Materials Submitted to the National Research Council
Part 2: Chemical-Specific Examples

U.S. Environmental Protection Agency
Integrated Risk Information System Program

January 30, 2013

DISCLAIMER

This document is for review purposes only. It has not been formally disseminated by EPA. It does not represent and
should not be construed to represent any Agency determination or policy.


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Chemical-Specific Examples Demonstrating
Implementation of NRC's 2011 Recommendations

The following are intended to provide the NRC panel with examples of how the IRIS Program is
implementing the NRC recommendations included in the 2011 Review of the Environmental Protection
Agency's Draft IRIS Assessment of Formaldehyde. The examples are not to be construed as final
Agency conclusions and are provided for the sole purpose of demonstrating the IRIS implementation of
the NRC recommendations.


-------
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

31

32

33

34

35

36

DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

EXAMPLE 1 - Literature Search and Screening

This example demonstrates the implementation of an improved literature search strategy as described
in the "Identifying and Selecting Pertinent Studies" section of the draft Handbook for IRIS Assessment
Development. The literature search strategy used to identify the studies to be included in the draft
assessment, as well as the presentation of the literature search documentation, is shown below.

Literature search for Ethyl tert-butyl ether (ETBE)

1. Initial chemical-specific search conducted in online scientific databases

a.	Pubmed database (http://www.ncbi.nlm.nih.gov/pubmed/) searched (1/8/13) for all
articles on Ethyl tert-butyl ether using the following search string:

"ETBE" OR "Ethyl tert-butyl ether" OR "2-Ethoxy-2-methyl-propane" OR "ethyl tertiary
butyl ether" OR "ethyl tert-butyl oxide" OR "tert-butyl ethyl ether" OR "ethyl t-butyl
ether" OR "637-92-3"

Search returned: 188 articles

b.	Toxline and DART searched (1/8/13) using the ToxNet database
(http://toxnet.nlm.nih.gov/) using the following search string excluding PubMed
records:

"ETBE" OR "Ethyl tert-butyl ether" OR "2-Ethoxy-2-methyl-propane" OR "ethyl tertiary
butyl ether" OR "ethyl tert-butyl oxide" OR "tert-butyl ethyl ether" OR "ethyl t-butyl
ether" OR "637-92-3"

Search returned: 110 articles (110 from Toxline; 0 from DART)

c.	TSCATS 2 (http://vosemite.epa.gov/oppts/epatscat8.nsf/ReportSearch7openform) was
searched using the CAS number 637-92-3 for the EPA receipt dates of 1/01/2004-
01/01/2013 since Toxline searches TSCATS through 2003.

Search returned: 1 article

d.	Web of Science database

(http://apps.webofknowledge.com/WOS GeneralSearch input.do?highlighted tab=WO
S&product=WOS&last prod=WOS&SID=lDg72P6B9iG5G14Nd7L&search mode=Genera
ISearch) was searched (1/8/13) using the following search string with lemmatization
"on":

"ETBE" OR "Ethyl tert-butyl ether" OR "2-Ethoxy-2-methyl-propane" OR "ethyl tertiary
butyl ether" OR "ethyl tert-butyl oxide" OR "tert-butyl ethyl ether" OR "ethyl t-butyl
ether" OR "637-92-3"

Search returned: 490 articles

e.	Proquest database
(http://search.proquest.com/environmentalscience/index?accountid=102841) was
searched (1/8/13) using the following search string including only scholarly journals:

2


-------
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

31

32

33

34

35

DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

"ETBE" OR "Ethyl tert-butyl ether" OR "2-Ethoxy-2-methyl-propane" OR "ethyl tertiary
butyl ether" OR "ethyl tert-butyl oxide" OR "tert-butyl ethyl ether" OR "ethyl t-butyl
ether" OR "637-92-3"

Search returned: 389 articles

2.	Total articles found: 1178 articles

a. 514 were duplicates and removed by EPA's HERO search.

3.	664 unique articles imported into an EndNote library from the HERO web site.

a.	27 references identified as reviews by EndNote query for "review"

b.	Review references removed from list and manually screened.

c.	2 review references chosen for "snowball" search

i.	McGregor, D. (2007). "Ethyl tertiary-butyl ether: a toxicological review." Critical
Reviews in Toxicology 37(4): 287-312.

ii.	de Peyster, A. (2010). "Ethyl t-butyl ether: Review of reproductive and
developmental toxicity." Birth Defects Research. Part B: Developmental and
Reproductive Toxicology 89(3): 239-263.

4.	108 cited references from 2 reviews were identified using Web of Science

a.	16 references were duplicates and removed by EndNote upon import

b.	92 unique references were imported into the EndNote library

5.	Title and abstract screened manually within the EndNote library for relevance and excluded
from further consideration for development of the hazard identification in the Toxicological
Review based on the following criteria:

a.	Biodegradation/environmental fate (69)

b.	Chemical analysis/fuel chemistry (323)

c.	Study on non ETBE chemical (105)

d.	Non-relevant exposure (2)

e.	Policy papers (27)

f.	Duplicate, society abstracts, reviews/commentary, case studies, miscellaneous (135)

g.	Foreign language (5)

h.	Risk assessment (1)

6.	62 articles verified by full text review. No articles removed

7.	17 unpublished studies conducted by the Japanese Petroleum Energy Center were provided via
direct correspondence. Studies were identified from a submitted report identified in the full
text screen. All studies were screened for relevance and none were removed.

8.	79 articles were grouped into broad categories and were evaluated for study quality in the next
step ("considered" studies).

3


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

1 Figure 1-1. Literature search documentation for ETBE

	«

OJ	

* b

¦O ^

	LO



-514

f

664 Articles

-27

637 Articles

+92

-16

729 Articles

1

-667

62 Articles



t

J

+1? >

| 79 Articles

Duplicates

Duplicates

Title & Abstract Screening

69 biodegradation/ environmental fate

323 chemical analysis/fuel chemistry

105 studies of other chemicals

2 non-relevant exposure paradigms

27 policy/current practice papers

135 duplicates, case studies, society abstracts, review/

commentary, pilot studies.

5 foreign language

1 risk assessment

FullText Screening:

Studies sorted according to broad categories

Studies moved to Study Quality Evaluation step ("Considered" studies)
Conducted January, 2013

4


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

1	EXAMPLE 2 -Evaluation and Display of Individual Studies

2	This example demonstrates the tables used to evaluate the pertinent studies (including epidemiology

3	and animal toxicology studies) identified through the literature search and screening step with respect

4	to potential methodological considerations¦, as described in the "Evaluation and Display of Individual

5	Studies" section of the draft Handbook for IRIS Assessment Development. This section will likely be

6	expanded upon in the assessment, but the tables below serve as examples of the table format used to
1	present the study evaluation results.

5


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Table 2-1. Evaluation of observational epidemiology studies of diethyl phthalate - sexual differentiation effects (gray shading indicates a
potential weakness or limitation of the study)

Reference,









Analysis and





Participant

Exposure





Presentation of

Sample



Setting and

Selection,

Measure and

Outcome

Consideration of

Results (Estimate

Size;

Evaluation of Major

Design

Follow-up

Comparability Range

Measure

Likely Confounding

and Variability)

Power

Limitations

Anogenital Distance

Suzuki etal. Recruitment process
(2011).	not described.

Japan. Birth Enrolled at prenatal
cohort	visit (mean 29 weeks

gestation)

120 of 344 enrollees
excluded because did
not delivery at study
hospital

Internal

comparison

group

Maternal urine

Anogenital

Gestational age, birth

Described as not

(9 - 40 weeks;

distance,

order, maternal age,

associated (details not

mean 29

measured at

maternal smoking and

reported)

weeks), MEP,

birth (1-3 days);

environmental tobacco



75th percentile

blinded to

smoke exposure



= 32 ng/mL (44

exposure.

(stepwise regression);



ng/mL with SG

Protocol

Used SG-corrected urine



correction)

described; 23

concentrations





assessors;







reliability







measures not







reported.





Maternal urine

Anogenital

Adjusted for weight

Percent change per

(mean 29

distance,

percentile and age. Did

interquartile increase in

weeks), MEP,

measured at

not adjust for MBP or

metabolite and p-value;

75th percentile

ages 2 - 36

MEHP.

also presented as

= 437 ng/mL

months; blinded



metabolite distribution



to exposure.



by 3 categories of



Multiple



anogenital distance.



assessors(3







sites); reliability







measures not







reported





n = 111
male
infants

Relatively low, narrow
exposure range.

Swan, 2008;
Swan et al.,
2005; 2003.
United States
(3 sites).

Birth cohort
(first follow-
up)

Standardized
recruitment process
(Sept 1999 - Aug
2002). 85% of cohort
agreed to be
recontacted. Eligible
if pregnancy ended
in live birth, was
currently 2-36
months of age, and
lived within 50 miles
of study center. 72%
of eligibles
participated in
follow-up; 75% of
participants had
maternal urine
sample and complete
physical exam data
(21 enrollees
excluded because
AGD exam not
considered reliable
(child too active; 2
declined)	

Internal

comparison

group

n=106 Is age-size adjustment
boys adequate (considering
potential temporal
changes in exposure)?

6


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Table 2-1. Evaluation of observational epidemiology studies of diethyl phthalate - sexual differentiation effects (gray shading indicates a
potential weakness or limitation of the study)

Reference,











Analysis and





Participant



Exposure





Presentation of

Sample



Setting and

Selection,



Measure and

Outcome

Consideration of

Results (Estimate

Size;

Evaluation of Major

Design

Follow-up

Comparability

Range

Measure

Likely Confounding

and Variability)

Power

Limitations

Cryptorchidism or Testicular Position













Main et al.,

Cases identified

Cases and

Breast milk

Cryptorchidism,

Analyzed separately by

SE and exact p-value for

n=62

Exposure measure

2006; Boisen

through standardized

controls well-

samples

at birth or 3

country and combined;

difference not given,

cases,

may not reflect in

et al., 2004.

examination; all

matched by

collected 1-3

months; blinded

no other variables

but p > 0.40

n=68

utero exposure; breast

Denmark and

births at two

maternal

months of age,

to exposure.

addressed



controls

pump use could

Finland.

university hospitals

characteristics

MEP, upper

Coordination







increase MEP levels

Nested case-

(one per country).



range not

and training of









control study

1997-2001



reported

assessors









within birth

(Denmark); 1997-





discussed;









cohort

1999 (Finland)





borderline cases
reviewed by two
assessors









Swan, 2008;

See entry above

Internal

Maternal urine

One or both

Did not adjust for MBP or

Described as not

n=119

Outcome seen in 10%

Swan et al.,



compa rison

(mean 29

testicles not

MEHP

associated (details not

boys

of the study sample;

2005; 2003.



group

weeks), MEP,

"normal" or



reported)



unclear what this

United States





75th percentile

"normal







represents from

(3 sites).





