Health Risk
Assessment/Characterization of the
Drinking Water Disinfection Byproduct
Chloroform
This Document Was Prepared By:
Toxicology Excellence for Risk Assessment
4303 Hamilton Avenue
Cincinnati, OH 45223
Under The Direction Of:
Health and Ecological Criteria Division
Office of Science and Technology
Office of Water
U.S. Environmental Protection Agency
Washington, DC 20460
Under Purchase Order No.
8W-0767-NTLX
November 4,1998
-------
FOREWORD
This document is provided for the consideration of EPA policy makers and does
not embody any decision by EPA policy makers. The purpose of this document is to
provide scientific support and rationale for the hazard identification and dose-response
information pertaining to chronic oral exposure to chloroform. It is not intended to be a
comprehensive treatise on the chemical or toxicology of chloroform. Matters considered
in this risk 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.
An earlier draft of this document underwent external peer review by three
independent experts and experts within EPA. The charge to external peer reviewers and
their comments are presented in the Appendix. Reviewers’ comments were considered in
preparing the version of this document released for public comment. Issues raised by the
public during the comment period have been addressed in the final version of this
document. Specifically, commenters stated that human studies were not considered in the
chloroform assessment, and that risk to fetuses, infants, and children was not taken into
consideration. To address these concerns, an expanded discussion of the human studies
has been added. Concerns about risk to fetuses, infants, arid children have been addressed
by specifically discussing developmental, young animal, and reproductive toxicity of
chloroform; by comparing the systemic toxicity of chloroform in these groups; and by
considering differences between these groups and adults in the ability to activate
chloroform to a toxic metabolite. A discussion of interactions among disinfectant
byproducts and risk assessment for exposure to chloroform as part of a mixture has also
been added in response to public comments. New papers noted by public commenters
have also been included in this document. Additionally the appropriateness of the 80%
RSC assumption was raised in the public comments. EPA is re-evaluating the assumption of
an 80% RSC from ingestion of chloroform in drinking water and considering data which
indicates that exposure to chloroform via inhalation and dermal exposure can potentially
contribute a substantial percentage of the overall exposure to chloroform depending on the
activity patterns of individuals.
2
-------
TABLE OF CONTENTS
FOREWORD
TABLEOFCONTENTS 3
LIST OF TABLES AND FIGURES 5
1. INTRODUCTION 6
2. SUMMARY OF MAJOR CONCLUSIONS IN RISK CHARACTERIZATION 8
2.1 CHARACTERIZATION OF HAZARD 8
2.1.1 T0xIc0KINETICS 8
2.1.2 HEALTH EFFECTS OF ExPosuREs HUMANs 11
2.1.3 HEALTH EFFECTS OF ExPosuRE ni EXPERIMENTAL SYSTEMS 14
2.1.4 CHILDREN’SRISKISSUES 18
2.1.5 GENERAL MECHANISM OF ToxIcITY 22
2.1.6 POTENTIAL INTERACTIONS 26
2.2 CHARACTERIZATION OF DOSE RESPONSE 27
2.2.1 QUANTIFICATION OF NONCARCINOGENIC EFFECTS 27
2.2.2 QUANTIFICATION OF CARCINoGENIC EFFECTS 28
2.3 CHARACTERIZATION OF ExPosuR 32
3. RISK CONCLUSIONS AND COMPARISONS 34
3.1 K x LINES OF EVIDENCE FOR CRITICAL EFFECT 34
3.1.1 OVERALLCONCLUSION 34
3.1.2 STRENGTHS AND WEAKNESSES OF THE EVIDENCE 35
3.1.3 WEIGHT OF EVIDENCE: KEY CONCLUSIONS, ASSUMPTIONS AND DEFAIJLTS 35
3.1.4 SIGNIFICANT IssuEs AND UNCERTAINTIES 36
3.1.5 ALTERNATIvE CONCLUSIONS 37
3.2 LIKELIHOOD OF HUMAN HARM 38
3.3 DOSE RESPONSE ASSESSMENT/CHARACTERIZATION 39
3.3.1 OVERALL CONCLUSIONS 39
3.3.2 STRENGTHS AND WEAKNESSES OF T}{E DATA AVAILABLE FOR ANALYSIS 39
3
-------
333 SELECTION OF STUDY, ISSUES OF ROUTE, FREQUENCY & DuRATION OF
EXPOSURE 40
33.4 STRENGTHS & WEAKNESSES OF THE ASSESSMENT: ISSUES & UNCERTAINTIES 41
3.3.5 BASIS OF ASSUMPTIONS AND DEFAULTS 42
3.3.6 ALTERNATIVE APPROACHES 43
3.4 RISK CHARACTERIZATION SUMMARY 43
4. REFERENCES 47
4
-------
LIST OF TABLES AND FIGURES
TabLe 1. Summary of quantification of toxicological effects for chloroform Page 31
Figure 1. Graphical presentation of data and extrapolations Page 58
5
-------
1. Introduction
Chloroform (CHCI 3 ) is a clear, colorless volati. liquid with a nonirritating odor
and a sweet taste (Hardie, 1964 and Windholz, 1976). Some important physical and
chemical properties of chloroform are summarized in U.S. EPA (1994). Chloroform was
used as an anesthetic as early as 1847 but is no longer employed for this purpose. It is
manufactured and used as a solvent and as an intermediate in the production of
refrigerants, plastics and other solvents (U.S. EPA 1980).
Because of its volatility, chloroform has the potential for evaporation from water
or other sources. Chloroform is stable in water, but light, aeration or the presence of
metals such as iron promote degradation (Hardie, 1964). Chloroform is formed by the
action of hypochiorous acid on endogenous organic molecules (e.g., humic or fulvic
acids) present in the water.
A number of national drinking water surveys performed in the United States
between 1975 and 1981 revealed that chloroform was detectable in a majority of systems
using a surface water source. Average levels usually ranged from 20 to 90 ugfL (U.S.
EPA 1975a, 1975b; Brass et a!. 1977; U.S. EPA 1985). Chloroform was also detectable
in systems using groundwater as a source, but usually at lower levels (ito 10 ug/L). This
is presumably because surface water typically contains higher levels of organic precursors
than groundwater, and generally requires more extensive chlorination than groundwater.
Concentration values have also been noted to vary as a function of season, being higher in
warm weather and lower in cold weather (Wallace et al. 1987, 1988), probably due to
reaction kinetics.
In 1994, U.S. EPA proposed the National Primary Drinking Water Regulations for
Disinfectants and Disinfectant Byproducts (D/DBP). To support the regulation, EPA
developed the following six Criteria Documents: Chlorine; Chlorine Dioxide, Chlorite,
and Chlorate; Chloramine; Trihalomethanes; Chlorinated Acids, Aldehydes, Ketones, and
Alcohols; and Bromate. The Agency is scheduled to publish the Final DIDBP Rule in
November 1998. Since the 1994 proposal, several new studies on DIDBPs have become
available. In order to provide an opportunity for public review and comment on these
new data and how they might influence the proposed rule, EPA (October 10, 1997)
published a text entitled “Summary of New Health Effects Data on Drinking Water
Disinfectants and Disinfectant Byproducts (D/DBPs) for the Notice of Data Availability
(NODA)” (U.S. EPA, 1997a).
6
-------
In 1996. EPA proposed revisions to the 1986 U.S. EPA Guidelines for Carcinogen
Assessment (U S EPA 1996). The proposed Guidelines include a new weight-of-
evidence approach that emphasizes understanding the mode of action, conditions of
expression of carcinogenicity (e.g., route and magnitude of exposure), and consideration
of all other relevant data. The 1996 proposed Guidelines include several default
procedures (linear, nonlinear, or both), rather than relying on the Linear Multi-Stage
(LMS) model as the only default for extrapolation of dose-response relationships.
In 1996, EPA co-sponsored an International Life Sciences Institute (ILSI) project
in which an expert panel was convened and charged with the following objectives:
• review the available database relevant to the carcinogenicity of chloroform and
dichloroacetic acid (DCA), excluding exposure and epidemiology data;
• consider how end points related to the mode of carcinogenic action can be applied in
the hazard and dose-response assessment;
• use guidance provided by the 1996 EPA Proposed Guidelin s for Carcinogen
Assessment to develop recommendations for appropriate approaches for risk
assessment; and
• provide a critique of the risk assessment process and comment on issues encountered
in applying the proposed EPA Guidelines (ILSI, 1997).
The panel was made up of 10 expert scientists from academia, industry, government, and
the private sector. It should be emphasized that the ILSI (1997) report does not represent
a risk assessment per se for chloroform (or DCA) but provides recommendations on how
to proceed with a risk assessment for these two chemicals.
The ILSI (1997) expert panel considered a wide range of information on
chloroform, including rodent tumor data, metabolism/toxicokinetic information,
cytotoxicity, genotoxicity, and cell proliferation data. Based on its analysis of the data,
the panel considered oxidative metabolism to be the predominant pathway of metabolism
for chloroform, and that exposure to chloroform resulted in recurrent or sustained toxicity
as a consequence of oxidative generation of highly tissue reactive and toxic metabolites
[ i.e., phosgene and hydrochloric acid (HC1)], which in turn would lead to regenerative
cell proliferation. The panel considered this mode of action of chloroform as a key
influence on the carcinogenic process, and concluded that chloroform was not acting
through a direct DNA reactive mechanism. The ILSI report noted that the weight-of-
evidence for the mode of action was stronger for the mouse kidney and liver responses,
and more limited, but still supportive, for the rat kidney tumor responses. Thus, the ILSI
(1997) panel viewed chloroform as a likely carcinogen to humans above a certain dose
7
-------
range. but considered it unlikely to be carcinogenic below a certain dose range. indicating
that “This mechanism is expected to involve a dose-response relationship which is
nonlinear and probably exhibits an exposure threshold.” The panel recommended the
margin of exposure approach (U.S. EPA, 1996) as appropriate for quantifying the cancer
risk associated with exposure to chloroform.
The intent of this document is to integrate information from hazard identification,
dose response assessment and exposure assessment, and synthesize an overall conclusion
about the risk of chloroform that is informative and useful for decision makers. This
document relies on secondary sources (e.g., U.S. EPA, 1994; U.S. EPA, 1997a) and
certain key published literature, and considers the recommendations of the ILSI panel
report (ILSI, 1997).
This report applies the principles of the 1996 proposed Guidelines to evaluate the
new science that has emerged for chloroform. This report will strive to be clear,
transparent, reasonable, and consistent with other risk characterizations of the U.S. EPA,
using guidelines established by EPA (U.S. EPA, 1995) and others (Ohanian et al., 1997).
EPA asked 3 external experts to review this report. These comments were considered in
revising this text and are found in the Appendix. The revised document was released for
public comment, and new data and issues raised by the public commenters are addressed
in this final report.
2. Summary of Major Conclusions In Risk Characterization
2.1 Characterization of Hazard
2.1.1 Toxicokinetics
U.S. EPA (1994 and 1997a) and ILSI (1997) summarized data pertinent to the
toxicokinetics of chloroform in part, as follows. Measurements of gastrointestinal
absorption of trihalomethanes such as chloroform in mice, rats and monkeys indicate that
absorption is rapid (peak blood levels at 1 hour) and extensive (64% to 98%). Limited
data indicate that gastrointestinal absorption of chloroform is also rapid and extensive (at
least 90%) in humans. Most studies of trihalomethane absorption have used oil-based
vehicles and gavage dosing. One study in rats found higher chloroform blood levels
following oral gavage administration of chloroform in water than after administration of
chloroform in an oil vehicle. This was interpreted as being due to higher absorption from
water than from oil, but the possible influence of differences in first-pass metabolism was
8
-------
not taken into account Derrnal absorption of chloroform in water by rats and hairless
guinea pigs is rapid and extensive. Absorption has also been reported in humans
dermally exposed to chloroform in water.
Absorbed chloroform appears to distribute widely throughout the body. For
example, chloroform was detected in a number of postmortem tissues from humans, with
highest levels (5 to 68 ug/kg) in body fat and lower levels (1 to 10 ugfkg) in kidney, liver,
and brain. Radiolabeled chloroform was detected in a variety of tissues following oral
dosing of rats and mice, with somewhat higher levels in stomach, liver, blood, and kidney
than in lung, muscle, or brain. Chloroform was also rapidly distributed, after
intraperitoneal injection, to the liver, kidney, and blood of male B6C3F 1 mice. Peak
radioactivity levels in all three tissues were achieved within 10 minutes of dosing, and
returned to background levels within 3 hours. Chloroform crosses the placenta and is
detected in fetal tissues following inhalation exposure of pregnant rats.
Chloroform is extensively metabolized by both humans and animals. The main
site of metabolism is the liver, but metabolism also occurs in the kidney. Recent studies
have investigated the cytochrome P450 (CYP) isoenzymes responsible for trihalomethane
(THM) metabolism. Chloroform is metabolized by both CYP2E1 and CYP2B 1;
however, due to different affinities of these enzymes for chloroform, metabolism by
CYP2E I predominates at low doses. As doses increase, CYP2B 1 begins to metabolize
chloroform (Nakajima, 1995).
Both the oxidative and reductive metabolism of chloroform is mediated by these
cytochrome P-450s. The oxidative pathway requires NADPH and oxygen, whereas the
reductive pathway can utilize NADPH or NADH and is inhibited by oxygen. In the
presence of oxygen (oxidative metabolism), the reaction product is trichioromethanol
(CCL 3 OH), which then decomposes to yield a reactive compound such as phosgene
(CC1 2 O). Phosgene is a reactive species, and may undergo a variety of reactions,
including adduct formation with various cellular nucleophiles (histidine, tyrosine,
methionine, and inositol), hydrolysis to carbon dioxide, or glutathione-dependent
reduction to yield carbon monoxide. Reaction with nucleophiles in proteins or
phospholipids can disable the affected molecules (enzymes, signal molecules, membrane
phospholipids, etc.) which may lead to cytotoxicity or disruption of intracellular
signaling. If oxygen is lacking (reductive metabolism), the metabolic reaction products
appear to be free radical species such as dichloromethyl radical (CHCI 2 .). l’his radical is
extremely reactive forming covalent adducts with a variety of cellular molecules and
causing lipid peroxidation.
9
-------
Although both oxidative and reductive pathways for chloroform metabolism have
been described, the oxidative pathway is more important biologically. Reductive
metabolism of chloroform to dichloromethyl radical appears to be a significant route of
toxicity only at high doses and in induced animals. At high in vitro doses with male rat
kidney microsomes, chloroform metabolism occurs primarily via the reductive pathway
(Gemma et al., 1996a). Phospholipid adducts formed from the reductive metabolism of
chloroform have also been observed in vivo (Gemma et al., 1996b), and the degree of
reductive metabolism was a function of the oxygen tension, the chloroform dose, and the
strain and species of the animal. However, when the relative importance of these
pathways was evaluated by analysis of phospholipid adduct patterns (since the position of
the adducted group on the phospholipid is an indicator of its source) a predominant role
of oxidative metabolism in chloroform activation was seen at a low tissue dose, even at
5% a concentration the authors considered to be representative of that found in the
liver and kidney. At this, oxygen level and a high tissue dose, phospholipid adducts
indicative of both oxidative and reductive metabolism were found in extracts from the
liver and kidneys of B6C3F I mice. At the high tissue dose and 5% P°2’ extracts from the
liver of Sprague-Dawley and Osborne Mendel rats produced adducts indicative of both
oxidative and reductive metabolism, while kidney extracts produced primarily adducts
related to reductive metabolism. It should be noted, however, that this was an in vitro
assay using indirect indicators of oxidative and reductive metabolism. Under conditions
of 20% oxygen partial pressure, less than 25% of phospholipid adducts in livers of
B6C3F 1 mice and the kidneys of DBA2 mice were to fatty acid tails, indicating that
oxidative metabolism predominates at this oxygen level; other studies were conducted
under anoxic conditions, but a hypoxic atmosphere was not investigated (Ade, 1994).
Overall, the pattern of protein and lipid binding in the kidney microsomes correlated with
hormonal status only under aerobic conditions. Therefore, oxidative metabolism of
chloroform is implicated in the observed gender specificity of its kidney toxicity. These
studies indicate that reductive metabolism may occur at low (but physiologic) oxygen
tension in the liver and kidney, but only at high dose levels.
ILSI (1997) noted that the centrilobular region of the liver (the primary region for
chloroform metabolism in the liver) is physiologically hypoxic. That document stated
that oxygen partial pressures in the liver range from I to 60 mm Hg (0.1-8% PO2)’ with a
mean of approximately 20 mm Hg (2.6% PO2) and the lowest values in the centrilobular
region. Gemma et al. (1996a) also considered a P°2 of 5% to be representative of
physiological kidney oxygen levels. However, ILSI (1997) noted that “a large amount of
circumstantial evidence argues against the significance of the anaerobic pathway to
chloroform under normal conditions.” This circumstantial evidence includes the
following: (1) Macromolecular binding (resulting from reductive metabolism) following
10
-------
chlorotorm administration accounts for only a very small portion of the delivered dose.
(2) Compared to other haloalkanes, chloroform is relatively ineffective as a source of free
radicals (3) The closely related compound carbon tetrachioride (Cd 4 ) is a potent
hepatotoxicant and is metabolized almost exclusively by the reductive pathway, but the
available data indicate that carbon tetrachioride carcinogenicity does not result from
direct mutagenicity. (4) Chloroform and carbon tetrachloride differ substantially in the
mechanisms of action related to the necrotic lesion and the kinetics of damage. Based on
these considerations, ILSI (1997) concluded that “free radicals do not play a significant
role in chloroform toxicity or carcinogenicity.”
Both in vivo and in vitro studies indicate that the pattern of trihalomethane
metabolism may differ between animal species and sexes. In vivo, mice have been found
to metabolize trihalomethanes to carbon dioxide more extensively than do rats (40% to
80% versus 4% to 18%). In vitro, the capacity for reductive metabolism of
trthalomethanes has been found to be greater in hepatic microsomes from mice than rats.
In addition, both total metabolism and the formation of covalent adducts in renal
microsomes have been found to be greater in male mice than female mice. These
metabolic differences may explain some of the important toxicological differences that
have been noted between sexes and species.
Excretion of chloroform occurs primarily via the lungs, with greater amounts
excreted unchanged as dose increases and metabolism is saturated. In humans,
approximately 90% of an oral dose of radiolabeled chloroform was exhaled as the end
metabolite, carbon dioxide, or as the parent compound, chloroform. Levels in the urine
were below the limit of detection (0.1%). In mice and rats, 45% to 88% of an oral dose of
chloroform was excreted from the lungs either as chloroform or as carbon dioxide, with
1% to 5% excreted in the urine. Intraperitoneal injection of rats with 36 C1-chloroform
resulted in the appearance of both inorganic and organic forms of chloride in the urine,
but the total amount was not quantified.
No data were located regarding the bioaccumulation and retention of chloroform
following chronic exposure. However, based on the rapid metaliolism and excretion of
chloroform, along with the low levels of chloroform in human autopsy samples, marked
accumulation and retention is not anticipated.
2.1.2 Health Effects of Exposures in Humans
EPA (1994 and 1997a) also summarized the health effects of exposures in humans
as follows. In a case study of a young man who ingested 4 ounces of chloroform (a dose
11
-------
of aboul 2.500 mg’kg), prominent clinical findings included jaundice, an enlarged liver.
increased serum levels of bilirubin, alkaline phosphatase (AP), and serum glutamic
oxaloacetic transferase (SGOT), along with albuminuria, glucosuria, ketonuria and the
presence of red cells and granular casts in the urine. These observations indicated that in
humans, as in other animals, the liver and kidneys are the organs most affected by acute
chloroform ingestion.
