PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
EPA Document # EPA-740-R1-8012
March 2020
Office of Chemical Safety and
Pollution Prevention
Draft Risk Evaluation for
Asbestos
Systematic Review Supplemental File:
Data Quality Evaluation of Human Health Hazard Studies:
Mesothelioma and Lung Cancer Studies
March 2020
AEPA
United States
Environmental Protection Agency
1
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Table of Contents
1. Asbestos-specific evaluation criteria for epidemiologic studies 3
1.1. Rationale for asbestos-specific evaluation criteria 3
1.2. Evaluation Criteria for Epidemiological Studies: Asbestos Exposure and Lung Cancer
Health Outcome 6
1.3. Evaluation Criteria for Epidemiological Studies: Asbestos Exposure and Mesothelioma
Health Outcome 17
2. Data Evaluation of Lung Cancer Studies 28
2.1. Table of studies evaluated for asbestos exposure and lung cancer incidence 28
2.2. Data Evaluation Scoring Sheets: Lung Cancer Outcome 30
2.2.1. Epidemiology evaluation results of the South Carolina, US cohort studies on asbestos exposure and
lung cancer incidence 30
2.2.2. Epidemiology evaluation results of the Chongqing, China mining cohort studies on asbestos
exposure and lung cancer incidence 36
2.2.3. Epidemiology evaluation results of the Balangero, Italy cohort of studies on asbestos exposure and
lung cancer incidence 41
2.2.4. Epidemiology evaluation results of the North Carolina, US cohort studies on asbestos exposure and
lung cancer incidence 46
2.2.5. Epidemiology evaluation results of the Salonit Anhovo, Slovenia cohort studies on asbestos
exposure and lung cancer incidence 51
2.2.6. Epidemiology evaluation results of the Quebec, Canada cohort studies on asbestos exposure and
lung cancer incidence 55
2.2.7. Epidemiology evaluation results of the Chongqing, China textile worker cohort studies on asbestos
exposure and lung cancer incidence 60
3. Data Quality Evaluation of Mesothelioma Data Sources 65
3.1. Data Evaluation Scoring Sheets: Mesothelioma Outcome 65
3.1.1. Epidemiology evaluation results of the Quebec, Canada cohort studies on asbestos exposure and
mesothelioma incidence 65
3.1.2. Epidemiology evaluation results of the South Carolina, US cohort studies on asbestos exposure and
mesothelioma incidence 69
3.1.3. Epidemiology evaluation results of the North Carolina, US cohort studies on asbestos exposure and
mesothelioma incidence 73
References 77
2
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
1. Asbestos-specific evaluation criteria for epidemiologic studies
1.1. Rationale for asbestos-specific evaluation criteria
For the first 10 Toxic Substances Control Act (TSCA) chemicals, a general set of study
evaluation criteria was developed. These evaluation criteria were not tailored to any specific
exposure or outcome. In the Problem Formulation step of the asbestos assessment, it was
accepted that asbestos was a known cause of lung cancer and mesothelioma, and that the purpose
of the systematic review would be the identification of studies which could inform the estimation
of an exposure-response function allowing for the derivation of an asbestos inhalation unit risk
for these two cancer sites combined. While there is also evidence that asbestos exposure is
associated with an increased risk of laryngeal and ovarian cancer, there is inadequate data for
exposure-response analyses. For the reasons described below, the study domains of exposure,
outcome, study participation, potential confounding, and analysis were further tailored to the
specific needs of evaluating asbestos studies for their potential to provide information on the
exposure-response relationship between asbestos exposure and mortality from lung cancer and
from mesothelioma (see sections L2 and 1.3).
In terms of evaluating exposure information, asbestos is unique among these first 10 TSCA
chemicals as it is a fiber and has a long history of different exposure assessment methodologies.
For mesothelioma, this assessment is also unique with respect to the impact of the timing of
exposure relative to the cancer outcome as the time since first exposure plays a dominant role in
modeling risk. The most relevant exposures for understanding mesothelioma risk were those that
occurred decades prior to the onset of cancer, and subsequent cancer mortality. Asbestos
measurement methodologies have changed over those decades; from early measurement of total
dust particles measured in units of million particles per cubic foot of air (mppcf) (by samplers
called midget impingers), to fibers per milliliter (f/ml), or the equivalent fibers per cubic
centimeter (f/cc) (where fiber samples were collected on membrane filters and the f/ml or f/cc
was measured using phase contrast microscopy (PCM) analysis of the filters). In several studies
encompassing several decades of asbestos exposure, matched samples from midget impingers
and membrane filters were compared to derive job- (or location-) specific factors allowing for
the conversion of earlier midget impinge measurements to estimate PCM measurement of
asbestos air concentrations. While some studies were able to provide these factors for specific
locations and jobs, other studies were only able to derive one factor for all jobs and locations.
The use of such data has allowed asbestos researchers to investigate the risk of asbestos and
3
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
successfully model lung cancer and mesothelioma mortality over several decades of evaluation
(U.S. EPA. 2014. 1988). Thus, the general exposure evaluation criteria were adjusted to be
specific to exposure assessment methodologies such as midget impingers and PCM with
attention to the use of job-exposure-matrices (JEMs) to reconstruct workers' exposure histories
and the reporting of key metrics needed to derive exposure-response functions for lung cancer
and mesothelioma.
In terms of evaluating the quality of outcome information, lung cancer is relatively
straightforward to evaluate as an outcome. Specific International Classification of Disease (ICD)
codes for lung cancer have existed for the entire time period of the studies making it possible to
identify cases from mortality databases. On the other hand, there was no diagnostic code for
mesothelioma in the ICD prior to the introduction of the 10th revision (ICD-10) which was not
implemented in United States until 1999. Before ICD-10, individual researchers had to go
beyond ICD codes and generally search original death certificates for mention of mesothelioma.
Thus, the general outcome evaluation criteria were adjusted to be specific to mesothelioma and
outcome ascertainment strategies.
Mesothelioma is a very rare cancer. As noted by U.S. EPA (2014). the "Centers for Disease
Control and Prevention estimated the death rate from mesothelioma, using 1999 to 2005 data, as
approximately 23.2 per million per year in males and 5.1 per million per year in females (CDC.
2009)." While extremely rare, the overwhelmingly dominant cause of mesothelioma is asbestos
exposure (Tossavainen, 1997) making the observance of mesothelioma in a population a sentinel
for asbestos exposure. It is critical to understand that the prevailing risk model for mesothelioma
models is an absolute risk model of mesothelioma mortality which assumes there is no risk at
zero exposure (U.S. EPA. 1988; Peto et al.. 1982; Peto. 1978). This use of an absolute risk model
is in stark contrast to the standard use of a relative risk model for lung and other cancers. For the
relative risk model, the risk of lung cancer in an asbestos exposed population would multiply the
background risk in an unexposed population, and consideration of study quality would be the
evaluation of the comparison population. There is, however, no background risk in developing
mesothelioma in an unexposed population. As a result, no comparison population was needed to
estimate the absolute risk among people exposed to asbestos, and therefore, criteria including
comparison population were adjusted for mesothelioma.
In terms of evaluating potential confounding variables, the potential confounding section
recognized that there are both direct and indirect methods for controlling for some confounders -
4
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
specifically that methodologies involving internal comparisons in a working population may
indirectly control for smoking and other factors assuming these factors do not vary with asbestos
exposure concentrations in the workplace. In contrast to lung cancer, mesothelioma is much
simpler to evaluate for potential confounding as chest radiation is the only other known risk
factor that could lead to mesothelioma, and this rare exposure is unlikely to be a confounder.
In terms of analysis, the evaluation criteria needed to be adapted for both mesothelioma and lung
cancer. For mesothelioma, the Peto model (Peto et al.. 1982; Peto. 1978) was traditionally used
for summary data published in the literature (U.S. EPA. 1988). so only modeling using the Peto
model by the authors, or the presentation of sufficient information to fit the Peto model post hoc
were considered acceptable. For lung cancer, a wider selection of statistical models was
acceptable, with the preference generally given to modeling that used individual data in the
analysis. Grouped data modeling would also reported but would be carried forward to the
summary only if no individual data modeling were available.
Lastly for Asbestos, studies from the same cohort were evaluated collectively to assess the
overall quality of the data collected from the cohort across all years of follow-up. This was done
to consider all information from a cohort that was presented across multiple studies as a whole.
For example, the most recent article for a cohort may not have presented the details necessary to
fully evaluate the number one domain criterion (Study Participation), whereas the first or
subsequent studies out of the cohort may have filled in data gaps.
5
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
1.2. Evaluation Criteria for Epidemiological Studies: Asbestos
Exposure and Lung Cancer Health Outcome
Confidence Level
(Score)
Description
Selected
Score
Domain 1. Study Participation
Metric 1. Participant selection (selection, performance biases)
Instructions: To meet criteria for confidence ratings for metrics where 'AND' is included, studies must
address both conditions where 'AND' is stipulated. To meet criteria for confidence ratings for metrics
where 'OR' is included studies must address at least one of the conditions stipulated. In Metrics 3 and 4,
criteria that must be met concurrently are enclosed in parentheses and linked with an indented 'AND.'
High
(score = 1)
For all studv tvves: All kev elements of the studv desisn are rcDortcd (e.s..
setting, participation rate described at all steps of the study, inclusion and
exclusion criteria, and methods of participant selection or case
ascertainment)
AND
The reported information indicates that selection in or out of the study (or
analysis sample) and participation was not likely to be biased (i.e., the
exposure-outcome distribution of the participants is likely representative of
the exposure-outcome distributions in the population of persons eligible for
inclusion in the study).
Medium
(score = 2)
For all studv tvves: Some kev elements of the studv desisn were not
present but available information indicates a low risk of selection bias (i.e.,
the exposure-outcome distribution of the participants is likely
representative of the exposure-outcome distributions in the population of
persons eligible for inclusion in the study).
Low
(score = 3)
For all studv tvves: Kev elements of the studv desisn and information on
the population (e.g., setting, participation rate described at most steps of the
study, inclusion and exclusion criteria, and methods of participant selection
or case ascertainment) are not reported [STROBE checklist 4, 5 and 6 (Von
Elm etal.. 2008)1.
Unacceptable (score
= 4)
For all studv tvves: The reported information indicates that selection in or
out of the study (or analysis sample) and participation was likely to be
significantly biased (i.e., the exposure-outcome distribution of the
participants is likely not representative of the exposure-outcome
distributions of the population of persons eligible for inclusion in the
study).
Not rated/applicable
Do not select for this metric.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important
elements such as relevance]
Metric 2. Attrition (missing data/attrition/exclusion, reporting biases)
High
(score = 1)
For cohort studies: There was minimal subiect loss to follow lid durins the
study (or exclusion from the analysis sample) and outcome and exposure
data were largely complete
OR
Loss of subjects (e.g., incomplete outcome data) or missing exposure and
outcome data was adequately* addressed (as described below) and reasons
6
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
( 0111'idcMice I.cm'I
(Score)
Description
wciv documented when liLiiiian suburb «civ rcnio\ cd from a slud\ (Vl_l\
2015).
AND
Missing data have been imputed using appropriate methods (e.g., multiple
imputation methods), and characteristics of subjects lost to follow up or
with unavailable records are not significantly different from those of the
study participants (NTP. 2015).
For case-control studies and cross-sectional studies: There was minimal
subject withdrawal from the study (or exclusion from the analysis sample)
and outcome data and exposure were largely complete.
OR
Any exclusion of subjects from analyses was adequately* addressed (as
described below), and reasons were documented when subjects were
removed from the study or excluded from analyses (NTP. 2015).
*NOTE for all study types: Adequate handling of subject attrition can include:
Use of imputation methods for missing outcome and exposure data; reasons
for missing subjects unlikely to be related to outcome (for survival data,
censoring was unlikely to introduce bias); missing outcome data balanced in
numbers across study groups, with similar reasons for missing data across
groups.
Medium
(score = 2)
For cohort studies: There was moderate subject loss to follow up during
the study (or exclusion from the analysis sample) or outcome and exposure
data were nearly complete.
AND
Any loss or exclusion of subjects was adequately addressed (as described in
the acceptable handling of subject attrition in the high confidence category)
and reasons were documented when human subjects were removed from a
study.
For case-control studies and cross-sectional studies: There was moderate
subject withdrawal from the study (or exclusion from the analysis sample),
but outcome and exposure data were largely complete
AND
Any exclusion of subjects from analyses was adequately addressed (as
described above), and reasons were documented when subjects were
removed from the study or excluded from analyses (NTP. 2015).
Low
(score = 3)
For cohort studies: The loss of subjects (e.g., loss to follow up, incomplete
outcome or exposure data) was moderate and unacceptably handled (as
described below in the unacceptable confidence category) (Source: OHAT).
OR
Numbers of individuals were not reported at important stages of study (e.g.,
numbers of eligible participants included in the study or analysis sample,
completing follow-up, and analyzed). Reasons were not provided for non-
participation at each stage [STROBE Checklist Item 13 (Von Elm et al..
2008)1.
For case-control and cross-sectional studies: The exclusion of subjects
from analyses was moderate and unacceptably handled (as described below
in the unacceptable confidence category).
OR
Numbers of individuals were not reported at important stages of study (e.g.,
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Confidence Level
(Score)
Description
Selected
Score
numbers of eligible participants included in the study or analysis sample,
completing follow-up, and analyzed). Reasons were not provided for non-
oarticioation at each stase 1 STROBE Checklist Item 13 (Von Elm et al..
2008)].
Unacceptable (score
= 4)
For cohort studies: There was larse subiect attrition durins the studv (or
exclusion from the analysis sample).
OR
Unacceptable handling of subject attrition: reason for missing outcome data
likely to be related to true outcome, with either imbalance in numbers or
reasons for missing data across study groups; or potentially inappropriate
application of imputation (Source: OH AT).
For case-control and cross-sectional studies: There was larse subiect
withdrawal from the study (or exclusion from the analysis sample).
OR
Unacceptable handling of subject attrition: reason for missing outcome data
likely to be related to true outcome, with either imbalance in numbers or
reasons for missing data across study groups; or potentially inappropriate
application of imputation.
Not rated/applicable
Do not select for this metric.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important
elements such as relevance]
Metric 3. Comparison Group (selection, performance biases)
High
(score = 1)
For ALL studv tvves: Anv differences in baseline characteristics of sroiros
were considered as potential confounding or stratification variables and
were thereby controlled by statistical analysis (Source: OHAT).
OR
For cohort and cross-sectional studies: Kev elements of the studv desisn
are reported (i.e., setting, inclusion and exclusion criteria, and methods of
participant selection), and indicate that groups were similar (e.g., recruited
from the same eligible population with the same method of ascertainment
and within the same time frame using the same inclusion and exclusion
criteria, and were of similar ase and health status) (NTP. 2015).
For case-control studies: Kev elements of the studv desisn are rcDortcd
indicate that that cases and controls were similar (e.g., recruited from the
same eligible population with the number of controls described, and
elisibilitv criteria and are recruited within the same time frame (NTP.
2015).
For studies revortins Standardized Mortalitv Ratios (SMRs) or
Standardized Incidence Ratios (SLRs): Ase. sex (if aDDlicablc). and race
(if applicable) adjustment or stratification is described and choice of
reference population (e.g.. general population) is reported.
Medium
(score = 2)
For cohort studies and cross-sectional studies: There is onlv indirect
evidence (e.g., stated by the authors without providing a description of
methods) that groups are similar (as described above for the high
confidence rating).
For case-control studies: There is indirect evidence (i.e.. stated bv the
authors without providing a description of methods) that cases and controls
are similar (as described above for the high confidence rating).
8
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Confidence l.c\cl
(Score)
Description
Sclcclcd
Score
For studies rcporlinu SMIis or SIlis: Auc. iC\ (il applicable), and race (if
applicable) adjustment or stratification is not specifically described in the
text, but results tables are stratified by age and/or sex (i.e., indirect
evidence); choice of reference population (e.g., general population) is
reported.
Low
(score = 3)
For cohort and cross-sectional studies: There is indirect evidence (i.e..
stated by the authors without providing a description of methods) that
groups were not similar (as described above for the high confidence rating).
AND
Differences between the exposure groups are not adequately controlled for
in the statistical analysis.
For case-control studies. There is indirect evidence (i.e.. stated bv the
authors without providing a description of methods) that cases and
controls were not similar (as described above for the high confidence
rating).
AND
The characteristics of cases and controls are not rcDortcd (Source: (NTP.
2015).
AND
Differences in groups is not adequately controlled for in the statistical
analysis.
For studies revortins SMRs or SIRs: Indirect evidence of a lack of
adjustment or stratification for age or sex (if applicable); indirect evidence
that choice of reference population (e.g., general population) is
inappropriate.
Unacceptable*
(score = 4)
For cohort studies: Subiects in all exposure groups were not similar
OR
(Information was not reported to determine if participant groups were
similar 1 STROBE Checklist 6 (Von Elm et al.. 2008)
AND
Potential differences in exposure groups were for a factor that was related to
the outcome and not controlled for in the statistical analysis.)
OR
(Subjects in the exposure groups had very different participation/response
rates (NTP. 2015).
AND
Participation rates were related to exposure and outcome)
For case-control studies: (Controls were drawn from a verv dissimilar
DODiilation than cases or recruited within verv different time frames (NTP.
2015).
AND
Potential differences in the case and control groups were not controlled for
in the statistical analysis.)
OR
Rationale and/or methods for case and control selection, matching criteria
including number of controls per case (if relevant) were not reported
[STROBE Checklist 6 (Von Elm et al.. 2008)1.
For cross-sectional studies: (Subiects in all exposure uroups were not
similar, recruited within very different time frames, or had very different
participation/response rates (NTP. 2015).
AND
9
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Confidence Level
(Score)
Description
Selected
Score
Potential differences in exposure groups were not controlled for in the
statistical analysis.)
OR
Sources and methods of selection of participants in all exposure groups
were not rcDortcd 1 STROBE Checklist 6 (Von Elm et al.. 2008)1.
For studies reporting SMRs or SIRs: Lack of adjustment or stratification
for both age and sex (if applicable), race (if applicable), and calendar time
or choice of reference population (e.g., general population) is not reported.
Not rated/applicable
Do not select for this metric.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important elements
such as relevance]
Domain 2. Exposure Characterization
Metric 4. Measurement of Exposure (Detection/measurement/information, performance biases)
High
(score = 1)
For all studv tvves: Quantitative estimates of exposure were consistentlv
assessed (i.e., using the same method and sampling time-frame) during
multiple time periods and using either PCM or TEM.
OR
A combination of methods were used over time (i.e., midget
impinger, PCM or TEM), but side by side sampling and analyses
were conducted to develop appropriate conversion criteria.
AND
For an occupational population, contains detailed employment records and
quantitative estimates of exposure using either PCM or TEM which allows
for construction of job-matrix for entire work history of exposure (i.e..
Cumulative or peak exposures, and time since first exposure).
Medium*
(score = 2)
For all studv tvves: (Exposure was assessed durine one time oeriod but this
time period is judged to be reasonably representative of the entire study
time period.
