Leonid Kopylev

Physiologically Based Pharmacokinetic Model Use in Risk Assessment—Why Being Published Is Not Enough

A panel of experts in physiologically based pharmacokinetic (PBPK) modeling and relevant quantitative methods was convened to describe and discuss model evaluation criteria, issues, and choices that arise in model application and computational tools for improving model quality for use in human health risk assessments (HHRAs). Although publication of a PBPK model in a peer-reviewed journal is a mark of good science, subsequent evaluation of published models and the supporting computer code is necessary for their consideration for use in HHRAs. Standardized model evaluation criteria and a thorough and efficient review process can reduce the number of review and revision iterations and hence the time needed to prepare a model for application. Efficient and consistent review also allows for rapid identification of needed model modifications to address HHRA-specific issues. This manuscript reports on the workshop where a process and criteria that were created for PBPK model review were discussed along with other issues related to model review and application in HHRA. Other issues include (1) model code availability, portability, and validity; (2) probabilistic (e.g., population-based) PBPK models and critical choices in parameter values to fully characterize population variability; and (3) approaches to integrating PBPK model outputs with other HHRA tools, including benchmark dose modeling. Two specific case study examples are provided to illustrate challenges that were encountered during the review and application process. By considering the frequent challenges encountered in the review and application of PBPK models during the model development phase, scientists may be better able to prepare their models for use in HHRAs.

Nonparametric Bayesian Methods for Benchmark Dose Estimation

The article proposes and investigates the performance of two Bayesian nonparametric estimation procedures in the context of benchmark dose estimation in toxicological animal experiments. The methodology is illustrated using several existing animal dose-response data sets and is compared with traditional parametric methods available in standard benchmark dose estimation software (BMDS), as well as with a published model-averaging approach and a frequentist nonparametric approach. These comparisons together with simulation studies suggest that the nonparametric methods provide a lot of flexibility in terms of model fit and can be a very useful tool in benchmark dose estimation studies, especially when standard parametric models fail to fit to the data adequately.

Approaches to cancer assessment in EPA's Integrated Risk Information System

The U.S. Environmental Protection Agencys (EPA) Integrated Risk Information System (IRIS) Program develops assessments of health effects that may result from chronic exposure to chemicals in the environment. The IRIS database contains more than 540 assessments. When supported by available data, IRIS assessments provide quantitative analyses of carcinogenic effects. Since publication of EPAs 2005 Guidelines for Carcinogen Risk Assessment, IRIS cancer assessments have implemented new approaches recommended in these guidelines and expanded the use of complex scientific methods to perform quantitative dose-response assessments. Two case studies of the application of the mode of action framework from the 2005 Cancer Guidelines are presented in this paper. The first is a case study of 1,2,3-trichloropropane, as an example of a chemical with a mutagenic mode of carcinogenic action thus warranting the application of age-dependent adjustment factors for early-life exposure; the second is a case study of ethylene glycol monobutyl ether, as an example of a chemical with a carcinogenic action consistent with a nonlinear extrapolation approach. The use of physiologically based pharmacokinetic (PBPK) modeling to quantify interindividual variability and account for human parameter uncertainty as part of a quantitative cancer assessment is illustrated using a case study involving probabilistic PBPK modeling for dichloromethane. We also discuss statistical issues in assessing trends and model fit for tumor dose-response data, analysis of the combined risk from multiple types of tumors, and application of life-table methods for using human data to derive cancer risk estimates. These issues reflect the complexity and challenges faced in assessing the carcinogenic risks from exposure to environmental chemicals, and provide a view of the current trends in IRIS carcinogenicity risk assessment.

Low levels of exposure to libby amphibole asbestos and localized pleural thickening.

Objective: To explore the relationship between low levels of exposure to Libby amphibole asbestos (LAA) and pleural abnormalities, specifically localized pleural thickening (LPT). Methods: Three studies presenting the risks associated with quantitative LAA exposure estimates were reviewed, paying particular attention to lower exposure ranges. Results: Studies reviewed were conducted among workers exposed to LAA at mining and milling operations in Libby, Montana, at a vermiculite processing facility in Marysville, Ohio, and community residents exposed to LAA from a vermiculite processing facility in Minneapolis, Minnesota. Pleural abnormalities were evaluated using radiographs. Despite differences in study populations and design, each study found that cumulative inhalation LAA exposure was associated with increased risk of LPT even at low levels of exposure. Conclusions: Inhalation exposure to LAA is associated with increased risk of LPT even at the lowest levels of exposure in each study.

Localized pleural thickening: Smoking and exposure to Libby vermiculite

There is limited research on the combined effects of smoking and asbestos exposure on risk of localized pleural thickening (LPT). This analysis uses data from the Marysville cohort of workers occupationally exposed to Libby amphibole asbestos (LAA). Workers were interviewed to obtain work and health history, including ever/never smoking and chest X-rays. Cumulative exposure estimates were developed on the basis of fiber measurements from the plant and work history. Benchmark concentration (BMC) methodology was used to evaluate the exposure–response relationship for exposure to LAA and a 10% increased risk of LPT, considering potential confounders and statistical model forms. There were 12 LPT cases among 118 workers in the selected study population. The mean exposure was 0.42 (SD=0.77) fibers/cc-year, and the prevalence of smoking history was 75.0% among cases and 51.9% among non-cases. When controlling for LAA exposure, smoking history was of borderline statistical significance (P-value=0.099), and its inclusion improved model fit, as measured by Akaikes Information Criterion. A comparison of BMC estimates was made to gauge the potential effect of smoking status. The BMC was 0.36 fibers/cc-year, overall. The BMC for non-smokers was approximately three times as high (1.02 fibers/cc-year) as that for the full cohort, whereas the BMC for smokers was about 1/2 that of the full cohort (0.17 fibers/cc-year).

