Bayesian hierarchical joint modeling of repeatedly measured continuous and ordinal markers of disease severity: Application to Ugandan diabetes data.

Modeling of correlated biomarkers jointly has been shown to improve the efficiency of parameter estimates, leading to better clinical decisions. In this paper, we employ a joint modeling approach to a unique diabetes dataset, where blood glucose (continuous) and urine glucose (ordinal) measures of disease severity for diabetes are known to be correlated. The postulated joint model assumes that the outcomes are from distributions that are in the exponential family and hence modeled as multivariate generalized linear mixed effects model associated through correlated and/or shared random effects. The Markov chain Monte Carlo Bayesian approach is used to approximate posterior distribution and draw inference on the parameters. This proposed methodology provides a flexible framework to account for the hierarchical structure of the highly unbalanced data as well as the association between the 2 outcomes. The results indicate improved efficiency of parameter estimates when blood glucose and urine glucose are modeled jointly. Moreover, the simulation studies show that estimates obtained from the joint model are consistently less biased and more efficient than those in the separate models.

A constrained maximum likelihood approach to evaluate the impact of dose metric on cancer risk assessment: application to ß-chloroprene.

ß-Chloroprene (2-chloro-1,3-butadiene, CD) is used in the manufacture of polychloroprene rubber. Chronic inhalation studies have demonstrated that CD is carcinogenic in B6C3F1 mice and Fischer 344 rats. However, epidemiological studies do not provide compelling evidence for an increased risk of mortality from total cancers of the lung. Differences between the responses observed in animals and humans may be related to differences in toxicokinetics, the metabolism and detoxification of potentially active metabolites, as well as species differences in sensitivity. The purpose of this study was to develop and apply a novel method that combines the results from available physiologically based kinetic (PBK) models for chloroprene with a statistical maximum likelihood approach to test commonality of low-dose risk across species. This method allows for the combined evaluation of human and animal cancer study results to evaluate the difference between predicted risks using both external and internal dose metrics. The method applied to mouse and human CD data supports the hypothesis that a PBK-based metric reconciles the differences in mouse and human low-dose risk estimates and further suggests that, after PBK metric exposure adjustment, humans are equally or less sensitive than mice to low levels of CD exposure.

Historical cohort study of US man-made vitreous fiber production workers: I. 1992 fiberglass cohort follow-up: initial findings.

This 1986 to 1992 update and expansion of an earlier historical cohort study examined the 1946 to 1992 mortality experience of 32,110 workers employed for 1 year or more during 1945 to 1978 at any of 10 US fiberglass (FG) manufacturing plants. Included are (1) a new historical exposure reconstruction for respirable glass fibers and several co-exposures (arsenic, asbestos, asphalt, epoxy, formaldehyde, polycyclic aromatic hydrocarbons, phenolics, silica, styrene, and urea); and (2) a nested, matched case-control study of 631 respiratory system cancer (RSC) deaths in male workers during 1970 to 1992 with interview data on tobacco smoking history. Our findings to date from external comparisons based on standardized mortality ratios (SMRs) in the cohort study provide no evidence of excess mortality risk from all causes combined, all cancers combined, and non-malignant respiratory disease. Also, excluding RSC, we observed no evidence of excess mortality risk from any of the other cause-of-death categories considered. For RSC among the total cohort, we observed a 6% excess (P = 0.05) based on 874 deaths. Among long-term workers (5 or more years of employment) we observed a not statistically significant 3% excess based on 496 deaths. Among the total cohort, we observed increases in RSC SMRs with calendar time and time since first employment, but these were less pronounced among long-term workers. RSC SMRs were not related to duration of employment among the total cohort or long-term workers. In an externally controlled analysis of male workers at risk between 1970 and 1992, we observed no association between RSC SMRs and increasing exposure to respirable FG. Our findings to date from internal comparisons based on rate ratios in the case-control study of RSC were limited to analyses of categorized study variables with and without adjustment for smoking. On the basis of these analyses, the duration of exposure and cumulative exposure to respirable FG at the levels encountered at the study plants did not appear to be associated with an increased risk of RSC. RSC risk also did not seem to increase with time since first employment. There is some evidence of elevated RSC risk associated with non-baseline levels of average intensity of exposure to respirable glass, but when adjusted for smoking this was not statistically significant, and there was no apparent trend with increasing exposure. This same pattern of findings was observed for duration of exposure, cumulative exposure, and average intensity of exposure to formaldehyde. None of the other individual co-exposures encountered in the study plants appeared to be associated with an increased risk of RSC. The primary focus of ongoing analyses is to determine the extent to which our present findings are robust to alternative characterizations of exposure.

