All countries computed external mortality comparisons based on national and/or local mortality rates. Except for Austria, which analyzed mortality in relation to only Co exposure, and the UK, which did not perform an exposure-response analysis, each country evaluated exposure-response relationships for lung cancer mortality in relation to lagged and/or unlagged Co, Ni, and W exposures. Each country also included internal comparisons (worker-to-worker) of lung cancer mortality using similar relative risk regression or Cox regression models. Each country also attempted to adjust risk estimates for lung cancer for potential confounding by smoking using either direct or indirect methods. Case-control studies that attempted to obtain information on lifetime smoking history via direct interview with cases or controls were conducted in the US, Sweden, and the UK with limited success due to low study member participation rates. Legal restrictions in Germany and Austria precluded any direct contact with cases or controls.
Exposure Estimation for Pooled Analysis
Details of the exposure assessment conducted by UIC are reported elsewhere.8 In brief, quantitative job-exposure matrices consisting of job classes and exposure estimates were constructed for W, Co, and Ni for the period 1952–2014 (jobs held prior to 1952 were assigned 1952 exposures). Job classes were created from knowledge of production processes, information obtained from plant personnel, and a review of anonymized work history record job combinations (derived from job and department titles, job and department codes, and other relevant identifying information) before generation of the exposure estimates. Because some operations and years were not fully represented by the US industrial hygiene (IH) data alone and the operations across plants, countries, and companies were similar, data from the nine EU plants and the initially eligible 12 US plants were pooled. The UIC “global” exposure estimates were used in the pooled US/EU cohort analysis reported here.
Based upon production information from plant personnel, separate Ni job-exposure matrices were constructed for Austria and Sweden. Nickel was used in the Austrian plant through 2005. Nickel was used in one Swedish plant from 1965 and in two Swedish plants from 1970. For these three plants, Ni exposures were set to zero for those years it was absent; the “global” estimates were used for all years Ni was present.
The country-specific exposure-related analyses reported by Austria,11 Germany,10 and Sweden12,13 were based on independent exposure assessments conducted in those countries. The UK analyzed exposure measurements but did not perform an exposure assessment or exposure-response analyses due to limited available IH measurements.9
Construction of Worker Exposure Indices
We considered both quantitative and qualitative indices of exposure in our pooled analysis of mortality risks from lung cancer. These analyses included all but 385 workers with unknown work history information.
For each study plant, we linked job classes and corresponding time-specific quantitative exposure estimates for W, Co, and Ni to the detailed study member work histories. This enabled the construction of working lifetime exposure profiles for each study member. The individual exposure profiles were used to compute three summary measures or metrics of exposure study for each study member: Duration of Exposure (Dur) = the sum of the days spent in jobs (in years); Cumulative Exposure (Cum) = number of days in each job and the estimated average daily exposure (in unit-years); and Average Intensity of Exposure (AIE) = the ratio of Cum to Dur (in units). The notation used to describe these summary measures is Agent_MeasureX, where X denotes the number of categories. For example, Tungsten_AIE4 refers to the average intensity of W exposure based on four categories.
Alternative characterizations of the Agent_Metrics were also computed using a 15-year lag period as described in detail by Youk et al.14 Lagging attempts to characterize only the most etiologically relevant exposures and is particularly relevant for examining chronic diseases, such as cancer, associated with long latency periods. A 15-year lag is commonly used for solid tumors such as lung cancer, the focus of this investigation (eg, Hauptmann et al15).
Because the level of W and/or Co and/or Ni “powder” exposure is related to whether a person worked in pre- or post-sintering jobs, we evaluated lung cancer risks among workers in relation to one of four categories: pre-sintering only jobs, post-sintering only jobs, mixed pre-/post-sintering jobs, and no pre-/post-sintering jobs. Although we were interested in examining lung cancer risk among workers who had exposure to W with and without concurrent exposure to Co or Ni, this analysis was possible in only one US plant site as discussed in the US cohort paper7 and thus was not pursued in the pooled analysis.
Pooled Analysis of General Mortality Patterns, External Comparisons
We examined the total and cause-specific mortality experiences of the pooled cohort from January 1, 1952, to December 31, 2014. Cohort analyses were performed using the Occupational Cohort Mortality Analysis Program (OCMAP-PLUS).16 Using a modified life-table procedure, the person-years at risk contributed by each study member were jointly classified by country, plant, race (US only), sex, age group, calendar time, year of hire, duration of employment (DOE), and time since first employment (TSFE). Analyses were classified generally by worker type, as the mortality experience of long-term workers may be less confounded by the unhealthy behavior and lifestyle patterns often observed among short-term workers.17,18 We defined short-term workers as those whose total DOE was less than 1 year (worker type 1) or less than 5 years (worker type 5), and long-term workers as having a total DOE of 1 or more or 5 or more years, respectively. In the analysis by worker type, the first 1 or 5 years of employment for long-term workers were excluded from the corresponding short-term worker person-year counts.
