Robert L. Sielken

Exact Confidence Intervals for Linear Combinations of Variance Components in Nested Classifications

Abstract Methodology is proposed for the construction of exact confidence intervals on nonnegative linear combinations of variance components from nested classification models. Examples are given for the one-fold and two-fold classifications. The robustness of these confidence intervals to model breakdown is also discussed.

Development of a cancer-based chronic inhalation reference value for hexavalent chromium based on a nonlinear-threshold carcinogenic assessment

The carcinogenicity of hexavalent chromium(CrVI) is of significant interest to regulatory agencies for the protection of public health and to industry. Additionally, the mode of action (MOA) and conditions under which CrVI may induce carcinogenicity (e.g., reductive capacity considerations) have recently been the subject of significant scientific debate. Epidemiological data supported by data relevant to the carcinogenic MOA support considering nonlinear-threshold carcinogenic assessments for comparison to default linear low-dose extrapolation approaches. This study reviews epidemiological studies available in the scientific literature and conducts additional statistical dose-response analyses to identify potential carcinogenic thresholds and points of departure (PODs) in the context of supportive MOA information for a nonlinear-threshold inhalation carcinogenic assessment. Dosimetric adjustments and application of appropriate uncertainty factors (total UF of 30) to the selected cumulative exposure POD results in a cancer-based chronic inhalation reference value (ReV) of 0.24 μgCrVI/m(3). This chronic ReV is 300 times higher than the 1 in 100,000 excess cancer risk air concentration of 8E-04 μg/m(3) based on USEPAs unit risk factor.

Development of an inhalation unit risk factor for hexavalent chromium

A unit risk factor (URF) was developed for hexavalent chromium (CrVI). The URF is based on excess lung cancer mortality in two key epidemiological studies of chromate production workers. The Crump et al. (2003) study concerns the Painesville, OH worker cohort, while Gibb et al. (2000) regards the Baltimore, MD cohort. A supporting assessment was also performed for a cohort from four low-dose chromate plants (Leverkusen and Uerdingen, Germany, Corpus Christi, TX, Castle Hayne, NC). For the Crump et al. (2003) study, grouped observed and expected number of lung cancer mortalities along with cumulative CrVI exposures were used to obtain the maximum likelihood estimate and asymptotic variance of the slope (β) for the linear multiplicative relative risk model using Poisson regression modeling. For the Gibb et al. (2000) study, Cox proportional hazards modeling was performed with optimal exposure lag and adjusting for the effect of covariates (e.g., smoking) to estimate β values. Life-table analyses were used to develop URFs for each of the two key studies, as well as for supporting and related studies. The two key study URFs were combined using weighting factors relevant to confidence to derive the final URF for CrVI of 2.3E-03 per μgCrVI/m(3).

Some issues in the quantitative modeling portion of cancer risk assessment.

Several questions should be asked in order to determine the relevance and scientific merit of a quantitative cancer risk assessment. Twenty such questions are formulated here and briefly discussed. These questions are intended to identify important issues and serve as a checklist for risk managers and developers of quantitative risk assessments. Among the many factors involved in these questions are the carcinogenic response, quantal response models, time to response, competing risks, model shapes, goodness of fit, dose scale, high-to-low-dose extrapolation, consistency across different subjects, animal-to-human extrapolation, route-to-route extrapolation, exposure durations and patterns, short-term tests, consistency with human epidemiological data, human exposures, statistical variability, impacts of assumptions and policy decisions and value judgments, risk characterization, upper and lower bounds, and keeping pace with scientific advances.

An updated inhalation unit risk factor for arsenic and inorganic arsenic compounds based on a combined analysis of epidemiology studies

The United States Environmental Protection Agency (USEPA) developed an inhalation unit risk factor (URF) of 4.3E-03 per μg/m(3) for arsenic in 1984 for excess lung cancer mortality based on epidemiological studies of workers at two smelters: the Asarco smelter in Tacoma, Washington and the Anaconda smelter in Montana. Since the USEPA assessment, new studies have been published and exposure estimates were updated at the Asarco and Anaconda smelters and additional years of follow-up evaluated. The Texas Commission on Environmental Quality (TCEQ) has developed an inhalation URF for lung cancer mortality from exposures to arsenic and inorganic arsenic compounds based on a newer epidemiology study of Swedish workers and the updates of the Asarco and Anaconda epidemiology studies. Using a combined analysis approach, the TCEQ weighted the individual URFs from these three epidemiology cohort studies, to calculate a final inhalation URF of 1.5E-04 per μg/m(3). In addition, the TCEQ also conducted a sensitivity analysis, in which they calculated a URF based on a type of meta-analysis, and these results compared well with the results of the combined analysis. The no significant concentration level (i.e., air concentration at 1 in 100,000 excess lung cancer mortality) is 0.067μg/m(3). This value will be used to evaluate ambient air monitoring data so the general public in Texas is protected against adverse health effects from chronic exposure to arsenic.