AND
Exposure was assessed using a combination of midget impingers, PCM,
and/or TEM measurements, but side by side sampling and analyses were
not conducted for all operations and thus there is a lack of confidence in the
conversion factors.)
OR
For an occupational study population, contains detailed employment records
and quantitative estimates of exposure using a combination of midget
impingers and PCM or TEM measurements for only a portion of
participant's work history of exposure (i.e., only early years or later years),
such that extrapolation of the missing years is required
Low
(score = 3)
For all studv tvves: Exposure was estimated solelv usins professional
judgement.
OR
Exposure was directly measured and assessed using a quantitative method
other than PCM or TEM and conversion factors were not determined.
Unacceptable (score
= 4)
For all studv tvves: There was no Quantitative measure or estimate of
exposure.
OR
Methods used to quantify the exposure were not well defined, and sources
of data and detailed methods of exposure assessment were not reported
[STROBE Checklist 7 and 8 (Von Elm et al.. 2008)1.
10
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Confidence Level
(Score)
Description
Selected
Score
OR
There is evidence of substantial exposure misclassification that would
significantly bias the results.
Not rated/applicable
Do not select for this metric.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important elements
such as relevance]
Metric 5. Exposure levels (Detection/measurement/information biases)
High
(score = 1)
Do not select for this metric.
Medium
(score = 2)
For all studv tvves: The ranse and distribution of exposure is sufficient or
adeauate to develop an cxDosurc-rcsDonsc estimate (Coooer et al.. 2016).
AND
Reports 3 or more levels of exposure (referent group + 2 or more) or an
exposure-response model using a continuous measure of exposure.
Low
(score = 3)
For all studv tvves: The ranse of exposure in the population is limited.
OR
Reports 2 levels of exposure (e.g., exposed/unexposed)) (Cooper) (Source:
IRIS).
Unacceptable (score
= 4)
For all studv tvves: The ranse and distribution of exposure are not adeauate
to determine an exposure-response relationship (Cooper et al.. 2016).
OR
No description is provided on the levels or range of exposure.
Not rated/applicable
Do not select for this metric.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important elements
such as relevance]
Metric 6. Temporality
High
(score = 1)
For all studv tvves: The studv presents an appropriate temporalitv between
exposure and outcome (i.e. the exposure precedes the disease).
AND
The interval between the exposure (or reconstructed exposure) and the
outcome is sufficiently long considering the latency of the disease (i.e.
study follow-up is more than 15 years for lung cancer) (Lakind et al.. 2014).
Medium
(score = 2)
For all studv tvves: Temporalitv is established, but it is unclear whether
there is adequate follow-up for consideration of latency (i.e., only 10 years
of follow-up) (Lakind et al.. 2014).
Low
(score = 3)
For all studv tvves: The temporalitv of exposure and outcome is uncertain
(5-10 years).
OR
There is inadequate follow-up of the cohort considering the latency period.
Unacceptable
(score = 4)
For all studv tvves: Studv lacks an established time order, such that
exposure is not likelv to have occurred prior to outcome (Lakind et al..
2014).
OR
There was inadequate follow-up of the cohort for the expected latency
period (<5 years).
OR
Sources of data and details of methods of assessment were not sufficiently
reported (e.g. duration of follow-up, periods of exposure, dates of outcome
11
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Confidence Level
(Score)
Description
Selected
Score
ascertainment, etc.) Source: STROBE Checklist 8 (Von Elm et al.. 2008)).
Not rated/applicable
Do not select for this metric.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important elements
such as relevance]
Domain 3. Outcome Assessment
Metric 7. Outcome measurement or characterization (detection/measurement/information, performance,
reporting biases)
High
(score = 1)
For all studv tvves: The outcome was assessed usins one or a
combination of the following well-established methods:
o Lung cancer cases confirmed by histological or cytological means
(including subtypes of lung cancer)
o ICD-10 C34 (lung and bronchus with or without C33 (trachea)
o ICD-9 (5-digit code) 162.2-162.9 or
o ICD-8 (4-digit code) 162.1 or
o ICD-7 (4-digit code) 162.1 and 163
o ICD-9 (3-digit code) 162
o ICD-8 (3-digit code) 162
o ICD-7 (3-digit code) 162 and 163
Medium
(score = 2)
For all studv tvves: Althoush authors state thev identified luns cancer
cases they did not report the ICD codes.
Low
(score = 3)
Do not select for this metric.
Unacceptable
(score = 4)
For all studv tvves: Anv sclf-rcDortcd information.
OR
Study lacks individual assessment of lung cancer (i.e., lung cancer is
assessed as a combination of cancer types, excluding lung and bronchus or
trachea).
Not
rated/applicable
Do not select for this metric
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important elements
such as relevance]
Metric 8. Reporting Bias
High
(score = 1)
For all studv tvves: Luns cancer findinss are reported in the
abstract, results or discussion. Effect estimates are reported with
confidence intervals and/or standard errors, number of
cases/controls or exposed/unexposed reported for each analysis, to
be included in exposure-response analysis or fully tabulated during
data extraction and analyses (NTP. 2015).
Medium
(score = 2)
For all study types: All of the study's findings (primary and secondary)
outlined in the abstract, results or discussion (that are relevant for the
evaluation) are reported but not in a way that would allow for detailed
extraction (e.g., results were discussed in the text but accompanying data
were not shown).
Low
(score = 3)
For all studv tvves: Luns cancer outcomes outlined in the methods,
abstract, and/or introduction (that are relevant for the evaluation) have not
been reported. (NTP. 2015).
Unacceptable
(score = 4)
Do not select for this metric.
12
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Confidence Level
(Score)
Description
Selected
Score
Not
rated/applicable
Do not select for this metric.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important elements
such as relevance]
Domain 4. Potential Confounding/Variable Control
Metric 9. Covariate Adjustment (confounding)
High
(score = 1)
For all studv tvves: ADDrooriatc adjustments or explicit considerations
were made for potential confounders (e.g. age, sex, SES, race, etc.)
(excluding co-exposures, which are evaluated in metric 11) in the final
analyses through the use of statistical models to reduce research-specific
bias, including matching, adjustment in multivariate models, stratification,
or other methods that were arorouriatelv iustified (NTP. 2015).
For Studies reporting SMRs or SIRs: Adjustments are described and
results are age-, race-, and sex-adjusted (or stratified) if applicable.
Medium
(score = 2)
For all studv tvves: There is indirect evidence that aDDroDriatc adjustments
were made (i.e., considerations were made for primary covariates
(excluding co-exposures) and potential confounders adjustment) without
providing a description of methods.
OR
The distribution of potential confounders (excluding co-exposures) did not
differ significantly between exposure groups or between cases and controls.
OR
The major potential confounders (excluding co-exposures) were
appropriately adjusted (e.g., SMRs, SIRs, etc.) and any not adjusted for are
considered not to appreciably bias the results (e.g., smoking rates in an
occupational cohort are expected to be generally similar in different
departments and thus confounding by smoking is unlikely when internal
analyses are applied).
For Studies reporting SMRs or SIRs: Indirect evidence that results are ase.
sex-, and race-adjusted (or stratified) if applicable.
Low
(score = 3)
For all studv tvves: There is indirect evidence (i.e.. no description is
provided in the study) that considerations were not made for potential
confounders adjustment in the final analvses (NTP. 2015).
AND
The distribution of primary covariates (excluding co-exposures) and
potential confounders was not reported between the exposure groups or
between cases and controls (NTP. 2015).
For Studies reporting SMRs or SIRs: Results are ase-. race-. OR sex-
adjusted (or stratified) if applicable (i.e., if 2 or all should have been
adjusted).
Unacceptable
(score = 4)
For all studv tvves: The distribution of potential confounders differed
significantly between the exposure groups.
AND
Confounding was demonstrated and was not appropriately adjusted for in
the final analvses (NTP. 2015).
For Studies reporting SMRs or SIRs: No discussion of adjustments.
Results are not adjusted for age, sex, and race (or stratified) if applicable.
Not
rated/applicable
Do not select for this metric.
13
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Confidence Level
(Score)
Description
Selected
Score
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important elements
such as relevance]
Metric 10. Covariate Characterization (measurement/information, confounding biases)
High
(score = 1)
For all studv tvves: Potential confounders (e.e. aee. sex. SES. race, etc.)
and were assessed using valid and reliable methodology where appropriate
(e.g., validated questionnaires, biomarker).
Medium
(score = 2)
For all studv tvves: A less-established method was used to assess
confounders (excluding co-exposures) and no method validation was
conducted against well-established methods, but there was little to no
evidence that that the method had poor validity and little to no evidence of
confounding.
Low
(score = 3)
For all studv tvves: The confounder assessment method is an insensitive
instrument or measure or a method of unknown validity.
Unacceptable
(score = 4)
For all studv tvves: Confounders were assessed using a method or
instrument known to be invalid.
Not
rated/applicable
Covariates were not assessed.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important
elements such as relevance]
Metric 11. Co-exposure Confounding (measurement/information, confounding biases)
High
(score = 1)
Do not select for this metric.
Medium
(score = 2)
For all studv tvves: Anv co-exoosures to Dollutants that are not the tarset
exposure that would likely bias the results were not likely to be present.
OR
Co-exposures to pollutants were appropriately measured or either directly or
indirectly adjusted for.
Low
(score = 3)
For cohort and cross-sectional studies: There is direct evidence that there
was an unbalanced provision of additional co-exposures across the primary
study groups, which were not appropriately adjusted for.
For case-control studies: There is direct evidence that there was an
unbalanced provision of additional co-exposures across cases and controls,
which were not appropriately adjusted for, and significant indication a
biased exposure-outcome association.
Unacceptable
(score = 4)
Do not select for this metric.
Not
rated/applicable
Enter 'NA' and do not score this metric.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important elements
such as relevance]
Domain 5. Analysis
Metric 12. Study Design and Methods
High
(score = 1)
Do not select for this metric.
Medium
(score = 2)
For all studv tvves: The studv desisn chosen was aDDroDriatc for the
research question.
AND
14
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Confidence Level
(Score)
Description
Selected
Score
The study uses an appropriate statistical method to address the research
question(s) (e.g.. Cox and Poisson regression for cohort studies and logistic
regression analysis for case-control studies.
Low
(score = 3)
Do not select for this metric.
Unacceptable
(score = 4)
For all studv tvves: The studv desisn chosen was not aDDroDriatc for the
research question.
OR
Inappropriate statistical analyses were applied to assess the research
questions.
Not
rated/applicable
Do not select for this metric.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important elements
such as relevance]
Metric 13. Statistical power (sensitivity)
High
(score = 1)
Do not select for this metric.
Medium
(score = 2)
For cohort and cross-sectional studies: The number of Darticioants are
adequate to detect an effect in the exposed population and/or subgroups of
the total population.
OR
The paper reported statistical power high is enough (> 80%) to detect an
effect in the exposure population and/or subgroups of the total population.
For case-control studies: The number of cases and controls are adeauate to
detect an effect in the exposed population and/or subgroups of the total
population.
OR
The paper reported statistical power was high enough (> 80%) to detect an
effect in the exposure population and/or subgroups of the total population.
Low
(score = 3)
Do not select for this metric.
Unacceptable
(score = 4)
For cohort and cross-sectional studies: The number of Darticioants is
inadequate to detect an effect in the exposed population and/or subgroups of
the total population and the study was negative.
For case-control studies: The number of cases and controls are inadeauate
to detect an effect in the exposed population and/or subgroups of the total
population and the study was negative.
Not
rated/applicable
Do not select for this metric.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important elements
such as relevance]
Metric 14. Reproducibility of analyses [adapted from Blettner et al. (2001)1
High
(score = 1)
Do not select for this metric.
Medium
(score = 2)
For all studv tvves: The description of the analvsis is sufficient to
understand precisely what has been done and to be conceptually
reproducible with access to the analytic data.
Low
For all studv tvves: The description of the analysis is insufficient to
15
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Confidence Level
(Score)
Description
Selected
Score
(score = 3)
understand what has been done and to be reproducible OR a description of
analyses are not present (e.g., statistical tests and estimation procedures
were not described, variables used in the analysis were not listed,
transformations of continuous variables (e.g. logarithmic) were not
explained, rules for categorization of continuous variables were not
presented, exclusion of outliers was not elucidated and how missing values
are dealt with was not mentioned).
Unacceptable
(score = 4)
Do not select for this metric.
Not
rated/applicable
Do not select for this metric.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important
elements such as relevance]
Metric 15. Statistical Models (confounding bias)
High
(score = 1)
Do not select for this metric.
Medium
(score = 2)
For all studv tvves: The model or method for calculating the risk estimates
(e.g., odds ratios, SMRs, SIRs) is transparent (it is stated how/why variables
were included or excluded)
AND
Model assumptions were met.
Low
(score = 3)
For all studv tvves: The statistical model buildins process is not fullv
appropriate
OR
Model assumptions were not met
OR
A description of analvses is not present TSTROBE Checklist 12e (Von Elm
et al.. 2008)].
Unacceptable
(score = 4)
Do not select for this metric.
Not
rated/applicable
Enter 'NA' if the study did not use a statistical model.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important
elements such as relevance]
16
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
1.3. Evaluation Criteria for Epidemiological Studies: Asbestos
Exposure and Mesothelioma Health Outcome
Confidence Level
(Score)
Description
Selected
Score
Domain 1. Study Participation
Metric 1. Participant selection (selection, performance biases)
Instructions: To meet criteria for confidence ratings for metrics where 'AND' is included, studies must
address both conditions where 'AND' is stipulated. To meet criteria for confidence ratings for metrics
where 'OR' is included studies must address at least one of the conditions stipulated.
High
(score = 1)
For all studv tvves: All kev elements of the studv desisn are rcDortcd (e.e..
setting, participation rate described at all steps of the study, inclusion and
exclusion criteria, and methods of participant selection or case
ascertainment)
AND
The reported information indicates that selection in or out of the study (or
analysis sample) and participation was not likely to be biased (i.e., the
exposure-outcome distribution of the participants is likely representative of
the exposure-outcome distributions in the population of persons eligible for
inclusion in the study).
Medium
(score = 2)
For all studv tvves: Some kev elements of the studv design were not
present but available information indicates a low risk of selection bias (i.e.,
the exposure-outcome distribution of the participants is likely representative
of the exposure-outcome distributions in the population of persons eligible
for inclusion in the study).
Low
(score = 3)
For all studv tvves: Kev elements of the studv desisn and information on
the population (e.g., setting, participation rate described at most steps of the
study, inclusion and exclusion criteria, and methods of participant selection
or case ascertainment) are not reported [STROBE checklist 4, 5 and 6 (Von
Elm etal.. 2008)1.
Unacceptable (score
= 4)
For all studv tvves: The reported information indicates that selection in or
out of the study (or analysis sample) and participation was likely to be
significantly biased (i.e., the exposure-outcome distribution of the
participants is likely not representative of the exposure-outcome
distributions of the population of persons eligible for inclusion in the study).
Not rated/applicable
Do not select for this metric.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important
elements such as relevance]
Metric 2. Attrition (missing data/attrition/exclusion, reporting biases)
High
(score = 1)
For cohort studies: There was minimal subiect loss to follow lid durine the
study (or exclusion from the analysis sample) and outcome and exposure
data were largely complete
OR
Loss of subjects (e.g., incomplete outcome data) or missing exposure and
outcome data was adequately* addressed (as described below) and reasons
were documented when human subiects were removed from a studv (NTP.
2015).
AND
Missing data have been imputed using appropriate methods (e.g., multiple
imputation methods), and characteristics of subjects lost to follow up or
with unavailable records are not significantly different from those of the
study participants (NTP. 2015).
17
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Confidence Level
(Score)
Description
Selected
Score
For case-control studies and cross-sectional studies: There was minimal
subject withdrawal from the study (or exclusion from the analysis sample)
and outcome data and exposure were largely complete.
OR
Any exclusion of subjects from analyses was adequately* addressed (as
described below), and reasons were documented when subjects were
removed from the studv or excluded from analvses (NTP. 2015).
*NOTE for all studv tvves: Adeauate handline of subiect attrition can include:
Use of imputation methods for missing outcome and exposure data; reasons
for missing subjects unlikely to be related to outcome (for survival data,
censoring was unlikely to introduce bias); missing outcome data balanced in
numbers across study groups, with similar reasons for missing data across
groups.
Medium
(score = 2)
For cohort studies: There was moderate subiect loss to follow lid durine
the study (or exclusion from the analysis sample) or outcome and exposure
data were nearly complete.
AND
Any loss or exclusion of subjects was adequately addressed (as described in
the acceptable handling of subject attrition in the high confidence category)
and reasons were documented when human subjects were removed from a
study.
For case-control studies and cross-sectional studies: There was moderate
subject withdrawal from the study (or exclusion from the analysis sample),
but outcome and exposure data were largely complete
AND
Any exclusion of subjects from analyses was adequately addressed (as
described above), and reasons were documented when subjects were
removed from the study or excluded from analvses (NTP. 2015).
Low
(score = 3)
For cohort studies: The loss of subiects (e.e.. loss to follow lid.
incomplete outcome or exposure data) was moderate and unacceptably
handled (as described below in the unacceptable confidence category)
(Source: OH AT).
OR
Numbers of individuals were not reported at important stages of study
(e.g., numbers of eligible participants included in the study or analysis
sample, completing follow-up, and analyzed). Reasons were not provided
for non-participation at each staee TSTROBE Checklist Item 13 (Von Elm
etal.. 2008)1.
For case-control and cross-sectional studies: The exclusion of subiects
from analyses was moderate and unacceptably handled (as described
below in the unacceptable confidence category).
OR
Numbers of individuals were not reported at important stages of study
(e.g., numbers of eligible participants included in the study or analysis
sample, completing follow-up, and analyzed). Reasons were not provided
for non-participation at each staee TSTROBE Checklist Item 13 (Von Elm
et al.. 2008)].
Unacceptable (score
= 4)
For cohort studies: There was laree subiect attrition durine the studv (or
exclusion from the analysis sample).
OR
18
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Confidence Level
(Score)
Description
Selected
Score
Unacceptable handling of subject attrition: reason for missing outcome data
likely to be related to true outcome, with either imbalance in numbers or
reasons for missing data across study groups; or potentially inappropriate
application of imputation (Source: OH AT).
For case-control and cross-sectional studies: There was large subiect
withdrawal from the studv (or exclusion from the analysis sample).
OR
Unacceptable handling of subject attrition: reason for missing outcome data
likely to be related to true outcome, with either imbalance in numbers or
reasons for missing data across study groups; or potentially inappropriate
application of imputation.
Not rated/applicable
Do not select for this metric.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important
elements such as relevance]
Metric 3. Comparison Group (selection, performance biases) [See special instructions for mesothelioma
studies in "Not rated/applicable"!*
High
(score = 1)
For ALL studv tvves: Anv differences in baseline characteristics of groiros
were considered as potential confounding or stratification variables and
were thereby controlled by statistical analysis (Source: OHAT).
