CONTEXT:We had available records on over 300 workers evaluated with the beryllium bronchoalveolar lavage lymphocyte proliferation test (BeBALLPT) at three expert chronic beryllium disease (CBD) diagnostic centers.OBJECTIVE:The objective was to describe the contribution of the BeBALLPT to classification of workers with respect to beryllium sensitization (BeS) and beryllium-induced lung inflammation.METHODS:Company records were used to identify beryllium workers who had undergone diagnostic bronchoscopy with BeBALLPT. Clinical, work and smoking information was abstracted from electronic and paper databases. We analyzed factors influencing BeBALLPT outcome, and its relation to blood-determined BeS and granulomatous inflammation.RESULTS:Positive BeBALLPTs contributed evidence of BeS in subjects without prior positive beryllium blood lymphocyte proliferation tests (BeBLPTs) and of pulmonary inflammation in persons without granulomata evident on lung biopsy. Positive BeBALLPTs were associated with positive BeBLPTs and more strongly with granulomata. The rate of both positive BeBALLPT and granulomata increased with time worked through 4 years and were lower in smoking subjects. The false negative rate of the BeBALLPT was 20%.CONCLUSION:A positive BeBALLPT is closely linked to the presence of granulomata on lung biopsy and can be considered as an indicator of lung inflammation in addition to BeS. The ability to use BeBALLPT as a substitute for the more risky lung biopsy is limited by the BeBALLPT false negative rate and lack of information on the false positive rate. It is not recommended that a positive BeBALLPT be considered sufficient evidence for both lung inflammation and BeS.

Purpose: This study explores how highly correlated time variables (occupational cohort time scales) contribute to confounding and ambiguity of interpretation. Methods: Occupational cohort time scales were identified and organized through simple equations of three time scales (relational triads) and the connections between these triads (time scale web). The behavior of the time scales was examined when constraints were imposed on variable ranges and interrelationships. Results: Constraints on a time scale in a triad create high correlations between the other two time scales. These correlations combine with the connections between relational triads to produce association paths. High correlation between time scales leads to ambiguity of interpretation. Conclusions: Understanding the properties of occupational cohort time scales, their relational triads, and the time scale web is helpful in understanding the origins of otherwise obscure confounding bias and ambiguity of interpretation.

