotoxicity in vivo are mixed, and because no clear explanation for
the internal disagreement between studies is apparent, EPA considers it prudent to assume that
DCA might be genotoxic, at least under in vivo exposure levels that are associated with
detectable increases in tumor incidence. Whether DCA is genotoxic at lower doses is not
known.
98
-------
Hepatocytotoxicity and Regenerative Hyperplasia
Dichloroacetic acid causes focal or widespread hepatic necrosis in mice at high doses,
and regenerative hyperplasia occurs in some animals in which a tumorigenic response is
observed. However, hepatotoxicity and regenerative hyperplasia have not been observed in rats.
In addition, hepatotoxicity and regenerative hyperplasia have not been observed in mice exposed
to low doses of DCA that were associated with tumor formation. For instance, DeAngelo et al.
(1999) reported an increased cumulative incidence and multiplicity of hepatic tumors in male
mice that exhibited no apparent cytotoxic or regenerative response to DCA. Further, treatment
of rodents with DCA for longer than two weeks decreases cell replication rates (Stauber and
Bull, 1997), indicating that any regenerative hyperplasia occurring shortly after initiation of
treatment is not sustained. The data indicate that this mode of action is not likely to play a role
in DCA-induced hepatocarcinogenicity.
Promotion of Spontaneous Mutations
There are several sets of observations suggesting that DCA may be acting through a
tumor promotion mechanism, including inhibition of proliferation of normal hepatocytes and
stimulated proliferation of c-Jun positive hepatocytes in vitro, and increased hepatic tumor
multiplicity in mice exposed to DCA followed by phenobarbital. The majority of cancer
bioassays, however, indicate that DCA is a complete carcinogen, because it is capable of
inducing cancer when administered alone, both at high doses in short-term assays (50-60 weeks),
and at lower doses with longer exposure periods (> 100 weeks).
Depression of Apoptosis
A proposed general mechanism of tumor promoters is the decreased apoptosis of initiated
cells in a tissue by a promoting compound, thereby allowing the outgrowth of cells previously
programmed to die. Snyder et al. (1995) showed that DCA decreased spontaneous apoptosis in
liver cells of mice exposed to DCA (0.5 or 5 g/L) for up to 30 days. Further, Stauber and Bull
(1997) determined that DCA treatment induced primarily small eosinophilic lesions or foci (1-
100 cells), an observation confirmed by Miller et al. (2000) in their study of DCA-induced tumor
growth. A proposed mechanism of carcinogenesis is that DCA is causing the formation of these
small lesions through suppression of apoptosis.
99
-------
In summary, available data are not adequate to indicate which mode(s) of action is
responsible for the hepatic tumor response of rats and mice to DC A exposure.
6.2. DOSE-RESPONSE
OralRfD
Available data suggest that all of the characteristic noncancer effects in humans and
animals occur with similar dose-response patterns, with effect levels of about 25 mg/kg-day or
higher. However, the dog appears to be somewhat more susceptible, at least with respect to
testicular effects in the male. In this case, clear effects have been noted at a dose of 12.5 mg/kg-
day (Cicmanec et al., 1991), while no NOAEL has been established. Besides testicular toxicity,
neurological changes, hepatic vacuolization, and increased liver weight were observed in males
and females at 12.5 mg/kg-day. The basis for the increased sensitivity of the dog to testicular
toxicity is not certain, but it may be due to relatively low capacity to metabolize DCA and clear
it from the plasma. However, since nothing is known about the mechanism of the testicular
toxicity and the metabolism of DCA is complex, especially after repeated dosing, there is no
basis for concluding that metabolism alone is responsible for testicular effects in dogs at low
doses. Using the LOAEL of 12.5 mg/kg-day identified in the dog, and applying an uncertainty
factor of 3000 to account for use of a LOAEL, extrapolation from animals to humans, and
potential inter-human variability in sensitivity, a duration adjustment and database deficiencies,
an oral RfD of 4.2E-03 mg/kg-day has been derived. Data from study by Cicmanec et al. (1991)
could not be reliably evaluated using the BMD approach.
The overall confidence in the RfD is medium. No adverse effects have been reported in
humans at doses lower than 25 mg/kg-day, but systematic investigations of potential hepatic or
reproductive effects in therapeutically-treated humans have not been performed. Metabolic
effects are seen in humans at oral doses as low as 10 mg/kg-day. Limited toxicokinetic data
indicate that humans clear single DCA doses from the plasma more rapidly than dogs, but data
on comparative metabolism in humans and dogs after multiple doses are lacking, and it has yet to
be established whether the various aspects of DCA toxicity are due to the parent compound or
one or more metabolites.
Oral Cancer Risk
100
-------
The cancer risk from ingestion of DCA was quantified based on a dose-response study in
male mice (DeAngelo et al., 1999). The cumulative incidence of hepatic total tumor incidence
(carcinoma plus adenoma) in the test animals was well-described by several different
dichotomous models, with the multistage model yielding the best fit. Based on this model, the
BMD (the dose that caused a 10% increase in extra risk) was 6.86 mg/kg-day, and the BMDL
was 2.1 mg/kg-day. In accordance with guidelines (U.S. EPA, 1999), the BMDL was used as
the point-of-departure (POD) for quantifying cancer risk. Because the mode of action by which
DCA increases cancer risk is not understood, extrapolation to low dose was performed by
assuming a no-threshold linear dose-response curve between the origin and the POD. This
yields a cancer slope factor of 0.048 (mg/kg-day)"1.
Dichloroacetic acid is a likely human carcinogen that lacks a cohesive mode of action.
This assessment is based on the strength of the evidence in several animal bioassays and is
supported by mechanistic data that suggest a complex etiology for tumor development. There
are a number of independent studies reporting consistently positive carcinogenic results at
roughly comparable doses, site concordance for tumor formation between two species, consistent
observations in different species and sexes, and clear evidence of a dose-response relationship.
The data on mechanism implicate more than one type of cellular change in the origin of tumors
along with defects in intra- and inter-cellular communication pathways. Accordingly, the use of
a linear extrapolation of dose is appropriate in quantifying the cancer risk for DCA.
Inhalation RfC and Cancer Risk
There are no studies of inhalation exposure to DCA. DCA has low volatility, and
inhalation exposure to DCA is not believed to be a significant exposure pathway for most
people. Therefore, no inhalation RfC or unit risk value have been derived.
101
-------
7. REFERENCES
Ames, BN; Gold, LS. (1991a) Too many rodent carcinogens: mitogenesis increases
mutagenesis. Science 249:970-971.
Ames, BN; Gold, LS. (1991b) Mitogenesis, mutagenesis and animal cancer tests. In:
Butterworth, BE; Slaga, TJ; Farland, W; McClain, M, eds. Chemically induced cell
proliferation. Implications for risk assessment: proceedings of the chemically induced cell
proliferation conference; November 29-December 2, 1989; Austin, TX. New York: Wiley-Liss,
Inc.; pp. 1-20.
Anandarajah, K; Kiefer, PM; Donohoe, BS; et al. (2000) Recruitment of double bond isomerase
to serve as a reductive dehalogenase during biodegradation of pentachlorophenol. Biochemistry
39:5303-5311.
Anderson, WB; Board, PG; Gargano, B; et al. (1999) Inactivation of glutathione transferase
Zeta by dichloroacetic acid and other fluorine-lacking a-haloalkanoic acids. Chern Res Toxicol
12:1144-1149.
Anna, CH; Maronpot, RR; Pereira, MA; et al. (1994) Ras proto-oncogene activation in
dichloroacetic acid-, trichloroethylene- and tetrachloroethylene-induced liver tumors in B6C3FJ
mice. Carcinogenesis 15:2255-2261.
Austin, EW; Parrish, JM; Kinder, DH; et al. (1996) Lipid peroxidation and formation of 8-
hydroxydeoxyguanosine from acute doses of halogenated acetic acids. Fundam Appl Toxicol
31:77-82.
Bannasch, P. (1986) Modulation of carbohydrate metabolism during carcinogenesis. Cancer
Detect Preven 9:243-249.
Barton, HA; Bull, R; Schultz, I; et al. (1999) Dichloroacetate (DCA) dosimetry: interpreting
DCA-induced liver cancer dose response and the potential for DCA to contribute to
trichloroethylene-induced liver cancer. Toxicol Lett 106:9-21.
Benane, SG; Blackman, CF; House, DE. (1996) Effect of perchloroethylene and its metabolites
on intercellular communication in clone 9 rat liver cells. J Toxicol Environ Health 48:427-437.
Bhat, HK; Kanz, MF; Campbell, GA; et al. (1991) Ninety-day toxicity study of chloroacetic
acids in rats. Fundam Appl Toxicol 17:240-253.
Blackburn, AC; Coggan, M; Tzeng, H-F; et al. (2001) GSTZld: a new allele of glutathione
transferase zeta and maleylacetoacetate isomerase. Pharmacogenetics 11:671-678.
102
-------
Blackburn, AC; Tzeng, H-F; Anders, MW; et al. (2000) Discovery of a functional
polymorphism in human glutathione transferase zeta by expressed sequence tag database
analysis. Pharmacogenetics 10:49-57.
Board, PG; Baker, RT; Chelvanayagam, G; et al. (1997) Zeta, a novel class of glutathione
transferases in a range of species from plants to humans. Biochem J 328:929-935.
Briining, T; Vamvakas, S; Makropoulos, V; et al. (1998) Acute intoxication with trichloroethane:
clinical symptoms, toxicokinetics, metabolism, and development of biochemical parameters for
renal damage. Toxicol Sci 41: 157-65.
Bruschi, SA; Bull, RJ. (1993) In vitro cytoxicity of mono-, di-, and trichloroacetate and its
modulation by hepatic peroxisome proliferation. Fundam Appl Toxicol 21:366-375.
Bull, RJ. (2000) Mode of action of liver tumor induction by trichloroethylene and its
metabolites, trichloroacetate and dichloroacetate. Env Health Perspect 108:241-259.
Bull, RJ; Sanchez, JJVI; Nelson, MA; et al. (1990) Liver tumor induction in B6C3Fj mice by
dichloroacetate and trichloroacetate. Toxicology 63:341-359.
Carter, JH; Carter, FEW; DeAngelo, AB. (1995) Biochemical, pathologic and morphometric
alterations induced in male B6C3Fj mouse liver by short-term exposure to dichloroacetic acid.
Toxicol Lett 81:55-71.
Carter, JH; Carter, HW; Deddens, JA; et al. (2003) A 2-year dose-response study of lesion
sequences during hepatocellular carcinogenesis in the male B6C3Fj mouse given the drinking
water chemical dichloroacetic acid. Environ Health Perspect 111:53-64.
Chang, LW; Daniel, FB; DeAngelo, AB. (1992) Analysis of DNA strand breaks induced in
rodent liver in vivo, hepatocytes in primary culture, and a human cell line by chlorinated acetic
acids and chlorinated acetaldehydes. Environ Mol Mutagens 20:277-88.
Chen, CW. (2000) Biologically based dose-response model for liver tumors induced by
trichloroethylene. Environ Health Perspect 108:335-342.
Cicmanec, JL; Condie, LW; Olson, GR; et al. (1991) 90-day toxicity study of dichloroacetate in
dogs. Fundam Appl Toxicol 17:376-389.
Clontech. (2000) New Atlas mouse cancer 1.2 array. CLONTECHniques. BD Biosciences
Clontech, San Jose, CA. Available from:
Cochat, P; Deloraine, A; Rotily, M; et al. (1995) Epidemiology of primary hyperoxaluria type
1. Societe de Nephrologie and Societe de Nephrologie Pediatrique Nephrol Dial Transplant
10:3-7.
103
-------
Cornett, R; James, MO; Henderson, GN; et al. (1999) Inhibition of gluathione S-transferase
and tyrosine metabolism by dichloroacetate: a potential unifying mechanism for its altered
biotransformation and toxicity. Biochem Biophys Research Comm 262:752-756.
Cornett, R; Yan, Z; Henderson, G; et al. (1997) Cytosolic biotransformation of dichloroacetic
acid (DCA) in the Sprague-Dawley rat. Fundam Appl Toxicol 36 (supplXX):318.
Cosby, NC; Dukelow, WR. (1992) Toxicology of maternally ingested trichloroethylene (TCE)
on embryonal and fetal development in mice and of TCE metabolites on in vitro fertilization.
Fundam Appl Toxicol 19:268-74.
Counts, JL; Goodman, JI. (1994) Hypomethylation of DNA: an epigenetic mechanism involved
in tumor promotion. Mol Carcinog 11:185-188.
Counts, JL; Goodman, JI. (1995) Alterations in DNA methylation may play a variety of roles in
carcinogenesis. Cell 83:13-15.
Crabb, DW; Yount, EA; Harris, RA. (1981) The metabolic effects of dichloroacetate.
Metabolism 30:1024-1039.
Curry, SH; Chu, PI; Baumgartner, TG; et al. (1985) Plasma concentrations and metabolic
effects of intravenous sodium dichloroacetate. Clin Pharmacol Therap 37:89-93.
Curry, SH; Lorenz, A; Chu, PI; et al. (1991) Disposition and pharmacodynamics of
dichloroacetate (DCA) and oxalate following oral DCA doses. Biopharm Drug Disp 12:375-
390.
Daniel, FB; DeAngelo, AB; Stober, JA; et al. (1992) Hepatocarcinogenicity of chloral hydrate,
2-chloroacetaldehyde, and dichloroacetic acid in male B6C3Fj mouse. Fundam Appl Toxicol
19:159-168.
Davis, ME. (1986) Effect of chloroacetic acids on the kidneys. Environ Health Perspect
69:209-214.
Davis, ME. (1990) Subacute toxicity of trichloroacetic acid in male and female rats.
Toxicology 63:63-72.
DeAngelo, A B; Daniel, FB; McMillan, L; et al. (1989) Species and strain sensitivity to the
induction of peroxisome proliferation by chloroacetic acids. Toxicol Appl Pharmacol 101:285-
298.
DeAngelo, AB; Daniel, FB; Most, BM; et al. (1996) The carcinogenicity of dichloroacetic acid
in the male Fischer 344 rat. Toxicology 114:207-221.
104
-------
DeAngelo, AB; Daniel, FB; Stober, JA; et al. (1991) The carcinogenicity of dichloroacetic acid
in the male B6C3FJ mouse. Fundam Appl Toxicol 16:337-347.
DeAngelo, AB; George, MH; House, DE. (1999) Hepatocarcinogenicity in the male B6C3Fj
mouse following a life-time exposure to dichloroacetic acid in the drinking water: dose-response
determination and modes of action. J Toxicol Environ Health 58:485-507.
DeMarini, DM; Perry, E; Sheldon, ML. (1994) Dichloroacetic acid and related compounds:
induction of prophage in E. coli and mutagenicity and mutation spectra in Salmonella TA 100.
Mutagenesis 9:429-437.
Epstein, DL; Nolen, GA; Randall, JL; et al. (1992) Cardiopathic effects of dichloroacetate in
the fetal Long-Evans rat. Teratology 46:225-235.
Evans, OB; Stacpoole, PW. (1982) Prolonged hypolactatemia and increased total pyruvate
dehydrogenase activity by dichloroacetate. Biochem Pharmacol 31:1295-1300.
Everhart, JL; Kurtz, DT; McMillan, JM. (1998) Dichloroacetic acid induction of peroxisome
proliferation in cultured hepatocytes. J Biochem Mol Toxicol 12:351-359.
Fernandez-Canon, JM; Penalva, MA. (1998) Characterization of a fungal maleylacetoacetate
isomerase gene and identification of its human homologue. J Biol Chem 273:329-327.
Fernandez-Canon, JM; Hejna, J; Reifsteck, C; et al. (1999) Gene structure, chromosomal
location, and expression pattern of maleylacetoacetate isomerase. Genomics 58:263-269.
Fernandez-Canon, JM; Baetscher, MW; Finegold, M; et al. (2002) Maleylacetoacetate
isomerase (MAII/GSTZ)-deficient mice reveal a glutathione-dependant nonenzymatic bypass in
tyrosine catabolism. Mol Cell Biology 22(13):4943-4951.
Ferreira-Gonzalez, A; DeAngelo, AB; Nasim, S; et al. (1995) Ras oncogene activation during
hepatocarcinogenesis in B6C3Fj male mice by dichloroacetic and trichloroacetic acid.
Carcinogenesis 16:495-500.
Fox, AW; Yang, X; Murli, H; et al. (1996) Absence of mutagenic effects of sodium
dichloroacetate. Fundam Appl Toxicol 32:87-95.
Fuscoe, JC; Afshari, AJ; George, MH; et al. (1996) In vivo genotoxicity of dichloroacetic acid:
evaluation with the mouse peripheral blood micronucleus assay and the single cell gel assay.
Environ Mol Mutagen 27:1-9.
Giller, S; Le Curieux, F; Erb, F; et al. (1997) Comparative genotoxicity of halogenated acetic
acids found in drinking water. Mutagenesis 12:321-328.
105
-------
Gonzalez-Leon, A; Bull, RJ. (1996) Continuous treatment with dichloroacetate in drinking
water alters its metabolism in rats but not mice. Fundam Appl Toxicol (Suppl 30; Abstract
11080):217.