= 437 ng/mL

retractile" at







clinical perspective

Birth cohort







clinical exam
(ages 0- 36
months); blinded
to exposure.
Multiple
assessors (3
sites); reliability
measures not
reported









Infant Hormone Levels















Lin et al.,

Pregnant women

Internal

Maternal urine

Cord blood

Gestational age, maternal

Beta, but no SE,

n=81 boys,

Limited analysis

2011; Wang

seen in prenatal

comparison

(3rd trimester,

hormone levels;

age, gravidity, smoking,

reported for regression

74 girls



et al.,2004.

clinic (>18 weeks

group

28-36 weeks),

blinded to

body mass index, ever

analyses (continuous





Taiwan. Birth

gestation) and



MEP, 95th

exposure

oral contraceptive use,

measures)





cohort

intending to deliver
in that hospital
invited to participate
(singleton births, no
medical

complications). Dec
2000—Nov 2001.



percentile-241
ng/mL (346
ug/g creatinine)



other phthalate
metabolites (stepwise
regression); Used
creatinine-adjusted
concentrations







7


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Table 2-1. Evaluation of observational epidemiology studies of diethyl phthalate - sexual differentiation effects (gray shading indicates a
potential weakness or limitation of the study)

Reference,









Analysis and





Participant

Exposure





Presentation of

Sample



Setting and

Selection,

Measure and

Outcome

Consideration of

Results (Estimate

Size;

Evaluation of Major

Design

Follow-up

Comparability Range

Measure

Likely Confounding

and Variability)

Power

Limitations

275 of 430 women in
cohort provided
urine sample; 120 of
275 excluded
because of missing
cord blood or other
data; did not differ
by age, body mass
index, smoking or
alcohol use

Mainetal.,	See entry above.	Internal

2006	Cases and controls	comparison

Denmark and	combined forthis	group

Birth cohort	analysis

Breast milk
samples
collected 1-3
months of age,
MEP, upper
range not
reported	

Serum hormone
levels at 3
months; blinded
to exposure

Analyzed separately by
country and combined;
no other variables
addressed

Spearman correlation	n=130

coefficients and p-	boys

values. Did not adjust
for MBP

Exposure measure
may not reflect in
utero exposure; breast
pump use could
increase MEP levels

Gender-Related Play

Swan et al.

See entry above. 128

Internal

Maternal urine

Pre-school

Covariates considered:

2010; Swan

of 334 eligible not

compa rison

(mean 29

Activities

creatinine concentration,

et al., 2005;

found; 56 of found

group

weeks), MEP,

Inventory (24

child's sex, maternal age,

2003.

did not participate.



75th percentile

items,

parental education,

United States

Higher percentage of



= 437 ng/mL

completed by

number of same and

(4 sites -

mothers of



(based on

parents;

opposite sex siblings,

Iowa added

participating families



earlier

instrument used

clinic location, parental

2002-2005).

were white (88%



publications)

in previous

attitude regarding sex-

Second

compared with 78%)





studies of direct

atypical play; kept in

follow-up of

and completed





and indirect

model if >10% change in

birth cohort.

college (73%





measures of

effect estimate (retained:



compared with 68%)





testosterone);

maternal age, boy's age,









blinded to

parental education,









exposure

parental attitude, and











education-attitude











interaction)

Described as not
associated (details not
reported)

n=74 boys,
71 girls

8


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Table 2-2. Evaluation of observational epidemiology studies of diethyl phthalate - neurobehavioral effects (gray shading indicates a
potential weakness or limitation of the study)

Reference,









Analysis and





Participant

Exposure





Presentation of

Sample



Setting and

Selection,

Measure and

Outcome

Consideration of

Results (Estimate

Size;

Evaluation of

Design

Follow-up

Comparability Range

Measure

Likely Confounding

and Variability)

Power

Major Limitations

Engel et al.,
2009; Wolff
et al., 2008.
United States
(Mt Sinai,
New York).
Birth cohort.

Seen for prenatal
care at Mt Sinai
hospital or two
private practices and
delivered at Mt Sinai.
Singleton,
primiparous
pregnancies,
delivered May 1998
-July 2001. 475
initially recruited;
404 of these eligible
(28 left area; 19
refused; 28
miscellaneous other
reasons). Outcome
not measured in 93
of 404 enrollees
(excluded if in NICU,
only in hospital on
weekend, parent
refused, baby not
testable, or study
personnel

unavailable). Of the
311 with outcome
data, 295 also had
urine sample.

Models were
restricted to
observations with
values >20 mg/dl_

Internal
compa rison
group

Maternal urine,

Brazelton

Covariates considered

Beta and 95% CI for

MEP (25-40

Neonatal

included maternal age,

summation of low

weeks, mean

Behavioral

race, marital status,

molecular weight

32), 75th

Assessment Scale

education, cesarean

metabolites (MEP,

percentile 1, 025

(7 domains; 28

delivery, delivery

MBP, MiBP and

ng/mL

behavioral items

anesthesia, infant age,

MMP)



and 18 primitive

infant sex, infant





reflexes); 4

jaundice, maternal





trained examiners

smoking, alcohol,





(no information

caffeine, and illicit drug





on agreement);

use, urinary creatinine,





blinded to

examiner, and maternal





exposure

urinary organophosphate







levels. Dropped from







model if <10% change in







Beta coefficient







compared with full







model. Also examined







interaction by sex of







child.



n=295 Data presented only
for summation of low
molecular weight
metabolites

Engel et al.,

See Engel et al.

Small

Maternal urine,

Behavior Rating

Covariates considered

Beta and 95% CI for n=177

2010; Wolff

(2009) for cohort

differences in

MEP (25-40

Inventory of

based on relation with

summation of low

et al., 2008.

description. BASC F

education level

weeks, mean 32),

Executive

phthalates metabolites

molecular weight

United States

scores > 3 excluded

and age in non-

MEP (distribution

Function *86

and outcomes. Also

metabolites

(Mt Sinai,

because of

participants at

not given but

items, 8

examined interaction by

(continuous, tertiles

New York).

questionable validity

follow-up

assumed similarto

subscales);

sex of child. Adjusted for

MEP); Beta and p <

Follow-up(s)

(n=2); 25 scores of 2

compared with

other studies from

Behavior

race, sex, education and

0.05 denoted for MEP

9


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Table 2-2. Evaluation of observational epidemiology studies of diethyl phthalate - neurobehavioral effects (gray shading indicates a
potential weakness or limitation of the study)

Reference,











Analysis and





Participant



Exposure





Presentation of

Sample



Setting and

Selection,



Measure and

Outcome

Consideration of

Results (Estimate

Size;

Evaluation of

Design

Follow-up

Comparability

Range

Measure

Likely Confounding

and Variability)

Power

Major Limitations

of birth

or 3 reviewed and 12

participants,

this cohort)

Assessment

marital status of primary

(continuous)





cohort.

excluded because of

but little



System for

caretaker, and urinary









concerns about

difference in



Children (BASC,

creatinine









language (n=2),

MEP between



130 items, 9











random responses

groups.



scales, parent











(n=7), or overly

Internal



ratings); used in











negative or

comparison



previous studies











unrealistic evaluation

group



of executive











(n=3). Internal





functioning and











comparison group





behavior; blinded

















to exposure









Miodovnik et

See Engel et al.

Higher

Maternal urine,

Social

Covariates considered:

Beta and 95% CI

n=137



al., 2011;

(2009) for cohort

proportion of

MEP (25-40

Responsiveness

maternal age, maternal

(continuous MEP)





Wolff et al.,

description. 137 of

lower

weeks, mean

Scale (65items,

IQ, marital status at the







2008.

original 404

education

32), MEP, 75th

completed by

time of follow-up,







United States

completed 7-9 year

(
-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

1	Table 2-3. Evaluation of animal toxicology studies for chemical X

2

Reference
(Species)

Exposure
Quality

Test Subjects

Study Design

Toxicity Endpoints

Data and
Statistics

Reporting

Smith et al. (1984)

(Monkey)

++

++

Note: N=20

++

Note: 102 wk study

++

Not applicable

++

Jones et al. (1986)

(Mouse)

+ co-exposure
likely

+ N=5; variable ages at
onset of exposure
across groups

++

Note: 13 wk study

Potential sampling bias;
No observer blinding
indicated; protocols
incompletely reported

+ data represents
pooled sexes

+ Surgical
procedures not
reported

Gray et al. (2012)

(Rat)

Test article and
exposure
methods not
specified

Bacterial infection
noted in animal colony;
N= 3 litters; males only;
overt maternal toxicity

No randomization across
litters into treatment groups;

testing during exposure
expected to confound results;
acute exposure

++

Not applicable

Results data not
reported

Criteria for the six categories developed based on the chemical and hazard type in question. In this example: gray box = examination of relevant study details identified
potential limitations that could influence interpretation of the study's results; '+' = criteria not completely met or potential issues identified, but unlikely to directly
affect study interpretation; ++ = criteria determined to be completely met. Text accompanying summary table would explain key study details informing these
determinations.

11


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

1	EXAMPLE 3 - Evidence Tables

2	This example demonstrates the evidence table format to be used for presenting the epidemiological

3	and toxicological evidence available for endpoint-specific hazards¦, as described in the "Evaluation and

4	Display of Individual Studies" section of the draft Handbook for IRIS Assessment Development.

5	• Human Evidence

6

Table 3-1. Evidence pertaining to male reproductive effects of diethyl phthalate in
humans

Reference and Study Design3

Results

Reproductive hormones

Meeker et al., 2009 (United States; Boston) (Tier 1)
425 male partners seen in subfertility clinic, mean age 36
years, 2000-2004

Serum, steroidal and gonadotropin hormones

Urine sample, median (90th percentile) MEP 153 (1376)

ng/mL

Beta (95% CI) for In-MEP in relation to hormone
0.0 = no effect
Testosterone 8.87 (-7.18, 24.9)

Estradiol 0.71 (-0.97, 2.40
1.0 = no effect
Free androgen index 1.04 (0.99,1.09)
FSH 0.98(0.91,1.06)
LH 0.98 0.91,1.04)

Adjusted for age, body mass index, smoking, season and time
of sample collection (and time squared), dilution ranking

Jonsson et al., 2005 (Sweden) (Tier 2)

234 men ages 18-21 years (military service)

Serum, steroidal and gonadotropin hormones.

Urine samples, median (95th percentile) MEP 240 (4400)

ng/mL; 83 (1600 nmol/mmol creatinine)

Mean difference (95% CI), highest compared with lowest quartile
of MEP

Testosterone (nM) -0.3 (-2.3,1.8)

Free testosterone (T/SHBG) 0.06 (-0.05, 0.2)
Estradiol (pM) 1.8 (-4.2, 7.7)
FSH (IU/L) 0.5 (-0.5, 0.6)

LH (IU/L) 0.7(0.1,1.2)

(Positive difference indicates lower value in highest exposure
quartile)

Abstinence time and smoking evaluated as confounders

Sperm parameters

Hauser et al., 2007, 2006 (United States; Boston) (Tier 1)
463 male partners seen in subfertility clinic, mean age 36
years, 2000-2004 [n=379 for damage measures]

Semen analysis, sperm damage measures analyzed
Urine sample, median (75th, 95th percentile) MEP 158
(535, 2214) ng/mL (specific-gravidity adjusted)

OR (95% CI), by metabolite quartile of MEP

Concentration Motility Morphology
MEP (< 20 x 10e/mL) (< 50% motile) (<4% normal)

1	(low) 1.0 (referent) 1.0 (referent) 1.0 (referent)

2	1.5 (0.7,3.6) 1.1 (0.6, 1.9) 0.8 (0.4, 1.6)

3	1.0(0.4,2.5) 0.8(0.5,1.5) 0.7(0.3,1.3)

4	(high) 1.2(0.5,3.0) 1.0(0.6,1.8) 0.5(0.3,1.1)
(trend p) (0.94) (0.84) (0.07)
Logistic regression adjusted for age, abstinence time, and
smoking.