We computed expected numbers of deaths by multiplying average annual country, race (US only)-, sex-, age-, and time-specific standard population death rates by the person-years at risk in the corresponding country, race (US only)-sex-age-time intervals. In the US cohort, person-years and observed deaths for study members of unknown race were assigned to the white category. Expected numbers of deaths were computed using standard populations, which included both national rates and regional rates based on the local areas from which the respective plant workforces were largely drawn. Appendix B shows the regional groups used for each of the five countries and 17 plant sites studied. For countries whose regional rates were unavailable for the study period, we applied the earliest regional rates available to the previous time periods in that region. For the US, we obtained mortality rates from the Mortality and Population Data System.16 Corresponding mortality rates for the four EU countries were provided by the respective investigators in the format of the US rates. Appendix A describes some of the modifications needed to conform the EU rates to the US rate format. Appendix C shows the cause of death categories and corresponding International Classification of Disease (ICD) codes used in our analyses.
To account for geographic variability, the analysis of general mortality patterns, in particular for lung cancer mortality, focused primarily on the regional comparisons. Regional comparisons usually provide the most valid external mortality comparisons by helping to adjust for the social, cultural, and economic factors related to disease, and at some level geographic differences in smoking habits. Standardized mortality ratios (SMRs) and their 95% confidence intervals (CI) were computed for the total pooled cohort and selected subgroups. Statistically significant deviations of the SMR below and above 1.00 were identified using Poisson distribution probabilities.19
Pooled Analysis of Lung Cancer Mortality, Internal Comparisons
We used relative risk (RR) regression modeling to investigate the dependence of the internal cohort mortality rates (modeled as time to death) for lung cancer on combinations of the exposure variables considered, with adjustment for potential confounding factors. For each cause of death, risk sets were explicitly constructed from the cohort data using the RISKSET program module in OCMAP-PLUS.16 Risk sets were matched on exact age at death (event time), sex and on year of birth (±1-year caliper) to control for cohort effects. Time-dependent exposures were evaluated for each study member at each event time they were at risk.
Multiplicative Cox RR regression models of the form λ(t) = λ0(t) expx[(t)β] were fitted to the internal cohort rates20,21 using the conditional logistic regression programs in Stata 1422 and SAS software 9.423 (SAS Institute, Cary, NC) were used to estimate β (beta) from the explicitly constructed risk sets. To parallel the descriptive SMR analysis of mortality in relation to exposure, categorized forms of the covariates were considered. The statistical significance of each main effect (expressed as a global P value) was assessed with a likelihood ratio statistic. For the quantitative exposure variables, a test for linear trend was also conducted (expressed as a trend P value). Using the continuous form of the exposure metrics, we also evaluated main effects with a test of the β (linear slope) coefficient derived from the model. We considered both unlagged and 15-year lagged versions of all exposure metrics. We conducted all tests on RRs and SMRs at the two-sided 0.05 significance level, and no formal adjustment was made for multiple comparisons.
We evaluated SMRs and RRs by study factor to determine if any study factors were statistically significant predictors of lung cancer mortality, and thus potential cofactors in our exposure-response modeling. RRs are not shown for sex, age at risk and time period at risk as these factors were used to match lung cancer cases to controls in the internal comparisons. We also did not compute RRs by race category as this information was unavailable in the four EU study countries.
Indirect Adjustment for Potential Confounding by Smoking
We used the methods described by Richardson24 and Richardson et al25 to indirectly adjust RRs for lung cancer among long-term (1+ years) workers for potential confounding by smoking. We chose the cause of death category, “ischemic heart disease” as the smoking-related cause of death category that is not associated with work in the hardmetal industry (Table 3, SMR = 1.01 for long-term (1+ years) workers).
Description of Pooled Cohort
Table 2 shows the distribution of the pooled cohort by selected study factors. The final cohort consisted of 32,354 persons (73.2% male) who contributed 798,330 person-years at risk. About 70% of the cohort worked longer than 1 year and 45% worked longer than 5 years. Sweden comprised almost one-half the total cohort (n = 14,768); the smallest cohort came from the UK (n = 1520). Sweden also had the highest proportion of female workers (49.3%). About 40% of the cohort was born after 1960, making them 55 years old or younger at the end of study, and nearly one-third of the cohort was hired at age 35 or older. More than half the cohort was hired after 1979. At the end of the study period, 24,674 study members (76.3%) were assumed alive and 7187 (22.2%) were known to be dead. Cause of death was ascertained for 7122 or 99.1% of the deaths. Only 1.5% of the cohort was untraceable. Nearly 20% of the cohort was employed for 20+ years and 68.9% had 20 or more years’ TSFE. Although 33.3% of the cohort had more than 5 years’ employment with more than 20 years of follow-up, only 18.3% of the cohort had more than 10 years of employment and more than 30 years of follow-up.