Gonzalez-Leon, A; Schultz, IR; Bull, RJ. (1997a) Species differences in the toxicokinetics of
dichloroacetate and trichloroacetate in F344 rats and B6C3FJ mice after prolonged
administration in drinking water. Fundam Appl Toxicol (Suppl 36):33.
Gonzalez-Leon, A; Schultz, IR; Xu, G; et al. (1997b) Pharmacokinetics and metabolism of
dichloroacetate in the F344 rat after prior administration in drinking water. Toxicol Appl
Pharmacol 146:189-195.
Gonzalez-Leon, A; Merdink, JL; Bull, RJ; et al. (1999) Effect of pretreatment with
dichloroacetic or trichloroacetic acid in drinking water on the pharmacokinetics of a subsequent
challenge dose in B6C3FJ mice. Chem-Biol Interact 123: 239-253.
Harrington-Brock, K; Doerr, CL; Moore, MM. (1998) Mutagenicity of three disinfection by-
products; di- and trichloroacetic acid and chloral hydrate in L5178Y/TK +/- - 3.7.2C mouse
lymphoma cells. Mutat Res 413:265-276.
Haworth, S; Lawlor, T; Mortelmans, K; et al. (1983) Salmonella mutagenicity results for 250
chemicals. Environ Mut 5 (Suppl 1): 1-142.
Herbert, V; Gardner, A; Colman, N. (1980) Mutagenicity of dichloroacetate, an ingredient of
some formulations of pargamic acid (vitamin B15). Am J ClinNutr 33:1179-1182.
Herren-Freund, SL; Pereira, MA; Khoury, DK; et al. (1987) The carcinogenicity of
trichloroethylene and its metabolites, trichloroacetic acid and dichloroacetic acid, in mouse liver.
Toxicol Appl Pharm 90:183-189.
Hers, HG; van Hoof, F; deBarsey, T. (1989) Glycogen storage diseases. In: The metabolic basis
of inherited disease. Scriver, CR; Beaudet, AL; Sly, WS; et al., eds. Sixth Edition. New York:
MacGraw Hill Information Service Co. pp. 425-452.
Hunter, ES; Rogers, EH; Schmid, JE; et al. (1996) Comparative effects of haloacetic acids in
whole embryo culture. Teratology 54:57-64.
IARC (International Agency for Research on Cancer). (1979) IARC Monographs on the
evaluation of carcinogenic risks to humans. Some halogenated hydrocarbons. Volume 20, pp.
491-514.
IARC. (1995) IARC Monographs on the evaluation of carcinogenic risks to humans. Dry
cleaning, some chlorinated solvents and other industrial chemicals. Volume 63, pp. 271-314.
106
-------
IARC. (2002) IARC Monographs on the evaluation of carcinogenic risks to humans. Working
group announcements. Some drinking water disinfectants and contaminants, including arsenic.
Volume 84, pp. 15-22; October 2002.
ILSI (International Life Sciences Institute). (1997) An evaluation of EPA's proposed guidelines
for carcinogenic risk assessment using chloroform and dichloroacetate as case studies: report of
an expert panel. Health and Environmental Sciences Institute, Washington D.C.
IRIS (Integrated Risk Information System), U.S. Environmental Protection Agency. (1996) IRIS
summary document for dichloroacetic acid. March 1996.
James, MO; Yan, Z; Cornett, R; et al. (1998) Pharmacokinetics and metabolism of
[14C]dichloroacetate in male Sprague-Dawley rats. Identification of glycine conjugates,
including hippurate, as urinary metabolites of dichloroacetate. Drug Metab Disp 26:1134-1143.
Johnson, R; Spiegelman, B; Hanahan, D; et al. (1996) Cellular transformation and malignancy
by ras require c-jun. Mol Cell Bio 16(8):4504-4511.
Kato-Weinstein, J; Lingohr, MK; Orner, GA; et al. (1998) Effects of dichloroacetate on
glycogen metabolism in B6C3FJ mice. Toxicology 130:141-54.
Katz, R; Tai, CN; Diener, RM; et al. (1981) Dichloroacetate, sodium: 3-month oral toxicity
studies in rats and dogs. Toxicol Appl Pharmacol 57:273-287.
Kerbey, AL; Randle, PJ; Cooper, RH; et al. (1976) Regulation of pyruvate dehydrogenase in rat
heart. Mechanism of regulation of proportions of dephosphorylated and phosphorylated enzyme
by oxidation of fatty acids and ketone bodies and of effects of diabetes: role of coenzyme A,
acetyl-coenzyme A and reduced and oxidized nicotinamide-adenine dinucleotide. Biochem J
154:327-348.
Kim, H; Haltmeier, P; Klotz, JB; et al. (1999) Evaluation of biomarkers of environmental
exposures: Urinary haloacids associated with ingestion of chlorinated drinking water. Environ
Res 80:187-195.
LaBerge, C; Lescault, A; Tanguay, RM. (1986) Hereditary tyrosinemias (type I): a new vista on
tyrosine toxicity. Adv Exp Med Biol 206:209-221.
Lantum, HBM; Board, PG; Anders, MW. (2002) Kinetics of the biotransformation of
maleylacetone and chlorofluoroacetate by polymorphic variants of human glutathione transferase
zeta(hgstZl-l). Chem Res Toxicol 15:957-963.
Lapeyre, JN; Becker, JJ. (1981) DNA methylation and methylase levels in normal and malignant
mouse hepatic tumor tissue. Carcinogenesis 2:873-878.
107
-------
Larson, JL; Bull, RJ. (1992) Metabolism and lipoperoxidative activity of trichloroacetate and
dichloroacetate in rats and mice. Toxicol Appl Pharmacol 115:268-277.
Lash, LH; Parker, JC. (2001) Hepatic and renal toxicities associated with perchloroethylene.
Pharm Rev 53:177-208.
Leavitt, SA; DeAngelo, AB; George, MH; et al. (1997) Assessment of the mutagenicity of
dichloroacetic acid in lacl B6C3FJ mouse liver. Carcinogenesis 18:2101-2106.
Lewis, RJ, Sr. (ed.). (1997) Hawley's Condensed Chemical Dictionary. 13th edition. New York:
John Wiley & Sons, Inc.
Lin, EL; Mattox, JK; Daniel, FB. (1993) Tissue distribution , excretion, and urinary metabolites
of dichloroacetic acid in the male Fischer 344 rat. J Toxicol Environ Health 38(1): 19-32.
Linder, RE; Klinefelter, GR; Strader, LF; et al. (1997) Spermatotoxicity of dichloroacetic acid.
Reprod Toxicol ll(5):681-86.
Lipscomb, JC; Mahle, DA; Brashear, WT; et al. (1995) Dichloroacetic acid: metabolism in
cytosol. Drug Metab Disp 23:1202-1205.
Lukas, G; Vyas, KH; Brindle, SD; et al. (1980) Biological disposition of sodium dichloroacetate
in animals and humans after intravenous administration. J Pharm Sci 69:419-421.
Maloney, EK; Waxman, DJ. (1999) Trans-activation of PPARa and PPARy by structurally
diverse environmental chemicals. Toxicol Appl Pharmacol 161:209-218.
Marnett, GN; Kurd, MK; Hollstein, MC; et al. (1985) Naturally occurring carbonyl compounds
are mutagens in Salmonella tester strain TA 104. Mutat Res 148:25-34.
Mather, GG; Exon, JH; Roller, LD. (1990) Subchronic 90-day toxicity of dichloroacetic and
trichloroacetic acid in rats. Toxicology 64:71-80.
Matsuda, H; Ose, Y; Nagase, T; et al. (1991) Mutagenicity of ozonation and chlorination of
products from p-hydroxybenzaldehyde. Sci Total Envir 103:141-149.
The Merck Index (1996), 12th edition. Whitehouse Station, NJ: Merck & Co., Inc.
Michal, G (ed.). (1999) Biochemical Pathways. New York: John Wiley & Sons, Inc..
Miller, JH; Minard, K; Wind, RA; et al. (2000) In vivo MRI measurements of tumor growth
induced by dichloroacetate: implications for mode of action. Toxicology 145:115-25.
Montgomery, R; Conway, TW; Spector, AA. (1990) Biochemistry - a case-oriented approach.
Baltimore, MD: The C.V. Mosby Company.
108
-------
Moore, GW; Swift, LL; Rabinowitz, D; et al. (1979) Reduction of serum cholesterol in two
patients with homozygous familial hypercholesterolemia by dichloroacetate. Atherosclerosis
33:285-293.
Moore, MM; Harrington-Brock, K. (2000) Mutagenicity of trichloroethylene and its metabolites:
implications for the risk assessment of trichloroethylene. Env Health Perspect 108:215-223.
Moser, VC; Phillips, PM; McDaniel, KL; et al. (1999) Behavioral evaluation of the
neurotoxicity produced by dichloroacetic acid in rats. Neurotox Teratol 21:719-731.
National Research Council. (1983) Risk assessment in the federal government: managing the
process. Washington, D.C: National Academy Press.
Nelson, MA; Bull, RJ. (1988) Induction of strand breaks in DNA by trichloroethylene and
metabolites in rat and mouse liver in vivo. Toxicol Appl Pharm 94:45-54.
Nelson, MA; Lansing, AJ; Sanchez, IM; et al. (1989) Dichloroacetic acid and trichloroacetic
acid induced DNA strand breaks are independent of peroxisome proliferation. Toxicology
58:239-248.
Ono, Y; Somiya, I; Kawamura, M. (1991) The evaluation of genotoxicity using DNA repairing
test for chemicals produced in chlorination and ozonation processes. Wat Sci Tech 23:329-338.
Parrish, JM; Austin, EW; Stevens, DK; et al. (1996) Haloacetate-induced oxidative damage to
DNA in the liver of male B6C3Fj mice. Toxicology 110:103-111.
Pereira, MA. (1996) Carcinogenic activity of dichloroacetic acid and trichloroacetic acid in the
liver of female B6C3FJ mice. Fundam Appl Toxicol 31:192-199.
Pereira, MA; Phelps, JB. (1996) Promotion by dichloroacetic acid and trichloroacetic acid of N-
methyl-N-nitrosourea-initiated cancer in the liver of female B6C3Fj mice. Cancer Lett 102:133-
141.
Pereira, MA; Kramer, PM; Conran, PB; et al. (2001) Effect of chloroform on dichloroacetic acid
and trichloroacetic acid-induced hypomethylation and expression of the c-myc gene and on their
promotion of liver and kidney tumors in mice. Carcinogenesis 22:1511-1519.
Polekhina, G; Board, PG; Blackburn, AC; et al. (2000) Glutathione transferase zeta-
maleylacetoacetate isomerase (1FW1). Weizmann Institute (Israel) - Atlas of Macromolecules:
Protein database .
Polekhina, G; Board, PG; Blackburn, AC; et al. (2001) Crystal structure of maleylacetoacetate
isomerase/glutathione transferase zeta reveals the molecular basis for its remarkable catalytic
promiscuity. Biochemistry 40:1567-1576.
109
-------
Ribes, G; Valette, G; Loubatieres-Mariani, MM. (1979) Metabolic effects of sodium
dichloroacetate in normal and diabetic dogs. Diabetes 28:852-857.
Richmond, RE; Carter, JH; Carter, HW; et al. (1995) Immunohistochemical analysis of
dichloroacetic acid (DCA)-induced hepatocarcinogenesis in male Fischer (F344) rats. Cancer
Lett 92:67-76.
Saghir, SA; Schultz, IR. (2002) Low-dose pharmacokinetics and oral bioavailability of
dichloroacetate in naive and GST ^-depleted rats. Environ Health Perspect 110:757-763
Saillenfait, AM; Langonne, I; Sabate, JP. (1995) Developmental toxicity of trichloroethylene,
tetrachloroethylene and four of their metabolites in rat whole embryo culture. Arch Toxicol
70:71-82.
Sakai, M; Matsushimia, Y; Nishizawa, M; et al. (1995) Suppression of rat glutathione
transferase P expression by peroxisome proliferators: interactions between Jun and peroxisome
proliferator- activated receptor-a. Cancer Res 55: 5370-5376.
Sanchez, IM; Bull, RJ. (1990) Early induction of reparative hyperplasia in B6C3FJ mice treated
with dichloroacetate and trichloroacetate. Toxicology 64:33-46.
Sasaki, Y; Endo, R. (1978) Mutagenicity of aldehydes in Salmonella. Mutat Res 54: 251-251.
Sayato, Y; Nakamuro, K; Ueno, H. (1987) Mutagenicity of products formed by ozonation of
naphthoresorcinol in aqueous solutions. Mutat Res 189:217-222.
Schroeder, M; DeAngelo, AB; Mass, MJ. (1997) Dichloroacetic acid reduces Ha-ras codon 61
mutations in liver tumors from female B6C3Fj mice. Carcinogenesis 18:1675-1678.
Schultz, IR; Merdink, JL; Gonzalez-Leon, A; et al. (2002) Dichloroacetate toxicokinetics and
disruption of tyrosine catabolism in B6C3F1 mice: dose-response relationships and age as a
modifying factor. Toxicology 173:229-247.
Sheehan, D; Meade, G; Foley, VM; et al. (2001) Structure functions and evolution of glutathione
transferases: implications for classification of nonmammalian members of an ancient enzyme
superfamily. Biochem J 360:1-16.
Smith, MK; Randall, JL; Read, EJ; et al. (1992) Developmental toxicity of dichloroacetate in the
rat. Teratology 46:217-233.
Snyder, RD; Pullman, J; Carter, JH; et al. (1995) In vivo administration of dichloroacetic acid
suppresses spontaneous apoptosis in murine hepatocytes. Cancer Res 55:3702-3705.
Spruijt, L; Naviaux, RK; McGowan, KA; et al. (2001) Nerve conduction changes in patients
with mitochondrial diseases treated with dichloroacetate. Muscle Nerve 24:916-924.
110
-------
Stacpoole, PW. (1989) The pharmacology of dichloroacetate. Metabolism 38:1124-1144.
Stacpoole, PW; Greene, YJ. (1992) Dichloroacetate. Diabetes Care 15:785-791.
Stacpoole, PW; Moore, GW; Kronauser, DM. (1978) Metabolic effects of dichloroacetate in
patients with diabetes mellitus and hyperlipoproteinemia. N Engl J Med 298:526-530.
Stacpoole, PW; Henderson, GN; Yan, Z; et al. (1998a) Pharmacokinetics, metabolism and
toxicology of dichloroacetate. Drug Metab Rev 30: 499-539.
Stacpoole, PW; Barnes, CL; Hurbanis, MD; et al. (1998b) Treatment of congenital lactic
acidosis with dichloroacetate. Arch Disease Childhood 77:535-541.
Stauber, AJ; Bull, RJ. (1997) Differences in phenotype and cell replicative behavior of hepatic
tumors induced by dichloroacetate (DCA) and trichloroacetate (TCA). Toxicol Appl Pharmacol
144:235-46.
Stauber, AJ; Bull, RJ; Thrall, BD. (1998) Dichloroacetate and trichloroacetate promote clonal
expansion of anchorage-independent hepatocytes in vivo and in vitro. Toxicol Appl Pharmacol
150:287-294.
Stevens, DK; Eyra, RJ; Bull, RJ. (1992) Adduction of hemoglobin and albumin in vivo by
metabolites of trichloroethylene, trichloroacetate, and dichloroacetate in rats and mice. Fundam
Appl Toxicol 19:336-342.
Stroger, R; Kubicka, P; Liu, C; et al. (1993) Maternal specific methylation of the imprinted
mouse Igf2r locus identifies the expressed locus as carrying the imprinted signal. Cell 73:61-71.
Tanguay, RM; Jorquera, R; Poudrier, J; et al. (1996) Tyrosine and its catabolites: from disease to
cancer. Acta Biochim Pol 43:209-216.
Tao, L; Li, K; Kramer, PM; et al. (1996) Loss of heterozygosity on chromosome 6 in
dichloroacetic acid and trichloroacetic acid-induced liver tumors in female B6C3FJ mice. Cancer
Lett 108:257-261.
Tao L; Kramer, PM; Ge, R; et al. (1998) Effect of dichloroacetic acid and trichloroacetic acid on
DNA methylation in liver and tumors of female B6C3FJ mice. Toxicol Sci 43:139-144.
Tao, L; Yang, S; Xie, M; et al. (2000) Effect of trichloroethylene and its metabolites,
dichloroacetic acid and trichloroacetic acid, on the methylation and expression of c-Jun and c-
Myc protooncogenes in mouse liver: prevention by methionine. Toxicol Sci 54:399-407.
Thai, SF; Allen, JW; DeAngelo, AB; et al. (2003) Altered gene expression in mouse livers after
dichloroacetic acid exposure. Mutat Res 543(2): 167-180.
Ill
-------
Thai, SF; Allen, JW; DeAngelo, AB; et al. (2001) Detection of early gene expression changes by
differential display in the livers of mice exposed to dichloroacetic acid. Carcinogenesis 22:1317-
1322.
long, Z; Board, PG; Anders, MW. (1998a) Glutathione transferase zeta-catalyzed
biotransformation of dichloroacetic acid and other a-haloacids. Chem Res Toxicol 11:1332-
1338.
