1 NCEA-C-1763
2 November 2006
3
Toxicological Reviews of Cyanobacterial
Toxins: Cylindrospermopsin
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23 National Center for Environmental Assessment
24 Office of Research and Development
25 U.S. Environmental Protection Agency
26 Cincinnati, OH 45268
-------
1 DISCLAIMER
2
3
4 This report is an external draft for review purposes only and does not constitute Agency
5 policy. Mention of trade names or commercial products does not constitute endorsement or
6 recommendation for use.
ii DRAFT: DO NOT CITE OR QUOTE
-------
1 TABLE OF CONTENTS
2
3
4 Page
5
6 LIST OF TABLES v
7 LIST OF FIGURES vi
8 LIST OF ACRONYMS vii
9 PREFACE viii
10 AUTHORS, CONTRIBUTORS AND REVIEWERS ix
11 ACKNOWLEDGMENTS x
12
13 1. INTRODUCTION 1
14
15 2. CHEMICAL AND PHYSICAL INFORMATION 3
16
17 3. TOXICOKINETICS 4
18 3.1. ABSORPTION 4
19 3.2. DISTRIBUTION 4
20 3.3. METABOLISM 5
21 3.4. ELIMINATION 5
22 3.5. PHYSIOLOGICALLY BASED TOXICOKINETIC MODELS 6
23
24 4. HAZARD IDENTIFICATION 7
25 4.1. STUDIES IN HUMANS - EPIDEMIOLOGY, CASE REPORTS,
26 CLINICAL CONTROLS 7
27 4.2. ACUTE, SHORT-TERM, SUBCHRONIC AND CHRONIC
28 STUDIES AND CANCER BIO AS SAYS IN ANIMALS - ORAL
29 AND INHALATION 8
30 4.2.1. Oral Exposure 9
31 4.2.1.1. Acute Studies 9
32 4.2.1.2. Short-Term Studies 10
33 4.2.1.3. Subchronic Studies 11
34 4.2.1.4. Chronic Studies 13
35 4.2.2. Inhalation Exposure 13
36 4.3. REPRODUCTIVE/DEVELOPMENTAL STUDIES - ORAL AND
37 INHALATION 13
38 4.4. OTHER STUDIES 14
39 4.4.1. Effects by Parenteral Exposure 14
40 4.4.1.1. Studies of Purified Cylindrospermopsin 14
41 4.4.1.2. Cell Extract Studies 14
42 4.4.2. Immunotoxicity 15
43 4.4.3. Tumor Initiation 15
44 4.4.4. Genotoxicity 16
iii DRAFT: DO NOT CITE OR QUOTE
-------
1 TABLE OF CONTENTS cont.
2
3
4 Page
5
6 4.5. MECHANISTIC DATA AND OTHER STUDIES IN SUPPORT OF
7 THE MODE OF ACTION 16
8 4.5.1. Liver Toxicity 16
9 4.5.2. Kidney Toxicity 18
10 4.5.3. Interactions with DNA and RNA 19
11 4.5.4. Structure-Activity Relationships 20
12 4.6. SYNTHESIS AND EVALUATION OF MAJOR NONCANCER
13 EFFECTS 21
14 4.6.1. Oral 21
15 4.6.2. Inhalation 23
16 4.6.3. Mode of Action Information 23
17 4.7. WEIGHT-OF-EVIDENCE EVALUATION AND CANCER
18 CHARACTERIZATION 24
19 4.7.1. Summary of Overall Weight-of-Evidence 24
20 4.7.2. Synthesis of Human, Animal and Other Supporting Evidence 24
21 4.8. SUSCEPTIBLE POPULATIONS AND LIFE STAGES 24
22 4.8.1. Possible Childhood Susceptibility 24
23 4.8.2. Possible Gender Differences 25
24 4.8.3. Other Possible Susceptible Populations 25
25
26 5. DOSE-RESPONSE ASSESSMENTS 26
27 5.1. NARRATIVE DESCRIPTION OF THE EXTENT OF THE
28 DATABASE 26
29 5.2. ORAL REFERENCE DOSE (RfD) 26
30 5.2.1. Data Considered in Deriving Reference Values 26
31 5.2.2. Acute Duration 26
32 5.2.2.1. Choice of Principal Study and Critical Effect - with
33 Rationale and Justification 27
34 5.2.3. Short-Term Duration 27
35 5.2.3.1. Choice of Principal Study and Critical Effect-with
36 Rationale and Justification 27
37 5.2.4. Subchronic Duration 27
38 5.2.4.1. Choice of Principal Study and Critical Effect - with
39 Rationale and Justification 27
40 5.2.4.2. Methods of Analysis - Including Models (PBPK, MBD, etc.) ...28
41 5.2.4.3. RfD Derivation - Including Application of UFs 28
42 5.2.5. Chronic Duration 32
43 5.2.5.1. Choice of Principal Study and Critical Effect - with
44 Rationale and Justification 32
45 5.2.6. Route-to-Route Extrapolation 33
46 5.3. INHALATION REFERENCE CONCENTRATION (RfC) 33
47 5.4. CANCER ASSESSMENT 33
iv DRAFT: DO NOT CITE OR QUOTE
-------
1 TABLE OF CONTENTS cont.
2
3
4 Page
5
6 6. MAJOR CONCLUSIONS IN THE CHARACTERIZATION OF HAZARD
7 AND DOSE RESPONSE 34
8 6.1. HUMAN HAZARD POTENTIAL 34
9 6.2. DOSE RESPONSE 34
10
11 7. REFERENCES 35
DRAFT: DO NOT CITE OR QUOTE
-------
1 LIST OF TABLES
2
3
4 No. Title Page
5
6 4-1 Tumor Initiating Activity of C. raciborskii Extracts 16
7
8 4-2 Summary Results of Major Oral Toxicity Studies of Pure
9 Cylindrospermopsin in Experimental Animals 22
10
11 5-1 Relative Kidney Weights in Mice Exposed to Purified
12 Cylindrospermopsin for 11 Weeks 29
13
14 5-2 Summary of Benchmark Dose Modeling (Relative Kidney Weight) 30
vi DRAFT: DO NOT CITE OR QUOTE
-------
1 LIST OF FIGURES
2
3
4
5 No. Title Page
6
7 2-1 Chemical Structure of Cylindrospermopsin 3
8
9 5-1 Linear Model Fit to Relative Kidney Weight Data (High Dose Group Dropped) 31
vii DRAFT: DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
AIC
BMD
BMDL
BMDS
BMR
CYP450
DMSO
EPA
GFR
HPLC
i.p.
LD50
LOAEL
NOAEL
PBPK
POD
RfC
RfD
ROS
THP
TPA
UF
LIST OF ABBREVIATIONS
Aikake's Information Criteria
Benchmark dose
Statistical lower confidence limit on the benchmark dose
Benchmark dose software
Benchmark response
Cytochrome P-450
Dimethyl sulfoxide
Environmental Protection Agency
Glomerular filtration rate
High performance liquid chromatography
Intraperitoneal
Dose lethal to 50% of the population
Lowest-observed-adverse-effect level
No-observed-adverse-effect level
Physiologically based pharmacokinetic
Point of departure
Reference concentration
Reference dose
Reactive oxygen species
Tamm-Harsfall protein
O-Tetradecanoylphorbol 13-acetate
Uncertainty factor
Vlll
DRAFT: DO NOT CITE OR QUOTE
-------
1 PREFACE
2
3
4 The Safe Drinking Water Act (SDWA), as amended in 1996, requires the Environmental
5 Protection Agency (EPA) to publish a list of contaminants that, at the time of publication, are not
6 subject to any proposed or promulgated national primary drinking water regulations, are known
7 or anticipated to occur in public water systems, and may require regulations under SDWA. This
8 list, known as the Contaminant Candidate List (CCL), was first published in 1998 and then again
9 in 2005. The 1998 and 2005 CCLs include "cyanobacteria (blue-green algae), other freshwater
10 algae, and their toxins" as microbial contaminants.
11
12 In 2001, a meeting was held among EPA, researchers from the drinking water industry,
13 academia and government agencies with expertise in the area of fresh water algae and their
14 toxins. The goal of this meeting was to convene a panel of scientists to assist in identifying a
15 target list of algal toxins that are likely to pose a health risk in source and finished waters of the
16 drinking water utilities in the U.S. Toxin selection was based on four criteria: health effects,
17 occurrence in the United States, susceptibility to drinking water treatment and toxin stability.
18 Cylindrospermopsin was identified at this meeting as being a toxin of high priority based on
19 those criteria.
20
21 The National Center for Environmental Assessment has prepared this Toxicological
22 Review of Cyanobacterial Toxins: Cylindrospermopsin as one in a series of dose-response
23 assessments to support the health assessment of unregulated contaminants on the CCL. The
24 purpose of this document is to compile and evaluate the available data regarding
25 Cylindrospermopsin toxicity to aid the Office of Water in regulatory decision making. It is not
26 intended to be a comprehensive treatise on the chemical or toxicological nature of
27 Cylindrospermopsin.
28
29 In Section 6, Major Conclusions in the Characterization of Hazard and Dose Response,
30 EPA has characterized its overall confidence in the quantitative and qualitative aspects of the
31 hazard and dose response by addressing knowledge gaps, uncertainties, quality of data and
32 scientific controversies. The discussion is intended to convey the limitations of the assessment
33 and to aid and guide the Office of Water in the ensuing steps of the human health risk assessment
34 of Cylindrospermopsin.
ix DRAFT: DO NOT CITE OR QUOTE
-------
1 AUTHORS, CONTRIBUTORS AND REVIEWERS
2
3
4 AUTHORS
5
6 Belinda Hawkins, Ph.D. (Chemical Manager)
7 National Center for Environmental Assessment
8 Office of Research and Development
9 U.S. Environmental Protection Agency
10 Cincinnati, OH
11
12 Stephen Bosch
13 Syracuse Research Corporation
14 Syracuse, NY
15
16 Marc Odin
17 Syracuse Research Corporation
18 Syracuse, NY
19
20 David Wohlers
21 Syracuse Research Corporation
22 Syracuse, NY
23
24
25 REVIEWERS
26
27 INTERNAL EPA REVIEWERS
28
29 Joyce Donohue, Ph.D.
30 Office of Water
31 Washington, DC
32
33 Elizabeth Hilborn, D.V.M., M.P.H.
34 National Health and Environmental Effects Research Laboratory
35 Office of Research and Development
36 Research Triangle Park, NC
37
38 James Sinclair, Ph.D.
39 Office of Water
40 Cincinnati, OH
41
42 Jeff Swartout
43 National Center for Environmental Assessment
44 Office of Research and Development
45 Cincinnati, OH
46
DRAFT: DO NOT CITE OR QUOTE
-------
1 EXTERNAL REVIEWERS
2
3
4 ACKNOWLEDGMENTS
5
xi DRAFT: DO NOT CITE OR QUOTE
-------
1 1. INTRODUCTION
2
3
4 This toxicological review presents background and justification for hazard and dose-
5 response assessments of cylindrospermopsin. U.S. Environmental Protection Agency (EPA)
6 toxicological reviews may include oral reference doses (RfD) and inhalation reference
7 concentrations (RfC) for chronic and less-than-lifetime exposure durations and a carcinogenicity
8 assessment.
9
10 The RfD and RfC provide quantitative information for use in risk assessments for health
11 effects known or assumed to be produced through a nonlinear (possibly threshold) mode of
12 action. These reference values are defined as an estimate of an exposure, designated by duration
13 and route, to the human population (including susceptible subgroups), that is likely to be without
14 an appreciable risk of adverse effects. Reference values may be derived for acute (<24 hours),
15 short-term (up to 30 days), subchronic (up to 10% of average lifespan) and chronic (up to
16 lifetime) exposures, all considered to be continuous exposures throughout the duration specified.
17 A reference value is derived from a BMDL (a statistical lower confidence limit on the
18 benchmark dose), a no-observed-adverse-effect level (NOAEL), lowest-observed-adverse-effect
19 level (LOAEL) or other suitable point of departure with uncertainty/variability factors applied to
20 reflect limitations of the data used. The RfD is expressed in units of mg/kg-day, and the RfC in
21 units of mg/m3.
22
23 The carcinogenicity assessment provides information on the carcinogenic hazard
24 potential of the substance in question and quantitative estimates of risk from oral exposure and
25 inhalation exposure. The information includes a weight-of-evidence judgment of the likelihood
26 that the agent is a human carcinogen and the conditions under which the carcinogenic effects
27 may be expressed. Quantitative risk estimates are presented in three ways. The slope factor is
28 the result of application of a low-dose extrapolation procedure and is presented as the risk per
29 mg/kg-day. The unit risk is the quantitative estimate in terms of either risk per |J,g/L drinking
30 water or risk per |J,g/m3 air breathed. Another form in which risk is presented is a drinking water
31 or air concentration providing cancer risks of 1 in 10,000; 1 in 100,000; or 1 in 1,000,000.
32
33 Development of these hazard identification and dose-response assessments for
34 cylindrospermopsin has followed the general guidelines for risk assessment as set forth by the
35 National Research Council (NRC, 1983). EPA guidelines that were used in the development of
36 this assessment include the following: Guidelines for the Health Risk Assessment of Chemical
37 Mixtures (U.S. EPA, 1986a), Guidelines for Mutagenicity Risk Assessment (U.S. EPA, 1986b),
3 8 Guidelines for Developmental Toxicity Risk Assessment (U.S. EPA, 1991), Guidelines for
39 Reproductive Toxicity Risk Assessment (U.S. EPA, 1996), Guidelines for Neurotoxicity Risk
40 Assessment (U.S. EPA, 1998a), Guidelines for Carcinogen Assessment (U.S. EPA, 2005a),
41 Supplemental Guidance for Assessing Susceptibility from Early-Life Exposure to Carcinogens
42 (U.S. EPA, 2005b), Recommendations for and Documentation of Biological Values for Use in
43 Risk Assessment (U.S. EPA, 1988), (proposed) Interim Policy for Particle Size and Limit
44 Concentration Issues in Inhalation Toxicity (U.S. EPA, 1994a), Methods for Derivation of
45 Inhalation Reference Concentrations and Application of Inhalation Dosimetry (U.S. EPA,
46 1994b), Use of the Benchmark Dose Approach in Health Risk Assessment (U. S. EPA, 1995),
DRAFT: DO NOT CITE OR QUOTE
-------
1 Science Policy Council Handbook: Peer Review (U.S. EPA, 1998b, 2000a), Science Policy
1 Council Handbook. Risk Characterization (U.S. EPA, 2000b), Benchmark Dose Technical
3 Guidance Document (U.S. EPA, 2000c) and Supplementary Guidance for Conducting Health
4 Risk Assessment of Chemical Mixtures (U.S. EPA, 2000d) and^4 Review of the Reference Dose
5 and Reference Concentration Processes (U.S. EPA, 2002).
6
7 Literature searches were conducted for studies relevant to the derivation of toxicity and
8 carcinogenicity values for cylindrospermopsin. The following databases were searched:
9 MEDLINE (PubMed), TOXLINE, BIOSIS, CANCERLIT, TSCATS, CCRIS, DART/ETIC,
10 EMIC, GENETOX, HSDB and RTECS. The relevant literature was reviewed through May
11 2006.
DRAFT: DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
2. CHEMICAL AND PHYSICAL INFORMATION
Cylindrospermopsin is a naturally occurring toxin produced by particular strains of
Cylindrospermopsis raciborskii and at least four other freshwater cyanobacterial species,
including Umezakia natans, Aphanizomenon ovalispomm, Anabaena bergii and Raphidiopsis
curvata (Fastner et al., 2003). The chemical structure of Cylindrospermopsin was not elucidated
until 1992. It consists of a tricyclic guanidine moiety combined with hydroxymethyluracil
(Figure 2-1) (Humpage and Falconer, 2003; Ohtani et al., 1992), has a molecular formula of
CisH^iNsOyS and a molecular weight of 415.43 (Lewis, 2000). It is zwitterionic (i.e., a dipolar
ion with localized positive and negative charges) (Ohtani et al., 1992). Deoxycylindro-
spermopsin, an analog of Cylindrospermopsin in which the hydroxyl group on the uracil bridge
(C-7) has been removed, has been isolated from C. raciborskii and R. curvata (Li et al., 2001;
Norris et al., 1999). Another structural variant of Cylindrospermopsin, 7-epicylindrospermopsin,
was isolated from A. ovalisporum (Banker et al., 2000).
Figure 2-1. Chemical Structure of Cylindrospermopsin*
* Conformations of steriocenters within the structure are indicated as either R or S. The numbers 7 and 12 indicate
carbon positions for identification purposes.
Cylindrospermopsin is a white powder that is highly soluble in water (Ohtani et al., 1992;
Sigma, 2006). It is also soluble in dimethylsulfoxide (DMSO) and methanol (Sigma, 2006).
Cylindrospermopsin is chemically stable in sunlight, at high temperatures and through a wide
range of pH values (Chiswell et al., 1999). Additional chemical and physical property data are
not available in the open literature for Cylindrospermopsin (HSDB, 2006; Lewis, 2000; O'Neil,
2001). This substance is produced on a small scale for research purposes (Sigma, 2006).
DRAFT: DO NOT CITE OR QUOTE
-------
1 3. TOXICOKINETICS
2
3
4 3.1. ABSORPTION
5
6 No quantitative data were located regarding the rate or extent of absorption of
7 cylindrospermopsin in humans or animals following oral, inhalation or dermal exposure.
8 Absorption of cylindrospermopsin from the gastrointestinal tract of mice is demonstrated by the
9 induction of hepatic and other systemic effects in 14-day and 11-week oral toxicity studies of
10 pure cylindrospermopsin (Humpage and Falconer, 2003; Shaw et al., 2000, 2001) (see Section
11 4.2.1).
12
13 3.2. DISTRIBUTION
14
15 No information was located regarding the tissue distribution of cylindrospermopsin
16 following oral, inhalation or dermal exposure. The distribution and elimination of
17 intraperitoneally (i.p.) administered 14C-cylindrospermopsin (>95% pure; extracted and purified
18 from lyophilized C. raciborskii cells) in normal saline was studied in male Quackenbush mice in
19 a series of experiments using sublethal and lethal dose levels of the chemical (Norris et al.,
20 2001). In one experiment, four mice were given a single sublethal dose of 0.1 mg/kg, and urine
21 and feces were collected for the following 48 hours. Most of the 14C was eliminated in the urine
22 and feces, as discussed in Section 3.4. Analysis of liver, kidneys and spleen at 48 hours showed
23 mean 14C recovery of 13.1% of the dose in the liver and <1% in the other tissues. Total recovery
24 of radiolabel was 85-90% of the administered dose in each of the four mice.
25
26 The second experiment reported by Norris et al. (2001) included 12 mice administered a
27 single 0.2 mg/kg dose of 14C-cylindrospermopsin, which is the approximate median lethal i.p.
28 dose (Norris et al., 2001). 14C content was determined in the urine and feces in all animals after
29 12 and 24 hours, and in the liver, kidneys and spleen in five mice that were euthanized after 5-6
30 days due to toxicity (effects not specified) and after 7 days in the surviving 7 mice that had no
31 signs of toxicity. Most of the 14C was eliminated in the urine and feces, as discussed in Section
32 3.4. The overall mean recoveries of 14C in the liver, kidneys and spleen after 5-7 days were 2.1,
33 0.15 and <0.1% of the dose, respectively. Comparison of data from four mice with signs of
34 toxicity and four mice without signs of toxicity showed no clear relationship between toxicity
35 and patterns of tissue distribution, although a trend toward decreased liver retention in the
36 surviving mice was suggested.
37
38 Norris et al. (2001) reported a third experiment, in which excretion and tissue distribution
39 were assessed in four mice that were given a 0.2 mg/kg i.p. dose of 14C-cylindrospermopsin and
40 evaluated after 6 hours (Norris et al., 2001). 14C was detected in all tissues that were examined
41 (liver, kidney, heart, lung, spleen, blood and bile), but occurred predominantly in the liver and
42 kidneys (20.6 and 4.3% of the dose, respectively). Approximately 60% of the administered dose
43 of 14C was eliminated in the urine and feces (see Section 3.4).
44
DRAFT: DO NOT CITE OR QUOTE
-------
1 3.3. METABOLISM
2
3 The distribution and elimination of i.p administered 14C-cylindrospermopsin (>95% pure;
4 extracted and purified from lyophilized C. raciborskii cells) in saline was studied in a series of
5 mouse experiments (Norris et al., 2001), as detailed in Sections 3.2 and 3.4. Urine, fecal, liver
6 and kidney samples from these studies were extracted with methanol to precipitate proteins, and
7 the 14C in the supernatant was fractionated using high performance liquid chromatography
8 (HPLC) for the detection of metabolites. No attempt was made to fractionate or otherwise
9 identify the 14C in the protein precipitate. Analysis of methanol extracts of urine samples
10 collected for 12 hours following a single dose of 0.1 mg/kg (4 mice) or 0.2 mg/kg (12 mice)
11 suggested that a large part (72%) of the excreted 14C was present as cylindrospermopsin (as
12 determined by retention times). Some (-23.5%) of the urinary 14C was detected in protein
13 precipitated by the methanol, suggesting the presence of a protein-bound metabolite. The
14 authors did not indicate whether the level of protein in the urine was normal or abnormal. Most
15 (94.3%) of the 14C in an aqueous extract of the feces had the same retention time as
16 cylindrospermopsin, but only one mouse dosed with 0.2 mg/kg was tested. Analysis of liver
17 tissue showed the presence of 14C in both methanol extract and protein precipitate. When
18 fractionated by HLPC, the extracted 14C had the same elution characteristics seen in some of the
19 urine methanol extracts, suggesting the presence of the same metabolite. The authors could not
20 rule out the possibility that the non-extractable 14C in the liver was protein-bound
21 cylindrospermopsin itself, although the evidence for metabolic activation of cylindrospermopsin
22 in other studies (Runnegar et al., 1995; Shaw et al., 2000) suggested that it might also be a
23 metabolite. The methanol-extractable metabolite was not found in kidney tissue. No
24 identification of metabolites was performed.
25
26 There is evidence indicating that the hepatic cytochrome P-450 (CYP450) enzyme
27 system is involved in the metabolism and toxicity of cylindrospermopsin. As discussed in
28 Section 4.5.1, pretreatment of hepatocytes with known inhibitors of CYP450 diminished the in
29 vitro cytotoxicity of cylindrospermopsin (Froscio et al., 2003; Runnegar et al., 1995). Similarly,
30 pretreatment of mice with a CYP450 inhibitor protected against the acute lethality of
31 cylindrospermopsin (Norris et al., 2002). Additionally, a main target of cylindrospermopsin
32 toxicity is the periacinar region of the liver, which is where CYP450-mediated xenobiotic
33 metabolism occurs (Shaw et al., 2000, 2001).
34
35 3.4. ELIMINATION
36
37 No information was located regarding the elimination of cylindrospermopsin following
38 oral, inhalation or dermal exposure. The elimination of i.p administered 14C-cylindrospermopsin
39 (>95% pure; extracted and purified from lyophilized C. raciborskii cells) in saline was studied in
40 male Quackenbush mice in a series of experiments using sublethal and lethal dose levels of the
41 chemical (Norris et al., 2001). In one experiment, four mice were given a single sublethal dose
42 of 0.1 mg/kg, and urine and feces were collected for the following 48 hours. The mean
43 cumulative excretion of 14C in the first 12 hours after dosing was 62.8% of the administered dose
44 in the urine and 15.5% in the feces. There was little additional excretion of 14C in either the
45 urine or feces following 12 additional hours. The 15.5% mean fecal excretion value reflects a
46 very high fecal excretion in one of the four animals (nearly 60% of the dose compared to <5% in
DRAFT: DO NOT CITE OR QUOTE
-------
1 the other mice); the authors considered the possibility that the high value in the one animal
2 resulted from the injection entering the upper gastrointestinal tract, but concluded that this
3 possibility was unlikely given the injection technique used, the recovery of 4.7% of the injected
4 dose in the liver after 48 hours and a similarly high fecal excretion of 14C in another animal in
5 the third experiment in this study (discussed below). Total mean recovery in the urine, feces,
6 liver, kidneys and spleen was 85-90% of the 14C dose in each of the four mice.
7
8 The second experiment reported by Norris et al. (2001) included 12 mice administered a
9 single 0.2 mg/kg dose of 14C-cylindrospermopsin, which is the approximate median lethal i.p.