OR
For cohort and cross-sectional studies: Kev elements of the studv design
are reported (i.e., setting, inclusion and exclusion criteria, and methods of
participant selection), and indicate that groups were similar (e.g., recruited
from the same eligible population with the same method of ascertainment
and within the same time frame using the same inclusion and exclusion
criteria, and were of similar age and health status) (NTP. 2015).
For case-control studies: Kev elements of the studv design are reported
indicate that that cases and controls were similar (e.g., recruited from the
same eligible population with the number of controls described, and
eligibility criteria and are recruited within the same time frame (NTP.
2015).
For studies reporting Standardized Mortalitv Ratios (SMRs) or
Standardized Incidence Ratios (SLRs): Age. sex (if aDDlicablc). race (if
applicable), and calendar time adjustment or stratification is described and
choice of reference population (e.g., general population) is reported.
Medium
(score = 2)
For cohort studies and cross-sectional studies: There is onlv indirect
evidence (e.g., stated by the authors without providing a description of
methods) that groups are similar (as described above for the high
confidence rating).
For case-control studies: There is indirect evidence (i.e.. stated bv the
authors without providing a description of methods) that cases and controls
are similar (as described above for the high confidence rating).
For studies revortins SMRs or SLRs: Age. sex (if applicable), race (if
applicable), and calendar time adjustment or stratification is not specifically
described (i.e., indirect evidence) in the text, but results tables are stratified
by age, sex (if applicable), race (if applicable); choice of reference
population (e.g., general population) is reported.
Low
(score = 3)
For cohort and cross-sectional studies'. There is indirect evidence (i.e..
stated by the authors without providing a description of methods) that
groups were not similar (as described above for the high confidence rating).
19
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Confidence Level
(Score)
Description
Selected
Score
AND
Differences between the exposure groups are not adequately controlled for
in the statistical analysis.
For case-control studies: There is indirect evidence (i.e.. stated bv the
authors without providing a description of methods) that cases and controls
were not similar (as described above for the high confidence rating).
AND
The characteristics of cases and controls are not rcDortcd (Source: (NTP.
2015).
AND
Differences in groups is not adequately controlled for in the statistical
analysis.
For studies reporting SMRs or SIRs. Indirect evidence of a lack of
adjustment or stratification for age, sex (if applicable), race (if applicable),
and calendar time; or indirect evidence that choice of reference population
(e.g., general population) is inappropriate.
Unacceptable (score
= 4)
For cohort studies: Subiects in all cxoosurc urouDS were not similar
OR
(Information was not reported to determine if participant groups were
similar 1 STROBE Checklist 6 (Von Elm et al.. 2008)
AND
Potential differences in exposure groups were for a factor that was related to
the outcome and not controlled for in the statistical analysis.)
OR
(Subjects in the exposure groups had very different participation/response
rates (NTP. 2015).
AND
Participation rates were related to exposure and outcome.)
For case-control studies: (Controls were drawn from a verv dissimilar
DODiilation than cases or recruited within verv different time frames (NTP.
2015).
AND
Potential differences in the case and control groups were not controlled for
in the statistical analysis.)
OR
Rationale and/or methods for case and control selection, matching criteria
including number of controls per case (if relevant) were not reported
rSTROBE Checklist 6 (Von Elm et al.. 2008)1.
For cross-sectional studies: (Subiects in all cxoosurc aroiios were not
similar, recruited within very different time frames, or had very different
DarticiDation/rcsDonsc rates (NTP. 2015).
AND
Potential differences in exposure groups were not controlled for in the
statistical analysis.)
OR
Sources and methods of selection of participants in all exposure groups
were not rcDortcd 1 STROBE Checklist 6 (Von Elm et al.. 2008)1.
For studies reporting SMRs or SIRs: Lack of adiustment or stratification
for both age, sex (if applicable), race (if applicable), and calendar time; or
choice of reference population (e.g., general population) is not reported.
Not rated/applicable
For mesothelioma studies, a comparison population is not required, as
20
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Confidence Level
(Score)
Description
Selected
Score
EPA's interest is in the absolute risk and not the relative risk. All studies of
mesothelioma allowing for evaluation of absolute risk should be labeled as
"Not rated / not applicable"
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important elements
such as relevance]
Domain 2. Exposure Characterization
Metric 4. Measurement of Exposure (Detection/measurement/information, performance biases)
High
(score = 1)
For all studv tvves: Quantitative estimates of exposure were consistentlv
assessed (i.e., using the same method and sampling time-frame) during
multiple time periods and using either PCM or TEM.
OR
A combination of methods were used over time (i.e., midget impinger, PCM
or TEM), but side by side sampling and analyses were conducted to develop
appropriate conversion criteria.
AND
For an occupational population, contains detailed employment records and
quantitative estimates of exposure using either PCM or TEM which allows
for construction of job-matrix for entire work history of exposure (i.e..
Cumulative or peak exposures, and time since first exposure).
Medium
(score = 2)
For all studv tvves: (Exposure was assessed durine one time oeriod but this
time period is judged to be reasonably representative of the entire study
time period.
AND
Exposure was assessed using a combination of midget impingers, PCM
and/or TEM measurements, but side by side sampling and analyses were
not conducted for all operations and thus there is a lack of confidence in the
conversion factors.)
OR
For an occupational study population, contains detailed employment records
and quantitative estimates of exposure using a combination of midget
impingers and PCM or TEM for only a portion of participant's work history
of exposure (i.e., only early years or later years), such that extrapolation of
the missing years is required.
Low
(score = 3)
For all studv tvves: Exposure was estimated solelv usins professional
judgement.
OR
Exposure was directly measured (e.g., midget impinger) and assessed using
a quantitative method other than PCM or TEM and conversion factors were
not determined.
Unacceptable (score
= 4)
For all studv tvves: There was no Quantitative measure or estimate of
exposure.
OR
Methods used to quantify the exposure were not well defined, and sources
of data and detailed methods of exposure assessment were not reported
[STROBE Checklist 7 and 8 (Von Elm et al.. 2008)1.
OR
There is evidence of substantial exposure misclassification that would
significantly bias the results.
Not rated/applicable
Do not select for this metric.
Reviewer's
/Document concerns, uncertainties, limitations, and deficiencies and anv
21
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Confidence Level
(Score)
Description
Selected
Score
comments
additional comments that may highlight study strengths or important elements
such as relevance]
Metric 5. Exposure levels (Detection/measurement/information biases)
High
(score = 1)
Do not select for this metric.
Medium
(score = 2)
For all studv tvves: The ranse and distribution of exposure is sufficient or
adequate to develop an exposure-response estimate (Coooer et al.. 2016).
Low
(score = 3)
For all studv tvves: The ranse of exposure in the population is limited.
Unacceptable (score
= 4)
For all studv tvves: The ranse and distribution of exposure are not
sufficient or adequate to determine an exposure-response relationship
(Cooper et al.. 2016).
OR
No description is provided on the levels or ranse of exposure.
Not rated/applicable
Do not select for this metric.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important elements
such as relevance]
Metric 6. Temporality
High
(score = 1)
For all studv tvves: The studv presents an appropriate temporalitv between
exposure and outcome (i.e. the exposure precedes the disease).
AND
The interval between the exposure (or reconstructed exposure) and the
outcome is sufficiently long considering the latency of the disease (i.e.
studv follow-up is more than 20 vears for mesothelioma) (Lakind et al..
2014).
Medium
(score = 2)
For all studv tvves: Temporalitv is established, but it is unclear whether
there is adequate follow-up for consideration of latency (i.e., only 15-20
years of follow-up) (Lakind et al.. 2014).
Low
(score = 3)
For all studv tvves: The temporalitv of exposure and outcome is uncertain
(10-15 years).
OR
There is inadequate follow-up of the cohort considering the latency period.
Unacceptable (score
= 4)
For all studv tvves: Studv lacks an established time order, such that
exposure is not likelv to have occurred prior to outcome (Lakind et al..
2014).
OR
There was inadequate follow-up of the cohort for the expected latency
period (<10 years).
OR
Sources of data and details of methods of assessment were not sufficiently
reported (e.g. duration of follow-up, periods of exposure, dates of outcome
ascertainment, etc.) Source: STROBE Checklist 8 (Von Elm et al.. 2008).
Not rated/applicable
Do not select for this metric.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important elements
such as relevance]
Domain 3. Outcome Assessment
Metric 7. Outcome measurement or characterization (detection/measurement/information, performance,
reporting biases)
22
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Confidence Level
(Score)
Description
Selected
Score
High
(score = 1)
For all studv tvpes: The outcome was assessed using one or a
combination of the following well-established methods:
o Mesothelioma cases confirmed by histological or cytological means
(including subtypes of mesothelioma) and/or
o ICD-10 codes (3 digit) C45 or (4 digit) C45.x (C45.0. C45.1, C45.2,
('45.7. C45.9)
o All fields on the death certificates of cohort searched for
'mesothelioma'
o Appropriate Pre-ICD 10 codes supplemented by additional evidence
(e.g. pathology/autopsy) see Table 1 of (KodyIcy et al.. 2011)
Medium
(score = 2)
For all studv tvves: Examined death certificates searched for
mesothelioma for pre-ICD-10 codes that include pleura, peritoneum and
site unspecified (ICD code 199)
Low
(score = 3)
Do not select for this metric.
Unacceptable (score
= 4)
For all studv tvves: Numbers of outcome events or summary measures
were not reported (Source: STROBE Checklist 15 (Von Elm et al.. 2008)
OR
Only pre ICD-10 codes (without additional information) were used for
ascertainment of mesothelioma.
OR
Examined death certificates searched for mesothelioma for codes that
included only pleura and/or peritoneum
OR
Study lacks individual assessment of mesothelioma (i.e. mesothelioma is
assessed as a combination with other cancer types, excluding lung and
bronchus or trachea)
OR
Any self-reported information
Not rated/applicable
Do not select for this metric.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important elements
such as relevance]
Metric 8. Reporting
Bias
High
(score = 1)
For all studv tvves: Mesothelioma findings are reported in the abstract,
results or discussion. Effect estimates are reported with confidence
intervals and/or standard errors, number of cases/controls or
exposed/unexposed reported for each analysis, to be included in exposure-
response analysis or fully tabulated during data extraction and analyses
(NTP. 2015).'
Medium
(score = 2)
For all study types: All of the study's findings (primary and secondary)
outlined in the abstract, results or discussion (that are relevant for the
evaluation) are reported, but not in a way that would allow for detailed
extraction (e.g., results were discussed in the text but accompanying data
were not shown).
Low
(score = 3)
For all studv tvves: Mesothelioma outcomes outlined in the methods,
abstract, and/or introduction (that are relevant for the evaluation) have not
been reported. (NTP. 2015).
Unacceptable (score
= 4)
Do not select for this metric.
Not rated/applicable
Do not select for this metric.
23
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Confidence Level
(Score)
Description
Selected
Score
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important elements
such as relevance]
Domain 4. Potential Confounding/Variable Control
Metric 9. Covariate Adjustment (confounding) [See special instructions for mesothelioma studies in "Not
rated/applicable"! *
High
(score = 1)
For all studv tvves: AoDroDriatc adjustments or explicit considerations
were made for potential confounders (e.g. age, sex, SES, race, etc.)
(excluding co-exposures, which are evaluated in metric 11) in the final
analyses through the use of statistical models to reduce research-specific
bias, including matching, adjustment in multivariate models, stratification,
or other methods that were aroronriatelv iustified (NTP. 2015)..
For Studies reporting SMRs or SIRs: Adjustments are described and
results are age-, race-, and sex-adjusted (or stratified) if applicable.
Medium
(score = 2)
For all studv tvves: There is indirect evidence that aDDroDriatc adjustments
were made (i.e., considerations were made for primary covariates
(excluding co-exposures) and potential confounders adjustment) without
providing a description of methods.
OR
The distribution of potential confounders (excluding co-exposures) did not
differ significantly between exposure groups or between cases and controls.
OR
The major potential confounders (excluding co-exposures) were
appropriately adjusted and any not adjusted for are considered not to
appreciably bias the results (e.g., smoking rates in an occupational cohort
are expected to be generally similar in different departments and thus
confounding by smoking is unlikely when internal analyses are applied).
For Studies reporting SMRs or SIRs: Indirect evidence that results are ase.
sex-, and race-adjusted (or stratified) if applicable.
Low
(score = 3)
For all studv tvves: There is indirect evidence (i.e.. no description is
provided in the study) that considerations were not made for potential
confounders adjustment in the final analvses (NTP. 2015).
AND
The distribution of primary covariates (excluding co-exposures) and
potential confounders was not reported between the exposure groups or
between cases and controls (NTP. 2015).
For Studies reporting SMRs or SIRs: Results are ase-. race-. OR sex-
adjusted (or stratified) if applicable (i.e., if 2 or all should have been
adjusted).
Unacceptable (score
= 4)
For all studv tvves: The distribution of potential confounders differed
significantly between the exposure groups.
AND
Confounding was demonstrated and was not appropriately adjusted for in
the final analvses (NTP. 2015).
For Studies reporting SMRs or SIRs: No discussion of adjustments.
Results are not adjusted for age, sex, and race (or stratified) if applicable.
Not rated/applicable
For mesothelioma studies, evaluations of potential confounders are not
required as there are few other causes of mesothelioma (zeolites, viruses,
therapeutic or diagnostic radiation) and none that are likely to be correlated
in a dose-dependent manner with asbestos. Evaluation of potential
24
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Confidence Level
(Score)
Description
Selected
Score
confounding in mesothelioma studies should be labeled as "Not
rated/applicable".
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important
elements such as relevance]
Metric 10. Covariate Characterization (measurement/information, confounding biases) [See special
instructions for mesothelioma studies in "Not rated/applicable"!*
High
(score = 1)
For all studv tvves: Potential confounders (e.e. aee. sex. SES. race, etc.)
were assessed using valid and reliable methodology where appropriate (e.g.,
validated questionnaires, biomarker).
Medium
(score = 2)
For all studv tvves: A less-established method was used to assess
confounders (excluding co-exposures) and no method validation was
conducted against well-established methods, but there was little to no
evidence that that the method had poor validity and little to no evidence of
confounding.
Low
(score = 3)
For all studv tvves: The confounder assessment method is an insensitive
instrument or measure or a method of unknown validity.
Unacceptable (score
= 4)
For all studv tvves: Confounders were assessed usins a method or
instrument known to be invalid.
Not rated/applicable
Covariates were not assessed.
OR
Metric 9 is rated "Not applicable"
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important
elements such as relevance]
Metric 11. Co-exposure reliability (measurement/information, confounding biases) [See special instructions
for mesothelioma studies in "Not rated/applicable"!*
High
(score = 1)
Do not select for this metric.
Medium
(score = 2)
For all studv tvves: Anv co-exoosures to Dollutants that are not the tarset
exposure that would likely bias the results were not likely to be present.
OR
Co-exposures to pollutants were appropriately measured or either directly or
indirectly adjusted for.
Low
(score = 3)
For cohort and cross-sectional studies: There is direct evidence that there
was an unbalanced provision of additional co-exposures across the primary
study groups, which were not appropriately adjusted for.
For case-control studies: There is direct evidence that there was an
unbalanced provision of additional co-exposures across cases and controls,
which were not appropriately adjusted for, and significant indication a
biased exposure-outcome association.
Unacceptable (score
= 4)
Do not select for this metric.
Not rated/applicable
For mesothelioma, there are no established risk factors other than exposure
to asbestos, therefore no known co-exposures are of concern. Enter 'NA'
and do not score this metric.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important elements
such as relevance]
Domain 5. Analysis
Metric 12. Study Design and Methods
25
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Confidence Level
(Score)
Description
Selected
Score
High
(score = 1)
Do not select for this metric.
Medium
(score = 2)
For all studv tvves: The studv desisn chosen was aDDroDriatc for the
research question.
AND
The study uses an appropriate statistical method to address the research
question(s) (e.g.. Cox and Poisson regression for cohort studies, logistic
regression analysis for case-control studies.
Low
(score = 3)
Do not select for this metric.
Unacceptable (score
= 4)
For all studv tvves: The studv desisn chosen was not aDDroDriatc for the
research question.
Not rated/applicable
Do not select for this metric.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important elements
such as relevance]
Metric 13. Statistical power (sensitivity) [See special instructions for mesothelioma studies in "Not
rated/applicable"! *
High
(score = 1)
Do not select for this metric.
Medium
(score = 2)
For cohort and cross-sectional studies: The number of participants are
adequate to detect an effect in the exposed population and/or subgroups of
the total population.
OR
The paper reported statistical power high is enough (> 80%) to detect an
effect in the exposure population and/or subgroups of the total population.
For case-control studies: The number of cases and controls are adeauate to
detect an effect in the exposed population and/or subgroups of the total
population.
OR
The paper reported statistical power was high enough (> 80%) to detect an
effect in the exposure population and/or subgroups of the total population.
Low
(score = 3)
Do not select for this metric.
Unacceptable (score
= 4)
For cohort and cross-sectional studies: The number of participants is
inadequate to detect an effect in the exposed population and/or subgroups of
the total population and the study was negative.
For case-control studies: The number of cases and controls are inadeauate
to detect an effect in the exposed population and/or subgroups of the total
population and the study was negative.
Not rated/applicable
For mesothelioma, EPA is primarily interested in the presentation of data
collected in the study, rather than the statistical analysis. EPA will pool data
across asbestos studies to conduct for the analysis of mesothelioma risk.
Therefore, the power of individual studies will not be considered. This
metric may be marked as not rated/applicable.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important elements
such as relevance]
Metric 14. Rcnroducibilitv of analvses lad anted from Blettner et al. (2001)1 [See SDecial instructions for
mesothelioma studies in "Not rated/applicable"!*
High
Do not select for this metric.
26
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Confidence Level
(Score)
Description
Selected
Score
(score = 1)
Medium
(score = 2)
For all studv tvves: The description of the analvsis is sufficient to
understand precisely what has been done and to be conceptually
reproducible with access to the analytic data.
Low
(score = 3)
For all studv tvves: The description of the analvsis is insufficient to
understand what has been done and to be reproducible OR a description of
analyses are not present (e.g., statistical tests and estimation procedures
were not described, variables used in the analysis were not listed,
transformations of continuous variables (e.g. logarithmic) were not
explained, rules for categorization of continuous variables were not
presented, exclusion of outliers was not elucidated and how missing values
are dealt with was not mentioned).
Unacceptable (score
= 4)
Do not select for this metric.
Not rated/applicable
For mesothelioma, EPA is primarily interested in the presentation of data
collected in the study, rather than the statistical analysis. If individual data
elements (e.g., time since first exposure, number of person-years, etc.) are
present in the study that will allow EPA to conduct its own analysis, this
metric may be marked as not rated/applicable.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important
elements such as relevance]
Metric 15. Statistical Models (confounding bias) [See special instructions for mesothelioma studies in "Not
rated/applicable"! *
High
(score = 1)
Do not select for this metric.
Medium
(score = 2)
For all studv tvves: The model or method for calculating the risk estimates
(e.g., odds ratios, SMRs, SIRs) is transparent (it is stated how/why variables
were included or excluded).