The mapping and sequencing of the human genome has resulted in an explosion of information, which has led in several instances to improved capability for detecting diseases or increased susceptibility to disease, treatment of diseases, and the identification of individuals at increased risk for adverse reactions to pharmaceuticals and environmental or workplace chemicals. Some genetic tests are already commercially available, for example, tests which screen for variations in genes that metabolize certain pharmaceuticals and others that identify individuals at increased risks of specific types of cancer. Genetic screening* offers the prospect of a new era for prevention and treatment and a growing array of effective new interventions.1,2 Genetic screening has also been accompanied by some misunderstanding, mistrust, and fear that it could be used inappropriately. Indeed, certain previous uses of genetic screening have been inconsistent with good ethical standards and sound scientific practice and have led some to advocate that genetic screening be treated as a separate category with special safeguards. Passage of the Genetic Information Nondiscrimination Act (GINA) of 2008 formalized many of these concerns, thereby establishing strict restrictions and guidelines for the use of such testing in occupational settings.3,4,5 Historically, the American College of Occupational and Environmental Medicine (ACOEM) has taken the position that genetic screening was not conceptually different from other types of medical testing or screening and that adherence to existing ethical standards, good scientific practices, and laws regulating medical confidentiality protected the rights of the individual appropriately, while allowing the new information to be used to further safeguard the health of individuals in the workplace and elsewhere. From a scientific perspective, ACOEM still regards genetic screening as conceptually similar to other types of medical screening. Nevertheless, from a legal perspective, ACOEM recognizes that many of the potential uses of genetic screening for workplace safety programs are now legally prohibited in numerous jurisdictions. Genetic screening may be offered only on a voluntary basis, and test results may not be used to determine work practices or conditions of employment. It is imperative that practitioners of occupational and environmental medicine be well grounded in the relevant ethical, legal, social, and scientific considerations and be prepared to offer sound advice to employees, employers, insurance companies, and regulatory agencies. For example, GINA permits voluntary testing (“genetic monitoring”) that provides information about the biological effects of toxic substances in the workplace. Similarly, GINA permits voluntary testing for genetic information in otherwise appropriate research programs, so long as the resulting genetic information is not used in a manner that impacts the tested individual's conditions of employment. Although the application of genetic screening in the workplace has been limited to date (and is now severely restricted), the ethical considerations of such testing in the workplace (and elsewhere) have been extensively examined. ACOEM endorses the following guiding principles: Genetic screening must be conducted with consideration of the law, medical ethical standards, and good scientific practices. Until extensively validated, genetic screening is a form of human investigation and subject to the appropriate ethical and scientific controls. Due consideration should be given to the quality and reliability of the screening tests and the predictive value of the results. Caution should be exercised in the use and interpretation of screening tests. If performed, genetic screening should always be accompanied by an opportunity to discuss the meaning of the results with an appropriately trained health professional. When consulted by employees about decisions to have genetic testing performed (or seeking explanations of results performed apart from the workplace), occupational physicians should candidly share, in line with these principles, the limitations of testing and the uncertainties regarding interpretation of results. ACOEM also notes that GINA may complicate ethical considerations for occupational physicians and others committed to the health and safety of workers and workplaces. It seems reasonable to expect that, in the future, some forms of genetic testing will provide a basis for more effective methods to ensure the health of individual workers, but that preventive actions taken on the basis of such testing might violate GINA. In such situations, both acting on the basis of genetic information to better protect the worker and not acting on that information, and thereby failing to protect the worker, would violate standards of ethical conduct. ACOEM hopes that such potential conflicts can be preemptively resolved without recourse to litigation and the federal court system. ACOEM is the preeminent medical organization that champions the health and safety of workers, workplaces, and environments. The College represents nearly 4500 physicians and other allied health professionals who are specialists in the field of occupational and environmental medicine. ACKNOWLEDGMENTS This statement was developed by ACOEM's Task Force on Genetic Screening in the Workplace under the auspices of the Council of Scientific Advisors. The statement was peer-reviewed by members of the Council and was approved by the ACOEM Board of Directors on November 8, 2014. This document updates ACOEM's 2010 statement on Genetic Screening in the Workplace.

Objective:To describe how smoking correction factors based on comparing worker smoking prevalence with population smoking prevalence are biased if applied to an occupational incidence cohort. Methods:Relative rates of smoking for shorter-tenure workers derived from occupational cohort lung cancer studies were applied to incidence and prevalence population tenure distributions to calculate relative smoking estimates. Results:High smoking rates in short-tenure workers have little effect on prevalent worker rates (relative smoking estimates, 1.04 and 1.02) and much larger effect in occupational incidence populations (relative smoking estimates, 1.58 and 1.21), which have a much higher proportion of short tenure-workers. Conclusions:Smoking correction estimates derived from surveys of smoking habits in prevalent workers may introduce bias when applied to incidence workers because of very different proportions of short-tenure workers (length-time biased sampling).

Objectives Inhalation beryllium exposures are associated with sensitisation, however dermal exposures are also important. In a previous study, we identified strong correlations between dermal-air, dermal- surface, and air- surface measurements. The aim of this study was to investigate workplace factors associated with exposures using mixed-effects models and structural equation modelling (SEM). Method Beryllium was measured in personal air, on gloves, and on surfaces at three manufacturing facilities. Predictor variables included substance and activity emission potential (REACH classification), dilution, segregation, PPE, personal behaviour, and work shift. Results The mixed model described 57 and 59% of total variance for air and dermal, respectively. The total variance explained by the SEM model for air and dermal was 0.51 and 0.48% respectively. In both models activity and substance emission potential, surface contamination, dilution, and personal behaviour were significant predictors of air concentrations (p ≤ 0.05); and surface contamination and air concentrations were significant predictors of dermal loading on cotton gloves (p ≤ 0.05). However, work shift and personal behaviour were predictive of dermal loading in the SEM (p ≤ 0.03), but not in the mixed model. In addition, the SEM reported a parameter estimate for air concentration as a predictor of dermal loading that was an order of magnitude higher than in the mixed model. Conclusions Although SEM requires relatively large sample sizes, it is useful for modelling multiple, correlated dependent variables. In addition, full-information maximum likelihood (FIML) methods can be used in SEM to include missing predictor variable data. Although we found both models to be useful, SEM has the potential to illustrate indirect pathways of outcome variables.