10 dose (Norris et al., 2001). Five of the 12 dosed animals died within 5-6 days (signs of toxicity
11 not reported). 14C content was determined in the urine and feces in all animals after 12 and 24
12 hours. Results were similar to those obtained with a sublethal dose (reported above), except that
13 there was some continued urinary and fecal excretion over the second 12 hours of the monitoring
14 period. The mean cumulative urinary and fecal excretion of 14C was 66.0 and 5.7% of the dose
15 within 12 hours, and 68.4 and 8.5% of the dose within 24 hours, respectively. The mean total
16 recovery in the urine and feces after 24 hours was 76.9% of the administered dose. The overall
17 mean recoveries of 14C in the liver, kidneys and spleen after 5-7 days were 2.1, 0.15 and <0.1%
18 of the administered dose, respectively. Comparison of data from four mice with signs of toxicity
19 and four mice without signs of toxicity showed no clear relationship between toxicity and
20 patterns of excretion, although trends toward increased urinary excretion and decreased fecal
21 excretion in surviving mice were suggested.
22
23 Norris et al. (2001) reported a third experiment in which four mice were given a 0.2
24 mg/kg i.p. dose of 14C-cylindrospermopsin and evaluated for 6 hours (Norris et al., 2001). The
25 mean cumulative urinary and fecal excretion of 14C after 6 hours was 48.2 and 11.9% of the
26 administered dose, respectively. One of the four mice showed more than 40% of the dose in the
27 feces (additional data not reported).
28
29 3.5. PHYSIOLOGICALLY BASED TOXICOKINETIC MODELS
30
31 No physiologically based toxicokinetic models have been developed for
32 cylindrospermopsin.
DRAFT: DO NOT CITE OR QUOTE
-------
1 4. HAZARD IDENTIFICATION
2
3
4 4.1. STUDIES IN HUMANS - EPIDEMIOLOGY, CASE REPORTS, CLINICAL
5 CONTROLS
6
7 An outbreak of a hepatoenteritis-like illness occurred in 148 residents of aboriginal
8 descent in the Palm Island community in Queensland, Australia, in 1979 (Blyth, 1980; Griffiths
9 and Saker, 2003). The total number of people exposed was not reported. Of the 148 cases, 138
10 were children (mean age 8.4 years, range 2-16 years, 41% male and 59% female) and 10 were
11 adults (sex and age not reported). The majority of the cases in the outbreak, called the "Palm
12 Island mystery disease," required hospitalization. The clinical symptoms included fever,
13 headache, vomiting, profuse bloody diarrhea, hepatomegaly and renal damage as indicated by
14 loss of water, electrolytes, proteins, ketones and carbohydrates. Many of the individuals required
15 intravenous therapy for electrolyte imbalance and, in some cases, for hypovolemic and acidotic
16 shock. The findings may indicate increased susceptibility of children unless the 138 children
17 were from the households of the 10 adults (not indicated) or if there was differential exposure
18 between the children and the adults (not indicated); the child:adult ratio is approximately 14:1.
19 A few days prior to the outbreak, the major drinking water supply for the island, Solomon Dam
20 reservoir, had been treated with unreported levels of copper sulfate to control a dense algal
21 bloom; only households connected to the reservoir were affected by the illness. Retrospective
22 analyses, including epidemiological and ecological assessments, implicated the predominant
23 cyanobacterial species in the reservoir, C. raciborskii, as the likely source of the illness (Griffiths
24 and Saker, 2003; Hawkins et al., 1985). Intraperitoneal injection of cell extracts of C. raciborskii
25 from the reservoir caused damage to the liver, kidneys and other organs in mice (Hawkins et al.,
26 1985), and the toxin was later identified as cylindrospermopsin (Ohtani et al., 1992). Some
27 symptoms of acute oral exposure to high concentrations of copper sulfate, including headache,
28 nausea, vomiting and diarrhea (HSDB, 2006), are similar to those observed during the outbreak.
29 The only information that was located regarding a potential role of the copper sulfate treatment
30 in the outbreak is an indication that its algalcidal mode of action, cell lysis, could have
31 contributed to the release of cylindrospermopsin and other cellular toxins into the water
32 (Griffiths and Saker, 2003).
33
34 Hayman (1992) investigated reports of disease (sometimes called "Barcoo fever") in the
35 Australian outback dating back as far as 1887. He concluded that the reported symptoms were
36 similar to those of the Palm Island mystery disease and that they might have been caused by
37 exposure to C. raciborskii. No additional information was located regarding effects in humans
38 known or suspected to be associated with exposure to cylindrospermopsin.
39
40 An outbreak of acute liver failure occurred in patients at a renal dialysis clinic in Caruaru,
41 Brazil (Carmichael et al., 2001). Following routine hemodialysis treatment during a week in
42 February 1996, 116 of 131 patients experienced headache, eye pain, blurred vision, nausea and
43 vomiting. Subsequently, 100 of the affected patients developed acute liver failure and, of these,
44 76 died. Analysis of the clinic's water treatment system (samples of carbon, sand and
45 cation/anion exchange resin from in-house filters) for microcystins and cylindrospermopsin
46 showed the presence of both cyanotoxins. Analyses of blood sera and liver samples revealed
DRAFT: DO NOT CITE OR QUOTE
-------
1 microcystins, but not cylindrospermopsin, although the method used to extract
2 cylindrospermopsin from these samples may have been inadequate. Based on a comparison of
3 victims' symptoms and liver pathology using animal studies of microcystins and
4 cylindrospermopsin, it was concluded that the major contributing factor to death of the dialysis
5 patients was intravenous exposure to microcystins.
6
7 The skin irritant potential of cylindrospermosin was evaluated using skin-patch testing in
8 humans (Pilotto et al., 2004). Both whole and lysed preparations of laboratory-grown C.
9 raciborskii cells were applied to the skin of 50 adult volunteers using adhesive patches divided
10 into 10 individual filter pad-containing chambers. Each volunteer was exposed to one patch for
11 whole cells and one patch for lysed cells with each patch containing six cell concentrations, two
12 positive controls (1 and 5% solutions of sodium lauryl sulfate), and two negative controls
13 (culture media and an empty patch). The concentrations (densities) of cells were consistent with
14 those found in C. raciborskii-contaming water bodies used for recreational water activities.
15 Patches were removed after 24 hours and erythematous reactions were graded as 0 (no reaction
16 or erythema), 1 (minimal or very weak spotty erythema), 2 (mild diffuse erythema), 3 (moderate
17 diffuse erythema) or 4 (severe diffuse erythema with edema) by a dermatologist blinded to the
18 cell type and concentration. The distribution of clinical gradings by patch type (control/active),
19 cell type and cell concentration was assessed using logistic regression modeling. Due to a
20 relatively small number of high-level gradings, each observation was dichotomized into no
21 reaction (grade 0) and a positive reaction (1, 2, 3 or 4) prior to modeling. The subjects were
22 more likely to have skin reactions to the active patches than to the negative control patches for
23 both whole cells (odds ratio (OR) = 2.13, 95% confidence interval (CI) 1.79-4.21, p<0.001) and
24 lysed cells (OR = 3.41, 95% CI 2.00-5.84, p<0.001). The mean percentages of subjects having a
25 reaction were 20% (95% CI 15-31%) for all subjects (n=50) and 11% (95% CI 6-18%) for
26 subjects not reacting to negative controls (n=39). The irritation was mild and resolved within 24
27 to 72 hours. There was no evidence of a statistically significant increasing dose-response
28 relationship between skin reactions and increasing cell concentrations for either whole or lysed
29 cells, although there was a slight reduction in response with increasing cell concentration for the
30 whole cells (OR = 0.966, 95% CI 0.936-0.997, p = 0.03). Additionally, there was no evidence
31 for a threshold effect (i.e., a particular concentration above which there were frequent or strong
32 reactions).
33
34 4.2. ACUTE, SHORT-TERM, SUBCHRONIC AND CHRONIC STUDIES AND
35 CANCER BIOASSAYS IN ANIMALS - ORAL AND INHALATION
36
37 Toxicity studies in animals have been performed using pure cylindrospermopsin isolated
38 and purified from cell extracts of C. raciborskii or other cylindrospermopsin-producing
39 cyanobacteria. Studies have also been performed in which the administered material consisted of
40 whole cell extracts, lyophilized (freeze-dried) cells in suspension and cell-free extracts of
41 sonicated freeze-dried cells. These studies are included in this report because they contribute
42 salient information to the overall toxicological database for cylindrospermopsin. However, due
43 to confounding factors discussed below, the studies of cell extracts are not useful for dose-
44 response assessment of cylindrospermopsin and are considered supplemental information for
45 hazard identification.
46
DRAFT: DO NOT CITE OR QUOTE
-------
1 Most of the cell extract studies were performed using laboratory cultures of
2 cyanobacterial cells, but there is no clear means of predicting the cylindrospermopsin content in
3 a particular extract. Studies with cylindrospermopsin and other cyanobacterial toxins indicate
4 that growth conditions can significantly contribute to the level of toxin produced by a given
5 species and strain, and that toxin concentration can also vary depending on the method used to
6 produce a material for toxicological testing (Chiswell et al., 1999; WHO, 1999). The
7 extracellular fraction of cylindrospermopsin can sometimes exceed the intracellular fraction
8 (Griffiths and Saker, 2003). For example, at different stages of a C. raciborskii bloom,
9 extracellular cylindrospermopsin ranged from 19 to 98% of the total amount in water (Chiswell
10 et al., 1999). Similarly, during a bloom of A. ovalisporum, >85% of the cylindrospermopsin was
11 extracellular (Shaw et al., 1999). In these studies, intracellular concentration of
12 cylindrospermopsin was determined by taking the difference between the concentration in a
13 sample of filtered water and the concentration in a sample of water that was frozen to release the
14 toxin contained in the cells. Extracts obtained by removing intact cells may or may not contain
15 toxin or may have variable amounts of toxin. For example, Falconer et al. (1999) found that the
16 cylindrospermopsin content in four different batches of cell-free extracts of C. raciborskii varied
17 from 1.3 to 5.4 mg/g extract. Additionally, cell extracts containing cylindrospermopsin can also
18 contain other potentially toxic substances. The 24-hour i.p. LD50 (dose lethal to 50% of the
19 population) of purified cylindrospermopsin in male CH3 mice was 2.1 mg/kg (Ohtani et al.,
20 1992), whereas the value for a cell extract in male Swiss mice was 0.29 mg/kg (Hawkins et al.,
21 1997), nearly an order of magnitude lower. Hawkins et al. (1997) proposed that the difference in
22 potency could reflect the presence of other toxins in the cell extract that were not present in the
23 purified cylindrospermopsin (see Section 4.4.1).
24
25 4.2.1. Oral Exposure
26
27 4.2.1.1. Acute Studies
28
29 4.2.1.1.1. Studies of Purified Cylindrospermopsin
30
31 No information regarding the acute oral toxicity of purified cylindrospermopsin was
32 identified in the materials reviewed for this document.
33
34 4.2.1.1.2. Cell Extract Studies
35
36 Twelve male MF1 mice were administered a saline suspension of freeze-dried C.
37 raciborskii cells (strains PHAWT/M or PHAWT/1) by gavage in single reported doses ranging
38 from 4.4 to 8.3 mg/kg (cylindrospermopsin-equivalent), and observed for the following 8 days
39 (Seawright et al., 1999). The following dose levels were tested (one mouse per level except as
40 noted): 4.4, 5.3, 5.7 (two mice), 5.8, 6.2, 6.5, 6.7, 6.8, 6.9, 8.0 and 8.3 mg/kg; there was no
41 control group. Eight of the 12 mice died. The lowest lethal dose was 4.4 mg/kg, the highest
42 nonlethal dose was 6.9 mg/kg and the average lethal dose was approximately 6 mg/kg. Deaths
43 occurred 2-6 days after treatment, and histological examinations showed effects that included
44 fatty liver with periacinar coagulative necrosis, acute renal tubular necrosis, atrophy of the
45 lymphoid tissue of the spleen and thymus, subepicardial and myocardial hemorrhages in the
DRAFT: DO NOT CITE OR QUOTE
-------
1 heart and ulceration of the esophageal section of the gastric mucosa. Some of the animals also
2 developed thrombohemorrhagic lesions in one or both eye orbits.
3
4 An aqueous suspension of a cell-free extract of freeze-dried and sonicated C. raciborskii
5 cells (strain AWT 205) was administered to an unspecified number of male Swiss mice in a
6 single gavage dose of 1400 mg extract/kg (Falconer et al., 1999). The cylindrospermopsin
7 content of the extract was not specified, but ranged from 1.3 to 5.4 mg/g extract in concurrent i.p.
8 experiments, indicating that the cylindrospermopsin-equivalent gavage dose was likely in the
9 range of 1.8-7.6 mg/kg. This dose level was not fatal, but caused severe liver and kidney
10 pathology. Histological changes were not detailed, but patterns of damage were reported to be
11 similar to those observed following i.p. administration (see Section 4.4.1). Additional
12 information on the design and results of the oral study were not provided.
13
14 Another gavage study reported that the minimum lethal dose of a saline extract of freeze-
15 dried C. raciborskii cells (strain AWT 205) in Swiss mice was 2500 mg extract/kg (Falconer and
16 Humpage, 2001). Based on a reported cylindrospermopsin content of 5.5 mg/g extract, the
17 equivalent dose of cylindrospermopsin was 13.8 mg/kg.
18
19 Groups of four Quackenbush mice were administered a cell-free extract of freeze-dried
20 and sonicated C. raciborskii cells (strain AWT 205) in water in a single gavage dose of 0, 1, 2, 4,
21 6 or 8 mg cylindrospermopsin/kg and observed for the following 7 days (Shaw et al., 2000,
22 2001). All animals were evaluated for gross pathological and histological (liver, kidney, spleen,
23 heart, lungs and thymus) changes. Hepatic effects were observed at all dose levels, as shown by
24 foamy hepatocellular cytoplasmic changes at 1 and 2 mg/kg, lipid infiltration with some
25 hepatocyte necrosis in the periacinar region at 4 mg/kg, and uniformly pale and mottled livers
26 with lipid infiltration throughout and cell necrosis mainly in the periacinar region at 6 mg/kg.
27 Mortality occurred in 2/4 mice at 6 mg/kg (in 5 days) and 4/4 mice at 8 mg/kg (in 24-48 hours).
28 Additional information on the experimental design and results was not reported.
29
30 4.2.1.2. Short-Term Studies
31
32 4.2.1.2.1. Studies of Purified Cylindrospermopsin
33
34 Groups of four Quackenbush mice were administered purified cylindrospermopsin by
35 daily gavage for 14 days (Shaw et al., 2000, 2001). The cylindrospermopsin was purified (purity
36 not reported) from an extract of freeze-dried C. raciborskii cells (strain AWT 205). All animals
37 were evaluated for gross pathological and histological (liver, kidney, spleen, heart, lungs and
38 thymus) changes. The authors identified the following effect levels: a NOAEL of 0.05 mg
39 cylindrospermopsin/kg-day and a LOAEL of 0.15 mg cylindrospermopsin/kg-day for lipid
40 infiltration in the liver, and a NOAEL of 0.3 mg cylindrospermopsin/kg-day (highest tested dose)
41 for lymphophagocytosis in the spleen. Additional information on the experimental design and
42 results was not reported.
43
10 DRAFT: DO NOT CITE OR QUOTE
-------
1 4.2.1.2.2. Cell Extract Studies
2
3 Groups of four Quackenbush mice were administered an aqueous cell-free extract of
4 freeze-dried and sonicated C. raciborskii cells (strain AWT 205) by daily gavage for 14 days
5 (Shaw et al., 2000, 2001). All animals were evaluated for gross pathological and histological
6 (liver, kidney, spleen, heart, lungs and thymus) changes. The authors identified the following
7 effect levels: a NOAEL of 0.05 mg cylindrospermopsin/kg-day and a LOAEL of 0.15 mg
8 cylindrospermopsin/kg-day for lipid infiltration in the liver, and a LOAEL of 0.05 mg
9 cylindrospermopsin/kg-day for lymphophagocytosis in the spleen. Additional information on the
10 experimental design and results was not reported.
11
12 4.2.1.2.3. Other Studies
13
14 Six Quackenbush mice and two Wistar rats were exposed for 21 days to drinking water
15 containing 800 |J,g/L cylindrospermopsin (Shaw et al., 2000, 2001). The water was "sourced"
16 from a dammed impoundment. The reported approximate daily dose based on water
17 consumption was 0.2 mg cylindrospermopsin/kg-day in both species. Gross pathological and
18 histological (liver, kidney, spleen, heart, lungs and thymus) examinations showed no effects,
19 indicating that 0.2 mg/kg-day was a NO AEL in the rats and mice. Additional information on the
20 experimental design and results was not reported.
21
22 4.2.1.3. Subchronic Studies
23
24 4.2.1.3.1. Studies of Purified Cylindrospermopsin
25
26 Groups of male Swiss albino mice (10 per dose, 6 in the highest dose group) were
27 administered purified cylindrospermopsin in water by gavage in doses of 0, 30, 60, 120 or 240
28 |j,g/kg-day for 11 weeks (Humpage and Falconer, 2003). The cylindrospermopsin was purified
29 (purity not reported) from an extract of freeze-dried C. raciborskii cells (strain AWT 205).
30 Endpoints monitored throughout the study included food and water consumption and body
31 weight. A clinical examination that focused on physiological and behavioral signs of toxicity
32 was conducted after 9 weeks of exposure. Hematology (all animals; red cell counts, hemoglobin,
33 packed cell volume, and white cell total and differential counts), serum chemistry (five
34 mice/group except all six mice at the high dose; total protein, albumin, globulin, glucose,
35 creatinine, urea, total bilirubin, total bile acids, cholesterol, triglycerides, sodium, potassium,
36 calcium, bicarbonate, creatinine kinase, alanine and aspartate aminotransferases [ALT and AST,
37 respectively], and alkaline phosphatase) and urine (five mice/group excluding high dose; specific
38 gravity, protein, glucose, ketones, creatinine, sodium, potassium, chloride, calcium, bicarbonate,
39 phosphate, pH, volume and presence of blood) evaluations were performed near or at the end of
40 the treatment period. Postmortem examinations included organ weights (liver, spleen, kidneys,
41 adrenal glands, heart, testis, epididymis and brain) and comprehensive histological evaluations.
42 The histological examinations were conducted in accordance with Organization for Economic
43 Cooperation and Development recommendations and performed on the following tissues: liver,
44 kidney, heart, lungs, thymus, thyroid, trachea, salivary glands, adrenal glands, epididymis, testis,
45 prostate, gall bladder, esophagus, stomach, duodenum/small intestine, large intestine, pancreas,
11 DRAFT: DO NOT CITE OR QUOTE
-------
1 spleen, urinary bladder, eyes, lymph nodes, aorta, cerebrum, cerebellum, spinal cord (cervical,
2 thoracic and lumbar) and peripheral nerve.
3
4 No deaths were reported. The mean final body weight was 7-15% higher than controls in
5 all dose groups, but the increases were not dose-related and were statistically significant only at
6 30 and 60 |j,g/kg-day (Humpage and Falconer, 2003). There were no significant changes in food
7 consumption; however, water intake was significantly reduced in all dose groups (data not
8 reported). Relative kidney weight was increased in a significant, dose-related manner beginning
9 at 60 |j,g/kg-day (12-23% greater than controls), while relative liver weight was significantly
10 increased only at the high dose of 240 |j,g/kg-day (13% greater than controls). Information on
11 absolute kidney and liver weights was not reported. Absolute testis weights were significantly
12 increased at >60 |j,g/kg-day (data not reported), but these changes were not significant when
13 normalized to body weight. The hematology, serum chemistry and urine evaluations showed no
14 clear exposure-related changes in any endpoint (including serum indicators of liver injury),
15 except for significant decreases in urine protein concentrations (g/mmol creatinine) at >120
16 |j,g/kg-day and urine specific gravity at 240 |j,g/kg-day (data presented graphically). The
17 postmortem examinations showed "minor increases in histopathological damage to the liver" at
18 >120 |j,g/kg-day and proximal renal tubular damage at 240 |j,g/kg-day, but additional information
19 regarding the type, severity and incidences of the liver and kidney lesions was not reported.
20
21 Cylindrospermopsin is known to inhibit protein synthesis in the liver (see Section 4.5.1).
22 Serum albumin, a major product of liver protein synthesis, was not decreased in this study
23 (Humpage and Falconer, 2003), but the most sensitive effects, decreased urinary protein at >120
24 |j,g/kg-day and increased relative kidney weight at >60 |j,g/kg-day, are both potential indicators of
25 suppressed protein synthesis. As hypothesized by the authors, the decrease in urinary protein is
26 consistent with decreased availability of protein and the increase in kidney weight may reflect a
27 compensatory hyperplasia, such that the kidney, as a protein-synthesizing organ, is stimulated to
28 grow in an attempt to maintain homeostasis in response to a chemically-induced decrease in
29 protein synthesis. Information supporting the hypothesis that the decrease in urinary protein
30 excretion reflects a specific effect of Cylindrospermopsin on protein synthesis, as well as the
31 possibility that it reflects a functional change in the nephron, is discussed in Section 4.5.2.
32 Because the renal effects observed by Humpage and Falconer (2003) are consistent with a known
33 mode of action of Cylindrospermopsin, and plausibly represent part of the progression of effects
34 leading to toxicity (Section 4.5.2), they are considered to be adverse. This study, therefore,
35 identifies aNOAEL and LOAEL of 30 and 60 |j,g/kg-day, respectively.
36
37 4.2.1.3.2. Cell Extract Studies
38
39 Groups of male Swiss albino mice (10 per dose except 12 controls and 5 at high-dose)
40 were exposed to a cell-free extract of sonicated and frozen C. raciborskii cells (strain AWT 205)
41 in the drinking water at reported Cylindrospermopsin doses of 0, 216, 432 or 657 |j,g/kg-day for
42 10 weeks (doses based on actual water consumption) (Humpage and Falconer, 2003). Food and
43 water consumption and body weight were measured throughout the study. Urinalyses (12
44 unspecified parameters) were performed after 5 and 10 weeks. Serum chemistry (15 unspecified
45 parameters) evaluations and examinations of unspecified major organs (organ weight, gross
12 DRAFT: DO NOT CITE OR QUOTE
-------
1 pathology and histopathology) were performed at the end of the exposure period. Hematology
2 was not evaluated.
3
4 Final body weights were significantly reduced at 432 and 657 |j,g/kg-day (9 and 7% less
5 than controls, respectively), and relative liver and kidney weights were significantly increased in
6 a dose-related manner at 216-657 |j,g/kg-day (27-47 and 30-43% greater than controls,
7 respectively). Other statistically significant effects included increased serum total bilirubin at
8 >216 |j,g/kg-day, decreased serum total bile acids at >216 |j,g/kg-day and decreased urine protein
9 concentration (g/mmol creatinine) at >432 |j,g/kg-day. There were no clear exposure-related
10 changes in any other serum or urine endpoints and no additional indicators of liver or kidney
11 injury. Results of the postmortem pathology examinations were not reported. The low dose of
12 216 |j,g/kg-day is a LOAEL for this study, based on increased relative liver and kidney weights,
13 increased serum bilirubin and decreased serum bile acids. An increase in serum bilirubin is
14 indicative of liver dysfunction or bile duct blockage as it reflects the ability of the liver to take
15 up, process and secrete bilirubin into the bile. Serum bile acids can be decreased due to an
16 inhibition of bile acid synthesis or an interference with bile acid resorption in the gastrointestinal
17 tract; bile acids are synthesized from cholesterol in the liver, conjugated, excreted in the bile and
18 resorbed in the ileum.
19
20 Quackenbush mice were administered drinking water containing a cell-free extract of
21 freeze-dried and sonicated C. raciborskii cells (strain AWT 205) for 90 days (Shaw et al., 2000,
22 2001). Gross pathological and histological (liver, kidney, spleen, heart, lungs and thymus)
23 examinations showed no effects at dose levels as high as 0.15 mg cylindrospermopsin/kg-day
24 (the highest tested dose), indicating that a NOAEL of 0.15 mg/kg-day was identified. Additional
25 information on the experimental design and results was not reported. The 0.15 mg/kg-day
26 NOAEL in Quackenbush mice is only slightly below the 216 |j,g/kg-day (0.22 mg/kg-day)
27 LOAEL for liver and kidney effects in the 10-week study with Swiss mice summarized above
28 (Humpage and Falconer, 2003); however, the LOAEL is based on different measured endpoints
29 (liver and kidney weights, serum bilirubin and serum bile acids) than the NOAEL
30 (histopathology).