Workers exposed to chloroform by inhalation at levels of 112 to 1,158 mg/rn 3 for 1
or more years complained of nausea, lassitude, dry mouth, flatulence, thirst, depression,
irritability, and “scalding” urination, but clinical examination and tests of liver function
(serum enzyme levels) failed to detect any abnormalities. Inhalation exposure of workers
to chloroform at levels of about 10 to 1,000 mg/rn 3 for Ito 4 years was reported to be
associated with an increased incidence of viral hepatitis and enlarged liver.
Although a number of studies have investigated the potential association between
chlorinated drinking water and cancer or developmental effects, only a few of these
evaluated the degree of association with chloroform exposure. Some epidemiological
studies suggest there may be an association between ingesting chlorinated water and
increased cancer mortality rates. However, because chlorinated drinking water contains a
number of disinfectant byproducts, and, depending on the source, may contain other
contaminants, it is usually not possible to associate any increased cancer rate with a
specific contaminant. Other limitations common to most or all of the studies include (1)
the potential for chloroform to serve as a surrogate measure of other contaminants that
might be truly responsible for the observation in the study, (2) insufficient historical
exposure information, (3) potential misclassification of exposure due to use of county-
wide or region-wide exposure information, and (4) insufficient accounting for migration.
In the most thorough evaluation of the potential link between chloroform in
drinking water and cancer, Doyle et al. (1997) reported an association between
chloroform ingestion in drinking water and the development of colon cancer in a study of
28,237 postmenopausal Iowa women; there was no association with the other cancer
types evaluated. (Note that the study population was very specific and was not
representative of the population of colon cancer cases in the U.S.) In this cohort study,
women aged 5 5-69 completed a mail survey in 1986 on medical history, anthropomorphic
data, and risk factors. The cohort members were followed for cancer incidence through
the state health registry, and for cancer mortality through the National Death Index and
questionnaires mailed in 1987, 1989, and 1992. Based on the follow-up surveys, out-
migration was estimated at less than 1% annually. Drinking water source was determined
in the 1989 mail survey. Only women drinking municipal water or private well water for
12
-------
more than 10 years ere included in this study. Exposure to four trihalomethanes
(chloroform, bromodichioromethane, dibromochioromethane, and bromo form) was based
on state-wide surveys in 1979 and 1986 of municipal water supplies, and exposure levels
were linked to cohort members on a community basis. Cohort members for which there
was no trihalomethane exposure information, who reported a change in residence, or who
reported a prior diagnosis of cancer (other than skin cancer) were excluded from the
study. All relative risks were adjusted for such potential confounders as smoking status,
fruit and vegetable intake, and age.
The authors focused on the relationship between chloroform exposure and cancer,
since chloroform was the most commonly occurring trihalomethane, had the broadest
concentration range, and correlates well with the concentration of the other
trihalomethanes. Private well users were excluded from the analysis, since
trihalomethane levels were not available, bringing the size of the cohort analyzed to
19,199. For the 1986 exposure data, there was a dose-related increase in colon cancer
with increasing chloroform levels, with relative risks of 1.06, 1.39, and 1.68 at 1-2 ug/L,
3-13 ugfL, and 14-287 ugIL, respectively (trend test p
-------
controlled for source type. population density, marital status, age, and year of death
(Lavvrence et a!, 1984). In a study of 400 residents of Telemark, Norway, age 50-59
years, the prevalence of colorectal polyps was not associated with the concentration of
chloroform in the drinking water (Hoff et al., 1992).
A few studies have also investigated the potential association between chloroform
in drinking water and developmental effects. As discussed in U.S. EPA (1994), Kramer
et al. (1992) conducted a population-based case-control analysis in Iowa to determine if
exposure to trihalomethanes was associated with low birthweight, prematurity, or
intrauterine growth retardation. After adjusting for maternal age, number of previous
children, marital status, education, adequacy of prenatal care, and maternal smoking, the
authors found a statistically significant association between exposure to water chloroform
levels of at least 10 ugIL and intrauterine growth retardation. The association remained
when the only water source was deep wells, a source unlikely to be contaminated with
pesticides and other chemicals, aside from disinfection byproducts. The study authors
noted that chloroform may have been a marker for other organic halides. A recent
epidemiological study found an association between spontaneous abortions and drinking
water levels of total trihalomethanes (TI-IMs) or of the TI-IM bromodichioromethane
(BDCM) (Waller et al., 1997). No association with chloroform levels was found.
No human data were located regarding whether chloroform exposure may result in
greater risk for any human subpopulation than for the general population. However,
reasonable predictions of sensitive populations can be made based on animal data. Based
on the induction ofCYP2E1 by ethanol, alcohol consumption may increase toxicity in the
liver. Based on potentiation of chloroform-induced hepatotoxicity in a rat model of
diabetes, diabetics may be also more sensitive to the liver effects. People with pre-
existing kidney or liver damage may also be more sensitive, since they would be expected
to have a lower functional reserve capacity. On the other hand, decreased liver function
might result in decreased capacity for chloroform metabolism, which would result in
decreased sensitivity.
Based on a detailed consideration of the animal data on the systemic and
developmental toxicity of chlorofonn, and of the metabolism of chloroform at different
ages, fetuses, infants, and children are not considered sensitive subpopulations for
chloroform toxicity. The supporting data are addressed in Section 2.1.3 in the context of
the animal data.
14
-------
2.1.3 Health Effects of Exposure in Experimental Systems
Data from animal studies also are summarized in EPA (1994 and 1997a) as
follows. Large oral doses of chloroform are lethal to laboratory animals, with acute LD 50
values ranging from 119 to 2,000 mg/kg. Death from acute high-dose chloroform
exposure was usually found to be due to central nervous system depression and cardiac
effects, and was usually accompanied by histopathological changes in the liver and
kidney.
Acute oral exposure to sublethal doses of chloroform can also produce effects on
the liver, kidney, and central nervous system. In mice, single oral doses of 60 to
89 mg/kg produced kidney damage, with doses of 140 to 250 mg/kg producing liver
damage. Organ damage was characterized by fatty infiltration, cellular necrosis,
vacuolization, enzyme level changes, and/or organ weight changes. Ataxia and sedation
were noted in mice receiving 500 mg/kg chloroform.
Short-term exposures of laboratory animals to chloroform have been observed to
cause effects on the liver, kidney, central nervous system, and immune system. Hepatic
effects, including organ weight changes, elevated serum enzyme levels, and
histopathological changes were reported in mice and/or rats administered 37 to
290 mg/kg-day chloroform. Kidney effects, characterized by decreased para
aminohippurate (PAH) uptake, histopathological changes and organ weight changes, were
reported in mice and/or rats administered 37 to 148 mg/kg-day chloroform.
The predominant effects of longer-term oral exposure to chloroform occur in the
liver and kidney. The effects on these two organs are similar to those described for short-
term exposures. Hepatic effects were reported in mice, rats and dogs administered 15 to
180 mg/kg-day chloroform. In general, these dose ranges are slightly lower than those
reported to cause effects following short-term exposures.
Data concerning the developmental effects of chloroform indicate toxicity to the
mother and fetus at high doses and suggest that reproductive and developmental toxicity
may occur as well. Signs of maternal toxicity (decreased body weight and changes in
organ weight) were reported in rats, rabbits and/or mice administered 50 to 100 mg/kg-
day chloroform. Fetotoxicity, as indicated by decreased fetal body weights, was evident
in the offspring of rats administered 121 to 400 mg/kg-day chloroform. Delayed
ossification and sternebral aberrations have been reported in rats and/or rabbits
administered 20 to 200 mg/kg-day chloroform. Statistically significant malformations
and variations (cleft plate, imperforate anus, acaudia, delayed ossification) have been
15
-------
obser ed in inhalation studies in which mice and/or rats were exposed to 30 or 100 ppm
chloroform (either 147 or 488 mg/rn 3 , respectively).
The overall evidence regarding chloroform genotoxicity is mostly negative. In
vitro, chloroform has yielded mixed but mainly negative results in a number of assays of
mutagenic activity, including 30 of 34 point mutation assays. Some of these results,
however, are inconclusive because of inadequacies in experimental protocols, especially
in the failure to use an appropriate (reconstituted) activation system or to take precautions
to prevent the escape of volatilized chloroform (U.S. EPA 1985; Rosenthal, 1987). In
vitro assays for early evidence of DNA damage (sister chromatid exchanges or DNA
damage in yeast) tend to give positive results, but these endpoints may not be indicative
of DNA alkylation and mutation. For example, the sister chromatid exchange assay has a
low specificity for predicting carcinogenesis (i.e, a high rate of false positives compared
to results of the rodent cancer bioassay). Thus, the weight of the evidence indicates that
chloroform is not a DNA-reactive mutagen.
Recent studies also found that chloroform was negative in bacterial mutagenesis
assays (Roldan-Arjona and Pueyo, 1993; LeCurieux et at., 1995), although a positive
response was seen at very high doses with a strain engineered to produce endogenous
glutathione S-transferase (Pegram et al., 1997). Chloroform was negative in in vivo and
in vitro UDS assays (Larson Ct al., 1994d). Chloroform was reported as positive in the
presence of S9 in an SCE assay in a rat leukemia cell line (Fujie et at., 1993). Positive
results were obtained in a mouse micronucleus test (Shelby and Witt, 1995). Butterworth
et al. (1998) found no statistically significant increases in mutant frequency in lad
transgenic B6C3F 1 mice exposed to chloroform via inhalation, although the chloroform-
exposed groups exhibited a consistent, dose-related increase over controls. This system
has the advantage of ease in identification of in vivo mutations. Exposure was to a
nonhepatotoxic (10 ppm), mildly hepatotoxic (30 ppm), or overtly hepatotoxic (90 ppm)
concentration, and mutant frequencies were evaluated after 10, 30, 90, or 180 days of
exposure.
Chloroform also causes the development of both benign and malignant tumors in
animals. The following summary derives in part from ILSI (1997). Several bioassays
have been performed that sought to characterize the potential for liver carcinogenicity in
mice. When given by corn oil gavage, chloroform-provided a positive carcinogenic
response in the male and female mouse liver (NCI, 1976).’ Time-weighted average doses
As described in EPA (1998), in a gavage bioassay (NCI, 1976), Osborne-Mendel rats and B6C3FI mice
were treated with chloroform in corn oil 5 times/week for 78 weeks. Fifty male rats received 90 or 125
16
-------
in this studs’ ere 0. 138, and 277 mg/kg-day for males and 0, 238, and 477 mg/kg-day
for females Corresponding incidences of liver tumors were 1/18 (6%), 18/50(36%), and
44/45 (98%) for males, and 0/20 (0%), 36/45 (80%), and 39/41(95%) in females. Mice
dosed by drinking water (Jorgenson et al., 1985)2 or by inhalation (Matsushima, 1994),
however, failed to exhibit a liver tumor response even though the doses were similar to
those of the corn oil gavage study, From these studies, it appears that neither the daily
dose, nor the cumulative dose of chloroform, was predictive of tumor outcome in the
drinking water study in mice. The inhalation study was negative in the liver despite a
concentration escalation strategy used to achieve final concentrations that exceeded
acutely lethal concentrations by several fold.
As also discussed in ILSI (1997), several bioassays have been performed that
sought to characterize the potential for liver carcinogenicity in rats. The only study in
which an increased incidence of liver tumors was observed was conducted with
mg/kg-day; females initially were treated with 125 or 250 mg/kg-day for 22 weeks and 90 or 180 mg/kg-
day thereafter. Male mice received 100 or 200, raised to 150 or 300 mg/kg-day at 18 weeks, females
were dosed with 200 or 400, raised to 250 or 500 mg/kg-day. A significant increase in kidney epithelial
tumors was observed in male rats and highly significant increases in hepatocellular carcinomas in mice
of both sexes. Liver nodular hyperplasia was observed in low-dose male mice not developing
hepatocellular carcinoma.
2 As described in EPA (1998), Jorgenson et al. (1985) administered chloroform (pesticide quality and
distilled) in drinking water to male Osborne-Mendel rats and female B6C3FI mice at concentrations of
0, 200, 400, 900, and 1800 mgfL for 104 weeks. These concentrations were reported by the author to
correspond to doses of 0, 19, 38, 81, and 160 mg/kg-day for rats and 0, 34, 65, 130, and 263 mg/kg-day
for mice. A significant increase in renal tumors in rats was observed in the highest dose group. The
increase was dose related. The liver tumor incidence in female mice was not significantly increased.
This study was specifically designed to measure the effects of low doses of chloroform in drinking water.
This recent, and as yet unpublished, work of Matsushima( 1994) states that groups of 50 male and 50
female F344 rats were exposed to chloroform vapor, 6 hours per day, 5 days per week, for 104 weeks.
Exposure concentrations were 0, 10.30 or 90 ppm. Groups of 50 male and 50 female BDFI mice were
exposed on the same schedule. The final exposure concentrations for the mice were 0, 5, 30, or 90 ppm.
Preliminary studies showed that BDFI mice (especially males) were initially sensitive to the toxic effects
of chloroform, but that exposed animals later developed resistance to the chemical. Therefore, animals
in the 30 ppm groups (both males and females) were exposed in a series of increasing concentrations, as
follows: 5 ppm for 2 weeks, 10 ppm for 2 weeks, and 30 ppm for 100 weeks. Similarly, animals in the
90 ppm groups were exposed to 5 ppm for 2 weeks, 10 ppm for 2 weeks, 30 ppm for 2 weeks, and 90
ppm for 98 weeks. A preliminary report suggests treatment related kidney neoplasia only in male mice,
with a small but statistically significant increase in liver neoplasia in females if the incidences of
carcinomas and adenomas were combined.
17
-------
chloroform administered in drinking water, and the tumors were found only in female
Wistar rats, not in males (Tumasonis et a!., 1987). The interpretation of these results is
complicated by the small size of the control group and the longer survival of the treated
(185 weeks) versus control females (145 weeks). Other drinking water, corn oil gavage,
and inhalation studies in various rat strains failed to demonstrate an increased incidence
of liver neoplasia
As also discussed by ILSI (1997), chloroform exposure produced positive kidney
tumor responses in both male rats and mice, but not in females of either species. For
example, a significant increase in renal tumors in male BDF 1 mice was seen following
inhalation exposures (Matsushima, 1994) and in male ICI mice exposed to chloroform
either in an arachis oil base or in toothpaste (Roe et al., 1979). Although renal tumors
were not found in male B6C3F 1 mice following chloroform exposure by corn oil gavage
(Nd, 1976) or drinking water (Jorgenson et al., 1985), positive responses for renal
tumors were found in male Osbome-Mendel rats. The incidences of all kidney tumors in
this strain of rats was 1/50 (2%), 6/3 13 (2%), 7/148 (5%), 3/48 (6%), and 7/50 (14%) for
exposures of 0, 200, 400, 900, and 1800 ppm of water, respectively. F344 and Sprague-
Dawley rats did not, however, show this tumor response. Thus, responses in kidney
tumors appear to vary with route of exposure, administration vehicle, and strain of rodent.
This makes it difficult to develop conclusive statements regarding potential strain
differences in tumor response. Nonetheless, it is clear that when a response is observed,
males are more responsive than females. This sex-specific difference is consistent with
the observation that kidney metabolism of chloroform is higher in males than in females.
Male F344 rats were exposed to chloroform in drinking water for 100 weeks at
concentrations of 0, 900, or 1,800 ppm as described in a recent, and as yet unpublished,
work (DeAngelo et al., 1995). Interim sacrifices of groups of 6 animals were performed
at 26, 52 and 78 weeks, and groups of 50 animals were scheduled for the 100 week
sacrifice. At each time point, the liver and kidney were examined for gross and
microscopic lesions. In the liver, with the exception of midzonal vacuolization (probably
due to fat accumulation), no lesions other than those normally associated with aging rats
were reported at any of the sacrifice periods. A preliminary report suggested an increase
in Liver tumors at the 1800 ppm dose. Kidney tumors were not found at either dose.
2.1.4 Children’s Risk Issues
Three questions were considered in evaluating whether fetuses and children are
more sensitive than adults to chloroform exposure:
18
-------
• Are fetuses and children more susceptible than adults to the systemic toxicity of
chloroform?
• Does chloroform causes reproductive and deve pmental effects at doses below
those causing systemic toxicity (including cancer)?
• How does the ability of fetuses and children to metabolize chloroform to a toxic
metabolite compare with that of adults?
Addressing the first question, whether fetuses and children are more sensitive than
adults to the hepatotoxic effects of chloroform, is complicated by the absence of studies
in which systemic effects in young rats were completely evaluated. An ideal study to
examine this question would include exposure in utero, during lactation, and between
weaning and reproductive age, with histopathological evaluation shortly after weaning
and before reproductive age. Such a study design would eliminate the potential for repair
during adulthood of any damage sustained by-the juvenile, and would detect systemic
effects of chloroform on the developing organism. (Teratogenic effects of chloroform are
detected by developmental studies, as discussed with regard to the second question.)
However, in the absence of such a study for chloroform, some comparisons can be made
based on the available studies. in particular, direct comparisons can be made between
two mouse studies in which chloroform was administered by corn oil gavage. In the first
study (NT?, 1988), CD-i (ICR)BR mice were exposed to chloroform in utero, during
lactation, and then by gavage as young rats through “young” adulthood. Mild to
moderate liver histopathology (degeneration of centrolobular hepatocytes, accompanied
by occasional single cell necrosis) was observed in females at 41 mg/kg-day, the only
dose at which systemic effects were evaluated. No adverse effects on fertility or
reproduction of the F 1 generation were observed. Although there was no significant
effect on sperm quality (sperm motility, density, and percent abnormal sperm) of the F 1
males, vacuolar degeneration of ductal epithelium in the cauda epididymidis was
observed at 41 mg/kg-day. Thus, the only dose tested in this study, 41 mg/kg-day, was a
LOAEL for liver histopathology.
In the second study (Bull et a!., 1986), 6-8 week old B6C3F1 mice were exposed
to 60, 130, or 270 mg/kg-day for 90 days by gavage in either corn oil or Emulphor (an
emulsifier). This discussion focuses on the corn oil gavage results, to facilitate direct
comparison to the NTP (1988) study. Liver histopathology (extensive vacuolation of the
liver accompanied by lipid accumulation) was observed at a dose of 60 mg/kg-day after
chloroform was administered by corn oil gavage, but higher corn oil gavage doses had
corresponding less accumulation of lipid. A 10% decrease in body weight and a 20%
(females) to 30% (males) increase in relative liver weight were also observed at the low
dose, with larger effects at the higher doses. Based on these effects, the study LOAEL
19
-------
could be considered to be the low dose, 60 mg/kg-day. (This LOAEL differs from that in
the Trihalomethanes Criteria Document [ U.S. EPA, 1994]. That document considered
130 mg/kg-day to be a NOAEL, apparently because of the differences in toxicity evoked
by chloroform when administered by corn oil or Emuiphor gavage, and the related
difficulty in making judgments regarding the critical effect.)