Low
(score = 3)
For all studv tvves: The statistical model buildins process is not fullv
appropriate
OR
Model assumptions were not met
OR
A description of analvses is not dresent 1 STROBE Checklist 12e (Von Elm
et al.. 2008)].
Unacceptable (score
= 4)
Do not select for this metric.
Not rated/applicable
For mesothelioma, EPA is primarily interested in the presentation of data
collected in the study, rather than the statistical analysis. If individual data
elements (e.g., time since first exposure, number of person-years, etc.) are
present in the study that will allow EPA to conduct its own analysis, this
metric may be marked as not rated/applicable.
Reviewer's
comments
[Document concerns, uncertainties, limitations, and deficiencies and any
additional comments that may highlight study strengths or important
elements such as relevance]
27
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
2. Data Evaluation of Lung Cancer Studies
2.1. Table of studies evaluated for asbestos exposure and lung cancer
incidence
Study Cohort
Author, Year
HERO ID
South Carolina,
US
(Berman and Crump, 2008)
626405
(Brown et al., 1994)
3081832
(Cole etal., 2013)
3078261
(Dement et al., 1983b)
67
(Dement and Brown, 1994a)
3094565
(Dement et al., 1994)
3081766
(Dement and Brown, 1994b)
3081783
(Edwards et al., 2014)
3078061
Million el ill.. Zii|2)
I247KM
(Hein et al., 2uu7)
709498
(Loomis et al., 2012)
1257856
(SRC. 2019c)
5080236
(Stavner et al., 1997)
3081241
(Stavner et al., 2008)
2604140
Qinghai, China
- miners
(Wane etal.. 2012)
2572504
(Wane etal.. 2013)
2548289
(Wang et al., 2014)
2538846
Balangero, Italy
(Piolatto et al., 1990)
3082492
(Pira et al., 2009)
2592425
(Pira et al., 2017)
5060134
(Rubino et al., 1979)
178
North Carolina,
US
(Berman and CrumD, 2008)
626405
(Dement et al., 2008)
626406
Million el al.. 2<)I2)
I247KM
(Loomis et al., 2009)
3079232
(Loomis et al., 2010)
2225695
(Loomis et al., 2012)
1257856
M.oomis el ill . 2<)|1))
5160027
(SRC, 2019a)
5080241
Salonit
Anhovo,
Slovenia
(Dodic Fikfak. 2003)
3080279
(Dodic Fikfak et al., 2007)
3079664
28
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study Cohort
Author, Year
HERO ID
Quebec,
Canada
(Berman and Cruma 2008)
626405
(Gibbs and Lachance, 1972)
3580825
(Liddell et al., 1997)
3081408
(Liddell et al., 1998)
3081200
(Liddell and Armstrong, 2002)
3080504
(Mcdonald et al.. 1993a)
3081910
(Mcdonald et al., 1993b)
3081911
(SRC. 2019b)
5080232
(Vacek, 1998)
3081118
Chongqing,
China -
asbestos
products factory
including
textiles
(Courtice et al., 2016)
3520560
(Deng et al., 2012)
2573093
(Wang et al., 2014)
2538846
(Yano et al., 2001)
3080569
Shaded rows indicate studies used for derivation of Inhalation Unit Risk (IUR).
29
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
2.2. Data Evaluation Scoring Sheets: Lung Cancer Outcome
2.2.1. Epidemiology evaluation results of the South Carolina, US
cohort studies on asbestos exposure and lung cancer incidence
Study
This cohort evaluation represents all identified publications pertaining to the South Carolina cohort of textile
reference:
workers. A full list of related naners reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Weighting
Factor
Weighted
Score
a. Setting and methods of case ascertainment were
rcDortcd in (Dement et al.. 1983b) v. 422 and (Hcin
et al.. 2007) p 617. Participant selection and
inclusion/exclusion criteria varied by study and
analysis. The initial cohort consisted of white men
a
e
o
employed for at least one month in a production job
.&
o
at the South Carolina plant between 1/1/1940 and
12/31/1965 ((Dement et al.. 1983b) d 422).
"H
es
Subsequent analyses added non-white men and/or
women ((Stavner et al.. 19971 (Dement et al.. 1994)
High
i
0.400
0.400
O
1
CLh
(Brown et al.. 19941 (Elliott et al.. 2012). (Edwards
5
et al.. 2014s). (Cole et al.. 20131 (Hein et al.. 2007V).
Xfl
b. Selection in or out of the study was based on 1)
employment in production job during designated
time frame and b) availability of necessary data
(birth and hire dates; work history; vital status).
These criteria are unlikely to result in biased subject
participation.
30
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This cohort evaluation represents all identified publications pertaining to the South Carolina cohort of textile
workers. A full list of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
2. Attrition
Attrition/missing data exclusions were reported in
(Dement et al.. 1983b) (a 423 and Table 1) and
(Hem et al.. 2007) (o 618 and Table Is). The number
of eligible workers and number excluded due to
incomplete work histories was not reported in the
sources reviewed. (Dement et al.. 1983b) (d 422)
reports that each worker was assigned a card at hire
on which was tracked the date of birth, sex, race,
and SSN, and job or department changes throughout
the career were recorded on the card; these were
used in the cohort establishment. The suggestion is
that all cohort members had complete work
histories: however. (Hein et al.. 2007) (d 624)
reported that the study was limited by incomplete
lifetime work histories. Vital status was unknown
for 2.1% of the original 1261 cohort members, and
cause of death was unknown for 5.5% of the deaths
in 1975. At the 2001 follow uo ((Hein et al.. 2007)).
vital status was unknown for 8.6% of the larger
cohort of 3072, and cause of death was unknown for
3.9%. The latter paper cited high rate of loss to
follow-up as a limitation of the study. Thus, loss to
follow up is judged to be moderate, and the lack of
information on number eligible and number with
incomplete work histories leads to low confidence.
Low
3
0.400
1.200
3. Comparison Group
Any differences in baseline characteristics (e.g.,
age, sex, race) were controlled by statistical analysis
((Dement et al.. 1983b). v. 422: (Hein et al.. 2007).
v> 617). In (Hein et al.. 2007) (u 617). it is reported
that birth cohort was used in statistical analysis as a
surrogate for smoking. Setting, inclusion and
exclusion criteria, and methods of participant
selection are reported ((Dement et al.. 1983b). v.
422 and (Hein et al.. 2007) o 617). and these sussest
that the groups were recruited from the same
eligible population with the same method of
ascertainment and within the same time frame using
the same inclusion and exclusion criteria.
High
1
0.200
0.200
31
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This cohort evaluation represents all identified publications pertaining to the South Carolina cohort of textile
workers. A full list of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
a
#o
¦B
t3
4. Measurement of Exposure
(Dement et al.. 1983a) describes the plant processes
and the exposure estimation methods. More than
6000 air samples obtained at the plant from 1930 to
1975 were analyzed by PCM to yield chrysotile
concentrations (fibers >5 um/mL: (Hein et al..
20071 p 617). Exposure concentrations were
estimated by department, job, and time period;
individual cumulative exposure assessed using the
modeled concentrations and JEM (methods outlined
in (Dement et al.. 1983a)). In 2008. an updated JEM
was developed to estimate fiber size-specific
exposure estimates (based on TEM analysis of
archived samples) in (Dement et al.. 2008).
High
i
0.400
0.400
cS
~-
es
-fl
U
£
o
o.
UJ
5. Exposure
levels
Exposure-response relationships were developed
(see Fieure 2 and Table VIII of (Dement et al..
1983b). Table 3 of (Hein et al.. 2007). (Elliott et al..
2012) see Table 2. A total of 6 cumulative exposure
levels are analvzed in Table 3 of (Hein et al.. 2007).
Medium
2
0.200
0.400
6. Temporality
Temporality was established (exposure preceded
death). Exposure response analysis in initial cohort
((Dement et al.. 1983b)) was restricted to
individuals with at least 15 years follow up since
first employment (p 426). The longest follow-up
time for the cohort was at least 36 years (1965-
2001; (Hein et al.. 2007) p 617). A ten vear las time
was used in the analvses in (Hein et al.. 2007) (p
617). In (Elliott et al.. 2012). the assessment of the
SC cohort also used a ten year lag time (Table 2).
High
1
0.400
0.400
Outcome Assessment
7. Outcome measurement or
characterization
Lung cancer deaths (underlying and contributing
cause) were determined from the National Death
Index Plus (1979 and later; (Hein et al.. 2007) p
617) or death certificates (before 1979; sources of
certificates not specified; (Dement et al.. 1983b) p
422). ICD in effect at time of death was used
((Dement et al.. 1983b) p 422). Deaths before 1979
were coded manually by a nosologist. ICD codes
162 and 163 (trachea, bronchus, and lung) were
considered luns cancers ((Dement et al.. 1983b)
Tables II, III, IV, VIII, XI).
High
1
0.667
0.667
32
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
This cohort evaluation represents all identified publications pertaining to the South Carolina cohort of textile
reference:
workers. A full list of related naners reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Weighting
Factor
Weighted
Score
Lung cancer deaths and person-years at risk by
in
exposure category are reported in Table 3 of (Hcin
m
et al.. 2007): luns cancer cases and person-vears at
s?
risk are reported in Table 1 of ffilliott et al.. 2012).
Rate ratio estimates are reported with CIs in (Hcin
High
i
0.333
0.333
Oh
et al.. 2007) and ffilliott et al.. 2012): coefficients
&
for Poisson regression are reported without SE in
00
Table 2 of ffilliott et al.. 2012). but rate ratios with
CIs are also reported for the same models.
Analyses in (Hcin et al.. 2007) and ffilliott et al..
2012) were adjusted for age. sex. race, decade of
"H
follow-up and birth cohort. No adjustment was
I
made for smokine in (Dement et al.. 1983b). which
=3
could bias SMR analyses. However, the authors of
"o
'-a'
<-
(Dement et al.. 1983b) used available information
c
&
on smoking rates among cohort members to
Medium
2
0.500
1.000
o
U
a
compare with rates in U.S. white males; patterns
cd
>
(Table XID were similar between the groups. (Hein
cS
"5
1
U
et al.. 2007) and ffilliott et al.. 2012) evaluated birth
as
cohort as a surrogate for smoking; lack of direct
consideration of smoking is not likely to bias
C
internal analysis in an occupational cohort.
C
5
C
(Dement et al.. 1983b) (p 422) reports that each
£
o
3
worker was assigned a card at hire on which was
©
u
cd
N
included the date of birth, sex, and race, presumably
o
O
o
collected on US Public Health Service
questionnaires in 1964 and 1971 and from medical
records were used to estimate patterns of smoking in
the cohort and compared with US White males
(comparison group for SMR analyses). Data were
not available for all cohort members.
33
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This cohort evaluation represents all identified publications pertaining to the South Carolina cohort of textile
workers. A full list of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
11. Co-exposure Confounding
Small amounts of crocidolite yarn were used
between 1950s and 1975, but the total quantity was
reported to be small (2000 lbs total vs 6-8 million
lbs/yr of chrysotile during that time period). In
addition, workers did not card, spun, or twist the
crocidolite; a single loom was used; and weaving
was performed wet, which minimized exposure to
crocidolite ((Hein et al.. 2007) d 616). The
distribution of this co-exposure relative to chrysotile
exposure was not evaluated, and no effort was made
to adjust for this co-exposure. However, available
information suggests the coexposure would be
negligible relative to chrysotile exposures.
Medium
2
0.250
0.500
12.
Study
Design
and
Study design was retrospective cohort; Poisson
regression used for internal analvses in (YHein et
al.. 2007) o 617) and (Elliott et al.. 2012) d 386.
Medium
2
0.400
0.800
13.
Statistical
power
None of the related studies report power
calculations; however, statistically significant
exposure-response relationships suggest there was
adequate power to detect the effect.
Medium
2
0.200
0.400
Analysis
14. Reproducibility of analyses
Analvsis description in (Hcin et al.. 2007) aooears
to be complete. Statistical tests and estimation
procedures and variables considered are reported
((Hcin et al.. 2007) v 617). For internal analvses.
cumulative exposure was treated as a continuous
variable and/or partitioned into categories with
aooroxi match eaual numbers of deaths ((Hcin et al..
2007) d 617 and footnote to Table 3). There were no
variable transformations, outlier exclusions, or
imputation of missing values. Analysis description
in (Elliott et al.. 2012) aooears to be complete.
Statistical tests and estimation procedures and
variables considered are reported (Elliott et al..
2012) v 386. Cumulative exposure was analvzed as
a continuous variable (Elliott et al.. 2012). v 386.
No variable transformations were reported.
Medium
2
0.200
0.400
15.
Statistical
Models
Statistical models and methods are described in
detail later publications (see (Hein et al.. 2007) d
617-618 and (Elliott et al.. 2012) v> 386. including
how variables were included or excluded.
Medium
2
0.200
0.400
Sum of scores:
5
8
High: >=1 and <1.7
Medium: >=1.7 and <2.3
Low: >=2.3 and <=3
Overall Score = Sum of Weighted Scores/Sum of Metric
Weighting Factors:
1.6
Overall
Score:
Nearest
tenth:
1.6
34
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This cohort evaluation represents all identified publications pertaining to the South Carolina cohort of textile
workers. A full list of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
Overall Quality Level:
High
Study
Quality
Comment:
The reviewer agreed with this study's overall quality level.
35
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
2.2.2. Epidemiology evaluation results of the Chongqing, China
mining cohort studies on asbestos exposure and lung cancer
incidence
Study
reference:
This cohort evaluation represents all identified publications pertaining to the cohort of miners in Chongqing,
China. A full list of related naners reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
1. Participant
selection
The key elements of the study design are reported
((Wane et al.. 2013) a 2. and (Wane et al.. 2012) a
20). All male workers employed for at least one
year in the chrysotile mine were included in the
study (participation is not likely to be biased).
High
i
0.400
0.400
a
#o
.2-
2. Attrition
There was no subject loss to follow up during the
study; outcome and exposure data were complete
((Wane et al.. 2013) v>. 2. and (Wans et al.. 2012) a
20).
High
i
0.400
0.400
"H
es
a.
¦a
5
Tfl
3. Comparison Group
Inclusion criteria and the methods of participant
selection were reported. All subjects were recruited
from the same eligible population within the same
time frame. In studies reoortine SMRs ((Warm et
al.. 2012) a 407. and (Wane et al.. 2013) a 3). the
choice of a reference population is reported (based
on age-specific national mortality data for males).
However, data on cause-specific mortality data were
limited (rates of 1990 and 2004 were used to
correspond to periods of 1981-1995 and 1996-2006,
respectively ((Wans et al.. 2012) p. 411).
High
i
0.200
0.200
36
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This cohort evaluation represents all identified publications pertaining to the cohort of miners in Chongqing,
China. A full list of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
Exposure Characterization
4. Measurement of Exposure
Periodic asbestos dust measurements were
available from 1984 to 1995 ((Warm et al.. 2013)
p.2). In 2006, additional measurements in various
workshops were Dcrformcd ((Warm et al.. 2013) v.2.
(Wans et al.. 2014) a 120. and (Wans et al.. 2012)
p. 406); these samples were also analyzed by TEM.
Paired samples from 1991 (using simultaneous
gravimetric and membrane filter methods) from the
main workshop only were used to define the
relationship between dust and fiber concentrations
((Wans et al.. 2013) v. 2). From these data, (and
using all periodically measured data at different
workshops), average fiber concentrations by
workshop/job were calculated. Cumulative
individual exposures were estimated as the product
of (fiber concentration at a specific workshop/job) x
(duration of iob) ((Wans et al.. 2013) o. 3 and
(Wans et al.. 2014) v. 120). Side-bv-side analyses
were not conducted for all operations or at all time
points (i.e. systematic dust/fiber data were not
available: (Wans et al.. 2012) o. 409). There were
no exposure data prior to 1984. The study authors
acknowledge that there may have been exposure
misclassification based on these estimations, but the
misclassification was likely to be non-differential
((Wans et al.. 2013) a 7 and (Wans et al.. 2014) a
123).
Medium
2
0.400
0.800
5. Exposure
levels
The range and distribution of exposure is sufficient
to develop and exposure-response estimate. The
Wans et al. 2014 studv ((Wans et al.. 2014) v>. 122)
reports 4 levels of exposure (referent + 3);
cumulative exposures were categorized quartiles for
analyses of lung cancer deaths.
Medium
2
0.200
0.400
6. Temporality
The study establishes appropriate temporality; the
interval between exposure and outcome is long
enoush considerins latency of the disease ((Wans et
al.. 2013) t). 1. (Wans etal.. 2014) t>. 119. and
(Wans et al.. 2012) a 406). The cohort was
followed for 26 years.
High
1
0.400
0.400
37
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This cohort evaluation represents all identified publications pertaining to the cohort of miners in Chongqing,
China. A full list of related papers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
C
O
(J
5
o
1 o
3 .a
CD *-h
C cd
o ^
The Wang et al. 2013 study ((Wang et al.. 20131 p.
2) indicates that causes of death were obtained from
hospitals and verified with the death registry; the
study indicated that "there are consistent diagnostic
criteria for cancers in China, largely based on
clinical manifestations and pathological
confirmation or biopsy." The study cites that SMRs
for "lung cancer" included cancers of the lung,
trachea, bronchus, and other thoracic neoplasm,
encompassing ICD-10 C37 and ICD-10 C38 in
addition to ICD-10 C34 (lung and bronchus) and
C33 (trachea). In the Wang et al. 2014 publication
((Wang et al.. 2014) p. 120), ICD codes
corresponding to lung cancer were not provided.
Medium
0.667
1.333
m
ef
o
a
-w
o
a.
S3
y
S3
o
O
In studies reporting SMRs ((Wang et al.. 2013) p.3
and (Wang et al.. 2012) p. 407), final analyses were
adjusted for smoking, age at entry, and/or
employment years. In the later study ((Wang et al..
2014) p. 121), the final risk estimate model was
adjusted for age and smoking. There were no
adjustments for sex, because the cohort consisted of
only males. Demographic data from Wang et al.
2012 ((Wang et al.. 2012) p. 408) and Wang et al.
2014 ((Wang etal.. 2014) p. 121) suggest that the
distribution of confounders was similar among
miners and controls (although there was some
indication that miners may have a relatively low
SES compared to the general population; (Wang et
al.. 2012) p. 411). Smoking was slightly more
prevalent in miners than controls ((Wang et al..
2012) p. 410), but analyses were adjusted for
smoking.
Medium
0.500
1.000
38
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
This cohort evaluation represents all identified publications pertaining to the cohort of miners in Chongqing,
reference:
China. A full list of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
Covariates were assessed using reliable
C
methodology. Vital status information was obtained
_o
from personnel records. Information on smoking
cd
N
habits and verification of occupational history was
&
obtained from workers or their immediate relatives
g
(if deceased) throush personal contact ((Warm et al..