Historical cohort study of US man-made vitreous fiber production workers: II. Mortality from mesothelioma.

As part of our ongoing mortality surveillance program for the US man-made vitreous fiber (MMVF) industry, we examined mortality from malignant mesothelioma using data from our 1989 follow-up of 3478 rock/slag wool workers and our 1992 follow-up of 32,110 fiberglass workers. A manual search of death certificates for 1011 rock/slag wool workers and 9060 fiberglass workers revealed only 10 death certificates with any mention of the word "mesothelioma." A subsequent review of medical records and pathology specimens for 3 of the 10 workers deemed two deaths as definitely not due to mesothelioma and one as having a 50% chance of being caused by mesothelioma. Two other deaths, for which only medical records were available, were given less than a 50% chance of being due to mesothelioma. Eight of the 10 decedents had potential occupational asbestos exposure inside or outside the MMVF industry. We also estimated the mortality risk from malignant mesothelioma in the cohort using two cause-of-death categorizations that included both malignant and benign coding rubrics. Using the more comprehensive scheme, we observed overall deficits in deaths among the total cohort and fiberglass workers and an overall excess among rock/slag wool workers. The excess in respiratory system cancer is largely a reflection of elevated lung cancer risks that we attributed mainly to confounding by smoking, to exposures outside the MMVF industry to agents such as asbestos, or to one or more of the several co-exposures present in many of the study plants (including asbestos). The second scheme, which focused on pleural mesothelioma in time periods when specific malignant mesothelioma coding rubrics were available, classified only one cohort death as being caused by malignant mesothelioma, compared with 2.19 expected deaths (local county comparison). We conclude that the overall mortality risk from malignant mesothelioma does not seem to be elevated in the US MMVF cohort.

Historical cohort study of US man-made vitreous fiber production workers: III. Analysis of exposure-weighted measures of respirable fibers and formaldehyde in the nested case-control study of respiratory system cancer.

The most recent findings of our nested case-control study of respiratory system cancer (RSC) among male fiberglass workers showed some evidence of elevated RSC risk associated with non-baseline levels of average intensity of exposure (AIE) to respirable fibers (RFib). When adjusted for smoking, this was not statistically significant, and no trend was apparent with increasing levels of exposure. Similar findings for RSC were noted for both cumulative exposure (Cum) and AIE to formaldehyde (FOR). In this reanalysis of our nested case-control study, we explored a possible exposure-response relationship between RSC and exposure to RFib or FOR using exposure weighting as an alternative characterization of exposure. Because of the uncertainties in selecting an appropriate exposure-weighting scheme, a range of plausible time lags and unlagged/lagged time windows was considered. As in the initial analysis of the nested case-control study, RFib and FOR exposures were categorized at the deciles of the RSC case distribution. For none of the exposure weighting schemes considered did we observe an increasing RSC risk with increasing levels of RFib_Cum or RFib_AIE. The exposure-weighted estimated risk ratios (RR) for both RFib_Cum and RFib_AIE were generally lower than those obtained from an unweighted model. For FOR_Cum, RRs were generally lower for the time-lagged and unlagged time window models than for the unweighted models, although some decile-specific RRs were higher for the lagged time window models. The exposure-weighted RRs for FOR_AIE were generally lower than the unweighted RRs for all of the weighting schemes considered. This reanalysis in terms of categorized exposures reveals no exposure-response relationships that were undetected in the original analysis where unweighted exposure measures were used. In the schemes considered, exposure weighting generally reduced the estimated risk of RSC.

Historical cohort study of US man-made vitreous fiber production workers: IV. Quantitative exposure-response analysis of the nested case-control study of respiratory system cancer.

As part of the 1992 update of an historical cohort study of 32,110 workers employed for at least 1 year in any of 10 US fiberglass manufacturing plants, a nested case-control study was done in which data on tobacco smoking were obtained for 631 male case subjects with respiratory system cancer (RSC) and 570 control subjects matched on age and year of birth. In this more extensive analysis of the nested case-control data, we provide a detailed assessment of the most prominent findings from the initial report. We expand the scope of the analysis to consider quantitative measures of exposure to respirable fibers (RFib), formaldehyde (FOR), and silica (Sil) and consider these and other exposures together in the same model. We investigate the functional form of possible exposure-response relationships between RSC risk, RFib, and FOR. In addition, we address the statistical issues of collinearity, effect modification, and potential confounding by coexposures. All analyses are adjusted for smoking. Neither measure of exposure to RFib (average intensity of exposure or cumulative exposure) was statistically significantly associated with RSC risk in any of the hundreds of fractional polynomial models considered. This more extensive analysis has substantiated our initial finding of no apparent exposure-response relationship between RSC risk and either cumulative or average intensity of exposure to RFib at the levels experienced by these workers. This study provides some evidence of increased RSC risk among workers at the higher observed levels of average intensity of exposure to FOR and/or Sil. No positive associations were identified between RSC risk and any of the other exposures considered in this case-control study.