Damage measures, Beta (95% CI) associated with interquartile
range increase

Comet extent Tail distribution

(nm) (nm) %DNA tail
MEP 6.06(0.941,12.3) 2.72(0.46,5.00) -0.26 (-2.52,2.02)
linear regression, adjusted for age and smoking

12


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Table 3-1. Evidence pertaining to male reproductive effects of diethyl phthalate in
humans

Reference and Study Design3

Results

Pant et al., 2008 (India) (Tier 2)

300 men, mean age 29 years (100 fertile, 200 infertile),
urban and rural
Semen analysis

DEP concentration in semen for fertile group, mean (±SE)
0.64 (± 0.24) in rural, 0.74 (± 0.04) |ag/mL in urban areas

Pearson correlation coefficient between semen DEP and sperm
parameter:

r (p-value)

Sperm concentration -0.19 (p< 0.05)
Sperm motility (%) 0.03
Morphology (% abnormal -0.02
Damage (Chromatin integrity) 0.07
(all other p-values > 0.05; exact value not reported)

Zhang et al., 2006 (China) (Tier 2)

52 men seen in Shanghai Institute of Planned Parenthood
Research clinics, mean age 32 years
Semen analysis

DEP concentration in semen, mean 0.47 mg/L

Spearman correlation coefficient between semen DEHP and
sperm parameter:

r (p-value)

Sperm concentration -0.25 (0.15)
Sperm motility (%) -0.13 (0.45)
Sperm rate of malformations 0.19 (0.28)

Jonsson et al., 2005 (Sweden) (Tier 2)

234 men ages 18-21 years (military service)

Semen analysis

Urine samples, median (95th percentile) MEP 240 (4400)
ng/mL, or 83 (1600 nmol/mmol creatinine)

Vlean difference (95% CI), highest compared with lowest quartile MEP
Sperm concentration (x 106/mL) 5.0 (-15. 25)

Sperm motility (%) -0.4 (-6.4, 5.6)

Sperm damage (chromatin integrity) 0.8 (-2.8, 4.4)
(Positive difference indicates lower value in highest exposure quartile)
Abstinence time and smoking evaluated as confounders

Liu et al., 2012 (China) (Tie 2)

97 male partners seen in subfertility clinic, mean age 32
years

Semen analysis

Urine sample, median (66th percentile) MEP 12.6 (21.3)
ng/mL

OR (95% CI), by metabolite tertile

Concentration Motility
MEP (<20 x 10e/mL) (<50% motile)

1	1.0 (referent) 1.0 (referent)

2	1.4(0.2,8.8) 0.7(0.2,1.9)

3	1.5(0.2,9.6) 0.4(0.1,1.2)

(trend p) (0.66) (0.10)

adjusted for age, body mass index, abstinence time, smoking,
alcohol use

Infertility

Tranfo et al., 2012 epub (Italy) (Tier 2)

Case-control study, 56 couples from assisted
reproduction center, n=56 control couples (parents),
mean age 39 years in both groups.

Case-control comparison

Urine samples, median (95th percentile) MEP 52 (651)
Mg/g creatinine (controls); slightly higher in women than
men

Comparison between MEP levels in cases and controls

Mann-Whitney
test p-value
Females <0.001
Males <0.001
Additional details of sex-stratified results not provided

Pant et al., 2008 (India) (Tier 2)

300 men, mean age 29 years (100 fertile, 200 infertile),
urban and rural

DEP concentration by fertility status (based on partners
who had conceived within 1 year of attempting
pregnancy)

DEP concentration in semen for fertile group, mean (±SE)
0.64 (± 0.24) in rural, 0.74 (± 0.04) |ag/mL in urban areas

DEP concentration in semen, mean ± SE (p-value for difference
between fertile and infertile):

Rural: Fertile Infertile

0.64 ± 0.24 0.74 ±0.04
Urban: Fertile Infertile

1.13 ±0.11 3.11 ±0.26 (p < 0.05)

a "Tier" reflects evaluation of confidence in study results based on evaluation of risk of specific types of bias. (In the
assessment, details of evaluation will be shown in Supplemental Information tables and text).

1

13


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

1	• Animal evidence

2

Table 3-2. Evidence pertaining to female reproductive effects of diethyl phthalate in
animals

Reference and Study Design

Results

Fertility and birth outcomes

Fujii et al. (2005)

No. of implantations (percent change compared to control)

Rat (Sprague Dawley); 21-



0

51/56

255/267

1297/1375

24/sex/group

F0 parental



2%

1%

1%

0, 600, 3,000, 15,000 ppm (0, 40,

females



197,1016 mg/kg-day in F0 males;
0, 51, 255,1297 mg/kg-day in F0

F1 parental
females

-

0%

4%

3%

females; 0, 46, 222,1150 mg/kg-

Fertility Index (percent change compared to control)

day in F1 males; 0, 56, 267,1375



0

51/56

255/267

1297/1375

mg/kg-day in F1 females)
Diet

F0 parental
females

-

0%

4%

0%

105 days for F0 and F1 parental
males

F1 parental
females

-

0%

0%

0%

119 days for F0 and F1 parental

Gestation length (days) (percent change compared to control)

females

(exposure through 10 weeks
premating + 3 weeks mating +

F0 parental
females

0

51/56
0%

255/267
0%

1297/1375
-1%

weaning)

F1 parental
females

-

0%

0%

-1%*



No. of pups delivered (percent change compared to control)





0

51/56

255/267

1297/1375



F0 parental
females

-

-1%

1%

1%



F1 parental
females

-

4%

7%

2%

Hardin et al. (1987)

Mouse (Swiss); 50 females/group

(percent change compared to control)





0



4500
0%

0, 4500 mg/kg-day

Gavage

GD6-GD13

No. of live pups/litter



Percent survival

-



-4%

Birth weight



-



-6%

Howdeshell et al. (2008)

(percent change compared to control)

Rat (Sprague Dawley); 3-5





0 100

300

600 900

female (dams)/group and 9

No. of implantations

5%

3%

4% 13%

control dams

No. of live fetuses

7%

5%

-6% 16%

0,100, 300, 600, 900 mg/kg-day

Total resorptions

0%

0% 325%* 0%

Gavage
GD8-GD18

Fetal mortality (%)

0%

0% 283%* 0%

14


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Table 3-2. Evidence pertaining to female reproductive effects of diethyl phthalate in
animals

Reference and Study Design

Results

NTP (1984)

Mouse (Swiss); 20/sex/group

(percent change compared to control)

















0, 0.25, 1.25, 2.5%

F0 females





0

0.25

1.25



2.5

Diet

















7 days premating + 98 days

No. of live pups/litter



-

23%*

14%



3%

cohabitation + 21 days
segregation (126 days total) (F0

Live pup weight





-

-2%

-2%



1%

males and females), and
0, 2.5% (0, 3640 mg/kg-day) in

F1 females





0





2.5



utero + lactation, and then in the

No. of live pups/litter



-





-14%

*

diet through a 7-day mating

















period at 74±10 days old (F1

Fertility index (%)



-





0%



females were allowed to deliver
litters)

Live pup weight





-





-3%



NTP (1988)

(percent change compared to control)

Rat (Sprague Dawley); 31-32



0



198



1909



3214

females (dams)/group

Corpora lutea





4%



2%



1%

0, 0.25, 2.5, 5% (0, 198, 1909,

per dam









3214 mg/kg-day)
Diet

Implantation
sites per litter

-



4%



1%



2%

GD6-GD 15

Resorptions
per litter
Percent

resorptions per
litter

Live fetuses per
litter

-



5%
2%
4%



13%
7%
-2%



-11%
-18%
3%

Singh et al. (1972)



Untreated

0.506



1.012



1.686

Rat (Sprague Dawley); 5 time-

No. of corpora

60



65



59



57

mated females/group

lutea







0, 0.506, 1.012, 1.686 mL/kg

No. of

0



28



0





Intraperitoneal

resorptions







z

injections on GD5,10, and 15

No. of live

59



35



57



54

(termination on GD 20)

fetuses







15


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Table 3-2. Evidence pertaining to female reproductive effects of diethyl phthalate in
animals

Reference and Study Design

Results

Anogenital distance

Fujii et al. (2005)

Rat (Sprague Dawley), 21-

(percent change compared to control)







24/sex/group

























0, 600, 3,000, 15,000 ppm (0, 40,
197,1016 mg/kg-day in F0 males;

Females



0

40-56

197-267

1016-1375

0, 51, 255,1297 mg/kg-day in F0













females; 0, 46, 222,1150 mg/kg-

F1 pups at PND 0



-

-5%

-5%

1%

day in F1 males; 0, 56, 267,1375













mg/kg-day in F1 females)
Diet

F1 pups at PND 4



-

-3%

-2%

-1%

105 days for F0 and F1 parental













males,

119 days for F0 and F1 parental

F2 pups at PND 0



-

-2%

0%

-1%

females (exposure through 10













weeks premating + 3 weeks

F2 pups at PND 4



-

-1%

-1%

-2%

mating + weaning)













Reproductive organ weights

Fujii et al. (2005)

Absolute ovary weight (percent change compared to control)

Rat (Sprague Dawley), 21-





0

51/56

255/267

1297/1375

24/sex/group

F0



-

-4%

-10%

-6%

0, 600, 3,000, 15,000 ppm (0, 40,

F1



-

1%

2%

4%

197,1016 mg/kg-day in F0 males;

F1 pup



-

4%

-8%

-4%

0, 51, 255,1297 mg/kg-day in F0

F2 pup



-

0%

0%

-4%

females; 0, 46, 222,1150 mg/kg-

Relative ovary weight (percent change compared to control)

day in F1 males; 0, 56, 267,1375
mg/kg-day in F1 females)

F0



0

51/56
-5%

255/267
-8%

1297/1375
-5%

Diet

F1



-

0%

0%

2%

105 days for F0 and F1 parental

F1 pup



-

7%

-3%

17%

males,

F2 pup



-

-3%

-3%

0%

119 days for F0 and F1 parental

Absolute uterus weight (percent change compared to control)

females (exposure through 10
weeks premating + 3 weeks

F0



0

51/56
2%

255/267
4%

1297/1375
-4%

mating + weaning)

F1



-

4%

7%

-1%



F1 pup



-

3%

7%

-22%*



F2 pup



-

-11%

-17%

-27%*



Relative uterus weight (percent change compared to control)







0

51/56

255/267

1297/1375



F0



-

0%

6%

-3%



F1



-

4%

4%

0%



F1 pup



-

5%

9%

-5%



F2 pup



-

-12%

-17%

-20%*

16


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Table 3-2. Evidence pertaining to female reproductive effects of diethyl phthalate in
animals

Reference and Study Design

Results

Pereira et al. (2007c)

Relative ovary weight (percent compared to control)

Rat (Wistar); 6/sex/group

F0 parental females





Fl adult females

0, 50 ppm (F0) (0, 2.85 mg/kg-

0

2.85



0 1.425

day)









0, 25 ppm (Fl) (0,1.425 mg/kg-









day)

-

40%*



23%*

Diet









150 days/generation









NTP (1984)

Mouse (Swiss); 20/sex/group
0, 0.25, 1.25, 2.5%

Ovary weight (percent change compared to control)



0

3640

Diet

Absolute





-3%

7 days premating + 98 days
cohabitation + 21 days
segregation (126 days total) (F0

Relative





3%

Uterus weight (percent change compared to control)

males and females), and



0

3640

0, 2.5% (0, 3640 mg/kg-day) in
utero + lactation, and then in the

Absolute





-4%

diet through a 7 day mating









period at 74±10 days old (Fl
females were allowed to deliver

Relative





-4%

litters)









^Statistically significant (p < 0.05) based on analysis of data by study authors.
Percent change compared to control = treated value - control value x 100

control value

1

17


-------
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

31

32

33

34

35

36

37

DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

EXAMPLE 4 - Evidence Integration

The example below demonstrates the integration of evidence from epidemiology studies in order to
draw conclusions about the hazards associated with chemical exposure to humans¦, as described in the
"Evaluating the Overall Evidence of Each Effect" section of the draft Handbook for IRIS Assessment
Development.

Example of Synthesis of Epidemiology Studies Evaluating Associations with
Lymphohematopoietic Cancers in Formaldehyde-Exposed Populations

In subsequent sections, the evidence of an association for each cancer-subtype in relation to
formaldehyde exposure was evaluated using a weight-of-evidence approach as outlined in the U.S.
EPA's Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a), and described in general terms in
the IRIS preamble. Causal considerations follow from the Bradford-Hill (1965) aspects of causality
and include consistency, strength of association, specificity, temporality, evidence of an exposure-
response relationship, and biological plausibility. Potential sources of bias were also considered,
including selection bias, information bias, and confounding.

This following example, currently under development, includes a draft evaluation of
evidence for one of the cancer-subtypes under consideration in the draft IRIS assessment of
Formaldehyde.