Inhalation of beryllium is associated with the development of sensitization; however, dermal exposure may also be important. The primary aim of this study was to elucidate relationships among exposure pathways in four different manufacturing and finishing facilities. Secondary aims were to identify jobs with increased levels of beryllium in air, on skin, and on surfaces; identify potential discrepancies in exposure pathways, and determine if these are related to jobs with previously identified risk.Beryllium was measured in air, on cotton gloves, and on work surfaces. Summary statistics were calculated and correlations among all three measurement types were examined at the facility and job level. Exposure ranking strategies were used to identify jobs with higher exposures.The highest air, glove, and surface measurements were observed in beryllium metal production and beryllium oxide ceramics manufacturing jobs that involved hot processes and handling powders. Two finishing and distribution facilities that handle solid alloy products had lower exposures than the primary production facilities, and there were differences observed among jobs. For all facilities combined, strong correlations were found between air-surface (rp ≥ 0.77), glove-surface (rp ≥ 0.76), and air-glove measurements (rp ≥ 0.69). In jobs where higher risk of beryllium sensitization or disease has been reported, exposure levels for all three measurement types were higher than in jobs with lower risk, though they were not the highest. Some jobs with low air concentrations had higher levels of beryllium on glove and surface wipe samples, suggesting a need to further evaluate the causes of the discrepant levels.Although such correlations provide insight on where beryllium is located throughout the workplace, they cannot identify the direction of the pathways between air, surface, or skin. Ranking strategies helped to identify jobs with the highest combined air, glove, and/or surface exposures. All previously identified high-risk jobs had high air concentrations, dermal mass loading, or both, and none had low dermal and air. We have found that both pathways are relevant.[Supplementary materials are available for this article. Go to the publisher's online edition of Journal of Occupational and Environmental Hygiene for the following free supplemental resource: a file describing the forms of beryllium materials encountered during production and characteristics of the aerosols by process areas.]

BACKGROUND:In 2000, a manufacturer of beryllium materials and products introduced a comprehensive program to prevent beryllium sensitization and chronic beryllium disease (CBD). We assessed the program's efficacy in preventing sensitization 9 years after implementation.METHODS:Current and former workers hired since program implementation completed questionnaires and provided blood samples for the beryllium lymphocyte proliferation test (BeLPT). Using these data, as well as company medical surveillance data, we estimated beryllium sensitization prevalence.RESULTS:Cross-sectional prevalence of sensitization was 0.7% (2/298). Combining survey results with surveillance results, a total of seven were identified as sensitized (2.3%). Early Program workers were more likely to be sensitized than Late Program workers; one of the latter was newly identified. All sensitization was identified while participants were employed. One worker was diagnosed with CBD during employment.CONCLUSIONS:The combination of increased respiratory and dermal protection, enclosure and improved ventilation of high-risk processes, dust migration control, improved housekeeping, and worker and management education showed utility in reducing sensitization in the program's first 9 years. The low rate (0.6%, 1/175) among Late Program workers suggests that continuing refinements have provided additional protection against sensitization compared to the program's early years.

Objective: Common variation is a statistical process-control term for variability associated with usual operating conditions. Special variation occurs when usual operating conditions are disrupted. The objective was to explore the implications for preventive occupational medicine practice of common and special variation in air-level exposure. Methods: Illustrations are derived from US and UK beryllium facility databases. Results: Special variation may be missed in finite sampling sets, giving a very inaccurate indication of the highest air levels experienced on the job. Depending on the toxicologic model, failure to assess special variation influences the meaningfulness of aspects of occupational prevention, from medical surveillance through risk management. Conclusions: Jobs and tasks should be characterized for special variation in addition to traditional air sampling. Both special variation and common variation should be considered in occupational medicine preventive practice.