Tong, Z; Board, PG; Anders, MW. (1998b) Glutathione transferase Zeta catalyses the
oxygenation of the carcinogen dichloroacetic acid to glyoxylic acid. Biochem J 331:371-374.
Toth, GP; Kelty, KC; George, EL; et al. (1992) Adverse male reproductive effects following
subchronic exposure of rats to sodium dichloroacetate. Fundam Appl Toxicol 19:57-63.
Tsai, WH; DeAngelo, AB. (1996) Responsiveness of hepatocytes from dichloroacetic acid or
phenobarbital treated mice to growth factors in primary culture. Cancer Lett 99:177-183.
Tzeng, H-F, Blackburn, AC; Board, PG; et al. (2000) Polymorphism- and species-dependent
inactivation of glutathione transferase zeta by dichloroacetate. Chem Res Toxicol 13: 231-236.
U.S. EPA. (U.S. Environmental Protection Agency) (1986a) Guidelines for carcinogen risk
assessment. Federal Register 51(185):33992-34003.
U.S. EPA. (1986b) Guidelines for the health risk assessment of chemical mixtures. Federal
Register 51(185):34014-34025.
U.S. EPA. (1986c) Guidelines for mutagenicity risk assessment. Federal Register
51(185):34006-34012.
U.S. EPA. (1988) Recommendations for and documentation of biological values for use in risk
assessment. EPA 600/6-87/008. Available from: National Technical Information Service
(NTIS), Springfield, VA; PB 88-179874/AS (February 1988).
U.S. EPA. (1991a) Guidelines for developmental toxicity risk assessment. Federal Register
56(234):63798-63826.
U.S. EPA. (1991b) Toxicology of the chloroacetic acids by-products of the drinking water
disinfection process. II. The comparative carcinogenicity of dichloroacetic and trichloroacetic
acid: implication for risk assessment. Document No. HERL-0820. Research Triangle Park, NC:
Health Effects Research Laboratory, U.S. EPA.
U.S. EPA. (1992) Guidelines for exposure assessment. Federal Register 57:22888-22938.
U.S. EPA. (1994a) Final draft for the drinking water criteria document on chlorinated
acids/aldehydes/ketones/alcohols. March 31, 1994.
112
-------
U.S. EPA. (1994b) Interim policy for particle size and limit concentration issues in inhalation
toxicity: notice of availability. Federal Register 59(206):53799.
U.S. EPA. (1994c) Methods for derivation of inhalation reference concentrations and application
of inhalation dosimetry. EPA/600/8-90/066F.
U.S. EPA. (1994d) Peer review and peer involvement at the U.S. Environmental Protection
Agency. Memorandum from U.S. EPA Administrator, Carol M. Browner, dated June 7, 1994.
Available at: .
U.S. EPA. (1995) The use of the benchmark dose approach in heath risk assessment.
EPA/630/R-94/007.
U.S. EPA. (1996a) Proposed guidelines for carcinogen risk assessment. Federal Register
61(79):17960-18011.
U.S. EPA. (1996b) Reproductive toxicity risk assessment guidelines. Federal Register
61(212):56274-56322.
U.S. EPA. (1998a) Proposed guidelines for neurotoxicity risk assessment. Federal Register
63(93):26926-26954.
U.S. EPA. (1998b) Science policy council handbook: peer review. Prepared by the Office of
Science Policy, Office of Research and Development, Washington, DC. EPA 100-B-98-001.
U.S. EPA. (1998c) Dichloroacetic acid: carcinogenicity identification characterization summary.
NCEA-W-0372.
U.S. EPA. (1998d) National primary drinking water regulations: disinfectants and disinfection
byproducts. Final Rule. Federal Register 63:69406-69407. December 16, 1998.
U.S. EPA (1998e) Estimating noncancer health risk reduction benefits: a proposed method and
case study. Prepared by Abt Associates for the Office of Water under EPA Contract 68-C6-0059.
Work Assignment 0-14.
U.S. EPA. (1999) Guidelines for carcinogen risk assessment. Review draft, NCEA-F-0644,
July. Risk Assessment Forum, Washington, DC. Available at: .
U.S. EPA. (2000a) Science policy council handbook: peer review. Second edition. Prepared by
the Office of Science Policy, Office of Research and Development, Washington, DC. EPA 100-
B-00-001.
U.S. EPA. (2000b) Science policy council handbook: risk characterization. Prepared by the
Office of Science Policy, Office of Research and Development, Washington, DC.
113
-------
EPA 100-B-00-002.
U.S. EPA. (2000c) Benchmark dose technical support document. External review draft. Office
of Research and Development, Risk Assessment Forum, Washington, DC. EPA/630/R-00/001.
U.S. EPA. (2000d) Supplementary guidance for conducting health risk assessment of chemical
mixtures. Office of Research and Development, Risk Assessment Forum, Washington, DC.
EPA/630/R-00/002.
U.S. EPA. (2003) Draft final guidelines for carcinogen risk assessment. Risk Assessment
Forum, Washington, DC. NCEA-F-0644A. Available at: < http://www.epa.gov/ncea/raf/
cancer2003.htm>
Velazquez, SF. (1995) Activation of the H-ras oncogene by drinking water disinfection by-
products. GRA&I, 1-43 (NTIS/PB95-200515).
Ward, KW; Roger, EH; Hunter, ES. (2002) Comparative pathogenesis of haloacetic acid and
protein kinase inhibitor embryotoxicity in mouse whole embryo culture. Toxicol Sci 53:118-126.
Waskell, L. (1978) A study of the mutagenicity of anesthetics and their metabolites. MutatRes
57:141-143.
Wempe, MF; Anderson, WB; Tzeng, HF; et al. (1999) Glutathione transferase zeta-catalyzed
biotransformation of deuterated dihaloacetic acids. Biochem Biophys Res Comm 261:779-783.
Whitehouse, S; Cooper, RH; Randle, PJ. (1974) Mechanism of activation of pyruvate
dehydrogenase by dichloroacetate and other halogenated carboxylic acids. Biochem J 141:761-
774.
WHO (World Health Organization). (2000) Environmental Health Criteria Monograph 216.
Disinfectants and Disinfectant By-product. Geneva, Switzerland.
Xu, G; Stevens, DK; Bull, RJ. (1995) Metabolism of bromodichloroacetate in B6C3F1 in mice.
Drug MetabDispos 23:1412-1416.
Yamaguchi, T; Nakagawa, K. (1993) Mutagenicity of formation of oxygen radicals by trioses
and glyoxal derivatives. Agric Biol Chem 47:2461-2465.
Yount, EA; Felten, SY; O'Connor, BL; et al. (1982) Comparison of the metabolic and toxic
effects of 2-chloropropionate and dichloroacetate. J Pharmacol Exp Ther 222:501-508.
Zhou, Y-C; Waxman, DJ. (1998) Activation of peroxisome proliferator-activated receptors by
chlorinated hydrocarbons and endogenous steroids. Environ Health Perspect 106:983-988.
114
-------
APPENDIX A
RESPONSE TO PEER REVIEW SUMMARY DOCUMENT
-------
APPENDIX A
External Peer Review—Summary of Comments and Disposition
The draft Toxicological Review and the IRIS Summary for Dichloroacetic Acid have
undergone internal peer review performed by scientists within EPA and a more formal external
peer review performed by scientists in accordance with EPA guidance on peer review (U.S. EPA,
1994d). Comments made by the internal reviewers were addressed prior to submitting the
documents for external peer review and are not part of this appendix. The external peer reviewers
were tasked with providing written answers to general questions on the overall assessment and on
chemical-specific questions in areas of scientific controversy or uncertainty. A summary of
significant comments made by the external reviewers and EPA's response to these comments
follows.
Disposition of Specific Charge Questions
Question 1. Does the documentation successfully communicate the essential components
and findings of the source documents?
Comment: One reviewer thought that the examination of the database across studies and
discussions of modes of action should have involved a more substantial and critical discussion,
making use of other information from the basic sciences and the toxicology of other chemicals.
Response: Revisions to the report have focused on increasing the background information for a
number of the mechanistic studies in Section 4.4.1. and providing more integration of data in the
synthesis sections (Sections 4.5. and 4.6.) of the Toxicological Review.
Comment: One reviewer suggested that there be more discussion regarding inconsistencies in
results across similar studies.
Response: Additional discussion regarding inconsistencies across similar studies, particularly in
regards to the genotoxicity studies in Section 4.4.2, has been added.
Comment: One reviewer was concerned that the external review draft appeared to deviate from
the Agency's cancer guidelines and other precedents by establishing the dual hazard classification
in the case where the dose-response relationship is uncertain ("likely to be a carcinogen at high
A-l
-------
exposure levels, but cannot be classified at exposure levels that are associated with
environmentally relevant exposure conditions"). The reviewer indicated that this was
inconsistent with the Agency's March 1998 Carcinogenicity Identification Characterization
Summary and recommended that this change be reviewed.
Response: The cancer classification has been changed and is no longer a "dual classification."
Question 2. Are there any significant publications that are not included in the Toxicological
Review document?
Comment: One reviewer noted several papers related to the metabolism of DC A that seem to
have been overlooked, including:
Austin, EW and Bull, RJ (1997) Effect of pretreatment with dichloroacetate or trichloroacetate
on the metabolism of bromodichloroacetate. J Toxicol Environ Health 52:367-383.
Austin, EW; Parrish, JM; Kinder, DH; et al. (1996) Lipid peroxidation and formation of 8-
hydroxyguanosine from acute doses of halogenated acetic acids. Fundam Appl Pharmacol 31:77-
82.
Lingohr, MK; Thrall, BD; Bull, RJ. (2001) Effects of dichloroacetate (DCA) on serum insulin
levels and insulin-controlled signaling proteins in livers of male B6C3F1 mice. Tox Sci 59:178-
184.
Schultz, IR; Merdink, JL; Gonzalez-Leon, A; et al. (1999) Comparative toxicokinetics of
chlorinated and brominated haloacetates in F344 rats. Toxicol Appl Pharmacol 158:103-114.
Xu, X; Stevens, DK; Bull, RJ. (1995) Metabolism of bromodichloroacetate in B6C3F1 mice.
DrugMetab andDisp 23:1412-1416
The reviewer suggested that EPA conduct a more expansive literature search to include the
technically more correct terms: dichloroacetate, haloacetates, and haloacetic acids.
Response: The suggested papers were obtained and reviewed. In addition, a literature search was
conducted covering the period from the original literature search (1998) to the present. Papers
identified from that search were also retrieved and reviewed. Some, but not all of the papers
suggested by the reviewer and identified in the new literature search were added to the
Toxicology Review. Studies that were not assimilated (Austin and Bull [1997]and Lingohr et al.
[2001]) were not added to the report because they did not significantly augment the DCA
A-2
-------
discussions in the revised report.
Comment: One reviewer recommended including the following study in the Toxicological
Review:
Spruijt, L; Naviaux, RK; McGowan, KA; et al. (2001) Nerve conduction changes in patients with
mitochondrial diseases treated with dichloroacetate. Muscle Nerve 24(7):916-924.
The reviewer also commented that the study should be considered for inclusion in Table 5-1,
Figure 5-1, and for comparison in the reference-dose calculations.
Response: Data from the paper was included in the section for noncarcinogenic, systemic effects.
The data were also included in Table 5-1 and in Figure 5-1.
Comment: One reviewer recommended that the following recent papers on DCA mechanism of
action be included:
Pereira, MA; Kramer, PM; Conran, PB; et al. (2001) Effect of chloroform on dichloroacetic acid
and trichloroacetic acid-induced hypomethylation and expression of the c-myc gene and on their
promotion of liver and kidney tumors in mice. Carcinogenesis 22(9):1511-9
Thai, SF; Allen, JW; DeAngelo, AB; et al. (2001) Detection of early gene expression changes by
differential display in the livers of mice exposed to dichloroacetic acid. Carcinogenesis
22(8): 1317-22.
Response: The studies have been included in the document.
Comment: One reviewer wanted additional discussion of the potential for DCA to mediate the
Carcinogenesis of other chlorinated compounds (e.g., per- and trichloroethylene) because data
show that these compounds are metabolized into DCA in vivo (see the IARC Monograph [IARC,
1995] on perchloroethylene for discussion).
Response: The draft external peer review document discusses briefly the relevance of metabolism
of perchloroethylene and trichloroethylene to DCA and resultant Carcinogenesis. A significant
amount of text has not been devoted to this subject, because some studies indicate that the amount
of DCA produced from trichloroethylene exposure (and possibly tetrachloroethylene) is small
(acute exposure to a large dose; Briining et al., 1998) and within the amount that would be readily
metabolized by GSTZ (Barton et al., 1999). The EPA is currently reevaluating these chemicals as
A-3
-------
part of the IRIS file, the metabolism of these compounds to DCA will be discussed in those
documents when completed.
Question 3. Are there any problems with the quantification of the DeAngelo et al. (1999)
data and the use of that data to quantify the carcinogenicity of DCA?
Comment: One reviewer noted that the DeAngelo et al. (1999) data are probably the most
appropriate as a basis for modeling cancer risk, but the Pereira et al. (1996) data should have been
considered because it used female mice whereas the DeAngelo et al. (1999) only used male mice.
The reviewer indicated that although other studies in male mice have fewer doses, it would still
be useful to point out that there is general agreement across many studies identifying the amount
of DCA that is required to consistently induce cancer in male mice. Therefore, the reviewer feels
that the DeAngelo study is a surrogate for a larger data base.
Response: The DeAngelo et al. (1999) study was chosen because it represented a longer duration
and because a comparison between the two studies of tumor multiplicity and the development of
carcinomas indicate that female mice are less susceptible to the carcinogenic effects of DCA than
are the males (e.g., they do not develop as many tumors or as many carcinomas as males at
comparable doses).
Comment: One reviewer noted that the wrong data sets were used in the benchmark dose
modeling effort for the cancer endpoint, and the correct data sets did not appear in the
Toxicological Review for either the rat or mouse experiment. The reviewer stated that the main
data set for each species would be the number of animals with tumors: (1) at the site of interest,
(2) of the same embryonic origin, and (3) that are malignant or have the potential to become
malignant. In this particular case, any animal with a hepatocellular adenoma or carcinoma would
be counted.
Response: An error in the draft report was corrected and the dose-response was modeled on the
number of animals with hepatocellular adenomas or carcinomas.
Question 4. The toxicological papers have been arranged in reverse chronological order
within specific sections. Does this cause any problem to you as a reader of the document?
Comment: One reviewer indicated that the format did not make it easy to discern how
A-4
-------
information has developed on the toxicodynamics of DCA's effects and that the document
occasionally indicate that a result was confirmed in the first paper put out on the subject.
Response: EPA has changed the order of the articles in the section to reflect groupings by effect
and by animal model. The change is intended to help facilitate comments that compare and
contrast results from similar studies, and better describe changes in thinking regarding mode of
action. The format of presentation has also been altered so that the discussions are more
integrated than they were with the study-by-study approach.
Comment: One reviewer preferred the reverse chronological order in which the toxicity studies
were reviewed since it allowed the reader to consider the most recent citations first. In this
reviewer's opinion, more recent studies tend to be of higher quality, particularly from the
standpoint of more standardized and complete protocols, and should be the primary focus.
Response: Although EPA has revised the format for reporting the studies, an attempt has been
made to insure that the data are presented in a cohesive, reader-friendly order.
Comment: One reviewer found the format to be distracting of the study-by-study descriptions
under Sections 4.1. to 4.4.
Response: It is difficult to synthesize the results of so many studies without losing details
regarding methodology. In EPA's revisions to the document, the details of the individual studies
have been maintained even though the format has been altered.
Question 5. Do you have any technical disagreement with the information presented?
Comment: One reviewer thought that the external review draft did not adequately address the
modes of action and that some hypotheses were maintained in the document that were no longer
supported in the scientific literature. Further discussion of alternate explanations, such as
distribution toward small lesions, potentially due to suppression of apoptosis, should be added.
Response: EPA has made extensive revision to the External Review Draft of the DC A
Toxicological Profile. Some of these revisions were initiated as a response to the reviewer's
comments. The discussions of the mechanistic and genotoxicity studies have been expanded and
new data regarding precursor lesions found in the liver have been added to the Toxicological
A-5
-------
Review.
Comment: One reviewer indicated that some genotoxicity sections implied a greater degree of
uncertainty than was warranted regarding whether DCA should be considered genotoxic. This
reviewer found the genotoxicity data on DCA neither convincing nor supportive of a weight-of-
evidence based conclusion that DCA is mutagenic. The reviewer believed that the mode of action
for DCA should still be considered uncertain, despite the broad ranging investigations.
Response: EPA has rewritten the genotoxicity section to reflect the preponderance of evidence
that DCA is nongenotoxic, except at high doses, in the models assessed.
Comment: One reviewer objected to the term "prevalence" used in the document and
recommended text changes to replace the term.
Response: The recommended changes have been made.
Comment: One reviewer noted that the ancillary statistics on the occurrence of adenomas and
carcinomas in the same animal could be removed from the tables and strongly suggested that the
next revision of the document be reviewed by experts at the National Center for Environmental
Assessment (NCEA).
Response: The data have been removed. The document has been reviewed by NCEA and other
individuals familiar with benchmark dose methodology. No specific critical comments regarding
the approach were raised.