Quantitative cancer risk assessment based on NIOSH and UCC epidemiological data for workers exposed to ethylene oxide

The most recent epidemiological data on individual workers in the NIOSH and updated UCC occupational studies have been used to characterize the potential excess cancer risks of environmental exposure to ethylene oxide (EO). In addition to refined analyses of the separate cohorts, power has been increased by analyzing the combined cohorts. In previous SMR analyses of the separate studies and the present analyses of the updated and pooled studies of over 19,000 workers, none of the SMRs for any combination of the 12 cancer endpoints and six sub-cohorts analyzed were statistically significantly greater than one including the ones of greatest previous interest: leukemia, lymphohematopoietic tissue, lymphoid tumors, NHL, and breast cancer. In our study, no evidence of a positive cumulative exposure-response relationship was found. Fitted Cox proportional hazards models with cumulative EO exposure do not have statistically significant positive slopes. The lack of increasing trends was corroborated by categorical analyses. Cox model estimates of the concentrations corresponding to a 1-in-a-million extra environmental cancer risk are all greater than approximately 1ppb and are more than 1500-fold greater than the 0.4ppt estimate in the 2006 EPA draft IRIS risk assessment. The reasons for this difference are identified and discussed.

Sequencing with Setup Costs by Zero-One Mixed Integer Linear Programming

Abstract The classical problem of sequencing jobs on a single processor is given a new formulation. Included in the formulation are sequence-dependent setup costs and times, inventory costs for completed jobs, and penalty costs for completion delays. The formulation results in a zero-one mixed integer linear programming problem.

Butadiene cancer exposure-response modeling: based on workers in the styrene-butadiene-rubber industry: total leukemia, acute myelogenous leukemia, chronic lymphocytic leukemia, and chronic myelogenous leukemia.

Cox regression is used to estimate exposure-response models (with cumulative 1,3-butadiene (BD) ppm-years as the exposure metric) based on the most recent data and validated exposure estimates from UABs study of North American workers in the styrene-butadiene-rubber industry. These data are substantially updated from those in USEPAs 2002 risk assessment. The slope for cumulative BD ppm-years is not statistically significantly different than zero for CML, AML, or, when any one of eight exposure covariates is added to the model, for all leukemias combined (total leukemia). For total leukemia, the EC(1/100,000) is approximately 0.15 BD environmental ppm and the corresponding unit risk factor is approximately 0.00007 per BD environmental ppm. The excess risk for CML is approximately 15-fold less than for total leukemia. The maximum likelihood estimates suggest that there is no excess risk for AML from cumulative BD ppm-years. For CLL, the slope is statistically significantly different than zero. The excess risk for CLL is approximately 2.5-fold less than for total leukemia. For both total leukemia and CLL, the slope is not statistically significantly different than zero when the exposure-response modeling is based on the person-years with cumulative BD ppm-years less than or equal to 300 ppm-years.

Development of a unit risk factor for nickel and inorganic nickel compounds based on an updated carcinogenic toxicity assessment

The TCEQ has developed a URF for nickel based on excess lung cancer in two epidemiological studies of nickel refinery workers with nickel species exposure profiles most similar to emissions expected in Texas (i.e., low in sulfidic nickel). One of the studies (Enterline and Marsh, 1982) was used in the 1986 USEPA assessment, while the other (Grimsrud et al., 2003) is an update to an earlier study (Magnus et al., 1982) used by USEPA. The linear multiplicative relative risk model with Poisson regression modeling was used to obtain maximum likelihood estimates and asymptotic variances for cancer potency factors (β) using cumulative nickel exposure levels versus observed and expected lung cancer mortality (Enterline and Marsh, 1982) or lung cancer incidence cases (Grimsrud et al., 2003). Life-table analyses were then used to develop URFs from these two studies, which were combined using weighting factors relevant to confidence to derive the final URF for nickel of 1.7E-04 per μg/m³. The de minimis air concentration corresponding to a 1 in 100,000 extra lung cancer risk level is 0.059 μg/m³. The TCEQ will use this conservative value to protect the general public in Texas against the potential carcinogenic effects from chronic exposure to nickel.

Variance Estimation Based on a Superpopulation Model in Two-Stage Sampling

Abstract Estimators for finite population parameters and their variances in two-stage sampling have been developed by using the linear least-squares prediction approach in a recent article by Royall (1976). This article considers a special case of the superpopulation model assumed by Royall and uses a new technique involving linear combinations of the sample observations to estimate the variances of these estimators. An exact confidence interval for the finite population total is calculated for the case in which all clusters have an equal number of elements and an equal number of elements are sampled from each selected cluster.