1.3.1.1.1. Hodgkin Lymphoma

Hodgkin lymphoma is a specific type of lymphohematopoietic cancer originating from white
blood cells. Historically, the diagnosis of Hodgkin lymphoma (previously called Hodgkin's disease)
used in epidemiologic studies has been ascertained from death certificates according to the version
of the International Classification of Diseases (ICD) in effect at the time of study subjects' deaths
[i.e., ICD-8 and ICD-9: Code 201 (WHO, 1967; 1977)].

Epidemiologic evidence

Evidence describing the association between formaldehyde exposure and the specific risk of
Hodgkin lymphoma was available from 13 epidemiologic studies - one case-control study (Gerin et
al., 1989) and 12 cohort studies (Beane Freeman etal., 2009; Pinkerton etal., 2004; Coggon etal.,
2003; Andjelkovich et al., 1995; Hansen and Olsen, 1995; Hall et al., 1991; Hayes et al., 1990;
Matanoski, 1989; Robinson etal., 1987; Stroup etal., 1986; Walrath and Fraumeni, 1984; 1983b).
Study details are provided in the evidence table for Hodgkin lymphoma (Table 4-1).

Causal Evaluation

The evidence of an association was evaluated using a weight-of-evidence approach as
outlined in the U.S. EPA's Guidelines for Carcinogen Risk Assessment (U.S. EPA, 2005a). The
epidemiologic data on Hodgkin lymphoma provided the strongest evidence regarding causation

18


-------
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

31

32

33

34

35

36

37

38

DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

with respect to two particular considerations: 1) the strong evidence of an exposure-response
relationship observed in the single largest cohort study; and 2) the inconsistent pattern of risks
across studies - many of which had fewer than 5 exposed cases.

Conclusion

Conclusion not available until draft is completed.

Consistency of the observed association

The results of the 13 studies were not consistent The study of the largest cohort of
formaldehyde-exposed workers (Beane Freeman et al., 2009) reported an elevated risk of dying
from Hodgkin lymphoma for the cohort as a whole (SMR=1.42; 95% CI: 0.96-2.1) and a pronounced
increase in risk among those workers with the highest peak formaldehyde exposures (RR=3.96;
95% CI: 1.31-12.02). However, the results of the other 12 studies were more consistent, with the
absence of an effect of formaldehyde exposure on the risk of developing and dying from Hodgkin
lymphoma.

Compared with other lymphohematopoietic cancers, the survival rate for Hodgkin
lymphoma is relatively high and mortality is rare. This rarity results in very low statistical power
and may have contributed to the apparently discordant results. Aside from the Beane Freeman et
al. (2009) study which reported 25 deaths from Hodgkin lymphoma, only two other cohort studies
observed more than five deaths from Hodgkin lymphoma, Coggon et al. (2003), which reported 6
observed deaths against 8.5 expected deaths, and Hansen and Olsen (1995), which reported 12
observed deaths against 12.2 expected deaths. The case-control study (Gerin etal., 1989) observed
only 8 cases of Hodgkin lymphoma and did not report an elevated risk associated with working in
formaldehyde-exposed jobs.

The study results presented in Table 4-1 (by publication date) detail all of the reported
associations between exposures to formaldehyde and the risks of developing and dying from
lymphatic leukemia. Results are plotted in Figure 4-1.

Strength of the observed association

Summary effect estimates for the association between formaldehyde exposure and Hodgkin
lymphoma were highly variable and the risk of developing or dying from Hodgkin lymphoma were
predominantly less than one (unity) and ranged from zero to 3.33.

While the summary effect estimate from the study by Beane Freeman et al. (2009) was
RR=1.42 (95% CI: 0.96, 2.10), the strength of the association was substantially stronger among
those workers exposed to the highest peak levels (RR=3.96). Beane Freeman etal. (2009) further
showed plots presenting the RR from the internal analyses for each endpoint and for each year of
follow-up. The association of Hodgkin lymphoma with formaldehyde exposure is not only seen for
the complete 2004 follow-up when the average length of follow-up was 42 years, but throughout

19


-------
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

31

32

33

34

35

36

37

38

DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

the cohort experience (see Beane Freeman et al., 2009; Figure 1). These plots show that during the
1970's and 1980's, the RR«8 and remained elevated at about RR = 4 through the end of follow-up in
2004.

Specificity of the observed association

Specificity refers to an increased inference of causality if a single cause is associated with a
single effect or disease (Hill, 1965). An example of specificity is seen with respect to a specific
infectious disease caused by a specific virus. Based on an understanding that many agents cause
cancer at multiple sites (e.g., tobacco), specificity is generally not considered to be a necessary
condition for making causal inferences regarding cancer.

Nonetheless, the specificity of the diagnoses of cancer is important - especially for
lymphohematopoietic cancers, which are heterogeneous in nature and arise from different cell
lines. This point concerning specificity was not discussed in Hill's paper on causality (1965). In an
epidemiology study, increasing the specificity of a diagnosis is likely to increase the precision of an
observed association because the exposure, if it is causally associated, is relevant to the cases under
study (e.g., cases are not diluted with diagnoses that are not relevant to the exposure). In this
section, only the specific diagnosis of Hodgkin lymphoma was considered. The most specific level
of Hodgkin lymphoma diagnosis that is commonly reported across the epidemiologic literature has
been based on the first three digits of the Eighth or Ninth Revision of the ICD code (i.e., Hodgkin's
disease ICD-8/9: 201).

Temporal relationship of the observed association

Only one study (Beane Freeman et al., 2009) reported on analyses of the temporal
relationship showing that risks were highest 15-25 years since first formaldehyde exposure. Such
a pattern is consistent with the expected time-course of disease and mortality following exposure to
formaldehyde; however, this finding with respect to formaldehyde is without corroboration for
Hodgkin lymphoma.

Exposure-response relationship

An exposure-response relationship showing increasing effects associated with greater
exposure strongly suggests cause and effect, especially when such relationships are also observed
for duration of exposure (USEPA, 2005a: p. 2-14). None of the studies evaluated the effect of
duration of formaldehyde exposure on the mortality risk of Hodgkin lymphoma. There were only
two studies that evaluated any form of exposure-response for increasing measures of formaldehyde
exposure (Coggon et al., 2003; Beane Freeman et al., 2009). Coggon et al. (2003) reported a lower
risk of dying from Hodgkin lymphoma among 'highly' exposed workers based on a single death.

Beane Freeman etal. (2009) reported a clear exposure-response relationship between
increasing levels of peak formaldehyde and increased risk of dying from Hodgkin lymphoma among

20


-------
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

31

32

33

34

35

36

37

38

DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

exposed workers (p=0.01). Compared to exposed workers in the lowest exposure category of peak
exposure, those in the middle category were at more than threefold higher risk (RR=3.30; 95% CI:
1.04,10.50) while those workers in the highest category were at fourfold higher risk (RR=3.96;
95% CI: 1.31,12.02). Beane Freeman et al. (2009) also reported an exposure-response relationship
between increasing levels of average formaldehyde intensity and increased risk of dying from
Hodgkin lymphoma among exposed workers (p=0.05).

Biologic plausibility

The reader is referred to the section on mode of action for lymphohematopoietic cancers.

Potential impact of selection bias, information bias, confounding bias, and chance

Selection bias is an unlikely bias in the epidemiologic studies of Hodgkin lymphoma as the
case-control study evaluated exposure status without regard to outcome status and had a
participation level of 83% and each of the cohort studies included at least 72% of eligible
participants and lost fewer than 9% of participants over the course of mortality follow-up.

The healthy-worker effect and the healthy-worker survivor effect could obscure a truly
larger effect of formaldehyde exposure in analyses based on "external" comparisons with mortality
in the general population (Walrath and Fraumeni 1983b; 1984; Stroup et al. 1986; Matanoski 1989;
Hayes et al., 1990; Hall et al., 1991; Robinson et al., 1987; Andjelkovich et al., 1995; Hansen and
Olsen, 1995; Coggon et al., 2003; Pinkerton et al., 2004; Beane Freeman et al., 2009), but would not
influence analyses using "internal" or matched comparison groups (Gerin et al., 1989; Beane
Freeman et al., 2009).

Information bias is unlikely to have resulted in bias away from the null; however, random
measurement error or non-differential misclassification is almost certain to have resulted in some
bias toward the null among these studies of Hodgkin lymphoma.

Chemical exposures that have not been independently associated with Hodgkin lymphoma
are not expected to confound results. The main support for a suggestive association of
formaldehyde exposure with increased risk of Hodgkin lymphoma is from the results for peak
exposures reported by Beane Freeman et al. (2009) who specifically examined the potential for
confounding from 11 substances including benzene and found that controlling for these exposures
did not meaningfully change the results. This provides evidence against potential confounding by
these co-exposures. There does not appear to be any evidence of confounding that would provide
an alternative explanation for the observed association of formaldehyde exposure with increased
risk of Hodgkin lymphoma reported by Beane Freeman et al. (2009).

The reported results for the risk of Hodgkin lymphoma associated with exposure to
formaldehyde were inconsistent There were 12 small studies, each with 12 or fewer exposed cases
and only 44 exposed cases among them, showing a consistent pattern of risks across studies
indicating a lack of an association. However, the single largest study in terms of study population

21


-------
1

2

3

4

5

6

7

8

9

10

11

DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

and number of formaldehyde exposed cases (n=25) showed increased risks of Hodgkin lymphoma
mortality as a cohort compared to the general population (SMR=1.42; 95% CI: 0.96, 2.10) and
statistically significant increased risks with increasing levels of peak exposure (p-trend among
exposed workers =0.01). The evidence of an association with peak exposures reported by Beane
Freeman et al. (2009) suggests an association whose risk increases with greater exposure.
However, there was only one statistically robust observation of an exposure-response relationship
showing increased risks with peak exposures and this finding is tempered by the lack of
corroborative epidemiologic evidence.

Conclusion

Conclusion not available until draft is completed.

22


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE
All Studies Reporting Hodgkin Lymphoma Risk Estimates

Individual-level exposure assessment

Beane Freeman

i

Industrial Workers


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE
Table 4-1. Epidemiologic studies of formaldehyde exposure and risk of Hodgkin lymphoma

Study

Reference: Beane Freeman et al. (2009)
with supplemental online tables.

Population: 25,619 workers employed
at 10 formaldehyde using or
formaldehyde producing plants in the
U.S. followed from either the plant start-
up or first employment through 2004.
Deaths were identified from the
National Death Index with remainder
assumed to be living. Vital status was
97.4% complete and only 2.6% lost to
follow-up.

Outcome definition: Death certificates
used to determine underlying cause of
death from Hodgkin disease (ICD-8:
201).

Design: Prospective cohort mortality
study with external and internal
comparison groups.

Analysis: RRs estimated using Poisson
regression stratified by calendar year,
age, sex, and race; adjusted for pay
category compared to workers in lowest
exposed category. Lagged exposures
were evaluated to account for cancer
latency.

SMRs calculated using sex, age, race, and
calendar-year-specific U.S. mortality
rates.

Related studies:

Blair et al. (1986)

Hauptmann et al. (2003)

Multiple exposure metrics including peak,
average, and cumulative exposures were
evaluated using categorical and
continuous data.

Go-exposures: Exposures to 11 other
compounds were identified and evaluated
as potential confounders.

Results: Effect estimate (95% CI) [# of
cases]

[2]
[6]
[8]

[11]

[2]

[10]

P]
[6]

[2]

[14]

m

[4]

Duration of exposure

No evidence of association (data not shown).

Exposures

Exposure assessment: Individual-level
exposure estimates based on job titles,
tasks, visits to plants by study industrial
hygienists, and monitoring data from
1966 through 1980.

Median time weighted average (over 8
hours) =0.3ppm (range 0.01—4.3).

Median cumulative exposure=0.6 ppm-
years (range 0—107.4).