£
o
2013) d. 2 and (Wane et al.. 2012) d. 407). The
Wans et al. 2014 studv ((Wans et al.. 2014) p. 120)
High
i
0.250
0.250
indicated that vital status was obtained through
>
follow-up and links to death certificates and using
o
U
structured questionnaires. Although individual
©
smoking status information was available,
information on the duration and/or intensity of
smoking was not available.
The members of the cohort were workers at an
asbestos mine in China. There was no evidence that
"S
there was an unbalanced provision of co-exposures
S3
amons exposure sroups ((Wans et al.. 2013) p. 7.
'a
(Wans et al.. 2014) p. 123). At least one of the
u
studies ((Wans et al.. 2013) p. 7) indicated that
a
s
workers generally stayed with the mine for a
Medium
2
0.250
0.500
C/3
O
lifetime, with little opportunity to change jobs
X
(making exposure to other occupational carcinogens
1
o
U
unlikely). Two of the studies ((Wans et al.. 2012) o.
406 and (Wans et al.. 2013) p. 2) indicated no
detection of tremolite (amphibole contamination <
0.1%, the limit of detection).
The study design was appropriate to address the
2
research question. Cox and/or Poisson regression
analvses were used ((Wans et al.. 2013) p. 3. (Wans
et al.. 2014) t>. 121. and (Wans et al.. 2012) d. 407).
"3
The Wans et al. 2012 studv ((Wans et al.. 2012) o.
.£P
407) and Wans et al. 2014 studv ((Wans et al..
2014) p. 121) used Cox proportional hazard models
Medium
2
0.400
0.800
&
Q
to obtain hazard ratios for lung cancer mortality in
&
relation to asbestos exposure (cumulative exposure
es
fi
CO
for the latter study). The Wang et al. 2013 study
<
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This cohort evaluation represents all identified publications pertaining to the cohort of miners in Chongqing,
China. A full list of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
14. Reproducibility of
analyses
The methods used to estimate exposures were not
described in a way that would facilitate
reproducibility (string references were cited in
(Warm et al.. 2013) o. 2 and (Wans et al.. 2014) o.
120 as supporting evidence for dust to fiber
concentration conversions). In general, the statistical
analyses used were described, including variables
used in the analyses.
Medium
2
0.200
0.400
15.
Statistic
al
Models
The methods used to calculate risk estimates (SMRs
and HRs) were adeauatelv described ((Warm et al..
2013s) p. 3 and (Wane et al.. 2014^ a 1211
Medium
2
0.200
0.400
High: >=1 and <1.7
Medium: >=1.7 and <2.3
Low: >-2.3 and <-3
Sum of scores:
5
8.0166
Overall Score = Sum of Weighted Scores/Sum of Metric
Weighting Factors:
1.6033
Overall
Score:
Nearest
tenth:
1.6
Overall Quality Level:
High
Study
Quality
Comment:
The reviewer agreed with this study's overall quality level.
40
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
2.2.3. Epidemiology evaluation results of the Balangero, Italy
cohort of studies on asbestos exposure and lung cancer
incidence
Study
reference:
This evaluation represents all identified publications pertaining to the Balangero, Italy cohort of miners. A full list
of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
Study Participation
1. Participant selection
Subjects included men from the Balangero mine
worker cohort that were employed in an Italian
asbestos mine. The initial cohort (YRubino et al..
19791 us 188) consisted of 952 men employed
between 1/1/1930 and 12/31/1965, with at least 30
calendar days' employment during that period.
Mortality data were collected from 1/1/1946 to
12/31/1975. Workers for which vital status could
not be acertained and a small number of contract
workers employed intermittently were excluded. In
the first follow-up, 1058 workers were included that
had worked at least one year between 1946 and
1987 and mortality follow-up was extended through
12/31/1987 (YPiolatto etal.. 19901 m 810). In
subseauent follow-iiDS (YPira et al.. 2009) us 805.
and (Pira et al.. 2017)). subiects included 1056 men
from the Balangero mine worker cohort employed
between 1930 and 1990, and mortality records were
evaluated though 2003 and 2014, respectively.
Records were not available between 1987 and 1990,
when the mine closed, so workers employed in 1987
were assumed to be employed through 1990 unless
they died during that period. Additional details in
the most recent following indicated that the initial
cohort included 1182 men; the 126 excluded
subjects were contract workers, those employed <1
yr, those with inconsistencies in data, and those
known to have died orior to 1946 fPira et al.. 2017)
Pg 558.
High
i
0.400
0.400
2. Attrition
In the most recent follow-up, study authors report
that one of the strengths of the study is low
proportion of subiects lost to follow-up (Pira et al..
2017) us 562. Loss to follow -uD was 2% in the
initial cohort (Rubino et al.. 1979). 3% in the first
follow-uD (YPiolatto et al.. 1990). us 810). and 4%
in the most recent follow -iiDS fPira et al.. 2009) us
805; fPira et al.. 2017) m 559.
High
i
0.400
0.400
41
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This evaluation represents all identified publications pertaining to the Balangero, Italy cohort of miners. A full list
of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
3. Comparison Group
The most complete data on comparison groups is
available from the most recent follow-uo (Pira et al..
2017). General do nidation mortality rates usins the
whole country from 1955 until 1980 and specifically
the Piedmont Region (where the mine is located)
from 1981 onwards (no regional rates available
prior to 1981). The 1955-1959 rates were applied to
1946-1954 period (no available data); this may have
led to an underestimate of expected deaths which
may have showed and increased rate during this
period. Expected numbers of deaths (overall and
selected cancers) were computed using age-specific
and calendar-year-specific (5-year categories) male
death rates (Pira et al.. 2017) us 559. The onlv
deviation from this was in the first follow-up, which
used national mortality rates were for the entire
follow-up oeriod (throueh 1987) (Piolatto et al..
1990) us. 811). In the initial studv on this cohort
((Rubino et al.. 1979). m 189). an additional case-
control study was performed in which 5 age-
matched controls were selected at random; they
were confirmed alive at the time of death for the
matched case. No details on what population
provided controls. The evaluation is based on the
cohort mortality study only, as this was the analysis
carried throueh the 3 follow-uo studies (Pira et al..
2017; Pira et al.. 2009; Piolatto et al.. 1990).
High
i
0.200
0.200
Exposure Characterization
4. Measurement of Exposure
Most complete report of exposure assessment is in
initial cohort studv ((Rubino et al.. 1979) oe. 189).
Chrysotile fiber counts were first measured in 1969
using membrane filter collection and phase contrast
microscopy (frequency not reported). To estimate
exposure from 1946-1969, factory records on daily
production, equipment used, characteristics of the
job and number of hours/day were used (this
method has considerable limitations due to basis on
mean values for large job categories and no
allowance for changes in weather). Simulated and
measured data were made comparable by using
weighting factors (e.g., more dusty operation for 1-2
hr/d compared with longer working hours in the
past). Less detailed information was included in
follow-up reports ((Piolatto et al.. 1990) m 810;
(Pira et al.. 2017). pe 558-559)
Medium
2
0.400
0.800
42
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This evaluation represents all identified publications pertaining to the Balangero, Italy cohort of miners. A full list
of related papers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
5. Exposure levels
In the initial cohort ((Rubino et al.. 1979s). Table 8).
exposure was reported as up to 100 fiber/yr or >100
fiber/year. In the follow-ups, exposure was reported
as <100 fiber/mL-yr, 100-<400 fiber/mL-yr, and
>=400 fiber/ml-vr ((Piolatto et al.. 19901 Table 3;
(Pira et al.. 20091 Table 2; (Pira et al.. 20171
Tables 3-4)
Medium
2
0.200
0.400
6. Temporality
Evaluation is based on the most recent publication,
which is has the longest follow-up period for this
cohort (follow up of the 1946-1990 cohort through
20141 (YPira et al.. 20171 oe 5591 The first two
studies on this cohort (YRubino et al.. 1979) vs.. 188.
(Piolatto et al.. 1990) us 811) have inadeauate
follow-up duration for lung cancer (<15 years).
High
1
0.400
0.400
7.
Outcome
measureme
nt or
Lung cancer mortality was assessed based on death
certificate cause of death according to ICD rubrics
162/163 (YPira et al.. 20171 Table 1: (Rubino et al..
1979) ps 189; (Piolatto et al.. 1990) ps 189).
High
1
0.667
0.667
Outcome Assessment
8. Reporting Bias
Overall SMRs plus 95% CIs for lung cancer are
reported for the initial study and all 3 follow-ups
with 95% CI values in Table 2 of the most recent
follow-uD ((Pira et al.. 2017)). Luns and oleural
cancers are grouped together for the SMR from the
original study. The most recent follow-up also
reports RRs with confidence intervals for lung
cancer mortality (Table 4; (Pira et al.. 2017)). The
case-control rcDort in the initial studv lYRubino et
al.. 1979). Table 51 did not include confidence
intervals. The evaluation is based on the SMR
analyses carried forward in the follow-ups and the
RR analyses conducted in the most recent follow-
up.
High
1
0.333
0.333
il Confounding/Variable
Control
9. Covariate Adjustment
SMR was stratified by age and calendar year (5-yr
categories). Onlv males were included. ((Rubino et
al.. 1979) vs. 189; (Piolatto et al.. 1990) us 811;
(Pira et al.. 2009). us 806; (Pira et al.. 2017). os.
559. In the most recent follow-up, data on smoking
was limited to 14.5% of the cohort, but the
prevelance of smoking in this subset of the cohort
was comparable to that of the general male
population ((Pira et al.. 2017) ps 562).
High
1
0.500
0.500
c
o
a.
10.
Covariat
e
Charact
Empirical data obtained from employment records.
Smoking information was obtained from medical
records (when available).
High
1
0.250
0.250
43
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This evaluation represents all identified publications pertaining to the Balangero, Italy cohort of miners. A full list
of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
11. Co-
exposure
Confounding
No adjustments for potential coexposures were
described. Dust identified as primarily chrysotile, no
amphibole fibers dectected, but a fibrous silicate
(balangeroite) was detected (0.2-0.5% of total
sample) ((Pira et al.. 2017) pe 560).
Low
3
0.250
0.750
12. Study Design and Methods
For this retrospective cohort there is an initial study
(YRubino et al.. 1979)) and 3 foHow-lid studies
(YPiolatto et al.. 19901 (Pira et al.. 20091 (Pira et
al.. 2017)). Evaluation is based on the most recent
follow-uD ((Pira et al.. 20171 oe. 5591 in which
SMRs were calculated for entire cohort as well as
based on indicators of asbestos exposure (duration
of exposure, age at first exposure, years since first
exposure, years since last exposure, period at first
exposure, and cumulative dust exposure). RRs were
also calculated using Poisson regression. In this
recent follow-up, cohort members contributed to
person-time of observation starting 1 year after first
employment (or 1946), and ended at death, date of
last contact for those lost to follow-up, 85th
birthday, or December 31, 2014.
Medium
2
0.400
0.800
Analysis
13. Statistical power
The evaluation is based on the most recent follow-
ut> ((Pira et al.. 2017). us 516). which has a 90%
statistical power to detect a SMR of 2.0 for lung
cancer among workers with <100 fiber/mL-years
cumulative exposure (determined SMR [95% CI]
was 0.82 [0.44-1.40]; 13 deaths). The power of the
analysis for <25 fibers/mL-years cumulative
exposure was reported as "low" (determined SMR
[95% CI] was 2.40 [0.49-7.01]; 3 deaths). The
power for analyses at higher exposure levels (>=100
fiber/mL-years) was not reported.
Medium
2
0.200
0.400
14.
Reproducibility
of analyses
The evaluation is based on the most recent follow-
ut> ((Pira et al.. 2017)). For SMR calculations,
expected deaths were not reported for SMRs from
selected causes according to indicators of asbestos
exposure. All other relevant data are reported in
Tables 1-4.
Medium
2
0.200
0.400
15. Statistical Models
The evaluation is based on the most recent follow-
ut> ((Pira et al.. 2017). oe. 559). SMRs were
calculated for entire cohort as well as based on
indicators of asbestos exposure (duration of
exposure, age at first exposure, years since first
exposure, years since last exposure, period at first
exposure, and cumulative dust exposure). RRs were
calculated using Poisson regression.
Medium
2
0.200
0.400
44
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This evaluation represents all identified publications pertaining to the Balangero, Italy cohort of miners. A full list
of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
High: >=1 and <1.7
Medium: >=1.7 and <2.3
Low: >-2.3 and <-3
Sum of scores:
5
7.1
Overall Score = Sum of Weighted Scores/Sum of Metric
Weighting Factors:
1.42
Overall
Score:
Nearest
tenth:
1.4
Overall Quality Level:
High
Study
Quality
Comment:
The reviewer agreed with this study's overall quality level.
45
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
2.2.4. Epidemiology evaluation results of the North Carolina, US
cohort studies on asbestos exposure and lung cancer incidence
Study
reference:
This evaluation represents all identified publications pertaining to the North Carolina cohort of textile workers. A
full list of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
Study Participation
1. Participant selection
a. Setting and methods of case ascertainment were
reported in (Loomis et al.. 2009s). a 535-536.
Participant selection and inclusion/exclusion criteria
varied by study and analysis. Although there were 4
plants in the cohort, exposure data were available
only for three of the four, so exposure-response
analyses were limited to these three plants. Original
selection criteria rcDortcd in (Loomis et al.. 2009) v
536 (participants had to work at least 1 day between
1950 and 1973) and p 539 (participants excluded
due to missine data). (Elliott et al.. 2012s) evaluated
a subset of the cohort that worked >30 days during
the same time frame, b. Selection in or out of the
study was based on 1) employment in production
job during designated time frame and b) availability
of necessary data (birth and hire dates; work history;
vital status). These criteria are unlikely to result in
biased subject participation.
High
i
0.400
0.400
2. Attrition
Attrition/missing data exclusions were reported in
both (Loomis et al.. 2009) (a 539) and (Elliott et al..
2012) see oe 386. The orieinal cohort was 5770
persons; 373 workers at plant 2 were excluded due
to lack of exposure data at this plant, 1596 were
excluded due to incomplete work histories (at
department level) or non-oroduction iobs ((Loomis
et al.. 2009) v. 539). Final cohort for cxDOSiirc-
response analyses was 3803. Vital status was
unknown for 241 of the 3803 (6%) cohort members
(suggesting moderate loss to follow up). The
subaroiiD evaluated in (Elliott et al.. 2012)
consisted of 3082 subjects (excluded persons who
worked <30 days); the proportion for whom vital
status was missing was not reported for the
subgroup.
Medium
2
0.400
0.800
46
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This evaluation represents all identified publications pertaining to the North Carolina cohort of textile workers. A
full list of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
3. Comparison Group
There is potential for selection bias. All plants were
subject to surveillance program that removed
workers from exposure if they developed x-ray
changes attributable to dust exposure (typical
change was pneumoconiosis). Study authors
reported that some x-ray changes are associated
with higher lung cancer risk. Thus, the surveillance
program could have selected workers at greater risk
of lung cancer for lower cumulative exposure
((Loomis et al.. 2009) o. 542. and (Elliott et al..
2012) pg. 388.
Medium
2
0.200
0.400
a
#o
¦B
t5
cS
~-
es
£
U
4. Measurement of Exposure
Air samples were available for 3 plants covering
period from 1935 to 1986 (459 <1950; 1674 from
1950-1969. and 1287 from 1970 forward: (Loomis
et al.. 2009). v. 536). Measurements used immneer
before 1964 and PCM thereafter; paired and
concurrent samples between 1964 and 1971 were
used to relate impinger to PCM-equivalent
concentrations. Air samples were not collected
yearly, so mean PCM-equivalent concentrations
were estimated by plant, department, job, and time
period usine multivariate mixed models ((Loomis et
al.. 2009). d. 536). Individual cumulative exposure
assessed using the modeled concentrations and JEM
((Loomis et al.. 2009). t> 536); details of JEM
reported in (Dement et al.. 2008).
High
1
0.400
0.400
£
o
o.
UJ
5. Exposure
levels
Exposure-response relationships were developed
(see Table 5 of (Loomis et al.. 2009) and (Elliott et
al.. 2012) see Table 2). A total of 5 cumulative
exposure levels are analyzed in Table 5 of (Loomis
etal.. 20091
Medium
2
0.200
0.400
6.
Temporality
Temporality was established (exposure preceded
death). The follow-up time was at least 30 years
(1973-2003); lag times of 0, 10, 20, and 30 years
were analyzed ((Loomis et al.. 2009) Table 5 and
(Elliott et al.. 2012) see Table 2).
High
1
0.400
0.400
Outcome Assessment
7. Outcome
measurement or
characterization
Lung cancer deaths (underlying or immediate cause
or other significant condition at time of death) were
determined from the National Death Index Plus
(1979 and later) or state records (before 1979).
Specific ICD codes were not reported, but ICD in
effect at time of death was used ((Loomis et al..
2009). o 536). Deaths before 1979 were coded
manually by a nosologist.
Medium
2
0.667
1.333
47
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This evaluation represents all identified publications pertaining to the North Carolina cohort of textile workers. A
full list of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
8. Reporting Bias
Lung cancer findings are reported in abstract,
results, and discussion of kev publications ((Loomis
et al.. 2009) and ffilliott et al.. 2012)). Luns cancer
deaths and person-years at risk by exposure
cateeorv are rcDortcd in Table 5 of (Loomis et al..
2009): luns cancer cases and Dcrson-vears at risk
are rcDortcd bv olant in Table 1 of ffilliott et al..
2012). Rate ratio estimates are reported with CIs in
(Loomis et al.. 2009) and ffilliott et al.. 2012):
coefficients for Poisson regression are reported
without SE in Table 2 of ffilliott et al.. 2012). but
rate ratios with CIs are also reported for the same
models.
High
i
0.333
0.333
Potential Confounding/Variable Control
9. Covariate
Adjustment
Analyses in (Loomis et al.. 2009) and ffilliott et al..
2012) were adiusted for ase. sex. race, decade of
follow-up and birth cohort. No adjustment was
made for smoking, which could bias SMR analyses
(reported in (Loomis et al.. 2009)) but is not likelv
to bias internal analysis in an occupational cohort
(e.e.. exposure-response analyses in (Loomis et al..
2009) and ffilliott et al.. 2012).
Medium
2
0.500
1.000
10. Covariate
Characterization
While not specified, information on covariates
included in the analyses were likely obtained from
same sources as vital status/cause of death. Subjects
with missing hire or birth date were excluded
((Loomis et al.. 2009). v 539). Smokins information
was available for <15% of the cohort.
Medium
2
0.250
0.500
11. Co-exposure
Confounding
One plant used a limited amount of amosite between
1963 and 1976 ((Loomis et al.. 2009). v. 536).
Adjustment for this coexposure was not possible
because none of the lung cancer deaths were among
workers involved in activities using amosite
((Loomis et al.. 2009). p. 539)
Medium
2
0.250
0.500
Analysi
s
12.
Study
Design
and
Study design was retrospective cohort; Poisson
regression used for internal analyses in (Loomis et
al.. 2009) (t>. 537) and ffilliott et al.. 2012) (t>. 386).