OBJECTIVE:Common variation is a statistical process-control term for variability associated with usual operating conditions. Special variation occurs when usual operating conditions are disrupted. The objective was to explore the implications for preventive occupational medicine practice of common and special variation in air-level exposure.METHODS:Illustrations are derived from US and UK beryllium facility databases.RESULTS:Special variation may be missed in finite sampling sets, giving a very inaccurate indication of the highest air levels experienced on the job. Depending on the toxicologic model, failure to assess special variation influences the meaningfulness of aspects of occupational prevention, from medical surveillance through risk management.CONCLUSIONS:Jobs and tasks should be characterized for special variation in addition to traditional air sampling. Both special variation and common variation should be considered in occupational medicine preventive practice.

OBJECTIVES Exposure-response relations for beryllium sensitization (BeS) and chronic beryllium disease (CBD) using aerosol mass concentration have been inconsistent, although process-related risks found in most studies suggest that exposure-dependent risks exist. We examined exposure-response relations using personal exposure estimates in a beryllium worker cohort with limited work tenure to minimize exposure misclassification. METHODS The population comprised workers employed in 1999 with six years or less tenure. Each completed a work history questionnaire and was evaluated for immunological sensitization and CBD. A job-exposure matrix was combined with work histories to create individual estimates of average, cumulative, and highest-job-worked exposure for total, respirable, and submicron beryllium mass concentrations. We obtained odds ratios from logistic regression models for exposure-response relations, and evaluated process-related risks. RESULTS Participation was 90.7% (264/291 eligible). Sensitization prevalence was 9.8% (26/264), with 6 sensitized also diagnosed with CBD (2.3%, 6/264). A general pattern of increasing sensitization prevalence was observed as exposure quartile increased. Both total and respirable beryllium mass concentration estimates were positively associated with sensitization (average and highest job), and CBD (cumulative). Increased sensitization prevalence was identified in metal/oxide production, alloy melting and casting, and maintenance, and for CBD in melting and casting. Lower sensitization prevalence was observed in plant-area administrative work. CONCLUSIONS Sensitization was associated with average and highest job exposures, and CBD was associated with cumulative exposure. Both total and respirable mass concentrations were relevant predictors of risk. New process-related risks were identified in melting and casting and maintenance.

OBJECTIVE:Beryllium mine and ore extraction mill workers have low rates of beryllium sensitization and chronic beryllium disease relative to the level of beryllium exposure. The objective was to relate these rates to the solubility and composition of the mine and mill materials.METHOD:Medical surveillance and exposure data were summarized. Dissolution of BeO, ore materials and beryllium hydroxide, Be(OH)(2) was measured in synthetic lung fluid.RESULT:The ore materials were more soluble than BeO at pH 7.2 and similar at pH 4.5. Be(OH)(2) was more soluble than BeO at both pH. Aluminum dissolved along with beryllium from ore materials.CONCLUSION:Higher solubility of beryllium ore materials and Be(OH)(2) at pH 7.2 might shorten particle longevity in the lung. The aluminum content of the ore materials might inhibit the cellular immune response to beryllium.

Various Be-containing micro-particle suspensions were equilibrated with simulated lung fluid (SLF) to examine their dissolution behavior as well as the potential generation of nanoparticles. The motivation for this study was to explore the relationship between dissolution/particle generation behaviors of Be-containing materials relevant to Be-ore processing, and their epidemiologically indicated inhalation toxicities. Limited data suggest that BeO is associated with higher rates of beryllium sensitization (BS) and chronic beryllium disease (CBD) relative to the other five relevant materials studied: bertrandite-containing ore, beryl-containing ore, frit (a processing intermediate), Be(OH)₂ (a processing intermediate), and silica (control). These materials were equilibrated with SLF at two pH values (4.5 and 7.2) to reflect inter- and intra-cellular environments in lung tissue. Concentrations of Be, Al, and Si in SLF increased linearly during the first 20 days of equilibration, and then rose slowly, or in some cases reached a maximum, and subsequently decreased. Relative to the other materials, BeO produced relatively low Be concentration in solution at pH 7.2; and relatively high Be concentration in solution at pH 4.5 during the first 20 days of equilibration. For both pH values, however, the Be concentration in SLF normalized to Be content of the material was lowest for BeO, demonstrating that BeO was distinct among the four other Be-containing materials in terms of its persistence as a source of Be to the SLF solution. Following 149 days of equilibration, the SLF solutions were fractionated using flow-field flow fractionation (FlFFF) with detection via ICP-MS. For all materials, nanoparticles (which were formed during equilibration) were dominantly distributed in the 10-100 nm size range. Notably, BeO produced the least nanoparticle-associated Be mass (other than silica) at both pH values. Furthermore, BeO produced the highest Be concentrations in the size range corresponding to < 3 kDa (determined via centrifugal ultrafiltration), indicating that in addition to persistence, the BeO produced the highest concentrations of truly dissolved (potentially ionic) Be relative to the other materials. Mass balance analysis showed reasonable sample recoveries during FFF fractionation (50-100%), whereas recoveries during ICP-MS (relative to acidified standards) were much lower (5-10%), likely due to inefficiencies in nebulizing and ionizing the nanoparticles.