Internal comparisons:

Peak exposure
1994 Follow-up:

Highest peak RR=3.30 (0.98-11.10)
(p-trend=0.04)
2004 Follow-up:

Peak exposure

Level 1 RR=0.67 (0.12-3.6)
Level 2 RR=1.00 (Ref. value)
Level 3 RR=3.30 (1.04-10.50)
Level 4 RR=3.96 (1.31-12.02)
p-trend (exposed) = 0.01;
p-trend (all) = 0.004

Average intensity

Level 1 RR=0.53 (0.11-2.66)
Level 2 RR=1.00 (Ref. value)
Level 3 RR=2.48 (0.84-7.32)
Level 4 RR=1.61 (0.73-3.39)
p-trend (exposed) = 0.05;
p-trend (all) = 0.03

Cumulative exposure

Level 1 RR=0.42 (0.09-2.05)
Level 2 RR=1.00 (Ref. value)
Level 3 RR=1.71 (0.66-4.38)
Level 4 RR=1.30 (0.40-4.19)
p-trend (exposed) = 0.08;
p-trend (all) = 0.06

Duration and timing: Exposure period
from <1946—1980. Median length of
follow-up: 42 years. Duration and timing
since first exposure were evaluated.

Variation in exposure:

For all variations in exposure:

Level 1 (unexposed)

Peak exposure:

Level 2 (>0 to <2.0 ppm)

Level 3 (2.0 to <4.0 ppm)

Level 4 (>4.0 ppm)

Average intensity:

Level 2 (>0 to <0.5ppm)

Level 3 (0.5 to <1.0 ppm)

Level 4 (>1.0 ppm)

Cumulative exposure:

Level 2 (>0 to <1.5 ppm-yrs)

Level 3 (1.5 to <5.5 ppm-yrs)

Level 4 (>5.5 ppm-yrs)

Time since first exposure

>0—15 yrs RR=1.00 (Ref. value)
>15-25 yrs RR=1.54 (0.42-5.62)
>25-35 yrs RR<1.54
>35 yrs RR<1.54

External comparisons:

SAlRlJnexposed

= 0.70 (0.17-2.80)
SMRE^posed = 1.42 (0.96-2.10)

[2]
[25]

24


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Table 4-1. Epidemiologic studies of formaldehyde exposure and risk of Hodgkin lymphoma

Study

Exposures

Results: Effect estimate (95% GI) [# of
cases]

Reference: Pinkerton et al. (2004)

Population: 11,039 workers in 3 U.S.
garment plants exposed for at least 3
months. Women comprised 81.7% of
the cohort. Vital status was followed
through 1998 with 98.3% completion
and only 1.7% lost to follow-up.

Outcome definition: Death certificates
used to determine both the underlying
cause of death (UCOD) as well as all
contributing multiple causes of death
(MCOD) from Hodgkin's disease (ICD:
201).

Design: Prospective cohort mortality
study with external comparison group.

Analysis: SMRs calculated using sex,
age, race, and calendar-year-specific U.S.
mortality rates. Results presented here
are UCOD unless otherwise noted.

Related studies:

Stayner et al. (1985)

Stayner et al. (1988)

Exposure assessment: Individual-level
exposure estimates for 549 randomly
selected workers during 1981 and 1984.
Geometric 8-hr time-weighted average
exposures ranged from 0.09—0.20 ppm.
Overall geometric mean concentration of
formaldehyde was 0.15 ppm, (GSD 1.90
ppm). Area measures showed constant
levels without peaks. Historically earlier
exposures may have been substantially
higher.

Duration and timing: Exposure period
from 1955—1983. Median duration of
exposure was 3.3 years. More than 40%
exposures <1963. Median time since first
exposure was 31.7 years. Duration and
timing since first exposure were
evaluated.

Variation in exposure: Not evaluated

Go-exposures: Study population
specifically selected because industrial
hygiene surveys at the plants did not
identify any chemical exposures other
than formaldehyde that were likely to
influence findings.

External comparisons:

SMR=0.55 (0.07-1.98) [2]

Reference: Coggon et al. (2003)

Population: 14,014 British men
employed in 6 chemical industry
factories which produced formaldehyde.
Cohort mortality followed from 1941
through 2000. Vital status was 98.9%
complete and only 1.1% lost to follow-
up.

Outcome definition: Death certificates
used to determine cause of deaths from
Hodgkin's disease (TCD-9: 201).

Design: Cohort mortality study with
external comparison group.

Analysis: SMRs based on English and
Welsh age- and calendar-year-specific
mortality rates.

Related studies:

Acheson et al. (1984)

Gardner et al. (1993)

Exposure assessment: Exposure
assessment based on data abstracted from
company records. Jobs categorized as
background, low, moderate, high, or
unknown levels.

Duration and timing: Occupational
exposure during 1941—1982. Duration
and timing since first exposure were not
evaluated.

Variation in exposure:

Time weighted average exposure
Level 1 (low)

Level 2 (moderate)

Level 3 (high)

Go-exposures: Not evaluated. Potential
low-level exposure to styrene, ethylene
oxide, epichlorhydrin, solvents, asbestos,
chromium salts, and cadmium.

External comparisons:

SMR=0.70 (0.26-1.53) [6]

Within-study external comparisons:

Worked in 'High' exposure jobs

SMR=0.36 (0.01-2.01) [1]

25


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Table 4-1. Epidemiologic studies of formaldehyde exposure and risk of Hodgkin lymphoma

Study

Exposures

Results: Effect estimate (95% GI) [# of
cases]

Reference: Andjelkovich et al. (1995)

Cohort mortality study of 3,929
automotive industry iron foundry
workers exposed from 1960—1987 and
followed through 1989. SMRs calculated
using sex-, age-, race-, and calendar-year-
specific U.S. mortality rates.

Exposure assessment based on review of
work histories by an industrial hygienist.

External comparisons:

SMRlJnexposed = 0.70 (0.01—3.88)
SMRsxposed = 0.72 (0.01—4.00)

[i]
[i]

Reference: Hansen and Olsen (1995)

Population: 2,041 men with cancer who
were diagnosed during 1970—1984 and
whose longest work experience occurred
at least 10 years before cancer diagnosis.
Identified from the Danish Cancer
Registry and matched with the Danish
Supplementary Pension Fund.
Ascertainment considered complete.
Pension record available for 72% of
cancer cases.

Outcome definition: Hodgkin's disease
(TCD-7: 201) listed on Danish Cancer
Registry file.

Design: Proportionate incidence study
with external comparison group.

Exposure assessment: Individual
occupational histories including industry
and job tide established through company
tax records to the national Danish
Product Register.

Subject were considered to be exposed to
formaldehyde if: 1) they had worked in an
industry known to use more than 1 kg
formaldehyde per employee per year; and
2) subjects longest single work experience
(job) in that industry since 1964 was >10
years prior to cancer diagnosis

All subjects were stratified based on job
tide as either low exposure (white-collar
worker), above background exposure
(blue-collar worker), or unknown (job
tide unavailable).

External comparisons:

Overall (exposure to formaldehyde >10
prior to cancer diagnosis)

SPIR=1.0 (0.5-1.7)

years
[12]

Analysis: Standardized proportionate
incidence ratio calculated as the
proportion of cases for a given cancer in
formaldehyde-associated companies
relative to the proportion of cases for
the same cancer among all employees in
Denmark. Adjusted for age and calendar
time.

Duration and timing: Exposure period
not stated. Based on date of diagnosis
during 1970—1984, and the requirement
of exposure more than 10 years prior to
diagnosis, the approximate period was
1960-1974.

Variation in exposure: Not evaluated.
Go-exposures: Not evaluated.





Reference: Hall et al. (1991)

Cohort mortality study of 4,512
pathologists from the Royal College of
Pathologists and the Pathological Society
of Great Britain from 1974—1987. Vital
status obtained from the census, a
national health registry, and other
sources. SMRs developed from the
English and Welsh populations.

Presumed exposure to formaldehyde
tissue fixative.

External comparisons:

SMR= 1.21 (0.03-6.71)

[1]

Related studies:

Harrington and Shannon (1975)
Harrington and Oaks (1984)







26


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Table 4-1. Epidemiologic studies of formaldehyde exposure and risk of Hodgkin lymphoma

Study

Exposures

Results: Effect estimate (95% GI) [# of
cases]

Reference: Matanoski (1989)

Population: 3,644 deceased U.S. male
pathologists, derived from membership
rolls of the American Association of
Pathologists and Bacteriologists (1900-),
the American Society for Experimental
Pathology (1913-), and the American
Medical Association (1912-1950).
Mortality was followed through 1978.
Death certificates obtained for 94% of
potential study subjects (n=3,425), 3%
from obituary notices (n=101) and 3%
presumed dead (n=118).

Outcome definition: Death certificates
and obituary notices used to determine
cause of death from Hodgkin's disease
(TCD-8: 201).

Design: Prospective mortality cohort
study with two external comparison
groups. The first comparison group was
the U.S. male population. The second
comparison group was comprised of
members of a professional society of
psychiatrists.

Analysis: SMRs calculated using sex,
race, age, and calendar-year-expected
deaths from the U.S. population and
psychiatrists.

Exposure assessment: Presumed
exposure to formaldehyde tissue fixative.

Duration and timing: Occupational
exposure preceding death during 1900—
1978. Duration and timing since first
exposure were not evaluated.

Variation in exposure: Not evaluated.

Go-exposures: Not evaluated.

External comparisons:

Compared to the U.S. male population
SMR=0.36 (0.04-1.31) [2]

Compared to the psvchiatrists

SMR=0.34 (0.06-1.12)1 [2]

fNote: EPA derived CIs using the Mid-P
Method (See Rothman and Boice, 1979)

Reference: Hayes et al. (1990)

Population: 4,046 deceased U.S. male
embalmers and funeral directors, derived
from licensing boards and funeral
director associations in 32 states and the
District of Columbia who died during
1975—1985. Death certificates obtained
for 79% of potential study subjects
(n=6,651) with vital status unknown for
21%.

Outcome definition: Death certificates
and licensing boards used to determine
cause of death from Hodgkin's disease
(TCD-8: 201).

Design: Proportionate mortality cohort
study with external comparison group.

Analysis: PMRs calculated using sex,
race, age, and calendar-year-expected
deaths from the U.S. population.

Exposure assessment: Presumed
exposure to formaldehyde tissue fixative.
Exposure based on occupation which was
confirmed on death certificate. Authors
subsequendy measured personal
embalming exposures ranging from 0.98
ppm (high ventilation) to 3.99 ppm (low
ventilation) with peaks up to 20 ppm.

Authors state that major exposures are to
formaldehyde and possibly gluteraldehyde
and phenol.

Duration and timing: Occupational
exposure preceding death during 1975—
1985. Of 115 deaths from
lymphohematopoietic cancer, 66 (57%)
were aged 60—74 years. Duration and
timing since first exposure were not
evaluated.

Variation in exposure: Not evaluated.
Go-exposures: Not evaluated.

External comparisons:

PMR=0.72 (0.15-2.10) [3]

27


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Table 4-1. Epidemiologic studies of formaldehyde exposure and risk of Hodgkin lymphoma

Study

Exposures

Results: Effect estimate (95% CI) [# of
cases]

Reference: Gerin et al. (1989)

Population: Male residents of Montreal,
Canada aged 35—70 years. 4,510 eligible
incident cancer cases were identified
during 1979—1985 from 19 major area
hospitals which report to the Quebec
Tumor Registry over 97% of all cancer
diagnoses from the Montreal area.
Interviews and questionnaires completed
for 3,726 subjects (83% of eligible cases).
18% of interviews were completed by
next-of-kin.

Outcome definition: Histologically
confirmed diagnosis of Hodgkin's
lymphoma (ICD: 201)

Design: Population-based case-control
study of 53 formaldehyde exposed men
with Hodgkin lymphoma. Cases were
compared with two groups; first, against
other cancer cases excluding those
diagnosed with lung cancer (n=2,599),
and second against 533 male population
controls selected from electoral list in
the Montreal area.