The similarity of effects at comparable corn oil gavage doses in the 2-generation
study (exposure of fetuses to young adults) and the 90-day study (exposure of young
adults to adults) indicates that there is no substantial additional sensitivity attributable to
exposure of the former group. However, two factors limit the strength of this conclusion:
(1) Different strains were evaluated; and (2) Neither study identified a NOAEL. On the
other hand, exposure of the generation in the NTP study was somewhat longer than the
90-day exposure in the subchronic study, strengthening the conclusion that fetuses and
children are not more sensitive.
The second question, whether reproductive and developmental effects occur at
doses below those causing systemic toxicity, can be evaluated by directly comparing
effect levels for reproductive/developmental effects and for systemic toxicity. No effects
of chloroform on reproductive function have been identified (NTP, 1988). Oral
developmental toxicity studies have found decreased fetal weight (Thompson et al., 1974)
and inhalation developmental studies have found an increased incidence of delayed
ossification (Baeder and Hofmann, 1991), but these effects occurred at doses above those
causing hepatotoxicity. Thus these effects do not constitute the critical effect, and so the
RID based on liver effects would be sufficiently protective. Based on an increased
incidence of fetuses with incompletely ossified skull bones in rabbits (Thompson et al.,
1974), the NOAEL for developmental toxicity is 35-50 mg/kg-day. It should be noted,
however, that a definitive NOAEL could not be identified in that study, due to the
absence of a clear dose-response. The developmental NOAEL is at least three times
higher than the LOAEL of 12.9 mg/kg-day for hepatotoxicity in dogs that forms the basis
for the RfD. Therefore, the RID based on hepatotoxicity in dogs is also adequate for
protection from developmental effects.
The third question concerns the relative ability of fetuses, children, and adults to
metabolize chloroform to a toxic metabolite. Because the toxicity of chloroform is
dependent on oxidative metabolism, primarily by cytochrome P450 CYP2E 1, studies on
CYP2E 1 levels in fetal and adult tissues were evaluated to determine whether fetuses
would be expected to be more sensitive than adults to the effects of chloroform. The
status of CYP2E 1 in fetuses remains unclear, with conflicting studies. Most of the
existing studies indicate that this enzyme is expressed in human adults but not in human
20
-------
fetuses, e\ en when measured using sensitive assays (reviewed in Hakkola eta!., 1998)
In these studies, !eve!s of both CYP2E1 protein and of the associated enzyme activity
were undetectable before birth, but rose rapidly shortly after birth, due to stabilization of
the C’{P2E1 protein. However, at least three studies indicate CYP2E1 is expressed in
fetal liver or cephalic tissue (Boutelet-Bochan et a!., 1997; Carpenter et al., 1996; Vieira
et a!., 1996). Boutelet-Bochan et al. (1997) detected low levels of Cyp2eI mRNA
transcription in human fetal brains (gestation days 52-117, or 7-17 weeks), and levels
tended to increase with gestational age. However, transcription was detected only using a
very sensitive assay (reverse transcriptase-polymerase chain reaction - RT-PCR) or the
moderately sensitive RNase protection assay. Transcription in fetal liver was much
lower, and was detectable in only two of six samples. Also using the RNase moderately
sensitive technique, Carpenter et al. (1996) found transcription of Cyp2el mRNA in the
liver of human fetuses at 19-24 weeks gestation, but not at 10 weeks gestation. Fetal liver
microsomes could metabolize the CYP2EI substrate ethanol, but at a rate only 12-27% of
adult liver microsomes. At least most of the observed activity was specific to CYP2E1,
since it was inhibited by an anti-CYP2E1 antibody. Like adult hepatocytes, fetal
hepatocytes exposed to ethanol had induced levels of CYP2E 1. Vieira et al. (1996) found
that CYP2EI protein could not be detected immunochemically in fetal human liver, and
there was only minimal evidence of Cyp2EJ mRNA or CYP2E 1 activity in fetal liver
microsomes. (The difference in assay results may be due to differences in sensitivity, or
to cross-reaction ofCYP1A1 activity.) The authors found, however, that CYP2E1
protein levels rise rapidly in the first few hours after birth, with a slow increase in protein
levels and in CYP2E1 RNA levels during childhood.
Thus, the overall human data show that if CYP2E 1 activity exists in human
fetuses, levels are much lower than those in adults. Regardless of fetal CYP2E 1
expression, the enzyme is rapidly induced upon birth. For this reason, children would be
expected to be capable of chloroform metabolism, although the amount of CYP2E1 may
be less than that present in the adult. Overall, the data on CYP2E1 activity provide no
evidence to suggest that children are more susceptible than adults.
The animal studies of developmental CYP2E 1 regulation provide uniform
evidence of the rapid induction of this gene soon after birth (Song et at., 1986; Umeno et
al., 1988; Schenkman et at., 1989; Ueno and Gonzalez, 1990). The idea that the enzyme
activity peaks before weaning with a gradual decrease to adult levels suggested by some
scientists, however, has not been consistently reported in the three studies which
compared expression over this period of time.
21
-------
For example. Schenkman et al. (1989) indicate that CYP2E1 protein is present in
low levels in neonates, rises to a peak level at age 2 weeks, and subsequently decreases to
adult levels by puberty. Analysis of protein levels quantified from western blots showed a
maximum at 2 weeks with decreasing levels at 4 and 12 weeks. The protein level at 12
weeks was approximately 50% of the level at 2 weeks. The authors did not provide a
statistical analysis of this result, but it appears from the error bars that the 2-week and 12-
week levels (but not 4 weeks levels) were significantly different.
Song eta!. (1986) conducted a similar analysis and reported a rapid transcriptional
induction of Cyp2el (P450j) within 1 week following birth which remained elevated
throughout 12 weeks. The authors did not quantitate the western blots, but visual
inspection indicates a small decline in protein levels by 12 weeks. However, in this same
study, enzyme activity gradually increased over time, reaching a maximum at adulthood.
Nor was an age-dependent decrease in mRNA levels observed.
Ueno and Gonzalez (1990) showed that extracts from 3 day old and 12 week old
rat liver, but not fetal or newborn rat liver were able to generate significant Cyp2el
transcription in vitro. The ability of the extract to drive transcription of Cyp2eI was
slightly greater at 12 weeks.
If the two-fold increase in Cyp2eI induction in animals where verified, its
importance in terms of chloroform toxicity would depend on the dose. Under low dose
conditions (for example, much lower than the Km) it is possible that an increase in the
level of enzyme would not have any effect on active metabolite formation since the
amount of chloroform, and not CYP2EI, would control the rate of the enzyme activity.
On the other hand, under saturating doses of chloroform, all the available enzyme would
be active, thus a two-fold increase in CYP2E1 could result in greater activation of the
compound. Additional analysis of the expected dose relative to the levels of enzyme
could help elucidate the potential for these differing scenarios to occur.
Taken together, these animal studies do not provide conclusive evidence of an
early period of increased enzymatic activity in young animals when compared with
adults. While the animal data remain unclear regarding the potential for a period of
increased CYP2EI activity above that in the adult, for humans, a gradual increase of
CYP2E 1 activity throughout out childhood with a maximum level at adulthood, as
described by Hakkola et al. (1998).
2.1.5 General Mechanism of Toxicity
22
-------
U S. EPA (1994 and 1997a) and ILSI (1997) summarized information on the
potential mechanism of toxicity for chloroform, as follows.
U.S. EPA (1994 and 1997a) summarized three lines of evidence that indicate that
chloroform metabolism is essential for toxicity: (1) the tissues that most actively
metabolize chloroform (liver, kidney) are also the chief target tissues; (2) chemical
treatments that increase or decrease metabolism also tend to increase or decrease toxicity
in parallel; and (3) species- and sex-related differences in metabolism are paralleled by
similar differences in toxicity. The detailed biochemical mechanism by which
chloroform metabolism leads to toxicity are not certain, but covalent binding of reactive
metabolites to cellular macromolecules is very likely a component. Such metabolites are
produced by oxidative metabolism to dthalocarbonyls and perhaps by reductive
metabolism to free radicals. Free radical production may also lead to cell injury by
inducing lipid peroxidation in cellular membranes.
ILSI (1997) also used information on strain-specific differences in chloroform
metabolism to explain apparently contradictory bioassay data. As noted in Section 2.1.3,
renal tumors were observed in male BDF 1 mice exposed to chloroform via inhalation
(Matsushima, 1994) and in male ICI mice exposed to chloroform in a toothpaste or
arachis oil vehicle (Roe et al., 1979), but not in male B6C3F1 mice administered
chloroform by corn oil gavage (NC!, 1976) or in drinking water (Jorgenson et al., 1985).
ILSI (1997) noted that males of the DBA strain, from which BDF1 mice were derived,
have higher tubular levels of the enzymes that bioactivate chloroform, and are much more
susceptible to chloroform-induced renal damage, than male C57BL mice, the parental
strain for B6C3F 1 mice. Thus, the observed data for kidney tumors in mice are consistent
with the hypothesis that the chloroform is metabolized to a cytotoxic compound, and that
the resulting toxicity and cell proliferation can result in the development of chloroform-
induced cancer.
Formation of DNA adducts has not been shown with chloroform exposure.
Although the formation of DNA adducts is the traditional hypothesis of tumor formation,
the weight of evidence favors the hypothesis that carcinogenesis may be related to
increased cell proliferation following direct tissue injury. This is the hypothesis
suggested by ELSI (1997) and Golden et al. (1997). However, this latter hypothesis has
not been definitively linked to chloroform carcinogenesis.
Several chemicals, including various ketones, dichioroacetic acid, and carbon
tetrachloride, potentiate the toxic effects of chloroform. The mechanism(s) of the
potentiation by ketones is not known, but appears to include a process other than
23
-------
rnduction of microsomal enzymes. The vehicle (corn oil versus aqueous) used for oral
dosing also affects toxicity, with toxicity generally being more severe followtng
administration in corn oil. The difference in toxicity may also be due to a difference in
dose rate of a bolus dose following the gavage (in oil) versus the more extended nature of
exposure in drinking water (several hours).
Larson and coworkers have conducted a series of studies (Larson et a!. 1993,
1994a, 1994b, 1994c, 1995a, 1995b, 1996;Templinetal. 1996a, 1996b, 1998)
investigating the relationship among chloroform exposure, cytotoxicity, regenerative cell
proliferation, and histopathology in rats and mice. Effects of oral exposure were
investigated in a series of short-term studies of 1 to 21 days; inhalation exposures were
conducted for 4 to 90 days. Based on earlier experiments, the study authors assumed that
all increases in cell division, as measured by the labeling index, were due to regenerative
cell proliferation, and thus an indirect measure of cytotoxicity. The labeling index is
defined as the percentage of cells in a tissue that are in S-phase (that portion of the cell
cycle when DNA synthesis occurs in preparation for cell division) within a specified time
interval. The value of the labeling index is usually low (<4%) in tissues such as liver and
kidney, but is increased by chemical treatments that result in cytolethality and
regenerative cell proliferation. - -
Key results from this set of studies are that chloroform administered by gavage in
corn oil generally resulted in increased labeling index in the liver and kidney of male and
female B6C3F 1 mice and F344 rats. The levels of the increases varied depending on the
sex, species and strain of rodent used, as well as the duration of exposure. There was,
however, a tendency for the degree of labeling to decline with the duration of dosing. An
increased labeling index was generally not observed in the liver or kidney of female
B6C3F 1 mice or male F344 rats administered chloroform in drinking water. (Male mice
and female rats were not tested using drinlcing water administration.) Following
inhalation exposure to chloroform, a significantly increased labeling index was observed
in the liver of male and female B6C3F I mice and F344 rats after exposure for as long as
90 days. The kidney labeling index was also significantly increased in male and female
B6C3F I mice, and F344 rats. Recent studies investigated cell proliferation in additional
strains used in chloroform bioassays. In male Osborne-Mendel rats (the strain arid sex
that form the basis for the chloroform cancer value derived in U.S. EPA, 1994, and in this
document) administered a single gavage dose of chloroform, there was no increase in the
liver labeling index, but there was a small increase in labeling index in the kidney; longer-
term cell proliferation assays have not been conducted in this strain. In a 90-day cell
proliferation assay conducted with male and female BDF 1 mice [ the strain used in the
inhalation bioassay of Matsushima (1994)], a significantly increased labeling index was
24
-------
obsen ed in the kidneys of males, but not females. A smaller increase in the labeling
index was observed in the livers of males and females at 7 weeks, but the increase in
males was not sustained through the end of the study.
Pereira (1994) reported that a small increase in the labeling index persisted for 159
days in the liver of female rats administered chloroform by corn oil gavage, with a LOEL
of 263 mg/kg-day. This study observed no increase in labeling index in female rats
administered chloroform in thinking water at doses up to 363 mg/kg-day for 5 to 159
days.
Chiu et al. (1996) compared the available data on short-term, subacute, and long-
term cytotoxicity with the tumor incidences in different strains and species of the exposed
rodents, as reported by published studies. Special attention was given to the subchronic
(Jorgenson and Rushbrook, 1980) and chronic data of the Jorgenson Ct a!. (1985) drinking
water study, because it was the principal study used by EPA (1994) to derive the oral
cancer risk estimate for chloroform. There were no treatment-related biochemical or
microscopic/gross histopathological changes in the kidney of the rats fter 30, 60, or 90
days of exposure to chloroform in drinking water. Chloroform carcinogenicity appeared
to be associated with cytotoxicity in some cases, but not in others, depending on the strain
and species of rodent tested. Data revealed that chloroform exposure, either by drinking
water or by corn oil gavage, induced kidney cancer in male Osborne-Mendel rats without
accompanying cytotoxicity at necropsy in the 2-year bioassays. In contrast, ILSI (1997)
reported that, based on a rereading of the available slides from the male rats in the
Jorgenson drinking water study, all high-dose rats exposed for 2 year had evidence of low
grade chronic renal tubule injury and regeneration, where the tissues were not
compromised by autolysis or diffuse di eases. Similar or minor lesions were observed in
high-dose rats exposed for 18 or 12 months, and at 900 ppm, but not at 400 ppm. Chronic
nephropathy was observed in all dose groups, including the controls, but the pathologist
was able to distinguish the chronic nephropathy from evidence of cytotoxicity and
regeneration.
Fox et a!. (1990) used a mutational analysis of the H-ras oncogene in liver tumors
in male C57BL/6 x C3HII-IE mice (i.e., B6C3F1 mice) to address the issue of the
mechanism of chloroform carcinogenicity. The spontaneous incidence of liver tumors in
males of this strain is high (20-30%). Mutations activating the oncogene were found in
64% of the spontaneous tumors analyzed. By contrast, only 5/24 (21%) of the liver
tumors from mice treated with chloroform (200 mg/kg by gavage in corn oil, twice
weekly for 1 year) had activated H-ras genes. About 59% of the liver tumors found in
mice treated with a known mutagenic carcinogen (benzidine hydrochloride) had an
25
-------
activated H-ras gene Because the number of tumors with activated H-ras genes in the
chloroform-treated group was not comparable to that expected spontaneously, the authors
suggested that the observed tumors with an H-ras mut ion occurred spontaneously, and
the chloroform-induced liver tumors occurred via a different mechanism.
Sprankle et al. (1996) examined the levels of gene expression of various
oncogenes in the liver of female B6C3F1 mice and kidneys of male F344 rats from the
study of Larson et a!. (1993). Mice received a single oral gavage dose of 350 mg/kg
chloroform in corn oil, and rats received a single oral gavage dose of 180 mg/kg
chloroform in corn oil. There were transient increases in mRNA levels for the myc and
fos growth control genes in the female mouse liver, but levels of Ha-ras, met, and
hepatocyte growth factor mRNA were comparable to control levels. In the male rat
kidney, there was a transient increase in levels of myc mRNA, but no strong effects on the
other growth control genes examined. The study authors stated that other cytotoxic
carcinogens induce a similar pattern of gene expression, and concluded that the changes
in myc andfos expression could play a role in chloroform-induced regenerative cell
proliferation.
Vorce and Goodman (1991) examined the methylation state of ras oncogenes in
chloroform-induced liver tumors in male B6C3F 1 mice. The mice were dosed by gavage
twice weekly for 1 year with 200 mg/kg chloroform in corn oil. Liver tumors were
reported in 80% of the animals, but the size of the treated group was not reported. The
study authors found that Ha-ras was hypomethylated in all chemical-induced and
spontaneous liver tumors examined, and sporadic hypomethylation of Ki-ras was
observed in the spontaneous liver tumors, as well as chloroform-induced tumors. There
was no effect on the methylation of myc, but there was some evidence of myc gene
amplification. Because hypomethylation appears to be necessary, although not sufficient,
for transcription, the study authors suggested that hypomethylation of ras genes may play
a role in liver tumor development in this strain.
Dees and Travis (1994) found that chloroform exposure (0.5—2.0%, v/v) resulted in
a slight hypermethylation of the p53 protein in both the RLE rat cell line, and in the
human sarcoma line Saos-2 transfected with the gene for the tumor suppressor p53.
Stronger hypermethylation was observed with the tumor promoter phorbol myristate
acetate, and with benzene and toluene. Because the compounds investigated have been
reported to stimulate protein kinase C, the study authors hypothesized that these
compounds increase p53 methylation by stimulating protein kinase C. The authors
suggested that this study and others investigating the possibility of protein kinase C-
26
-------
mediated tumor promotion by chloroform provide support for an alternative mechanism
for chloroform carcinogenicity.
2.1.6 Potential Interactions
As discussed in U.S. EPA (1994), a variety of chemicals have been found to
potentiate chloroform toxicity. Inducers of the forms of cytochrome P450 that metabolize
chloroform result in increased toxicity, due to the increased production of reactive
metabolites. Conversely, inhibitors or inactivators of P450 would decrease chloroform
toxicity. Several ketones and chemicals that are metabolized to ketones also increase
chloroform hepatotoxicity by a mechanism not solely related to enzyme induction.
Mechanisms proposed for this potentiation include an effect on calcium pump activity
and increased susceptibility of organelles. Suicide inactivators of CYP2E 1, which are
activated by the enzyme but bind covalently to it, include carbon tetrachloride, vinyl
chloride, and trichioroethylene. The extent of all of these interactions, however, has not
been well quantified.
The interaction between chloroform and dichioroacetic acid (DCA) or
trichioroacetic acid (TCA) is of greater interest, since DCA and TCA are also drinking
water disinfectant byproducts. (Both DCA and TCA are ketogenic compounds.) Davis
and colleagues have conducted a series of experiments investigating these interactions
(Davis, 1992; Yang and Davis, 1997a and 199Th). Pretreatment of fasted male and
female Sprague-Dawley rats with three gavage doses of 2.45 mmol/kg/dose in 24 hours
(total of 947 mg/kg) markedly increased the hepatotoxicity observed from a 3.12
mmol/kg (372 mg/kg) i.p. dose of chloroform. DCA alone was not hepatotoxic, and
minimal to no effects were seen at this dose of chloroform alone. A larger effect was
seen when rats were fasted prior to challenge with chloroform. The authors noted that
fasting depletes protein stores, induces P4502E1 activity, and decreases hepatic
glutathione content. In another study, nonfasted rats were administered DCA or TCA
under the same conditions at 0.92 or 2.45 mmol/kg/dose, followed three hours later by a
single i.p. dose of 75 mg/kg chloroform. Based on blood biochemistry, this chloroform
dose caused hepatotoxicity and nephrotoxicity in females, but only marginal liver and
kidney effects in males. DCA increased the chloroform hepatotoxicity and
nephrotoxicity, and TCA increased chloroform nephrotoxicity, but interactive effects
were seen only in females. In vitro data showed increased chloroform metabolism
following DCA treatment. Although small but statistically significant increases in total
cytochrome P450 levels and decreases in glutathione were observed in vivo, the authors
noted that the increase in toxicity in vivo could not be attributed to increased cytochrome
P450 levels, decreased hepatic glutathione, nor increased chloroform dose to the liver.