BACKGROUND:Up to 12% of beryllium-exposed American workers would test positive on beryllium lymphocyte proliferation test (BeLPT) screening, but the implications of sensitization remain uncertain.METHODS:Seventy two current and former employees of a beryllium manufacturer, including 22 with pathologic changes of chronic beryllium disease (CBD), and 50 without, with a confirmed positive test were followed-up for 7.4 +/-3.1 years.RESULTS:Beyond predicted effects of aging, flow rates and lung volumes changed little from baseline, while DLCO dropped 17.4% of predicted on average. Despite this group decline, only 8 subjects (11.1%) demonstrated physiologic or radiologic abnormalities typical of CBD. Other than baseline status, no clinical or laboratory feature distinguished those who clinically manifested CBD at follow-up from those who did not.CONCLUSIONS:The clinical outlook remains favorable for beryllium-sensitized individuals over the first 5-12 years. However, declines in DLCO may presage further and more serious clinical manifestations in the future. These conclusions are tempered by the possibility of selection bias and other study limitations.

Objective: We evaluated a workplace preventive program's effectiveness, which emphasized skin and respiratory protection, workplace cleanliness, and beryllium migration control in lowering beryllium sensitization. Methods: We compared sensitization prevalence and incidence rates for workers hired before and after the program using available cross sectional and longitudinal surveillance data. Results: Sensitization prevalence was 8.9% for the Pre-Program Group and 2.1% for the Program Group. The sensitization incidence rate was 3.7/1000 person-months for the Pre-Program Group and 1.7/1000 person-months for the Program Group. After making adjustments for potential selection and information bias, sensitization prevalence for the Pre-Program Group was 3.8 times higher (95% CI = 1.5 to 9.3) than the Program Group. The sensitization incidence rate ratio comparing the Pre-Program Group to the Program Group was 1.6 (95% CI = 0.8 to 3.6). Conclusions: This preventive program reduced the prevalence of but did not eliminate beryllium sensitization.

Objectives. In 2000, 7% of workers at a copper-beryllium facility were beryllium sensitized. Risk was associated with work near a wire annealing/pickling process. The facility then implemented a preventive program including particle migration control, respiratory and dermal protection, and process enclosure. We assessed the program's efficacy in preventing beryllium sensitization.Methods. In 2000, the facility began testing new hires (program workers) with beryllium lymphocyte proliferation tests (BeLPTs) at hire and at intervals during employment. We compared sensitization incidence rates (IRs) and prevalence rates for workers hired before the program (legacy workers) with rates for program workers, including program worker subgroups. We also examined trends in BeLPTs from a single laboratory.Results. In all, five of 43 legacy workers (IR=3.8/1,000 person-months) and three of 82 program workers (IR=1.9/1,000 person-months) were beryllium sensitized, for an incidence rate ratio (IRR) of 2.0 (95% confidence interval [CI] 0.5, 10.1). Two of 37 pre-enclosure program workers (IR=2.4/1,000 person-months) and one of 45 post-enclosure program workers (IR=1.4/1,000 person-months) were beryllium sensitized, for IRRs of 1.6 (95% CI 0.3, 11.9) and 2.8 (95% CI 0.4, 66.2), respectively, compared with legacy workers. Test for trend in prevalence rates was significant. Among 2,159 first-draw BeLPTs, during 95 months, we identified seven months when high numbers of redraws were required, with one possible misclassification in this facility.Conclusions. Fewer workers became sensitized after implementation of the preventive program. However, low statistical power due to the facility's small workforce prevents a definitive conclusion about the program's efficacy. These findings have implications for other copper-beryllium facilities, where program components may merit application.