Analysis: ORs calculated by levels of a
composite exposure index using logistic
regression controlling for age, ethnic
group, socio-economic status, smoking,
and dirtiness of jobs held (white vs. blue
collar).

Related studies:

Siemiatycki et al. (1987)

Exposure assessment: Individual-level
exposure estimates developed based on a
complete and detailed occupational
history ascertained by interviewers using a
standardized questionnaire. A team of
chemists and hygienists translated each
job into a list of potential formaldehyde
exposures based on their confidence
level, the frequency of exposure, and the
duration of exposure.

Duration and timing: Exposure period
based on occupational histories prior to
cancer diagnosis. Duration of exposure
was evaluated.

Variation in exposure: For cancer sites
with fewer than 30 cases exposed to
formaldehyde, results for the exposure
subgroups were not shown.

Co-exposures: Additional occupational
and non-occupational potential
confounders were included when the
estimated exposure-disease OR changed
by more than 10%.

External comparisons:

Compared to other cancers

OR=0.5 (0.2-1.2) [8]

Compared to population controls

OR=0.5 (0.2-1.4) [8]

28


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Table 4-1. Epidemiologic studies of formaldehyde exposure and risk of Hodgkin lymphoma

Study

Exposures

Results: Effect estimate (95% GI) [# of
cases]

Reference: Robinson et al. (1987)

Population: 2,283 plywood mill workers
employed at least one year during 1945—
1955 followed for mortality until 1977
with vital status for 98% and death
certificates for 97% of deceased.

Outcome definition: Death certificates
used to determine underlying cause of
death from Hodgkin's disease as coded
by trained nosologist using ICD-7:201.

Design: Prospective cohort mortality
study with external comparison group.
A subcohort of 818 men co-exposed to
formaldehyde and pentachlorophenol
were also evaluated.

Analysis: SMRs calculated using sex,
age, race, and calendar-year-specific U.S.
mortality rates.

Exposure assessment: Presumed
exposure to formaldehyde-based glues
used to manufacture and patch plywood.
Sub-cohort of 818 men co-exposed to
formaldehyde and pentachlorophenol
worked for one year or more in the
relevant exposure categories of veneer
pressing and drying, glue mixing, veneer
and panel gluing and patching.

Duration and timing: Exposures during
1945—1955. Duration and timing since
first exposure were not evaluated.

Variation in exposure:

Duration of exposure

Latency (time since first exposure)

Go-exposures: Pentachlorophenol

External comparisons:

Whole cohort of mill workers (n=2.283^
SMR= 1.11(0.20—3.50) [2]

Sub-cohort of hiphlv exposed workers (n=818")

SMR=3.33(0.59—10.49) [2]

Reference: Stroup et al. (1986)

Population: 2,239 white male members
of the American Association of
Anatomists from 1888—1969 who died
during 1925—1979. Death certificates
obtained for 91% with 9% lost to
follow-up.

Outcome definition: Hodgkin's disease
(ICD-8: 201) listed as cause of death on
death certificates.

Design: Cohort mortality study with
external comparison group.

Analysis: SMRs calculated using sex,
race, age, and calendar-year-expected
number of deaths from the U.S.
population.

Exposure assessment: Presumed
exposure to formaldehyde tissue fixative.

Duration and timing: Occupational
exposure preceding death during 1925—
1979. Median birth year was 1912. By
1979, 33% of anatomists had died.
Duration and timing since first exposure
were not evaluated.

Variation in exposure: Not evaluated.
Go-exposures: Not evaluated.

External comparisons:

SMR= 0 (0-2.0) [0]

29


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Table 4-1. Epidemiologic studies of formaldehyde exposure and risk of Hodgkin lymphoma

Study

Exposures

Results: Effect estimate (95% GI) [# of
cases]

Reference: Walrath and Fraumeni

(1984)

Population: 1,007 deceased white male
embalmers from the California Bureau
of Funeral Directing and Embalming
who died during 1925—1980. Death
certificates obtained for all.

Outcome definition: Hodgkin's
disease (ICD-8: 201) listed as cause of
death on death certificates.

Design: Proportionate mortality cohort
study with external comparison group.

Analysis: PMRs calculated using sex,
race, age and calendar-year-expected
number of deaths from the U.S.
population.

Exposure assessment: Presumed
exposure to formaldehyde tissue fixative.

Duration and timing: Occupational
exposure preceding death during 1916—
1978. Birth year ranged from 1847—1959.
Median age of death was 62 years. Most
deaths were among embalmers with
active licenses. Duration and timing since
first exposure were not evaluated.

Variation in exposure: Not evaluated.

Go-exposures: Not evaluated.

External comparisons:

Observed: 0 Hodgkin's disease deaths
Expected: 2.5 Hodgkin's disease deaths

PMR= 0 (0-1.20)f [0]

fNote: EPA derived CIs using the Mid-P
Method (See Rothman and Boice, 1979)

Reference: Walrath and Fraumeni

(1983b)

Population: 1,132 deceased white male
embalmers licensed to practice during
1902—1980 in New York who died
during 1925—1980 identified from
registration files. Death certificates
obtained for 75% of potential study
subjects (n=l,678).

Outcome definition: Hodgkin's disease
(ICD-8: 201) listed as cause of death on
death certificates.

Design: Proportionate mortality cohort
study with external comparison group.

Analysis: PMRs calculated using sex,
race, age, and calendar-year-expected
numbers of deaths from the U.S.
population.

Exposure assessment: Presumed
exposure to formaldehyde tissue fixative.

Duration and timing:

Occupational exposure preceding death
during 1902—1980. Median year of birth
was 1901. Median year of initial license
was 1931. Median age at death was 1968.
Expected median duration of exposure
was 37 years. Duration and timing since
first exposure were not evaluated.

Variation in exposure: Not evaluated.

Go-exposures: Not evaluated.

External comparisons:

Observed: 2 Hodgkin's disease deaths
Expected: 2.3 Hodgkin's disease deaths

PMR= 0.87 (0.15-2.87)1 [7|

fNote: EPA derived CIs using the Mid-P
Method (See Rothman and Boice, 1979)

1

30


-------
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

31

32

33

34

35

36

37

DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

EXAMPLE 5 - Selecting Studies for Derivation of Toxicity Values

The example below demonstrates the selection of studies for derivation of toxicity values from a group
of studies identified and evaluated as part of the hazard identification, as described in the "Dose-
Response Analysis" section of the draft Handbook for IRIS Assessment Development.

Summary of issues covered:

•	Overall description of the most suitable studies, given availability of a broader range of study
designs.

•	Summary of studies judged less suitable (or unsuitable); dismiss if possible or necessary
(e.g., non-developmental acute or short-term studies; studies with only one relatively high
dose).

Draft assessment text:

In Section 1.2.1, reproductive toxicities in male and female rodents were identified as
hazards and liver and kidney toxicities were identified as potential hazards of dipentyl phthalate
(DPP) exposure by the oral route. Studies within each effect category were evaluated using general
study quality characteristics (as discussed in Section 6 of the Preamble) to help inform the selection
of studies from which to derive toxicity values. Rationales for selecting studies and effects to
represent each of these hazards are summarized below. The first objective was to derive an overall
reference dose (RfD) for DPP. The second objective was to derive organ/system-specific reference
values for DPP for each of the effects identified as hazards, to facilitate aggregating effects across
phthalates when exposure is to a phthalate mixture.

A number of DPP studies supporting hazard identification were not considered for dose-
response assessment, due to study designs that were less relevant for developing reference values
for lifetime exposure and/or lacked evaluation of dose-response relationships (e.g., one dose level).
These studies mainly comprised non-developmental studies with short-term and acute exposures
(<10 daily doses) and evaluation of effects for <2 days following the last exposure, and included
mechanistic studies. Most were conducted at relatively high doses (>2,000 mg/kg-day), generally
using a single dose level, thus providing little information about dose-response relationships.
The remaining DPP studies were reproductive or developmental studies. The reproductive study of
Heindel et al. (1989; NTP, 1985), while including three dose levels (Task 2-continuous breeding
phase), was not considered for dose-response assessment because a very high response (90%
decrease in number of live pups per litter) was observed at the lowest dose tested, thus yielding
little information about the shape of the dose response. The rat study of Liu et al. (2005) was not
considered for dose-response assessment because it included only one dose level and because of
the availability of other rat studies that used multiple, lower-dose levels and assessed a number of
reproductive/developmental outcomes, including offspring mortality (Hannas etal., 2011;
Howdeshell etal., 2008).

31


-------
1

2

3

4

5

6

7

8

9

10

11

12

13

14

DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

The studies selected for dose-response assessment consisted of two gestational exposure
studies evaluating endpoints in rats exposed to DPP via gavage (Hannas et al., 2011; Howdeshell et
al., 2008). Male reproductive toxicity was demonstrated in both studies. Effects observed included
outcomes consistent with the "phthalate syndrome"—decreased fetal testicular testosterone
production, decreased anogenital distance (AGD) in male pups, and retention of nipples/areolae in
male pups after gestational exposure (Hannas etal., 2011; Howdeshell etal., 2008). Female
reproductive toxicity was also demonstrated following gestational exposure to DPP by increased
fetal/neonatal mortality (Hannas et al., 2011; Howdeshell et al., 2008). The Heindel et al. (1989;
NTP, 1985) 29-week mouse study, while employing only one dose level (following task 3 [crossover
mating phase] during week 19 to terminus of study in F0 male and female mice), represents the
only evidence available for informing liver and kidney hazard following oral exposure to DPP.

Thus, alterations in liver and kidney weights that were observed in adult mice exposed to DPP for
up to 29 weeks were considered for dose-response assessment (Heindel etal., 1989; NTP, 1985).

32


-------
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

31

32

33

34

35

36

37

DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

EXAMPLE 6 - Dose-Response Modeling Output

The example below demonstrates the presentation of dose-response modeling results and output as it
would appear in the supplemental information of IRIS assessmenst, as described in the "Dose-Response
Analysis" section of the draft Handbook for IRIS Assessment Development.

Benchmark Dose Modeling Summary

This appendix provides technical detail on dose-response evaluation and determination of
points of departure (POD) for relevant toxicological endpoints. The endpoints were modeled using
the U.S. EPA's Benchmark Dose Software (BMDS, version 2.2). The following sections describe
common practices used in evaluating the model fit and selecting the appropriate model for each
endpoint, as outlined in the Benchmark Dose Technical Guidance Document (U.S. EPA, 2012). In
some cases it may be appropriate to use alternative methods, based on statistical judgment;
exceptions are noted as necessary in the summary of the modeling results.

Noncancer Endpoints
Evaluation of Model Fit

For each dichotomous endpoint (see Table 6-1), BMDS dichotomous models were fitted to
the data using the maximum likelihood method. The following parameter restrictions were applied,
unless otherwise noted: for the log-logistic model, restrict slope > 1; for the gamma and Weibull
models, restrict power > 1; for the multistage models, restrict beta's > 0. Each model was tested for
goodness-of-fit using a chi-square goodness-of-fit test (x2 p-value < 0.10 indicates lack of fit). Other
factors were also used to assess model fit, such as scaled residuals, visual fit, and adequacy of fit in
the low-dose region and in the vicinity of the benchmark response (BMR).

For each continuous endpoint, BMDS continuous models were fitted to the data using the
maximum likelihood method. The following parameter restrictions were applied, unless otherwise
noted: for the polynomial models, restrict the coefficients bl and higher to be nonnegative or
nonpositive if the direction of the adverse effect is upward or downward, respectively; for the Hill,
power, and exponential models, restrict power > 1. Model fit was assessed by a series of tests as
follows. For each model, first the homogeneity of the variances was tested using a likelihood ratio
test (BMDS Test 2). If Test 2 was not rejected (x2 p-value > 0.10), the model was fitted to the data
assuming constant variance. If Test 2 was rejected (x2 p-value < 0.10), the variance was modeled as
a power function of the mean, and the variance model was tested for adequacy of fit using a
likelihood ratio test (BMDS Test 3). For fitting models using either constant variance or modeled
variance, models for the mean response were tested for adequacy of fit using a likelihood ratio test
(BMDS Test 4, with x2 p-value < 0.10 indicating inadequate fit). Other factors were also used to
assess the model fit, such as scaled residuals, visual fit, and adequacy of fit in the low-dose region
and in the vicinity of the BMR.