Medium
2
0.400
0.800
48
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This evaluation represents all identified publications pertaining to the North Carolina cohort of textile workers. A
full list of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
13. Statistical power
None of the related studies report power
calculations. Authors ((Loomis et al.. 20091 v. 541)
cite large size of cohort (3803 who worked for at
least 1 dav in olants with exposure data lYLoomis et
al.. 2009)1; 3082 who worked at least 30 davs in
olants with exposure data (Elliott et al.. 20121 hieh
proportion of subjects with vital status ascertained,
and long follow-up (30+ yrs; total 124,029 person-
vears workins at least 1 dav lYLoomis et al.. 2009)1;
100742 person-years working at least 30 days
(Elliott et al.. 2012) as strengths.
Medium
2
0.200
0.400
14. Reproducibility of analyses
Analvsis description in (Loomis et al.. 2009)
appears to be complete. Statistical tests and
estimation procedures and variables considered are
reported ((Loomis et al.. 2009) o 537-538 and
supplemental file); cumulative exposure categories
for internal analysis define quantiles of exposure
amons cases (footnote to Table 5 in (Loomis et al..
2009)). There were no variable transformations,
outlier exclusions, or imputation of missing values.
Analvsis description in (Elliott et al.. 2012) aDDcars
to be complete. Statistical tests and estimation
procedures and variables considered are reported
(Elliott et al.. 2012) us 386. Cumulative exposure
was analvzed as a continuous variable (Elliott et al..
2012) us 386. No variable transformations were
reported.
Medium
2
0.200
0.400
15. Statistical Models
Statistical models and methods are described in each
oublication (see (Loomis et al.. 2009) t> 537-538 and
(Elliott et al.. 2012) v> 386). including how variables
were included or excluded; supplemental file to
Loomis et al. (2009) orovides details of SMR
computation when race was unknown and further
explanation of the Poisson exposure-response
models. For (Elliott et al.. 2012). covariates were
assessed as confounders using a 10% change in
estimate method and as effect measure modifiers
using likelihood ratio test.
Medium
2
0.200
0.400
Sum of scores:
5
8.4666
High: >=1 and <1.7
Medium: >=1.7 and <2.3
Low: >=2.3 and <=3
Overall Score = Sum of Weighted Scores/Sum of Metric
Weighting Factors:
1.6933
Overall
Score:
Nearest
tenth:
1.7
49
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This evaluation represents all identified publications pertaining to the North Carolina cohort of textile workers. A
full list of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
Overall Quality Level:
Medium
Study
Quality
Comment:
The reviewer agreed with this study's overall quality level.
50
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
2.2.5. Epidemiology evaluation results of the Salonit Anhovo,
Slovenia cohort studies on asbestos exposure and lung cancer
incidence
Study
reference:
This cohort evaluation represents all publications pertaining to the Slovenian cohort of asbestos-cement workers.
A full list of related naners reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
a
#o
"s
.2-
1. Participant selection
This study included 58 histologically confirmed
incident lung cancer cases from the national cancer
registrar and 290 matched controls from a cohort of
6714 workers employed at Salonit Anhovo factory
after Dec 31 1946 who worked there for at least one
day between 1964 and 1994 (pg. 263-264). DOB,
gender, and year of hire (pre-1959 or post-1959,
based on better quality of exposure data after 1959)
were used as matching factors. Five controls closest
to the birth date were selected and had to be alive at
time of diagnosis (using national mortality registrar)
(pg. 263). The follow-up begins at 1964 because the
cancer registrar data were only available in a
computerized form from 1964 onward (registry
established in 1957) (pg. 263).
High
i
0.400
0.400
t:
es
a.
¦a
5
Tfl
2. Attrition
Initial study pop was 67 cases of lung cancer, 335
controls (5 controls/case) were selected based on
original. Nine cases were excluded because they
were hired prior to 1947 (a priori date cut-off;
factory was owned by Italians from 1921-1947 and
the Slovenian republic from 1947 onward), so those
45 matched controls were also excluded. This left
58 cases and 290 matched controls, (pg. 264)
Retention of 87% (58/67 cases) was characterized as
moderate subject exclusion.
Medium
2
0.400
0.800
3. Comparison
Group
Cases and controls were selected from the same
occupational cohort. Matched based on DOB,
gender, and year of hire (pre- or post-1959), (pg.
263) Controls were confirmed alive at age of
diagnosis for case. (pg. 263) Other demographics
were also similar (see Table 1, pg 264).
High
1
0.200
0.200
51
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This cohort evaluation represents all publications pertaining to the Slovenian cohort of asbestos-cement workers.
A full list of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
4. Measurement of Exposure
A total of 1030 air measurements from factory were
available from 1961-1995 using several different
monitoring methods, including a set of 78 paired
measurements using side-by-side gravimetric and
membrane filter methods. An early publication on
this cohort bv Dodic Fikfak (2003) (vs. 171)
indicates that gravimetric is a mass based method
(units of mg/m3) and the membrane filter method is
a fiber counting method (units of f/cm3).
Microscopy analysis was not described; therefore
TEM and PCM methods were likely not used.
Exposure Characterization
The non-parametric classification and regression
tree (CART) method was used to calculate
conversion factors for different combinations of
fiber type, product, and production method.
Exposure levels were measured at fixed locations
close to worker's breathing zones. Exposures were
estimated for missing years using previous or next
values (or average of both). Exposures for most
workers were based on measured exposure values in
work area. A few jobs did not have applicable air
sample measurements, and exposures were
estimated with JEM. The percentage of individuals
in the JEM group were not reported; however, study
authors state that subjects from this group were not
selected for the study, (pg. 263)
Low
3
0.400
1.200
5.
Exposur
e levels
Evaluated as dichotomous exposure definitions:
exposed/unexposed, above/below median, and
above/below 90th percentile, (pg. 263-264)
Low
3
0.200
0.600
6. Temporality
Average latency between start of employment and
diagnosis in cases was 24.9 years (pg. 264). Study
authors conducted evaluations with different latency
periods (0-15 yrs, 16-35 yrs, and >35 yrs) (pg. 263).
The primary analysis is exposure >15 years prior to
diagnosis (Table 4, pg. 266).
High
1
0.400
0.400
Outc
ome
Asse
ssme
7.
Outc
ome
meas
All cases were histologically confirmed incident
cases of primary lung cancer (pg. 263).
High
1
0.667
0.667
52
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This cohort evaluation represents all publications pertaining to the Slovenian cohort of asbestos-cement workers.
A full list of related naners reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
8. Reporting Bias
The primary analysis is limited to cases (and
matched controls) with exposure >15 years prior to
diagnosis of case (Table 4, pg. 266). The study does
not report the number of cases and controls exposed
for >15 years only. Additionally, exposure estimates
are for all years, 0-15 yr, 16-35 yr, and >35 yrs (data
are very limited for >35), separated by case and
control (Table 3, pg. 266). Based on data reporting,
it is not clear exactly which exposure-cutoff values
were used for median and 90th percentile analyses.
Medium
2
0.333
0.667
"o
~-
c
o
9. Covariate Adjustment
Matching variables of birth and gender were
included in models as covariates. The matching
factor of pre vs. post 1959 hire was also evaluated
as a potential covariate. However, a comparison of
analyses did not show evidence that pre/post 1959
hire introduced confounding, so this covariate was
not maintained in the primary analysis, (pg. 263).
Separate analyses were calculated for smokers and
non-smokers, but non-smoking population was very
small (pg. 263, 265). Logistic regression models
were adjusted for smoking (yes/no) (pg. 264). No
further covariate assessment/adjustments made
(except co-exposure, addressed in Metric 11).
Medium
2
0.500
1.000
U
3
es
10.
Cova
riate
Char
Covariates were empirical data obtained from
employment records (age, sex) (pg. 263).
High
1
0.250
0.250
c
es
%
c
¦3
c
5
£
C
O
U
."s
c
.If
%
S3
Amphibole asbestos: Amphibole exposure made up
for 10% of the total asbestos exposure (pg. 261).
Detailed records allowed the estimation of separate
individual exposure histories for different forms (pg.
261). Methods indicate that models were adjusted
for confounders (pg. 263), but did not specifically
indicate whether or not final model was adjusted for
amphibole asbestos exposure.
o
a.
o
O
%
VI
0
Oh
X
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This cohort evaluation represents all publications pertaining to the Slovenian cohort of asbestos-cement workers.
A full list of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
12. Study Design and
Methods
Study was a case-control design. Models of
exposure and risk were adjusted for confounding
using unconditional or conditional multivariate
logistic regression (pg. 263). Matching variables
were included as covariates in unconditional models
(pg. 263). Primary evaluation used OR calculations
to determine risk from cumulative exposure
estimates using 15-year latency for main analysis
(Table 4, pg 266).
Medium
2
0.400
0.800
Analysis
13. Statistical
power
Low would be selected if it was an option. No
statistically significant findings were observed
(Table 4, pg 266); study authors attribute this to low
statistical power-confidence intervals. Low
statistical power was reported as a limitation of the
study by study authors (pg 266-267).
Medium
2
0.200
0.400
14.
Reproduci
bility of
analyses
Median and 90th percentile cutoffs were not
explicitly reported for chrysotile asbestos for >15 yr
latency analysis. Number of cases and controls
included in >15 yr latency analysis not reported.
Low
3
0.200
0.600
15.
Statistical
Models
Logistic regression models were constructed for
each of the following dichotomous exposure
definitions: ever/never, above/below median, and
above/below 90th percentile (pg. 263-264).
Medium
2
0.200
0.400
Sum of scores:
5
9.1334
High: >=1 and <1.7
Medium: >=1.7 and <2.3
Low: >-2.3 and <-3
Overall Score = Sum of Weighted Scores/Sum of Metric
Weighting Factors:
1.8267
Overall
Score:
Nearest
tenth:
1.8
Overall Quality Level:
Medium
Study
Quality
Comment:
The reviewer agreed with this study's overall quality level.
54
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
2.2.6. Epidemiology evaluation results of the Quebec, Canada
cohort studies on asbestos exposure and lung cancer incidence
Study
reference:
This evaluation represents all identified publications pertaining to the Quebec, Canada cohort of miners. A full list
of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
a
#o
"s
1. Participant selection
Some details of the setting (e.g., the mine locations
and production quantities, but no description of the
facilities and exposure conditions) are provided in
(Liddell et al.. 1997) (d 14s); additional information
mav be available in Gibbs and Lachance (1972)
(cited as the source of the exposure assessment), but
no pdf was available in HERO at the time of
evaluation. Participant selection, inclusion/exclusion
criteria, case ascertainment, and participation at
each level are described in detail ((Liddell et al..
1997). v 14-16). Selection into the cohort was based
on a) male sex; b) birth year between 1890 and
1920; b) employment at the Thetford Mines or
Asbestos mine/mill or factory for at least one
month. Participant selection and participation were
not likely to be biased based on these criteria.
High
i
0.400
0.400
o.
'3
"H
es
a.
¦a
5
Tfl
2. Attrition
Numbers of subjects lost to follow up are reported
in (Liddell et al.. 1997). Table 4; there it is rcDortcd
that a total of 1138/10918 (-10%) were lost to
follow-up. No information comparing the
demographics or exposure of those lost to follow up
to the study subjects was located in the six selected
HERO IDs. Missing data were not imputed;
however, censoring of survival data is unlikely to
introduce bias. This level of attrition is considered
to be moderate.
Medium
2
0.400
0.800
3. Comparison Group
Internal analysis was reported only in the nested
case-control studv in Liddell et al. (1998) (ORs in
Table 4). Most of the Dublications (see (Liddell et
al.. 1997) v> 18; Liddell and Armstrong (2002) d 9)
reported SMR analyses using age- sex- and calendar
year (5 year intervals) -adjusted general population
(Quebec when available, or Canada for earlier time
periods) mortality rates as the comparison group.
No adjustment for race was made.
Medium
2
0.200
0.400
Exposure
Character
ization
4.
Measurem
entof
Exposure
Exposure levels measured exclusively using midget
immnser ((Liddell et al.. 1997). o 17) and
conversion factors were not determined (based on
review of the six selected HERO IDs).
Low
3
0.400
1.200
55
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This evaluation represents all identified publications pertaining to the Quebec, Canada cohort of miners. A full list
of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
5. Exposure
levels
Cumulative exposure-response relationships were
developed (see for example Table 8 of (Liddell et
al.. 1997)). A total of 7 cumulative exoosure levels
(exposure through age 55) are reported in Table 8 of
(Liddell et al.. 1997): in Liddell and Armstrong
(2002). 10 exposure levels are reported in Table 5.
Medium
2
0.200
0.400
6. Temporality
Temporality was established (exposure preceded
death), and there was adequate follow-up for
consideration of latency. Of the total cohort of
10918 men, 6415 were still employed when the
cohort was first established in 1966, and follow up
extended to 1992 ((Liddell et al.. 1997) d 15). Entrv
into the cohort was restricted to men born between
1891 and 1920 ((Liddell et al.. 1997) d 15). so the
youngest subjects at the end of follow up were 72
years old. There was no indication in the studies
reviewed that the operations at Thetford or Asbestos
had ceased at any point during the follow up time.
High
1
0.400
0.400
C
S
X
5«
5«
5«
<
a>
B
7. Outcome
measurement or
characterization
ICD code 162 (ICD-9) used. Cause of death was
obtained from death certificate (or other "reliable"
information, primarily from hospitals); these were
available for over 98% of the cohort followed
throueh 1992 ((Liddell et al.. 1997). o 16). Cause-
specific death rates (referent group for SMRs) were
available only for deaths from 1950 forward.
High
1
0.667
0.667
(J
5
o
8.
Reporti
ng Bias
In most analyses SMRs are reported without CI
estimates (e.e.. Table 8 in (Liddell et al.. 1997);
Tables 3-5 of Liddell and Armstrong (2002)).
Medium
2
0.333
0.667
Potential
Confounding/V ari
able Control
9. Covariate
Adjustment
SMRs were calculated using gender-specific rates
across 16 ase categories ((Liddell et al.. 1997) v
18). In Liddell et al. (1998). a detailed analysis of
the effect of smoking on risk estimates was
dresented. Vacek (1998) also included an analysis
of the impact of smoking. No adjustment for race
was made.
Medium
2
0.500
1.000
56
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This evaluation represents all identified publications pertaining to the Quebec, Canada cohort of miners. A full list
of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
10. Covariate Characterization
Information on sex and birth date or age at first
employment were recorded from personnel records
in 1966 and reviewed and/or corrected during
subseauent analyses ((Liddell et al.. 1997) v 14-15).
No information on race of cohort members was
located in any of the 8 selected HERO IDs. A
questionnaire was administered in 1970 to obtain
smoking histories; subjects still living completed the
questionnaires (99.6% of 6583 men alive completed
their own questionnaires), and proxies provided the
information for deceased subjects (for 90% of those
who died after 1950: (Liddell et al.. 1997) d 18).
Each subject was assigned to the smoking category
in which his response placed him at the time of the
Questionnaire ((Liddell et al.. 1997) o 18). A total of
891 questionnaires, mostly completed by proxies,
were judged unreliable and the subjects omitted
from analyses that considered smokins ((Liddell
and Armstrong. 2002) o 7). While the methods to
assess potential confounders were not validated,
there is little indication that the methods had poor
validity.
Medium
2
0.250
0.500
11. Co-exposure Confounding
Liddell et al. (1997) (u 33) reoorts that analysis of
fibers in the lungs of workers in the Quebec industry
showed higher levels of tremolite fibers than
chrysotile fibers, especially at the Thetford mines.
On d 34. Liddell et al. (1997) states that the
chrysotile produced in Quebec may be contaminated
not only with tremolite but with other amphibole
fibers. Liddell and Armstrone (2002) (d 8) reoorts
that the Thetford mines complex had "more
substantial" tremolite contamination than the mine
and mill at Asbestos, and suggests that it is therefore
desirable to analyze the sites separately. None of the
8 selected HERO IDs provided quantitative
estimation of the degree of contamination. Thus,
there is no information to indicate how the co-
exposure may have been distributed across cohort
members and/or its relationship to chrysotile
exposure. It is possible that additional information is
available in Gibbs and Lachance (1972). which is
cited as the source of the JEM and exposure
assessment, but no pdf was available in HERO at
the time of evaluation.
Low
3
0.250
0.750
57
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This evaluation represents all identified publications pertaining to the Quebec, Canada cohort of miners. A full list
of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
12. Study
Design and
Methods
Study design was retrospective cohort. Poisson
regression used in some analyses (Liddell and
Armstrong (2002) d. 8) but not others. Liddell et al.
(1998) reported a nested case-control studv within
the cohort and used conditional logistic regression
Medium
2
0.400
0.800
13. Statistical
power
Power calculations were not reported. The number
of participants was sufficient to detect an effect
(statistically significant association reported in
Vacek (1998Y). However, manv of the studies did
not report statistical significance or confidence
intervals for calculated SMRs (see for example
Tables 8 and 10 of Liddell et al. (1997)).
Medium
2
0.200
0.400
Analysis
14. Reproducibility of analyses
The analysis description varied by study. This
metric is rated based exclusively on the description
in Liddell et al. (19971 which is based on the
combined longest follow up and largest population.
The effect estimation methods in this paper were
reported (p. 18-19), but the rules for cumulative
exposure categorization (as shown in Tables 7 and
8) were not reported. There were no variable
transformations or outlier exclusions, and no true
statistical analyses (CIs were not reported for the
SMRs).
Low
3
0.200
0.600
15. Statistical Models
Liddell et al. (1997) did not include anv statistical
models per se (effect estimates calculated as SMRs
= O/E without CIs). Liddell et al. (1998) rcDortcd a
nested case-control study within the cohort and used
conditional logistic regression. 90% Confidence
intervals were estimated from the regression on the
assumption that the regression coefficients are
normally distributed; no information on whether
assumptions were met was nrovided. Liddell and
Armstrone (2002) orovidcd detailed description of
statistical models (p 8-9) but did not describe model
assumptions or whether they were met.
Low
3
0.200
0.600
Sum of scores:
5
9.5834
High: >=1 and <1.7
Medium: >=1.7 and <2.3
Low: >-2.3 and <-3
Overall Score = Sum of Weighted Scores/Sum of Metric
Weighting Factors:
NA
Overall
Score:
Nearest
tenth:
NA
Overall Quality Level:
Low
58
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This evaluation represents all identified publications pertaining to the Quebec, Canada cohort of miners. A full list
of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
Study
Quality
Comment:
The reviewer downgraded this study's overall quality rating. They noted: Lack of PCM or TEM-equivalent
exposure estimates and potentially significant co-exposure to tremolite or other amphiboles. Note: The original
calculated score for this study was 1.9. This value is not presented above because the final rating was changed based
on professional judgement.
59
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
2.2.7. Epidemiology evaluation results of the Chongqing, China
textile worker cohort studies on asbestos exposure and lung
cancer incidence
Study
reference:
This evaluation represents all identified publications pertaining to the Chongqing, China cohort of textile workers.
A full list of related papers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
1. Participant selection
The key elements of the study design are reported.