27
-------
As noted for the examples of DCA and TCA, contaminants that interact with
chloroform by would have the effect of shifting the dose-response curve for chloroform to
the left. A large enough shift of the curve could move a given exposure toward the linear
region of the dose-response curve for cytotoxicity, and hence carcinogenicity. Clearly
such high exposures should be avoided.
In the low-dose range associated with environmental exposures, toxicity data on
DCA and chloroform interactions are not available. Data on DCA exposure of humans is
limited to information on drinking water levels. As reported by ILSI (1997), IARC
(1995) reported that levels of DCA in drinking water are generally less than 100 ugfL,
corresponding to a dose of approximately 0.003 mg/kg-day. As noted in Section 2.3, total
mean chloroform intake via ingestion, inhalation, and dermal contact is approximately
0.002 mg/kg-day, with total intake of 0.01 mg/kg-day estimated for individuals
consuming tap water containing relatively high levels of chloroform. As noted in the
Technical Support Document on Risk Assessment of Chemical Mixtures (U.S. EPA,
1988), mechanistic considerations suggest that “additivity may be a plausible assumption
in the low-dose region because thresholds for many types of interactions are expected to
exist.” Thus, it is reasonable to conclude, in the absence of data to the contrary, that DCA
and chloroform interactions at low dose would be additive. This conclusion is further
supported by recognizing that the public is exposed to more than just DCA and
chloroform in drinking water, and that not all of these interactions will be synergistic. In
fact, some of them are expected to be antagonistic.
2.2 Characterization of Dose Response
2.2.1 Quantification of Noncarcinogenic Effects
U.S. EPA (1994) developed Health Advisories and a Reference Dose (RfD) for the
noncarcinogenic effects of chloroform. The basis for these values is discussed in the
Drinking Water Criteria Document on Trihalomethanes (U.S. EPA, 1994) and on IRIS
(U.S. EPA, 1998).
A One-day Health Advisory (HA) value for chloroform of 4 mgfL was calculated
from an acute oral No-Observed-Adverse-Effect Level (NOAEL) of 35 mg/kg in mice. A
NOAEL value of 35 mg/kg-day, identified in pregnant rabbits dosed by gavage on days 6
to 15 of gestation, was used to calculate a Ten-day HA value of 4 mgIL. No adequate
data were located for calculating Longer-term HA values for chloroform, so the Drinking -
Water Equivalent Level (DWEL) of 0.4 mg/L (based on the RID, see below) may be
28
-------
taken as a conservative Longer-term HA for adults, and the adjusted DWEL (0 1 mg/L) as
a conservative Longer-term HA for children.
A Reference Dose (RID) of 0.01 mg/kg-day was based on a Lowest-Observed-
Adverse-Effect Level (LOAEL) for minimal liver injury (e.g., slightly elevated SGOT
levels and an increased number of fatty cysts in the liver) of 15 mg/kg-day identified in
7.5-year study in dogs (Heywood et a!., 1979). The critical study was of chronic duration,
used a fairly large number of dogs, and measured multiple endpoints; however, only two
treatment doses were used and no NOAEL was determined. An uncertainty factor of
1000 was used; subfactors of 10 were judged appropriate for interspecies extrapolation,
intraspecies variability, and LOAEL to NOAEL adjustment. U.S. EPA judged the
confidence in the study as medium. Confidence in the data base was considered medium
to low; several studies support the choice of a LOAEL, but a NOAEL was not found.
Confidence in the RID was also considered medium to low. After adjusting for an adult
consuming 2 liters of tap water per day for a 70 kg adult, and applying a relative source
contribution of 80% because most exposure is likely to come from drinking water, the
MCLG is estimated to be:
MCLG Based on RID for hepatotoxicity 0.01 mg/kg-day x 70 kg x 0.8/2 L/day
= 0.3 mg(L(rounded)
2.2.2 Quantification of Carcinogenic Effects
Chloroform has been reported to be carcinogenic in several different chronic
animal studies, increasing the frequency of liver tumors in male and female mice
administered chloroform by gavage in oil, but not in female mice administered
chloroform in drinking water, and increasing the incidence of kidney tumors in male rats
and certain strains of male mice. The U.S. EPA (1985) reviewed the evidence on the
carcinogenicity of chloroform and ranked it as a Group B2 carcinogen (probable human
carcinogen). Because the formation of liver tumors in mice appears to be dependent upon
the use of an oil vehicle, U.S. EPA (1987) recommended that the calculation of the cancer
risk estimate for chloroform be based on the incidence of renal tumors in male rats
exposed to chloroform in drinking water from the Jorgenson Ct al. (1985) study. This is
the current position on EPA’s Integrated Risk information System (U.S. EPA, 1998), and
is the position of this document as well.
Using EPA’s 1986 cancer guidelines, the potency of chloroform from this study is
estimated to be 6.1 x iO (mg/kg-day)’ by applying the LMS model. This results in a
unit risk value of 1.7 x i0 ’ (ugfL)’ and a corresponding drinking water concentration of
29
-------
6 ugh at an upper bound excess cancer risk level of 1 x 10.6. Using the body weight to the
3/4 power conversion instead of the body weight to the 2/3 conversion, the potency
calculated using the LMS model is 4.0 x iO (mg/kg day)*
Using recommendations for linear extrapolation from the U.S. EPA’s draft cancer
guidelines (U.S. EPA, 1996), an ED 10 and an LED 10 of 37 and 23 mg/kg-day,
respectively, were calculated in U.S. EPA (1997b) using the same data set. 4 Using the
LED 10 the resulting potency is 4.3 x i0 (mg/kg-day)* The resulting unit risk value is
1.2 x iO (ug(L) and the corresponding drinking water concentration is 8 ugIL at an
upper limit excess cancer risk level of I x 10 . Use of the ED 10 instead of the LED 10
results in values that are approximately 60% higher.
In view of the weight of evidence that chloroform may induce tumors by a
nonlinear mechanism (ILSI, 1997), a margin of exposure approach for dose-response
analysis might be employed. U.S. EPA (1997b) calculated an ED 10 of 37 mg/kg-day and
an LED 10 of 23 mg/kg-day) from tumors in the kidney of Osborne-Mendel rats
administered chloroform in drinking water (Jorgenson et al., 1985). Dividing these by an
estimated margin of exposure (MOE) can result in an exposure from which recommended
water concentration may be derived.
U.S. EPA (1996) considers several areas of uncertainty that should be addressed
with any MOE analysis. These areas are the slope of the dose response curve, the nature
of the response modeled, the nature and extent of human variability, the persistence of the
agent in the body, and the human sensitivity to the critical effect as compared with
experimental animals. In the case of chloroform, the overall MOE might be as high as
1000, for these areas combined as discussed briefly below.
The use of a 10-fold default factor for intra-human variability is appropriate and
could be recommended in lieu of specific data on differences in dynamics among
individuals, and based on expected differences in the metabolism of chloroform due to
differences in the CYP2E1 enzyme (e.g., Lucas et a!., 1993; Stephans et al., 1994). The
use of a additional 10-fold default factor for inter-species variation is appropriate and
could be supported as per recommendation in the 1996 guidelines (U.S. EPA, 1996). The
use of a final 10-fold factor is appropriate for the remaining uncertainties associated with
the mode of carcinogenic action understanding, the slope of the dose response curve, and
Except that values were determined using the adjustment of 3/4ths power of body weight between rats
and humans, rather than the 2i3rds power of body weight currently on IRIS (U.S. EPA, 1998).
30
-------
the lack of chloroform persistence in the body For chloroform, the slope of the dose
response curve for kidney tumors is shallow in comparison to curves for other chemicals,
but not necessarily similar endpoints (Dourson and Stara, l983). Perhaps contrary to
what one might expect, a shallower curve is expected to be associated with more need for
margin of exposure. This is because a shallower slope means that the nonlinearity in the
dose response curve is less quickly achieved as high dose is extrapolated to low dose
(when compared to chemicals with steeper dose response curves). (In other words, for
two chemicals with nonlinear modes of action and the same LED 10 , the dose to the
experimental animal below which the tumorigenic action would not be expected to occur
would be lower for the chemical with shallower slope.) In contrast, the lack of
persistence of chloroform in the body is supported by the known rapid clearance and
excretion of chloroform when compared to other chemicals. Thus, lack of persistence
indicates that more MOE is not needed. When these areas of uncertainty are folded
together, an overall factor of 10 is reasonable.
Therefore, 37 or 23 mg/kg-day is divided by a MOE of 1000, giving 0.03 7 and
0.023 mg/kg-day, respectively. After adjusting for a 70 kg adult consuming 2 liters of tap
water per day, and applying a relative source contribution of 80% (EPA assumes that
drinking water is the predominant source of chloroform intake), the MCLG based on
tumor responses is estimated to be:
MCLG for Chloroform (Based on LED 10 for Tumor Response) =
0.023 mg/kg-day x 70 kg x 0.8/2 L/day = 0.6 mgfL (rounded)
MCLG for Chloroform (Based on ED 10 for Tumor Response) =
0.037 mg/kg-day x 70 kg x 0.8/ 2 L/day = 1 mg/L (rounded)
Alternatively, an ED 10 or LED 10 , for the same endpoint, development of tumors in
the kidney of Osborne-Mendel rats administered chloroform in drinking water (Jorgenson
et al., 1985) could be expressed in terms of the tissue dose, rather than the administered
dose. For example, ILSI (1997) proposed an ED 10 or LED 10 of7l or 59 (respectively)
mg/hr/liter of liver, based on the maximum rate of metabolism of chloroform tG
phosgene. If a human physiologically based pharmacokinetic (PBPK) model were
The slope of the probit, log administered-dose response curve for the kidney tumors evoked by
chloroform in the Jorgenson et al. (1985) study is approximately 0.6. An appropriate comparison of
steepness in this dose response curve would perhaps be to cancer slopes from other chemicals. However,
while numerous values of the “slope factor” q1 have been calculated, little information is available on
the slope of cancer dose-response curves in the range of the experimental data. In the absence of such
data, the lethality curves of Dourson and Stara (1983) were used for comparison.
31
-------
a ailable. such a model could be used to con ert the ED 0 or LED 1 ( estimated b ILSI
(1997) based on tissue dose to human equivalent doses The construction of a human
PBPK model is still needed; this is an area for future work
It might also be appropriate to base the point of departure on cytotoxicity and
regenerative cell proliferation endpoints, since the dose response for tumor incidence and
increased cell proliferation were similar. The ILSI panel (ILSI, 1997) noted that the
cytotoxicity data have been reported as means ±SD. In order to determine an ED 10 or
LED 10 for cell proliferation, it would be necessary to define an increase in cell
proliferation that would be considered to be adverse, and then determine the incidence of
adverse responses based on individual animal data. In the absence of such data, tumor
incidence was considered an appropriate endpoint for analysis in the observable ange, in
light of the similarity of the dose-response curves for the two endpoints.
Table I summarizes the quantification of noncarcinogenic and carcinogenic effects
for chloroform.
Table 1. Summary of Quantification of Toxicological Effects for Chloroform
Health Advisory (HA), DWEL, Risk Specific Concentration (RSC) or Concentration mg/L
MOE-based concentration
One-day HA for 10-kg child 4
Ten-day HA for 10-kg child 4
Longer-term HA for 10-kg child 0 1
Longer-term HA for 70-kg adult 0.4
DWEL based on RID 0 4
MCLGbasedonRfD 03or007
U S EPA 1986 concentration at 10 risk level 0 006
U S EPA 1996 concentration at 10 risk level 0.008
MCLG based on MOE from tumor endpoint at LED 1 O & exposure dose 0 6
MCLG based on MOE from tumor endpoint at LED 10 & tissue dose future effort
MCLG based on MOE from cytotoxicity at LED 10 and tissue dose future effort
32
-------
4 Estimates ot concentrations incorporate consumption assumptions of 2 liters of water per day and 70 kg
body eight for adults, and 1 liter of water per day and 10kg child body weight The MCLG
concentrations further include an 80% or 20% relative source contribution of drinking water to the
overall chloroform exposure
33
-------
2.3 Characterization of Exposure
U S. EPA (1994, 1997a) summarized information on human exposure to
chloroform. Information below has been extracted from this document, unless otherwise
noted
Chloroform is found in virtually all treated drinking water; however,
concentrations vary widely depending on the type of water treatment, locale, time of year,
and source of the drinking water. Chloroform concentrations in drinking water have
ranged from less than 0.5 to 550 ugfL (ppb). Concentrations of all chloroform in drinking
water is generally lower when the raw water is derived from ground water sources rather
than surface water sources.
Chloroform has been detected in food at concentrations ranging from non-
detectable to 830 ng/g (ppb). Chloroform is approved by FDA as an indirect food
additive for use as a component of adhesives or polycarbonate resins used in food
packaging.
Chloroform is ubiquitous in air, although the concentrations are highly variable
depending on the ambient environment. Chloroform concentrations tend to be higher in
indoor air compared to outdoor air because of the confined space and release of
chloroform from various indoor sources. Chloroform concentrations in personal air and
outdoor air ranged from 0.06 to 215 ug/m 3 (0.01—44 ppb) , and from 0.04 to 21.5 ug/m 3
(0.008—4.4 ppb), respectively. One major source of chloroform in indoor air appears to
be from tap water that releases chloroform when used for showers or washing. One study
indicated that concentrations of chloroform in shower stall air samples during a 10-minute
shower ranged from 10 to 500 ug/m 3 (2.05—1 12 ppb). The absorbed inhalation and
dermal doses were 0.24 and 0.23 ugfkg/day, respectively, for a combined absorbed
chloroform dose from a 1 0-minute shower of 0.47 ug/kg/day.
The use of chlorine to disinfect swimming pools and hot tubs also results in the
release of chloroform to the overlying air. One study indicated that chloroform
concentrations in swimming pool and hot tub water ranged from less than 1 to 530 ugfL.
Concentrations of chloroform in the air two meters above the pool water ranged from 0.1
to 260 ug/m 3 (0.2—53 ppb).
Chloroform exhaled in breath is related to body burden of chloroform and recent
exposure to chloroform in air or water. Background chloroform concentrations measured
in breath have ranged from 0.22 to 5.06 ug/m 3 (0.05—1.04 ppb), and reported breath
34
-------
concentrations after a 10-minute shower ranged from 6 to 21 uglm 3 (1.23—4 3 ppb)
Chloroform has also been detected in the blood, milk and adipose tissue of humans.
Chloroform concentrations in blood have ranged from less than 0.1 to greater than
25 ug/L (ppb). Chloroform has been detected in the milk of 7 of 49 lactating women
living in industrial areas; however, actual concentrations were not reported.
A number of authors have calculated inhalation and ingestion rates based on
various assumptions. ILSI (1997) summarizes some of these estimates and provides
some of their own which are reported here.
ILSI (1997) estimated mean intake of chloroform from indoor air for the general
population to be 0.3 to 1.2 ugfkg body weight per day. This is based on a daily inhalation
volume for adults of 22 m 3 , a mean body weight for males and females of 64 kg, the
assumption that 20 out of 24 hr are spent indoors (IPCS, 1994),.and mean levels of
chloroform in indoor air (1 to 4 ug/m 3 ).
As stated by ILSI (1997), individuals may be exposed to elevated concentrations
of chloroform (from chlorinated tap water) during showering (Jo et a!., 1990a, l990b).
Based on assumptions of an absorption efficiency from the respiratory tract of 0.77, a
breathing rate of 0.014 m 3 /min for a 70-kg adult, a measured mean concentration in
shower air of 157 ug chloroform/rn 3 and a ratio of body burden resulting from dermal
exposure to that of inhalation exposure of 0.93, these authors estimated that the average
intake of chloroform (inhalation and dermal absorption) was 0.5 uglkg body weight per
shower for a person weighing 70 kg.
ILSI (1997) stated that based on a review of relevant estimates, Maxwell et a!.
(1991) concluded that the ratio of the dose of chloroform received over a lifetime from
inhalation to that received from ingestion of drinking water is probably in the range of
0.6-1.5 but could be as high as 5.7. The ratio of the dose received dermally to that
received orally over a lifetime from drinking water was considered to be approximately
0.3 but could be as high as 1.8.
ILSI (1997) estimated mean intake of chloroform from drinking water for the
general population is less than 0.5 ug/kg body weight per day. This is based on a daily
volume of ingestion for adults of 1.4 liters and a mean body weight for males and females
of 64 kg (IPCS, 1994), with mean levels of chloroform in drinking water (generally < 20
ug/liter). As discussed by Bauer (1981), actual levels of exposure may be less than those
estimated on the basis of mean levels in drinking water since most of the chloroform
would be expelled from drinking water that is heated before consumption (tea, coffee,
35
-------
soups, sauces) For example, approximately 96% of the total volatile halogenated
hydrocarbon fraction was eliminated in water boiling for 5 mm, whereas 50-90% was
eliminated upon heating at 70-90°C (Bauer, 1981). (1 tlogenated hydrocarbons
eliminated in this way could of course be inhaled.) It should be noted, however, that
owing to the wide variations in concentrations of chloroform in water supplies, intake
from drinking water could be considerably greater than estimated here for some segments
of the general population.
ILSI (1997) estimated daily intake of chloroform from foodstuffs to be
approximately 1 ug/kg body weight per day. This is based on a daily volume of ingestion
of solid foodstuffs for reference adults of 1.5 kg and a mean body weight for males and
females of 64 kg (IPCS, 1994), and the mean level and percentage detection of
chloroform in foodstuffs in the market-basket survey reported by Daft (1989).
As stated also by ILSI (1997), based on estimates of mean exposure from various
media, therefore, the general population is exposed to chloroform principally in food,
drinking water, and indoor air in approximately equivalent amounts. The estimated
intake from outdoor air is considerably less. The total estimated mean intake is
approximately 2 uglkg body weight per day. For some individuals living in dwellings
supplied with tap water containing relatively high concentrations of chloroform,
estimated total intakes from drinking water through ingestion, inhalation, and dermal
contact are up to 10 ug/kg body weight per day. These multi-media exposure estimates
can be used as combined route inputs with a human PBPK model to. estimate metabolized
dose in target tissues.
3. Risk Conclusions and Comparisons
3.1 Key Lines of Evidence for Critical Effect
3.1.1 Overall Conclusion
Chloroform causes the development of cancer in several animal species by a
mechanism that is believed to be non-linear in the low-dose range. Chloroform will likely
be carcinogenic to humans by all routes of exposure if a sufficient dose is administered.
Human data are insufficient to determine chloroform’s potential carcinogenicity.
Experimental animal studies are somewhat variable in both tumors evoked and magnitude
of response. The principal studies for the determination of chloroform’s tumorigenic
effect and low dose extrapolation are NCI (1976), in which chloroform was administered
by gavage in oil to male and Osbome-Mendel rats and B6C3FI mice, and Jorgenson et
36
-------
al (1985), in which chloroform was administered in drinking water to male Osborne-
Mendel rats and female B6C3FI mice.