Comment: One reviewer questioned a statement in the original draft regarding the absence of
reductive dechlorination in humans; the statement was made in the draft based on the apparent
lack of evidence of thiodiacetic acid or monochloroacetic acid excretion in humans. The reviewer
indicated that the basis for the statement was a set of clinical observations on adults with diabetes
by a research group out of the University of Florida at Gainesville. Further, the reviewer felt that
the reader should be aware that such conclusions were based on clinical observations in adults in
a specialized population exposed to a variety of other medications. The issue of heterogeneity
and the potential for a subpopulation for which this pathway might be important should be
considered.
A-6
-------
Response: The entire pharmacokinetic section has been rewritten and there is no longer mention
of reductive chlorination not being a relevant metabolic pathway in humans. There are human
data indicating that there was excretion of monochloroacetic acid in one human subject.
Accordingly, the original text was incorrect.
Comment: One reviewer indicated that the original draft review document gave the reader the
impression that multiple doses of DC A were metabolized similarly to one dose. The reviewer
indicated that studies currently suggest that humans may not be able to metabolize multiple doses
of DC A readily and suggested that the report should not focus so much on single-dose studies
when comparing differences between species in the ability to metabolize DCA.
Response: The entire pharmacokinetic section has been rewritten. The document now includes
considerable discussion on the results of high doses and pretreatment on DCA metabolism
including discussion of half-life.
Comment: With regard to the discussion concerning the change in the elimination half-life with
multiple dosing, one reviewer indicated that the 1991 paper by Curry should be discussed in the
paragraph on page 5, lines 24-34.
Response: The comment has been addressed in the revisions to the toxicokinetic section of the
Toxicological Review.
Comment: One reviewer suggested that the Toxicological Review notes that the human studies
with DCA were performed on individuals who were therapeutically treated with DCA, and thus
were an unhealthy population. No analytic epidemiological studies have been performed to date,
nor have any of the studies had sufficient power to detect carcinogenicity. Clear statements
notifying the reader to that effect should be added.
Response: The recommended changes have been made.
Comments: One reviewer provided the following comments regarding the selection criteria for
the application of benchmark dose modeling to noncancer data sets: (1) a reference for the criteria
provided on page 57 would be useful for the reader and (2) the exclusion of effects which are
reversible can be questioned, especially in cases where the environmental exposures may last a
lifetime, and especially in cases where the effects are as severe as observed in the study in
A-7
-------
question [on page 49, at lines 24-25]. The reviewer also indicated that the nature and severity of
the endpoint and the length of exposure should be carefully considered in such an exercise.
Response: The recommended change for point (1) above has been made. EPA has changed the
text regarding the neurotoxicity effects in rats as suggested by point (2) and removed the "slow
reversibility" comment. It is maintained that the study should not be used for the development of
a RfD since it did not identify a NOAEL and the lowest LOAEL was higher than that identified
for testicular effects in the Cicmanec et al. (1991) study.
Comments: One reviewer provided the following comments with regard to the presentation of
data on genotoxicity:
- Table 4-2 indicated all mutagenicity studies by Herbert et al. (1980) are negative or
equivocal. Yet, in this reviewer's opinion, the authors concluded that "Dichloroacetate
demonstrates low grade mutagenicity in the Ames Salmonella/mammalian microsome
mutagenicity test."
Table 4-2 indicated the findings from DeMarini et al. (1994) in the microscreen prophage-
induction assay as +/-, but the published study clearly states that DCA was genotoxic in
that assay.
- Table 4-3 did not present multiple positive results from the Fuscoe et al. (1996) study,
excluding a small but statistically significant increase in the frequency of micronucleated
normochromatic erythrocytes after 10 weeks of exposure, and a positive finding by
alkaline single cell gel electrophoresis indicating cross-linking in addition to the
frequency of micronucleated polychromatic erythrocytes. The reviewer suggested that the
text indicate that these authors also coadministered vitamin E and found it did not affect
DNA damage induction by dichloroacetic acid.
- The Ono et al. (1991) and Waskell et al. (1978) studies on DNA repair should be included
in Table 4-2.
It should be noted in the Toxicological Review that Harrington-Brock et al. (1998) found
the potency of DCA similar to the classic mutagen ethylmethanesulfonate.
The genotoxicity tables should indicate that Chang et al. (1992) found that 5 and 10
mmole/kg DCA produced a small amount of strand breakage in mice (7% at 4 hr). Strand
breaks from continuous exposure to DCA were also elevated slightly in the mouse. Table
4-3 should indicate that the splenocytes and epithelial cells in the Chang et al. (1992)
study were derived from the stomach and duodenum.
The reviewer suggested an interpretation of the transgenic mice data indicating that a
A-8
-------
direct genotoxic effect would be time independent and that since the findings were not
observed at 4 and 10 weeks but were at 60 weeks, the mutational events might be
secondary to toxicological changes in the liver. The reviewer said this interpretation did
not take into account the greater DCA exposure resulting from the prolonged exposure.
- The original draft indicated that DCA has been consistently negative in all standard DNA
cross-linking studies conducted and provided a secondary reference as justification. The
reviewer indicated that the primary references should be cited.
The analysis of data regarding mutagenicity involving study counts of numbers of
positives and negatives was superficial (page 30). The reviewer indicated that a more
analytic presentation, providing the context for the results in terms of endpoints examined,
and power of studies to detect weak effects, was needed.
Response: Changes to the text on genotoxicity have been made to present a more accurate
presentation of positive and negative results, as well as to compare and contrast findings from
similar studies.
Question 6. Do you agree with the selection of the critical study and effect or the assignment
of uncertainty factors for determination of the RfD?
Comment: One reviewer agreed with the choice of the Cicmanec et al. (1991) study as the
principal study and the critical endpoint but was uncomfortable with the justification for not
modeling the non-monotonic data using benchmark dose (BMD) response. The issue is where to
select the point of departure in such a case, and this should be stated in the document. In this
reviewer's opinion, a better argument for not performing a BMD analysis is that the endpoint is
trivial and not interpretable in a health sense (e.g., weight gain at low doses of DCA in DeAngelo
et al., [1996]). The reviewer indicated that RfD values developed by a BMD analysis and the
NOAEL/LOAEL approach would be comparable because different uncertainty factors could be
used.
Response: EPA has modified the text to reflect the recommended comment regarding point of
departure for non-monotonic data modeled using BMD techniques. EPA has not added the
changes regarding the comparability of RfDs developed by a BMD analysis and more traditional
NOAEL/LOAEL approach because the changes would complicate this section of the document.
Revisions to the discussions on the derivation of the total uncertainty factor have been made,
however.
A-9
-------
Comment: One reviewer indicated that is was appropriate to use the Cicmanec et al. study (1991)
as the critical study and testicular degeneration as the critical effect for calculating the RfD.
Further, the basis for using a NOAEL/LOAEL approach rather than the BMD was well-
documented and explained. The reviewer believed the factor of 10 for the interspecies
extrapolation could be reduced based on an assessment of the metabolic differences between the
dog and humans, indicating that the dog may be more sensitive than humans, rather than the
opposite. The reviewer also indicated that it appears to be appropriate to use the DeAngelo et al.
(1999) data for determining the dose-response for the cancer assessment.
Response: The uncertainty factor for the extrapolation from human to animals based on the
metabolic data on the effects of preexposure on metabolism in humans and animals has been
reduced. The data on similarities between humans and animals with regard to DCA metabolism
justify a decrease in the interspecies uncertainty factor to 3 rather than 10. However, on close
examination of the metabolic data in dogs, EPA now feels they are not particularly strong. EPA
added a 3-fold factor for data base uncertainty because of the lack of a multigeneration study of
DCA. Accordingly, the total uncertainty factor remains at 3,000.
Comment: One reviewer thought the derivation of the noncancer RfD was clearly laid out and
specific calculations were well justified. Further, the reasoning leading to the exclusion of the
benchmark dose calculations from selection of the RfD was also appropriate. The reviewer had
reservations regarding the selection of uncertainty factors and suggested that the potential exists
for wide human variability (e.g., Curry et al., [1991] and the GSTZ polymorphism and other
factors noted in the Toxicological Review). Therefore, these findings, which are based on limited
observations, increase the concern (the small number of individuals studied and the limited types
of observations made biases the investigation toward a false negative). Further, the LOAEL was
identified at a high response - 80% of the dogs in the critical study were found with testicular
degeneration and suggested that a factor of 10 may not be sufficient to predict from an 80%
incidence what might be observed as a NOAEL in a study of reasonable size (e.g., with 50
animals per dose group).
Response: The 10-fold uncertainty factor for intraspecies variability has been retained. The
discussion on why this factor was selected and the discussion on sensitive populations have been
expanded to provide greater support for this decision. The UF of 10 for the use of a LOAEL was
also retained, however a threefold uncertainty factor was added for data base uncertainty due to
the lack of a multigeneration study of reproductive toxicity which would include consideration of
A-10
-------
the functional impact of the testicular effects on reproduction.
Question 7. Do you have any suggestions for improving the presentation of information that
are compatible with IRIS SOPs?
Comment: One reviewer believed that the mode of action section did not include enough critical
examination of conflicting data in available studies. Further, additional efforts should be included
regarding how the mode of action data have been developed for DCA.
Response: The text has been modified to improve the handling of mode of action data.
Comment: One reviewer suggested the following changes: (1) reformat the tumor incidence
tables; (2) include a schematic of hypothesized metabolism of DCA; (3) provide introductory
paragraphs before each major section orienting the reader to the nature of the studies described
and providing context; and (4) change the section regarding human studies to give the reader
better understanding that the studies were not designed to provide evidence of carcinogenicity.
Further, a very brief synopsis of the drinking water studies showing an association between
drinking water and cancer could also be added to the text, with the necessary caveats indicating
the limitations of these studies with regard to establishing a causal relationship between DCA and
human cancer.
Response: The recommended changes have been made with the following two exceptions: the
cancer studies involving drinking water exposure have not been added and introductory
paragraphs to every section in the hazard identification section have not been added due to space
limitations. However, the format of each section has been revised to allow for greater integration
of the data and to provide opportunities, in some cases, for cross-study comparisons.
A-ll
-------
APPENDIX B
SUMMARY OF STUDIES ON DCA TOXICITY AND
APPLICABILITY FOR BMD ANALYSIS
-------
APPENDIX B
Summary of Studies on DCA Toxicity and Applicability for BMD Analysis
DATA SUM M ARY
E nd p o in t
M etab ol.
H ep atic
N eu r o.
R ep r o.
Species
Human
Rat
Dog
Human
Mouse
Rat
Dog
Human
Rat
Dog
Rat
Dog
R ef er ence
Stacpoole et al., 1978
M oore et al., 1979
Evans and Stacpoole, 1982
Davis, 1990
Davis, 1986
Katz et al., 1981
Yount et al., 1982
Ribes et al., 1979
Katz et al., 1981
Stacpoole et al., 1998
Sanchez and Bull, 1990
Bull et al., 1990
DeAngelo et al., 1991
Daniel et al., 1992
M ather et al., 1990
DeAngelo et al., 1996
Katz et al., 1981
Smith et al., 1992
Toth et al., 1992
Yount et al., 1982
Cicmanec et al., 1991
Cicmanec et al., 1991
M oore et al., 1979
Spruijt et al., 2001
Stacpoole et al., 1998
Stacpoole et al., 1978
Yount et al., 1982
Katz et al., 1981
Cicmanec et al., 1 991
Katz et al., 1981
Smith et al., 1992
Linder et al., 1997
Toth et al., 1992
DeAngelo et al., 1996
Yount et al., 1982
Katz et al., 1981
Katz et al., 1981
Cicmanec et al., 1 991
D u ratio n
5-16 wks
7 days
1 day
2 weeks
3 m onths
3 m onths
7 days
13 weeks
? m onths
2 weeks
2 m onths
1 year
2 years
3 m onths
100 weeks
3 m onths
gestatio n
10 w eeks
3 m onths
3 m onths
3 m onths
1 year
<5 years
6-7 days
3 m onths
3 m onths
3 m onths
13 weeks
2 weeks
1 0 w eeks
100 weeks
3 m onths
13 weeks
1 3 w eeks
3 m onths
G rps
1
1
2
3
5
4
2
4
1
4
7
3
3
2
4
3
4
4
4
2
4
4
1
1
1
1
4
4
4
4
5
4
3
4
4
4
NOAEL
120
150
57
7.6
3.9
40.2
14
14
3.6
125
LO AEL
43
50
100
125
323
150
50
25
190
100
140
77
88
35.5
125
140
31
323
12.5
12.5
50
50
25
43
323
125
12.5
50
140
54
31
40.2
323
500
50
12.5
Effect
Decreased serum glucose, lactate
Decreased cholesterol
Decreased serum lactate
Decreased serum lactate and glucose
Altered serum metabolites
Decreased serum metabolites
Decreased serum metabolites
Increased serum enzymes
Increased glycogen, focal necrosis
Increased glycogen
Increased glycogen
Increased relative liver weight
Increased liver weight, necrosis
Increased liver weight, glycogen
Increased liver weight
Increased liver weight
H ep atom eg aly
Increased liver weight
Liver histopathology
Tingling, slowed NCV, poor reflexes
Peripheral neuropathy, slowed NCV
Sedation, peripheral neuropathy
Sedation, peripheral neuropathy
Slowed NCV, hindlimb weakness
Histological brain lesions
Vacuolar changes in brain
Vacuolar changes in brain
Decreased fetal wt, increased resorptions
Multiple effects on sperm formation
Decreased epididymal weight
Increased testicular weight
Testicular degeneration
Testicular degeneration
Prostate atrophy, testicular changes
Incidence of testicular lesions
SELECTION CRITERIA
Qua ntita tive
>2
groups?
no
no
no
yes
yes
yes
no
no
yes
no
yes
yes
yes
yes
no
yes
yes
yes
yes
yes
no
yes
yes
no
no
no
no
no
yes
yes
yes
yes
yes
yes
yes
no
yes
yes
yes
L
<10*LL?
yes
yes
yes
yes
yes
no
no
no
yes
yes
no
yes
no
yes
yes
yes
yes
no
no
yes
no
yes
yes
yes
yes
yes
yes
no
no
yes
yes
no
yes
yes
yes
no
no
yes
yes
N
-------
APPENDIX C
BENCHMARK DOSE-RESPONSE FOR
NONCANCER ENDPOINTS
-------
Noncancer BMD Modeling Results
Cicmanec et al., 1991
Quantal Quadratic Model $Revision: 2.1 $ $Date: 2000/02/26 03:38:55 $
Input Data File: H:\OW\DCA\FINALBMD\NONCANCE\CICMANEC\CICMANEC.(d)
Gnuplot Plotting File: H:\OW\DCA\FINALBMD\NONCANCE\CICMANEC\CICMANEC.plt
Wed Feb 21 14:33:51 2001
HMDS MODEL RUN
The form of the probability function is:
P [response] = background + (1-background)* [1-EXP(-slope*dose*2)]
Dependent variable = Affected
Independent variable = Dose
Total number of observations = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial (and Specified) Parameter Values
Background = 0.0833333
Slope = 0.000462557
Power = 2 Specified
Asymptotic Correlation Matrix of Parameter Estimates
( *** xhe model parameter(s) -Background -Power
have been estimated at a boundary point, or have been specified by the user,
and do not appear in the correlation matrix )
Slope
Slope
1
Variable
Background
Slope
Parameter Estimates
Estimate
0
0.0103004
Std. Err.
NA
0.00572423
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
Deviance Test DF
Log(likelihood)
-2 .50201
-2.50201
-12.2173
1. 04799e-006
19.4305
P-value
0.0002227
AIC: 7.00403
Goodness of Fit
Dose
0.0000
12 . 5000
39. 5000
72.0000
Est. Prob.
0.0000
0 .8000
1 .0000
1.0000
Expected
0.000
4 .000
5 .000
5.000
Observed
0
4
5
5
Size
5
5
5
5
Scaled
Residual
0
-4 . 68e-006
0. 0007239
0
Chi-square =0.00 DF = 3
Benchmark Dose Computation
P-value = 1.0000
C-l
-------
Specified effect =
Risk Type =
Confidence level =
BMD =
BMDL =
0 .1
Extra risk
0.95
3 .19824
2.09228
Linder et al., 1997 (A2 Run)
Hill Model. $Revision: 2.1 $ $Date: 2000/10/11 21:21:23 $
Input Data File: H:\OW\DCA\FINALBMD\NONCANCE\LINDER97\COUNT\LIND97A2.(d)
Gnuplot Plotting File: H:\OW\DCA\FINALBMD\NONCANCE\LINDER97\COUNT\LIND97A2.plt
Wed Feb 21 15:59:00 2001
BMDS MODEL RUN
The form of the response function is:
Y [dose] = intercept + v*dose*n/(k*n + dose^n)
Dependent variable = MEAN
Independent variable = Dose
Power parameter restricted to be greater than 1
The variance is to be modeled as Var(i) = alpha * mean(i)
rho
Total number of dose groups = 6
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
alpha = 0.639066
rho = 1.35894
intercept = 224
v = -138
n = 1.55811
k = 214.237
Asymptotic Correlation Matrix of Parameter Estimates
alpha
rho
intercept
V
n
k
alpha
1
-1
-0. 032
0.12
0.24
-0.1
rho
-1
1
0. 028
-0.12
-0.24
0. 1
intercept
-0.032
0. 028
1
-0.72
-0.46
-0 .39
V
0.12
-0. 12
-0. 72
1
0.79
-0.2
n
0.24
-0.24
-0.46
0.79
1
-0.2
k
-0.1
0. 1
-0 .39
-0.2
-0.2
1
Variable
alpha
rho
intercept
v
n
k
Parameter Estimates
Estimate Std. Err.