33


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

1	Model Selection

2	For each endpoint, the BMDL estimate (95% lower confidence limit on the benchmark dose

3	[BMD], as estimated by the profile likelihood method) and Akaike information criterion (AIC) value

4	were used to select a best-fit model from among the models exhibiting adequate fit If the BMDL

5	estimates were "sufficiently close," that is, differed by at most threefold, the model selected was the

6	one that yielded the lowest AIC value. If the BMDL estimates were not sufficiently close, the lowest

7	BMDL was selected as the POD.

8

9	Table 6-1. Noncancer endpoints selected for dose-response modeling for 1,2,4-

10	trimethylbenzene

11

Species (generation) / Sex
Endpoint

Doses and Effect Data

Korsak (1996)

Rat (Wistar) / Male

Dose (mg/kg-d)

0

123

492

1230

CNS: Pawlick (seconds)

No. of animals
Mean ± SD

9

15.4 ±5.8

10

18.2 ±5.7

9

27.6 ±4.6

10

30.1 ±6.1

CNS: RotoRod

Incidence /Total

0/10

1/10

2/10

4/10

12

13	Modeling Results

14	Below are tables, graphs, and BMDS output summarizing the modeling results for each

15	endpoint modeled.

16

17	Table 6-2. Summary of BMD modeling results for CNS: Pawlick in male Wistar rats exposed

18	to 1,2,4-trimethylbenzene by inhalation for 3 months (Korsak, 1996); BMR = 1 SD change

19	from the control mean

20

34


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Model3

Goodness of fit

BMD1sd
(mg/m3)

BMDL1sd
(mg/m3)

Basis for Model Selection

p-value

AIC

Exponential (M2)b
Exponential (M3)

0.01

181.65

646

512

Only exponential model 4 provided an
adequate fit, so it was selected.

Exponential (M4)

0.35

173.57

150

80.8

Exponential (M5)

NAC

174.68

200

89.7

Hill

NAC

174.68

186

88.6

Polynomial l°d
Polynomial 2°
Polynomial 3°
Power

0.02

178.58

508

380

aConstant variance models are presented (BMDS Test 2 p-value = 0.84), with the selected model in bold. Scaled
residuals for the selected model for doses 0,123, 492, and 1230 mg/kg-d were 0.36, -0.65, 0.53, and -0.19,
respectively.

bFor exponential model 4, the estimate of d was 1 (boundary). The models in this row reduced to exponential
model 2.

°X2 test had insufficient degrees of freedom.

dFor the power model, the power parameter estimate was 1 (boundary of parameter space). For the polynomial 2°
and 3° models, the b2 and b3 coefficient estimates were 0 (boundary of parameter space). The models in this row
reduced to the polynomial 1° model.

1

Exponential Model 4 with 0.95 Confidence Level

dose

2	08:28 09/01 2011

3	Figure 6-1. Plot of mean response by dose, with the fitted curve for exponential

4	model 4 with constant variance. BMR = 1 SD change from the control mean; dose

5	shown in mg/kg-day.

6

35


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

1	Exponential Model. (Version: 1.7; Date: 12/10/2009)

2

3	The form of the response function is: Model 4: Y[dose] = a * [c-(c-l) * exp{-b * dose}]

4

5	A constant variance model is fit.

6

7	Parameter Estimates

Variable

Estimate

Default Initial
Parameter Values

Inalpha

3.35713

3.3338

rho

0

0

a

14.756

14.63

b

0.00266447

0.00210148

c

2.10364

2.16029

d

1

1

8

9	Table of Data and Estimated Values of Interest

Dose

N

Obs Mean

Est Mean

Obs Std Dev

Est Std Dev

Scaled Resid

0

9

15.4

14.76

5.8

5.358

0.361

123

10

18.2

19.31

5.7

5.358

-0.653

492

9

27.6

26.65

4.6

5.358

0.531

1230

10

30.1

30.43

6.1

5.358

-0.193

10

11	Likelihoods of Interest

Model

Log(likelihood)

# Param's

AIC

A1

-82.34222

5

174.6844

A2

-81.92912

8

179.8582

A3

-82.34222

5

174.6844

R

-98.61903

2

201.2381

4

-82.78544

4

173.5709

12

13	Tests of Interest

Test

-2*log( Likelihood
Ratio)

Test df

p-value

Test 1 (Does response and/or variances differ among Dose
levels, A2 vs. R)

33.38

6

<0.0001

Test 2 (Are Variances Homogeneous, A2 vs. Al)

0.8262

3

0.8432

Test 3 (Are variances adequately modeled, A2 vs. A3)

0.8262

3

0.8432

Test 4 (Does the model for the Mean fit, A3 vs. fitted)

0.8864

1

0.3464

14

15	Benchmark Dose Computation

16

17	BMR = 1 estimated standard deviation from the control mean

18

19	BMD = 149.743

20

21	BMDL at the 95% confidence level = 80.7575

22

23

36


-------
1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Table 6-3. Summary of BMD modeling results for incidence of CNS: RotoRod in male Wistar
rats exposed to 1,2,4-trimethylbenzene by inhalation for 3 months (Korsak, 1996); BMR =
10% extra risk



Goodness of fit

BMD10

BMDL10



Model3

p-value

AIC

(mg/m3)

(mg/m3)

Basis for Model Selection

Gamma"









Of the models that provided an

Multistage 1°









adequate fit, the log-logistic model was

Multistage 2°

0.93

32.33

229

129

selected based on lowest BMDL (BMDLs

Multistage 3°









differed by more than threefold).

Weibull











Log-Logistic

0.97

32.16

194

93.9



Logistic

0.60

35.53

529

342



Probit

0.63

35.40

490

318



Log-Probit

0.58

35.40

426

233



aSelected model in bold. Scaled residuals for the selected model for doses 0,123, 492, and 1230 mg/kg-d were 0,
0.43, -0.15, and -0.09, respectively.

bFor the gamma and Weibull models, the power parameter estimates were 1 (boundary of parameter space). For
the multistage 2° and 3° models, the b2 and b3 coefficient estimates were 0 (boundary of parameter space). The
models in this row reduced to the multistage 1° model.

Log-Logistic Model with 0.95 Confidence Level

dose

08:33 09/01 2011

Figure 6-2. Plot of incidence rate by dose, with the fitted curve for the log-
logistic model. BMR = 10% extra risk; dose shown in mg/kg-day.

Log-logistic Model (Version: 2.13; Date: 10/28/2009)

The form of the probability function is: P[response] = background+(l-background)/[l+EXP(-intercept-
slope*Log(dose))]

Slope parameter is restricted as slope >= 1

37


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE
1 Parameter Estimates

Variable

Estimate

(Default) Initial
Parameter Values

Background

0

0

Intercept

-7.46289

-7.46166

Slope

1

1

2

3	Analysis of Deviance Table

Model

Log(likelihood)

# Param's

Deviance

Test d.f.

p-value

Full model

-14.985

4







Fitted model

-15.0832

1

0.196433

3

0.9782

Reduced model

-18.5491

1

7.12817

3

0.06792

4

5	AIC: 32.1664

6

7	Goodness of Fit Table

Dose

Est. Prob.

Expected

Observed

Size

Scaled Resid

0

0.0000

0.000

0

10

0.000

123

0.0659

0.659

1

10

0.434

492

0.2202

2.202

2

10

-0.154

1230

0.4138

4.138

4

10

-0.089

8

9	ChiA2 = 0.22 d.f. = 3 P-value = 0.9743

10

11	Benchmark Dose Computation

12

13	BMR = 10% extra risk

14

15	BMD = 193.575

16

17	BMDL at the 95% confidence level = 93.947

18

19

20	Cancer Endpoints

21	For each endpoint (see Table 6-4), multistage cancer models, with coefficients restricted to

22	be non-negative (beta's > 0), were fitted to the data using the maximum likelihood method. Each

23	model was tested for goodness-of-fit using a chi-square goodness-of-fit test (x2 p-value < 0.051

24	indicates lack of fit). Other factors were used to assess model fit, such as scaled residuals, visual fit,

25	and adequacy of fit in the low-dose region and in the vicinity of the BMR.

26	For each endpoint, the BMDL estimate (95% lower confidence limit on the BMD, as

27	estimated by the profile likelihood method) and AIC value were used to select a best-fit model from

28	among the models exhibiting adequate fit If the BMDL estimates were "sufficiently close," that is,

29	differed by less than threefold, the model selected was the one that yielded the lowest AIC value. If

30	the BMDL estimates were not sufficiently close, the lowest BMDL was selected as the POD.

31

1 A significance level of 0.05 is used for selecting cancer models because the model family (multistage) is
selected a priori [Benchmark Dose Technical Guidance Document, U.S. EPA, 2012).

38


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

1	Table 6-4. Cancer endpoints selected for dose-response modeling for diisononyl

2	phthalate (DINP)

3

Species / Sex
Endpoint

Doses and Effect Data

Moore (1998b)

Mice (B6C3Fi)/ Female

Dose (mg/kg-d)

0

15.89

47.30

127.47

263.72

Hepatocellular adenoma
or carcinoma

Incidence /Total

3/70

5/68

10/68

11/67

33/70

4

5	Modeling Results

6	The modeling results are summarized below.

7

8	Table 6-5. Summary of BMD model results for increased incidence of hepatocellular

9	carcinomas and adenomas combined in female B6C3Fi mice exposed to DINP in the

10	diet for 2 years (Moore, 1998b); BMR = 10% extra risk

11

Model3

Goodness of fit

BMD10hed

BMDL10hed

Basis of model selection

p-value

AIC

Multistage 1°

0.30

281.78

55.6

42.6

All models provided an adequate fit. The
multistage-cancer 4° model was selected
based on lowest AIC.

Multistage 2°

0.25

282.65

82.0

45.2

Multistage 3°

0.34

282.04

87.7

47.1

Multistage 4°

0.39

281.73

88.7

48.4

aSelected model in bold. Scaled residuals for the selected model for doses 0,15.89, 47.30,127.47, and 263.72 mg/kg-
d were -0.48, 0.02,1.10, -0.66, and 0.06, respectively. The cancer slope factor for the selected model was 0.1 /48.4 =
0.00206.

39


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Multistage Cancer Model with 0.95 Confidence Level

dose

10:50 04/20 2011

1

2	Figure 6-3. Plot of incidence rate by dose, with the fitted curve for the

3	multistage-cancer 1° model. BMR = 10% extra risk; dose shown in mg/kg-day.

4

5	Multistage Cancer Model. (Version: 1.9; Date: 05/26/2010)

6

7	The form of the probability function is: P[response] = background + (l-background)*[l-EXP(

8	-betal*doseAl-beta2*doseA2-beta3*doseA3-beta4*doseA4)]

9

10	The parameter betas are restricted to be positive

11

12	Parameter Estimates

Variable

Estimate

(Default) Initial
Parameter Values

Background

0.0559152

0.0659908

Beta(l)

0.00114845

0.000880186

Beta(2)

0

0

Beta(3)

0

0

Beta(4)

5.58108e-011

6.9501e-011

13

14

Analysis of Deviance Table

Model

Log(likelihood)

# Pa ram's

Deviance

Test d.f.

P-value

Full model

-136.965

5







Fitted model

-137.865

3

1.79969

2

0.4066

40


-------
1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Reduced model

-162.082

50.233

AIC: 281.73
Goodness of Fit Table

Dose

Est. Prob.