In 2001, NIOSH published their results of a case-control study1 of beryllium exposure and lung cancer nested in a cohort of beryllium workers. Two additional analyses2,3 and several commentaries4–7 on methods have been reported. This article summarizes the findings and methodologic challenges of this nested case-control study. Sanderson et al1 used a case-control design nested in with this occupational cohort, density selection of controls by attained age. Exposure was characterized as employment duration, cumulative exposure, average exposure, and maximum exposure; it was truncated at the age of death of the lung cancer case. Cases were compared with controls without adjustment for covariates. Unlagged exposure was not importantly increased in cases. With 10- and 20-year lagging of exposures, cases had much higher exposures than controls. Levy et al2 recognized that case-control differences in age at hire would introduce confounding of lagged exposure. Using an incorrect theory as to the genesis of the case-control differences, this analysis subselected controls on attained age by limiting case-control differences in attained age to 3 years. The 3.9-year difference in age at hire between cases and controls narrowed to 1.6. All the case-control differences in 10-year lagged exposure attenuated. Schubauer-Berrigan et al3 addressed date-of-birth and age-at-hire confounding by adding these variables to the analysis. The result was to attenuate the odds ratios for cumulative exposure at a lag of 10 and 20 years and for average exposure and maximum exposure at a lag of 20 years, but there was minimal decrease in the odds ratios for average exposure and maximum exposure at a lag of 10 years. We address several questions raised by these analyses and the commentaries. Are date of birth and age at hire confounders of the relationship between lung cancer and beryllium exposure in the 3 nested case-control analyses?1–3 Date of birth and age at hire do not confound the unlagged analyses, but they do confound when exposure is lagged. Date of birth and age at hire are both associated with lung cancer status and lagged exposure, but not with unlagged exposure. The original NIOSH analysis does not address confounding by date of birth and age at hire, whereas the 2 subsequent analyses do. Why are date of birth and age at hire associated with lung cancer case status? Date of birth is associated with lung cancer due to the rapid increase in lung cancer rates in persons born from 1870 through 1910 and maintained in those born from 1910 to 1930.3 Age at hire becomes secondarily associated with lung cancer because of the strong negative association (correlation −0.87) of age at hire with date of birth. The differences between cases and controls in date of birth and age at hire are not the result in mismatching of cases and controls on attained age,3,5–7 as suggested by Levy et al2 and Deubner et al.4 Why are date of birth and age at hire not associated with unlagged exposure? This is a data-determined fact. In controls the correlations of age at hire with unlagged cumulative and average exposure are −0.005 and −0.05, respectively. Thus, age at hire or date of birth are not confounders, even though they are associated with lung cancer status, because they are not associated with unlagged exposure. Why are date of birth and age at hire associated with lagged exposure? In controls the correlations of age at hire with cumulative and average exposure at lag of 20 years are each −0.60. In lagged analyses, only exposure occurring before the lag-age (case age at death minus latency) is counted. For all exposure metrics, lagged exposure becomes 0 when lag-age is less than age at hire (“lagged out”). The older the age at hire, the more likely exposure is to lag to 0, and so age at hire becomes negatively associated with lagged exposure. The association of date of birth with lagged exposure is secondary through the association of date of birth with age at hire. The association of date of birth and age at hire with case status and lagged exposure makes both date of birth and age at hire confounders of lung cancer and lagged exposure. Is there another form of confounding in these studies? Use of lagged out subjects as a referent group could be a source of confounding if lagged-out subjects were very different from other subjects. In these studies subjects who lag out at 10 years have a mean date of birth of 1891, compared with 1911 for other subjects, and a mean age at hire of 54, compared with 35 years. These differences are due to lung cancer causing death at older ages; thus, 10 year lag-age (age at death –10 years) is also older, and age at hire must be greater than lag-age. The earlier date of birth is due to the negative association of date of birth with age at hire. In analyses with quartile exposure categories the lowest quartile receives subjects who “lag out.” In the analyses by Sanderson et al1 and Schubauer-Berrigan et al,3 lagged-out subjects comprise about one-third of the lowest exposure quartile at lag 10, and almost the entire lowest exposure quartile at lag 20. Lagged-out subjects also differ in that they cannot manifest