2.94166 8.45724
1.09046 0.564024
236.288 9.20067
-158.548 16.23
1.47093 0.377618
179.513 37.1316
Table of Data and Estimated Values of Interest
Dose N Obs Mean Obs Std Dev Est Mean Est Std Dev Chi*2 Res.
0 8 224 40 236 33.8 -0.364
C-2
-------
18
54
160
480
1440
248
208
165
106
86
32
25
21
35
11
231
213
164
108
84 .8
33 .4
31. 9
27.6
22
19.3
0 .507
-0 .161
0.0468
-0.0883
0. 0609
Model Descriptions for likelihoods calculated
Model Al: Yij = Mu(i) + e(ij)
Var{e(ij)
Model A2:
Yij = Mu(i) + e(ij!
Var{e(ij)} = Sigma(i)*2
Model A3: Yij = Mu(i) + e(ij)
Var{e(ij)} = alpha*(Mu(i))
Model R:
Var{e(i)
Yi = Mu + e(i)
Likelihoods of Interest
Model Log (likelihood) DF
Al -182.377837 7
A2 -175.765302 12
A3 -181.171992 8
fitted -182.888696 6
R -225.464414 2
Explanation of Tests
AIC
378.755674
375.530604
378.343984
377.777393
454.928827
Test 1: Does response and/or variances differ among Dose levels? (A2 vs. R)
Test 2: Are Variances Homogeneous? (Al vs A2)
Test 3: Are variances adequately modeled? (A2 vs. A3)
Test 4: Does the Model for the Mean Fit? (A3 vs. fitted)
Test
Test 1
Test 2
Test 3
Test 4
Tests of Interest
-2*log(Likelihood Ratio) Test df
99 .3982
13 .2251
10.8134
3.43341
10
5
4
2
p-value
<.0001
0.02136
0.02874
0.1797
The p-value for Test 1 is less than .05. There appears to be a
difference between response and/or variances among the dose levels
It seems appropriate to model the data
The p-value for Test 2 is less than .05.
model appears to be appropriate.
The p-value for Test 3 is less than .05.
different variance model.
A non-homogeneous variance
You may want to consider a
The p-value for Test 4 is greater than .05.
to adequately describe the data.
Benchmark Dose Computation
Specified effect = 1
The model chosen seems
Risk Type =
Confidence level =
BMD =
Estimated standard deviations from the control mean
0.95
73.8032
Warning: optimum may not have been found. Bad completion code in Optimization routine.
BMDL computation failed.
C-3
-------
Linder et al., 1997 (B Run)
Hill Model. $Revision: 2.1 $ $Date: 2000/10/11 21:21:23 $
Input Data File: H:\OW\DCA\FINALBMD\NONCANCE\LINDER97\MOTILITY\LINDE97B.(d)
Gnuplot Plotting File: H:\OW\DCA\FINALBMD\NONCANCE\LINDER97\MOTILITY\LINDE97B.plt
Wed Feb 21 16:05:43 2001
BMDS MODEL RUN
The form of the response function is:
Y [dose] = intercept + v*dose*n/ (k*n + dose^n)
Dependent variable = MEAN
Independent variable = Dose
rho is set to 0
Power parameter restricted to be greater than 1
A constant variance model is fit
Total number of dose groups = 6
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
alpha
rho
intercept
v
n
k
110.154
0 Specified
72
-66
1.05789
190.476
Asymptotic Correlation Matrix of Parameter Estimates
alpha
rho
intercept
v
n
k
alpha
1
0
0
0
0
0
rho
0
1
0
0
0
0
intercept
0
0
1
0
0
0
n
0
0
0
0
1
0
k
0
0
0
0
0
1
Parameter Estimates
Variable
alpha
rho
intercept
v
n
k
Estimate
124 . 88
0
74 .658
-67. 6049
1 .86631
177.165
Std.
Err.
1
1
1
1
1
1
Table of Data and Estimated Values of Interest
Dose
0
18
54
160
480
1440
Obs Mean
Obs Std Dev
Est Mean
Est Std Dev Chi*2 Res.
72
74
72
41
20
6
12
8
9
13
17
7
74 . 7
73 . 7
68
44 .1
16 .2
8 .38
11 .2
11 .2
11.2
11.2
11.2
11.2
-0.238
0 .0247
0.356
-0.274
0 .344
-0 .213
Model Descriptions for likelihoods calculated
Model Al: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma*2
Model A2: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma(i)*2
Model R:
Yi = Mu
C-4
-------
Var{e(i)}
Likelihoods of Interest
Model Log(likelihood) DF AIC
Al -138.103400 7 290.206800
A2 -133.640412 12 291.280823
fitted -139.856491 5 289.712983
R -186.484888 2 376.969776
Test 1: Does response and/or variances differ among dose levels
(A2 vs. R)
Test 2: Are Variances Homogeneous (Al vs A2)
Test 3: Does the Model for the Mean Fit (Al vs. fitted)
Tests of Interest
Test -2*log(Likelihood Ratio) Test df p-value
Test 1 96.763 10 <.0001
Test 2 8.92598 5 0.1121
Test 3 3.50618 2 0.1732
The p-value for Test 1 is less than .05. There appears to be a
difference between response and/or variances among the dose levels. It seems appropriate to
model the data.
The p-value for Test 2 is greater than .05. A homogeneous variance model appears to be
appropriate here.
The p-value for Test 3 is greater than .05. The model chosen appears
to adequately describe the data
Benchmark Dose Computation
Specified effect = 1
Risk Type = Estimated standard deviations from the control mean
Confidence level = 0.95
BMD = 74 .3974
BMDL = 46 .8201
Mather et al., 1990
Hill Model. $Revision: 2.1 $ $Date: 2000/10/11 21:21:23 $
Input Data File: H:\OW\DCA\FINALBMD\NONCANCE\MATHER\MATHER4.(d)
Gnuplot Plotting File: H:\OW\DCA\FINALBMD\NONCANCE\MATHER\MATHER4.plt
Wed Feb 21 16:28:03 2001
BMDS MODEL RUN
The form of the response function is:
Y [dose] = intercept + v*dose*n/(k*n + dose^n)
Dependent variable = MEAN
Independent variable = Dose
rho is set to 0
Power parameter restricted to be greater than 1
A constant variance model is fit
Total number of dose groups = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
alpha = 0.00948683
rho = 0 Specified
intercept = 3.8
C-5
-------
V =
n =
k =
2 .62
0 .318455
584.908
user,
Asymptotic Correlation Matrix of Parameter Estimates
( *** xhe model parameter(s) -n
have been estimated at a boundary point, or have been specified by the
and do not appear in the correlation matrix )
alpha rho intercept v
alpha 1000
rho 0100
intercept
v
k
Variable
alpha
rho
intercept
v
n
k
0 0
0 0
0 0
Parameter Estimates
Estimate
0 .0116056
0
3 .86856
3 .26005
1
96.1449
1 0
0 1
0 0
Std. Err.
1
1
1
1
NA
1
k
0
0
0
0
1
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Table of Data and Estimated Values of Interest
Dose
Obs Mean
Obs Std Dev
Est Mean
Est Std Dev Chi*2 Res.
0
3 .9
35.5
345
10
10
10
10
3 . 8
4 .08
4 .73
6 .42
0 .1
0 .1
0.09
0.1
3 .87
4
4 .75
6 .42
0 .108
0 .108
0.108
0.108
-0. 636
0. 783
-0.164
0.0176
Model Descriptions for likelihoods calculated
+ e(ij)
2
+ e (i j )
(i)*2
e(i)
Model Al : Yij = Mu(i)
Var{e(ij)} = Sigma
Model A2 : Yi j = Mu (i )
Var{e(ij)J
Model R: Yi = Mu
Var{e (i) }
Degrees of freedom for Test Al vs fitted <= 0
Likelihoods of Interest
Model Log (likelihood) DF AIC
Al 75.183917 5 -140.367835
A2 75.264219 8 -134.528438
fitted 69.125316 5 -128.250632
R -21.361630 2 46.723261
Test 1:
Test 2 :
Test 3 :
Does response and/or variances differ among dose levels
(A2 vs. R)
Are Variances Homogeneous (Al vs A2)
Does the Model for the Mean Fit (Al vs. fitted)
Tests of Interest
Test -2*log(Likelihood Ratio) Test df p-value
Test 1 193.091 6 <.0001
Test 2 0.160603 3 0.9837
Test 3 12.1172 0 NA
The p-value for Test 1 is less than .05. There appears to be a
difference between response and/or variances among the dose levels.
It seems appropriate to model the data
C-6
-------
The p-value for Test 2 is greater than .05. A homogeneous variance
model appears to be appropriate here
NA - Degrees of freedom for Test 3 are less than or equal to 0. The Chi-Square
test for fit is not valid
Benchmark Dose Computation
Specified effect = 1
Risk Type
Confidence level =
BMD =
BMDL =
Estimated standard deviations from the control mean
0.95
3.28571
3.00326
Toth et al., 1992 (A Run)
Linear Model. $Revision: 2.1 $ $Date: 2000/10/11 17:51:39 $
Input Data File: H:\OW\DCA\FINALBMD\NONCANCE\TOTH92\COUNT\TOTH92A.(d)
Gnuplot Plotting File: H:\OW\DCA\FINALBMD\NONCANCE\TOTH92\COUNT\TOTH92A.plt
Wed Feb 21 15:02:38 2001
BMDS MODEL RUN
The form of the response function is:
Y [dose] = beta_0 + beta_l*dose + beta_2*dose*2 + ...
Dependent variable = MEAN
Independent variable = Dose
Signs of the polynomial coefficients are not restricted
The variance is to be modeled as Var(i) = alpha*mean(i)*rho
Total number of dose groups = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
alpha = 1
rho = 0
beta_0 = 638.2
beta 1 = -2.14309
Variable
alpha
rho
beta_0
beta 1
Parameter Estimates
Estimate Std. Err.
0.00134237 0.0067082
2.64948 0.801147
639.904 30.5081
-2.17205 0.320949
Asymptotic Correlation Matrix of Parameter Estimates
alpha
rho
beta_0
beta 1
alpha
1
-1
-0.015
0.017
rho
-1
1
0. 015
-0.017
beta_0
-0.015
0.015
1
-0.87
Table of Data and Estimated Values of Interest
Dose
Res .
0
31.25
62 .5
125
Obs Mean
Obs Std Dev
Est Mean
beta_l
0.017
-0.017
-0.87
1
Est Std Dev
19
18
18
19
630
583
503
368
205
137
164
91 .6
640
572
504
368
191
165
139
92
-0. 985
1.2
-0.149
-0 .0821
C-7
-------
Model Descriptions for likelihoods calculated
Model Al : Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma*2
Model A2 : Yi j = Mu (i ) + e (i j )
Var{e(ij)} = Sigma (i)*2
Model A3: Yij = Mu(i) + e(ij)
Var{e(ij)} = alpha* (Mu (i)
Model R: Yi = Mu + e (i)
Var{e (i) }
Likelihoods of Interest
Model Log (likelihood) DF AIC
Al -408.202121 5 826.404241
A2 -402.269507 8 820.539015
A3 -403.341780 6 818.683559
fitted -403.457293 4 814.914586
R -422.268100 2 848.536200
Explanation of Tests
Test 1: Does response and/or variances differ among Dose
levels? (A2 vs. R)
Test 2: Are Variances Homogeneous? (Al vs A2)
Test 3: Are variances adequately modeled? (A2 vs. A3)
Test 4: Does the Model for the Mean Fit? (A3 vs. fitted)
Tests of Interest
Test -2*log (Likelihood Ratio) Test df p-value
Test 1 39.9972 6 <.0001
Test 2 11.8652 3 0.007859
Test 3 2.14454 2 0.3422
Test 4 0.231027 2 0.8909
The p-value for Test 1 is less than .05. There appears to be a difference between response
and/or variances among the dose levels It seems appropriate to model the data.
The p-value for Test 2 is less than .05. A non- homogeneous variance
model appears to be appropriate .
The p-value for Test 3 is greater than .05. The modeled variance appears to be appropriate
here .
The p-value for Test 4 is greater than .05. The model chosen seems to adequately describe
the data.
Benchmark Dose Computation
Specified effect = 450
Risk Type = Point risk
Confidence level = 0.95
BMD = 87.4306
BMDL = 75.7201
Toth et al., 1992 (C Run)
Hill Model. $Revision: 2.1 $ $Date: 2000/10/11 21:21:23 $
Input Data File: H:\OW\DCA\FINALBMD\NONCANCE\TOTH92\MOTILITY\TOTH92C.(d)
Gnuplot Plotting File: H:\OW\DCA\FINALBMD\NONCANCE\TOTH92\MOTILITY\TOTH92C.plt
Wed Feb 21 15:30:34 2001
BMDS MODEL RUN
The form of the response function is:
Y [dose] = intercept + v*dose*n/(k*n + dose^n)
-------
Dependent variable = MEAN
Independent variable = Dose
rho is set to 0
Power parameter restricted to be greater than 1
A constant variance model is fit
Total number of dose groups = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
alpha
rho
intercept
v
n
k
115.908
0
54.6
-27. 5
5. 16921
59.6104
Specified
Asymptotic Correlation Matrix of Parameter Estimates
alpha
rho
intercept
v
n
k
alpha
1
0
0
0
0
0
rho
0
1
0
0
0
0
intercept
0
0
1
0
0
0
Variable
alpha
rho
intercept
v
n
k
Parameter Estimates
Estimate
109.538
0
54 .6
-27.8588
6.017
60.7743
Std. Err.
1
1
1
1
1
1
n
0
0
0
0
1
0
k
0
0
0
0
0
1
Table of Data and Estimated Values of Interest
Dose N Obs Mean Obs Std Dev Est Mean
Est Std Dev Chi*2 Res.
0 15
31.25 14
62.5 17
125 19
54.6
54 . 1
39. 5
27.1
10.2
11 .2
12
9.8
54.6
54 . 1
39. 5
27.1
10.5
10 .5
10 .5
10.5
3.54e-007
3 . 56e-009
7.21e-008
-2.48e-007
Model Descriptions for likelihoods calculated
Model Al : Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma*2
Model A2 : Yi j = Mu (i ) + e (i j )
Var{e(ij)} = Sigma (i)*2
Model R: Yi = Mu + e(i)
Var{e(i) }
Degrees of freedom for Test Al vs fitted <= 0
Likelihoods of Interest
Model Log (likelihood) DF
Al -185.128961 5
A2 -184.702137 8
fitted -185.128962 5
R -211.659863 2
AIC
380.257923
385.404274
380 .257923
427.319726
Test 1: Does response and/or variances differ among dose levels
(A2 vs. R)
Test 2: Are Variances Homogeneous (Al vs A2)
Test 3: Does the Model for the Mean Fit (Al vs. fitted)
Tests of Interest
Test -2*log(Likelihood Ratio) Test df p-value
C-9
-------
Test 1 53.0618 6 <.0001
Test 2 0.853649 3 0.8366
Test 3 5.57208e-007 0 NA
The p-value for Test 1 is less than .05. There appears to be a
difference between response and/or variances among the dose levels.
It seems appropriate to model the data
The p-value for Test 2 is greater than .05. A homogeneous variance
model appears to be appropriate here.
NA - Degrees of freedom for Test 3 are less than or equal to 0. The Chi-Square
test for fit is not valid.
Benchmark Dose Computation
Specified effect = 1
Risk Type = Estimated standard deviations from the control mean
Confidence level = 0.95
BMD = 55.8547
BMDL = 40.3906
C-10
-------
APPENDIX D
BENCHMARK DOSE-RESPONSE FOR CANCER ENDPOINTS
DeAngelo et al., 1999 (5 Doses)
-------
Cancer BMD Modeling Results-5 Doses
$Revision: 2.2 $ $Date: 2001/03/14 01:17:00 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY DOCUMENTS\DCA\FINAL
MODELING\5DOSEGAMMA.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY DOCUMENTS\DCA\FINAL
MODELING\5DOSEGAMMA.plt
Sat Aug 03 09:50:30 2002
DCA DeAngelo (1999) Five Doses
The form of the probability function is:
P[response]= background+(1-background)*CumGamma[siope*dose,power]
where CumGamma(.) is the cumulative Gamma distribution function
Dependent variable = Either
Independent variable = HED
Power parameter is restricted as power >=1
Total number of observations = 5
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial (and Specified) Parameter Values
Background = 0.362745
Slope = 0.0528134
Power = 1.3
Asymptotic Correlation Matrix of Parameter Estimates
Background Slope Power
Background 1 0.19 0.26
Slope 0.19 1 0.98
Power 0.26 0.98 1
Parameter Estimates
Variable
Background
Slope
Power
Estimate
0.351081
0.211892
4.06938
Std. Err.