Monte Carlo Analysis of Impact of Underascertainment of Mesothelioma Cases on Underestimation of Risk

The accuracy of cancer mortality data varies across different cancers. Mortality records and death certificates may not always reflect the true cause of death for various reasons (e.g., misdiagnosis, improper recording on the death certificate, miscoding of the cause of death recorded on the death certificate). Mesothelioma, a rare tumor which is caused by exposure to asbestos, is a special case. Until 1999 when the 10 th revision of the International Classification of Diseases (ICD-10) introduced a specific mesothelioma code, mesothelioma deaths were coded to other causes (e.g., cancer of the pleura, cancer of other or ill-specified sites). Even after the introduction of this mesothelioma code, researchers have shown that estimates of mesothelioma mortality based on death certificates are still biased downward. This article reviews available literature with quantitative information on mesothelioma underascertainment, in particular on different rates of underestimation for pleural and peritoneal mesotheliomas, and suggests two approaches to estimating downward bias in absolute risk estimates due to mesothelioma underascertainment. The choice of approach used depends on whether the information on the proportion of peritoneal mesotheliomas is available. Both approaches are demonstrated and evaluated using a cohort of asbestos workers from Libby, MT. The methods developed in this article may be used in analyses of other asbestos cohorts and in methodologies to predict future mesothelioma burden in populations. Similar approaches can be used to assess the impact of underascertainment of other cancers on risk estimates of other chemicals.

Parameters of a Dose‐Response Model Are on the Boundary: What Happens with BMDL?

It is well known that, under appropriate regularity conditions, the asymptotic distribution for the likelihood ratio statistic is chi(2). This result is used in EPAs benchmark dose software to obtain a lower confidence bound (BMDL) for the benchmark dose (BMD) by the profile likelihood method. Recently, based on work by Self and Liang, it has been demonstrated that the asymptotic distribution of the likelihood ratio remains the same if some of the regularity conditions are violated, that is, when true values of some nuisance parameters are on the boundary. That is often the situation for BMD analysis of cancer bioassay data. In this article, we study by simulation the coverage of one- and two-sided confidence intervals for BMD when some of the model parameters have true values on the boundary of a parameter space. Fortunately, because two-sided confidence intervals (size 1-2alpha) have coverage close to the nominal level when there are 50 animals in each group, the coverage of nominal 1-alpha one-sided intervals is bounded between roughly 1-2alpha and 1. In many of the simulation scenarios with a nominal one-sided confidence level of 95%, that is, alpha= 0.05, coverage of the BMDL was close to 1, but for some scenarios coverage was close to 90%, both for a group size of 50 animals and asymptotically (group size 100,000). Another important observation is that when the true parameter is below the boundary, as with the shape parameter of a log-logistic model, the coverage of BMDL in a constrained model (a case of model misspecification not uncommon in BMDS analyses) may be very small and even approach 0 asymptotically. We also discuss that whenever profile likelihood is used for one-sided tests, the Self and Liang methodology is needed to derive the correct asymptotic distribution.

Authors' response: A systematic review of the association between pleural plaques and changes in lung function.

We welcome the opportunity to respond to Goodman et al ,1 and to correct their misperceptions about our paper.2 As noted in their letter, Kerper et al 3 also recently analysed lung function decrements associated with pleural plaques. While the methodological details of our publications differed somewhat, the identified literature and the conclusions regarding magnitude of effect on lung function were well aligned. We found statistically significant 2–4% decrements in lung function in people exposed to asbestos with pleural plaques relative to asbestos-exposed people without abnormalities. Kerper et al 3 reported 3–5% decrements. It is not clear why Kerper et al 3 chose to ignore differences in study size: all studies were considered equally in their analysis, despite sample sizes ranging from tens to thousands, and a summary estimate was not calculated. With respect to the specific points raised, …

Approaches to Calculation of Average Exposure in Analysis of Epidemiologic Cohorts Using Large Arylonitrile Cohort As An Example

ȠObjectives: To explore two different approaches to calculate average exposure in occupational cohorts using a large occupational cohort as an example. The data for occupational cohort exposed to acrylonitrile was collected and analyzed previously by NCI; outcome was lung cancer. Methods: Both approaches use cumulative exposure as the numerator. As the denominator, one uses the duration of exposure, while the other uses the length of employment. The former approach is used when detailed exposure history is available, and the latter is used when exposure history is less detailed. The differences are investigated for a large occupational cohort. Results: With restricting the cohort to only those with enough latency for lung cancer, the cumulative exposure divided by the length of employment is a significant predictor of the lung cancer mortality, while cumulative exposure divided by the duration of exposure (average intensity) is not. Analysis is shown not to be positively confounded by smoking.