31
32 4.2.1.4. Chronic Studies
33
34 No information regarding the chronic oral toxicity of cylindrospermopsin was identified
35 in the materials reviewed for this document.
36
37 4.2.2. Inhalation Exposure
38
39 No information regarding the inhalation toxicity of cylindrospermopsin was identified in
40 the materials reviewed for this document.
41
42 4.3. REPRODUCTIVE/DEVELOPMENTAL STUDIES - ORAL AND INHALATION
43
44 No information regarding the reproductive or developmental toxicity of
45 cylindrospermopsin was identified in the materials reviewed for this document.
46
13 DRAFT: DO NOT CITE OR QUOTE
-------
1 4.4. OTHER STUDIES
2
3 4.4.1. Effects By Parenteral Exposure.
4
5 4.4.1.1. Studies of Purified Cylindrospermopsin
6
7 Acute lethality values have been determined for Cylindrospermopsin purified from
8 extracts of cultured C. raciborskii or U. natans cells (Ohtani et al., 1992; Shaw et al., 2000,
9 2001; Terao et al., 1994). In male CHS mice, 24-hour and 5- to 6-day LD50 values of 2.1 and 0.2
10 mg/kg, respectively, have been reported for a single i.p. dose of purified Cylindrospermopsin
11 (purity not reported) (Ohtani et al., 1992). Another study found that a single 0.2 mg/kg i.p. dose
12 of purified Cylindrospermopsin (purity not reported) caused 50% moribundity after 31 hours in
13 Quackenbush mice (Shaw et al., 2000, 2001). The main pathological findings in the moribund
14 animals were lipid infiltration and cell necrosis in the liver. Terao et al. (1994) also found that
15 the liver was the main target of toxicity in male ICR mice administered a single 0.2 mg/kg i.p.
16 dose of purified Cylindrospermopsin (purity not reported), although treatment-related lesions
17 were additionally noted in the thymus, kidney and heart. A time series of ultrastructural tissue
18 examinations indicated four sequential phases of liver changes: inhibition of protein synthesis,
19 membrane proliferation, fat droplet accumulation and cell death.
20
21 4.4.1.2. Cell Extract Studies
22
23 The results of acute i.p. studies of extracts of freeze-dried and sonicated C. raciborskii
24 cells are generally similar to those of the i.p. studies of purified Cylindrospermopsin. A single
25 0.2 mg/kg cylindrospermopsin-equivalent dose caused 50% moribundity in Quackenbush mice
26 after 98 hours (Shaw et al., 2000, 2001). Other single-dose LD50 values, expressed as
27 cylindrospermopsin-equivalent doses, included 24-hour and 7-day values of 0.29 and 0.18
28 mg/kg, respectively, in male Swiss mice (Hawkins et al., 1997). This 24-hour LDso was lower
29 than the 24-hour i.p. LD50 of 2.1 mg/kg for purified Cylindrospermopsin in mice (Ohtani et al.,
30 1992), leading the authors to suggest that the extract contained more than one toxin. The liver
31 was the main target organ in the extract studies, although lesions also occurred in other tissues,
32 including kidney, adrenal gland, lung and intestine (Hawkins et al., 1985, 1997; Shaw et al.,
33 2000,2001).
34
35 A single dose i.p. LDso value of 64 mg freeze-dried culture/kg was determined in mice
36 observed for 24 hours (Hawkins et al., 1985). Falconer et al. (1999) assessed the acute lethality
37 and liver and kidney effects of four different batches of cell-free extracts of sonicated freeze-
38 dried C. raciborskii cells in male Swiss albino mice treated by single i.p. injection. Reported
39 24-hour and 7-day LDso values for the four batches were 50-110 and 20-65 mg extract/kg,
40 respectively. The Cylindrospermopsin content in the four batches varied from 1.3 to 5.4 mg/g
41 extract, indicating that the cylindrospermopsin-equivalent LD50 values were 0.07-0.6 mg/kg
42 (24-hour) and 0.03-0.4 mg/kg (7-day). Liver damage was characterized by cellular vacuolation,
43 intercellular spaces and darker nuclear and cytoplasmic staining. Kidney damage included
44 proximal tubule epithelial necrosis and presence of proteinaceous material in the distal tubules.
45 There was no clear correlation between Cylindrospermopsin batch concentration and the LD50
14 DRAFT: DO NOT CITE OR QUOTE
-------
1 values or severity of liver or kidney lesions, leading the study authors to suggest that more than
2 one toxin was present in the extract.
3
4 4.4.2. Immunotoxicity
5
6 No information was located regarding effects of cylindrospermopsin on immune
7 function, although immune system tissues appear to be a target of short-term, high-level
8 exposures. Massive necrosis of lymphocytes occurred in the cortical layer of the thymus of male
9 ICR mice given a single 0.2 mg/kg i.p. dose of cylindrospermopsin purified (purity not reported)
10 from cultured U. natans cells (Terao et al., 1994). Effects observed in MF1 mice administered a
11 single gavage dose of a suspension of freeze-dried C. raciborskii cells, in the lethal dose range of
12 4.4-8.3 mg cylindrospermopsin/kg, included atrophy in lymphoid tissue of the spleen (follicular
13 lymphocyte loss due to lymphophagocytosis) and thymus (degeneration and necrosis of cortical
14 lymphocytes) (Seawright et al., 1999). These effects were considered by the study authors to be
15 normal responses of the immune system to the stress of severe intoxication. Lympho-
16 phagocytosis was observed in the spleen of Quackenbush mice exposed to a cell-free extract of
17 freeze-dried and sonicated C. raciborskii cells by gavage at a nonlethal dose level of 0.05 mg
18 cylindrospermopsin/kg-day for 14 days (Shaw et al., 2000, 2001).
19
20 4.4.3. Tumor Initiation
21
22 The tumor initiating activity of cylindrospermopsin was tested in male Swiss mice using
23 O-tetradecanoylphorbol 13-acetate (TPA) as the promoter (Falconer and Humpage, 2001). Mice
24 were administered a gavage dose of saline (27 mice) or 500 mg/kg of a saline extract of fireeze-
25 dried C. raciborskii cells (strain AWT 205) (34 mice) every other week for three doses. Other
26 groups received a single dose of 1500 mg extract/kg (14 mice) or two doses of 1500 mg
27 extract/kg separated by 2 weeks (17 mice). Most (70%) of the 2 x 1500 mg extract/kg group
28 died within 1 week of the second dose, leaving five survivors for use in the rest of the study.
29 Based on a reported cylindrospermopsin content of 5.5 mg/g extract, the cylindrospermopsin-
30 equivalent doses in the 500 and 1500 mg extract/kg groups were 2.75 and 8.25 mg/kg,
31 respectively. Two weeks after the final dose, the saline and 500 mg extract/kg groups were
32 divided into subgroups of 13-18 mice that were fed liquid food containing TPA dissolved in
33 DMSO, or food containing DMSO alone, for 24 hours twice weekly for 30 weeks. All of the
34 mice in both 1500 mg extract/kg groups were similarly exposed to TPA-containing liquid food
35 (no 1500 mg/kg mice were exposed to food containing DMSO alone). Histological examinations
36 of the liver, kidneys, spleen and grossly abnormal organs were performed on all groups at the
37 end of the 30-week promotion period. Neoplastic changes were found in none of the 27 control
38 mice and in a total of 5 cylindrospermopsin-treated mice, a difference that was not statistically
39 significant. There was no pattern to the neoplasic changes, as they occurred in different animals,
40 target organs and treatment groups, as detailed in Table 4-1.
15 DRAFT: DO NOT CITE OR QUOTE
-------
Table 4-1. Tumor Initiating Activity of C. raciborskii Extracts
Oral Treatment (mg extract/kg)
Saline/DMSO
Saline/TPA
3 x 500/DMSO
3 x 500/TPA
1 x 1 500/TPA
2xl500/TPA
Number of Mice
14
13
18
16
14
5
Histological Finding*
No neoplasia observed
No neoplasia observed
1 hepatocellular carcinoma
1 lymphoma
No neoplasia observed
2 hepatocellular dysplastic foci
1 fibroblastic osteosarcoma
No neoplasia observed
2
o
J
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
* All findings were in different animals
Source: Falconer and Humpage (2001)
4.4.4. Genotoxicity
Purified cylindrospermopsin caused an increase in the frequency of micronuclei in the
human lymphoblastoid cell line WIL2-NS (Humpage et al., 2000). Both centromere-positive
and centromere-negative micronuclei were induced, suggesting that whole chromosome loss, as
well as DNA strand breaks, contributed to the in vitro cytogenetic damage. DNA strand
breakage was also observed in the liver of Balb/c mice following a single 0.2 mg/kg i.p. dose of
purified cylindrospermopsin (Shen et al., 2002). Covalent binding of cylindrospermopsin or a
metabolite to DNA (adduct not identified) was detected in the liver of Quackenbush mice given a
single i.p. injection of a cell-free extract of C. raciborskii (dose levels not reported) (Shaw et al.,
2000). Purified cylindrospermopsin caused cell growth inhibition and altered cell morphology,
but no apoptosis or DNA strand breaks, in Chinese hamster ovary Kl cells in vitro (Fessard and
Bernard, 2003).
4.5. MECHANISTIC DATA AND OTHER STUDIES IN SUPPORT OF THE MODE
OF ACTION
4.5.1. Liver Toxicity
The liver is widely regarded as the main target of cylindrospermopsin toxicity, and
consequently, most mechanistic studies have assessed hepatic endpoints. The specific
mechanism for the liver toxicity is not clearly understood, although it has generally been
considered to involve cylindrospermopsin-induced inhibition of protein synthesis.
16
DRAFT: DO NOT CITE OR QUOTE
-------
1 Cylindrospermopsin was shown to be a potent inhibitor of protein synthesis in an in vitro rabbit
2 reticulocyte globin synthesis assay (Terao et al., 1994). Ultrastructural liver changes in mice
3 treated with a single 0.2 mg/kg i.p. dose of purified Cylindrospermopsin had features in common
4 with those dosed with the protein synthesis inhibitor cycloheximide, particularly detachment of
5 ribosomes from the rough endoplasmic reticulum, suggesting that protein synthesis inhibition
6 plays a role in Cylindrospermopsin hepatotoxicity in vivo (Terao et al., 1994). However, unlike
7 the liver in the cycloheximide-dosed mice, the liver of those treated with Cylindrospermopsin
8 showed membrane proliferation, fat droplet accumulation and reduced amount of total P450 in
9 microsomes, indicating that mechanisms other than protein synthesis inhibition must also
10 contribute to Cylindrospermopsin toxicity.
11
12 Cylindrospermopsin-induced depletion of mouse hepatic glutathione was demonstrated in
13 vivo (Norris et al., 2002), although the study authors did not consider the effect to be of sufficient
14 magnitude to represent the primary mechanism of Cylindrospermopsin toxicity.
15 Cylindrospermopsin also caused decreased glutathione levels, as well as decreased synthesis of
16 glutathione and protein, in cultured rat hepatocytes (Runnegar et al., 1994, 1995, 2002).
17 Inhibition of glutathione synthesis was the predominant mechanism for the reduction in
18 glutathione; other mechanisms, including increased consumption of glutathione, increased
19 formation of oxidized glutathione, increased glutathione efflux, hidden forms of glutathione,
20 decreased glutathione precursor availability and decreased cellular ATP were effectively ruled
21 out (Runnegar et al., 1995). Glutathione depletion occurred at non-toxic Cylindrospermopsin
22 concentrations and preceded the onset of observable toxicity at higher concentrations (Runnegar
23 et al., 1994). Pretreatment with the CYP450 inhibitor, a-naphthoflavone, partially protected
24 against cytotoxicity and cellular glutathione depletion, indicating involvement of the CYP450
25 enzyme system in Cylindrospermopsin metabolism and that one or more metabolites might be
26 more active than the parent compound in inhibiting glutathione synthesis (Runnegar et al., 1995).
27 In vitro studies in mouse hepatocytes provided no indication that reductions in glutathione levels
28 by Cylindrospermopsin led to increased levels of reactive oxygen species (ROS) (Humpage et al.,
29 2005).
30
31 Cylindrospermopsin induced time- and concentration-dependent toxicity and inhibition of
32 protein synthesis in hepatocytes isolated from male Swiss mice (Froscio et al., 2003). The
33 broad-spectrum CYP450 inhibitors proadifen (SKF525A) and ketoconazole diminished the
34 induction of cytotoxicity by Cylindrospermopsin, but did not diminish the inhibition of protein
35 synthesis. These findings suggest that the cytotoxic effects of Cylindrospermopsin might be
36 linked more to CYP450-mediated bioactivation than to inhibition of protein synthesis by the
37 parent compound. Similarly, pretreatment of male Quackenbush mice with the broad-spectrum
38 CYP450 inhibitor piperonyl butoxide protected against the acute lethality of Cylindrospermopsin
39 (Norris et al., 2002). In a study using inhibitors of specific CYP450 isoforms, furafylline
40 (CYP1A2) and omeprazole (CYP3A4 and CYP2C19) protected against Cylindrospermopsin
41 cytotoxicity in an in vitro mouse hepatocyte system; unspecified inhibitors of CYPs 2A6, 2D6
42 and 2E1 were not found to be cytoprotective (Humpage et al., 2005). Additional support for the
43 involvement of CYP450 in the hepatotoxicity of Cylindrospermopsin is the finding that liver
44 histopathology is mainly induced in the region (periacinar) where CYP450-catalyzed xenobiotic
45 metabolism occurs (Shaw et al., 2000, 2001).
46
17 DRAFT: DO NOT CITE OR QUOTE
-------
1 4.5.2. Kidney Toxicity
2
3 No studies were located that specifically investigated the involvement of protein
4 synthesis inhibition or other modes of action in cylindrospermopsin-induced toxicity in the
5 kidney or other non-hepatic target tissues (e.g., spleen and thymus). As detailed in Section
6 4.2.1.3.1, the kidney was the most sensitive target in mice that were exposed to
7 cylindrospermopsin by daily gavage for 11 weeks (Humpage and Falconer, 2003). Renal effects
8 in the mice included increased relative kidney weight at >60 |j,g/kg-day, decreased urinary
9 protein at >120 |j,g/kg-day and decreased urine specific gravity and proximal renal tubular
10 lesions at 240 |j,g/kg-day. The authors hypothesized that the decrease in urinary protein is
11 consistent with decreased availability of protein and that the increase in kidney weight may
12 reflect a compensatory hyperplasia, such that the kidney, as a protein-synthesizing organ, is
13 stimulated to grow in an attempt to maintain homeostasis in response to a cylindrospermopsin-
14 related decrease in protein synthesis. Information supporting the hypothesis that the decrease in
15 urinary protein excretion reflects a specific effect of cylindrospermopsin on protein synthesis, as
16 well as the possibility that it reflects a functional change in the nephron, is discussed below.
17 Also discussed is evidence suggesting a dose-severity progression of kidney effects.
18
19 Potential mechanisms for a decrease in urinary protein include a decrease in glomerular
20 filtration (i.e., filtered load) of protein, an increase in resorption of filtered protein and a decrease
21 in secretion of nephrogenic protein. A decrease in glomerular filtration of protein (e.g., |j,g
22 protein/day) could result from a decrease in serum protein concentration or a decrease in
23 glomerular filtration rate (mL/day, GFR). The predominant serum protein in urine of healthy
24 animals (e.g., mice, rats and humans) is albumin (-50% of serum proteins in urine). In the
25 Humpage and Falconer (2003) study, serum albumin concentration increased in mice exposed to
26 cylindrospermopsin, and serum creatinine (a marker of GFR) apparently was unchanged; it was
27 measured but not discussed in the results. Therefore, it is unlikely that glomerular filtration of
28 serum proteins decreased in response to cylindrospermopsin (if a change occurred, it is likely to
29 have been an increase in the rate of filtration of albumin). Furthermore, serum proteins normally
30 account for approximately 15% of total urinary protein (Pesce and First, 1979). The decrease in
31 urinary excretion of protein observed in Humpage and Falconer (2003) was substantially larger
32 than this (-50%), indicating that the decrease in urinary protein cannot derive solely from a
33 decrease in excretion (i.e., glomerular filtration) of serum proteins.
34
35 No information is presented in Humage and Falconer (2003) that would allow an
36 assessment of tubular resorption of filtered protein (e.g., plasma-to-urine clearance of protein,
37 excretion of low-molecular weight proteins such as p2|J,globulin or retinal binding protein).
38
39 In healthy mammals, the dominant protein in urine (-50%) is the nephrogenic Tamm-
40 Horsfall protein (THP, uromucoid) (Bachmann et al., 1991, 2005). In the absence of a decrease
41 in filtration or increased resorption of filtered serum protein, the substantial decrease in urinary
42 protein (i.e., -50%) observed by Humpage and Falconer (2003) would almost certainly have to
43 involve decreased excretion of THP, since it is the predominant protein in urine. Although there
44 are numerous possible mechanisms for an acute change in THP excretion (Bachman et al., 1991),
45 long-term maintenance of lower (i.e., steady-state) rate of urinary excretion of THP requires a
46 decreased rate of synthesis of THP (Bachman et al., 1991, 2005; Schoel and Pfleiderer, 1987).
18 DRAFT: DO NOT CITE OR QUOTE
-------
1 THP is synthesized exclusively in the thick ascending limb of the loop of Henle (TAL);
2 therefore, a sustained change in THP excretion is likely to reflect a functional change in this
3 region of the nephron. Increases and decreases in THP have been observed in various kidney
4 diseases, and in association with experimental treatments that induce hypertrophy of the TAL,
5 including increased dietary protein (Bachmann et al., 1991). Depletion of THP from the kidney
6 may, in itself, be adverse. Mice deficient in THP (i.e., THP knockout mice) display impaired
7 urine concentrating ability, up-regulation of distal nephron electrolyte transport proteins and
8 increased susceptibility to urinary tract infections (Bachmann et al., 2005; Bates et al., 2004).
9 The decrease in urine specific gravity in animals exposed to cylindrospermopsin in the Humpage
10 and Falconer (2003) study may be indicative of impaired urine concentrating ability and,
11 possibly, related to impaired function of the TAL (i.e., impairment of transport activity in this
12 region of the nephron impairs urine concentrating ability) and/or decreased synthesis of THP.
13
14 Additional kidney effects in the Humpage and Falconer (2003) mouse study included
15 proximal renal tubular damage (type and severity of lesions not reported) at the high dose.
16 Clinical effects in the Palm Island outbreak in which humans apparently ingested drinking water
17 containing elevated levels of cylindrospermopsin included renal damage, as indicated by loss of
18 water, electrolytes, proteins, ketones and carbohydrates (Blyth, 1980; Griffiths and Saker, 2003)
19 (Section 4.1). Proteinuria would be expected with proximal tubular damage, as this is the site of
20 resorption of filtered protein. Proteinuria was not observed by Humpage and Falconer (2003),
21 but information on the type and severity of the tubular damage was not reported. Proteinuria did
22 occur in the humans, although other mechanisms could have caused it (e.g., glomerular injury
23 will produce high molecular weight proteinuria). The evidence for proximal tubular damage and
24 functional impairment (e.g., proteinuria, glucosuria) together strengthen the argument that the
25 kidney is a target of cylindrospermopsin and, when considered with decreased protein excretion
26 at lower doses, suggests a dose-severity progression.
27
28 4.5.3. Interactions with DNA and RNA
29
30 Based on structural characteristics of cylindrospermopsin (its nucleoside structure and
31 potentially reactive guanidine and sulfate groups), it has been speculated that cylindrospermopsin
32 may exert its toxic effects via pathways that include reactions with DNA and/or RNA (see
33 Humpage et al., 2000; Shen et al., 2002). Covalent binding between DNA and
34 cylindrospermopsin, or a metabolite, occurred in mouse liver in vivo (Shaw et al., 2000). DNA
35 adducts were detected, but not identified, using the 32P-postlabeling assay; this involved
36 extraction of the DNA, hydrolysis into individual nucleotides, labeling of the nucleotides using
37 32P-ATP, separation of adducted nucleotides using two-dimensional thin layer chromatography
38 and visualization of adduct spots by autoradiography. Cylindrospermopsin also induced DNA
39 strand breakage in mouse liver in vivo (Shen et al., 2002) and increases in micronuclei occurred
40 in treated binucleated cells of the WIL2-NS lymphoblastoid cell-line (Humpage et al., 2000).
41 Two mechanisms were suggested for causing the cytogenetic damage: one at the level of DNA to
42 induce strand breaks and the other at the level of kinetochore/spindle function to induce loss of
43 whole chromosomes (Humpage et al., 2000; Shen et al., 2002). The broad-spectrum CYP450
44 inhibitors omeprazole and SKF525A inhibited cylindrospermopsin-induced DNA damage in
45 primary cultured mouse hepatocytes at subcytotoxic concentrations, suggesting that CYP-derived
19 DRAFT: DO NOT CITE OR QUOTE
-------
1 metabolites are responsible for cylindrospermopsin genotoxicity and that genotoxicity is a
2 primary effect of the chemical (Humpage et al., 2005).
3
4 Cylindrospermopsin-induced up-regulation of the tissue transglutaminase (tTGase) gene
5 was demonstrated in liver RNA of Balb/c mice following i.p. injection of a single 100 ng/kg
6 dose of cylindrospermopsin (Shen et al., 2003). tTGase is a unique member of the TGase (EC
7 2.3.2.13) family that catalyzes the post-translational modification of proteins via Ca2+-dependent
8 cross-linking reactions (Shen et al., 2003). The up-regulation of tTGase can lead to liver injury
9 (Grenard et al., 2001; Mirza et al., 1997), and has been implicated in diverse biological
10 processes, such as induction of apoptosis (Piacentini et al., 2002; Zhang et al., 1995), cell death
11 and differentiation (Shen et al., 2003; Fesus et al., 1987) and adhesion and morphological
12 changes of cells (Shen et al., 2003; Akimov and Belkin, 2001).
13
14 4.5.4. Structure-Activity Relationships
15
16 Natural cylindrospermopsin, synthetic (racemic) cylindrospermopsin and selected
17 synthetically-produced cylindrospermopsin structural analogues were assessed for effects on
18 protein synthesis in both the rabbit reticulocyte lysate system and cultured rat hepatocytes
19 (Runnegar et al., 2002). No significant differences were observed in levels of protein synthesis
20 inhibition elicited by natural cylindrospermopsin and its diol analogue, indicating that the sulfate
21 group might not be a necessary component of cylindrospermopsin-induced protein synthesis
22 inhibition. Additionally, the orientation of the hydroxyl group at C7 in the carbon bridge does
23 not appear to be important, since the C7 epimer of cylindrospermopsin and its corresponding diol
24 exhibited protein synthesis inhibition similar to that elicited by synthetic (racemic)
25 cylindrospermopsin. The cyclopentyl ring and the methyl and hydroxyl groups on the adjacent
26 hexyl ring may be important structural features, because the analogue lacking these features was
27 500-1000-fold less effective in the inhibition of protein synthesis.
28
29 The uracil portion of cylindrospermopsin appears to play an important role in
30 cylindrospermopsin toxicity. Banker et al. (2001) found that the acute lethality of
31 cylindrospermopsin to mice was eliminated by chlorination or partial cleavage of the uracil
32 moiety (resulting in 5-chloro-cylindrospermopsin and cylindrospermic acid, respectively), as
33 shown by a 5-day i.p. LD50 value of 0.2 mg/kg for cylindrospermopsin and 10-day i.p. LD50
34 values of >10 mg/kg for 5-chloro-cylindrospermopsin and >10 mg/kg for cylindrospermic acid.
35
36 Deoxycylindrospermopsin, an analogue of cylindrospermopsin isolated and purified from
37 C. raciborskii, was tested for toxicity in male white Quackenbush mice treated by i.p. injection
38 (Norris et al., 1999). Deoxycylindrospermopsin did not appear to be toxic during 5 days
39 following administration of a 0.8 mg/kg dose, whereas Ohtani et al. (1992) reported a 5- to 6-day
40 i.p. LDso value of 0.2 mg/kg for cylindrospermopsin in male CD3 mice. Although this
41 comparison suggests that deoxycylindrospermopsin is significantly less toxic than
42 cylindrospermopsin, differences in study designs (e.g., the use of different strains of mice) could
43 have contributed to the difference in toxicity.