0.0526875
0.137073
2.80924
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
Log(likelihood)
-85.1782
-85.4032
-111.914
Deviance Test DF
0.449935
53.471
P-value
0.7985
<.0001
D-l
-------
AIC:
176.806
Goodness of
Dose Est,
0.
1.
13.
26.
47.
0000
3000
2000
5000
5000
0,
0,
0,
0,
0,
Chi-square =
. Prob.
.3511
.3512
.5420
.8713
.9931
0.31
Fit
Expected
17.
11.
13.
30.
20.
DF =
,554
,589
,550
,497
,854
2
Scaled
Observed
18
11
14
30
21
P-value
Size
50
33
25
35
21
= 0.8582
Residual
0.
-0.
0.
-0.
0.
1321
2148
1808
2509
3829
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0.1
Extra risk
0.95
8.45355
2.54711
Logistic Model $Revision: 2.1 $ $Date: 2000/02/26 03:38:20 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY DOCUMENTS\DCA\FINAL
MODELING\5DOSELOGISTIC.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY DOCUMENTS\DCA\FINAL
MODELING\5DOSELOGISTIC.plt
Sat Aug 03 09:55:12 2002
DCA DeAngelo (1999) Five Doses
The form of the probability function is:
P[response] = I/[1+EXP(-intercept-slope*dose)]
Dependent variable = Either
Independent variable = HED
Slope parameter is not restricted
Total number of observations = 5
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
background = 0 Specified
intercept = -0.766205
slope = 0.0938121
Asymptotic Correlation Matrix of Parameter Estimates
D-2
-------
intercept
slope
*** The model parameter(s) -background
have been estimated at a boundary point, or have been specified by
the user, and do not appear in the correlation matrix )
intercept
1
-0.58
slope
-0.58
1
Parameter Estimates
Variable
intercept
slope
Model
Full model
Fitted model
Reduced model
AIC:
Estimate
-0.734827
0.09509
Std. Err.
0.227439
0.0171984
Analysis of Deviance Table
Log(likelihood)
-85.1782
-86.057
-111.914
176.114
Deviance Test DF
1.75755
53.471
P-value
0.6242
<.0001
Goodness of Fit
Dose Es
0.0000
1.3000
13.2000
26.5000
47.5000
Chi-sguare =
Benchmark Dos
Specified effect
Risk Type
Confidence level
BMD
BMDL
t
0
0
0
0
0
e
. Prob. E
.3241
.3518
.6272
.8563
.9777
1.30
Computation
=
Extra
=
3.
2.4
xpect
16.
11.
15.
29.
20.
DF =
0.1
risk
0.95
0997
293
ed
207
609
681
971
532
3
Observed Size
18 50
11 33
14 25
30 35
21 21
P-value = 0.7280
Scaled
Residual
0.5418
-0.2219
-0.6952
0.01387
0.6916
Logistic Model $Revision: 2.1 $ $Date: 2000/02/26 03:38:20 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY DOCUMENTS\DCA\FINAL
MODELING\5DOSELOGLOGISTIC.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY DOCUMENTS\DCA\FINAL
MODELING\5DOSELOGLOGISTIC.plt
Sat Aug 03 09:57:01 2002
DCA DeAngelo (1999) Five Doses
The form of the probability function is:
D-3
-------
P[response] = background+(1-background)/[1+EXP(-intercept-slope*Log(dose))]
Dependent variable = Either
Independent variable = HED
Slope parameter is not restricted
Total number of observations = 5
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
User has chosen the log transformed model
Default Initial Parameter Values
background = 0.36
intercept = -5.15172
slope = 2.06624
Asymptotic Correlation Matrix of Parameter Estimates
slope
0.23
-0.99
1
background
background
intercept
slope
1
-0.27
0.23
intercept
-0.27
1
-0.99
Variable
background
intercept
slope
Model
Full model
Fitted model
Reduced model
Parameter Estimates
Estimate
0.352308
-9.86897
3.46213
Std. Err.
0.0530284
3.76293
1.18873
Analysis of Deviance Table
Log(likelihood)
-85.1782
-85.726
-111.914
Deviance Test DF
P-value
1.0956
53.471
0.5782
<.0001
AIC: 177.452
Goodness of Fit
D
0.
1.
13.
26.
47.
lose
0000
3000
2000
5000
5000
Est,
0,
0,
0,
0,
0,
. Prob.
.3523
.3524
.5348
.8796
.9810
Exped
17,
11,
13,
30,
20,
;ed 01
.615
.629
.369
.786
.600
userved '
18
11
14
30
21
Size
50
33
25
35
21
Scaled
Residu
0.
-0.
0.
-0.
0.
.al
1139
2292
2528
4082
6383
Chi-sguare =0.70
DF = 2
P-value = 0.7034
D-4
-------
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Probit Model $Revision: 2.1 $ $Date: 2000/02/26 03:38:53 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY DOCUMENTS\DCA\FINAL
MODELING\5DOSELOGPROBIT.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY DOCUMENTS\DCA\FINAL
MODELING\5DOSELOGPROBIT.pit
Sat Aug 03 09:58:44 2002
DCA DeAngelo (1999) Five Doses
The form of the probability function is:
P[response] = Background
+ (1-Background) * CumNorm(Intercept+Slope*Log(Dose)) ,
where CumNorm(.) is the cumulative normal distribution function
Dependent variable = Either
Independent variable = HED
Slope parameter is not restricted
Total number of observations = 5
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
User has chosen the log transformed model
Default Initial (and Specified) Parameter Values
background = 0.36
intercept = -2.62547
slope = 1.05552
Asymptotic Correlation Matrix of Parameter Estimates
background
intercept
slope
background
1
-0.27
0.23
intercept
-0.27
1
-0.99
slope
0.23
-0.99
1
Parameter Estimates
Variable
background
intercept
slope
Estimate
0.351713
-5.92321
2.07572
Std. Err.
0.0527934
2.15789
0.677696
D-5
-------
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood)
-85.1782
-85.5499
-111.914
177.1
Deviance Test DF
0.743422
53.471
P-value
0.6896
<.0001
Goodness of
Dose
0.
1.
13.
26.
47.
0000
3000
2000
5000
5000
Est,
0,
0,
0,
0,
0,
. Prob.
.3517
.3517
.5366
.8771
.9881
Fit
Scaled
Expected Observed Size
17,
11.
13,
30,
20,
.586
.607
.415
.697
.751
18
11
14
30
21
50
33
25
35
21
Residual
0.
-0.
0.
-0.
0.
1227
2211
2344
3588
5019
Chi-square = 0.50
DF = 2
P-value = 0.7790
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0.1
Extra risk
0.95
9.35739
4.27367
Multistage Model. $Revision: 2.1 $ $Date: 2000/08/21 03:38:21 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY DOCUMENTS\DCA\FINAL
MODELING\5DOSEMULTISTAGE2.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY DOCUMENTS\DCA\FINAL
MODELING\5DOSEMULTISTAGE2.pit
Sat Aug 03 10:00:45 2002
DCA DeAngelo (1999) Five Doses
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*dose/xl-beta2*dose/x2) ]
The parameter betas are restricted to be positive
Dependent variable = Either
Independent variable = HED
Total number of observations = 5
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
D-6
-------
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
Background = 0
Beta(l) = 0
Beta(2) = 4.46493e+016
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(l)
have been estimated at a boundary point, or have been specified by
the user, and do not appear in the correlation matrix )
Background
Beta(2)
Background
1
-0.37
Beta(2)
-0.37
1
Parameter Estimates
Variable
Background
Beta(l)
Beta(2)
Estimate
0.347888
0
0.00223736
Std. Err.
0.0838624
NA
0.000611002
NA - Indicates that this parameter has hit a bound
implied by some ineguality constraint and thus
has no standard error.
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
Log(likelihood)
-85.1782
-85.3116
-111.914
Deviance Test DF
0.266877 3
53.471 4
P-value
0.9661
<.0001
AIC:
174.623
Dose
Goodness of Fit
Est._Prob. Expected Observed
Size
ChiA2 Res.
i :
i :
i :
i :
1
2
3
1
4
0
1
3
.0000
.3000
.2000
0,
0,
0,
.3479
.3503
.5584
17.
11.
13.
,394
,562
,960
18
11
14
50
33
25
0.
-0.
0.
,053
,075
,006
D-7
-------
26.5000 0.8645 30.257 30 35 -0.063
i: 5
47.5000 0.9958 20.912 21 21 1.004
Chi-square =0.18 DF = 3 P-value = 0.9809
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 6.86232
BMDL = 2.05186
Probit Model $Revision: 2.1 $ $Date: 2000/02/26 03:38:53 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY DOCUMENTS\DCA\FINAL
MODELING\5DOSEPROBIT.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY DOCUMENTS\DCA\FINAL
MODELING\5DOSEPROBIT.pit
Sat Aug 03 10:03:12 2002
DCA DeAngelo (1999) Five Doses
The form of the probability function is:
P[response] = CumNorm(Intercept+Slope*Dose),
where CumNorm(.) is the cumulative normal distribution function
Dependent variable = Either
Independent variable = HED
Slope parameter is not restricted
Total number of observations = 5
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial (and Specified) Parameter Values
background = 0 Specified
intercept = -0.442019
slope = 0.0558392
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -background
have been estimated at a boundary point, or have been specified by
the user, and do not appear in the correlation matrix )
intercept slope
D-8
-------
intercept
slope
Variable
intercept
slope
Model
Full model
Fitted model
Reduced model
AIC:
1 -0.6
-0.6 1
Parameter Estimates
Estimate Std. Err.
-0.449908 0.138144
0.0570853 0.00973486
Analysis of Deviance Table
Log(likelihood)
-85.1782
-85.7683
-111.914
175.537
Deviance Test DF
P-value
1.18017
53.471
0.7578
<.0001
Dose
0.0000
1.3000
13.2000
26.5000
47.5000
hi-square
Est.
0.
0.
0.
0.
0.
= 0.94
Goodness
of
Fit
Prob. Expected Observed Size
3264
3536
6193
8561
9881
DF = 3
16,
11.
15,
29,
20,
.319
.668
.482
.963
.751
P-value
18
11
14
30
21
= 0.8155
50
33
25
35
21
Scaled
Residual
0.5069
-0.2432
-0.6105
0.01799
0.502
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0.1
Extra risk
0.95
3.15928
2.53542
Quantal Linear Model $Revision: 2.2 $ $Date: 2000/03/17 22:27:16 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY DOCUMENTS\DCA\FINAL
MODELING\5DOSEQLINEAR.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY DOCUMENTS\DCA\FINAL
MODELING\5DOSEQLINEAR.plt
Sat Aug 03 10:04:30 2002
DCA DeAngelo (1999) Five Doses
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(-slope*dose)]
Dependent variable = Either
Independent variable = HED
D-9
-------
Total number of observations = 5
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial (and Specified) Parameter Values
Background = 0.362745
Slope = 0.0701811
Power = 1 Specified
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Power
have been estimated at a boundary point, or have been specified by
the user, and do not appear in the correlation matrix )
Background
Slope
Background
1
-0.33
Slope
-0.33
1
Parameter Estimates
Variable
Background
Slope
Model
Full model
Fitted model
Reduced model
AIC:
Estimate
0.32128
0.056121
Std. Err.
0.0525609
0.0112913
Analysis of Deviance Table
Log(likelihood)
-85.1782
-87.2055
-111.914
178.411
Deviance Test DF
4.05467
53.471
P-value
0.2556
<.0001
Goodness of
Dose
0.
1.
13.
26.
47.
0000
3000
2000
5000
5000
Est,
0,
0,
0,
0,
0,
. Prob.
.3213
.3690
.6764
.8466
.9528
Fit
Scaled
Expected Observed Size Residual
16.
12.
16.
29.
20.
,064
,178
,911
,631
,009
18
11
14
30
21
50
33
25
35
21
0.
-0
-1
0
5863
.425
.244
.173
1.02
Chi-sguare =3.14 DF = 3
Benchmark Dose Computation
P-value = 0.3701
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0.1
Extra risk
0.95
1.87738
1.37335
D-10
-------
Quantal Quadratic Model $Revision: 2.2 $ $Date: 2000/03/17 22:27:16 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY DOCUMENTS\DCA\FINAL
MODELING\5DOSEQQUADRATIC.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY DOCUMENTS\DCA\FINAL
MODELING\5DOSEQQUADRATIC.pit
Sat Aug 03 10:05:52 2002
DCA DeAngelo (1999) Five Doses
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(-slope*dose/x2)]
Dependent variable = Either
Independent variable = HED
Total number of observations = 5
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial (and Specified) Parameter Values
Background = 0.362745
Slope = 0.0014775
Power = 2 Specified
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Power
have been estimated at a boundary point, or have been specified by
the user, and do not appear in the correlation matrix )
Background
Slope
Background
1
-0.29
Slope
-0.29
1
Parameter Estimates
Variable
Background
Slope
Model
Full model
Fitted model
Reduced model
Estimate
0.34789
0.00223738
Std. Err.
0.0507344
0.000540864
Analysis of Deviance Table
Log(likelihood) Deviance Test DF
-85.1782
-85.3116 0.266877 3
-111.914 53.471 4
P-value
0.9661
<.0001
AIC:
174.623
D-ll
-------
Goodness of Fit
E
0.
1.
13.
26.
47.
lose
0000
3000
2000
5000
5000
Est,
0,
0,
0,
0,
0,
. Prob.
.3479
.3504
.5584
.8645
.9958
Exped
17,
11,
13,
30,
20,
;ed 01
.394
.562
.960
.257
.912
:>served '
18
11
14
30
21
Size
50
33
25
35
21
Scale
Resid
0
-0
0.
-0
0
d
ual
.1798
.2049
01597
.1271
.2972
Chi-square =0.18
DF = 3
P-value = 0.9809
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0 . 95
BMD = 6.86229
BMDL = 5.69087
Weibull Model $Revision: 2.2 $ $Date: 2000/03/17 22:27:16 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY DOCUMENTS\DCA\FINAL
MODELING\5DOSEWEIBULL.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY DOCUMENTS\DCA\FINAL
MODELING\5DOSEWEIBULL.pit
Sat Aug 03 10:06:59 2002
DCA DeAngelo (1999) Five Doses
The form of the probability function is:
P[response] = background + (1-background) * [1-EXP (-slope*dose/xpower) ]
Dependent variable = Either
Independent variable = HED
Power parameter is not restricted
Total number of observations = 5
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial (and Specified) Parameter Values
Background = 0.362745
Slope = 0.00510196
Power = 1.67901
Asymptotic Correlation Matrix of Parameter Estimates
Background Slope Power
D-12
-------
Background
Slope
Power
1
-0.31
0.29
-0.31
1
-1
0.29
-1
1
Variable
Background
Slope
Power
Parameter Estimates
Estimate Std. Err.
0.35075 0.0529332
0.00136633 0.00364981
2.15173 0.817672
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
AIC:
Log(likelihood)
-85.1782
-85.2928
-111.914
176.586
Deviance Test DF
P-value
0.229231
53.471
0.8917
<.0001
Goodness of Fit
Dose
0.
1.
13.
26.
47.
0000
3000
2000
5000
5000
Est,
0,
0,
0,
0,
0,
. Prob.
.3507
.3523
.5435
.8659
.9974
Expected Observed Size
17,
11.
13,
30,
20,
.537
.626
.587
.308
.946
18
11
14
30
21
50
33
25
35
21
Scaled
Residual
0.
-0.
0
-0.
0.
1371
2282
.166
1529
2319
Chi-sguare =0.18
DF = 2
P-value = 0.9160
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0 . 95
BMD = 7.53401
BMDL = 2.50369
D-13
-------
APPENDIX E
BENCHMARK DOSE-RESPONSE FOR CANCER ENDPOINTS
DeAngelo et al., 1999 (4 and 6 Doses)
-------
Cancer BMD Modeling Results - 4 and 6 Dose Groups
As discussed in Section 5.3, the BMDL was estimated using data on the numbers of animals
with either hepatocarcinoma or hepatoadenoma from the DeAngelo et al. (1999) study. The highest
dose group was excluded from the main analysis because the highest dose in the study (64.6 mg/kg-day
HED) resulted in reduced weight gain over the second half of the study and a greatly increased severity
of hepatic necrosis in test animals. Based on these findings, it was judged that the highest dose was at
or near the MTD.
For purposes of comparison, this appendix presents the results of BMD modeling when all
dose groups are included. In addition, BMDL modeling has been performed with the two highest dose
groups excluded. This approach was taken because the prevalence of combined hepatocarcinoma and
adenoma in the second highest dose group was 100%, and omitting this group (and the highest-dose
group) could result in a better model fit in the low-dose range.
Results of the BMD modeling using all six dose groups and the four lowest dose groups are
summarized in Tables E-l and E-2. When all six dose groups are used, the results are very similar to
those presented in Section 5.3 for modeling with the highest dose omitted. The multistage and quantal-
quadratic models estimate identical BMDs and have identical p-values and AICs. The BMDL
estimated using the multistage model with all six dose groups is again 2.1 mg/kg-day (rounded up from
2.08), essentially the same as that obtained when the high dose group was omitted (2.1 mg/kg-day,
rounded from 2.05). The BMDLs estimated from the good-fitting models (p-values greater than 0.8)
range from 2.08 (multistage) to 5.69 mg/kg-day (quantal-quadratic).