Historical cohort study of US man-made vitreous fiber production workers: V. Tobacco-smoking habits.

As part of our ongoing mortality surveillance program for the US man-made vitreous fiber industry, we surveyed a random sample of study members to estimate tobacco-smoking habits for the total cohort. Separate sampling frames were constructed for four study groups: male and female workers within the fiberglass and rock/slag wool subcohorts. The frames included all persons who had worked a year or more between 1945 and 1986 (with some exceptions), and who were alive as adults (18+ years) on January 1, 1980, the year the age distribution of the cohort most resembled the US comparison population. Subjects were randomly selected from the frames, and a structured telephone interview was administered to the subject or a proxy respondent between January 1995 and December 1997. Using survey data, we estimated the point prevalence of ever and current cigarette smoking on January 1, 1980, and made comparisons with other occupational groups and general populations. Overall response rates (interviews/targeted sample) were greater than 78% for each of the four study groups. From our estimates, we infer that male workers from both the fiberglass and rock/slag wool cohorts and female rock/slag wool workers had higher rates of ever smoking than the corresponding general populations of the United States and most of the states where the study plants were located. These findings suggest that at least part of the elevated externally standardized mortality ratios (US and regional rate-based) for respiratory system cancer noted among male subjects and the male-dominated total cohort in our previous cohort analyses were due to uncontrolled positive confounding by smoking.

Historical cohort study of US man-made vitreous fiber production workers: VI. Respiratory system cancer standardized mortality ratios adjusted for the confounding effect of cigarette smoking.

To date, the US cohort study of man-made vitreous fiber workers has provided no consistent evidence of a relationship between man-made vitreous fiber exposure and mortality from malignant or non-malignant respiratory disease. Nevertheless, there have been small, overall excesses in respiratory system cancer (RSC) among workers from the fiberglass and rock/slag wool production plants included in the study that were unexplained by estimated worker exposures to respirable fiber or other agents present in the plants. The present investigation was designed to provide a quantitative estimate of the extent to which the overall excess in RSC mortality observed at the total cohort level among male fiberglass and rock/slag wool workers is a result of the positive confounding effects of cigarette smoking. Because cigarette-smoking data were neither available nor obtainable at the individual level for all members of the fiberglass and rock/slag wool cohorts, we used the "indirect" method to adjust RSC standardized mortality ratios (SMRs) at the group (cohort and plant) level. Our adjustment suggested that cigarette smoking accounts for all of the 7% and 24% excesses in RSC observed, respectively, for the male fiberglass and rock/slag wool cohorts in the latest mortality updates. The same conclusion was reached regardless of which of several alternative formulations were used to adjust local rate-based RSC SMRs. We found that our smoking adjustments were robust with respect to several alternative characterizations and (with the exception of one fiberglass plant) produced adjusted RSC SMRs that were lower than their unadjusted counterparts. Further, all statistically significantly elevated unadjusted SMRs were reduced to not statistically significant levels. These results reaffirm that RSC SMRs based on US and local rates must take into account the potential confounding effects of cigarette smoking. They also suggest that the use of local county mortality rate-based SMRs may not help to adjust for cigarette smoking to the degree suggested by some investigators.

Historical cohort study of US man-made vitreous fiber production workers: VII. Overview of the exposure assessment.

Data and procedures used to reconstruct the history of exposures at each of the 15 plants (19 distinct sites) are presented. The assessment consisted of five steps: (1) develop a Technical History of operations, stable periods, and time points of changes relevant for exposures, and identify the presence of potentially confounding co-exposures; (2) develop a set of unique department-job names with descriptions and a Job Dictionary for all verbatim names in work histories; (3) collect all company and other exposure data (> 1600 observed), and develop quantitative fiber, formaldehyde, and silica exposure estimates; (4) integrate estimates with the Technical History to make Exposure Extrapolation Tables; and (5) use the the Tables with job data to develop an Exposure Matrix for each plant. Nineteen Exposure Matrices were made, with 82 to 621 lines, covering up to 54 years of operations.

Historical cohort study of US man-made vitreous fiber production workers: VIII. Exposure-specific job analysis.