For noncancer toxicity a Reference Dose (RID) was based on a Lowest-Observed-
Adverse-Effect Level (LOAEL) for minimal liver injury of 15 mg/kg-day identified in a
7.5-year study in dogs (Heywood et al., 1979). The critical study was of chronic duration,
used a fairly large number of dogs, and measured multiple endpoints. Confidence in
critical study is considered medium, and in its supporting data base is considered medium
to low. Overall confidence in the RID is also considered medium to low, meaning that
additional data may more likely change the value of the RID when compared to a high
confidence RID (U.S. EPA, 1998).
3.1.2 Strengths and Weaknesses of the Evidence
As summarized in part by ILSI (1997), chloroform produces tumor responses in
the liver and kidney, but the responses vary by route of exposure, sex and strain. In the
liver, chloroform causes an increase in tumors only .with corn oil gavage administration;
other routes of exposure at similar or higher doses failed to induce a carcinogenic
response in the liver. Chloroform also induces renal neoplasia in rats and mice, but this
response is limited to males in both species. Renal tumors were found in two strains of
mice (BDF I and id), while other strains failed to show a renal tumor response. This
strain-specific difference appears to be related to a higher metabolic capability for
chloroform bioactivation, and thus greater cytotoxicity, in the sensitive strains. In rats,
renal tumors were found in the Osborne-Mendel strain. Studies in Wistar, Sprague-
Dawley and F344 rats were negative for kidney tumors.
Several studies in animals support the conclusion that chloroform causes cancer
after oral exposure, and at least two studies are strong enough to support quantitative dose
response assessment (NCI, 1976; and Jorgenson et a!., 1985). Of these two, the
Jorgenson et a!. (1985) study uses drinking water, which is preferred route and vehicle of
administration. By contrast, the NCI (1976) bioassay used oil gavage;- interpretation of
the tumor response data from this study is complicated by the potential effect of the bolus
dosing, and the potential effect of the oil on toxicity. One unpublished animal study is
available that indicates tumor response after inhalation exposure (M tsushima, 1994). As
noted above, however, the tumorigenicity of chloroform varies with the dosing regimen
(e.g., drinking water versus gavage in oil), species, strain, and sex. The available
information in humans is uninformative with respect to the potential carcinogenicity of
chloroform.
37
-------
3.1.3 Weight of Evidence: Key Conclusions, Assumptions and Defaults
Chloroform is likely to cause tumors in humans by multiple routes of exposure.
The tumor-causing potential of chloroform has been demonstrated in both rats and in
mice. The very weak or absent mutagenic activity of chloroform, however, suggests that
the pathway to tumor formation involves an indirect production (or promotion) of
mutations. Careful analysis of chloroforn;-toxicokinetics and chloroform-induced
pathology indicates that the observed tumors result from the oxidative metabolism of
chloroform to phosgene and hydrochloric acid. These metabolites are cytotoxic in liver
and kidney where they are produced at relatively high rates. Recurrent and continuing
hyperplasia following these episodes of toxicity appears to be a necessary precursor for
hepatic and renal tumor formation in rodents. It should be noted, however, that cell
proliferation is a necessary, but not a sufficient condition for the development of
chloroform-induced tumors.
The evidence for this mechanism is strong for liver tumors and for kidney tumors
in mice. It is less strong for kidney tumors in rats, primarily due to the lack of data on
cell proliferation in Osborne-Mendel rats following long-term exposure. One key
assumption behind this conclusion is that the metabolism of chloroform in humans is
similar to that in rodents for the target organs of interest. The data generally support this
assumption qualitatively, based on similarities in the two main paths of metabolism, but
notable quantitative differences exist among animal species in the amounts of metabolites -
formed. Another key assumption in this analysis is that the rate of cell killing and
regrowth found in rodents would be similar to that in humans. This assumption is
reasonable, based on our understanding of the toxicity of chloroform, and based on the
similarity in target organs observed in the limited number of human case studies to those
seen in animals. Another assumption is that the use of a rodent model in lieu of human
data is a reasonable default. Much work in toxicology supports this procedure.
Another potential mode of action for the induction of tumors by chloroform might
be through metabolism in the absence of oxygen (i.e., a reductive pathway), depletion of
cellular defense mechanisms (e.g., glutathione), and.DNA or other macromolecular
damage. This mode of action does not appear likely in vivo, except at high doses or in
induced animals. If present, it is likely to be quantitatively less extensive than the
cytotoxic mode of action. It may also be, however, that chloroform causes tumor
formation by both modes of action. The reductive pathway might lead to a small amount
of DNA damage, and the cytotoxicity and regrowth prompted by the oxidative pathway
might stimulate both naturally occurring mutations and chloroform-induced DNA
damage.
38
-------
3.1.4 Significant Issues and Uncertainties
As summarized in part by ILSI (1997), there are some data deficiencies; the
remaining uncertainties for assessing chloroform are in defining the relationship of tissue
metabolism to toxicity, and in assessing the pharmacokinetic parameters for chloroform
metabolism in humans. Acquisition of additional information in these areas would
provide greater confidence in the conclusion that tumors associated with exposure to
chloroform are primarily a secondary consequence of marked cytotoxicity in direct
association with a period of sustained cell proliferation induced by a metabolite.
The weight of evidence for an obligatory role of cytotoxicity in chloroform
carcinogenicity is strongest for hepatic tumors in rats and mice, and for renal tumors in
mice. The evidence is more limited for renal tumors in rats, primarily due to the relative
paucity of data on intermediate endpoints (e.g., cell proliferation) in the strains where
tumors have been observed. Uncertainty could be reduced by acquisition of additional
information on cytotoxicity and proliferative response in the strain in which tumors were
observed (i.e., Osborne-Mendel rats) following long-term exposure to chloroform, -
Additional data on chronic (e.g, 2-year) cytotoxicity/proliferative response in the kidneys
of F344 rats might also contribute to greater confidence in the hypothesized mode(s) of
action.
Although the weight of evidence supports the claim that chloroform is not
mutagenic, one area which could be clarified by further work is whether any of the
metabolites of chloroform are DNA-reactive. Data available for phosgene indicate that,
while it is extremely reactive, it is likely to bind to other cellular components prior to
reaching the genetic material in the cell nucleus. No studies on the genotoxicity of
phosgene were located.
It would also be desirable to clarify whether the same pathways of metabolism
contribute to the potential for cytotoxicity in rodents and humans, specifically with
respect to CYP2EI. In mice, it is clear that this pathway is responsible for much (if not
all) of the cytotoxic responses in liver and kidney. CYP2EI activity in mice and rats has
been localized to the centhiobular region of the liver and to the cortex of the kidney. It is
not known whether other P450 isoenzymes contribute to the metabolism of chloroform in
humans. Data on the localization and levels of CYP2EI in potential target organs in
humans are not available. -
If the development of MOE is to be based on tissue doses, then some uncertainty
39
-------
exists in the choice of the appropriate tissue dose surrogate for the putative toxic
metabolite of chloroform (i.e., phosgene) as made by ILSI (1997). ILSI (1997) also did
not attempt to assess pharmacokinetic parameters (tissue metabolism rates and the size of
the regions of the tissues which have chloroform metabolizing enzymes) for chloroform
metabolism in humans. These parameters need to be evaluated before extending the
rodent PBPK model to humans.
3.1.5 Alternative Conclusions
Metabolism of chloroform in the absence of oxygen (i.e., reductive metabolism)
will lead to the production of free radicals. One of the body’s normal protections against
free radicals is the naturally occurring chemical glutathione. If glutathione is depleted by
excess production of free radicals, this may in turn lead to genotoxicity, which may be
responsible for chloroform’s tumor production. However, this pathway is recognized to
be a minor one at most, and would likely be of relevance only at high doses or in induced
animals.
3.2 Likelihood of Human Harm
Sensitive subgroups of humans to chloroform’s toxicity have not been identified.
However, people with increased levels of chloroform metabolism, such as resulting from
alcohol-related CYP2E1 induction, might be expected to have an increased susceptibility
to the effects of chloroform. People with pre-existing liver or kidney dysfi.inction would
also be expected to be more sensitive to the effects of chloroform, since they would have
a decreased functional reserve. Based on a comparison of metabolic capacity and
sensitivity to systemic effects in fetal and young rats versus adult rats, and based on the
doses at which systemic effects have been seen in adult rats versus developmental effects
in rats and rabbits, children and fetuses do not constitute a sensitive population.
The margin of exposure approach taken for chloroform is considered to be
protective of susceptible groups, including children. The mode of action understanding
for chloroform’s carcinogenic and cytotoxic effects is not age related and involves a
generalize mechanism of toxicity that is seen consistently across different species.
Furthermore, CYP2E 1 generated metabolites are key to chloroform’s mode of action, and
CYP2E 1 activity is lower in children compared to adults or at least not more than adults
(Casazza et al., 1994). Additionally, the margin of exposure analysis is designed to be
protective for sensitive populations. In the case of chloroform, an additional 10 factor
was used to account for variability between the average human response and the response
of more sensitive individuals. Note that the use of this factor presupposes that the
40
-------
variation in human response is more than 10-fold, since it accounts for differences in
average to sensitive, and not resistant to sensitive individuals.
The average human exposure to chloroform from air, water and food is
approximately 2 ugIkg/day (ILSI, 1997). This average exposure will not likely yield the
occurrence of tumors in humans because it is about 12 to 19-fold less than the estimated
MOE-based dose of 23 or 37 (based on c ther the LED 10 or ED 10 , respectively) ug/kg/day,
which is derived from the tum
-------
divided by the estimated average exposure of —2 ug/kg-day (ILSI, 1997). If the ED 10
were used, the resulting MOE would be —19,000, meaning that the estimated average
chloroform exposures in humans are —19,000-fold be’ w those doses which have been
estimated to cause a 10% increase in tumors in rodents.
For noncancer toxicity an Rfl) of 0.01 mg/kg-day was based on a LOAEL for
minimal liver injury in dogs (Heywood et al., 1979). Confidence in the critical study is
considered medium, and confidence in the supporting data base is considered medium to
low, resulting in overall medium to low confidence in the RiD (U.S. EPA, 1998).
3.3.2 Strengths and Weaknesses of the Data Available for Analysis
Several animal studies support the conclusion that chloroform causes cancer in
animals. The assumption that it is likely to be carcinogenic in humans is reasonable.
Differences among the animal studies can be explained in part by differences in type of
exposure (single large daily doses versus more continuous consumption over the course
of the day), or by differences in metabolism. Humans develop some of the same toxic
responses to chloroform as do experimental animals after acute exposures, but the human
data cannot be used to determine directly whether chloroform causes tumors in. humans.
Based on the weight of evidence, sufficient mode of action information is available to
support a nonlinear default dose response assessment. Although some data may suggest
that a linear default dose response assessment is appropriate, this approach is not
supported by the overall weight of evidence.
The use of tumor data as a basis for the development of an ED 10 and LED, 0 is
traditional and well supported. Although use of tumor precursor data (e.g., cell
proliferation) was addressed in the proposed cancer guidelines (U.S. EPA, 1996),
sufficient data for determining an ED, 0 or LED 10 for cell proliferation are not available.
Use of the ED, 0 or LED 10 based on tissue dose would improve the accuracy of the animal
to human extrapolation. However, pharmacokinetic calculations of human tissue doses
have not been performed, although the development of such calculations appears possible
given the available data on enzyme levels in human liver and kidney.
3.3.3 Selection of Study, Issues of Route, Frequency & Duration of Exposure
A traditional approach to the estimation of human dose from animal dose was
conducted (U.S. EPA, l997b) on the Jorgenson et. al. (1985) study, in which both benign
and malignant kidney tumors were evoked by oral exposure to chloroform in drinking
water. This calculation entails the estimation of the human dose in mg/kg-day by
42
-------
multipl ing the animal dose by a 1/4 power of body vveight ratio between animals and
humans (often referred to as a body weight to the 3/4 power scaling). The use of this
traditional approach allows easier comparisons with other carcinogens. Unfortunately, it
does not use all of the available mode of action information on chloroform.
An alternative and perhaps preferred conversion of animal dose to human dose
could be based on PBPK model. Such a model could be used to calculate estimates of
the target tissue doses of phosgene, a primary chloroform metabolite, in those regions of
the kidney or liver with high levels of chloroform metabolizing enzymes. A PBPK
analysis has been performed to develop tissue doses of phosgene for rats (ILSI, 1997).
Such analyses for humans would be needed before interspecies comparisons and low dose
extrapolation could be made.
The likely chloroform exposure to the general population will be in food, drinking
water and indoor air. Taken together, this combined exposure is more likely to be
continuous rather than episodic, and more likely to be long-term rather than short-term.
Single or episodic exposures may occur in the environment due to releases, or perhaps in
the workplace due to spills. Such exposures, associated risks and appropriate
management solutions are not discussed here.
The most appropriate information on which to base a risk characterization of
chloroform to the general population is from the extensive experimental animal data base,
as briefly summarized in this document and more extensively elsewhere (U.S. EPA,
1994; U.S. EPA 1997a; and ILSI, 1997). Human data on the toxicity of chloroform are
always more desirable than animal toxicity studies. However, from the narrow viewpoint
of relevance of risk assessment, such data are not available in sufficient quantities or
quality to yield credible assessments. The human data also can not at present be used to
judge the appropriateness of the estimates of risk from the experimental animals.
The experimental animal studies on which the dose response assessments are
based are from long-term exposures and continuous dosing (Jorgenson et al., 1985;
Heywood et a!., 1979). These two studies paid careful attention to determining the doses
associated with effects and targets of chloroform (i.e., in the liver and kidney) identified
by other animal studies and in the limited available human data. The exposure route in at
least one of these studies (drinking water, in Jorgenson et al., 1985) matches the expected
human exposure routes more closely than other rodent studies that used oil gavage,
thereby increasing the confidence in the resulting estimates of risk.
3.3.4 Strengths & Weaknesses of the Assessment: Issues & Uncertainties
43
-------
The weight of evidence suggests that chloroform induces tumors by a nonlinear
secondary mechanism and that a nonlinear dose response is reasonable. As per EPA
proposed guidelines (1996), a margin of exposure approach is therefore recommended.
The basis of this MOE is either the ED 10 of 37 mg/kg-day, or the LED 10 of 23 mg/kg-day,
as estimated by EPA (1 997b) from tumors in the kidney of Osborne-Mendel rats exposed
to chloroform in drinking water (Jorgenson et al., 1985). Division of either the ED 10 or the
LED 10 by an estimated margin of exposure (MOE) of 1000 yields doses in the
approximate range of 30 uglkg-day.
The weight of evidence for the mode of action of chloroform supports the expected
nonlinear behavior. For example, cytotoxicity and cell regrowth is nonlinear (ILSI,
1997). Moreover, the assumption of a linear dose response assessment for carcinogens is
often predicated on the strength of their mutagenicity, and in this regard chloroform is
equivocally negative.
Alternative measures of ED 10 or the LED 10 have been proposed (ILSI, 1997). For
example, the estimation of an ED 10 of 71.3 mg/liter of kidney cortex/hour for increased
incidence of kidney tumors in rats is well supported by the consistency of the dose
response data under other dosing protocols and in other species. Additional data on cell
proliferation in rats would be needed to strengthen this alternative assessment, since
proliferation is a key parameter in its use. A margin-of-exposure of less than 1000 fold
might be applied to this latter ED 10 because the use of a PBPK model could reduce some
of the uncertainties in the rat-to-human extrapolation.
The choice of a nonlinear approach to the dose respcmse assessment is reasonable
given the weight of the overall evidence for the induction of tumors by chloroform in
rodents. The assumption that such tumors would be evoked in humans by a similar mode
of action, given a sufficient dose, is also reasonable, based on our understanding of the
metabolism of chloroform by mammals in general, and the types of effects evoked by
chloroform in rats and humans specifically. Metabolism does vary among species,
however, which introduces some uncertainty into the extrapolation from rats to humans.
Alternative modes of action are theoretically possible, for example, tumors may be
evoked by gene mutations from free radicals via reductive metabolism after cellular
defense mechanisms, such as glutathione, are depleted. However, at best the contribution
of the reductive pathway to chloroform metabolism is quantitative much smaller than that
of the oxidative pathway (ILSI, 1997).
44
-------
It also may be thai tumors are being induced by both modes of action, i.e., gene
mutations by way of minor (at best) reductive metabolism and cytotoxicity and regrowth
by oxidative metabolism. This latter supposition should be carefully explored in future
research.
3.3.5 Basis of Assumptions and Defaults
Sufficient data are available on chloroform’s kinetics and dynamics in order to
postulate a default nonlinear approach to dose response assessment as per the proposed
EPA guidelines (1996). Standard assumptions in this default dose response assessment
include:
• the use of experimental animal data as a surrogate for humans,
• the use of kidney tumor response in rats (for cancer effects) and liver disease in dogs
(for noncancer effects) as meaningful for extrapolating to human disease,
• the conversion of experimental doses in either rats or dogs to humans by either a
314ths power of body weight adjustment or division by a 10-fold uncertainty factor,
• the use of factors based on a logarithmic scale (10, 3 or 1) with either the MOE or
RID that address additional scientific uncertainties in the overall data base,
• the use of one digit of arithmetic precision for the MCLGs and DWEL because our
understanding of the underlying biology is’unlikely to be more precise than this.
The use of these and similar assumptions is common practice in conducting dose response
assessments by other environmental and health agencies throughout the world.
3.3.6 Alternative Approaches
A default linear approach (also described in the proposed EPA 1996 guidelines)
could be suggested for comparison to the nonlinear approach recommended here. The
basis for this linear approach can be found in the traditional approach to the dose response
assessment of carcinogens, and is suggested by some because of the expected formation
of free radicals during the reductive metabolism of chloroform (as discussed above).
However, cellular defense mechanisms, such as glutathione, must be depleted before this
formation of free radicals can occur, and the weight of evidence suggests that the
reductive metabolism is quantitatively minor when compared to oxidative metabolism.
Comparison with the current assessment on EPA’s Integrated Risk Information System
(IRIS) is also possible. Comparisons of the resulting water concentrations were presented
above in Table 1. Figure 1 also shows the alternative approach.
45
-------
3.4 Risk Characterization Summary
En the 1994 proposed rule, EPA classified chlornform under the 1986 EPA
Guidelines for Carcinogen Risk ssessment as a Group B2; probable human carcinogen.
This classification was primarily based on sufficient evidence of carcinogenicity in
animals. Kidney tumor data in male Osborne-Mendel rats reported by Jorgenson et a!.
(1985) was used to estimate the carcinogenic risk. An MCLG of zero was proposed.
Because the mode of carcinogenic action was not understood at that time, EPA used the
linearized multistage model and derived a carcinogenicity potency factor for chloroform
of 6 X i0 (mg/kg-day) ’. The 95% upper bound limit lifetime cancer risk levels of l0 ,
i0 , and 10 were determined to be associated with concentrations of chloroform in
drinking water of 6, 60, and 600 ugIL. Since the 1994 rule, several new studies are
available providing insight into the mode of carcinogenic action for chloroform. EPA has
reassessed the cancer risk associated with chloroform exposure (U.S. EPA, 1997b) by
applying the principles of the 1996 EPA Proposed Guidelines for Carcinogen Risk
Assessment (U.S. EPA, 1996), which are considered scientifically consistent with the
Agency’s 1986 guidelines (U.S. EPA, 1986).
The recent report by [ LSI (1997) was fi.illy considered as well as the new science
that has emerged on chloroform since the 1994 proposed rule, Based on this new
information, chloroform is considered to be a likely human carcinogen by all routes of
exposure. Chloroform’s carcinogenic potential is indicated by animal tumor evidence
(induced liver tumors in mice and induced renal tumors in both mice and rats) from
inhalation and oral exposures, as well as metabolism, toxicity, mutagenicity and cellular
proliferation data that contributes to an understanding of its mode of carcinogenic action.