44
20 DRAFT: DO NOT CITE OR QUOTE
-------
1 4.6. SYNTHESIS AND EVALUATION OF MAJOR NONCANCER EFFECTS
2
3 4.6.1. Oral
4
5 Information on the health effects of cylindrospermopsin in humans is limited to
6 observations on the Australian Palm Island poisoning incident that involved acute and/or short-
7 term drinking water exposure to C. raciborskii, a non-infectious cyanobacterium (Blyth, 1980;
8 Griffiths and Saker, 2003). The clinical picture of the illness is well-defined and includes fever,
9 headache, vomiting, bloody diarrhea, hepatomegaly and kidney damage with loss of water,
10 electrolytes and protein, but no data are available on exposure levels of cylindrospermopsin that
11 induced these effects.
12
13 The preponderance of information on noncancer effects of cylindrospermopsin in animals
14 is available from oral and i.p. administration studies in mice that were exposed to purified
15 compound or extracts of C. raciborskii cells. These studies indicate that the liver and kidneys
16 are main targets of toxicity and that cylindrospermopsin also causes significant lesions in other
17 organs, particularly the spleen and thymus. Considering both animal and human kidney data, the
18 evidence suggests a dose-severity progression of renal effects ranging from decreased protein
19 synthesis at low doses to functional impairment at high doses. The cell extract studies provide
20 limited dose-response information for cylindrospermopsin because concentrations vary between
21 cultures and strains and, in some cases, may contain other toxins, as discussed in the introduction
22 to Section 4.2. The available oral toxicity studies of purified cylindrospermopsin are
23 summarized in Table 4-2.
24
25 No studies have been performed assessing the acute oral toxicity of purified
26 cylindrospermopsin. Studies in which mice were administered single gavage doses of
27 suspensions or cell-free extracts of C. raciborskii cells at near-lethal to lethal levels found severe
28 damage to the liver (fatty and necrotic changes), kidneys (acute tubular necrosis), spleen and
29 thymus (atrophy of lymphoid tissue), heart (hemorrhages) and gastric mucosa (ulceration of the
30 esophageal section) (Falconer et al., 1999; Seawright et al., 1999; Shaw et al., 2000, 2001).
31
32 A limited amount of information on the short-term oral toxicity of cylindrospermopsin is
33 available from inadequately reported 14- and 21-day studies.
34
35 Histological examinations of small numbers of mice that were administered daily gavage
36 doses of purified cylindrospermopsin for 14 days showed effects in the liver (fatty infiltration)
37 and spleen (lymphophagocytosis) (Shaw et al., 2000, 2001). Fatty infiltration in the liver was the
38 more sensitive effect based on a reported NOAEL of 0.05 mg/kg-day and LOAEL of 0.15
39 mg/kg-day. Small numbers of mice and rats were exposed to cylindrospermopsin for 21 days in
40 drinking water from a dammed impoundment at a reported approximate dose of 0.2 mg/kg-day
41 (Shaw et al., 2000, 2001). No histopathological changes were noted, indicating a NOAEL of 0.2
42 mg/kg-day in drinking water. The adequacy of the 14- and 21-day effect levels cannot be
43 assessed due to a lack of any additional reported information on the design and results of these
44 studies.
45
21 DRAFT: DO NOT CITE OR QUOTE
-------
Table 4-2. Summary Results of Oral Toxicity Studies of Pure Cylindrospermopsin in Experimental Animals*
Species
Sex
Average
Daily Dose
(mg/kg-day)
Exposure
Duration
NOAEL
(mg/kg-day)
LOAEL
(mg/kg-day)
Responses
Comments
Reference
Acute Exposure
No suitable acute studies are available
Short-Term Exposure
Mouse
NR
0.05,0.15,
0.3
(gavage)
14 days
0.05
0.15
Lipid infiltration in liver.
Low confidence in NOAEL and
LOAEL. A full report of this study
has not been published; this table
provides essentially all available
information on experimental design
and results.
Shaw et al.,
2000, 2001
Subchronic Exposure
Mouse
M
0, 0.03,
0.06,0.12,
0.24
(drinking
water)
1 1 weeks
0.03
0.06
Increased relative kidney
weight with decreased
urinary protein at >0. 12
mg/kg-day.
Well-designed study with endpoints
that included food and water
consumption, body weight, clinical
signs, hematology, serum chemistry,
urinalysis, organ weights (eight
organs) and histology
(comprehensive). Ten mice/level
(six in high dose group).
Humpage and
Falconer, 2003
Chronic Exposure
No suitable chronic studies are available.
* Oral studies using suspensions or cell-free extracts of C. raciborskii cells are discussed in Section 4.2.1.
NR = Not reported
22
DRAFT: DO NOT CITE OR QUOTE
-------
1 A comprehensive subchronic toxicity study was conducted in which mice were exposed
2 to five dose levels of purified cylindrospermopsin (0, 30, 60, 120 or 240 ng/kg) by daily gavage
3 for 11 weeks (Humpage and Falconer, 2003). Histopathological effects were observed in the
4 liver at >120 |j,g/kg-day ("minor increases in histopathological damage") and kidneys at 240
5 |j,g/kg-day (proximal tubular damage), but no other information on the lesions, including
6 incidence data, was reported. There were no changes in liver weight at doses below 240
7 |j,g/kg-day or serum indices of liver damage (e.g., serum ALT, AST and alkaline phosphatase) in
8 any of the dose groups. Relative kidney weight was increased at >60 |j,g/kg-day and urine
9 protein was decreased at >120 |j,g/kg-day. These effects are considered to be adverse because
10 they are consistent with a known mode of action of cylindrospermopsin (inhibition of protein
11 synthesis) and represent part of the spectrum of effects leading to toxicity, as discussed in
12 Section 4.5.2. Based on the increase in kidney weight, the subchronic NOAEL and LOAEL
13 values are 30 and 60 |j,g/kg-day, respectively.
14
15 No information was located regarding the chronic toxicity, neurotoxicity or
16 developmental/reproductive toxicity of cylindrospermopsin.
17
18 4.6.2. Inhalation
19
20 No information was located regarding the inhalation toxicity of cylindrospermopsin.
21
22 4.6.3. Mode of Action Information
23
24 The liver and kidneys appear to be the main targets of cylindrospermopsin toxicity. The
25 mechanism for liver toxicity is incompletely characterized, but involves inhibition of protein
26 synthesis (Froscio et al., 2003; Terao et al., 1994). Available evidence indicates that the protein
27 synthesis inhibition is not decreased by broad-spectrum CYP450 inhibitors, suggesting that it is
28 mediated by the parent compound (Froscio et al., 2003). Hepatocytotoxicity occurs at higher
29 levels of cylindrospermopsin and appears to be CYP450-dependent, indicating the involvement
30 of metabolites and other mechanisms (Froscio et al., 2003; Humpage et al., 2005; Norris et al.,
31 2002). Studies specifically investigating the inhibition of protein synthesis in the kidneys are not
32 available, although the results of the 11-week oral toxicity study in mice (Humpage and
33 Falconer, 2003) are consistent with an inhibition of protein synthesis. Effects in this study
34 included decreased urinary protein and, at a higher dose, proximal renal tubular lesions. As
35 discussed in Section 4.5.2, the decrease in urinary protein excretion at low doses could reflect a
36 specific effect of cylindrospermopsin on protein synthesis or, possibly, a functional change in the
37 nephron. The proximal renal tubular damage in mice (Humpage and Falconer, 2003), as well as
38 the clinical findings of renal insufficiency in the Palm Island human poisoning incident (Blyth,
39 1980; Griffiths and Saker, 2003), suggest that cytotoxic mechanisms may predominate in the
40 kidney at higher doses.
41
42 Genotoxic effects of cylindrospermopsin include DNA adduction and strand breakage in
43 mouse liver (Shaw et al., 2000; Shen et al., 2002) and micronuclei formation in a lymphoblastoid
44 cell line (Humpage et al., 2000). Broad spectrum CYP450 inhibitors inhibited
45 cylindrospermopsin-induced DNA damage in mouse hepatocytes at sub-cytotoxic concentrations
23 DRAFT: DO NOT CITE OR QUOTE
-------
1 (Humpage et al., 2005), suggesting that metabolites are responsible for cylindrospermopsin
2 genotoxicity and that genotoxicity is a primary effect of the chemical.
3
4 4.7. WEIGHT-OF-EVIDENCE EVALUATION AND CANCER
5 CHARACTERIZATION
6
7 4.7.1. Summary of Overall Weight-of-Evidence
8
9 No information is available on the carcinogenicity of cylindrospermopsin in humans, and
10 no cancer studies of purified cylindrospermopsin have been conducted in animals. A test of an
11 extract of C. raciborskii cells suggests that cylindrospermopsin has no tumor initiating activity in
12 mice (Falconer and Humpage, 2001). A limited amount of data indicate that cylindrospermopsin
13 or a metabolite can covalently bind to DNA (Shaw et al., 2000) and cause cytogenetic damage,
14 as shown by induction of micronuclei (Humpage et al., 2000) and DNA strand breakage (Shen et
15 al., 2002). In accordance with the Guidelines for Carcinogen Risk Assessment (U.S. EPA,
16 2005a), the weight of evidence descriptor for the carcinogenic hazard potential of
17 cylindrospermopsin is "Inadequate Information to Assess Carcinogenic Potential."
18
19 4.7.2. Synthesis of Human, Animal and Other Supporting Evidence
20
21 No information was located regarding the carcinogenicity of purified cylindrospermopsin
22 in humans or animals. There was no indication that cylindrospermopsin had tumor initiating
23 activity in a test in which mice were administered a cell-free extract of freeze-dried C.
24 raciborskii cells by gavage followed by oral exposure to the tumor promoter TPA (Falconer and
25 Humpage, 2001).
26
27 The nucleotide structure of cylindrospermopsin, as well as the presence of potentially
28 reactive guanidine and sulfate groups, suggests the possibility of interference with DNA and/or
29 RNA synthesis and induction of mutations. Covalent binding between DNA and
30 cylindrospermopsin (or a metabolite) (Shaw et al., 2000, 2001) and DNA strand breakage (Shen
31 et al., 2002) have been demonstrated in mouse liver, and micronuclei were induced in human
32 WIL2-NS lymphoblasts (Humpage et al., 2000). Although the available data indicate that DNA
33 strand breakage could be a key mechanism for cylindrospermopsin-induced cytogenetic damage
34 (Humpage et al., 2000; Shen et al., 2002), insufficient data are available to speculate on the
35 carcinogenic potential of cylindrospermopsin.
36
37 4.8. SUSCEPTIBLE POPULATIONS AND LIFE STAGES
38
39 4.8.1. Possible Childhood Susceptibility
40
41 As discussed in Section 4.1, cylindrospermopsin has been implicated in the Palm Island
42 outbreak of a hepatoenteritis-like illness in 148 Australians (Blyth, 1980; Griffiths and Saker,
43 2003). Of the 148 cases, 138 were children (mean age 8.4 years, range 2-16 years, 41% male
44 and 59% female) and 10 were adults (sex and age not reported). There are no reported
45 indications that the 138 children were from the households of the 10 adults or that the children
24 DRAFT: DO NOT CITE OR QUOTE
-------
1 and adults received different exposures, suggesting a possible increased sensitivity of children
2 (the child:adult ratio is approximately 14:1).
3
4 4.8.2. Possible Gender Differences
5
6 There is no information on possible gender differences in the disposition of, or response
7 to, cylindrospermopsin.
8
9 4.8.3. Other Possible Susceptible Populations
10
11 No data were located regarding populations that might be unusually susceptible to
12 cylindrospermopsin. It is conceivable that individuals with liver and/or kidney disease might be
13 more susceptible than the general population because of compromised detoxification
14 mechanisms in the liver and impaired excretory mechanisms in the kidney.
25 DRAFT: DO NOT CITE OR QUOTE
-------
1 5. DOSE-RESPONSE ASSESSMENTS
2
3
4 5.1. NARRATIVE DESCRIPTION OF THE EXTENT OF THE DATABASE
5
6 Studies on the absorption, tissue distribution, metabolism and elimination of
7 cylindrospermopsin following oral, inhalation or dermal exposure have not been performed.
8 Gastrointestinal absorption of cylindrospermopsin is indicated by the induction of systemic
9 effects in oral toxicity studies. Studies in which cylindrospermopsin was administered to mice
10 by acute i.p. injection indicate that it is largely distributed to the liver and rapidly eliminated in
11 the urine as unmetabolized compound. The liver and kidneys are the main targets of
12 cylindrospermopsin toxicity. Possible modes of action include inhibition of protein synthesis,
13 CYP450-mediated bioactivation to a reactive intermediate and covalent binding between parent
14 compound or a metabolite and DNA and/or RNA.
15
16 The only information on the toxicity of cylindrospermopsin in humans is from reports of
17 a poisoning outbreak that is attributed to the consumption of drinking water containing toxin-
18 producing C. raciborskii. Although the clinical picture of this hepatoenteritis-like illness is well
19 defined and includes bloody diarrhea, swollen liver and impaired kidney function, there are no
20 data on exposure levels that induced these effects.
21
22 Most of the available data on the toxicity of cylindrospermopsin in animals are available
23 from oral and i.p. studies that tested purified compound or extracts of C. raciborskii cells. These
24 studies are generally consistent in indicating that cylindrospermopsin causes lesions in the liver
25 and other organs, particularly the kidneys, spleen and thymus. The cell extract studies are not
26 useful for dose-response assessment of cylindrospermopsin due to the confounding factors
27 discussed in the introduction to Section 4.2. The database on oral toxicity of pure
28 cylindrospermopsin is limited by a small number of studies and insufficient reporting. No
29 studies have been performed assessing the acute oral toxicity of pure cylindrospermopsin. Data
30 on the short-term oral toxicity of pure compound are available from inadequately reported
31 14-day gavage and 21-day drinking water studies in mice and rats. The reports of these studies
32 identify NOAELs and LOAELs for histopathology, but the adequacy of these effect levels
33 cannot be verified due to a virtual lack of any additional information on the experimental designs
34 and results. Data on the subchronic oral toxicity of pure cylindrospermopsin are available from a
35 well-designed and reported 11-week study in mice that provides a suitable basis for derivation of
36 a subchronic oral RfD value. No chronic toxicity, reproductive toxicity, developmental toxicity
37 or carcinogenicity studies of pure cylindrospermopsin have been performed.
38
39 No information is available on the inhalation toxicity of cylindrospermopsin.
40
41 5.2. ORAL REFERENCE DOSE (RfD)
42
43 5.2.1. Data Considered in Deriving Reference Values
44
45 Data considered in deriving oral RfDs for each duration of exposure are summarized in
46 Table 4-1 (Section 4.6.1).
26 DRAFT: DO NOT CITE OR QUOTE
-------
1 5.2.2. Acute Duration
2
3 5.2.2.1. Choice of Principal Study and Critical Effect - with Rationale and
4 Justification
5
6 Derivation of an acute oral RfD for cylindrospermopsin is precluded by insufficient data.
7 The only information on the toxicity of cylindrospermopsin in humans is the outbreak of a
8 hepatoenteritis-like illness that is attributed to the consumption of drinking water containing C.
9 raciborskii (Blyth, 1980; Griffiths and Saker, 2003; Hawkins et al., 1985; Ohtani et al., 1992).
10 Although the clinical picture of the illness is well defined, measured or estimated exposure levels
11 have not been reported. No acute oral toxicity studies of purified cylindrospermopsin have been
12 performed in animals. Single-dose studies of suspensions or cell-free extracts of C. raciborskii
13 cells were conducted in mice, but only near-lethal to lethal dose levels were tested (Falconer et
14 al., 1999; Seawright et al., 1999; Shaw et al., 2000, 2001).
15
16 5.2.3. Short-Term Duration
17
18 5.2.3.1. Choice of Principal Study and Critical Effect - with Rationale and
19 Justification
20
21 Derivation of a short-term oral RfD for cylindrospermopsin is precluded by insufficient
22 data. The only information relevant to the short-term toxicity of cylindrospermopsin in humans
23 is qualitative data on the outbreak of the hepatoenteritis-like illness that is attributed to the
24 consumption of drinking water containing C. raciborskii (Blyth, 1980; Griffiths and Saker, 2003;
25 Hawkins et al., 1985; Ohtani et al., 1992). As discussed in Sections 4.2.1.2 and 4.5.1, a limited
26 amount of information is available on the short-term oral toxicity of cylindrospermopsin from
27 poorly reported 14-day gavage and 21-day drinking water studies (Shaw et al., 2000, 2001). The
28 14-day study reported a NOAEL of 0.05 mg/kg-day and LOAEL of 0.15 mg/kg-day for liver
29 fatty infiltration in mice, and the 21-day study reported a free-standing NOAEL of 0.2
30 mg/kg-day for histopathology in mice and rats. The appropriateness of these effect levels cannot
31 be assessed due to inadequate information on the design and results of the studies.
32
33 5.2.4. Subchronic Duration
34
35 5.2.4.1. Choice of Principal Study and Critical Effect - with Rationale and
36 Justification
37
38 The comprehensive 11-week subchronic study in mice (Humpage and Falconer, 2003),
39 detailed in Section 4.2.1.3, is the only subchronic study of purified cylindrospermopsin and
40 provides a suitable basis for RfD derivation. The LOAEL was 60 |j,g/kg-day for increased
41 relative kidney weight. At 120 |j,g/kg-day, there was a significant decrease in urinary protein
42 concentration and minor histopathological changes in the liver. Decreased urinary protein and
43 increased relative kidney weight are both potential indicators of suppressed protein synthesis, a
44 known mode of action of cylindrospermopsin. The decrease in urinary protein is consistent with
45 decreased availability of protein and the increase in kidney weight may reflect a compensatory
46 hyperplasia, such that the kidney, as a protein-synthesizing organ, is stimulated to grow in an
27 DRAFT: DO NOT CITE OR QUOTE
-------
1 attempt to maintain homeostasis in face of a chemical-related decrease in protein synthesis
2 (Humpage and Falconer, 2003). Information supporting this hypothesis, as well as the
3 possibility that the decrease in urinary protein excretion reflects a functional change in the
4 nephron, is discussed in Section 4.5.2. Because the changes are consistent with a known mode
5 of action and represent part of the progression of effects leading to toxicity, they are considered
6 to be adverse and indicate that the LOAEL and NOAEL are 60 and 30 |j,g/kg-day, respectively.
7
8 5.2.4.2. Methods of Analysis - Including Models (PBPK, BMD, etc.)
9
10 A point of departure (POD) can be determined using the kidney weight data and BMD
11 modeling, but BMD analysis of the urinary protein and histopathology data is precluded by
12 insufficient data. In particular, the urinary protein data are limited by inadequate reporting
13 (mean concentrations and errors are conveyed in a bar graph with no numerical values
14 specifically reported, no indication if the error bars represent standard deviation or standard
15 error, and no indication of numbers of animals) and the pathology findings are limited by a lack
16 of incidence data.
17
18 In accordance with current BMD technical guidance (U.S. EPA, 2000c), available
19 continuous-variable models in the EPA Benchmark Dose Software (BMDS version 1.3.2; linear,
20 polynomial, power and Hill models) were fit to the data for changes in mean relative kidney
21 weight shown in Table 5-1. Statistical tests in the BMDS showed that variance was
22 homogeneous across dose groups. BMDs and BMDLs were calculated using 1 standard
23 deviation above the control mean as the benchmark response level (BMR), while assuming
24 homogenous variance across groups. Using data from all dose groups, an adequate fit to the data
25 was obtained with the Hill model (Table 5-2), but the BMDS was not able to compute a BMDL.
26 After dropping the high dose group, there were insufficient degrees of freedom remaining to fit
27 the Hill model, but the linear model adequately fit the data and produced an estimated BMD of
28 43.1 |j,g/kg-day and BMDL of 33.1 |j,g/kg-day. The two-degree polynomial and power models
29 defaulted to the same linear model, albeit with lower p-value and/or higher Aikake's Information
30 Criteria (AIC) due to the greater number of parameters in these models. The BMD modeling
31 results are summarized in Table 5-2 and detailed in Appendix A, and the fit of the linear model
32 to the data is shown in Figure 5-1. The BMDL of 33.1 |j,g/kg-day is similar to the 30 |j,g/kg-day
33 NOAEL for increased kidney weight and is used as the POD for the RfD.
34
35 5.2.4.3. RfD Derivation - Including Application of UFs
36
37 The BMDL of 33.1 |j,g/kg-day for increased relative kidney weight was used as the point
38 of departure (POD) for the subchronic RfD. Dividing the BMDL of 33.1 |j,g/kg-day by a
39 composite uncertainty factor (UF) of 1000 results in a subchronic RfD for cylindrospermopsin of
40 3xlO"5 mg/kg-day.
41
28 DRAFT: DO NOT CITE OR QUOTE
-------
Table 5-1. Relative Kidney Weights in Mice Exposed to Purified Cylindrospermopsin for 1 1 Weeks
Relative kidney weight
(mean + standard deviation)
Oral Dose (|j,g/kg-day)
0
1.48 + 0.10
(10)a
30
1.57 + 0.14
(10)
60
1.66 + 0.16b
(9)
120
1.82 + 0.12C
(9)
240
1.78 + 0.17C
(6)
2 a Values in parentheses are the number of animals evaluated in each group
3 b Statistically significant difference from controls (p<0.05)
4 c Statistically significant difference from controls (p<0.001)
5 Source: Humpage and Falconer (2003)
29
DRAFT: DO NOT CITE OR QUOTE
-------
Table 5-2. Summary of Benchmark Dose Modeling (Relative Kidney Weight/
Model Fit to Means
df
p-Value for
Model Fit
AIC for
Fitted
Model
BMD
(|ig/kg-day)
BMDL
(Hg/kg-day)
Relative kidney weight, all dose groups
(p=0.59 for test of homogenous variance, indicating assumption of homogenous variance is
appropriate)
Linear
2-Degree polynomial (pos
betas)
Power (power >=1)
Hill (power >=1)
3
2
2
1
0.01
0.003
0.003
0.21
-120.85
-120.85
-116.85
-124.74
106.07
106.07
106.07
43.19b
76.56
76.56
76.56
NAC
Relative kidney weight, high dose group dropped
(p=0.52 for test of homogenous variance, indicating assumption of homogenous variance is
appropriate)
Linear
2-Degree polynomial (pos
betas)
Power (power >=1)
Hill (power >=1)
2
1
1
0
0.98
0.84
0.84
NAd
-116.47
-116.47
-112.47
-110.51
43.90
43.90
43.90
41.20
33.07
33.07
33.07
21.72
2 a Modeling conducted assuming homogenous variance and using BMR of 1 standard deviation
3 b Optimum BMD may not have been found (i.e. bad completion code in the BMDS optimization
4 routine)
5 c BMDL computation failed
6 d The Chi-Square test for fit is not valid due to insufficient degrees of freedom (df)
7 NA = not available
30
DRAFT: DO NOT CITE OR QUOTE
-------
Linear Model with 0.95 Confidence Level
1
2
o
J
4
5
cu
tn
c
WJ
cu
CC
(U
1.9
1.8
1.7
1.5
1.4
Linear
22:31 05/122005
BMDL
BMD
20
40
60
dose
80
100
120
Figure 5-1. Linear Model Fit to Relative Kidney Weight Data (High Dose Group Dropped)
Source: Humpage and Falconer (2003)
DRAFT: DO NOT CITE OR QUOTE
-------
1 Subchronic RfD = BMDL-UF
2 = 33.1 ng/kg-day-1000
3 = 0.00003 mg/kg-day or 3x105 mg/kg-day
4
5 The composite UF of 1000 includes a factor of 10 for interspecies extrapolation, a factor
6 of 10 to account for interindividual variability in the human population and a factor of 10 for
7 database limitations, as follows.
8
9 • A default 10-fold UF is used to account for the interspecies variability in extrapolating
10 from laboratory animals to humans. No information is available on the toxicity of
11 purified cylindrospermopsin in humans, and no data on toxicokinetic differences between
12 animals and humans in the disposition of ingested cylindrospermopsin are available.
13 • A 10-fold UF is used to account for variation in sensitivity within human populations
14 because there is insufficient information on the degree to which humans of varying
15 gender, age, health status or genetic makeup might vary in the disposition of, or response
16 to, ingested cylindrospermopsin. As discussed in Section 4.1, data from the Palm Island
17 outbreak of a hepatoenteritis-like illness (Blyth, 1980; Griffiths and Saker, 2003) suggest
18 a possible increased sensitivity of children to cylindrospermopsin.