UNLABELLED: All jobs held by a cohort of US man-made vitreous fiber production workers were analyzed for airborne fiber exposure. This exposure-specific job analysis was part of an exposure assessment for an epidemiologic study of mortality patterns, with particular focus on respiratory cancer, among 35,145 workers employed in 10 fiberglass and five rock or slag wool plants. The exposure assessment was conducted from the start-up date of each plant (1917 to 1946) to 1990. For the job analysis, 15,465 crude department names and 47,693 crude job titles were grouped into 1668 unique department and job pairs (UDJobs), which represented a job title linked to a specific department within each plant. Every UDJob was evaluated according to a set of job elements related to airborne fiber exposure. The distribution of the cohort person-years by UDJob and the job-exposure elements was then evaluated. The results show the main departments and jobs that employed the workers for each plant. The distribution of person-years varies across the job-exposure elements. The same job title was used in different departments within and across plants. When job titles not linked to departments were evaluated, the values of the job-exposure elements varied considerably across all plants and within plant. IN CONCLUSION: (1) exposure misclassification could occur if job title alone were used for the exposure assessment; (2) the job-exposure elements analysis provides an efficient way to identify major job determinants of exposure without relying on the more detailed, resource-intensive task-based approach; and (3) the evaluation of the cohort person-years by UDJobs and job-exposure elements is an effective way to identify which plants, departments, and jobs have sufficient information for making precise risk estimates in the broader epidemiologic study.

Reevaluation of lung cancer risk in the acrylonitrile cohort study of the National Cancer Institute and the National Institute for Occupational Safety and Health.

OBJECTIVES: The present study provides additional analyses of data obtained earlier on lung cancer risk among workers with acrylonitrile exposure. METHODS: The original authors provided the data. For total mortality and the cancer sites of a priori interest (lung, stomach, brain, breast, prostate, and the lymphatic and hematopoietic systems), standardized mortality ratios (SMR) and 95% confidence intervals (95% CI) were computed, the total United States and surrounding counties being used as standard populations. Regional rate-based SMR values were also computed between lung cancer and cumulative acrylonitrile exposure. RESULTS: Except for lung cancer, the external comparisons corroborated the earlier internal comparisons (no increased cancer mortality risk). For lung cancer, the external comparisons revealed death deficits for the unexposed workers (SMR 0.68, 95% CI 0.5-0.9) and all categories of acrylonitrile-exposed workers. The SMR obtained using external rates and the most exposed group (SMR 0.92. 95% CI 0.6-1.4) differed from the corresponding relative risk (RR) of the internal rates (RR 1.5, 95% CI 0.9-2.4). CONCLUSIONS: The analysis of the present study provides little evidence that acrylonitrile exposure increases the mortality risk of cancers of a priori interest, including lung cancer. The lung cancer findings of the external comparison differed from the earlier findings of the internal comparisons. Selection bias (as the healthy worker effect) was probably not responsible. Additional follow-up and analyses, especially of the unexposed workers with low lung cancer rates, may help elucidate the internal and external comparison differences. Results from both comparisons should be presented when the relative risks differ markedly, as both have advantages and disadvantages.

Mortality among the residents of the Three Mile Island accident area: 1979-1992.

The largest U.S. population exposed to low-level radioactivity released by an accident at a nuclear power plant is composed of residents near the Three Mile Island (TMI) Plant on 28 March 1979. This paper (a collaboration of The University of Pittsburgh and the Pennsylvania Department of Health) reports on the mortality experience of the 32,135 members in this cohort for 1979-1992. We analyzed standardized mortality ratios (SMRs) using a local comparison population and performed relative risk regression modeling to assess overall mortality and specific cancer risks by confounding factors and radiation-related exposure variables. Total mortality was significantly elevated for both men and women (SMRs = 109 and 118, respectively). All heart disease accounted for 43.3% of total deaths and demonstrated elevated SMRs for heart disease of 113 and 130 for men and women, respectively; however, when controlling for confounders and natural background radiation, these elevations in heart disease were no longer evident. Overall cancer mortality was similar in this cohort as compared to the local population (male SMR = 100; female SMR = 101). In the relative risk modeling, there was a significant effect for all lymphatic and hematopoietic tissue in males in relation to natural background exposure (p = 0.04). However, no trend was noted. We found a significant linear trend for female breast cancer risk in relation to increasing levels of TMI-related likely [gamma]-exposure (p = 0.02). Although such a relationship has been noted in other investigations, emissions from the TMI incident were significantly lower than in other documented studies. Therefore, it is unlikely that this observed increase is related to radiation exposure on the day of the accident. The mortality surveillance of this cohort does not provide consistent evidence that radioactivity released during the TMI accident has a significant impact on the mortality experience of this cohort to date. However, continued follow-up of these individuals will provide a more comprehensive description of the morbidity and mortality experience of the cohort.