Expected

Observed

Size

0

0.0559

3.914

70

15.89

0.0730

4.963

68

47.30

0.1061

7.214

10

68

127.47

0.1964

13.160

11

67

263.72

0.4676

32.732

33

70

ChiA2 = 1.88 d.f. = 2 P-value = 0.3915
Benchmark Dose Computation

BMR = 10% extra risk
BMD = 88.7294

BMDL at the 95% confidence level = 48.4306
BMDU at the 95% confidence level = 163.388

Taken together, (48.4306,163.388) is a 90% two-sided confidence interval for the BMD
Multistage Cancer Slope Factor = 0.00206481

41


-------
DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

1	EXAMPLE 7 - Considerations for Selecting Organ/System-Specific or

2	Overall T oxicity Values

3	The example below demonstrates the derivation or selection of an organ/system-specific toxicity

4	value for each organ or system affected by the agent, as well as an overall toxicity value for the agent

5	to represent lifetime human exposure levels where effects are not anticipated to occur¦, as described in

6	the "Dose-Response Analysis" section of the draft Handbook for IRIS Assessment Development.

7	Draft Assessment Text:

8	The candidate values presented in the table below are preliminary to the derivation of the

9	organ/system-specific reference values. These candidate values are considered individually in the

10	selection of a representative oral reference value for a specific hazard and subsequent overall RfD

11	for benzo[a]pyrene.

12

13	Table 7-1. Effects and corresponding derivation of candidate values

14

Endpoint and Reference

PODHEDa
(mg/kg-d)

POD type

UFa

UFh

ufl

UFS

ufd

Composite
UF

Candidate
value (mg/kg-d)

Developmental

Neurodevelopmental
alterations in rats
Chen et al. (2012)

0.09

BMDL1sd

10

10

1

1

3

300

3 x 10"4

Cardiovascular effects in rats
Jules et al. (2012)

0.15

LOAEL

3

10

10

1

3

1,000

2 x 10"4

Reproductive

Decreased ovary weight and
ovarian follicles in rats
Xu et al. (2010)

0.37

BMDL1sd

3

10

1

10

3

1,000

4 x 10"4

Decreased intratesticular
testosterone in rats
Zheng et al. (2010)

0.24

NOAEL

3

10

1

10

3

1,000

2 x 10"4

Decreased sperm count in mice
Mohamed et al. (2010)

0.15

LOAEL

3

10

10

10

3

10,000

Not calculated due
to UF > 3000b

Cervical epithelial hyperplasia
in mice

Gao et al. (2011a)

0.06

BMDL10

3

10

1

10

3

1,000

6 x 10"5

Immunological

Decreased thymus weight in
rats

Kroese et al. (2001)

1.9

BMDL1sd

3

10

1

10

3

1,000

2 x 10"3

Decreased serum IgM in rats
De Jong et al. (1999)

1.7

NOAEL

3

10

1

10

3

1,000

2 x 10"3

Decreased serum IgA in rats
De Jong et al. (1999)

5.2

NOAEL

3

10

1

10

3

1,000

5 x 10"3

42


-------
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

31

32

33

34

35

36

37

38

DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Endpoint and Reference

PODHEDa
(mg/kg-d)

POD type

UFa

UFh

UFl

UFS

ufd

Composite
UF

Candidate
value (mg/kg-d)

Decreased number of B cells in
rats

De Jong et al. (1999)

5.2

NOAEL

3

10

1

10

3

1,000

5 x 10"3

aHED PODs were calculated using BW3/4 scaling (U.S. EPA,2011) for adult animal studies (Chen et al., 2011;
Mohamed et al., 2010; Xu et al., 2010; and De Jong et al., 1999) but not for studies dosing early postnatal animals
(Chen et al., 2012).

bAs recommended in EPA's A Review of the Reference Dose and Reference Concentration Processes (U.S. EPA,
2002), the derivation of a reference value that involves application of the full 10-fold uncertainty factor in four or
more areas of extrapolation should be avoided.

UFa—A value of 3 (100.5 = 3.16, rounded to 3) was applied to account for uncertainty in characterizing
toxicodynamic differences between rats and humans when an HED was calculated using BW3/4 scaling as
uncertainty in characterizing toxicokinetic differences was accounted for through calculation of an HED
using a standard DAF consistent with EPA guidance (U.S. EPA, 2011b). A value of 10 was applied when
BW3/4 scaling was not employed to account for uncertainty in extrapolating from laboratory animals to
humans because of the absence of information to characterize either the toxicokinetic or toxicodynamic
differences between animals and humans following oral exposure to benzo[a]pyrene.

UFh—A value of 10 was applied to account for potentially susceptible individuals because adequate
information is not available to quantitatively estimate variability in human susceptibility. In the case of
benzo[a]pyrene, insufficient information is available to quantitatively estimate variability in human
susceptibility.

UFl—A value of 1 was applied when the POD is based on dose-response modeling or a NOAEL; 10 when the
POD is a LOAEL. In the case of benzo[a]pyrene, an UFl of 1 was applied for LOAEL-to-NOAEL extrapolation
because a BMR of a 1 SD change from the control mean in neurodevelopmental impairments was selected
under an assumption that it represents a minimal biologically significant response level.

UFs—A value of 1 was applied when dosing occurred during gestation or the early postnatal period that is
relevant to developmental effects (U.S. EPA, 1991a); 10 when the POD is based on a subchronic study
(studies in this table, other than the developmental toxicity studies, were 42-90 days in duration) to account
for the possibility that longer exposure may induce effects at a lower dose.

UFd—A value of 3 was applied to account for database deficiencies including the lack of a standard
multigenerational study or extended 1-generation study that includes exposure from premating through
lactation, considering that benzo[a]pyrene has been shown to affect fertility in adult male and female
animals by multiple routes of exposure (see Section 1.1.2). Also, the lack of a study examining functional
neurological endpoints following a more comprehensive period of developmental exposure (i.e., gestation
through lactation) is a data gap, considering human and animal evidence indicating altered neurological
development (see Section 1.1.1).

43


-------
1

2

3

4

5

6

7

8

9

10

11

DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Figure 7-1 presents graphically the candidate values, UFs, and PODs, with each bar
corresponding to one data set described in Table 7-1.

Neurodevelopmental
alterations in rats
(Chen et al., 2012)

>
L1J

o

Cardiovascular effects in rats
(Jules et al., 2012)

¦i- Ovary weight and
ovarian follicles in rats
(Xuet al., 2010)

L>
D
O

o

tx.

Q-

4, Intratesticular
testosterone in rats
Zheng et al. (2010)

nL- Sperm count in mice
(Mohamed et al., 2010)

Cervical epithelial hyperplasia
(Gaoet al., 2011)

•I Thymus weight in rats
(Kroeseet al., 2001)

<
u

u
o

_i

o
z

Z>

Serum IgM in rats
(De Jong et al., 1999)

-i- Serum IgA in rats
(De Jong et al., 1999)

4, Number of B cells in rats
(De Jong et al., 1999)

Composite UF
ACandidate value
• POD(HED)

0.00001 0.0001 0.001	0.01	0.1

Doses (mg/kg-d)

10

Figure 7-1. Candidate values with corresponding PODs and composite UFs.

Derivation of Organ/System-specific Reference Doses

Table 7-2 distills the candidate reference doses from Table 7-1 into a single value for each
organ or system. These organ or system-specific reference values may be useful for subsequent
cumulative risk assessments that consider the combined effect of multiple agents acting at a
common site.

44


-------
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

DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Table 7-2. Organ/system-specific RfDs and proposed overall RfD for benzo[a]pyrene

Effect

Basis

RfD (mg/kg-d)

Exposure
description

Confidence

Developmental

Neurodevelopmental alterations

3 x 10"4

Critical
window of
development
(postnatal)

MEDIUM

Reproductive

Decreased ovary weight and ovarian follicles

4 x 10"4

Subchronic

MEDIUM

Immunological

Decreased thymus weight and serum IgM

2 x 10"3

Subchronic

LOW

Proposed Overall RfD

Developmental toxicity

3 x 10"4

Critical
window of
development
(postnatal)

MEDIUM

Developmental Toxicity

The candidate value based on neurodevelopmental impairment in rats (Chen et al., 2012)
was selected as the organ/system-specific RfD representing developmental toxicity. This candidate
RfD was selected because it is associated with the application of the smaller composite UF and
because similar effects were replicated across other studies.

Reproductive Toxicity

The candidate RfD based on decreased ovary weight and ovarian follicle numbers in rats
from the Xu et al. (2010) study was selected as the organ/system-specific RfD representing
reproductive toxicity. The ovarian effects are supported by a large database of animal studies and
human studies of exposure to benzo[a]pyrene and PAH mixtures. The data supporting cervical
effects associated with oral benzo[a]pyrene exposure are limited to a single study; however, the
finding is supported by corollary findings after i.p. exposure and by studies in humans.

Immunotoxicity

The candidate RfDs based on decreased thymus weight (Kroese et al., 2001) and serum IgM
levels in rats (De Jong et al., 1999) were selected as the organ/system-specific RfD representing
immunotoxicity. The observed decreases in thymus weight, IgM and IgA levels, and number of
B cells associated with exposure to benzo[a]pyrene were determined to be representative of
immunotoxicity. In combination, these effects provide more robust evidence of immunotoxicity.
The candidate RfDs for decreased thymus weight (Kroese et al., 2001) and serum IgM levels in rats
(De Jong et al., 1999) were equal and provided the most sensitive candidate RfDs; thus, these
candidate RfDs were selected as the organ/system-specific RfDs representing immunotoxicity.

45


-------
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

31

32

33

34

35

36

DRAFT MATERIALS FOR REVIEW ONLY - DO NOT CITE OR QUOTE

Selection of the Proposed Overall Reference Dose

For benzo[a]pyrene, multiple organ/system-specific reference doses were derived for
effects identified as potential hazards from benzo[a]pyrene including developmental toxicity,
reproductive toxicity, and immunotoxicity. To estimate an exposure level below which effects from
benzo[a]pyrene exposure are not expected to occur, the lowest organ/system-specific RfD
(3 x 10"4 mg/kg-day) is proposed as the overall reference dose for benzo[a]pyrene. This value,
based on induction of neurodevelopmental alterations in rats exposed to benzo[a]pyrene during a
susceptible lifestage is supported by several animal and human studies (see Section 1.1.1).

The overall reference dose is derived to be protective of all types of effects for a given
duration of exposure and is intended to protect the population as a whole including potentially
susceptible subgroups (U.S. EPA, 2002). Decisions concerning averaging exposures over time for
comparison with the RfD should consider the types of toxicological effects and specific lifestages of
concern. Fluctuations in exposure levels that result in elevated exposures during these lifestages
could potentially lead to an appreciable risk, even if average levels over the full exposure duration
were less than or equal to the RfD.

Furthermore, certain exposure scenarios may require particular attention to the risk-
assessment population of interest in order to determine whether a reference value based on
toxicity following developmental exposure is warranted. For example, the use of an RfD based on
developmental effects may not be appropriate for a risk assessment in which the population of
interest is post-reproductive age adults.

Confidence Statement

A confidence level of high, medium, or low is assigned to the study used to derive the RfD,
the overall database, and the RfD itself, as described in Section 4.3.9.2 of EPA's Methods for
Derivation of Inhalation Reference Concentrations and Application of Inhalation Dosimetry (U.S. EPA,
1994).

Confidence in the principal study (Chen etal., 2012) is medium-to-high. The study design
included randomized experimental testing, blinded observations, culling of pups to account for
nutritional availability, treatment-randomization, and controls for litter and nursing bias. Some
informative experimental details were, however, omitted including the sensitivity of some assays at
the indicated developmental ages and lack of reporting gender-specific data for all outcomes.
Notably, these study limitations do not apply to the endpoint chosen to derive the RfD, and the
overall methods and reporting are considered sufficient Confidence in the database is medium,
primarily due to the lack of a multigenerational reproductive toxicity study given the sensitivity to
benzo[a]pyrene during development Reflecting medium-to-high confidence in the principal study
and medium confidence in the database, confidence in the RfD is medium.

46


-------