19 • A 10-fold UF is used to account for deficiencies in the database. There is no information
20 on the longer-term toxicity of cylindrospermopsin in humans. Other database
21 deficiencies include a lack of particular kinds of animal studies on purified
22 cylindrospermopsin, including a chronic study, subchronic or chronic studies in a second
23 species and reproductive and developmental toxicity studies.
24 The NOAEL/LOAEL approach and an UF of 1000 would also yield an RfD of 0.00003
25 mg/kg-day due to the similarity of the NOAEL and BMDL for increased kidney weight (30 and
26 33.1 |j,g/kg-day, respectively).
27
28 5.2.5. Chronic Duration
29
30 5.2.5.1. Choice of Principal Study and Critical Effect - with Rationale and
31 Justification
32
33 Derivation of a chronic oral RfD for cylindrospermopsin is precluded by insufficient
34 data. No information is available on the chronic toxicity of cylindrospermopsin by any route of
35 exposure. The 11-week study (Humpage and Falconer, 2003) used to derive the subchronic RfD
36 was considered for use in the derivation of a chronic RfD; however, this approach was rejected
37 due to the lack of information on the potential progression of cylindrospermopsin-induced
38 adverse effects with increased exposure duration. The use of the POD from the 11-week
39 subchronic study for the derivation of a chronic RfD would require the application of a
40 subchronic-to-chronic UF to account for the uncertainties involved in extrapolating across
41 exposure durations. The application of a full subchronic-to-chronic UF of 10, along with UFs of
42 10 in three other areas of uncertainty (interspecies UF, intraspecies UF, database UF), would
43 result in a total composite uncertainty factor of 10,000. A composite uncertainty factor of this
32 DRAFT: DO NOT CITE OR QUOTE
-------
1 magnitude suggests that the database is insufficient to support the derivation of an RfD for
2 chronic exposure; therefore, no chronic oral RfD is derived.
3
4 5.2.6. Route-to-Route Extrapolation
5
6 Derivation of acute, short-term, and chronic RfD values for cylindrospermopsin by route-
7 to-route extrapolation could not be considered due to a lack of inhalation data.
8
9 5.3. INHALATION REFERENCE CONCENTRATION (RfC)
10
11 No information is available on the toxicity of inhaled cylindrospermopsin.
12
13 5.4. CANCER ASSESSMENT
14
15 No dose-response or other information is available regarding the carcinogenicity of pure
16 cylindrospermopsin.
33 DRAFT: DO NOT CITE OR QUOTE
-------
1 6. MAJOR CONCLUSIONS IN THE CHARACTERIZATION OF
2 HAZARD AND DOSE RESPONSE
3
4
5 6.1. HUMAN HAZARD POTENTIAL
6
7 Cylindrospermopsin is a naturally occurring chemical produced by Cylindrospermopsis
8 (particularly C. raciborskii) and at least four other genera of freshwater cyanobacteria.
9 Toxicokinetic studies of Cylindrospermopsin have not been performed using natural routes of
10 exposure, but oral toxicity studies show that it is absorbed from the gastrointestinal tract, and i.p.
11 toxicokinetic studies indicate that it is mainly distributed to the liver and excreted in the urine as
12 unmetabolized compound. Main targets of Cylindrospermopsin toxicity include the liver and
13 kidneys, and possible modes of action include inhibition of protein synthesis, bioactivation to a
14 reactive intermediate and covalent binding of parent compound or a metabolite to DNA and/or
15 RNA.
16
17 The main information on the toxicity of Cylindrospermopsin in humans is from
18 qualitative reports of a hepatoenteritis-like illness that is attributed to the acute or short-term
19 consumption of drinking water containing C. raciborskii. The database on oral toxicity of
20 purified Cylindrospermopsin in animals is limited by a small number of studies and insufficient
21 reporting. No studies have been performed assessing the acute oral toxicity of pure
22 Cylindrospermopsin. Information on short-term oral toxicity is available from inadequately
23 reported 14- and 21-day studies in mice and rats. Data on the subchronic oral toxicity of pure
24 Cylindrospermopsin are available from a comprehensive 11-week study that identified NOAELs
25 and LOAELs for kidney and liver effects in mice. No chronic toxicity, reproductive toxicity,
26 developmental toxicity or carcinogenicity studies of Cylindrospermopsin have been conducted.
27 Testing following inhalation has not been performed.
28
29 6.2. DOSE RESPONSE
30
31 Kidney effects data in the 11-week toxicity study (Humpage and Falconer, 2003) provide
32 a suitable basis for deriving a subchronic oral RfD. The most sensitive effect in this study was
33 increased relative kidney weight; decreased urinary protein and minor histopathological damage
34 to the liver occurred at the next highest dose. As discussed in Section 4.5.2, increased kidney
35 weight and decreased urinary protein are consistent with suppressed protein synthesis, a known
36 mode of action of Cylindrospermopsin, and represent part of the progression of effects leading to
37 toxicity. Based on a BMDL of 33.1 |j,g/kg-day for increased relative kidney weight in mice, a
38 subchronic RfD of 3xlO"5 mg/kg-day was derived by dividing the BMDL by a UF of 1000. The
39 UF comprises component factors of 10 for interspecies extrapolation, 10 for interindividual
40 variability and 10 for database deficiencies. Acute, short-term and chronic oral RfDs could not
41 be derived due to inadequate data. Inhalation RfC derivation is precluded by the lack of data for
42 this route of exposure. There is inadequate information to evaluate the carcinogenicity of
43 Cylindrospermopsin.
34 DRAFT: DO NOT CITE OR QUOTE
-------
1 7. REFERENCES
2
3 Akimov, S.S. and A.M. Belkin. 2001. Cell surface tissue transglutaminase is involved in
4 adhesion and migration of monocytic cells on fibronectin. Blood. 98(5): 1567-1576.
5 Bachmann, S., A.B. Dawnay, N. Bouby and L. Bankir. 1991. Tamm-Horsfall protein excretion
6 during chronic alterations in urinary concentration and protein intake in the rat. Renal Physiol.
7 Biochem. 14(6):236-245.
8 Bachmann, S., K. Mutig, J. Bates et al. 2005. Renal effects of Tamm-Horsfall protein
9 (uromodulin) deficiency in mice. Am. J. Renal Physiol. 288(3):F559-F567.
10 Banker, R., B. Teltsch, A. Sukenik and S. Carmeli. 2000. 7-Epicylindrospermopsin, a toxic
11 minor metabolite of the cyanobacterium Aphanizomenon ovalisporum from Lake Kinneret,
12 Israel. J. Nat. Prod. 63(3):387-389.
13 Banker, R., S. Carmeli, M. Werman et al. 2001. Uracil moiety is required for toxicity of the
14 cyanobacterial hepatotoxin cylindrospermopsin. J. Toxicol. Environ. Health A. 62(4):281-288.
15 Bates, J.M., H.M. Raffi, K. Prasadan et al. 2004. Tamm-Horsfall protein knockout mice are
16 more prone to urinary tract infection: Rapid communication. Kidney Int. 65(3):791-797.
17 Blyth, S. 1980. Palm Island mystery disease. Med. J. Aust. 2(l):40-42.
18 Carmichael, W.W., S.M.F.O. Azevedo, IS. An et al. 2001. Human fatalities from
19 cyanobacteria: Chemical and biological evidence for cyanotoxins. Environ. Health Perspect.
20 109(7):663-668.
21 Chiswell, R.K., G.R. Shaw, G. Eaglesham et al. 1999. Stability of cylindrospermopsin, the
22 toxin from cyanobacterium, Cylindrospermopsis raciborskii: Effect of pH, temperature, and
23 sunlight on decomposition. Environ. Toxicol. 14(1):155-161.
24 Falconer, I.R. and A.R. Humpage. 2001. Preliminary evidence for in vivo tumour initiation by
25 oral administration of extracts of the blue-green alga Cylindrospermopsis raciborski containing
26 the toxin cylindrospermopsin. Environ. Toxicol. 16(2): 192-195.
27 Falconer, I.R., SJ. Hardy, A.R. Humpage et al. 1999. Hepatic and renal toxicity of the blue-
28 green alga (cyanobacterium) Cylindrospermopsis raciborski in male Swiss albino mice.
29 Environ. Toxicol. 14(1):143-150.
30 Fastner, J., R. Heinze, A.R. Humpage et al. 2003. Cylindrospermopsin occurrence in two
31 German lakes and preliminary assessment of toxicity and toxin production of
32 Cylindrospermopsis raciborskii (cyanobacteria) isolates. Toxicon. 42(3):313-321.
33 Fessard, V. and C. Bernard. 2003. Cell alterations but no DNA strand breaks induced in vitro
34 by cylindrospermopsin in CHO Kl cells. Environ. Toxicol. 18(5):353-359.
35 DRAFT: DO NOT CITE OR QUOTE
-------
1 Fesus, L., V. Thomazy and A. Falus. 1987. Induction and activation of tissue transglutaminase
2 during programmed cell death. FEES Lett. 224(1): 104-108.
3 Froscio, S.M., A.R. Humpage, P.C. Burcham and I.R. Falconer. 2003. Cylindrospermopsin-
4 induced protein synthesis inhibition and its dissociation from acute toxicity in mouse
5 hepatocytes. Environ. Toxicol. 18(4):243-251.
6 Grenard, P., S. Bresson-Hadni, S. El Alaoui et al. 2001. Transglutaminase-mediated cross-
7 linking is involved in the stabilization of extracellular matrix in human liver fibrosis. J. Hepatol.
8 35(3):367-375.
9 Griffiths, DJ. and M.L. Saker. 2003. The Palm Island mystery disease 20 years on: A review of
10 research on the cyanotoxin cylindrospermopsin. Environ. Toxicol. 18(2):78-93.
11 Hawkins, P.R., M.T.C. Runnegar, A.R.B. Jackson and I.R. Falconer. 1985. Severe
12 hepatotoxicity caused by the tropical cyanobacterium (blue-green alga) Cylindrospermopsis
13 raciborskii (Woloszynska) Seenaya and Subba Raju isolated from a domestic water supply
14 reservoir. Appl. Environ. Microbiol. 50(5): 1292-1295.
15 Hawkins, P.R., N.R. Chandrasena, GJ. Jones et al. 1997. Isolation and toxicity of
16 Cylindrospermopsis raciborskii from an ornamental lake. Toxicon. 35(3):341-346.
17 Hayman, J. 1992. Beyond the Barcoo - probable human tropical cyanobacterial poisoning in
18 outback Australia. Med. J. Aust. 157(11-12):794-796.
19 HSDB (Hazardous Substances Data Bank). 2006. Produced by the U.S. National Library of
20 Medicine (NLM), Toxicology Data Network (TOXNET), Bethesda, Maryland. Accessed
21 January 2, 2006 at http://toxnet.nlm.nih.gov/cgi-bin/sis/htmlgen7HSDB.
22 Humpage, A.R. and I.R. Falconer. 2003. Oral toxicity of the cyanobacterial toxin
23 cylindrospermopsin in male Swiss albino mice: Determination of no observed adverse effect
24 level for deriving a drinking water guideline value. Environ. Toxicol. 18(2):94-103.
25 Humpage, A.R., M. Fenech, P. Thomas and I.R. Falconer. 2000. Micronucleus induction and
26 chromosome loss in transformed human white cells indicate clastogenic and aneugenic action of
27 the cyanobacterial toxin, cylindrospermopsin. Mutat. Res. 472:155-161.
28 Humpage, A.R., F. Fontaine, S. Froscio, P. Burcham and I.R. Falconer. 2005.
29 Cylindrospermopsin genotoxicity and cytotoxicity: Role of cytochrome P-450 and oxidative
30 stress. J. Toxicol. Environ. Health, Part A. 68(9):739-753.
31 Lewis, R.J. 2000. Sax's Dangerous Properties of Industrial Materials, Vol. 1-3, 10th ed. John
32 Wiley & Sons Inc., New York, NY. p. 1061.
33 Li, R., W.W. Carmichael, S. Brittain et al. 2001. First report of the cyanotoxins
34 cylindrospermopsin and deoxycylindrospermopsin from Raphidiopsis curvata (cyanobacteria).
35 J. Phycol. 37(6): 1121-1126.
36 DRAFT: DO NOT CITE OR QUOTE
-------
1 Mirza, A., S-L. Liu, E. Frizell et al. 1997. A role for tissue transglutaminase in hepatic injury
2 and fibrogenesis, and its regulation by NF-KB. Am. J. Physiol. 272:G281-G288.
3 Norris, R.L., G.K. Eaglesham, G. Pierens et al. 1999. Deoxycylindrospermopsin, an analog of
4 cylindrospermopsin from Cylindrospermopsis raciborskii. Environ. Toxicol. 14(1): 163-165.
5 Norris, R.L.G., A. A. Seawright, G.R. Shaw et al. 2001. Distribution of 14C cylindrospermopsin
6 in vivo in the mouse. Environ. Toxicol. 16(6):498-505.
7 Norris, R.L.G., A.A. Seawright, G.R. Shaw et al. 2002. Hepatic xenobiotic metabolism of
8 cylindrospermopsin in vivo in the mouse. Toxicon. 40(4):471-476.
9 NRC (National Research Council). 1983. Risk Assessment in the Federal Government:
10 Managing the Process. National Academy Press, Washington, DC.
11 Ohtani, I, R.E. Moore andM.T.C. Runnegar. 1992. Cylindrospermopsin: A potent hepatotoxin
12 from the blue-green alga Cylindrospermopsis raciborskii. J. Am. Chem. Soc.
13 114(20):7941-7942.
14 O'Neil, M.J., Ed. 2001. The Merck Index - An Encyclopedia of Chemicals, Drugs, and
15 Biologicals, 13th ed. Merck and Co., Inc., Whitehouse Station, NJ.
16 Pesce, AJ. and M.R. First. 1979. Proteinuria. An integrated review. Marcell Dekker.
17 pp. 54-79
18 Piacentini, M., M.G. Farrace, L. Piredda et al. 2002. Transglutaminase overexpression
19 sensitizes neuronal cell lines to apoptosis by increasing mitochondrial membrane potential and
20 cellular oxidative stress. J. Neurochem. 81(5):1061-1072.
21 Pilotto, L., P. Hobson, M.D. Burch, G. Ranmuthugala, R. Attewell and W. Weightman. 2004.
22 Acute skin irritant effects of cyanobacteria (blue-green algae) in healthy volunteers. Aust. N. Z.
23 J. Public Health. 28(3):220-224.
24 Runnegar, M.T., S.M. Kong, Y-Z. Zhong et al. 1994. The role of glutathione in the toxicity of a
25 novel cyanobacterial alkaloid cylindrospermopsin in cultured rat hepatocytes. Biochem.
26 Biophys. Res. Commun. 201(1):235-241.
27 Runnegar, M.T., S.M. Kong, Y-Z. Zhong and S.C. Lu. 1995. Inhibition of reduced glutathione
28 synthesis by cyanobacterial alkaloid cylindrospermopsin in cultured rat hepatocytes. Biochem.
29 Pharmacol. 49(2):219-225.
30 Runnegar, M.T., C. Xie, B.B. Snider et al. 2002. In vitro hepatotoxicity of the cyanobacterial
31 alkaloid cylindrospermopsin and related synthetic analogues. Toxicol. Sci. 67(l):81-87.
32 Schoel, B. and G. Pfleiderer. 1987. The amount of Tamm-Horsfall protein in the human kidney,
33 related to its daily excretion. J. Clin. Chem. Clin. Biochem. 25(10):681-682.
37 DRAFT: DO NOT CITE OR QUOTE
-------
1 Seawright, A.A., C.C. Nolan, G.R. Shaw et al. 1999. The oral toxicity for mice of the tropical
2 cyanobacterium Cylindrospermopsin raciborskii (Woloszynska). Environ. Toxicol.
3 14(1):135-142.
4 Shaw, G.R., A. Sukenik, A. Livne et al. 1999. Blooms of the Cylindrospermopsin containing
5 cyanobacterium, Aphanizomenon ovalisporum (Forti), in newly constructed lakes, Queensland,
6 Australia. Environ. Toxicol. 14(1): 167-177.
7 Shaw, G.R., A.A. Seawright, M.R. Moore and P.K. Lam. 2000. Cylindrospermopsin, a
8 cyanobacterial alkaloid: Evaluation of its toxicologic activity. Ther. Drug Monit. 22(l):89-92.
9 Shaw, G.R., A.A. Seawright and M.R. Moore. 2001. Toxicology and human health implications
10 of the cyanobacterial toxin Cylindrospermopsin. In: Mycotoxins and Phycotoxins in Perspective
11 at the Turn of the Millennium, WJ. Dekoe, R.A. Samson, H.P. van Egmond et al., Ed. IUPAC
12 & AOAC International, Brazil, p. 435-443.
13 Shen, X., P.K.S. Lam, G.R. Shaw and W. Wickramasinghe. 2002. Genotoxicity investigation of
14 a cyanobacterial toxin, Cylindrospermopsin. Toxicon. 40(10):1499-1501.
15 Shen, X., G.R. Shaw, G.A. Codd et al. 2003. DNA microarray analysis of gene expression in
16 mice treated with the cyanobacterial toxin, Cylindrospermopsin. In: Proceedings of the Eighth
17 Canadian Workshop on Harmful Marine Algae, S.S. Bates, Ed. Fisheries and Oceans Canada,
18 Monkton, New Brunswick, p. 49-51. Available at
19 http://www.glf. dfo-mpo.gc.ca/sci-sci/cwhma-atcamn/8th_cwhma_proceedings.pdf
20 Sigma. 2006. Cylindrospermopsin from Cylindrospermopsis raciborskii. C9866. Sigma-
21 Aldrich, Inc., Saint Louis, MO. Accessed January 23, 2006 at
22 http://www.sigmaaldrich.com/catalog/search/ProductDetail?ProdNo=C9866&Brand=SIGMA.
23 Terao, K., S. Ohmori, K. Igarashi et al. 1994. Electron microscopic studies on experimental
24 poisoning in mice induced by Cylindrospermopsin isolated from blue-green alga Umezakia
25 natans. Toxicon. 32(7):833-843.
26 U.S. EPA. 1986a. Guidelines for the Health Risk Assessment of Chemical Mixtures. Fed. Reg.
27 51(185):34014-34025.
28 U.S. EPA. 1986b. Guidelines for Mutagenicity Risk Assessment. Fed. Reg.
29 51(185):34006-34012.
30 U.S. EPA. 1988. Recommendations for and Documentation of Biological Values for Use in
31 Risk Assessment. Prepared by the Office of Health and Environmental Assessment,
32 Environmental Criteria and Assessment Office, Cincinnati, OH for the Office of Solid Waste and
33 Emergency Response, Washington, DC. EPA 600/6-87/008. NTIS PB88-179874/AS.
34 U.S. EPA. 1991. Guidelines for Developmental Toxicity Risk Assessment. Fed. Reg.
35 56(234):63798-63826.
38 DRAFT: DO NOT CITE OR QUOTE
-------
1 U.S. EPA. 1994a. Interim Policy for Particle Size and Limit Concentration Issues in Inhalation
2 Toxicity Studies: Notice of Availability. Fed. Reg. 59(206):53799.
3 U.S. EPA. 1994b. Methods for Derivation of Inhalation Reference Concentrations and
4 Application of Inhalation Dosimetry. U.S. Environmental Protection Agency, Office of
5 Research and Development,Washington, DC. EPA/600/8-90/066F. NTIS PB2000-500023.
6 Available at http://www.epa.gov/iris/backgr-d.htm.
7 U.S. EPA. 1995. Use of the Benchmark Dose Approach in Health Risk Assessment. U.S.
8 Environmental Protection Agency, Risk Assessment Forum, Washington, DC.
9 EPA/630/R-94/007. NTIS PB95-213765. Available at http://www.epa.gov/iris/backgr-d.htm.
10 U.S. EPA. 1996. Guidelines for Reproductive Toxicity Risk Assessment. Fed. Reg.
11 61(212):56274-56322.
12 U.S. EPA. 1998a. Guidelines for Neurotoxicity Risk Assessment. Fed. Reg.
13 63(93):26926-26954.
14 U.S. EPA. 1998b. Science Policy Council Handbook: Peer Review. U.S. Environmental
15 Protection Agency, Office of Research and Development, Washington, DC. EPA/100/B-98/001.
16 NTIS PB98-140726.
17 U.S. EPA. 2000a. Science Policy Council Handbook: Peer Review, 2nd ed. U.S. Environmental
18 Protection Agency, Office of Research and Development, Washington, DC. EPA/100/B-00/001.
19 Available at http://www.epa.gov/iris/backgr-d.htm.
20 U.S. EPA. 2000b. Science Policy Council Handbook: Risk Characterization. U.S.
21 Environmental Protection Agency, Office of Research and Development, Washington, DC.
22 EPA/ 100/B-00/002. Available at http://www.epa.gov/iris/backgr-d.htm.
23 U.S. EPA. 2000c. Benchmark Dose Technical Guidance Document [External Review Draft].
24 U.S. Environmental Protection Agency, Risk Assessment Forum, Washington, DC.
25 EPA/630/R-00/001. Available at http://www.epa.gov/iris/backgr-d.htm.
26 U.S. EPA. 2000d. Supplementary Guidance for Conducting Health Risk Assessment of
27 Chemical Mixtures. U.S. Environmental Protection Agency, Office of Research and
28 Development, Washington, DC. EPA/630/R-00/002. Available at
29 http://www.epa.gov/iris/backgr-d.htm.
30 U.S. EPA. 2002. A Review of the Reference Dose and Reference Concentration Processes.
31 U.S. Environmental Protection Agency, Risk Assessment Forum, Washington, DC.
32 EPA/630/P-02/002F. Available at http://www.epa.gov/iris/backgr-d.htm.
33 U.S. EPA. 2005a. Guidelines for Carcinogen Risk Assessment. U.S. Environmental Protection
34 Agency, Risk Assessment Forum, Washington, DC. EPA/630/P-03/001B. Available at
35 http://www.epa.gov/iris/backgr-d.htm.
39 DRAFT: DO NOT CITE OR QUOTE
-------
1 U.S. EPA. 2005b. Supplemental Guidance for Assessing Susceptibility from Early-Life
2 Exposure to Carcinogens. U.S. Environmental Protection Agency, Risk Assessment Forum,
3 Washington, DC. EPA/630/R-03/003F. Available at http://www.epa.gov/iris/backgr-d.htm.
4 WHO (World Health Organization). 1999. Toxic Cyanobacteria in Water: A Guide to Their
5 Public Health Consequences, Monitoring, and Management, I. Chorus and J. Bartram, Ed. ISBN
6 0-419-23930-8. World Health Organization, Geneva, Switzerland. Available at
7 http://www.who.int/water_sanitation_health/resourcesquality/toxcy anobacteria.pdf
8 Zhang, L-X., KJ. Mills, M.L. Dawson et al. 1995. Evidence for the involvement of retinoic
9 acid receptor RARa-dependent signaling pathway in the induction of tissue transglutaminase and
10 apostosis by retinoids. J. Biol. Chem. 270(11):6022-6029.
40 DRAFT: DO NOT CITE OR QUOTE
-------
APPENDIX A
BENCHMARK DOSE MODELING RESULTS FOR CYLINDROSPERMOPSIN
A-l
-------
Part I.
Humpage and Falconer 2003
male mice treated with purified cylindrospermopsin
rel kidney wt
A-2
-------
Polynomial Model. Revision: 2.2 Date: 9/12/2002
Input Data File: C:\BMDS\DATA\CYLINDRO.(d)
Gnuplot Plotting File: C:\BMDS\DATA\CYLINDRO.plt
Thu May 12 22:04:04 2005
HMDS MODEL RUN
The form of the response function is:
Y[dose] = beta 0 + beta l*dose + beta 2*doseA2 + ...
Dependent variable = MEAN
Independent variable = dose
rho is set to 0
Signs of the polynomial coefficients are not restricted
A constant variance model is fit
Total number of dose groups = 5
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
alpha = 0.018741
rho = 0 Specified
beta_0 = 1.551
beta 1 = 0.00123333
Parameter Estimates
Variable
alpha
beta_0
beta 1
95.0% Wald Confidence Interval
Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
0.0215477 0.00459399 0.0125437 0.0305518
1.54205 0.0311925 1.48091 1.60318
0.00138389 0.000287871 0.000819671
0.00194811
Asymptotic Correlation Matrix of Parameter Estimates
alpha
beta_0
beta 1
alpha
1
2.1e-009
1.2e-010
beta_0
2.1e-009
1
-0.7
beta_l
1.2e-010
-0.7
1
Table of Data and Estimated Values of Interest
A-3
-------
Dose
Res .