Mortality among chemical plant workers exposed to acrylonitrile and other substances.

OBJECTIVES: To examine the association between exposure to acrylonitrile (AN) and cancer mortality by performing an independent and extended historical cohort study of workers from a chemical plant in Lima, Ohio included in a recent NCI-NIOSH study. METHODS: Subjects were 992 white males who were employed for three or more months between 1960 and 1996. We identified 110 deaths and cause of death for 108. Worker exposures were estimated quantitatively for AN and qualitatively for nitrogen products. Statistical analyses included U.S. and local county-based SMRs and internal relative risk regression of internal cohort rates. RESULTS: No statistically significant excess mortality risks were observed among the total cohort for the cancer sites implicated in previous studies: stomach, lung, breast, prostate, brain, and hematopoietic system. We observed a statistically significant bladder cancer excess based on four deaths (SMR=7.01, 95% CI=1.91-17.96) among workers not exposed to AN. Among 518 AN-exposed workers, we observed a not statistically significant excess of lung cancer based on external (SMR=1.32, 95% CI=.60-2.51) and internal (RR=1.98, 95% CI=.60-6.90) comparisons. Although the trends were not statistically significant, exposure-response analyses of internal cohort rates showed monotonically increasing lung cancer rate ratios with increasing AN exposure, with RRs exceeding 2.0 in the highest exposure categories. CONCLUSIONS: With the possible exception of lung cancer, this study provides little evidence that exposure to AN at levels experienced by Lima plant workers is associated with an increased risk of death from any cause including the implicated cancer sites.

Mortality patterns among workers exposed to acrylamide: 1994 follow up.

OBJECTIVE: To update the mortality experience of a cohort of 8508 workers with potential exposure to acrylamide at three plants in the United States from 1984-94. METHODS: Analyses of standardised mortality ratios (SMR) with national and local rates and relative risk (RR) regression modelling were performed to assess site specific cancer risks by demographic and work history factors, and exposure indicators for acrylamide and muriatic acid. RESULTS: For the 1925-94 study period, excess and deficit overall mortality risks were found for cancer sites of interest: brain and other central nervous system (CNS) (SMR 0.65, 95% confidence interval (95% CI) 0.36 to 1.09), thyroid gland (SMR 2.11, 95% CI 0.44 to 6.17), testis and other male genital organs (SMR 0.28, 95% CI 0.01 to 1.59), and cancer of the respiratory system (SMR 1.10, 95% CI 0.99 to 1.22); however, none was significant or associated with exposure to acrylamide. A previously reported excess mortality risk of cancer of the respiratory system at one plant remained increased among workers with potential exposure to muriatic acid (RR 1.50, 95% CI 0.86 to 2.59), but was only slightly increased among workers exposed or unexposed to acrylamide. In an exploratory exposure-response analysis of rectal, oesophageal, pancreatic, and kidney cancer, we found increased SMRs for some categories of exposure to acrylamide, but little evidence of an exposure-response relation. A significant 2.26-fold risk (95% CI 1.03 to 4.29) was found for pancreatic cancer among workers with cumulative exposure to acrylamide > 0.30 mg/m3.years; however, no consistent exposure-response relations were detected with the exposure measures considered when RR regression models were adjusted for time since first exposure to acrylamide. CONCLUSION: The contribution of 1115 additional deaths and nearly 60,000 person-years over the 11 year follow up period corroborate the original cohort study findings of little evidence for a causal relation between exposure to acrylamide and mortality from any cancer sites, including those of initial interest. This is the most definitive study of the human carcinogenic potential of exposure to acrylamide conducted to date.

OCMAP-PLUS: a program for the comprehensive analysis of occupational cohort data.

The Occupational Cohort Mortality Analysis Program (OCMAP) has been redesigned for optimal microcomputer use and extended to include new computing algorithms. The new program, OCMAP-PLUS, offers a comprehensive, flexible, and efficient analysis of incidence or mortality rates and standardized measures in relation to multiple and diverse work history and exposure measures. New features include executable code, minimization of memory requirements, disk file storage of person-day arrays, stratified analyses by geographic area, employment status and up to eight exposure variables, a data imputation algorithm for study members with unknown race, and enhanced algorithms for constructing several time-dependent exposure measures. New modules create grouped data files for Poisson and logistic regression and risk set files for use in relative risk regression analysis. The Mortality and Population Data System (MPDS) provides external comparison rates and proportional mortalities. Analysis from two recent cohort mortality studies illustrate several new features.