Although the precise mechanism of chloroform carcinogenicity has not been established.
the ILSI (1997) report is a reasonable scientific basis to support a putative mode of
carcinogenic action involving cytotoxicity produced by the oxidative generation of highly
reactive metabolites (phosgene and hydrochloric acid), followed by regenerative cell
proliferation as the predominant influence of chloroform on the carcinogenic process.
This supports a nonlinear approach to extrapolating low dose risk. The ILSI (1997)
report also discusses uncertainties associated with the chloroform assessment, which
include lack of data on cytotoxicity and cell proliferation responses in Osborne-Mendel
rats, mutagenicity data on chloroform metabolites, and the lack of comparative metabolic
data in humans. These data deficiencies raise some uncertainty about how chloroform
may influence tumor development at low doses.
Therefore, both a linear and nonlinear default approaches is applied to estimate
cancer risk associated with drinking water exposure to chloroform. The nonlinear default
or margin of exposure approach should be used in quantifying the cancer risk associated
46
-------
with chloroform exposure because the evidence is more compelling for a nonlinear mode
of carcinogenic action. The linear dose-response extrapolation approach appears overly
conservative in estimating low-dose risk, but nevertheless it is shown to account for
remaining uncertainties. [ It should be noted that the 1996 linear LED 10 approach and the
1986 LMS approach resulted in similar unit risk estimations, 1.2 x 10-7 (ug/L) and 1.7.
x 10-7 (ug1L) , respectively].
The tumor kidney response data in Osborne-Mendel rats from Jorgenson et a!.
(1985) are used to serve as the basis for the point of departure because a relevant route of
human exposure (i.e., drinking water) and multiple dose of chloroform (i.e., 5 doses
including zero) were used in this study. The ED 10 and LED 10 for kidney tumors in this
study were estimated to be 37 and 23 mg/kg-day, respectively (U.S. EPA, 1997b). These
values were adjusted to equivalent human doses using the body weight to the 3/4
interspecies scaling factor, as proposed in the 1996 EPA cancer guidelines (U.S. EPA,
1996). Consistent with the EPA 1996 proposed cancer guidelines, a 100-fold factor was
applied to account for the variability and uncertainty associated with intra- and
interspecies differences, in the absence of data specific to chloroform. A science policy
decision was made to apply an additional factor of 10 to account for the remaining
uncertainties associated with the mode of carcinogenic action understanding, and the
nature of the tumor dose response relationship being relatively shallow. The total factor
of 1000-fold represents an adequate margin of exposure that addresses inter-and intra-
species differences and uncertainties in the database. Other factors considered in
determining the adequacy of the margin of exposure include the size of the population
exposed, duration and magnitude of exposure, and persistence in the environment.
Taking these factors into consideration, a MOE of 1000 is still regarded as adequate.
Although a large population is chronically exposed to chlorinated drinking water,
chloroform is not bio-persistent and humans are exposed to very low levels of chloroform
in the drinking water, below those exposures needed to induce a cytotoxic response.
(Levels of occurrence are typically as high as —130 ugfL and median values are —20
ugfL.) Therefore, 37 or 23 mg/kg-day is divided by a MOE of 1000, giving 0.037 and
0.023 mg/kg-day, respectively. After adjusting for a 70 kg adult consuming 2 L of tap
water per day, and applying a relative source contribution of 80% (assuming that drinking
water is the predominant source of chloroform intake), the MCLG based on tumor
responses is estimated to be 0.6 mg/L based on the LED 10 , or I mgIL based on the ED 10 .
In the 1994 proposed rule, data from a chrome oral study in dogs were used to
derive the RID of 0.01 mgfkg/d (U.S. EPA, 1994). This Rif) is based on a LOAEL for
hepatotoxicity and application of an uncertainty factor of 1000. After adjusting for an
adult consuming 2 L of tap water per day for a 70 kg adult, and applying a relative source
47
-------
contribution o 80% or 20%. the MCLG is estimated to be
MCLG Based on RID for Hepatotoxicity =
0.01 mg/kg-day x 70 kg x 0.8 or 0.2 / 2LIday = 0.3 mg/L (rounded)
= 0.07 mg/L (rounded)
A MCLG based on hepatotoxicity (U.S. EPA, 1994) is more sensitive than the
values determined from the LED 10 (ED 10 approach for kidney tumorigenesis (0.6 mg(L or
1 mgfL), and is consistent with chloroform’s putative mode of action involving the
oxidative generation of reactive and toxic metabolites (phosgene and hydrochloric acid).
The exact mechanism by which these metabolites cause cytotoxicity is not known, but
plausible mechanisms can be hypothesized based on an understanding of important
nucleophilic targets in the cell. The electrophilic metabolite phosgene could react with
phosphotidyl inositols or tyrosine kinases, which in turn could potentially lead to
interference with signal transduction pathways thus leading to carcinogenesis.
Glutathione, free cysteine, histidine, methionine, and tyrosine are all potential reactants
for electrophilic agents. Likewise, it is also plausible that phosgene reacts with cellular
phospholipids, peptides, and proteins, resulting in generalized tissue injury.
Hepatotoxicity is the primary effect observed following chloroform exposure, and among
tissues studied for chioroform-oxidative metabolism, the liver was found to be the most
active (ILSI, 1997). It should be noted that the MCLG based on the RfD for
hepatotoxicity, 0.3 mg/L or 0.07 mg(L, and the MCLG based on the ED 10 for renal
tumorigenesis, I mg/L, falls within the 5 x iO to —2 x l0 range predicted for cancer
risk in the 1994 proposal using the LMS model. An MCLG based on protection against
cytotoxicity should be protective against carcinogenicity given that the putative mode of
action understanding for chloroform involves cytotoxicity as a key event preceding tumor
development.
Therefore, based on the available evidence, a nonlinear approach is considered for
estimating the carcinogenic risk associated with lifetime exposure to chloroform via
dnnking water. Given that hepatic injury is a primarily effect following chloroform
exposure, which is consistent with the mode of action understanding, the 1994 R±D based
on hepatotoxicity is considered a reasonable basis for the chloroform MCLG. It should
be noted that 70 or 300 ug(L equates to 1 to 5 x 10 cancer risk level using the LMS
model for kidney tumors.
48
-------
4. References
Ade, P., C. Guastadisegni, E. Testai, and L. Vittozzi. ‘994 Multiple activation of
chloroform in kidney microsomes from male and female DBA/2J mice. J. Biochem
Toxicol. 9(6): 289-295.
Baeder, C., and 1. Hofmann. 1991. Initinl submission-chloroform: supplementary
inhalation embryotoxicity study in Wistar rats (final report) with attachments and cover
letter dated 12/24/91. NTIS/0TS0535017. EPAIOTS Doc#88-920000566. Performed by
Hoechst Aktiengesellschaff, Germany, Sponsored by Hoechst AG and Dow Europe SA. Report
No. 91.0902.
Bauer, U. 1981. Human exposure to environmental chemicals: Investigations on volatile
organic halogenated compounds in water, air, food and human tissues. III.
Communication: Results of investigations. Zbl Bakt HygI Abt Orig. B 174: 200-237.
Boutelet-Bochan, H., Y. Huang, and M.R. Juchau. 1997. Expression ofCYP2EI during
embryogenesis and fetogenesis in human cephalic tissues: implications for the fetal
alcohol syndrome. Biochem. Biophys. Res. Commun. 238: 443-447.
Brass, H.J., M.A. Feige, T. Halloran, J.W. Mello, D. Munch, and R.T. Thomas. 1977.
The national organic monitoring survey: Samplings and analyses for purgeable organic
compounds. liii Pojasek, R.B., ed. Drinking water quality enhancement source
protection. Ann Arbor, MI: Ann Arbor Science, pp. 393-4 16.
Bull, R.J., J.M. Brown, E.A. Meirerhenry, T.A. Jorgenson, M. Robinson, and J.A. Stober.
1986. Enhancement of the hepatotoxicity of chloroform in B6C3F1 mice by corn oil:
Implications for chloroform carcinogenesis. Env. Health Perspect. 69: 49-58.
Butterworth, B.E., M.V. Templin, A.A. Constan, et al. 1998. Long-term mutagenicity
studies with chloroform and dimethylnitrosamine in female lad transgenic B6C3F mice.
Env. Mol. Mut. 31: 248-256.
Carpenter, S.P., J.M. Lasker, and J.L. Raucy. 1996. Expression, induction, and catalytic
activity of the ethanol-inducible cytochrome P450 (CYP2E1) in human fetal liver and
hepatocytes. Mol. Pharmacol. 49: 260-268.
Casazza et al. 1994. Serum acetone and liver acetone monoxygenase activity in pregnant
rats, fetuses, and neonates: reversible pretranslational reduction of cytochrome P4502E 1
during pregnancy. Archives of Biochemistry and Biophysics. 309: 111-1 16.
49
-------
Chiu. N . J. Orme-Zavaleta, A. Chiu, C. Chen, A. DeAngelo, W. Brattin, and J. Blancato.
1996. Characterization of cancer risk associated with exposure to chloroform. Environ
Carcino. and Ecotox. Rev. C 14(2): 81-104.
Daft, J.L. 1989. Determination of furnigants in fatty and non-fatty foods. J. Agric. Food
Chem. 37(2): 560-564.
Davis, M.E. 1992. Dichioroacetic acid and trichloroacetic acid increase chloroform
toxicity. J Toxicol Environ Health. 37: 139-148.
DeAngelo, A. et. al. 1995. Evaluation of the ability of chloroform administered in the
drinking water to enhance renal carcinogenesis in male F344 rats (letter summary from A.
DeAngelo to N. Chiu. October, 1995).
Dees, C., and C. Travis. 1994. Hyperphosphorylation of P53 induced by benzene,
toluene, and chloroform. Cancer Letters. 84(2): 117-123.
Dourson, M.L., and J.F. Stara. 1983. Regulatory history and experimental support of
uncertainty (safety) factors. Reg. Toxicol. Pharmacol. 3: 224-238.
Doyle, T.J., W. Zheng, J.R. Cerhan, et al. 1997. The association of drinking water
source and chlorination by-products with cancer incidence among postmenopausal
women in Iowa: A prospective cohort study. Am J Pub Health 87(7): 1168-1176.
Fox, T.R, A.M. Schumann, P.G. Watanabe, B.L. Yano, V.M. Maher, and J.J.
McCormick. 1990. Mutational analysis of the H-RAS oncogene in spontaneous
C57BL/6 x C3FLIHE mouse liver tumors and tumors induced with genotoxic and
nongenotoxic hepatocarcinogens. Cancer Res. 50(13): 4014-4019.
Fujie, K., T. Aoki, Y. Ito, and S. Maeda. 1993. Sister-chromatid exchanges induced by
trihalomethanes in rat erythroblastic cells and their suppression by crude catechin
extracted from green tea. Mutat Res. 300(3-4): 24 1-246.
Gemma, S., P. Ade, M. Sbraccia, E. Testal, and L. Vittozzi. 1996a. In vitro quantitative
determination of phospholipid adducts of chloroform intermediates in hepatic and renal
microsomes from different rodent strains. Environmental Toxicology and Pharmacology.
2(2-3): 233-242.
Gemma. S., S. Faccioli, P. Chieco, M. Sbraccia, E. Testal, and L Vittozzi. 1996b. In vivo
50
-------
CHC I3 bioactivation. toxicokinetics. toxicity, and induced compensatory cell
proliferation in B6C3F1 male mice. Toxicol. Appi. Pharmacol. 14 1(2): 394-402.
Golden R.J., S.E. Hoim, D.E. Robinson, P.H. Julkunen, and E.A. Reese. 1997.
Chloroform mode of action: Implications for cancer risk assessment. Regulatory
Toxicology and Pharmacology 26: 142-155.
Hakkola, J., 0. Pelkonen, M. Pasanen, and H. Raunio. 1998. Xenobiotic metabolizing
cytochrome P450 enzymes in the human feto-placental unit: Role in intrauterine toxicity.
Crit. Rev. Toxicol. 28(1): 35-72.
Hardie, D.W.F. 1964. Chlorocarbons and chiorohydrocarbons: Chloroform. j j Kirk,
D., and D.E. Othmer, eds. Encyclopedia of chemical technology, 2nd ed. New York:
John Wiley and Sons, Inc. Reviewed in U.S. EPA (1980a).
Heywood, R., R.J. Sortwell, P.R.B. Noel, A.E. Street, D.E. Prentice, FJ.C. Roe, P.F.
Wadsworth, A.N. Worden, and N.J. Van Abbe. 1979. Safety evaluation of toothpaste
containing chloroform: HI. Long-term study in beagle dogs. J. Environ. Pathol. Toxicol.
2: 835-85 1.
Hoff, G., I.E. Moen, P. Mowinckel, et al. 1992. Drinking water and the prevalence of
colorectal adenomas: An epidemiologic study in Telemark, Norway. Eur J Cancer Prey
1(6): 423-428.
International Agency for Research on Cancer (IARC). 1995. Monographs on the
evaluation of carcinogenic risks to humans. Vol. 63. Dry cleaning, some chlorinated
solvents, and other industrial chemicals. WHO. Lyon, France.
International Life Sciences Institute (ILSI). 1997. An evaluation of EPA’s proposed
guidelines for carcinogen risk assessment using chloroform and dichioroacetate as case
studies: Report of an expert panel. Washington, D.C. November.
IPCS (1994). Environmental health criteria 170: Assessing human health risks of
chemicals: derivation of guidance values for health-based exposure limits. Geneva,
International Programme on Chemical Safety, WHO.
Jo, W.K., C.P. Weisel and P.J. Lioy. 1990a. Chloroform exposure and the health risk
associated with multiple uses of chlorinated tap water. Risk Anal. 10(4): 58 1-585.
51
-------
Jo, W K, C.P. Weisel, and P.J. Lioy. 199Db. Routes of chloroform exposure and body
burden from showering with chlorinated tap water. Risk Anal. 10(4): 575-580.
Jorgenson, T.A., and C.J. Rushbrook. SRI International. 1980. Effects of chloroform in
the drinking water of rats and mice: Ninety-day subacute toxicity study. Menlo Park, CA:
United States Environmental Protection Agency. Contract No. 68-03-26 16. EPA-600/1
80-030.
Jorgenson, T.A., E.F. Meierhenry, C.J. Rushbrook, R.J. Bull and M. Robinson. 1985.
Carcinogenicity of chloroform in drinking water to male Osborne-Mendel rats and female
B6C3F1 mice. Fund. App!. Toxicol. 5: 760-769.
Kramer, M.D., C.F. Lynch, P. Isacson, and J.W. Hanson. 1992. The association of
waterborne chloroform with intrauterine growth retardation. Epidemiology. 3(5): 407-
413.
Larson, J.L., D.C. Wolf and B.E. Butterworth. 1993. Acute hepatotoxic and nephrotoxic
effects of chloroform in male F-344 rats and female B6C3F1 mice. Fundam. App!.
Toxicol. 20(3): 302-3 15.
Larson, J.L., D.C. Wolf and B.E. Butterworth. 1994a. Cytotoxicity and cell proliferation
in hepatocarcinogenicity of chloroform in Female B6C3FI mice: Comparison of
administration by gavage in corn oil vs. ad libitum in drinking water. Fundam. Appi.
Toxicol. 22: 90-102.
Larson, J.L., D.C. Wolf and B.E. Butterworth. 1994b. Induced cytolethality and
regenerative cell proliferation in the livers and kidneys of male B6C3F 1 mice given
chloroform by gavage. Fundam. Appi. Toxicol. 23(4): 537-43.
Larson, J.L., D.C. Wolf, K.T. Morgan, S. Mery, and B.E. Butterworth. l994c. The
toxicity of 1-week exposures to inhaled chloroform in female B6C3F1 mice and male
F344 rats. Fund. App!. Toxicol. 22(3): 431-446.
Larson, J.L., CS. Sprankle, and B.E. Butterworth. 1994d. Lack of chloroform-induced
DNA repair in vitro and in vivo in hepatocytes of female B6C3F 1 mice. Environ. Mol.
Mutagen. 23(2): 132-6.
Larson., J.L., D.C. Wolf, and B.E. Butterworth. 1995a. Induced regenerative cell
proliferation in livers and kidneys of male F344 rats. Toxicology 95: 73-86.
52
-------
Larson. J L , D C. Wolf, S. Mery, K.T. Morgan, and B E. Butterworth. 1995b. Toxicity
and cell proliferation in the liver, kidneys, and nasal passages of female F-344 rats,
induced by chloroform administered by gavage. Food Chem Toxicol. 33(6): 443-456.
Larson, J.L., M.V. Templin, D.C. Wolf, et a!. 1996. A 90-day chloroform inhalation
study in female and male B6C3F1 mice: Implications for cancer risk assessment.
Fundam. AppI Toxicol. 30: 118-137.
Lawrence, C.E., P.R. Taylor, B.J. Trock, and A.A. Reilly. 1984. Trihalomethanes in
drinking water and human colorectal cancer. JNCI. 72(3): 563-568.
Le Curieux, F., L. Gauthier, F. Erb, and D. Marzin. 1995. Use of the SOS chromotest,
the Ames-fluctuation test and the newt micronucleus test to study the genotoxicity of
four trihalomethanes. Mutagenesis. 10(4): 333-41.
Lucas D., F. Berthou, Y. Dréano, P. Lozac’h, A. Volant, and J.-F. Ménez. 1993.
Comparison of levels of cytochromes P-450, CYPIA2, CYP2EI, and their related
monooxygenase activities in human surgical liver samples. Alcohol. Clin. Exp. Res.
17(4): 900-905.
Matsushima, T. 1994. Inhalation carcinogenesis study of chloroform. Letter summary
from T. Matsushima of Japan Bioassay Laboratory to A. Chiu. August, 1994.
Maxwell M.I., D.E. Burmaster, and D. Ozbnoff. 1991. Trihalomethanes. Maximum
contaminant levels: The significance of inhalation, dermal exposures to chloroform in
household water. Regul. Toxicol. Pharmacol. 14: 297-312.
Nakajima, T., E. Elovaara, T. Okino, H.V. Gelboin, M. Kiockars, V. Riihimaki, T.
Aoyama, and H. Vainio. 1995. Different contributions of cytochrome P450 2E1 and
P450 2B1/2 to chloroform hepatotoxicity in rat. Toxicol. App!. Pharmacol. 133(2):
2 15-222.
NC!. 1976. National Cancer Institute. Report on carcinogenesis bioassay of chloroform.
Springfield, VA: NTIS PB-264018.
NTP. 1988. National Toxicology Program. Chloroform reproduction and fertility
assessment in CD-i mice when administered by gavage. Report by Gulati, D.K., E.
Hope, R.C. Mounce, et a!. Environmental Health Research and Testing, Inc., Lexington,
KY to National Toxicology Program, NTP- 89-0 18. NTIS PB89-148639.
53
-------
Ohanian. E.V , J,A Moore. JR. Fowle III, G.S. Omenn , S.C Lewis, G M. Gray, and
D W. North. 1997. Risk Characterization: A bridge to informed decision making.
Fundamental and Applied Toxicology. 39: 81-88.
Pegram, R.A., M.E. Andersen, S.H. Warren, T.M. Ross, and L.D. Claxton. 1997.