0
30
60
120
240
N
10
10
9
9
6
Obs
1.
1.
1.
1.
1.
Mean
48
57
66
82
78
Obs
0
0
0
0
Std Dev
0.1
.14
.16
.12
.17
Est
1.
1.
1.
1.
1.
Mean
54
58
63
71
87
Est
0.
0.
0.
0.
0.
Std Dev
147
147
147
147
147
ChiA
-
-0
0
-
2
1.34
.292
.714
2.29
1.57
Model Descriptions for likelihoods calculated
Model Al: Yij
Var{e(ij)}
Model A2: Yij
Var{e(ij) }
Model R: Yi
Var{e(i) }
Mu (i) + e (ij )
SigmaA2
Mu (i) + e (ij )
Sigma(i)A2
Mu + e (i)
SigmaA2
Likelihoods of Interest
Model
Al
A2
fitted
R
Log (likelihood)
68.148702
69.554943
62.424690
52.631671
DF
6
10
2
2
AIC
-124.297404
-119.109885
-120.849381
-101.263343
Test 1: Does response and/or variances differ among dose
levels
(A2 vs. R)
Test 2: Are Variances Homogeneous (Al vs A2)
Test 3: Does the Model for the Mean Fit (Al vs. fitted)
Test
Test 1
Test 2
Test 3
Tests of Interest
-2*log(Likelihood Ratio) Test df
33.8465
2.81248
11.448
p-value
<.0001
0.5897
0.009534
The p-value for Test 1 is less than .05. There appears
to be a
difference between response and/or variances among the
dose levels.
It seems appropriate to model the data
The p-value for Test 2 is greater than .05. A
homogeneous variance
model appears to be appropriate here
A-4
-------
The p-value for Test 3 is less than .05. You may want
to try a
different model
Benchmark Dose Computation
Specified effect = 1
Risk Type
Estimated standard deviations from the control mean
Confidence level =
BMD =
0.95
106.072
BMDL =
76.5569
Linear Model with 0.95 Confidence Level
1.9
o>
V)
c
o
Q.
v)
CD
o:
c
CD
1.7
1.6
1.5
1.4
Linear
BMD
BMP
50
100 150
dose
22:0405/122005
200
250
A-5
-------
Polynomial Model. Revision: 2.2 Date: 9/12/2002
Input Data File: C:\BMDS\DATA\CYLINDRO.(d)
Gnuplot Plotting File: C:\BMDS\DATA\CYLINDRO.plt
Thu May 12 22:27:06 2005
HMDS MODEL RUN
The form of the response function is:
Y[dose] = beta 0 + beta l*dose + beta 2*doseA2 + ...
Dependent variable = MEAN
Independent variable = dose
rho is set to 0
The polynomial coefficients are restricted to be positive
A constant variance model is fit
Total number of dose groups = 5
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
alpha =
rho =
beta_0 =
beta_l =
beta 2 =
0.018741
0
1.46523
0
0
Specified
Parameter Estimates
Variable Estimate
alpha 0.0215477
beta 0 1.54205
beta_l
beta 2
Std. Err.
0.00459399
0.0311925
0.00138389 0.000287871
0
NA
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
0.0125437 0.0305518
1.48091 1.60318
0.000819671
0.00194811
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Asymptotic Correlation Matrix of Parameter Estimates
alpha
beta_0
beta 1
alpha
1
1.4e-010
7.2e-011
beta_0
1.4e-010
1
-0.7
beta_l
7.2e-011
-0.7
1
A-6
-------
The following parameter(s) have been estimated at a
boundary
point or have been specified. Correlations are not
computed:
beta 2
Table of Data and Estimated Values of Interest
Dose
Res .
0
30
60
120
240
N
10
10
9
9
6
Obs
1.
1.
1.
1.
1.
Mean
48
57
66
82
78
Obs
0
0
0
0
Std Dev
0.1
.14
.16
.12
.17
Est
1.
1.
1.
1.
1.
Mean
54
58
63
71
87
Est
0.
0.
0.
0.
0.
Std Dev
147
147
147
147
147
ChiA2
-1
-0.
0.
2
-1
.34
292
714
.29
.57
Model Descriptions for likelihoods calculated
Model Al: Yij
Var{e(ij)}
Model A2: Yij
Var{e(ij) }
Model R: Yi
Var{e(i) }
Mu (i) + e (ij )
S i gma A 2
Mu (i) + e (ij )
Sigma(i)A2
Mu + e (i)
SigmaA2
Likelihoods of Interest
Model Log (likelihood)
Test 1:
levels
Test 2:
Test 3:
Al
A2
fitted
R
Does response
(A2 vs. R)
Are Variances
Does the Model
68.148702
69.554943
62.424690
52.631671
and/or variances
DF
6
10
2
2
differ
AIC
124.297404
119.109885
120.849381
101.263343
among dose
Homogeneous (Al vs A2 )
for the Mean Fit
(Al vs
. fitted)
Tests of Interest
Test
Test 1
Test 2
Test 3
-2*log(Likelihood Ratio) Test df
33.8465
2.81248
11.448
p-value
<.0001
0.5897
0.003267
A-7
-------
The p-value for Test 1 is less than .05. There appears
to be a
difference between response and/or variances among the
dose levels.
It seems appropriate to model the data
The p-value for Test 2 is greater than .05. A
homogeneous variance
model appears to be appropriate here
The p-value for Test 3 is less than .05. You may want
to try a
different model
Benchmark Dose Computation
Specified effect = 1
Risk Type = Estimated standard deviations from the control mean
Confidence level = 0.95
BMD = 106.072
BMDL = 76.5569
A-8
-------
Polynomial Model with 0.95 Confidence Level
1.9
o>
V)
c
o
Q.
w
CD
o:
c
CD
1.7
1.6
1.5
1.4
Polynomial
BMD
BMP
50
100
150
200
250
dose
22:2705/122005
Power Model. $Revision: 2.1 $ $Date: 2000/10/11 20:57:36 $
Input Data File: C:\BMDS\DATA\CYLINDRO.(d)
Gnuplot Plotting File: C:\BMDS\DATA\CYLINDRO.plt
Thu May 12 22:28:38 2005
HMDS MODEL RUN
The form of the response function is:
Y[dose] = control + slope * doseApower
Dependent variable = MEAN
Independent variable = dose
rho is set to 0
The power is restricted to be greater than or equal to 1
A constant variance model is fit
Total number of dose groups = 5
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
A-9
-------
Default Initial Parameter Values
alpha = 0.018741
Asymptotic
alpha
alpha 1
rho -1
control NA
slope NA
power NA
NA - This parameter's
Variable
alpha
rho
control
slope
power
Table of Data and
Dose N Obs M
Res .
0 10 1.48
30 10 1.57
60 9 1.66
120 9 1.82
240 6 1.78
rho = 0 Specified
control = 1.48
slope = 0.0132605
power = 0. 612843
Correlation Matrix of Parameter Estimates
rho control slope
- 1 NA NA
1 NA NA
NA NA NA
NA NA NA
NA NA NA
variance has been estimated at zero.
Parameter Estimates
Estimate Std. Err.
0.0215477 0.0399798
0 3.81654
1.54205 0.0380886
0.00138389 0.00299534
1 0.337065
Estimated Values of Interest
ean Obs Std Dev Est Mean Est Std Dev
0.1 1.54 0.147
0.14 1.58 0.147
0.16 1.63 0.147
0.12 1.71 0.147
0.17 1.87 0.147
power
NA
NA
NA
NA
NA
ChiA2
-0.423
-0.0924
0.238
0.762
-0.642
Model Descriptions for likelihoods calculated
Model Al: Yij = Mu(i) + e(ij)
Var{e(ij)} = SigmaA2
A-10
-------
Model A2: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma(i)A2
Model R: Yi = Mu + e(i)
Var{e(i)} = SigmaA2
Likelihoods of Interest
Model Log(likelihood) DF AIC
Al 68.148702 6 -124.297404
A2 69.554943 10 -119.109885
fitted 62.424690 4 -116.849381
R 52.631671 2 -101.263343
Test 1: Does response and/or variances differ among dose levels
(A2 vs. R)
Test 2: Are Variances Homogeneous (Al vs A2)
Test 3: Does the Model for the Mean Fit (Al vs. fitted)
Tests of Interest
Test -2*log(Likelihood Ratio) df p-value
Test 1 33.8465 8 <.00001
Test 2 2.81248 4 0.5897
Test 3 11.448 2 0.003267
The p-value for Test 1 is less than .05. There appears to be a
difference between response and/or variances among the dose levels.
It seems appropriate to model the data
The p-value for Test 2 is greater than .05. A homogeneous variance
model appears to be appropriate here
The p-value for Test 3 is less than .05. You may want to try a
different model
Benchmark Dose Computation
Specified effect = 1
Risk Type = Estimated standard deviations from the control mean
Confidence level = 0.95
BMD = 106.072
BMDL = 76.5569
A-ll
-------
Power Model with 0.95 Confidence Level
1.9
1.8
CD
V)
c
o 17
Q. I-'
CD
o:
ro 1.6
0
1.5
1.4
Power
BMD
BMD
50
100 150
dose
200
250
22:2805/122005
Hill Model. $Revision: 2.1 $ $Date: 2000/10/11 21:21:23 $
Input Data File: C:\BMDS\DATA\CYLINDRO.(d)
Gnuplot Plotting File: C:\BMDS\DATA\CYLINDRO.plt
Thu May 12 22:29:31 2005
HMDS MODEL RUN
The form of the response function is:
Y[dose] = intercept + v*doseAn/(kAn + doseAn)
Dependent variable = MEAN
Independent variable = dose
rho is set to 0
Power parameter restricted to be greater than 1
A constant variance model is fit
Total number of dose groups = 5
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
A-12
-------
Default Initial Parameter Values
alpha = 0.0183633
rho = 0 Specified
intercept = 1.48
v = 0.34
n = 0.959904
k = 63.3333
Asymptotic Correlation Matrix of Parameter Estimates
alpha rho intercept v n
alpha 10000
rho 01000
intercept 00100
v 0 0 0 1 0
n 0 0 0 0 1
k 0 0 0 0 0
Parameter Estimates
Variable Estimate Std. Err.
alpha 0.0172084 1
rho 0 1
intercept 1.4859 1
v 0.33478 1
n 2.46401 1
k 51.624 1
Table of Data and Estimated Values of Interest
Dose N Obs Mean Obs Std Dev Est Mean Est Std Dev
Res .
0 10 1.48 0.1 1.49 0.131
30 10 1.57 0.14 1.56 0.131
60 9 1.66 0.16 1.68 0.131
120 9 1.82 0.12 1.78 0.131
240 6 1.78 0.17 1.81 0.131
k
0
0
0
0
0
1
ChiA2
-0.0449
0.11
-0.183
0.279
-0.253
Model Descriptions for likelihoods calculated
Model Al: Yij = Mu(i) + e(ij)
Var{e(ij)} = SigmaA2
A-13
-------
Model A2: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma(i)A2
Model R: Yi = Mu + e(i)
Var{e (i) } = SigmaA2
Likelihoods of Interest
Model Log(likelihood) DF AIC
Al 68.148702 6 -124.297404
A2 69.554943 10 -119.109885
fitted 67.371885 5 -124.743770
R 52.631671 2 -101.263343
Test 1: Does response and/or variances differ among dose levels
(A2 vs. R)
Test 2: Are Variances Homogeneous (Al vs A2)
Test 3: Does the Model for the Mean Fit (Al vs. fitted)
Tests of Interest
Test -2*log(Likelihood Ratio) Test df p-value
Test 1 33.8465 8 <.0001
Test 2 2.81248 4 0.5897
Test 3 1.55363 1 0.2126
The p-value for Test 1 is less than .05. There appears to be a
difference between response and/or variances among the dose levels.
It seems appropriate to model the data
The p-value for Test 2 is greater than .05. A homogeneous variance
model appears to be appropriate here
The p-value for Test 3 is greater than .05. The model chosen appears
to adequately describe the data
Benchmark Dose Computation
Specified effect = 1
Risk Type = Estimated standard deviations from the control mean
Confidence level = 0.95
BMD = 43.1891
Warning: optimum may not have been found. Bad completion code in
Optimization routine.
BMDL computation failed.
A-14
-------
2
1.9
Hill Model
Hill
o>
V)
c
o
Q.
w
CD
o:
c
CD
1.7
1.6
1.5
1.4
BMP
22:2905/122005
50
100 150
dose
200
250
A-15
-------
Part II.
Humpage and Falconer 2003
male mice treated with purified cylindrospermopsin
rel kidney wt
drop high dose group
A-16
-------
Polynomial Model. Revision: 2.2 Date: 9/12/2002
Input Data File: C:\BMDS\DATA\CYLINDRO.(d)
Gnuplot Plotting File: C:\BMDS\DATA\CYLINDRO.plt
Thu May 12 22:31:50 2005
HMDS MODEL RUN
The form of the response function is:
Y[dose] = beta 0 + beta l*dose + beta 2*doseA2 + ...
Dependent variable = MEAN
Independent variable = dose
rho is set to 0
The polynomial coefficients are restricted to be positive
A constant variance model is fit
Total number of dose groups = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
alpha = 0.0172471
rho = 0 Specified
beta_0 = 1.484
beta 1 = 0.00282857
Parameter Estimates
Variable Estimate
alpha 0.0154483
beta_0 1.4838
beta 1 0.00283099
95.0% Wald Confidence Interval
Std. Err. Lower Conf. Limit Upper Conf. Limit
0.00354407 0.008502 0.0223945
0.0306522 1.42373 1.54388
0.000456935 0.00193541 0.00372656
Asymptotic Correlation Matrix of Parameter Estimates
alpha
beta_0
beta 1
alpha
1
-1.7e-010
1.4e-010
beta_0
-1.7e-010
1
-0.75
beta_l
1.4e-010
-0.75
1
A-17
-------
Table of Data and Estimated Values of Interest
Dose
Res .
0
30
60
120
N
10
10
9
9
Obs
1.
1.
1.
1.
Mean
48
57
66
82
Obs
0
0
0
Std Dev
0.1
.14
.16
.12
Est
1.
1.
1.
1.
Mean
48
57
65
82
Est
0.
0.
0.
0.
Std Dev
124
124
124
124
Chi
-0
0
-
A2
.0968
.0323
0.153
0.085
Model Descriptions for likelihoods calculated
Model Al: Yij
Var{e(ij)}
Model A2: Yij
Var{e(ij)}
Model R: Yi
Var{e(i)}
Mu (i) + e (ij )
SigmaA2
Mu (i) + e (ij )
Sigma(i)A2
Mu + e (i)
S i gma A 2
Likelihoods of Interest
Model
Al
A2
fitted
R
Log (likelihood)
60.255446
61.376237
60.234922
46.462327
DF
5
8
2
2
AIC
-110.510893
-106.752474
-116.469843
-88.924654
Test 1: Does response and/or variances differ among dose
levels
(A2 vs. R)
Test 2: Are Variances Homogeneous (Al vs A2)
Test 3: Does the Model for the Mean Fit (Al vs. fitted)
Test
Test 1
Test 2
Test 3
Tests of Interest
-2*log(Likelihood Ratio) Test df
29.8278
2.24158
0.0410496
6
3
2
p-value
<.0001
0.5238
0.9797
The p-value for Test 1 is less than .05. There appears
to be a
difference between response and/or variances among the
dose levels.
It seems appropriate to model the data
A-18
-------
The p-value for Test 2 is greater than .05. A
homogeneous variance
model appears to be appropriate here
The p-value for Test 3 is greater than .05. The model
chosen appears
to adequately describe the data
Benchmark Dose Computation
Specified effect = 1
Risk Type
Estimated standard deviations from the control mean
Confidence level =
BMD =
0.95
43.9038
BMDL =
33.0684
Linear Model with 0.95 Confidence Level
1.9
1.8
CD
V)
c
O
Q.
V)
CD
c
CD
CD
-Ig
' -°
1.5
1.4
Linear
BMDL iBM.D
22:31 05/122005
20 40 60
dose
80
100
120
A-19
-------
Polynomial Model. Revision: 2.2 Date: 9/12/2002
Input Data File: C:\BMDS\DATA\CYLINDRO.(d)
Gnuplot Plotting File: C:\BMDS\DATA\CYLINDRO.plt
Thu May 12 22:33:17 2005
HMDS MODEL RUN
The form of the response function is:
Y[dose] = beta 0 + beta l*dose + beta 2*doseA2 + ...
Dependent variable = MEAN
Independent variable = dose
rho is set to 0
The polynomial coefficients are restricted to be positive
A constant variance model is fit
Total number of dose groups = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
alpha =
rho =
beta_0 =
beta_l =
beta 2 =
0.0172471
0
1.47945
0.00314242
0
Specified
Parameter Estimates
Variable
alpha
beta_0
beta_l
beta 2
95.0% Wald Confidence Interval
Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
0.0154483 0.00354407 0.008502 0.0223945
1.4838 0.0306522 1.42373 1.54388
0.00283099 0.000456935
0.00193541
0
NA
0.00372656
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Asymptotic Correlation Matrix of Parameter Estimates
alpha
beta_0
beta 1
alpha
1
4.9e-011
9.5e-011
beta_0
4.9e-011
1
-0.75
beta_l
9.5e-011
-0.75
1
A-20
-------
The following parameter(s) have been estimated at a
boundary
point or have been specified. Correlations are not
computed:
beta 2
Dose
Res .
Table of Data and Estimated Values of Interest
N Obs Mean Obs Std Dev Est Mean
0
30
60
120
10
10
9
9
Est Std Dev
ChiA2
1.48
1.57
1.66
1.82
0
0
0
0.1
.14
.16
.12
1.
1.
1.
1.
48
57
65
82
0.
0.
0.
0.
124
124
124
124
-0
0
-
.
0
0
0968
0323
.153
.085
Model Descriptions for likelihoods calculated
Model Al: Yij
Var{e(ij)}
Model A2: Yij
Var{e(ij)}
Model R: Yi
Var{e(i)}
Mu (i) + e (ij )
SigmaA2
Mu (i) + e (ij )
Sigma(i)A2
Mu + e (i)
S i gma A 2
Likelihoods of Interest
Model Log (likelihood)
Test 1:
levels
Test 2:
Test 3:
Al
A2
fitted
R
Does response
(A2 vs. R)
Are Variances
Does the Model
60.255446
61.376237
60.234922
46.462327
and/or variances
DF
5
8
2
2
differ
AIC
110.510893
106.752474
116.469843
-88.924654
among dose
Homogeneous (Al vs A2 )
for the Mean Fit
(Al vs
. fitted)
A-21
-------
Tests of Interest
Test -2*log(Likelihood Ratio) Test df p-value
Test 1 29.8278 6 <.0001
Test 2 2.24158 3 0.5238
Test 3 0.0410496 1 0.8394
The p-value for Test 1 is less than .05. There appears
to be a
difference between response and/or variances among the
dose levels.
It seems appropriate to model the data
The p-value for Test 2 is greater than .05. A
homogeneous variance
model appears to be appropriate here
The p-value for Test 3 is greater than .05. The model
chosen appears
to adequately describe the data
Benchmark Dose Computation
Specified effect = 1
Risk Type = Estimated standard deviations from the control mean
Confidence level = 0.95
BMD = 43.9038
BMDL = 33.0684
A-22
-------
Polynomial Model with 0.95 Confidence Level
1.9
1.8
n Response
^j
eg 1-6
^
1.5
1.4
22:33 (
- ' ' _ ' .' ' -
polynomial
: ^~^-
\ " / 1
; T ..>--^ :
: ^T\^ 1 ;
^ 1 1 -L -
i ^ i ;
; 1 ;
'-- , , BMDL iBM.P. , , , , , , :
0 20 40 60 80 100 120
dose
D5/122005
Power Model. $Revision: 2.1 $ $Date: 2000/10/11 20:57:36 $
Input Data File: C:\BMDS\DATA\CYLINDRO.(d)
Gnuplot Plotting File: C:\BMDS\DATA\CYLINDRO.plt
Thu May 12 22:34:08 2005
HMDS MODEL RUN
The form of the response function is:
Y[dose] = control + slope * doseApower
Dependent variable = MEAN
Independent variable = dose
rho is set to 0
The power is restricted to be greater than or equal to 1
A constant variance model is fit
Total number of dose groups = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
A-23
-------
Default Initial Parameter Values
alpha = 0.0172471
rho = 0 Specified
control = 1.48
slope = 0.00345163
power = 0. 958769
Asymptotic Correlation Matrix of Parameter Estimates
alpha rho control slope
alpha 1 -0.99 0.25 -0.29
rho -0.99 1 -0.25 0.29
control 0.25 -0.25 1 -0.71
slope -0.29 0.29 -0.71 1
power 0.29 -0.28 0.66 -1
Parameter Estimates
Variable Estimate Std. Err.
alpha 0.0154483 0.0279175
rho 0 3.65432
control 1.4838 0.0410106
slope 0.00283099 0.00517499
power 1 0.372848
Table of Data and Estimated Values of Interest
Dose N Obs Mean Obs Std Dev Est Mean Est Std Dev
Res .
0 10 1.48 0.1 1.48 0.124
30 10 1.57 0.14 1.57 0.124
60 9 1.66 0.16 1.65 0.124
120 9 1.82 0.12 1.82 0.124
power
0.29
-0.28
0.66
-1
1
ChiA2
-0.0306
0.0102
0.051
-0.0283
Model Descriptions for likelihoods calculated
Model Al: Yij
Var{e(ij)}
Model A2: Yij
Var{e(ij)}
Mu (i) + e (ij )
SigmaA2
Mu (i) + e (ij )
Sigma(i)A2
A-24
-------
Model R: Yi = Mu + e(i)
Var{e(i)} = SigmaA2
Likelihoods of Interest
Model
Al
A2
fitted
R
Log(likelihood) DF
60.255446 5
61.376237 8
60.234922 4
46.462327 2
AIC
-110.510893
-106.752474
-112.469843
-88.924654
Test 1:
Test 2:
Test 3:
Does response and/or variances differ among dose levels
(A2 vs. R)
Are Variances Homogeneous (Al vs A2)
Does the Model for the Mean Fit (Al vs. fitted)
Tests of Interest
Test -2*log(Likelihood Ratio) df
Test 1 29.8278 6
Test 2 2.24158 3
Test 3 0.0410496 1
p-value
<.00001
0.5238
0.8394
The p-value for Test 1 is less than .05. There appears to be a
difference between response and/or variances among the dose levels.
It seems appropriate to model the data
The p-value for Test 2 is greater than .05. A homogeneous variance
model appears to be appropriate here
The p-value for Test 3 is greater than .05. The model chosen appears
to adequately describe the data
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
Estimated standard deviations from the control mean
0.95
43.9038
BMDL =
33.0684
A-25
-------
Power Model with 0.95 Confidence Level
1.9
1.8
CD
V)
c
o 17
Q. I-'
CD
o:
C! . —
«5 1.6
CD
S
1.5
1.4
22:34 (
: _ :
rower
[ ~r j
i ^^1
; T ^^-^ 1 ;
; T ^-^ \
'-- ^ /-^r "" ^ ^
; ^^ 1 ;
\ 1 ^
; , , BMDL IBMD. , , , , , , ;
0 20 40 60 80 100 120
dose
D5/122005
Hill Model. $Revision: 2.1 $ $Date: 2000/10/11 21:21:23 $
Input Data File: C:\BMDS\DATA\CYLINDRO.(d)
Gnuplot Plotting File: C:\BMDS\DATA\CYLINDRO.plt
Thu May 12 22:34:43 2005
HMDS MODEL RUN
The form of the response function is:
Y[dose] = intercept + v*doseAn/(kAn + doseAn)
Dependent variable = MEAN
Independent variable = dose
rho is set to 0
Power parameter restricted to be greater than 1
A constant variance model is fit
Total number of dose groups = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
A-26
-------
Default Initial Parameter Values
alpha =
rho =
intercept =
v =
n =
k =
0.0163951
0
1.48
0.34
0.68364
63.3333
Specified
Asymptotic Correlation Matrix of Parameter Estimates
alpha
rho
intercept
V
n
k
alpha
1
0
0
0
0
0
rho
0
1
0
0
0
0
intercept
0
0
1
0
0
0
V
0
0
0
1
0
0
n
0
0
0
0
1
0
k
0
0
0
0
0
1
Parameter Estimates
Variable
alpha
rho
intercept
V
n
k
Estimate
0.0154316
0
1.48
1.62004
1.08746
406.089
Std. Err
1
1
1
1
1
1
Table of Data and Estimated Values of Interest
R
-
Dose
es .