The fixed cohort is well-established. The inclusion
criteria were clearly specified (male workers
registered with the plant by January 1 1972,
employed for at least one \ car: Dens et al. (2012) v.
82 and Courtice et al. (2016s) o. 370). Workers with
cardiopulmonary disease, or those employed after
January 1 1972 were excluded. The reported
information indicates that selection into/out of the
study was not likely biased.
High
i
0.400
0.400
a
#o
"s
o.
'3
"H
es
a.
2. Attrition
There was minimal loss of subjects at follow-up.
The 2016 study (conducted 37 years after
establishment of the cohort) reported that 577 of
586 workers (99%) were successfully followed
throueh 2008 (Courtice et al. (2016) a 370).
Therefore, exposure and outcome data were largely
complete.
High
i
0.400
0.400
¦a
5
Tfl
3. Comparison Group
Inclusion criteria and the methods of participant
selection were reported. All subjects were recruited
from the same eligible population within the same
time frame. In the Wang et al. 2012 study reporting
SMRs (and RRs: (Wans et al.. 2014) o. 121). the
choice of a reference population is reported (based
on age-specific national mortality data for males).
However, data on cause-specific mortality data were
limited (rates of 1990 and 2004 were used to
correspond to periods of 1981-1995 and 1996-2006,
respectively. The Courtice et al. (2016) (d. 375)
used the lowest continuous cumulative exposure
category (i.e., members of the same cohort in the
lowest exposure quartile) as the reference group.
High
i
0.200
0.200
60
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This evaluation represents all identified publications pertaining to the Chongqing, China cohort of textile workers.
A full list of related papers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
Exposure Characterization
4. Measurement of Exposure
Asbestos dust measurements were available starting
in 1955 (Courtice et al. ("2016) 0.371. Dens et al.
(2012) v. 82). Startins in 1999. samples were also
analyzed by phase contrast and/or electron scanning
microscoDY (Courtice et al. (2016) d. 371. Dens et
al. (2012) d. 82). Paired sravimetric and membrane
filter samples from 1999 and 2002 ( Courtice et al.
(2016) v. 371. (Wans et al.. 2014) v. 120) or from
1999. 2002. and 2006 (Dens et al. (2012) v. 82)
were used to define the relationship between dust
and fiber concentrations. From these data, (and
using periodically measured data), average fiber
concentrations by job type/exposure area were
calculated. Studies for this cohort ((Wans et al..
2014) v. 120 and Dens et al. (2012) a 82) indicate
that conversion from dust to fiber concentrations
required log transformation of the paired samples
(the distribution of paired samples was positively
skewed). Individual cumulative exposures were
estimated as the product of (fiber concentration at a
specific workshoo/iob) x (duration of iob) (Courtice
et al. (2016) d.371. Dens et al. (2012) u. 82). There
were no exposure data prior to 1955; exposure was
assumed to be the same as the earliest time
measurement in 1955 (Courtice et al. (2016) v. 371).
There was also no samples for administration or rear
service workers; these groups were assumed to
belong in the lowest cumulative exposure group
(Courtice et al. (2016) o. 371). The lack of detailed
exposure information and the use of recent (since
1999) samples to convert to historical measurements
since 1955 are limitations of the studv (Courtice et
al. (2016) v. 375-376). In addition, workers mav
have been additionally exposed to chrysotile at
home (from spinnins: Courtice et al. (2016) v.
376). The study authors acknowledge that there may
have been exposure misclassification based on these
estimations, but the misclassification was likely to
be non-differential.
Medium
2
0.400
0.800
5. Exposure
levels
The range and distribution of exposure is sufficient
to develop and exposure-response estimate. The
Courtice et al. 2016 studv (Courtice et al. (2016) o.
375) reports 4 levels of exposure (referent + 3);
cumulative exposures were categorized into
quartiles for analyses of lung cancer deaths.
Medium
2
0.200
0.400
61
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
This evaluation represents all identified publications pertaining to the Chongqing, China cohort of textile workers.
reference:
A full list of related papers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Weighting
Factor
Weighted
Score
The study establishes appropriate temporality; the
interval between exposure and outcome is long
enough considering latency of the disease. The
.1?
cohort was followed for 35 vears (Dene et al. (2012)
p. 81). 26 vears (YWane et al.. 2014) p. 119) and 37
2
vears ( Courtice et al. (2016) p. 370). The studv bv
|
Dens et al. 2012 (Dens et al. (2012) p. 83)
High
i
0.400
0.400
370).
o
(J
Lung cancer findings are reported in the results. In
5
O
a
(3
bX)
the Wans et al. 2012 studv ((Wans et al.. 2012) d.
122), SMRs and hazard ratio data (with 95%
C
e
o
confidence intervals) were complete. The Courtice
High
i
0.333
0.333
et al. 2016 studv (Courtice et al. (2016) a 375)
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This evaluation represents all identified publications pertaining to the Chongqing, China cohort of textile workers.
A full list of related papers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
10. Covariate Characterization
Covariates were assessed using reliable
methodology. Vital status information was obtained
from Dcrsonnel records and interviews (Courtice et
al. (2016) a 370). The Wane et al. 2014 and
Coutice e et al. 2016 studies ((Warm et al.. 2014) o.
120 and Courtice et al. (2016) o. 370) indicated that
vital status was obtained through follow-up and
links to records at hospitals, death registry, and
using structured questionnaires. Information on
smoking habits and verification of occupational
history was obtained from workers or their
immediate relatives (if deceased) through personal
contact (Dens et al. (2012)). Althoush individual
smoking status information was available,
information on the duration and/or intensity of
smoking was not available.
High
i
0.250
0.250
11. Co-exposure Confounding
The members of the cohort were workers at
chrysotile products plant in China. There was no
evidence that there was an unbalanced provision of
co-exposures (other than asbestos) among exposure
sroups. The Courtice et al. (2016) studv indicated
that workers rarely changed jobs or between job
types (making exposure to other occupational
carcinogens unlikely). However, there is uncertainty
with respect to the purity of the chrysolite to which
the workers were exposed. Samples collected in
2006 showed evidence of tremolite contamination
(Courtice et al. (2016) o. 376). The studv authors
suggested that low incidences of mesothelioma in
this cohort suggest that amphibole contamination
was limited (Courtice et al. (2016) p. 376).
Medium
2
0.250
0.500
Analysis
12. Study Design and Methods
The study design (cohort) was appropriate to
address the research question. In the Deng et al.
2012 studv (Dens et al. (2012) a 81). Poisson
regression analyses was used to fit models (log-
linear, log-quadratic, power, additive relative risk
and categorical) to estimate relationships between
cumulative exposure and mortality from lung
cancer. The Wans et al. 2012 studv ((Wans et al..
2012) d. 407) used Cox Drooortional hazard models
to obtain hazard ratios for lung cancer mortality in
relation to cumulative asbestos exposure. The study
bv (Courtice et al. (2016) v>. 371) estimated
exposure-response relationships using Cox
proportional hazard models.
Medium
2
0.400
0.800
63
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This evaluation represents all identified publications pertaining to the Chongqing, China cohort of textile workers.
A full list of related papers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
13. Statistical
power
The number of participants (cohort size = 577
workers) was sufficiently large to detect an effect in
the exposed population. However, it was noted that
there was a relatively small number of nonsmokers
in the cohort (Dens et al. (2012) p. 86).
Medium
2
0.200
0.400
14. Reproducibility of
analyses
The methods used to estimate exposures were not
described in a way that would facilitate
reproducibility; a process "similar" to those in other
studies was cited (Courtice et al. C2016) v. 371) or a
strins of references was provided ((Wans et al..
2014) d. 120) for dust to fiber concentration
conversions. In general, the statistical analyses used
were described, including variables used in the
analyses.
Medium
2
0.200
0.400
15. Statistical Models
The data analysis section of the Deng et al. paper
(Dens et al. (2012) d. 83) describes the analvses
used in detail (including calculations, model
considerations, variables). The methods used to
calculate risk estimates (HRs) in Wang et al. 2014
were adequately described (YWans et al.. 2014) v.
121). In the Courtice et al. 2016 studv (Courtice et
al. (2016) v. 371). the methods used to senerate
HRs were briefly described (i.e. age included as a
time dimension for Cox proportional hazard
modeling).
Medium
2
0.200
0.400
Sum of scores:
5
6.85
High: >=1 and <1.7
Medium: >=1.7 and <2.3
Low: >-2.3 and <-3
Overall Score = Sum of Weighted Scores/Sum of Metric
Weighting Factors:
1.37
Overall
Score:
Nearest
tenth:
1.4
Overall Quality Level:
High
Study
Quality
Comment:
The reviewer agreed with this study's overall quality level.
64
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
3. Data Quality Evaluation of Mesothelioma Data Sources
3.1. Data Evaluation Scoring Sheets: Mesothelioma Outcome
3.1.1. Epidemiology evaluation results of the Quebec, Canada
cohort studies on asbestos exposure and mesothelioma
incidence
Study
reference:
This evaluation represents all identified publications pertaining to the Quebec, Canada cohort of miners. A full list
of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
Study Participation
1. Participant selection
Some details of the setting (e.g., the mine locations
and production quantities, but no description of the
facilities and exposure conditions) are provided in
Liddell et al. (1997) (v 14s); additional information
mav be available in Gibbs and Lachance (1972)
(cited as the source of the exposure assessment), but
no pdf was available in HERO at the time of
evaluation. Participant selection, inclusion/exclusion
criteria, case ascertainment, and participation at
each level are described in detail ((Liddell et al..
1997). v 14-16). Selection into the cohort was based
on a) male sex; b) birth year between 1890 and
1920; b) employment at the Thetford Mines or
Asbestos mine/mill or factory for at least one
month. Participant selection and participation were
not likely to be biased based on these criteria.
High
i
0.500
0.500
2. Attrition
Numbers of subjects lost to follow up are reported
in Liddell et al. (1997). Table 4; there it is rcDortcd
that a total of 1138/10918 (-10%) were lost to
follow-up. No information comparing the
demographics or exposure of those lost to follow up
to the study subjects was located in the eight
selected HERO IDs. Missing data were not imputed;
however, censoring of survival data is unlikely to
introduce bias. This level of attrition is considered
to be moderate.
Medium
2
0.500
1.000
3. Comparison
Group
Not applicable for mesothelioma studies
Not Rated
NA
NA
NA
65
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This evaluation represents all identified publications pertaining to the Quebec, Canada cohort of miners. A full list
of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
Exposure Characterization
4. Measurement of
Exposure
Exposure levels measured exclusively using midget
immneer (Liddell et al. (1997). d 17s) and
conversion factors were not determined (based on
review of the eisht selected HERO IDs). In (Berman
and CniniD. 20081 a sinsle conversion factor is
applied to all operation for estimation of equivalent
exposure concentrations.
Medium
2
0.400
0.800
5. Exposure levels
Cumulative exposure-response relationships were
developed (see for example Table 9 of Liddell et al.
(1997)). A total of 6 cumulative exposure levels
(exposure through age 55) are reported in Table 9 of
Liddell et al. (1997). (Berman and CruinD. 2008)
reports Km estimate for mesothelioma exposure-
response.
Medium
2
0.200
0.400
6. Temporality
Temporality was established (exposure preceded
death), and there was adequate follow-up for
consideration of latency. Of the total cohort of
10918 men, 6415 were still employed when the
cohort was first established in 1966, and follow up
extended to 1992 (Liddell et al. (1997) t> 15). Entrv
into the cohort was restricted to men born between
1891 and 1920 (Liddell et al. (1997) t> 15). so the
youngest subjects at the end of follow up were 72
years old. There was no indication in the studies
reviewed that the operations at Thetford or Asbestos
had ceased at any point during the follow up time.
High
1
0.400
0.400
Outcome Assessment
7. Outcome
measurement or
characterization
Mesothelioma cases post-1966 were identified via
examination of "all related clinical, biopsy, and
nccroDSY records" (Liddell et al. (1997) v 16).
High
1
0.667
0.667
8.
Reporting
Bias
Rate estimates by exposure level are reported
without CI estimates (e.e.. Table 9 in Liddell et al.
(1997)). (Berman and Crumo. 2008) reports Km
estimate for mesothelioma exposure-response data.
Medium
2
0.333
0.667
Potential
Confounding/V
ariable Control
9. Covariate
Adjustment
Not applicable for mesothelioma studies
Not Rated
NA
NA
NA
66
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This evaluation represents all identified publications pertaining to the Quebec, Canada cohort of miners. A full list
of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
10. Covariate
Characterization
Not applicable for mesothelioma studies
Not Rated
NA
NA
NA
11. Co-exposure
Confounding
Not applicable for mesothelioma studies
Not Rated
NA
NA
NA
Analysis
12. Study
Design and
Methods
Study design was retrospective cohort. Poisson
regression used in some analyses (Liddcll and
Armstrong (2002) d. 8) but not others. Aoorooriate
statistical method was used in (Bcrman and CruniD.
2008)
Medium
2
1.000
2.000
13. Statistical
power
Not applicable for mesothelioma studies
Not Rated
NA
NA
NA
14.
Reproducibility
of analyses
(Bcrman and CruiriD. 2008) reoorts Km estimate for
mesothelioma exposure-response.
Not Rated
NA
NA
NA
15. Statistical
Models
(Bcrman and CruniD. 2008) reoorts Km estimate for
mesothelioma exposure-response.
Not Rated
NA
NA
NA
High: >=1 and <1.7
Medium: >=1.7 and
<2.3
Low: >=2.3 and <=3
Sum of scores:
4
6.4334
Overall Score = Sum of Weighted Scores/Sum of Metric
Weighting Factors:
NA
Overall
Score:
Nearest
tenth:
NA
Overall Quality Level:
Medium
67
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This evaluation represents all identified publications pertaining to the Quebec, Canada cohort of miners. A full list
of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
Study
Quality
Comment:
The reviewer downgraded this study's overall quality rating. They noted: Lack of PCM or TEM-equivalent
exposure estimates and details/high uncertainty on conversion factor. Note: The original calculated score for this
study was 1.6. This value is not presented above because the final rating was changed based on professional
judgement
68
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
3.1.2. Epidemiology evaluation results of the South Carolina, US
cohort studies on asbestos exposure and mesothelioma
incidence
Study
This cohort evaluation represents all identified publications pertaining to the South Carolina cohort of textile
reference:
workers. A full list of related naners reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Weighting
Factor
Weighted
Score
a. Setting and methods of case ascertainment were
rcDortcd in (Dement et al.. 1983b). v. 422 and (Hcin
et al.. 2007). pe 617. Participant selection and
inclusion/exclusion criteria varied by study and
analysis. The initial cohort consisted of white men
a
e
o
-=3
O
employed for at least one month in a production job
at the South Carolina plant between 1/1/1940 and
.&
12/31/1965 ((Dement et al.. 1983b) d 422).
es
Oh
Subsequent analyses added non-white men and/or
women ((Stavner et al.. 1997). CDement et al..
High
i
0.500
0.500
O
1994). CBrown et al.. 1994). (Elliott et al.. 2012).
5
cd
CLh
(Edwards et al.. 2014). (Cole et al.. 2013). (Hcin et
tfl
al.. 2007)). b. Selection in or out of the studv was
based on 1) employment in production job during
designated time frame and 2) availability of
necessary data (birth and hire dates; work history;
vital status). These criteria are unlikely to result in
biased subject participation.
69
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
This cohort evaluation represents all identified publications pertaining to the South Carolina cohort of textile
reference:
workers. A full list of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Weighting
Factor
Weighted
Score
Attrition/missing data exclusions were reported in
(SRC. 2019cl (Dement et al.. 1983b^ (v. 423 and
Table 1) and (Hem et al.. 20071 see ps 618 and
Table 1. The number of eligible workers and
number excluded due to incomplete work histories
was not reported in the sources reviewed. ((Dement
et al.. 1983b) p 422) reports that each worker was
assigned a card at hire on which was tracked the
date of birth, sex, race, and SSN, and job or
department changes throughout the career were
recorded on the card; these were used in the cohort
O
establishment. The suggestion is that all cohort
£
members had complete work histories; however,
Low
0.500
1.500
((Hein et al.. 2007) ps 624) reported that the studv
(Hein et al.. 20071 vital status was
unknown for 8.6% of the larger cohort of 3072, and
cause of death was unknown for 3.9%. The latter
paper cited high rate of loss to follow-up as a
limitation of the study. Thus, loss to follow up is
judged to be moderate, and the lack of information
on number eligible and number with incomplete
work histories leads to low confidence.
3. Comparison
Group
Not applicable for mesothelioma studies
Not Rated
NA
NA
NA
(Dement et al.. 1983a) describes the plant processes
a
8
and the exposure estimation methods. More than
©
u
VI
6000 air samples obtained at the plant from 1930 to
N
¦e
Oh
X
1975 were analyzed by PCM to yield chrysotile
W
concentrations (fibers >5 um/mL) (Hein et al..
si
o
20071 pg 617. Exposure concentrations were
si
£
estimated by department, job, and time period;
High
1
0.400
0.400
U
s
individual cumulative exposure assessed using the
~-
=3
modeled concentrations and JEM (methods outlined
&
o
cd
in (Dement et al.. 1983a). In 2008. an updated JEM
o.
was developed to estimate fiber size-specific
exposure estimates (based on TEM analysis of
archived samples) in (Dement et al.. 20081
70
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This cohort evaluation represents all identified publications pertaining to the South Carolina cohort of textile
workers. A full list of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
5. Exposure
levels
(Bcrman and CruniD. 2008) reoorts Km estimate for
mesothelioma exposure-response data obtained
from the primary investigators for this cohort.
Medium
2
0.200
0.400
6. Temporality
Temporality was established (exposure preceded
death). (Hcin et al.. 2007) v 618 reoorts number of
years between first employment and death for one
of the 3 mesothelioma cases (~50 years). The other
two cases had been identified in an earlier analysis
(Dement et al.. 1994); that oaocr reported latencv
periods of 37 and 34 years for those cases.
High
1
0.400
0.400
Outcome Assessment
7. Outcome
measurement or
characterization
ICD-10 code C45 used to identify mesothelioma
cases after 1998; to identify earlier cases, death
certificates were reviewed for any mention of
mesothelioma (fHein et al.. 2007) v> 617).
High
1
0.667
0.667
8. Reporting
Bias
None of the publications reports exposure-response
information for mesothelioma: however (Bcrman
and CauriD. 2008) reoorts Km estimate for
mesothelioma exposure-response data obtained
from the primary investigators for this cohort.
Low
3
0.333
1.000
Potential Confounding/Variable Control
9. Covariate
Adjustment
Not applicable for mesothelioma studies
Not Rated
NA
NA
NA
10. Covariate
Characterization
Not applicable for mesothelioma studies
Not Rated
NA
NA
NA
11. Co-exposure
Confounding
Not applicable for mesothelioma studies
Not Rated
NA
NA
NA
71
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This cohort evaluation represents all identified publications pertaining to the South Carolina cohort of textile
workers. A full list of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
12. Study
Design and
Methods
Studv desisn was retrospective cohort (Hcin et al..