exposure-caused lung cancer through the age at which they are used in the analysis. This age is less than their age at hire plus the lung cancer latency period assumed for lagging. It is desirable to retain this group with 0 lagged exposure and 0 exposure-caused disease in the analysis as a (0,0) referent, because of concern about bias toward the null if it is excluded. However, this view might be modified if fatal confounding (not correctible in the analysis) is introduced. How is confounding by date of birth and age at hire addressed in the analyses By Levy et al2 and Schubauer-Berrigan et al3? Levy et al2 matched on attained age to reduce differences in date of birth and age at hire. There are valid objections to stratifying on attained age.5,6 However, it is not clear what portion of the attenuation of odds ratios in the report by Levy et al2 is related to narrower case-control differences of date of birth and age at hire and what portion is due to a bias toward the null. Schubauer-Berrigan et al3 added categories of date of birth or age at hire to the analysis to control confounding. With case-control differences of date of birth or age at hire of about 4 years, and substantial overlap in the distributions, it is reasonable to believe that confounding would be substantially controlled if the only source of confounding were these case-control differences. However, when the data are arranged in exposure categories, confounding from the use of the lagged-out subjects as referent is apparent. The nearly 20-year difference in date of birth and age at hire at lag 10, together with lack of overlap when age at hire is less than 35 years, raises the possibility that the odds ratio estimates reported by Schubauer-Berrigan et al3 might be affected by residual confounding. In conclusion, substantial progress has been made in using the NIOSH data to clarify the relationship between beryllium exposure and lung cancer. The second article2 is an improvement on the first article1 as it addressed confounding, albeit through incomplete control of age at hire using matching on a surrogate, (attained age) which may have introduced bias toward the null. The third article3 is an improvement on the first 21,2 because it correctly identified one source of confounding and addressed it directly. At this point it seems clear that in this beryllium worker cohort neither employment duration nor cumulative exposure to beryllium is positively associated with lung cancer. What is needed is an assessment of whether there is residual confounding in the third article3 and, if so, whether it contributes to the lag-10 findings for beryllium average and maximum exposure. If there is residual confounding related to the use of lagged-out subjects in the analysis as the referent, further thought is needed regarding how the lagged-out subjects can best be used in the analysis. We are grateful to the authors of the analyses and commentaries. Mismatching in incidence-density sampling on attained age does not necessarily cause case-control differences in date of birth and age at hire when these 3 are associated. However, neither does it prevent case-control differences in date of birth nor age at hire that arise by other mechanisms. The data in these analyses display a complex web of associations. Data selection, modification, and arrangement change these associations, so that thorough assessment for confounding is necessary. Our main reservation regarding the simulations presented by the 2 articles published in this issue of Epidemiology5,6 is that these simulations do not fully model important associations in the NIOSH data. Ultimately, the decision of whether to address confounding, the choice of the method for controlling confounding, and the adequacy of the control have to be evaluated in the context of associations encountered in a particular analysis. We are pursuing construction of simulations that closely model important associations in the NIOSH data to more completely understand how lagging exposure creates confounding, to identify when residual confounding may be likely, and to test methods to control confounding.

OBJECTIVE:The objective was to assess highly confounded patterns in a standardized mortality ratio (SMR) analysis of lung cancer in beryllium worker cohorts.METHODS:We used Cox proportional hazards single- and multi-variate models to assess confounding and the SMR patterns.RESULTS:We confirmed the lack of association of lung cancer with time worked. We could not confirm the original study's finding of lung cancer highly associated with earlier plants and or with workers hired in the 1940s compared to the 1950s. The pattern of higher rates of lung cancer with increasing latency was attenuated when covariates were added to the model. We could not exclude that the lower SMR and hazard ratios for workers hired in the 1960s might be related to assumed lower beryllium exposures.CONCLUSION:The patterns observed provide little support for an association of lung cancer with beryllium work factors. This result is likely due to the absence in the original study of a significant overall excess of lung cancer after smoking adjustment.