Table E-l. Results of benchmark dose carcinogenicity
including all (six) dose groups 1>2
Model
Multi-Stage(2)
Quantal-quadr.
Weibull
Gamma
Probit
Log-Probit
Logistic
Log-Logistic
BMD
6.86
6.86
7.54
8.49
3.15
9.46
3.08
9.37
BMDL
2.08
5.69
2.61
2.73
2.53
4.62
2.42
4.51
p-value
0.996
0.996
0.981
0.957
0.917
0.910
0.849
0.846
AIC
174.63
174.63
176.59
176.82
175.55
177.14
176.21
177.59
E-l
-------
iQuantal-linear
1.83
1.35
0.489
178.79
1. Data = DeAngelo, et al. (1999), animals with hepatocarcinoma or adenoma
2. BMD estimated using BMR = 0.10, BMDL estimated as 95% UCL
When the two highest dose groups are omitted, the multistage and quantal-quadratic model
again provide the best fits to the data (p-value = 0.962, AIC = 174.43). On the whole, the p-values
are slightly lower than when five or six dose groups are used. In this case, the multistage (as well as the
gamma, Weibull, and log-logistic models) predict a BMDL of 1 .7 mg/kg-day. The other four good-
fitting models (p-value above 0.7) predict BMDLs of 1.83 (log-probit), 2.59 (probit), 2.51 (logistic),
and 5.72 mg/kg-day (quantal-quadratic).
Table E-2. Results of benchmark dose carcinogenicity excluding the
two highest dose groups 1|2
Model
Multi-Stage(2)
Quantal-quadr.
Log-Probit
Log-Logistic
Gamma
Weibull
Probit
Logistic
Quantal-linear
BMD
6.96
6.96
8.28
7.95
7.49
6.78
3.30
3.28
2.17
BMDL
1.70
5.72
1.83
1.73
1.70
1.70
2.59
2.51
1.50
p-value
0.962
0.962
0.802
0.800
0.800
0.783
0.743
0.714
0.439
AIC
174.43
174.43
176.42
176.42
176.42
176.43
174.95
175.03
176.01
1. Data = DeAngelo, et al. (1999), animals with hepatocarcinoma or adenoma, omitting
two highest dose groups (control and three dosed groups)
2. BMD estimated using BMR = 0.10, BMDL estimated as 95% UCL
Including the highest dose group therefore has very little effect on the estimated BMDL.
Excluding the two highest dose groups, results in the bulk of the models estimate BMDLs that are 15 to
20% lower than the multistage model estimate when the single highest dose group is excluded. As
noted in Section 5.3, because of the lack of conclusive evidence concerning the carcinogenic mode of
action of DC A, the quantal-quadratic model, which assumes zero slope at zero dose, was not
considered to provide reliable BMDL estimates, despite its good fit to the data.
E-2
-------
Benchmark Modeling Results for Entire Data Set (6 Dose Groups)
$Revision: 2.2 $ $Date: 2001/03/14 01:17:00 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA_ALL_TUMORS.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA_ALL_TUMORS.pit
Thu Aug 08 10:57:03 2002
UBMDS MODEL RUN
The form of the probability function is:
P [response] = background+ (1 -background) *CumGamma [si ope* dose, power]
where CumGamma ( . ) is the cumulative Gamma distribution function
Dependent variable = Either
Independent variable = HED
Power parameter is restricted as power >=1
Total number of observations = 6
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial (and Specified) Parameter Values
Background = 0.362745
Slope = 0.0528134
Power = 1.3
Asymptotic Correlation Matrix of Parameter Estimates
Background
Slope
Power
Background
1
0.19
0.26
Slope
0.19
1
0.98
Power
0.26
0.98
1
Parameter Estimates
Variable
Background
Slope
Power
Estimate
0.351181
0.213843
4.10684
Std. Err.
0.0526892
0.135637
2.78823
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
Log(likelihood)
-85.1782
-85.4078
-117.777
Deviance Test DF
0.459312
65.1977
P-value
0.9277
<.0001
E-3
-------
AIC: 176.816
Goodness of Fit
E
0.
1.
13.
26.
47.
64.
lose
0000
3000
2000
5000
5000
6000
Est,
0,
0,
0,
0,
0,
0,
. Prob.
.3512
.3513
.5411
.8718
.9932
.9996
Exped
17,
11,
13,
30,
20,
10,
;ed 01
.559
.592
.529
.513
.858
.995
userved '
18
11
14
30
21
11
Size
50
33
25
35
21
11
Scale
Resid
0
-0
0
-0
0
0.
d
ual
.1306
.2159
.1891
.2593
.3782
06768
Chi-square =
0.31
DF = 3
P-value = 0.9573
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0.1
Extra risk
0.95
8.49415
2.73397
Logistic Model $Revision: 2.1 $ $Date: 2000/02/26 03:38:20 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA_ALL_TUMORS.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA_ALL_TUMORS.pit
Thu Aug 08 10:59:14 2002
UBMDS MODEL RUN
The form of the probability function is:
P[response] = I/[1+EXP(-intercept-slope*dose)]
Dependent variable = Either
Independent variable = HED
Slope parameter is not restricted
Total number of observations = 6
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
E-4
-------
Default Initial Parameter Values
background = 0 Specified
intercept = -0.493007
slope = 0.0690064
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -background
have been estimated at a boundary point, or have been specified by the
user, and do not appear in the correlation matrix )
intercept
slope
intercept
1
-0.57
slope
-0.57
1
Parameter Estimates
Variable
intercept
slope
Model
Full model
Fitted model
Reduced model
Estimate
-0.739275
0.0958841
Std. Err.
0.226806
0.0169284
Analysis of Deviance Table
Log(likelihood)
-85.1782
-86.105
-117.777
Deviance Test DF
1.85361
65.1977
P-value
0.7627
<.0001
AIC:
176.21
Goodness of Fit
Dose
0.
1.
13.
26.
47.
64.
0000
3000
2000
5000
5000
6000
Est,
0,
0,
0,
0,
0,
0,
. Prob.
.3232
.3510
.6286
.8584
.9784
.9957
Expected Observed Size
16,
11.
15,
30,
20,
10,
.158
.583
.716
.042
.547
.953
18
11
14
30
21
11
50
33
25
35
21
11
Scaled
Residual
-0
-0
-0.
0
0
0.557
.2127
.7103
02049
.6802
.2169
Chi-sguare =
1.37
DF = 4
P-value = 0.8494
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0.1
Extra risk
0.95
3.08205
2.42162
E-5
-------
Logistic Model $Revision: 2.1 $ $Date: 2000/02/26 03:38:20 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA_ALL_TUMORS.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA_ALL_TUMORS.pit
Thu Aug 08 10:58:15 2002
UBMDS MODEL RUN
The form of the probability function is:
P[response] = background+(1-background)/[1+EXP(-intercept-slope*Log(dose))]
Dependent variable = Either
Independent variable = HED
Slope parameter is restricted as slope >= 1
Total number of observations = 6
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
User has chosen the log transformed model
Default Initial Parameter Values
background = 0.36
intercept = -5.08987
slope = 2.02139
Asymptotic Correlation Matrix of Parameter Estimates
slope
0.24
-0.99
1
Parameter Estimates
background
background
intercept
slope
1
-0.28
0.24
intercept
-0.28
1
-0.99
Variable
background
intercept
slope
Model
Full model
Fitted model
Reduced model
Estimate
0.35305
-10.1683
3.56171
Std. Err.
0.0531767
3.73822
1.1728
Analysis of Deviance Table
Log(likelihood)
-85.1782
-85.7956
-117.777
Deviance Test DF
1.23487
65.1977
P-value
0.7447
<.0001
AIC:
177.591
E-C
-------
Goodness of
Dose
0.
1.
13.
26.
47.
64.
0000
3000
2000
5000
5000
6000
Est,
0,
0,
0,
0,
0,
0,
. Prob.
.3531
.3531
.5298
.8824
.9825
.9940
Fit
Scaled
Expected Observed Size
17,
11.
13,
30,
20,
10,
.653
.653
.245
.883
.632
.934
18
11
14
30
21
11
50
33
25
35
21
11
Residual
0.
-0.
0.
-0
0.
0.
1028
2377
3024
.463
6116
2568
Chi-square =
0.81
DF = 3
P-value = 0.8464
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0.1
Extra risk
0.95
9.37448
4.51405
Probit Model $Revision: 2.1 $ $Date: 2000/02/26 03:38:53 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA_ALL_TUMORS.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA_ALL_TUMORS.pit
Thu Aug 08 11:01:31 2002
UBMDS MODEL RUN
The form of the probability function is:
P[response] = Background
+ (1-Background) * CumNorm(Intercept+Slope*Log(Dose)
where CumNorm(.) is the cumulative normal distribution function
Dependent variable = Either
Independent variable = HED
Slope parameter is not restricted
Total number of observations = 6
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
E-7
-------
User has chosen the log transformed model
Default Initial (and Specified) Parameter Values
background = 0.36
intercept = -2.62162
slope = 1.0525
Asymptotic Correlation Matrix of Parameter Estimates
background intercept slope
background 1 -0.27 0.24
intercept -0.27 1 -0.99
slope 0.24 -0.99 1
Parameter Estimates
Variable
background
intercept
slope
Estimate
0.352048
-6.0128
2.10539
Std. Err.
0.0528377
2.11645
0.661561
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
Log(likelihood)
-85.1782
-85.5714
-117.777
Deviance Test DF
0.786451
65.1977
P-value
0.8527
<.0001
AIC:
177.143
Dose Est,
0.
1.
13.
26.
47.
64.
0000
3000
2000
5000
5000
6000
0,
0,
0,
0,
0,
0,
Chi-sguare =
Goodness of Fit
. Prob. Expected
.3520
.3520
.5340
.8785
.9889
.9981
0.54
17.
11.
13.
30.
20.
10.
DF =
,602
,618
,350
,746
,766
,980
3
Observed
18
11
14
30
21
11
P-value
Size
50
33
25
35
21
11
= 0.9104
Scaled
Residual
0.
-0.
0.
-0.
0.
0
1177
2251
2606
3859
4864
.143
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0.1
Extra risk
0.95
9.46131
4.6164
Multistage Model. $Revision: 2.1 $ $Date: 2000/08/21 03:38:21 $
-------
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA_ALL_TUMORS.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA_ALL_TUMORS.pit
Thu Aug 08 11:02:47 2002
UBMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*dose/xl-beta2*dose/x2) ]
The parameter betas are restricted to be positive
Dependent variable = Either
Independent variable = HED
Total number of observations = 6
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
Background = 0
Beta(l) = 5.62186e+017
Beta(2) = 1.98607e+016
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(l)
have been estimated at a boundary point, or have been specified by
the user, and do not appear in the correlation matrix )
Background Beta(2)
Background 1 -0.37
Beta(2) -0.37 1
Parameter Estimates
Variable Estimate Std. Err.
Background 0.347869 0.0838449
Beta(l) 0 NA
Beta(2) 0.00223812 0.000610001
NA - Indicates that this parameter has hit a bound
implied by some ineguality constraint and thus
E-9
-------
has no standard error.
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
AIC:
Log (likelihood) Deviance Test DF P-value
-85.
-85.
-117
174
1782
3123
.777
.625
0.26814 4 0.9918
65.1977 5 <.0001
Goodness of Fit
i
i
i
i
i
i
Dose Est
: 1
0.0000 0.
: 2
1.3000 0.
: 3
13.2000 0.
: 4
26.5000 0.
: 5
47.5000 0.
: 6
64.6000 0.
Chi-square =
. Prob.
3479
3503
5585
8646
9958
9999
0.18
Expected
17.393
11.561
13.961
30.260
20.912
10.999
DF = 4
Observed Size ChiA2 Res.
18 50 0.053
11 33 -0.075
14 25 0.006
30 35 -0.063
21 21 1.004
11 11 1.000
P-value = 0.9962
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 6.86115
BMDL = 2.07845
Probit Model $Revision: 2.1 $ $Date: 2000/02/26 03:38:53 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA_ALL_TUMORS.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA_ALL_TUMORS.pit
Thu Aug 08 11:03:53 2002
UBMDS MODEL RUN
The form of the probability function is:
P[response] = CumNorm(Intercept+Slope*Dose),
E-10
-------
where CumNorm(.) is the cumulative normal distribution function
Dependent variable = Either
Independent variable = HED
Slope parameter is not restricted
Total number of observations = 6
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial (and Specified) Parameter Values
background = 0 Specified
intercept = -0.366862
slope = 0.0471876
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -background
have been estimated at a boundary point, or have been specified by the
user, and do not appear in the correlation matrix )
intercept
intercept 1
slope -0.6
slope
-0.6
1
Parameter Estimates
Variable
intercept
slope
Model
Full model
Fitted model
Reduced model
AIC:
Dose Est,
0.0000 0,
1.3000 0,
13.2000 0,
26.5000 0,
47.5000 0,
64.6000 0,
Estimate
-0.45065
0.0572063
Analysis of Deviance
Log (likelihood) Deviance
-85.1782
-85.7748 1.19325
-117.777 65.1977
175.55
Goodness of Fit
. Prob. Expected Obser
.3261 16.306
.3534 11.661
.6196 15.490
.8566 29.982
.9883 20.754
.9994 10.994
Std. Err.
0.137888
0.00964233
Table
Test DF
4
5
ved Size
18
11
14
30
21
11
P-value
0.8792
<.0001
Scaled
Residual
50 0.511
33 -0.2406
25 -0.614
35 0.008589
21 0.4987
11 0.08041
Chi-sguare =
0.95
DF = 4
P-value = 0.9171
E-ll
-------
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0 . 95
BMD = 3.15463
BMDL = 2.53418
Quantal Linear Model $Revision: 2.2 $ $Date: 2000/03/17 22:27:16 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA_ALL_TUMORS.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA_ALL_TUMORS.pit
Thu Aug 08 11:05:32 2002
UBMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(-slope*dose)]
Dependent variable = Either
Independent variable = HED
Total number of observations = 6
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial (and Specified) Parameter Values
Background = 0.362745
Slope = 0.0422209
Power = 1 Specified
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Power
have been estimated at a boundary point, or have been specified by
the user, and do not appear in the correlation matrix )
Background Slope
Background 1 -0.32
Slope -0.32 1
Parameter Estimates
Variable Estimate Std. Err.
Background 0.319734 0.0523578
Slope 0.0576209 0.0110511
E-12
-------
Analysis of Deviance Table
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-85.1782
-87.3967
-117.777
Deviance Test DF
P-value
4.43708
65.1977
0.3501
<.0001
AIC: 178.793
Goodness of Fit
D
0.
1.
13.
26.
47.
64.
lose
0000
3000
2000
5000
5000
6000
Est,
0,
0,
0,
0,
0,
0,
. Prob.
.3197
.3688
.6821
.8522
.9559
.9836
Expect
15.
12.
17.
29.
20.
10.
;ed 01
,987
,171
,051
,829
,075
,819
userved '
18
11
14
30
21
11
Size
50
33
25
35
21
11
Scale
Resid
0
-0
0.
0
0
d
ual
.6105
.4226
-1.31
08159
.9838
.4289
Chi-square =
3.43
DF = 4
P-value = 0.4890
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0.1
Extra risk
0.95
1.82851
1.35289
Quantal Quadratic Model $Revision: 2.2 $ $Date: 2000/03/17 22:27:16 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA_ALL_TUMORS.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA_ALL_TUMORS.pit
Thu Aug 08 11:06:41 2002
UBMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(-slope*dose/x2)]
Dependent variable = Either
Independent variable = HED
Total number of observations = 6
Total number of records with missing values = 0
Maximum number of iterations = 250
E-13
-------
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial (and Specified) Parameter Values
Background = 0.362745
Slope = 0.000653574
Power = 2 Specified
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Power
have been estimated at a boundary point, or have been specified by
the user, and do not appear in the correlation matrix )
Background
Slope
Background
1
-0.29
Slope
-0.29
1
Parameter Estimates
Variable
Background
Slope
Model
Full model
Fitted model
Reduced model
Estimate
0.347869
0.00223812
Std. Err.
0.050728
0.000540196
Analysis of Deviance Table
Log(likelihood)
-85.1782
-85.3123
-117.777
Deviance Test DF
0.26814
65.1977
P-value
0.9918
<.0001
AIC: 174.625
Goodness of Fit
D
0.
1.
13.
26.
47.
64.
lose
0000
3000
2000
5000
5000
6000
Est,
0,
0,
0,
0,
0,
0,
. Prob.
.3479
.3503
.5585
.8646
.9958
.9999
Exped
17,
11,
13,
30,
20,
10,
;ed 01
.393
.561
.961
.260
.912
.999
userved '
18
11
14
30
21
11
Size
50
33
25
35
21
11
Scale
Resid
0
-0
0.