Glutath jane S-transferase-mediated mutagenicity of trihalomethanes in Salmonella
typhimurium: Contrasting results with bromodichioromethane and chloroform. Toxicol.
Appi. Pharmacol. 144: 183-188.
Pereira, M.A. 1994. Route of administration determines whether chloroform enhances or
inhibits cell proliferation in the liver ofB6C3FI mice. Fundam. Appl. Toxicol. 23(1):
87-92.
Roe, F.J.C., A.K. Palmer, A.N. Worden, and N.J. Van Abbe. 1979. Safety evaluation of
toothpaste containing chloroform: I. Long-term studies in mice. J. Environ. Pathol.
Toxicol. 2: 799-8 19.
Roldan-Arjona, 1., and C. Pueyo. 1993. Mutagenic and lethal effects of halogenated
methanes in the Ara test of Salmonella typhimuriwn: Quantitative relationship with
chemical reactivity. Mutagenesis 8 (2): 127-13 1.
Rosenthal, S.L. 1987. A review ofthe mutagenicity of chloroform. Environ. Molecular
Mutagen. 10: 211-226.
Schenkman, J.B., K.E. Thummel, and L.V. Favreau. 1989. Physiological and patho-
physiological alterations in rat hepatic cytochrome P-450. Drug Met. Rev. 20(2-4): 557-
584.
Shelby, M.D., and K.L. Witt. 1995. Comparison of results from mouse bone marrow
chromosome aberration and micronucleus tests. Environ. Mol. Mutagen. 25(4): 302-313.
Song, B-J, H.V. Gelboin, S-S. Park, C.S. Yang, and F.J. Gonzalez. 1986.
Complementary DNA and protein sequences of ethanol-inducible rat and human
cytochrome P-450s. J. Bid. Chem. 261(35): 16689-16697.
Sprankle, C.S., J.L. Larson, S.M. Goldsworthy, and B.E. Butterworth. 1996. Levels of
myc, fos, Ha-ras, met and hepatocyte growth factor mR.NA during regenerative cell
proliferation in female mouse liver and male rat kidney after a cytotoxic dose of
chloroform. Cancer Lett. 10 1(1): 97-106.
54
-------
Templin. MV., K C. Jamison. D.C. Wolf, K T. Morgan, and B.E. Butterworth. 1996a
Comparison of chloroform-induced toxicity in the kidneys, liver, and nasal passages of
male Osborne-Mendel and F-344 rats. Cancer Lett. 104(1): 7 1-8.
Templin, M.V., J.L. Larson, B.E. Butterworth, K.C. Jamison, J.R. Leininger, S. Mery,
K.T. Morgan, B.A. Wong, and D.C. Wolf. 1996b. A 90-day chloroform inhalation study
in F-344 rats: Profile of toxicity and relevance to cancer studies. Fund. Appi. Toxicol.
32: 109-125.
Templin, M.V., A.A. Constan, D.C. Wolf, et al. 1998. Patterns of chloroform-induced
regenerative cell proliferation in BDF 1 mice correlate with organ specificity and dose-
response of tumor formation. Carcinogenesis. 19(1): 187-193.
Thompson, D.J., S.D. Warner, and V.B. Robinson. 1974. Teratology studies on orally
administered chloroform in the rat and rabbit. Toxicol. Appl. Pharmacol. 29: 348-3 57.
Tumasonis, C.F., D.N. McMartin, and B. Bush. 1987. Toxicity of chloroform and
bromodichioromethane when administered over a lifetime in rats. J. Environ. Pathol.
Toxicol. Oncol. 7(4): 55-64. /
Ueno, T., and F.J. Gonzalez. 1990. Transcriptional control of the rat hepatic CYP2EI
gene. Molecular and Cellular Biology. 10(9): 4495-4505.
Umeno, M., B.J. Song, C. Korzak, et al. 1988. The rat P-45011E1 gene: Complete intron
and exon sequence, chromosome mapping, and correlation of developmental expression
with specific 5’ cytosine demethylation. J. Biol. Chem. 263(10): 4956-4962.
U.S. EPA. 1975a. Preliminary assessment of suspected carcinogens in drinking water.
Report to Congress, United States Environmental Protection Agency. Washington, DC.
EPA-56014-75-005 PB. 260961. Reviewed in NAS 1977.
U.S. EPA. 1975b. Region V Joint FederallState survey of organics and inorganics in
selected drinking water supplies. United States Environmental Protection Agency.
Reviewed in NAS, 1977.
U.S. EPA. 1980a. Ambient water quality criteria for chloroform. U.S. Environmental
Protection Agency, Office of Water Regulations and Standards. Washington, DC
EPA-440/5-80-033.
55
-------
U S EPA 1985. Health assessment document for chloroform. Final Report. U S.
Environmental Protection Agency, Office of Research and Development. Research
Triangle Park, NC EPA-600/8-84-004F.
U.S. EPA. 1986. Reference Values for Risk Assessment. Cincinnati, OH. U.S.
Environmental Protection Agency, Environmental Criteria and Assessment Office.
U.S. EPA. 1987. Office of Drinking Water Position on Chloroform Risk Assessment.
United States Environmental Protection Agency.
U.S. EPA. 1988. Technical Support Document on Risk Assessment of Chemical
Mixtures. EPAI600/8-90/064. Office of Research and Development, Washington DC.
U.S. EPA. 1994. Final Draft for the Drinking Water Criteria Document on
Trihalomethanes. U.S. Environmental Protection Agency, Office of Science and
Technology. Washington, D.C., April 8.
U.S. EPA. 1995. Carol Browner’s memorandum on EPA Risk Characterization
Program. U.S. Environmental Protection Agency, Office of the Administrator.
Washington, D.C., March 21.
U.S. EPA. 1996. Proposed Guidelines for Carcinogen Risk Assessment. U.S.
Environmental Protection Agency, Office of Research and Development. Washington,
DC. EPAJ600TP-92/003C.
U.S. EPA. 1997a. Summary of new data on trihalomethanes (THMS) for the notice of
data availability (NODA). U.S. Environmental Protection Agency, Office of Science and
Technology. Washington DC.
U.S. EPA. 1997b. Development of an ED 10 and LED 10 for chloroform. Personal
correspondence with Paul Pinsky, U.S. Environmental Protection Agency, National
Center for Environmental Assessment. Washington, D.C.
U.S. EPA. 1998. Integrated Risk Information System (IRIS). Online at
http://www.epa.gov/iriS. U.S. Environmental Protection Agency, National Center for
Environmental Assessment. Washington, D.C.
Vieira, I., M. Sonnier, and T. Cresteil. 1996. Developmental expression of CYP2EJ in
the human liver: Hypermethylation control of gene expression during the neonatal
56
-------
period Eur J Biochem 238: 476-83.
Vorce, R.L., and J I. Goodman. 1991. Hypomethylation of ras oncogenes in chemically
induced and spontaneous B6C3F1 mouse liver tumors. J. Toxicol Environ Health. 34(3):
367-84.
Wallace, L. 1988. Chloroform available on tap. Health Environ. 2(5): 3-4.
Wallace, L.A., E.D. Pellizzari, T.D. Hartwell, C. Sparacino, B. Whitmore, L. Sheldon, H.
Zelon, and R. Perritt. 1987. The TEAM study: Personal exposures to toxic substances in
air drinking water, and breath of 400 residents of New Jersey, North Carolina and North
Dakota. Environ. Research. 43: 290-307.
Windholz, M., ed. 1976. The Merck index. Rahway, NJ: Merck & Co., Inc.
Yang, H., and M.E. Davis. 1997a. Dichioroacetic acid pretreatment of male and female
rats increases chloroform metabolism in vitro. Toxicol. 124: 53-62.
Yang, H., and M.E. Davis. 1997b. Dichloroacetic acid pretreatment of male and female
rats increases chloroform-induced hepatotoxicity. Toxicol. 124: 63-72.
57
-------
APPENDIX: SUMMARY OF REVIEWERS’ COMMENTS
Charge to External Peer Reviewers
Statement of Work
TITLE : Peer review of Health Risk Characterization Report on Chloroform
BACKGROUND :
The mission of the United States Environmental Protection Agency’s (EPA) Office
of Water (OW) is to protect public health and the environment from adverse effects of
contaminants in media such as ambient wa r, drinking water, waste water, sewage sludge
and sediments. This procurement relates to the peer review of a health risk assessment on
the disinfection by product, chloroform. This risk assessment will be used in support of
EPA’s stage I disinfection by product rule which is scheduled to be final in November
1998. The Safe Drinking Water Act Amendments of 1996 emphasize that “the best peer
review science” be used in carrying out SDWA regulations.
PURPOSE :
A cancer risk assessmentlcharacterization document has been recently prepared
that cites and updates EPA’s 1994 assessment on chloroform. This 1998 document
considers a new cancer bioassay in rodents and applies the EPA 1996 proposed revisions
to its guidelines for carcinogen risk assessment.
TASK DESCRIPTION :
This purchase will procure a peer review on the 1998 EPA chloroform risk
characterization assessment report. EPA has attached the 1998 risk assessment
document, consisting of approximately 20-25 pages, on chloroform to be reviewed
(Attachment 1), as well as supporting materials, such as EPA’s 1996 guidelines for
carcinogen assessment (Attachment 2), EPA’s 1994 Criteria Document on chloroform
(Attachment 3), and hard copies of key studies (Attachment 4). The peer reviewer shall
submit written comments that are clear/transparent, and constructive. They ahall
comment on whether the document clearly and adequately discusses:
- the weight of the evidence
- the key lines of evidence
58
-------
- the mode of carcinogenic action understanding
- alternative hypotheses
- uncertainties in the risk assessment
- scientific basis for the risk assessment dose-response choice (i.e., linear versus
nonlinear default approaches)
The peer reviewers shall indicate where they are in agreement with the report and where
they disagree. If they disagree with any part of the report or find a weakness in the report,
they shall recommend explicit guidance on revising the report. They shall provide
comments that include an overall general summary on the acceptability and adequacy of
the risk assessment, and specific comments as needed (comment 1 on page X, paragraph
X, line X).
59
-------
Figure 1. Graphical Representation of Extrapolations (Data not shown; X axis is not to scale; see text for further details)
Extrapolation Range Observed Range
0
U)
C
0
0.
U)
0
10%
‘0%
00002 0.002 001 0023
RSO(l0 ) RID (o.em9n)
(O .3mQ 1)
MOE
4
1000- fold
37
ED 10
Dose (mg/kg/day)
Estimated Average
Human xposure
Confidence Limit on
Estimate
Projected Linear
23
LED 10
-------
Summary & Response of External Reviewer Comments
Health Risk Assessment/Characterization of the Drinking Water Disinfection
Byproduct Chloroform, March l3t , 1998
Comments below are either actual text or paraphrases from the reviewers’ comments.
Editorial revisions and specific suggestion s were invariably made in response to
comments by the reviewers. These comments and responses are not reviewed here.
James A. Swenberg, D.V.M., Ph.D.
University of North Carolina
Section, Comment and Response
2.1.1 Toxicokinetics
The summary should state more definitively that the oxidative pathway is
responsible for most metabolism. Excretion of chloroform by the lungs is dose
dependent, with greater amounts excreted unchanged at high doses.
Response: Summary was changed as suggested.
2.1.4 Mechanisms of Toxicity
The section on formation of DNA adducts needs revision. DNA adducts have not
been identified with chloroform. Very small amounts of “covalent binding” have been
reported. These are not the same. It may only represent metabolic incorporation from the
C- 1 pool. The reductive pathway is associated with free radical formation, which leads to
oxidative stress and lipid peroxidation. Such responses are highly nonlinear and are
enhanced by conditions that deplete cellular defenses such as glutathione. Under normal
conditions, a cell is well equipped to detoxify these free radicals. The Melnick paper
further confuses the issue by including brominated compounds that are generally
recognized as genotoxic and that have identifiable DNA adducts.
Response: Several clarifying changes were made in the text on these issues.
2.2.2 Quantification of Carcinogenic Effects
This section is rather confusing compared to the ILSI document. . I do not
understand the factor of 3 for slope, nor the factor of 10 for severity. The cancer
endpoints are associated with toxicity and are expected results for the cause and effect. If
cytotoxicity is used instead of tumors, a factor of 10 and the use of the EDIO may be
reasonable. This can be coupled with a factor of 10 for intra-human variability and a
factor of 3-10 for inter-species variation, providing a margin of exposure between 300
Responses to External Review
-------
and 1000. The ILSI document uses a factor of 30. This should be discussed. This section
of the document needs to be more transparent and its content presented as a range
throughout the document.
Response: Text was changed to add clarity but the range of MOE was not
specified. Rather a single value of 1000-fold was chosen.
3.1.2 Strengths and Weaknesses
This section is very brief. In comparison with the ILSI document, it is much less
informative and less transparent. There is no discussion of inconsistencies in cell
proliferation data and strain specific deficiencies in the data sets.
Response: The text was only changed slightly here. Reference is made to the ILSI
work. This text could be greatly expanded but such effort would increase the
overall length beyond that desired for this risk characterization text.
3.1.3 Key Conclusions, Assumptions and Defaults
Again, this section is very brief and less informative than the 1LSI document. The
last part of paragraph 1 is incorrect in calling the DNA damage that results from reductive
metabolism “direct”. It is associated with free radicals and is fully expected to be
nonlinear and highly correlated with depletion of cellular defense mechanisms.
Response: Changes were made to correct the first paragraph. This section was
combined with the weight of evidence described below.
3.1.4 Weight of Evidence
Once again, the document should not refer to the free radical mechanism for DNA
damage as “direct” and it should not be stated to be linear as a function of dose. I know
of no data that demonstrate this, but there are many data sets that speak against it.
Response: Changes were made to correct this section. This section was combined
with that described above.
3.1.7 Alternative Conclusions
this may lead to high dose genotoxicity from free radicals that could contribute
to chloroform’s carcinogenicity. This pathway is recognized to be a minor one that
would not be expected to contribute to low dose carcinogenicity.
Response: Text was changed to reflect the concept of minor pathway.
3.2.1 Overall Conclusions.
This secticrn automatically embraces 7 ugfkg-day as the MOE-based dose. This
Responses to External Review 2
-------
uses a factor of 3000. In my opinion, this should be given as a range using a MOE. ILSI
would go as low as 30. Thus, this section should be broadened to better reflect a range of
MOE.
Response: The range of a MOE was not specified, but an increase in transparency
was attempted. An overall MOE of 1000 was chosen as best representative of the
underlying data.
3.2.2 Human Susceptibility
This section only addresses differences in exposure. It needs to also address
genetic polymorphisms that change metabolism. The inter-human factor of 10 has been
used for MOE and should be put in context with this section.
Response: Section was changed as suggested.
3.3.1 Overall Conclusions
Again, this section only uses the MOE of 3000. It needs to be more inclusive of a
range.
Response: Again, the range of a MOE was not specified, but an increase in
transparency was attempted. An overall MOE of 1000 was chosen as best
representative of the underlying data.
3.3.7 Significant Issues and Uncertainties
As stated several times in this review, tumors would not be expected “to be evoked
in a linear fashion by gene mutations from free radicals via reductive metabolism”. Such
mechanisms are expected to be nonlinear and associated with high doses that deplete
cellular defense mechanisms.
The MOE should be discussed as a range, not as 3000. This number is highly
conservative and not well supported by the data presented.
Response: The section was revised to address the comments on the expected lack
of low dose linearity with tumors. Again, the range of a MOE was not specified,
but an increase in transparency was attempted. An overall MOE of 1000 was
chosen as best representative of the underlying data.
3.3.8 Alternative Approaches
Again, the free radical mechanism is not expected to be linear.
Response: Section was revised to reflect this comment
Responses to External Review .
-------
Lorenz Rhomberg, Ph.D.
11 Newbury Terrace
Newton, MA 02 159-2127
Comment:
Critical discussion of key matters are not included in the risk characterization text, such
as chloroform’s metabolism, the role of corn oil vehicle, the agent’s genotoxic potential,
and the central hypothesis that animal tumors are secondary to cytotoxicity and
compensatory cell proliferation in the target tissues. In sum, the Characterization does
not really document the Agency’s weighing of the evidence on these key questions.
Response: The commentor is correct in that the text does not go into length
regarding key matters. However, this text is intended to be a characterization of
the rislç geared more to risk managers. Previous manuscripts, appropriately
referenced in this text, are available for a filler explication of the key matters that
were only summarized here.
Comment:
The Characterization document contains no data. While it need not be comprehensive, it
would seem important to present (in tabular form) the key datasets upon which the
quantitative risk assessment calculations are based as well as other datasets that stand as
the most prominent alternatives. The administered doses and tumor counts of the key
Jorgenson male rat kidney tumor data are reported differently in Chiu et a!. (1996) and in
the [ LSI document, for instance, and it is important that the specific version used in the
final calculations (and the reasons for its variation from other versions) be documented.
Response: Tumor counts were added for the key studies. The chosen data set can
be viewed on EPA’s IRIS ( yw.epa.gov/iris) , and was appropriately cited.
Comment:
Similarly, there is no documentation of risk assessment calculations, either in dose metric
calculation or in dose-response analysis. The citation for the key dose-response analysis
(p. 19) to “personal correspondence” with Paul Pinsky of EPA ’s National Center for
Environmental Assessment is not adequate documentation for so central a calculation. It
makes it impossible for the reader (and for me as a reviewer) to verify the calculations or
to see what specific approaches were taken (for instance, regarding such matters as lower
bound determination, allowance for concurrent mortality, and degree of model).
Response: This document is the original source of the data set used for the cancer
analysis. Citation of a personal communication with the EPA scientist that did this
Responses to External Review 4
-------
analysis is appropriate.
Comment:
Elsewhere (p.34) it is implied that the cross-species equivalency of doses was assumed to
be based on surface area scaling (i.e., mg&g /day equivalency), although the
calculations are said to follow the draft Proposed Guidelines, which mandate mgfkg 314 /day
as an oral dosing default. The document should show key calculations so that such
matters can be checked and the calculation’s basis understood.
Response: The commentor is correct. The 3/4 power of body weight was used in
the assessment and is appropriately acknowledged in a footnote to the text.
Comment:
While reviews of the literature and the state of the science on areas such as epidemiology,
animal bioassays, metabolism and pharmacokinetics, and mechanisms of toxicity need
not be included in the Risk Characterization per Se, a succinct summary of the main issues
(as opposed to the main studies and their findings) would seem to be in order. In the case
of chloroform, for epidemiology, the main issues are attribution of water chlorination
effects to specific disinfection byproducts and the confounding of exposure to
chiorine-disinfected drinking water with other factors of potential importance to cancer
risks; for animal bioassays they are the inconsistencies among studies, the dependence on
route of administration, the assessment of cytotoxicity among bioassay subjects, and the
potential bearing of minor tumor responses other than those in kidney and liver; for
metabolism the issue is the extent of reductive metabolism in vivo; for pharmacokinetics
they are the incomplete understanding of the role of corn oil, the characterization of
metabolism in humans, and the question of the most appropriate internal dose measure for
understanding the relation of tissue toxic effects as a function of dosing regime; for
toxicity mechanisms the issues are the possibility of significant genotoxicity in the target
organs for carcinogenicity, the mechanistic role, if any, of reductive metabolism, the
correlation of cytotoxicity and tumors in the bioassays, the dose-response patterns for
cytotoxicity, the nature of the evidence that cytotoxicity is necessary for chloroform’s
tumorigenesis (and not just ancillary), and the link of quantitative measures of cell
proliferation to quantitative changes in cancer risk.