2007) and the aDDrooriatc statistical method was
used in (Bcrman and CruniD. 2008).
Medium
2
1.000
2.000
&
*5«
13. Statistical
power
Not applicable for mesothelioma studies
Not Rated
NA
NA
NA
es
C
<
14.
Reproducibility
of analyses
(Bcrman and CruniD. 2008) reoorts Km estimate for
mesothelioma exposure-response data obtained
from the primary investigators for this cohort.
Not Rated
NA
NA
NA
15. Statistical
Models
(Bcrman and CruniD. 2008) reoorts Km estimate for
mesothelioma exposure-response data obtained
from the primary investigators for this cohort.
Not Rated
NA
NA
NA
Sum of scores:
4
6.8667
High: >=1 and <1.7
Medium: >=1.7 and <2.3
Low: >-2.3 and <-3
Overall Score = Sum of Weighted Scores/Sum of Metric
Weighting Factors:
1.7167
Overall
Score:
Nearest
tenth:
1.7
Overall Quality Level:
Medium
Study
Quality
Comment:
The reviewer agreed with this study's overall quality level.
72
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
3.1.3. Epidemiology evaluation results of the North Carolina, US
cohort studies on asbestos exposure and mesothelioma
incidence
Study
This evaluation represents all identified publications pertaining to the North Carolina cohort of textile workers. A
reference:
full list of related papers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Weighting
Factor
Weighted
Score
a. Setting and methods of case ascertainment were
reported in (Loomis et al.. 2009). p. 535-536.
Participant selection and inclusion/exclusion criteria
varied by study and analysis. Although there were 4
plants in the cohort, exposure data were available
O
-=3
O
only for three of the four, so exposure-response
analyses were limited to these three plants. Original
selection criteria reported in (Loomis et al.. 2009) v
Oh
536 (participants had to work at least 1 day between
High
i
0.500
0.500
1950 and 1973) andp 539 (participants excluded
O
due to missine data). (Elliott et al.. 2012) evaluated
CLh
a subset of the cohort that worked >30 days during
the same time frame, b. Selection in or out of the
study was based on 1) employment in production
job during designated time frame and b) availability
of necessary data (birth and hire dates; work history;
vital status). These criteria are unlikely to result in
#o
biased subject participation.
es
.&
Attrition/missing data exclusions were reported in
((SRC. 2019a) n. 1). ((Loomis et al.. 2009) d. 539)
es
Q-
and (Elliott et al.. 2012) ps 386. The orieinal cohort
was 5770 persons; 373 workers at plant 2 were
5
excluded due to lack of exposure data at this plant,
Cfl
fl
o
1596 were excluded due to incomplete work
histories (at department level) or non-production
s
iobs ((Loomis et al.. 2009) p. 539). Final cohort for
Medium
2
0.500
1.000
<
exposure-response analyses was 3803. Vital status
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
This evaluation represents all identified publications pertaining to the North Carolina cohort of textile workers. A
reference:
full list of related papers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
(SRC. 2019a) reports air concentrations and
exposure duration by interval of TSFE. Air samples
were available for 3 plants covering period from
8
1935 to 1986 (459 <1950; 1674 from 1950-1969,
u
VI
and 1287 from 1970 forward: (Loomis et al.. 20091
Oh
X
W
p. 536). Measurements used impinger before 1964
and PCM thereafter; paired and concurrent samples
o
between 1964 and 1971 were used to relate
"H
impinger to PCM-equivalent concentrations. Air
Medium
2
0.400
0.800
S
samples were not collected yearly, so mean PCM-
to
C/3
equivalent concentrations were estimated by plant,
N
cd
1
department, job, and time period using multivariate
¦e
mixed models (YLoomis et al.. 2009). v. 536).
t5
Individual cumulative exposure assessed using the
es
modeled concentrations and JEM ((Loomis et al..
-fl
u
2009) p 536); details of JEM reported in (Dement et
B
al.. 2008).
s
%
o
o.
UJ
5. Exposure
levels
(SRC. 2019a) reoorts air concentrations and
exposure durations by interval of TSFE.
Medium
2
0.200
0.400
.1?
2
o
Temporality was established (exposure preceded
1
death). (SRC. 2019a) reports cases bv interval of
High
1
0.400
0.400
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
This evaluation represents all identified publications pertaining to the North Carolina cohort of textile workers. A
reference:
full list of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
(SRC. 2019a) provides mesothelioma cases and
person-years at risk by interval of TSFE, including
separate reporting of those assessed by ICD10.
Mesothelioma cases (with detail of those assessed
by ICD10) reported by employment duration in
o
^ o
Analysis
12. Study
Design and
Methods
Study design was retrospective cohort and SMR
analvsis was oerformed. (Loomis et al.. 2009)
Medium
2
1.000
2.000
75
-------
PEER REVIEW DRAFT. DO NOT CITE OR QUOTE
Study
reference:
This evaluation represents all identified publications pertaining to the North Carolina cohort of textile workers. A
full list of related DaDers reviewed during this evaluation are listed in Table 2.1.
Domain
Metric
Comments
Qualitative
Determination
Metric Score
Metric
Weighting
Factor
Weighted
Score
13. Statistical
power
Not applicable for mesothelioma studies
Not Rated
NA
NA
NA
14.
Reproducibility
of analyses
(SRC. 2019a) orovidcs individual data elements
allowing independent analysis
Not Rated
NA
NA
NA
15. Statistical
Models
(SRC. 2019a) orovidcs individual data elements
allowing independent analysis
Not Rated
NA
NA
NA
High: >=1 and <1.7
Medium: >=1.7 and
<2.3
Low: >=2.3 and <=3
Sum of scores:
4
6.1
Overall Score = Sum of Weighted Scores/Sum of Metric
Weighting Factors:
1.525
Overall
Score:
Nearest
tenth:
1.5
Overall Quality Level:
High
Study
Quality
Comment:
The reviewer agreed with this study's overall quality level.
76
-------
Asbestos
References
Berman. DW; Crump. KS. (2008). Update of potency factors for asbestos-related lung cancer and mesothelioma. Crit Rev Toxicol 38:1-
47. http://dx.doi.org/10.1080/104084408Q2276167
Blettner. M; Heuer. C: Razum. O. (2001). Critical reading of epidemiological papers. A guide. Eur J Public Health 11: 97-101.
Brown. DP: Dement. JM; Okun. A. (1994). Mortality patterns among female and male chrysotile asbestos textile workers. J Occup Med
36: 882-888.
CDC. (2009). Malignant mesothelioma mortality: United States, 19992005. MMWRMorb Mortal Wkly Rep 58: 393-396.
Cole. SR; Richardson. DB; Chu. H; Naimi. AI. (2013). Analysis of occupational asbestos exposure and lung cancer mortality using the g
formula. Am J Epidemiol 177: 989-996. http://dx.doi.org/10.1093/aie/kws343
Cooper. GS: Lunn. RM; Agerstrand. M; Glenn. BS: Kraft. AD: Luke. AM: Ratcliffe. JM. (2016). Study sensitivity: Evaluating the ability
to detect effects in systematic reviews of chemical exposures. Environ Int 92-93: 605-610.
http://dx.doi.Org/10.1016/i.envint.2016.03.017
Courtice. MN; Wang. X: Lin. S: Yu. IT: Berman. DW: Yano. E. (2016). Exposure-response estimate for lung cancer and asbestosis in a
predominantly chrysotile-exposed Chinese factory cohort. Am J Ind Med 59: 369-378. http://dx.doi.org/10.10Q2/aiim.22579
Dement. JM: Brown. DP. (1994a). Cohort Mortality and Case-Control Studies of White Male Chrysotile Asbestos Textile Workers (pp.
Asbestos and Cancers). (NIOSH/00224027). Dement, JM; Brown, DP.
Dement. JM: Brown. DP. (1994b). Lung cancer mortality among asbestos textile workers: a review and update. Ann Occup Hyg 38: 525-
532, 412.
Dement. JM: Brown. DP: Okun. A. (1994). Follow-up study of chrysotile asbestos textile workers: cohort mortality and case-control
analyses. Am J Ind Med 26: 431-447.
Dement. JM: Harris. RL. Jr: Symons. MJ: Shy. CM. (1983a). Exposures and mortality among chrysotile asbestos workers: Part 1:
Exposure estimates. Am J Ind Med 4: 399-419. http://dx.doi.org/10.1002/aiim.47000403Q3
Dement. JM: Harris. RL. Jr: Symons. MJ: Shy. CM. (1983b). Exposures and mortality among chrysotile asbestos workers: Part II:
Mortality. Am J Ind Med 4: 421-433. http://dx.doi.org/10.1002/aiim.47000403Q4
Dement. JM: Kuempel. ED: Zumwalde. RD: Smith. RJ: Stavner. LT: Loomis. D. (2008). Development of a fibre size-specific job-
exposure matrix for airborne asbestos fibres. J Occup Environ Med 65: 605-612. http://dx.doi.Org/10.l 136/oem.2007.033712
Deng. 0: Wang. X: Wang. M: Lan. Y. (2012). Exposure-response relationship between chrysotile exposure and mortality from lung
cancer and asbestosis. Occup Environ Med 69: 81-86. http://dx.doi.Org/10.l 136/oem.2011.064899
Dodic Fikfak. M. (2003). The amphibole hypothesisa nested case-control study of lung cancer and exposure to chrysotile and
amphiboles. Arh Hig Rada Toksikol 54: 169-176.
77
-------
Asbestos
Dodic Fikfak. M; Kriebel. D; Ouinn. MM; Eisen. EA; Wegman. DH. (2007). A case control study of lung cancer and exposure to
chrysotile and amphibole at a Slovenian asbestos-cement plant. Ann Occup Hyg 51: 261-268.
http://dx.doi.org/10.1093/annhyg/mem003
Edwards. JK; Cole. SR; Chu. H; Olshan. AF; Richardson. DB. (2014). Accounting for outcome misclassification in estimates of the effect
of occupational asbestos exposure on lung cancer death. Am J Epidemiol 179: 641-647.
Elliott. L; Loomis. D; Dement. J: Hein. MJ; Richardson. D; Stavner. L. (2012). Lung cancer mortality in North Carolina and South
Carolina chrysotile asbestos textile workers. Occup Environ Med 69: 385-390. http://dx.doi.org/10.1136/oemed-2011-100229
Gibbs. GW: Lachance. M. (1972). Dust exposure in chrysotile asbestos mines and mills of Quebec. Arch Environ Health 24: 189-197.
http://dx.doi.org/10.1080/00039896.1972.10666Q68
Hein. MJ: Stavner. LT; Lehman. E; Dement. JM. (2007). Follow-up study of chrysotile textile workers: Cohort mortality and exposure-
response. Occup Environ Med 64: 616-625. http://dx.doi.Org/10.l 136/oem.2006.031005
Kopylev. L; Sullivan. PA: Vinikoor. LC: Bateson. TF. (2011). Monte Carlo analysis of impact of underascertainment of Mesothelioma
cases on underestimation of risk. Open Epidemiol J 4: 45-53.
Lakind. JS: Sobus. J: Goodman. M; Barr. DB: Fuerst. P; Albertini. RJ; Arbuckle. T; Schoeters. G: Tan. Y; Teeguarden. J: Tornero-Velez.
R; Weisel. CP. (2014). A proposal for assessing study quality: Biomonitoring, Environmental Epidemiology, and Short-lived
Chemicals (BEES-C) instrument. Environ Int 73: 195-207. http://dx.doi.Org/10.1016/i.envint.2014.07.011
Liddell. FD; Armstrong. BG. (2002). The combination of effects on lung cancer of cigarette smoking and exposure in quebec chrysotile
miners and millers. Ann Occup Hyg 46: 5-13. http://dx.doi.org/10.1093/annhyg/mef008
Liddell. FD: Mcdonald. AD: Mcdonald. JC. (1997). The 1891-1920 birth cohort of Quebec chrysotile miners and millers: development
from 1904 and mortality to 1992. Ann Occup Hyg 41: 13-36. http://dx.doi.org/10.1016/S0003-4878(96)00044-0
Liddell. FD: Mcdonald. AD: Mcdonald. JC. (1998). Dust exposure and lung cancer in Quebec chrysotile miners and millers. Ann Occup
Hyg 42: 7-20.
Loomis. D; Dement. J; Richardson. D; Wolf. S. (2010). Asbestos fibre dimensions and lung cancer mortality among workers exposed to
chrysotile. Occup Environ Med 67: 580-584. http://dx.doi.Org/10.l 136/oem.2009.050120
Loomis. D; Dement. JM; Elliott. L; Richardson. D; Kuempel. ED; Stavner. L. (2012). Increased lung cancer mortality among chrysotile
asbestos textile workers is more strongly associated with exposure to long thin fibres. Occup Environ Med 69: 564-568.
http://dx.doi. org /10.1136/oemed-2012-100676
Loomis. D; Dement. JM; Wolf. SH; Richardson. DB. (2009). Lung cancer mortality and fibre exposures among North Carolina asbestos
textile workers. Occup Environ Med 66: 535-542. http://dx.doi.org/10.1136/oem.2008.044362
Loomis. D; Richardson. DB; Elliott. L. (2019). Quantitative relationships of exposure to chrysotile asbestos and mesothelioma mortality.
Am J Ind Med 62: 471-477. http://dx.doi.org/10.1002/aiim.22985
Mcdonald. JC; Liddell. FD; Dufresne. A; Mcdonald. AD. (1993a). The 1891-1920 birth cohort of Quebec chrysotile miners and millers:
78
-------
Asbestos
mortality 1976-88. Br JIndMed 50: 1073-1081.
Mcdonald. JC; Liddell. FD; Gibbs. GW; Eyssen. GE; Mcdonald. AD. (1993b). Dust exposure and mortality in chrysotile mining, 1910-
75. 1980. Br JIndMed 50: 1058-1072.
NTP. (2015). Handbook for conducting a literature-based health assessment using OHAT approach for systematic review and evidence
integration. U.S. Dept. of Health and Human Services, National Toxicology Program.
https://ntp.niehs.nih.gov/ntp/ohat/pubs/handbookian2015 508.pdf
Peto. J. (1978). The hygiene standard for chrysotile asbestos. Lancet 1: 484-489. http://dx.doi.org/10.1016/S0140-6736(78)90145-9
Peto. J: Seidman. H; Selikoff. IJ. (1982). Mesothelioma mortality in asbestos workers: implications for models of carcinogenesis and risk
assessment. Br J Cancer 45: 124-135. http://dx.doi.org/10.1038/bic.1982.15
Piolatto. G: Negri. E; La Vecchia. C: Pira. E; Decarli. A: Peto. J. (1990). An update of cancer mortality among chrysotile asbestos miners
in Balangero, northern Italy. Br J Ind Med 47: 810-814.
Pira. E; Pelucchi. C: Piolatto. PG: Negri. E; Bilei. T; La Vecchia. C. (2009). Mortality from cancer and other causes in the Balangero
cohort of chrysotile asbestos miners. Occup Environ Med 66: 805-809. http://dx.doi.Org/10.l 136/oem.2008.044693
Pira. E; Romano. C: Donat. F; Pelucchi. C: La Vecchia. C: Boffetta. P.. (2017). Mortality from cancer and other causes among Italian
chrysotile asbestos miners. Occup Environ Med 74: 558-563. http://dx.doi.Org/10.l 136/oemed-2016-103673
Rubino. GF; Piolatto. G: Newhouse. ML: Scansetti. G: Aresini. GA; Murray. R. (1979). Mortality of chrysotile asbestos workers at the
Balangero Mine, northern Italy. Occup Environ Med 36: 187-194.
SRC. (2019a). A Summary on the North Carolina Cohort of Asbestos Textile Workers.
SRC. (2019b). A Summary on the Quebec Cohort of Asbestos Miners and Millers.
SRC. (2019c). A Summary on the South Carolina Cohort of Asbestos Textile Workers.
Stavner. L; Kuempel. E; Gilbert. S: Hein. M; Dement. J. (2008). An epidemiological study of the role of chrysotile asbestos fibre
dimensions in determining respiratory disease risk in exposed workers. Occup Environ Med 65: 613-619.
http://dx.doi.Org/10.l 136/oem.2007.035584
Stavner. L; Smith. R; Bailer. J: Gilbert. S: Steenland. K; Dement. J: Brown. D; Lemen. R. (1997). Exposure-response analysis of risk of
respiratory disease associated with occupational exposure to chrysotile asbestos. Occup Environ Med 54: 646-652.
Tossavainen. A. (1997). ASBESTOS ASBESTOSIS AND CANCER THE HELSINKI CRITERIA FOR DIAGNOSIS AND
ATTRIBUTION HELSINKI FINLAND JANUARY 20-22 1997. Scand J Work Environ Health 23: 311-316.
U.S. EPA. (1988). IRIS summary for asbestos (CASRN 1332-21-4). Washington, DC: U.S. Environmental Protection Agency, Integrated
Risk Information System, http://www.epa.gov/iris/subst/0371.htm
U.S. EPA. (2014). Toxicological review of libby amphibole asbestos: In support of summary information on the Integrated Risk
Information System (IRIS) [EPA Report], (EPA/635/R-11/002F). Washington, DC: Integrated Risk Information System, National
Center for Environmental Assessment, Office of Research and Development.
79
-------
Asbestos
https://cfpub.epa.gov/ncea/iris/iris documents/documents/toxreviews/1026tr.pdf
Vacek. PM. (1998). Effects of the intensity and timing of asbestos exposure on lung cancer risk at two mining areas in Quebec. J Occup
Environ Med 40: 821-828.
Von Elm. E; Altman. DG: Egger. M; Pocock. SJ: G0tzsche. PC: Vandenbroucke. JP. (2008). The Strengthening the Reporting of
Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. J Clin Epidemiol 61:
344-349. http://dx.doi.Org/10.1016/i.iclinepi.2007.l 1.008
Wang. X: Lin. S: Yano. E; Qiu. H; Yu. IT: Tse. L; Lan. Y; Wang. M. (2012). Mortality in a Chinese chrysotile miner cohort. Int Arch
Occup Environ Health 85: 405-412. http://dx.doi.org/10.1007/sQ0420-011-0685-9
Wang. X: Lin. S: Yano. E; Yu. IT: Courtice. M; Lan. Y; Christiani. DC. (2014). Exposure-specific lung cancer risks in Chinese chrysotile
textile workers and mining workers. Lung Cancer 85: 119-124. http://dx.doi.Org/10.1016/i.lungcan.2014.04.011
Wang. X: Yano. E; Lin. S: Yu. IT: Lan. Y; Tse. LA: Qiu. H; Christiani. DC. (2013). Cancer mortality in Chinese chrysotile asbestos
miners: exposure-response relationships. PLoS ONE 8: e71899. http://dx.doi.org/10.1371/iournal.pone.0071899
Yano. E; Wang. ZM; Wang. XR; Wang. MZ; Lan. YJ. (2001). Cancer mortality among workers exposed to amphibole-free chrysotile
asbestos. Am J Epidemiol 154: 538-543.
80
------- |