0
30
60
120
N
10
10
9
9
Obs
1.
1.
1.
1.
Mean
48
57
66
82
Obs
0
0
0
Std Dev
0.1
.14
.16
.12
Est
1.
1.
1.
1.
Mean
48
57
66
82
Est
0.
0.
0.
0.
Std
124
124
124
124
Dev ChiA2
-2.33e-007
-4.42e-007
8.2e-007
-1.25e-008
Model Descriptions for likelihoods calculated
Model Al: Yij = Mu(i) + e(ij)
Var{e(ij)} = SigmaA2
A-27
-------
Model A2: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma(i)A2
Model R: Yi = Mu + e(i)
Var{e (i) } = SigmaA2
Warning: Likelihood for fitted model larger than the Likelihood for model
Al.
Likelihoods of Interest
Model Log(likelihood) DF AIC
Al 60.255446 5 -110.510893
A2 61.376237 8 -106.752474
fitted 60.255447 5 -110.510893
R 46.462327 2 -88.924654
Test 1: Does response and/or variances differ among dose levels
(A2 vs. R)
Test 2: Are Variances Homogeneous (Al vs A2)
Test 3: Does the Model for the Mean Fit (Al vs. fitted)
Tests of Interest
Test -2*log(Likelihood Ratio) Test df p-value
Test 1 29.8278 6 <.0001
Test 2 2.24158 3 0.5238
Test 3 -4.58279e-007 0 NA
The p-value for Test 1 is less than .05. There appears to be a
difference between response and/or variances among the dose levels.
It seems appropriate to model the data
The p-value for Test 2 is greater than .05. A homogeneous variance
model appears to be appropriate here
NA - Degrees of freedom for Test 3 are less than or equal to 0. The Chi-
Square
test for fit is not valid
Benchmark Dose Computation
Specified effect = 1
Risk Type = Estimated standard deviations from the control mean
Confidence level = 0.95
BMD = 41.1966
BMDL = 21.722
A-28
-------
Hill Model with 0.95 Confidence Level
Hill
CD
V)
c
CD
o:
1.9
1.8
* -,
1.7
1
' -
1.5
1.4
BMD
BMD
20
22:3405/122005
40
60
dose
80
100
120
A-29
-------
Part III.
Humpage and Falconer 2003
male mice treated with purified cylindrospermopsin
urinary protein levels
A-30
-------
Polynomial Model. Revision: 2.2 Date: 9/12/2002
Input Data File: C:\BMDS\DATA\CYLINDRO.(d)
Gnuplot Plotting File: C:\BMDS\DATA\CYLINDRO.plt
Thu May 12 23:15:09 2005
HMDS MODEL RUN
The form of the response function is:
Y[dose] = beta 0 + beta l*dose + beta 2*doseA2 + ...
Dependent variable = MEAN
Independent variable = dose
rho is set to 0
Signs of the polynomial coefficients are not restricted
A constant variance model is fit
Total number of dose groups = 5
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
alpha = 0.0467949
rho = 0 Specified
beta_0 = 4.00125
beta 1 = -0.0114583
Parameter Estimates
Variable Estimate
alpha 0.109717
beta 0 4.03293
beta 1
Std. Err.
0.0233918
0.070386
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
0.0638701 0.155564
3.89498 4.17088
-0.0120086 0.000649584
-0.0132818
-0.0107354
Asymptotic Correlation Matrix of Parameter Estimates
alpha
beta_0
beta 1
alpha
1
2.5e-009
9.1e-010
beta_0
2.5e-009
1
-0.7
beta_l
9.1e-010
-0.7
1
A-31
-------
Table of Data and Estimated Values of Interest
Dose
Res .
0
30
60
120
240
N
10
10
9
9
6
Obs
4.
3
3.
2.
1
Mean
25
.7
25
15
.5
Obs
0
0
0
Std Dev
0.3
.25
.05
0.2
.15
Est
4.
3.
3.
2.
1.
Mean
03
67
31
59
15
Est
0.
0.
0.
0.
0.
Std Dev
331
331
331
331
331
ChiA2
2
0.
-0.
2
.07
261
565
-4
.58
Model Descriptions for likelihoods calculated
Model Al: Yij = Mu(i) + e(ij)
Var{e(ij)} = SigmaA2
Model A2: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma(i)A2
Model R: Yi = Mu + e(i)
Var{e(i)} = SigmaA2
Likelihoods of Interest
Model Log(likelihood) DF AIC
Al 48.017412 6 -84.034824
A2 59.392540 10 -98.785081
fitted 26.616696 2 -49.233392
R -21.651321 2 47.302642
Test 1: Does response and/or variances differ among dose
levels
(A2 vs. R)
Test 2: Are Variances Homogeneous (Al vs A2)
Test 3: Does the Model for the Mean Fit (Al vs. fitted)
Test
Test 1
Test 2
Test 3
Tests of Interest
-2*log(Likelihood Ratio) Test df
162.1
22.7503
42.8014
p-value
<.0001
0.000142
<.0001
The p-value for Test 1 is less than .05. There appears
to be a
difference between response and/or variances among the
dose levels.
It seems appropriate to model the data
A-32
-------
The p-value for Test 2 is less than .05.
running a
non-homogeneous variance model
The p-value for Test 3 is less than .05.
to try a
different model
Consider
You may want
Benchmark Dose Computation
Specified effect = 1
Risk Type
Estimated standard deviations from the control mean
Confidence level =
BMD =
0.95
27.5832
BMDL =
22.9758
Linear Model with 0.95 Confidence Level
4.5
CD
V)
c
o
Q.
V)
CD
c
CD
CD
3.5
2.5
1.5
Linear
BMDL
BMD
0 50
23:1505/122005
100 150 200
dose
250
A-33
-------
Polynomial Model. Revision: 2.2 Date: 9/12/2002
Input Data File: C:\BMDS\DATA\CYLINDRO.(d)
Gnuplot Plotting File: C:\BMDS\DATA\CYLINDRO.plt
Thu May 12 23:16:16 2005
HMDS MODEL RUN
The form of the response function is:
Y[dose] = beta 0 + beta l*dose + beta 2*doseA2 + ...
Dependent variable = MEAN
Independent variable = dose
Signs of the polynomial coefficients are not restricted
The variance is to be modeled as Var(i) = alpha*mean(i)Arho
Total number of dose groups = 5
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
alpha = 0.0467949
rho = 0
beta_0 = 4.00125
beta 1 = -0.0114583
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
alpha 0.0612302 0.0136561 0.0344646 0.0879957
rho -0.214753 0.0359694 -0.285252 -0.144255
beta_0 4.26301 0.0535233 4.15811 4.36791
beta 1 -0.0177625 0.00084855 -0.0194257 -0.0160994
Asymptotic Correlation Matrix of Parameter Estimates
alpha rho beta_0 beta_l
alpha 1 0.12 O.Il -0.18
rho 0.12 1 -0.18 0.29
beta_0 0.11 -0.18 1 -0.75
beta 1 -0.18 0.29 -0.75 1
A-34
-------
Dose
Res .
Table of Data and Estimated Values of Interest
N Obs Mean Obs Std Dev Est Mean Est Std Dev
ChiA2
0
30
60
120
240
10
10
9
9
6
4.25
3.7
3.25
2.15
1.5
0.3
0.25
0.05
0.2
0.15
4.26
3.73
3.2
2.13
4.95e-008
0.212
0.215
0.218
0.228
1.51
-0.194
-0.444
0.724
0.243
2.44
Model Descriptions for likelihoods calculated
Model Al: Yij
Var{e(ij)}
Model A2: Yij
Var{e(ij)}
Model A3: Yij
Var{e(ij)}
Model R: Yi
Var{e(i)}
Mu (i) + e (ij )
S i gma A 2
Mu (i) + e (ij )
Sigma(i)A2
Mu (i) + e (ij )
alpha*(Mu(i))Arho
Mu + e(i)
SigmaA2
Likelihoods of Interest
Model Log(likelihood) DF AIC
Al 48.017412 6 -84.034824
A2 59.392540 10 -98.785081
A3 49.799962 7 -85.599924
fitted 33.618701 4 -59.237401
R -21.651321 2 47.302642
Test 1:
levels?
Test 2:
Test 3:
Test 4:
Explanation of Tests
Does response and/or variances differ among Dose
(A2 vs. R)
Are Variances Homogeneous? (Al vs A2)
Are variances adequately modeled? (A2 vs. A3)
Does the Model for the Mean Fit? (A3 vs. fitted)
Tests of Interest
Test
Test 1
Test 2
Test 3
Test 4
-2*log(Likelihood Ratio) Test df p-value
162.1
22.7503
19.1852
32.3625
8 <.0001
4 0.000142
3 0.0002503
3 <.0001
A-35
-------
The p-value for Test 1 is less than .05. There appears
to be a
difference between response and/or variances among the
dose levels
It seems appropriate to model the data
The p-value for Test 2 is less than .05. A
non-homogeneous variance
model appears to be appropriate
The p-value for Test 3 is less than .05. You may want
to consider a
different variance model
The p-value for Test 4 is less than .05. You may want
to try a different
model
Benchmark Dose Computation
Specified effect = 1
Risk Type = Estimated standard deviations from the control mean
Confidence level = 0.95
BMD = 11. 9223
BMDL computation failed.
A-36
-------
Linear Model with 0.95 Confidence Level
4.5
CD
V)
c
o
Q.
-------
Default Initial Parameter Values
alpha
rho
beta_0
beta_l
beta_2
beta 3
0.0467949
0
4.23439
-0.0148331
-5.51541e-005
2.89474e-007
Specified
Parameter Estimates
Variable Estimate
alpha
beta_0
beta_l
beta_2
beta 3
0.0424211
4.23509
-0.0149779
-5.33342e-005
2.84322e-007
95.0% Wald Confidence Interval
Std. Err. Lower Conf. Limit Upper Conf. Limit
0.0090442
0.0633648
0.00335279
4.06738e-005
1.17673e-007
0.0246948
4.1109
-0.0215492
-0.000133053
5.36875e-008
0.0601474
4.35929
-0.00840654
2.63851e-005
5.14956e-007
Asymptotic Correlation Matrix of Parameter Estimates
alpha
beta 0
beta 1
beta 2
beta 3
alpha
1
-1.8e-008
-4.3e-009
1.3e-009
1.5e-009
beta 0
-1.8e-008
1
-0.71
0.55
-0.48
beta 1
-4.3e-009
-0.71
1
-0.96
0.91
beta 2
1.3e-009
0.55
-0.96
1
-0.99
beta 3
1.5e-009
-0.48
0.91
-0.99
1
Dose
Res .
Table of Data and Estimated Values of Interest
N Obs Mean Obs Std Dev Est Mean Est Std Dev ChiA2
0
30
60
120
240
10
10
9
9
6
4.25
3.7
3.25
2.15
1.5
0.3
0.25
0.05
0.2
0.15
4.24
3.75
3.21
2.16
1.5
0.206
0.206
0.206
0.206
0.206
0.229
-0.698
0.643
-0.161
0.0141
Model Descriptions for likelihoods calculated
Model Al: Yij = Mu(i) + e(ij)
Var{e(ij)} = SigmaA2
Model A2: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma(i)A2
Model R: Yi = Mu + e(i)
Var{e (i) } = SigmaA2
A-38
-------
Likelihoods of Interest
Model Log(likelihood) DF AIC
Al 48.017412 6 -84.034824
A2 59.392540 10 -98.785081
fitted 47.522415 4 -87.044831
R -21.651321 2 47.302642
Test 1: Does response and/or variances differ among dose
levels
(A2 vs. R)
Test 2: Are Variances Homogeneous (Al vs A2)
Test 3: Does the Model for the Mean Fit (Al vs. fitted)
Tests of Interest
Test -2*log(Likelihood Ratio) Test df p-value
Test 1 162.088 8 <.0001
Test 2 22.7503 4 0.000142
Test 3 0.989994 1 0.3197
The p-value for Test 1 is less than .05. There appears
to be a
difference between response and/or variances among the
dose levels.
It seems appropriate to model the data
The p-value for Test 2 is less than .05. Consider
running a
non-homogeneous variance model
The p-value for Test 3 is greater than .05. The model
chosen appears
to adequately describe the data
Benchmark Dose Computation
Specified effect = 1
Risk Type = Estimated standard deviations from the control mean
Confidence level = 0.95
BMD = 13.1764
BMDL = 9.7619
A-39
-------
Polynomial Model with 0.95 Confidence Level
4.5
o
Q.
W
CD
o:
c
CD
CD
1.5
Polynomial
BMD
BMD
50
100 150
dose
200
250
23:1705/122005
Polynomial Model. Revision: 2.2 Date: 9/12/2002
Input Data File: C:\BMDS\DATA\CYLINDRO.(d)
Gnuplot Plotting File: C:\BMDS\DATA\CYLINDRO.plt
Thu May 12 23:18:39 2005
HMDS MODEL RUN
The form of the response function is:
Y[dose] = beta 0 + beta l*dose + beta 2*doseA2 + ...
Dependent variable = MEAN
Independent variable = dose
rho is set to 0
The polynomial coefficients are restricted to be negative
A constant variance model is fit
Total number of dose groups = 5
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
A-40
-------
Default Initial Parameter Values
alpha
rho
beta_0
beta_l
beta_2
beta 3
0.0467949
0
4.23439
-0.0148331
-5.51541e-005
0
Specified
Parameter Estimates
Variable
alpha
beta 0
beta 1
beta 2
beta 3
Estimate
0.109717
4.03293
-0.0120086
0
0
Std. Err.
0.0233918
0.070386
0.000649584
NA
NA
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
0.0638701 0.155564
3.89498 4.17088
-0.0132818 -0.0107354
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Asymptotic Correlation Matrix of Parameter Estimates
alpha beta 0 beta 1
alpha 1 -1.2e-(XT7 8.2e-OC>8
beta_0 -1.2e-007 1 -0.7
beta_l 8.2e-008 -0.7 1
The following parameter(s) have been estimated at a
boundary
point or have been specified. Correlations are not
computed:
beta 2 beta 3
Dose
Res .
Table of Data and Estimated Values of Interest
N Obs Mean Obs Std Dev Est Mean Est Std Dev ChiA2
0
30
60
120
240
10
10
9
9
6
4.25
3.7
3.25
2.15
1.5
0.3
0.25
0.05
0.2
0.15
4.03
3.67
3.31
2.59
1.15
Model Descriptions for likelihoods calculated
Model Al: Yij = Mu(i) + e(ij)
Var{e(ij)} = SigmaA2
0.331
0.331
0.331
0.331
0.331
2.07
0.261
-0.565
-4
2.58
A-41
-------
Model A2: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma(i)A2
Model R: Yi = Mu + e(i)
Var{e(i)} = SigmaA2
Likelihoods of Interest
Model Log(likelihood) DF AIC
Al 48.017412 6 -84.034824
A2 59.392540 10 -98.785081
fitted 26.616696 2 -49.233392
R -21.651321 2 47.302642
Test 1: Does response and/or variances differ among dose
levels
(A2 vs. R)
Test 2: Are Variances Homogeneous (Al vs A2)
Test 3: Does the Model for the Mean Fit (Al vs. fitted)
Test
Test 1
Test 2
Test 3
Tests of Interest
-2*log(Likelihood Ratio) Test df
162.1
22.7503
42.8014
p-value
<.0001
0.000142
<.0001
The p-value for Test 1 is less than .05. There appears
to be a
difference between response and/or variances among the
dose levels.
It seems appropriate to model the data
The p-value for Test 2 is less than .05. Consider
running a
non-homogeneous variance model
The p-value for Test 3 is less than .05. You may want
to try a
different model
Benchmark Dose Computation
Specified effect = 1
Risk Type
Estimated standard deviations from the control mean
Confidence level =
BMD =
0.95
27.5832
BMDL =
22.9758
A-42
-------
Polynomial Model with 0.95 Confidence Level
4.5
CD
V)
c
o
Q.
-------
Default Initial Parameter Values
alpha
rho
beta_0
beta_l
beta 2
0.0467949
0
4.31231
-0.0224959
4.45961e-005
Specified
Parameter Estimates
Variable Estimate
alpha 0.0480497
beta_0 4.30818
beta_l -0.0223839
beta 2 4.40376e-005
Std. Err.
0.0102442
0.0592565
0.00144605
5.86023e-006
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
0.0279713 0.068128
4.19204 4.42432
-0.0252181 -0.0195497
3.25518e-005 5.55235e-005
Asymptotic Correlation Matrix of Parameter Estimates
alpha
beta 0
beta 1
beta 2
alpha
1
3.3e-009
-6.2e-010
-7.7e-009
beta 0
3.3e-009
1
-0.75
0.62
beta 1
-6.2e-010
-0.75
1
-0.95
beta 2
-7.7e-009
0.62
-0.95
1
Dose
Res .
Table of Data and Estimated Values of Interest
N Obs Mean Obs Std Dev Est Mean Est Std Dev
ChiA2
0
30
60
120
240
10
10
9
9
6
4.25
3.7
3.25
2.15
1.5
0.3
0.25
0.05
0.2
0.15
4.31
3.68
3.12
2.26
1.47
0.219
0.219
0.219
0.219
0.219
-0.839
0.342
1.73
-1.45
0.306
Model Descriptions for likelihoods calculated
Model Al: Yij
Var{e(ij)}
Model A2: Yij
Var{e(ij) }
Model R: Yi
Var{e(i) }
Mu (i) + e (ij )
S i gma A 2
Mu (i) + e (ij )
Sigma(i)A2
Mu + e (i)
SigmaA2
A-44
-------
Likelihoods of Interest
Model Log(likelihood) DF AIC
Al 48.017412 6 -84.034824
A2 59.392540 10 -98.785081
fitted 44.781449 3 -83.562899
R -21.651321 2 47.302642
Test 1: Does response and/or variances differ among dose
levels
(A2 vs. R)
Test 2: Are Variances Homogeneous (Al vs A2)
Test 3: Does the Model for the Mean Fit (Al vs. fitted)
Test
Test 1
Test 2
Test 3
Tests of Interest
-2*log(Likelihood Ratio) Test df
162.1
22.7503
6.47193
p-value
<.0001
0.000142
0.03932
The p-value for Test 1 is less than .05. There appears
to be a
difference between response and/or variances among the
dose levels.
It seems appropriate to model the data
The p-value for Test 2 is less than .05. Consider
running a
non-homogeneous variance model
The p-value for Test 3 is less than .05. You may want
to try a
different model
Benchmark Dose Computation
Specified effect = 1
Risk Type
Estimated standard deviations from the control mean
Confidence level =
BMD =
0.95
9.98916
BMDL =
!.24167
A-45
-------
Polynomial Model with 0.95 Confidence Level
4.5
CD
V)
c
o
Q.
-------
Default Initial Parameter Values
alpha =
rho =
beta_0 =
beta_l =
beta 2 =
0.0467949
0
4.31231
-0.0224959
0
Specified
Parameter Estimates
Variable
alpha
beta_0
beta_l
beta 2
Estimate
0.109717
4.03293
-0.0120086
0
Std. Err.
0.0233918
0.070386
0.000649584
NA
95.0% Wald Confidence Interval
Lower Conf. Limit Upper Conf. Limit
0.0638701 0.155564
3.89498 4.17088
-0.0132818 -0.0107354
NA - Indicates that this parameter has hit a bound
implied by some inequality constraint and thus
has no standard error.
Asymptotic Correlation Matrix of Parameter Estimates
alpha
beta 0
beta 1
alpha
1
1.4e-008
2.1e-009
beta 0
1.4e-008
1
-0.7
beta 1
2.1e-009
-0.7
1
The following parameter(s) have been estimated at a
boundary
point or have been specified. Correlations are not
computed:
beta 2
Dose
Res .
Table of Data and Estimated Values of Interest
N Obs Mean Obs Std Dev Est Mean Est Std Dev ChiA2
0
30
60
120
240
10
10
9
9
6
4.25
3.7
3.25
2.15
1.5
0.3
0.25
0.05
0.2
0.15
4.03
3.67
3.31
2.59
1.15
0.331
0.331
0.331
0.331
0.331
2.07
0.261
-0.565
-4
2.58
Model Descriptions for likelihoods calculated
Model Al: Yij = Mu(i) + e(ij)
Var{e(ij)} = SigmaA2
A-47
-------
Model A2: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma(i)A2
Model R: Yi = Mu + e(i)
Var{e(i)} = SigmaA2
Likelihoods of Interest
Model Log(likelihood) DF AIC
Al 48.017412 6 -84.034824
A2 59.392540 10 -98.785081
fitted 26.616696 2 -49.233392
R -21.651321 2 47.302642
Test 1: Does response and/or variances differ among dose
levels
(A2 vs. R)
Test 2: Are Variances Homogeneous (Al vs A2)
Test 3: Does the Model for the Mean Fit (Al vs. fitted)
Test
Tests of Interest
-2*log(Likelihood Ratio) Test df
p-value
Test 1 162.088 8 <.0001
Test 2 22.7503 4 0.000142
Test 3 42.8014 2 <.0001
The p-value for Test 1 is less than .05. There appears
to be a
difference between response and/or variances among the
dose levels.
It seems appropriate to model the data
The p-value for Test 2 is less than .05.
running a
non-homogeneous variance model
Consider
The p-value for Test 3 is less than .05. You may want
to try a
different model
Benchmark Dose Computation
Specified effect = 1
Risk Type
Estimated standard deviations from the control mean
Confidence level =
BMD =
0.95
27.5832
BMDL =
22.9758
A-48
-------
Polynomial Model with 0.95 Confidence Level
4.5
CD
V)
c
o
Q.
-------
Default Initial Parameter Values
alpha = 0.0467949
rho = 0 Specified
control = 4.25
slope = -340.972
power = -0. 879496
Asymptotic Correlation Matrix of Parameter Estimates
alpha rho control slope
alpha NA NA NA NA
rho NA NA NA NA
control NA NA NA NA
slope NA NA NA NA
power NA NA NA NA
NA - This parameter's variance has been estimated at zero.
Parameter Estimates
Variable Estimate Std. Err.
alpha 0.109717 0.114352
rho 0 0.935809
control 4.03293 0.129954
slope -0.0120086 0.0133934
power 1 0.202697
Table of Data and Estimated Values of Interest
Dose N Obs Mean Obs Std Dev Est Mean Est Std Dev
Res .
0 10 4.25 0.3 4.03 0.331
30 10 3.7 0.25 3.67 0.331
60 9 3.25 0.05 3.31 0.331
120 9 2.15 0.2 2.59 0.331
240 6 1.5 0.15 1.15 0.331
power
NA
NA
NA
NA
NA
ChiA2
0.655
0.0825
-0.188
-1.33
1.05
Model Descriptions for likelihoods calculated
Model Al: Yij = Mu(i) + e(ij)
Var{e(ij)} = SigmaA2
A-50
-------
Model A2: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma(i)A2
Model R: Yi = Mu + e(i)
Var{e (i) } = SigmaA2
Likelihoods of Interest
Model Log(likelihood) DF AIC
Al 48.017412 6 -84.034824
A2 59.392540 10 -98.785081
fitted 26.616696 4 -45.233392
R -21.651321 2 47.302642
Test 1: Does response and/or variances differ among dose levels
(A2 vs. R)
Test 2: Are Variances Homogeneous (Al vs A2)
Test 3: Does the Model for the Mean Fit (Al vs. fitted)
Tests of Interest
Test -2*log(Likelihood Ratio) df p-value
Test 1 162.088 8 <.00001
Test 2 22.7503 4 0.000142
Test 3 42.8014 2 <.00001
The p-value for Test 1 is less than .05. There appears to be a
difference between response and/or variances among the dose levels.