Response: Several of these issues were further discussed as suggested by the
commentor, but others were given only cursory treatment, due in part to the overall
limitations of this text as a risk management tool.
Comment:
The current draft Risk Characterization refers to some of these issues (although not all of
them), but it tends to do so simply by citing the existence of the arguments. As a result,
when it comes time to make choices of analytical approaches, the reader is left without a
Responses to External Revtew 5
-------
strong sense of why the Agency is choosing the path it takes or how much confidence it
invests in its chosen approach. For example, the key section on Mechanism of Toxicity
(pp. 13-17) mentions various relevant studies and notr their differing implications, but it
takes no stand on the issue of whether chloroform’s observed animal carcinogenicity
should be considered secondary to (and causally dependent on) cytotoxicity in the target
organs. Yet immediately afterward, and without any discussion, the document states (p.
19) “In view of the weight of evidence that chloroform may induce tumors by a nonlinear
mechanism (ILSI, 1997), a margin of exposure approach for dose response analysis [ sic].”
(One can easily imagine this incomplete sentence having arisen from a hesitancy between
finishing it with “is chosen” versus “is considered,” with neither being chosen in the end.)
Response: Attempts were made to clarify the Agency’s position throughout this
text.
Comment:
In my view, something much bolder is needed, even if the outcome is to boldly declare
one’s uncertainty. As I argue below, I think the Agency would err if it put all its
confidence in the existing “nonlinear” dose-response method, but if it is to do so, it should
more explicitly declare its reasoning for why such a course is appropriate. After all, it
reverses the earlier Agency stance. As it stands, the conclusions are simply (albeit
hesitantly) declared without clearly stated reasons.
Response: Statements were made more boldly throughout the text that the
nonlinear approach to the assessment of chloroform’s carcinogenic risk at low
doses was consistent with more of the data than alternative hypotheses.
Comment:
Corn Oil Gavage - A better discussion is warranted since the difference in cytotoxicity
and in tumorigenesis between drinking water and oil gavage experiments is critical to
choosing the Jorgenson study and to arguments about the internal dosimeter that is most
appropriate. On p.6 it is noted that chloroform is actually absorbed more readily from
aqueous vehicles, and the reference to the role of first-pass metabolism seems
inappropriate since it applies equally to all oral exposures. To me, the magnitude of the
corn oil effect seems puzzling and confidence in other mechanistic explanations of
cytotoxicity hinge on the strength of the dose-rate explanation. It needs to be noted as an
issue.
Response: EPA (1987) has discussed this issue and decided to use the Jorgenson
study. This text merely confirms the previous decision. This discussion is
highlighted in EPA’s IRIS (EPA, 1998).
Comment:
Responses to External Review 6
-------
Inhalation - Even for drinking water contamination, exposure by inhalation can be a
major route of uptake (as noted briefly in the document’s exposure section). Yet
elsewhere the issue of inhalation is not mentioned, despite the existence of the positive
Japanese inhalation study. Is the earlier EPA inhalation unit risk for chloroform obviated,
or does it stand in the absence of an explicit revision? How are inhalation exposures, and
combined inhalation and oral exposures, to be assessed dosimetrically?
Response: The inhalation study is not yet published, but these issues should again
be considered after it is.
Comment:
Metabolism - There are several questions that bear further discussion. No clear statement
is given regarding the weight to be given to the possibility of reductive metabolism in
vivo to dichioromethyl radicals. What is the evidence that this occurs in vivo? If the free
radicals are thought to be potentially genotoxic, where is the evidence of such
genotoxicity? (After all, the genotoxicity tests are not specific to damage caused by
oxidative metabolism; the only question is whether metabolic patterns typical of the in
vivo situation are examined. Why is there no unscheduled DNA synthesis in the liver
when animals are exposed in vivo, for instance?)
The impact of having two P450 isozymes with activity toward chloroform is not
examined. If CYP2B 1/2 are active at high doses, what does this say about relative high-
and low-dose metabolic activation?
Response: The section on metabolism was revised to further discuss the weight of
evidence for the reductive metabolic pathway. Also, text was added to further
discuss the P450 metabolism.
Conmient:
The document tends to equate reductive metabolism with genotoxicity (and hence linear
dose-response approaches) and oxidative metabolism with cytotoxicity (and nonlinear
approaches). It implies that, if one doubts the operation of reductive metabolism,
nonlinear approaches are indicated. This seems inappropriate for several reasons. First,
as noted, the in vivo genotoxicity data do not refer to which reactive moiety may be
causing effects, and the lack of effects speaks equally to the genotoxic potential of
metabolically generated free radicals and metabolically generated phosgene. Second, free
radicals attack a variety of cellular macromolecules, notably lipids, and they may serve as
agents of cytotoxicity as well as phosgene. l’his effect may, however, have different
sensitivity to the rate of metabolism, an argument that plays importantly in the choice of
internal dosimeters. Third, genotoxicity is not the only reason to rely on a linear
approach to dose-response analysis, and lack of genotoxicity should not be deemed a
sufficient reason to undertake only a margin of exposure approach.
Responses to External Review 7
-------
Response: Text changes were made throughout the document in order to break the
unduly ngorous connection between reductive metabolism and lineanty, and
oxidative metabolism and non-linearity.
Comment:
As noted above, the key studies should have the data presented in tabular form. It should
be clear how issues such as concurrent m&tality are handled. In particular, the Jorgenson
study has some issues with early mortality.
Response: The data from the cancer bioassays were presented in the test. This
text relied on EPA’s ERJS file for current interpretation of these bioassays.
Comment:
At one point, EPA was making much of tumor responses in the Jorgenson study other
than the kidney tumors in rats (e.g., possible elevation of mouse lymphomas). The
current assessment presumes that only mouse liver and rat kidney tumors are at issue. A
clear statement of the basis for choosing which tumor responses warrant investigation is
in order.
Response: These issues have been discussed previously in EPA texts. No need
existed to repeat the earlier arguments.
Comment:
Characterization should have a concise summary (probably tabular) of the results on
target organ doses (examining the contending dose metrics) in different species and
organs following chloroform administration by different routes, as implied by the
pharmacokinetic modeling. This will aid in examining the question of how species
differences in response are explained by dosimetric differences.
Response: PBPK modeling was not used in this text, other than to reference the
ILSI report. Thus, a table of these values was not considered necessary.
Furthermore, this text is a risk characterization document, geared more for review
by risk managers. Technical details are appropriately referenced in other previous
work.
Comment:
Similarly, a table should summarize the results of target organ cytotoxicity in the animal
bioassays to aid in examining the question of the correlation of cytotoxicity and
tumorigenicity. Also, graphs of the results of the series of CUT studies on labeling index
at different exposures (similar to those shown in Chiu et a!. 1996) should be shown.
These should be x-y graphs, not bar graphs, so that the shape of the labeling index
response curve with exposure is evident.
Responses to External Review 8
-------
Response. These data are interesting and cited as part of the ILSI report.
However, they are not used here and would only add text to an already “full” risk
characterization document.
Comment:
The equation of genotoxicity with the reductive pathway, evident in the document is
misguided in my view, as argued earlier.
Response: Text was changed to reflect this concern.
Comment:
From the foregoing, I imagine it is evident that I feel that the “nonlinear” approach as set
out in the Proposed Guidelines is too readily invoked and that it poorly accommodates
what we know and do not know about the carcinogenic process, even when that process is
thought to be secondary to a nonlinear biological precursor effect such as cytotoxicity. In
criticizing the application of this approach to chloroform, I am not invoking genotoxicity
nor even denying cytotoxicity as the key mode of action.
The rationale behind the nonlinear method is that at some moderately low doses the dose
response curve for tumors becomes sufficiently steep in its decline with exposure level
that smaller exposures will be without meaningful risk. If human exposures are well
below those associated with tumors in animals, it is supposed that one can be assured that
the intervening steep drop is sufficient to render the low exposures safe. The unanswered
question is how far below is “well below” and what evidence can be adduced to judge the
adequacy of the size of the margin of exposure.
Logically, the approach should be to examine the dose-response pattern for the biological
effect to which carcinogenesis is deemed secondary, and which is presumed to have an
actual or practical threshold. In practice, the method assumes that this point of steep
decline to trivial risk levels will be found somewhere slightly below the ED 10 for tumors.
The Characterization document accepts this presumption quite readily; it calculates a
point of departure as the EDIO (or LED1O) for rat kidney tumors from the Jorgenson
drinking water study, and then concentrates its discussion to the size of uncertainty
factors that might be applied. Actual dose-response patterns are not examined, but they
are presumed to be sufficiently nonlinear. Also, the metabolic activation of chloroform is
stated by be (sic) “nonlinear as a function of dose” (p.28). A plot of these data do not
necessarily lead to this conclusion.
In sum, these data seem to provide a poor basis for invoking a method that presumes that
risk is highly nonlinear just below the ED1O for tumors. This shows that a good deal
more effort needs to be put into defining an appropriate point of departure, and that the
Responses to External Review 9
-------
adequacy of the point chosen needs to be examined and defended in a way that has not
yet been done.
Response: A number of interesting and important issues are raised in this
comment that apply to chloroform, but perhaps more importantly, to the ongoing
revisions of the cancer guidelines of EPA. Specific text changes were made in the
chloroform text to downplay the non-linearity with dose, especially in the area of
metabolism. However, the recommendation to use an MOE approach rather than a
linear extrapolation is still retained with chloroform, because, of the two available
methods, this seems to best fit the available data. This judgment is consistent with
EPA’s 1996 guidelines.
Comment:
I question whether the uncertainty factor approach is appropriate at all. Nonetheless, one
can exaniine the document’s reasoning for the size of the various factors.
3-FoldFactorfor Slope
Shallower slopes warrant larger margins, but how big should such a margin be?
Allowing 3 -fold below an ED 10 for slope (p. 19) with a probit slope of 0.62 leads to a
dose that has a projected risk of 2.5%. (A 10-fold factor only goes down to 0.3%, as
opposed to the linear met1 od’s 1.0%). Since this is the only factor that is designed to
address the non linearity and drop to acceptable doses, it would seem that in the present
case a factor of 100 might be just marginally enough (i.e., going down to a 10-4 risk on
the fitted nonlinear probit curve).
The point of the nonlinear method is to avoid extrapolating a fitted curve down to low
dose. But 3-fold below the ED1O is still within the observed range (it is about at the level
of the second dose group in the Jorgenson study) and is in the range where increasing
tumor risk is empirically evident.
I 0-Fold Factor for Severity of Endpoint
I do not understand the logic of this factor. It has not been the usual practice to assume
site concordance and to presume that the potential human endpoint will be for the same
tumor in terms of histological type and malignant aggressiveness. If the argument is that
the rat tumors were frequently benign, is the presumption that a factor needs to be
included in case the human tumors are less so? If the rat tumors were all malignant,
would this factor then be smaller (on the premise that human tumors could only lie less
malignant, not more)? I also do not understand why the “closeness” of the cytotoxicity
and tumorigenicity curves (an odd concept in itself, since one is continuous and the other
quantal) leads to a lesser factor.
Responses to External Review 10
-------
I 0-Fold Factor Intra-Human Variability
This presumably means for among-human variability (rather than intra-individual
variability). When assessing population risk, one should presumably not use this factor,
since it is to allow for particular vulnerability in individual risk in some people, but in a
population such high individual risks on the part of some will be balanced by lower than
average risks in others. The relative CYP2E I activity is a poor basis for judging such
variability, since it is not established that amount of metabolic activation is proportional
to this activity (although the PBPK models could examine this question).
I-Fold Factor Persistence
Since the animal studies are for lifetime exposure, it is not clear what a correction for
persistence is for. Presumably, this is covered by the dosimetry assumptions about
equivalently toxic dose rates in humans and animals.
I 0-Fold Factor Interspecies Variation
This has the same ambiguity that the analogous factor has in noncancer risk assessment.
Is the factor an adjustment for an expected sensitivity difference (as implied by the
discussion of the surface area adjustment to doses as obviating it) or is it an allowance for
the possibility that for this particular agent, humans are particularly susceptible owing to
some cause not covered by the dosimetry or other adjustments that are made? I note
again that the 2/3-power adjustment mentioned is at odds with the methodology specified
in the Guidelines Proposal. The reference to assumed kinetic differences is inappropriate,
since the factor is applied to account for general differences in the pace of physiological,
processes, including those that control the pharmacodynamic response.
The result of applying these factors is that the “acceptable” dose is only 40-fold higher
than the dose producing a 10-6 risk under the linear method. Given that the allowance for
slope is clearly too small, this raises the question of whether the nonlinear method is
meaningfully different from the linear method. It differs mostly in relying on ill-defined
arguments for how large the margin of exposure must be in order to be deemed safe.
Response: A number of interesting and important issues are also raised in this
comment that apply to chloroform, but perhaps more importantly, to the ongoing
revisions of the cancer guidelines of EPA. In part based on this comment, specific
text changes were made in the chloroform text to downplay the numerous
discussions of factors with various pieces of the data. The end result was the
choice of a 1 000-fold MOE. This choice is consistent with the overall amount of
available data for chloroform, but at the same time also accounts- for potential
sensitive populations. The use of MOE is new in EPA dose response assessment
Responses to External Review ii
-------
for carcinogens. Improvements to the process are likely to occur as additional
chemicals are considered for this procedure.
Comment:
It seems unfortunate not to explore the implications of internal dose measures more
thoroughly. A lot has been done in this arena, and an interesting analysis was conducted
in the ILSI Panel document. In that case, I had objections to the use of the ED1O as a
point of departure for similar reasons that I raise above. Nonetheless, the dose-response
analysis shows consistency among different data sets when internal doses are used,
allowing more confidence in the description of the relationship in the observed range. I
refer to my comments on the IS! document on this matter.
Response: This risk characterization text was intended to synthesize previous
work into a coherent picture for risk managers, and not to enhance the technical
work of the ILSI or other previous EPA deliberations. While our understanding
of the chloroform data base still can be improved, sufficient data are available to
develop a non-linear approach to the estimation of low dose cancer risk for
chloroform within EPA’s new cancer guidelines.
Comment:
The case of chloroform is complex, difficult, and challenging. It is a formidable task to
bring rigorous discussion of the many issues into a single, readable, useful risk
characterization. The authors of the Risk Characterization have made a good try, but they
have been hampered by the lack of defined stances on the issues and clearly articulated
bases for analytical approaches among the source material. The risk characterization is
supposed to summarize and communicate these judgments, not to make them de novo. I
am sympathetic with the challenge faced by the authors, but I am compelled to review the
document according to what ideally could be accomplished, and what ideally should be
covered in a fully developed risk assessment for a compound with a database as rich as
that of chloroform.
Many of my comments on the nonlinear method and the margin of exposure stem from
doubts about the method as set out in the Proposed Guidelines, especially regarding the
wisdom of using an ED1O as a point of departure.
I do not disagree markedly with the basic conclusions of the Risk Characterization
regarding the likely levels of human risk from exposures as actually experienced,
although I feel that the invocation of marked non linearity between the inimal response
levels and the human exposures is somewhat misleading. I would feature both the linear
and nonlinear approaches, with a fuller discussion of the ways in which each of these may
be misleading if taken too literally. It is a shame that the risk characterization has not
Responses to External Review 12
-------
examined the attempt by Evans and coworkers (1994, Regul. Toxicol. Pharmacol. 20 15
-36) to create a distributional approach to chloroform potency assessment, based on
expert judgment regarding the relative credence to be put on a variety of analytical
approaches, including linear and nonlinear, administered doses versus internal
pharmacokinetic doses, cytotoxicity and genotoxicity, and other factors. This paper
showed that no single approach suffices, but that a collection of approaches, each
weighted by its judged likelihood to be appropriate, captures the full weight of evidence
and span of possibilities.
I do feel that the document would substantially profit from more discussion of the pros
and cons of various approaches and a more forthright discussion of how the Agency
views each of the contentious and problematic issues. The document should show more
explicitly what was done and should discuss more thoroughly why things were done in
that way.
Response: These comments were used to change the text in several ways in order to
create an approximation of what the commentor intended, within the overall scope
and intent of a risk characterization document.
Responses to External Review 13
-------
R. Julian Preston, Ph.D.
Chemical Industry Institute of Toxicology
Comment:
As discussed there are no data that indicate that chloroform induces tumors in humans.
However, there is sufficient evidence showing that chloroform can induce tumors in
rodents (rats and mice) but that the particular response can vaiy with route of exposure,
administration vehicle and strain of rodent. Thus, establishing which data set is the most
appropriate one for use in estimating cancer risk to humans is not particularly easy. The
report utilizes the incidence of renal tumors in male rats exposed to chloroform in
drinking water (Jorgenson et al, 1985). The justification is not very clearly laid out,
although it would appear to be an appropriate choice, especially absent a published report
of the data from Matsushima at a!. (1994).
Response: Text was improved to further justify the chosen study. Further
justification can be found on EPA’s IRIS (EPA, 1998).
Comment:
While evidence for relevant effects in humans suggesting carcinogenic effects of
chloroform are not strong, the fact that chloroform induces male liver tumors in female
Wistar rats and kidney tumors in BDF1 male mice and Osborne-Mendel rats is suggestive
of the potential for human carcinogenicity in a qualitative sense. At the same time the
variability of response according to strain, especially absence in some strains, is worthy
of somewhat more discussion in the report. In particular, there is a need to discuss the
choice of data set for modeling more detail.
Response: These points were further discussed in the text as indicated.
Comment:
The very weak genotoxicity or absence of genotoxicity argues quite effectively that the
pathway to a tumor involves the indirect production of mutations, bearing in mind that
mutations have to be formed by some mechanism in order that a tumor can develop via its
multi-step process.
Response: Text was enhanced to address this issue. Background mutations were
suggested as one source of initiated cells.
Comment:
The discussion of mode of action in the report is generally clear and accurate. The rather
close relationship between regenerative cell proliferation as a result of cytotoxicity and
the induction of tumors suggests a link between the two. However, it needs to be made
Responses to External Review 14
-------
clearer that cell proliferation is perhaps a necessary step but clearly not a sufficient step
by itself for tumor induction. Something else is needed albeit weak genotoxicity,
oxidative DNA damage, other secondarily-produced DNA damages or even an
error-prone replication net result would seem to result in an overall dose response for
tumors that is non linear and perhaps reflected by the curve for cell proliferation. This
avoids the concern that is often expressed that tumors do not always arise when cell
proliferation is increased (and perhaps also the corollary that tumors arise absent cell
proliferation) by chloroform treatment.
Response: Text was added to clarify the distinction between the induction and the
subsequent growth of tumors induced by the cell proliferation. Text was also
added to further enhance the issue of oxidative DNA damage.
Comment:
This is a clearly written, concise report that follows a defined path to arrive at a risk
assessment for human rental tumors from drinking water exposure. The choice of the
rodent cancer data for the assessment needs to be justified more clearly. The mode of
action needs to describe a process whereby tumors will result, given that cancer is a
genetic disease. The Report successfully follows an approach consistent with EPA’s 1996
Proposed Guidelines for Carcinogen Assessment.
Response: The choice and justification of rodent model was more clearly
described; reference was made to EPA’s IRIS where further support is shown.
Statements were added to more clearly describe the cancer process as one that
involves several steps, one of which is genetic damage.
Responses to External Review 15
-------
------- |