-0
0
0
d
ual
.1801
.2047
01553
.1282
.2969
.0251
Chi-sguare =
0.18
DF = 4
P-value = 0.9962
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0.1
Extra risk
0.95
6.86116
5.69064
E-14
-------
Weibull Model $Revision: 2.2 $ $Date: 2000/03/17 22:27:16 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA_ALL_TUMORS.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA_ALL_TUMORS.pit
Thu Aug 08 11:07:43 2002
UBMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background) * [1-EXP (-slope*dose/xpower) ]
Dependent variable = Either
Independent variable = HED
Power parameter is restricted as power >=1
Total number of observations = 6
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial (and Specified) Parameter Values
Background = 0.362745
Slope = 0.0273825
Power = 1.10388
Asymptotic Correlation Matrix of Parameter Estimates
Background
Slope
Power
Background
1
-0.31
0.29
Slope
-0.31
1
-1
Power
0.29
-1
1
Parameter Estimates
Variable
Background
Slope
Power
Estimate
0.350763
0.00136197
2.15274
Std. Err.
0.0529253
0.00362597
0.814689
Analysis of Deviance Table
E-15
-------
Model
Full model
Fitted model
Reduced model
Log(likelihood)
-85.1782
-85.293
-117.777
Deviance Test DF
0.229541
65.1977
P-value
0.9727
<.0001
AIC:
176.586
Goodness of Fit
E
0.
1.
13.
26.
47.
64.
lose
0000
3000
2000
5000
5000
6000
Est,
0,
0,
0,
0,
0,
1.
. Prob.
.3508
.3523
.5434
.8660
.9975
.0000
Expect
17.
11.
13.
30.
20.
11.
;ed 01
,538
,626
,584
,309
,947
,000
userved '
18
11
14
30
21
11
Size
50
33
25
35
21
11
Scale
Resid
0
-0
0
-0
0
0.
d
ual
.1369
.2283
.1668
.1534
.2315
01242
Chi-square =
0.18
DF = 3
P-value = 0.9814
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0.1
Extra risk
0.95
7.53803
2.61177
Benchmark Dose Modeling Results Excluding Two Highest Dose Groups
(Four Dose Groups)
$Revision: 2.2 $ $Date: 2001/03/14 01:17:00 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA4HIGHEST.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA4HIGHEST.pit
Thu Jul 18 09:00:04 2002
BMDS MODEL RUN
The form of the probability function is:
P[response]= background+(1-background)*CumGamma[siope*dose,power],
E-16
-------
where CumGamma(.) is the cumulative Gamma distribution function
Dependent variable = Either
Independent variable = HED
Power parameter is restricted as power >=1
Total number of observations = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial (and Specified) Parameter Values
Background = 0.362745
Slope = 0.186832
Power = 3.70123
Asymptotic Correlation Matrix of Parameter Estimates
Background
Slope
Power
Background
1
0.17
0.22
Slope
0.17
1
0.99
Power
0.22
0.99
1
Parameter Estimates
Variable
Background
Slope
Power
Estimate
0.349291
0.16719
3.24212
Std. Err.
0.0525503
0.140528
2.7498
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
Log(likelihood)
-85.1782
-85.2105
-99.(
Deviance Test DF
0.0646347
27.8208
P-value
0.7993
<.0001
AIC:
176.421
Goodness of Fit
Dose I
0.0000
1.3000
13.2000
26.5000
Chi-sguare =
!st.
0,
0,
0,
0,
. Prob.
.3493
.3498
.5595
.8573
0.06
Expect
17.
11.
13.
30.
DF =
;ed
,465
,542
,987
,005
1
Observed
18
11
14
30
P-value
Size
50
33
25
35
= 0.7996
Scaled
Residual
0.1588
-0.198
0.005356
-0.002554
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
E-17
-------
Confidence level = 0.95
BMD = 7.49106
BMDL = 1.69767
Logistic Model $Revision: 2.1 $ $Date: 2000/02/26 03:38:20 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA4HIGHEST.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA4HIGHEST.pit
Thu Aug 08 10:40:29 2002
BMDS MODEL RUN
The form of the probability function is:
P[response] = I/[1+EXP(-intercept-slope*dose)]
Dependent variable = Either
Independent variable = HED
Slope parameter is not restricted
Total number of observations = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
background = 0 Specified
intercept = -0.724563
slope = 0.0880096
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -background
have been estimated at a boundary point, or have been specified by
the user, and do not appear in the correlation matrix )
intercept slope
intercept 1 -0.6
slope -0.6 1
Parameter Estimates
Variable Estimate Std. Err.
intercept -0.701555 0.230098
slope 0.0881974 0.0187968
Analysis of Deviance Table
E-18
-------
Model
Full model
Fitted model
Reduced model
AIC:
Log(likelihood)
-85.1782
-85.5143
-99.0886
175.029
Deviance Test DF
P-value
0.672212
27.8208
0.7145
<.0001
Goodness of Fit
0
1
13
26
Dose
.0000
.3000
.2000
.5000
Est,
0,
0,
0,
0,
. Prob.
.3315
.3573
.6136
.8369
Exped
16,
11.
15,
29,
;ed
.573
.792
.341
.293
Observed
18
11
14
30
Size
50
33
25
35
Scaled
Residu
0.
-0.
-0.
0.
.al
4286
2879
5508
3235
Chi-square =
0.67
DF = 2
P-value = 0.7137
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 3.27774
BMDL = 2.51045
Logistic Model $Revision: 2.1 $ $Date: 2000/02/26 03:38:20 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA4HIGHEST.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA4HIGHEST.pit
Thu Jul 18 09:01:00 2002
BMDS MODEL RUN
The form of the probability function is:
P[response] = background+(1-background)/[1+EXP(-intercept-slope*Log(dose))]
Dependent variable = Either
Independent variable = HED
Slope parameter is restricted as slope >= 1
Total number of observations = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
User has chosen the log transformed model
E-19
-------
Default Initial Parameter Values
background = 0.36
intercept = -4.76605
slope = 1.72426
Asymptotic Correlation Matrix of Parameter Estimates
slope
0.17
-0.99
1
background
background
intercept
slope
1
-0.21
0.17
intercept
-0.21
1
-0.99
Parameter Estimates
Variable
background
intercept
slope
Model
Full model
Fitted model
Reduced model
Estimate
0.349285
-8.17034
2.88043
Std. Err.
0.0524508
3.8887
1.26948
Analysis of Deviance Table
Log(likelihood)
-85.1782
-85.2103
-99.(
Deviance Test DF
0.0642027
27.8208
P-value
O.E
<.0001
AIC: 176.421
Goodness of Fit
Dose
0.0000
1.3000
13.2000
26.5000
Est,
0,
0,
0,
0,
. Prob.
.3493
.3497
.5597
.8572
Expected
17.464
11.539
13.993
30.003
Observed
18
11
14
30
Size
50
33
25
35
Scaled
Residual
0.1589
-0.1969
0.002802
-0.001644
Chi-sguare =
0.06
DF = 1
P-value = 0.8002
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 7.95414
BMDL = 1.73353
Probit Model $Revision: 2.1 $ $Date: 2000/02/26 03:38:53 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA4HIGHEST.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA4HIGHEST.pit
E-20
-------
Thu Aug 08 10:48:03 2002
BMDS MODEL RUN
The form of the probability function is:
P[response] = Background
+ (1-Background) * CumNorm(Intercept+Slope*Log(Dose)) ,
where CumNorm(.) is the cumulative normal distribution function
Dependent variable = Either
Independent variable = HED
Slope parameter is not restricted
Total number of observations = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
User has chosen the log transformed model
Default Initial (and Specified) Parameter Values
background = 0.36
intercept = -2.48935
slope = 0.92656
Asymptotic Correlation Matrix of Parameter Estimates
background
intercept
slope
background
1
-0.21
0.17
intercept
-0.21
1
-0.99
slope
0.17
-0.99
1
Parameter Estimates
Variable
background
intercept
slope
Model
Full model
Fitted model
Reduced model
Estimate
0.349398
-5.01493
1.76634
Std. Err.
0.0523345
2.34977
0.763476
Analysis of Deviance Table
Log(likelihood) Deviance Test DF
-85.1782
-85.2094 0.0623592 1
-99.0886 27.8208 3
P-value
0.8028
<.0001
AIC:
176.419
E-21
-------
Goodness of
Dose
0.
1.
13.
26.
0000
3000
2000
5000
Est,
0,
0,
0,
0,
. Prob.
.3494
.3494
.5600
.8571
Fit
Scaled
Expected Observed Size
17.
11.
14.
30.
,470
,530
,000
,000
18
11
14
30
50
33
25
35
Residual
0.
-0.
1572
1936
3.554e-005
-2.898e
:-005
Chi-square =
0.06
DF = 1
P-value = 0.8031
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0.1
Extra risk
0.95
8.27816
1.83164
Multistage Model. $Revision: 2.1 $ $Date: 2000/08/21 03:38:21 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA4HIGHEST.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA4HIGHEST.pit
Thu Jul 18 09:01:41 2002
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(
-betal*dose/xl-beta2*dose/x2) ]
The parameter betas are restricted to be positive
Dependent variable = Either
Independent variable = HED
Total number of observations = 4
Total number of records with missing values = 0
Total number of parameters in model = 3
Total number of specified parameters = 0
Degree of polynomial = 2
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
E-22
-------
Default Initial Parameter Values
Background = 0.345937
Beta(l) = 0.00210524
Beta(2) = 0.00208855
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Beta(l)
have been estimated at a boundary point, or have been specified by
the user, and do not appear in the correlation matrix )
Background
Beta(2)
Background
1
-0.39
Beta(2)
-0.39
1
Parameter Estimates
Variable
Background
Beta(l)
Beta(2)
Estimate
0.349311
0
0.00217333
Std. Err.
0.0845452
NA
0.000652057
NA - Indicates that this parameter has hit a bound
implied by some ineguality constraint and thus
has no standard error.
Model
Full model
Fitted model
Reduced model
Analysis of Deviance Table
Log(likelihood) Deviance Test DF
-85.1782
-85.2172 0.0779541 2
-99.0886 27.8208 3
P-value
0.9618
<.0001
AIC: 174.434
Goodness of Fit
Dose Est._Prob. Expected Observed
Size
ChiA2 Res.
1
0.
2
1.
3
13.
4
26.
0000
3000
2000
5000
0,
0,
0,
0,
.3493
.3517
.5544
.8586
17.
11.
13.
30.
,466
,606
,861
,050
18
11
14
30
50
33
25
35
0.
-0.
0.
-0.
,047
,081
,023
,012
Chi-sguare = 0.08 DF = 2
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0 . 95
BMD = 6.96267
BMDL = 1.69536
P-value = 0.9619
E-23
-------
Probit Model $Revision: 2.1 $ $Date: 2000/02/26 03:38:53 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA4HIGHEST.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA4HIGHEST.pit
Thu Aug 08 10:52:47 2002
BMDS MODEL RUN
The form of the probability function is:
P[response] = CumNorm(Intercept+Slope*Dose),
where CumNorm(.) is the cumulative normal distribution function
Dependent variable = Either
Independent variable = HED
Slope parameter is not restricted
Total number of observations = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial (and Specified) Parameter Values
background = 0 Specified
intercept = -0.429874
slope = 0.0538043
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -background
have been estimated at a boundary point, or have been specified by
the user, and do not appear in the correlation matrix )
intercept
slope
intercept
1
-0.61
Variable
intercept
slope
Model
slope
-0.61
1
Parameter Estimates
Estimate
-0.433568
0.0539456
Std. Err.
0.140472
0.0109001
Analysis of Deviance Table
Log(likelihood) Deviance Test DF
E-24
P-value
-------
Full model
Fitted model
Reduced model
AIC:
-85.1782
-85.4739
-99.0886
174.948
0.591454
27.8208
0.744
<.0001
E
0.
1.
13.
26.
lose
0000
3000
2000
5000
Est,
0,
0,
0,
0,
Goodns
. Prob.
.3323
.3581
.6097
.8404
5SS Of
Exped
16,
11,
15,
29,
Fit
;ed 01
.615
.819
.242
.413
userved '
18
11
14
30
Size
50
33
25
35
Scaled
Residu
0.
-0.
-0.
0.
.al
4158
2972
5093
2709
Chi-square =
0.59
DF = 2
P-value = 0.7430
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
BMDL =
0.1
Extra risk
0.95
3.2962
2.58591
Quantal Linear Model $Revision: 2.2 $ $Date: 2000/03/17 22:27:16 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA4HIGHEST.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA4HIGHEST.pit
Thu Jul 18 09:03:07 2002
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(-slope*dose)]
Dependent variable = Either
Independent variable = HED
Total number of observations = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial (and Specified) Parameter Values
Background = 0.362745
Slope = 0.0538938
E-25
-------
Power = 1 Specified
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Power
have been estimated at a boundary point, or have been specified by
the user, and do not appear in the correlation matrix )
Background
Slope
Background
1
-0.37
Slope
-0.37
1
Parameter Estimates
Variable
Background
Slope
Model
Full model
Fitted model
Reduced model
Estimate
0.328661
0.0485733
Std. Err.
0.0533696
0.011993
Analysis of Deviance Table
Log (likelihood)
-85.1782
-86.0023
-99.0886
Deviance Test DF
1.64815
27.8208
P-value
0.4386
<.0001
AIC:
176.005
Goodness of Fit
Scaled
Dose
0.
1.
13.
26.
0000
3000
2000
5000
Est.
0.
0.
0.
0.
, Prob.
,3287
,3697
,6464
,8147
Expected Observed Size
16,
12,
16,
28,
.433
.201
.161
.514
18
11
14
30
50
33
25
35
Residual
0.
-0.
-0.
0.
4718
4333
9038
6465
Chi-sguare =
1.65
DF = 2
P-value = 0.4393
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0 . 95
BMD = 2.16911
BMDL = 1.49987
Quantal Quadratic Model $Revision: 2.2 $ $Date: 2000/03/17 22:27:16 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA4HIGHEST.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA4HIGHEST.pit
Thu Jul 18 09:04:01 2002
E-26
-------
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background)*[1-EXP(-slope*dose/x2)]
Dependent variable = Either
Independent variable = HED
Total number of observations = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial (and Specified) Parameter Values
Background = 0.362745
Slope = 0.00203373
Power = 2 Specified
Asymptotic Correlation Matrix of Parameter Estimates
( *** The model parameter(s) -Power
have been estimated at a boundary point, or have been
specified by the user, and do not appear in the correlation
matrix )
Background
Slope
Background
1
-0.31
Slope
-0.31
1
Parameter Estimates
Variable
Background
Slope
Model
Full model
Fitted model
Reduced model
Estimate
0.349311
0.00217333
Std. Err.
0.0510221
0.000565508
Analysis of Deviance Table
Log(likelihood) Deviance Test DF
-85.1782
-85.2172 0.0779541 2
-99.0886 27.8208 3
P-value
0.9618
<.0001
AIC: 174.434
Goodness of Fit
Dose
0.0000
Est._Prob.
0.3493
Expected Observed
17.466 18
Size
Scaled
Residual
50
0.1585
E-27
-------
1.3000 0.3517 11.606 11 33 -0.2209
13.2000 0.5544 13.861 14 25 0.05602
26.5000 0.8586 30.050 30 35 -0.02422
Chi-square = 0.08 DF = 2 P-value = 0.9619
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0 . 95
BMD = 6.96267
BMDL = 5.72293
Weibull Model $Revision: 2.2 $ $Date: 2000/03/17 22:27:16 $
Input Data File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA4HIGHEST.(d)
Gnuplot Plotting File: C:\DOCUMENTS AND SETTINGS\06157\MY
DOCUMENTS\DCA\DCA4HIGHEST.pit
Thu Jul 18 09:04:51 2002
BMDS MODEL RUN
The form of the probability function is:
P[response] = background + (1-background) * [1-EXP (-slope*dose/xpower) ]
Dependent variable = Either
Independent variable = HED
Power parameter is restricted as power >=1
Total number of observations = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial (and Specified) Parameter Values
Background = 0.362745
Slope = 0.00220968
Power = 1.97468
Asymptotic Correlation Matrix of Parameter Estimates
Background Slope Power
Background 1 -0.3 0.28
E-28
-------
Slope
Power
-0.3
0.28
-1
1
Parameter Estimates
Variable
Background
Slope
Power
Estimate
0.34865
0.00248422
1.95791
Std. Err.
0.0531378
0.00736373
0.930113
Model
Full model
Fitted model
Reduced model
AIC:
Analysis of Deviance Table
Log(likelihood)
-85.1782
-85.2162
-99.0886
176.432
Deviance Test DF
0.0759528
27.8208
P-value
0.7829
<.0001
D
0.
1.
13.
26.
lose
0000
3000
2000
5000
Est,
0,
0,
0,
0,
Goodns
. Prob.
.3487
.3513
.5583
.8575
5SS Of
Exped
17,
11,
13,
30,
Fit
;ed 01
.433
.595
.956
.013
userved '
18
11
14
30
Size
50
33
25
35
Scaled
Residual
0.1684
-0.2168
0.01761
-0.006207
Chi-sguare =
0.08
DF = 1
P-value = 0.7832
Benchmark Dose Computation
Specified effect = 0.1
Risk Type = Extra risk
Confidence level = 0.95
BMD = 6.78011
BMDL = 1.6957
E-29
------- |