It seems appropriate to model the data
The p-value for Test 2 is less than .05. Consider running a
non-homogeneous variance model
The p-value for Test 3 is less than .05. You may want to try a
different model
Benchmark Dose Computation
Specified effect = 1
Risk Type = Estimated standard deviations from the control mean
Confidence level = 0.95
BMD = 27.5832
BMDL = 22.9758
A-51
-------
Power Model with 0.95 Confidence Level
4.5
3.5
CD
V)
c
o
Q. o
03 -J
CD
o:
ro 2.5
CD
1.5
Power
BMDU BMD
50
100 150
dose
200
250
23:21 05/122005
Power Model. $Revision: 2.1 $ $Date: 2000/10/11 20:57:36 $
Input Data File: C:\BMDS\DATA\CYLINDRO.(d)
Gnuplot Plotting File: C:\BMDS\DATA\CYLINDRO.plt
Thu May 12 23:21:44 2005
HMDS MODEL RUN
The form of the response function is:
Y[dose] = control + slope * doseApower
Dependent variable = MEAN
Independent variable = dose
rho is set to 0
The power is not restricted
A constant variance model is fit
Total number of dose groups = 5
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
A-52
-------
Default Initial Parameter Values
alpha = 0.0467949
rho = 0 Specified
control = 4.25
slope = -0.0336166
power = 0. 803617
Asymptotic Correlation Matrix of Parameter Estimates
alpha rho control slope
alpha 1 -0.97 -0.22 0.57
rho -0.97 1 0.22 -0.58
control -0.22 0.22 1 -0.67
slope 0.57 -0.58 -0.67 1
power 0.6 -0.62 -0.6 0.99
Parameter Estimates
Variable Estimate Std. Err.
alpha 0.0696661 0.0661961
rho 0 0.846718
control 4.30267 0.0817494
slope -0.0726613 0.0255026
power 0.676841 0.0658079
Table of Data and Estimated Values of Interest
Dose N Obs Mean Obs Std Dev Est Mean Est Std Dev
Res .
0 10 4.25 0.3 4.3 0.264
30 10 3.7 0.25 3.58 0.264
60 9 3.25 0.05 3.14 0.264
120 9 2.15 0.2 2.45 0.264
240 6 1.5 0.15 1.34 0.264
power
0.6
-0.62
-0.6
0.99
1
ChiA2
-0.2
0.468
0.41
-1.12
0.623
Model Descriptions for likelihoods calculated
Model Al: Yij = Mu(i) + e(ij)
Var{e(ij)} = SigmaA2
Model A2: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma(i)A2
A-53
-------
Model R: Yi = Mu + e(i)
Var{e(i)} = SigmaA2
Likelihoods of Interest
Model Log(likelihood) DF AIC
Al 48.017412 6 -84.034824
A2 59.392540 10 -98.785081
fitted 36.608912 4 -65.217824
R -21.651321 2 47.302642
Test 1: Does response and/or variances differ among dose levels
(A2 vs. R)
Test 2: Are Variances Homogeneous (Al vs A2)
Test 3: Does the Model for the Mean Fit (Al vs. fitted)
Tests of Interest
Test -2*log(Likelihood Ratio) df p-value
Test 1 162.088 8 <.00001
Test 2 22.7503 4 0.000142
Test 3 22.817 2 l.lle-005
The p-value for Test 1 is less than .05. There appears to be a
difference between response and/or variances among the dose levels.
It seems appropriate to model the data
The p-value for Test 2 is less than .05.
non-homogeneous variance model
The p-value for Test 3 is less than .05.
different model
Consider running a
You may want to try a
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
Estimated standard deviations from the control mean
0.95
6.72481
BMDL =
3.97902
A-54
-------
Power Model with 0.95 Confidence Level
4.5
4
o> 3-5
V)
c
o
Q. o
03 -J
CD
o:
ro 2.5
CD
1.5
Power
BMD
L BMP
50
100 150
dose
200
250
23:21 05/122005
Hill Model. $Revision: 2.1 $ $Date: 2000/10/11 21:21:23 $
Input Data File: C:\BMDS\DATA\CYLINDRO.(d)
Gnuplot Plotting File: C:\BMDS\DATA\CYLINDRO.plt
Thu May 12 23:22:53 2005
HMDS MODEL RUN
The form of the response function is:
Y[dose] = intercept + v*doseAn/(kAn + doseAn)
Dependent variable = MEAN
Independent variable = dose
rho is set to 0
Power parameter restricted to be greater than 1
A constant variance model is fit
Total number of dose groups = 5
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
A-55
-------
Default Initial Parameter Values
alpha =
rho =
intercept =
v =
n =
k =
0.0263306
0
4.25
-2.75
2.08208
80.4545
Specified
Asymptotic Correlation Matrix of Parameter Estimates
alpha rho intercept
alpha
rho
intercept
V
n
k
1
0
0
0
0
0
0
1
0
0
0
0
0
0
1
0
0
0
V
0
0
0
1
0
0
n
0
0
0
0
1
0
k
0
0
0
0
0
1
Variable
alpha
rho
intercept
v
n
k
Parameter Estimates
Estimate
0.0466116
0
4.22518
-3.35427
1.63976
94.2147
Std.
Err.
1
1
1
1
1
1
Dose
Res .
Table of Data and Estimated Values of Interest
N Obs Mean Obs Std Dev Est Mean
0
30
60
120
240
10
10
9
9
6
Est Std Dev
4.25
3.7
3.25
2.15
1.5
ChiA2
0
0.
0.
0
0.
.3
25
05
.2
15
4
3
3
2
1
.23
.78
.14
.22
.47
0
0
0
0
0
.216
.216
.216
.216
.216
0
-0
0
-0
0
.115
.369
.502
.323
.156
A-56
-------
Model Descriptions for likelihoods calculated
Model Al: Yij = Mu(i) + e(ij)
Var{e(ij)} = SigmaA2
Model A2: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma(i)A2
Model R: Yi = Mu + e(i)
Var{e(i)} = SigmaA2
Likelihoods of Interest
Model Log(likelihood) DF AIC
Al 48.017412 6 -84.034824
A2 59.392540 10 -98.785081
fitted 45.449933 5 -80.899866
R -21.651321 2 47.302642
Test 1: Does response and/or variances differ among dose levels
(A2 vs. R)
Test 2: Are Variances Homogeneous (Al vs A2)
Test 3: Does the Model for the Mean Fit (Al vs. fitted)
Tests of Interest
Test -2*log(Likelihood Ratio) Test df p-value
Test 1 162.088 8 <.0001
Test 2 22.7503 4 0.000142
Test 3 5.13496 1 0.02345
The p-value for Test 1 is less than .05. There appears to be a
difference between response and/or variances among the dose levels.
It seems appropriate to model the data
The p-value for Test 2 is less than .05. Consider running a
non-homogeneous variance model
The p-value for Test 3 is less than .05. You may want to try a
different model
Benchmark Dose Computation
Specified effect = 1
Risk Type = Estimated standard deviations from the control mean
Confidence level = 0.95
BMD = 18.4161
BMDL = 12.8257
A-57
-------
Hill Model with 0.95 Confidence Level
4.5
4
o
Q.
W
CD
o:
c
CD
1.5
Hill
BMDL
BMD
50
23:2205/122005
100 150
dose
200
250
A-58
-------
Part IV.
Humpage and Falconer 2003
male mice treated with purified cylindrospermopsin
urinary protein levels
drop high dose
A-59
-------
Polynomial Model. Revision: 2.2 Date: 9/12/2002
Input Data File: C:\BMDS\DATA\CYLINDRO.(d)
Gnuplot Plotting File: C:\BMDS\DATA\CYLINDRO.plt
Thu May 12 23:24:32 2005
HMDS MODEL RUN
The form of the response function is:
Y[dose] = beta 0 + beta l*dose + beta 2*doseA2 + ...
Dependent variable = MEAN
Independent variable = dose
rho is set to 0
Signs of the polynomial coefficients are not restricted
A constant variance model is fit
Total number of dose groups = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
alpha = 0.0503676
rho = 0 Specified
beta_0 = 4.25
beta 1 = -0.017381
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
alpha 0.0457633 0.0104988 0.025186 0.0663406
beta_0 4.24885 0.052757 4.14545 4.35225
beta 1 -0.017373 0.000786454 -0.0189144 -0.0158316
Asymptotic Correlation Matrix of Parameter Estimates
alpha beta_0 beta_l
alpha 1 2.2e-OC>9 le-OC>9
beta_0 2.2e-009 1 -0.75
beta 1 le-009 -0.75 1
A-60
-------
Table of Data and Estimated Values of Interest
Dose
Res .
0
30
60
120
N
10
10
9
9
Obs
4.
3
3.
2.
Mean
25
.7
25
15
Obs Std Dev
0.3
0.25
0.05
0.2
Est
4.
3.
3.
2.
Mean
25
73
21
16
Est
0
0
0
0
Std Dev
.214
.214
.214
.214
ChiA2
0.
-0.
0.
-0.
017
409
611
198
Model Descriptions for likelihoods calculated
Model Al: Yij
Var{e(ij)}
Model A2: Yij
Var{e(ij)}
Model R: Yi
Var{e(i)}
Mu (i) + e (ij )
SigmaA2
Mu (i) + e (ij )
Sigma(i)A2
Mu + e (i)
S i gma A 2
Likelihoods of Interest
Model
Al
A2
fitted
R
Log (likelihood)
39.893006
50.462856
39.601194
-10.831438
DF
5
8
2
2
AIC
-69.786011
-84.925712
-75.202387
25.662876
Test 1: Does response and/or variances differ among dose
levels
(A2 vs. R)
Test 2: Are Variances Homogeneous (Al vs A2)
Test 3: Does the Model for the Mean Fit (Al vs. fitted)
Test
Test 1
Test 2
Test 3
Tests of Interest
-2*log(Likelihood Ratio) Test df
122.589
21.1397
0.583624
6
3
2
p-value
<.0001
<.0001
0.7469
The p-value for Test 1 is less than .05. There appears
to be a
difference between response and/or variances among the
dose levels.
It seems appropriate to model the data
A-61
-------
The p-value for Test 2 is less than .05. Consider
running a non-homogeneous variance model
The p-value for Test 3 is greater than .05. The model
chosen appears to adequately describe the data
Benchmark Dose Computation
Specified effect =
Risk Type
Estimated standard deviations from the control mean
Confidence level =
BMD =
0.95
12.3135
BMDL =
10.1957
Linear Model with 0.95 Confidence Level
4.5
CD
V)
o 3.5
Q.
V)
CD
o:
CD
CD
2.5
Linear
BMDL!
BMP
20
40
60
dose
80
100
120
23:2405/122005
A-62
-------
Polynomial Model. Revision: 2.2 Date: 9/12/2002
Input Data File: C:\BMDS\DATA\CYLINDRO.(d)
Gnuplot Plotting File: C:\BMDS\DATA\CYLINDRO.plt
Thu May 12 23:26:08 2005
HMDS MODEL RUN
The form of the response function is:
Y[dose] = beta 0 + beta l*dose + beta 2*doseA2 + ...
Dependent variable = MEAN
Independent variable = dose
rho is set to 0
Signs of the polynomial coefficients are not restricted
A constant variance model is fit
Total number of dose groups = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
alpha = 0.0503676
rho = 0 Specified
beta_0 = 4.23636
beta_l = -0.0164394
beta 2 = -7.57576e-006
Parameter Estimates
Variable Estimate
alpha 0.0456423
beta_0 4.23685
beta 1 -0.016519
beta 2
95.0% Wald Confidence Interval
Std. Err. Lower Conf. Limit Upper Conf. Limit
0.0104711 0.0251194 0.0661652
0.0648402 4.10977 4.36394
0.00280362 -0.022014 -0.011024
-6.92416e-006 2.18203e-005
-4.96911e-005
3.58428e-005
Asymptotic Correlation Matrix of Parameter Estimates
alpha
beta_0
beta_l
beta 2
alpha
1
1.2e-010
-1.4e-013
l.le-010
beta_0
1.2e-010
1
-0.73
0.58
beta_l
-1.4e-013
-0.73
1
-0.96
beta_2
l.le-010
0.58
-0.96
1
A-63
-------
Dose
Res .
Table of Data and Estimated Values of Interest
N Obs Mean Obs Std Dev Est Mean Est Std Dev
ChiA2
0
30
60
120
10
10
9
9
4.25
3.7
3.25
2.15
0.3
0.25
0.05
0.2
4.24
3.74
3.22
2.15
0.214
0.214
0.214
0.214
0.195
-0.519
0.41
-0.0684
Model Descriptions for likelihoods calculated
Model Al: Yij
Var{e(ij)}
Model A2: Yij
Var{e(ij)}
Model R: Yi
Var{e(i)}
Mu (i) + e (ij )
SigmaA2
Mu (i) + e (ij )
Sigma(i)A2
Mu + e (i)
S i gma A 2
Likelihoods of Interest
Test 1:
levels
Test 2:
Test 3:
Model Log(likelihood) DF AIC
Al 39.893006 5 -69.786011
A2 50.462856 8 -84.925712
fitted 39.651475 3 -73.302950
R -10.831438 2 25.662876
Does response and/or variances differ among dose
(A2 vs. R)
Are Variances Homogeneous (Al vs A2)
Does the Model for the Mean Fit (Al vs. fitted)
Tests of Interest
Test
Test 1
Test 2
Test 3
-2*log(Likelihood Ratio) Test df
122.589
21.1397
0.483061
p-value
<.0001
<.0001
0.487
The p-value for Test 1 is less than .05. There appears
to be adifference between response and/or variances among the
dose levels.It seems appropriate to model the data
The p-value for Test 2 is less than .05. Consider
running a non-homogeneous variance model
The p-value for Test 3 is greater than .05. The model
chosen appears to adequately describe the data
A-64
-------
Benchmark Dose Computation
Specified effect = 1
Risk Type
Estimated standard deviations from the control mean
Confidence level =
BMD =
0.95
12.8637
BMDL =
9.74645
Polynomial Model with 0.95 Confidence Level
4.5
CD
V)
o 3.5
Q.
V)
CD
c
CD
CD
2.5
Polynomial
ftMDL
BMP
20
23:2605/122005
40
60
dose
80
100
120
A-65
-------
Polynomial Model. Revision: 2.2 Date: 9/12/2002
Input Data File: C:\BMDS\DATA\CYLINDRO.(d)
Gnuplot Plotting File: C:\BMDS\DATA\CYLINDRO.plt
Thu May 12 23:27:07 2005
HMDS MODEL RUN
The form of the response function is:
Y[dose] = beta 0 + beta l*dose + beta 2*doseA2 + ...
Dependent variable = MEAN
Independent variable = dose
rho is set to 0
The polynomial coefficients are restricted to be negative
A constant variance model is fit
Total number of dose groups = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
alpha = 0.0503676
rho = 0 Specified
beta_0 = 4.23636
beta_l = -0.0164394
beta 2 = -7.57576e-006
Parameter Estimates
95.0% Wald Confidence Interval
Variable Estimate Std. Err. Lower Conf. Limit Upper Conf. Limit
alpha 0.0456423 0.0104711 0.0251194 0.0661652
beta_0 4.23685 0.0648402 4.10977 4.36394
beta_l -0.016519 0.00280362 -0.022014 -0.011024
beta 2 -6.92416e-006 2.18203e-005 -4.96911e-005 3.58428e-005
Asymptotic Correlation Matrix of Parameter Estimates
alpha beta_0 beta_l beta_2
alpha 1 1.2e-010 -1.4e-013 l.le-010
beta_0 1.2e-010 1 -0.73 0.58
beta_l -1.4e-013 -0.73 1 -0.96
beta 2 l.le-010 0.58 -0.96 1
A-66
-------
Dose
Res .
0
30
60
120
Table of Data and Estimated Values of Interest
N Obs Mean Obs Std Dev Est Mean Est Std Dev ChiA2
10
10
9
9
4.25
3.7
3.25
2.15
0
0
0
.
.
0
.3
25
05
.2
4
3
3
2
.24
.74
.22
.15
0
0
0
0
.214
.214
.214
.214
0
-0
-0.
.195
.519
0.41
0684
Model Descriptions for likelihoods calculated
Model Al: Yij
Var{e(ij)}
Model A2: Yij
Var{e(ij)}
Model R: Yi
Var{e(i)}
Mu (i) + e (ij )
SigmaA2
Mu (i) + e (ij )
Sigma(i)A2
Mu + e (i)
S i gma A 2
Likelihoods of Interest
Test 1:
levels
Test 2:
Test 3:
Model
Al
A2
fitted
R
Log (likelihood)
Does response
(A2 vs .
R)
Are Variances
Does the
39.893006
50.462856
39.651475
-10.831438
and/or variances
Homogeneous
DF
5
8
3
2
differ
AIC
-69.786011
-84.925712
-73.302950
25.662876
among dose
(Al vs A2)
Model for the Mean Fit
(Al vs
. fitted)
Test
Test 1
Test 2
Test 3
Tests of Interest
-2*log(Likelihood Ratio) Test df
122.589
21.1397
0.483061
p-value
<.0001
<.0001
0.487
The p-value for Test 1 is less than .05. There appears to be a difference
between response and/or variances among the dose levels.
It seems appropriate to model the data
The p-value for Test 2 is less than .05. Consider running a non-homogeneous
variance model
The p-value for Test 3 is greater than .05. The model
chosen appears to adequately describe the data
A-67
-------
Benchmark Dose Computation
Specified effect = 1
Risk Type
Estimated standard deviations from the control mean
Confidence level =
BMD =
0.95
12.8637
BMDL =
10.2295
Polynomial Model with 0.95 Confidence Level
4.5
CD
V)
o 3.5
Q.
V)
CD
o:
c q
CD -J
CD
2.5
Polynomial
BMDL
BMP
20
23:2705/122005
40
60
dose
80
100
120
A-68
-------
Power Model. $Revision: 2.1 $ $Date: 2000/10/11 20:57:36 $
Input Data File: C:\BMDS\DATA\CYLINDRO.(d)
Gnuplot Plotting File: C:\BMDS\DATA\CYLINDRO.plt
Thu May 12 23:27:43 2005
HMDS MODEL RUN
The form of the response function is:
Y[dose] = control + slope * doseApower
Dependent variable = MEAN
Independent variable = dose
rho is set to 0
The power is restricted to be greater than or equal to 1
A constant variance model is fit
Total number of dose groups = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
alpha = 0.0503676
rho = 0 Specified
control = 4.25
slope = -22.4349
power = -0.494765
Asymptotic Correlation Matrix of Parameter Estimates
alpha
rho
control
slope
power
alpha
1
-0.98
-0.041
-0.0042
-0.014
rho
-0.98
1
0.042
0.0043
0.014
control
-0.041
0.042
1
-0.67
-0.62
slope
-0.0042
0.0043
-0.67
1
1
power
-0.014
0.014
-0.62
1
1
A-69
-------
Parameter Estimates
Variable
alpha
rho
control
slope
power
Estimate
0.0457244
0
4.24139
-0.0159238
1.01795
Std. Err
0.0552045
0.931145
0.0671298
0.00783392
0.100882
Table of Data and Estimated Values of Interest
Dose
Res .
0
30
60
120
N
10
10
9
9
Obs
4.
3
3.
2.
Mean
25
.7
25
15
Obs
0
0
Std Dev
0.3
.25
.05
0.2
Est
4.
3.
3.
2.
Mean
24
73
21
16
Est
0
0
0
0
Std Dev
.214
.214
.214
.214
Chi
0
-
-0
A2
.0403
0.157
0.172
.0426
Model Descriptions for likelihoods calculated
Model Al: Yij = Mu(i) + e(ij)
Var{e(ij)} = SigmaA2
Model A2: Yij = Mu(i) + e(ij)
Var{e(ij)} = Sigma(i)A2
Model R: Yi = Mu + e(i)
Var{e(i)} = SigmaA2
Likelihoods of Interest
Model Log(likelihood) DF AIC
Al 39.893006 5 -69.786011
A2 50.462856 8 -84.925712
fitted 39.617353 4 -71.234706
R -10.831438 2 25.662876
Test 1:
Test 2:
Test 3:
Does response and/or variances differ among dose levels
(A2 vs. R)
Are Variances Homogeneous (Al vs A2)
Does the Model for the Mean Fit (Al vs. fitted)
Test
Test 1
Test 2
Test 3
Tests of Interest
-2*log(Likelihood Ratio)
122.589
21.1397
0.551305
df
6
3
1
p-value
<.00001
9.847e-005
0.4578
A-70
-------
The p-value for Test 1 is less than .05. There appears to be a
difference between response and/or variances among the dose levels.
It seems appropriate to model the data
The p-value for Test 2 is less than .05. Consider running a
non-homogeneous variance model
The p-value for Test 3 is greater than .05. The model chosen appears
to adequately describe the data
Benchmark Dose Computation
Specified effect =
Risk Type
Confidence level =
BMD =
Estimated standard deviations from the control mean
0.95
12.8275
BMDL =
10.2065
Power Model with 0.95 Confidence Level
4.5
CD
I 3.5
Q.
V)
CD
o:
CD
2.5
Power
BMDL BMP
20
40
60
dose
80
100
120
23:2705/122005
A-71
-------
Hill Model. $Revision: 2.1 $ $Date: 2000/10/11 21:21:23 $
Input Data File: C:\BMDS\DATA\CYLINDRO.(d)
Gnuplot Plotting File: C:\BMDS\DATA\CYLINDRO.plt
Thu May 12 23:28:26 2005
HMDS MODEL RUN
The form of the response function is:
Y[dose] = intercept + v*doseAn/(kAn + doseAn)
Dependent variable = MEAN
Independent variable = dose
rho is set to 0
Power parameter restricted to be greater than 1
A constant variance model is fit
Total number of dose groups = 4
Total number of records with missing values = 0
Maximum number of iterations = 250
Relative Function Convergence has been set to: le-008
Parameter Convergence has been set to: le-008
Default Initial Parameter Values
alpha = 0.0273861
rho = 0 Specified
intercept = 4.25
v = -2.1
n = 1.7744
k = 62.7273
Asymptotic Correlation Matrix of Parameter Estimates
alpha rho intercept
0
0
1
0
0
0
alpha
rho
intercept
V
n
k
1
0
0
0
0
0
0
1
0
0
0
0
V
0
0
0
1
0
0
n
0
0
0
0
1
0
k
0
0
0
0
0
1
A-72
-------
Parameter Estimates
Variable
alpha
rho
intercept
v
n
k
Estimate
0.0457348
0
4.24125
-461.591
1.02098
23703.4
Std.
Err.
1
1
1
1
1
1
Table of Data and Estimated Values of Interest
R
-
Dose
es .
0
30
60
120
N
10
10
9
9
Obs
4.
3
3.
2.
Mean
25
.7
25
15
Obs
0
0
Std Dev
0.3
.25
.05
0.2
Est
4.
3.
3.
2.
Mean
24
73
21
16
Est
0
0
0
0
Std Dev
.214
.214
.214
.214
ChiA2
0.0409
-0.159
0.174
-0.0428
Model Descriptions for likelihoods calculated
Model Al: Yij
Var{e(ij)}
Model A2: Yij
Var{e(ij) }
Model R: Yi
Var{e(i) }
Mu (i) + e (ij )
SigmaA2
Mu (i) + e (ij )
Sigma(i)A2
Mu + e (i)
SigmaA2
Degrees of freedom for Test Al vs fitted <= 0
Likelihoods of Interest
Model Log(likelihood) DF
Al 39.893006 5
A2 50.462856 8
fitted 39.613009 5
R -10.831438 2
Test 1:
Test 2:
Test 3:
Test
AIC
-69.786011
-84.925712
-69.226018
25.662876
Does response and/or variances differ among dose levels
(A2 vs. R)
Are Variances Homogeneous (Al vs A2)
Does the Model for the Mean Fit (Al vs. fitted)
Tests of Interest
-2*log(Likelihood Ratio) Test df
p-value
A-73
-------
Test 1 122.589 6 <.0001
Test 2 21.1397 3 <.0001
Test 3 0.559993 0 NA
The p-value for Test 1 is less than .05. There appears to be a
difference between response and/or variances among the dose levels.
It seems appropriate to model the data
The p-value for Test 2 is less than .05. Consider running a
non-homogeneous variance model
NA - Degrees of freedom for Test 3 are less than or equal to 0. The Chi-
Square test for fit is not valid
Benchmark Dose Computation
Specified effect = 1
Risk Type = Estimated standard deviations from the control mean
Confidence level = 0.95
BMD = 12.8644
Warning: optimum may not have been found. Bad completion code in
Optimization routine.
BMDL computation failed.
A-74
-------
Hill Model
4.5
CD
V)
o 3.5
Q.
CD
o:
c q
CD °
CD
2.5
Hill
BMP
20
40
60
dose
80
100
120
23:2805/122005
A-75
------- |