Table 426,27 shows selected descriptive statistics for the Agent_Measures used in the study computed across all workers with known exposures. Figure 1 shows the mean Agent_AIEs by country. Figure 2 shows for all countries combined the mean Agent_AIEs by study time periods. Mean Tungsten, Cobalt, and Nickel_AIEs varied little across countries. Across the time period of the study, mean Agent_AIEs generally decreased at varying rates across time.
Table 4 and Fig. 1 also show that for each country the median and maximum Tungsten_AIE (mg/m3) and median and maximum Nickel_AIE (mg/m3), calculated using the exposure interval midpoints, were well below the 2016 US American Conference of Governmental Industrial Hygienists threshold limit values (TLVs) of 1.5 and 5.0 mg/m3, respectively.27 For Co, the median Cobalt_AIE (mg/m3) was below the 2016 TLV of 0.02 mg/m3 for workers in each study plant and overall, however, each country had some workers with maximum Cobalt_AIE values that exceeded the 2016 US TLV.
General Mortality Patterns: External Comparisons
Table 5 shows, for the total pooled cohort during the entire 1952–2014 study period, observed deaths and SMRs for all causes of death combined and selected cause-specific categories. We observed 7187 deaths overall yielding small but statistically significant (P < 0.01) 12% and 14% overall excesses in mortality compared with the national and regional experiences, respectively. Cause-specific SMRs based on national (SMR-N) and regional (SMR-R) rates were generally similar. We observed statistically significantly elevated overall SMRs for the categories: all malignant neoplasms (1857 deaths, SMR-R = 1.06), cancers of the buccal cavity and pharynx (63 deaths, SMR-R = 1.69), cancers of digestive organs and peritoneum (589 deaths, SMR-R = 1.10), respiratory system cancer (479 deaths, SMR-R = 1.19) including cancer of bronchus, trachea and lung termed “lung cancer” (459 deaths, SMR-R = 1.20), all heart disease (2041 deaths, SMR-R = 1.11) including ischemic heart disease (1575 deaths, SMR-R = 1.15), nonmalignant respiratory disease (NMRD) (499 deaths, SMR-R = 1.20) including influenza and pneumonia (171 deaths, SMR-R = 1.20), bronchitis, emphysema and asthma (chronic obstructive pulmonary disease) (224 deaths, SMR-R = 1.21), and emphysema (58 deaths, SMR-R = 1.44), cirrhosis of the liver (184 deaths, SMR-R = 1.34) and all external causes of death (748 deaths, SMR-R = 1.31) including accidents (368 deaths, SMR-R = 1.21) and the residual category “all other accidents” (271 deaths, SMR-R = 1.41), and homicides (162 deaths, SMR-R = 2.29). Overall statistically significant deficits in deaths were observed for breast (64 deaths, SMR-R = 0.78) and kidney cancer (39 deaths, SMR-R = 0.70). The SMR-R for Other NMRD (which includes the cause death “hardmetal lung disease”) was 1.18 (95% CI = 0.96–1.43) compared with a statistically significant deficit in deaths based on the national comparison (SMR-N = 0.81, 95% CI = 0.66–0.99).
Table 3 shows observed deaths and SMRs for all causes of death and selected cause-specific categories for short and long-term workers, where we use worker type 1 (1+ years of employment) as the default long-term worker definition here and in tables that follow. SMRs for short-term workers in this table do not include the less than 1 year person-years of long-term workers. Also, here and in tables that follow, SMRs were based only on the regional comparisons. Table 3 clearly shows that the many statistically significant mortality excesses noted among the total pooled cohort were limited to short-term workers. That is, with the exception of the sole category “homicides and other external causes” (69 deaths, SMR-R = 1.62), long-term workers revealed no statistically significantly elevated SMRs for any of the cause of death categories examined. Among long-term workers, we observed 287 lung cancer deaths yielding a small, not statistically significant SMR-R of 1.10. For NMRD and its subcategories, long-term workers revealed mostly deficits in deaths or slight, not statistically significantly excesses. This includes an SMR of 0.99 (95% CI = 0.76–1.26) for Other NMRD. Even more pronounced, long-term workers revealed mostly statistically significant deficits in deaths from all external causes and most of its subcategories, compared with statistically significant excesses among the total pooled cohort.
Table 6 shows observed deaths and SMRs for lung cancer by country and sex for short and long-term workers as defined in Table 3. For males and females combined, lung cancer SMRs among long-term workers are below or close to 1.00 in all countries except Sweden, which revealed a statistically significant 42% excess in lung cancer mortality (144 deaths, SMR-R = 1.42). Although small numbers of deaths precluded a full evaluation of lung cancer relative to country and sex, among long-term workers, lung cancer SMRs were higher among females in Germany, Sweden and the US. Among all long-term workers, we observed a statistically significant 34% excess in lung cancer deaths among females (66 deaths, SMR-R = 1.34) compared with a slight, not statistically significant 4% excess among male workers (221 deaths, SMR-R = 1.04).
Lung Cancer Mortality in Relation to Study Factors: External and Internal Comparisons
Table 7 shows for the total pooled cohort observed deaths for lung cancer and corresponding RRs and regional rate-based SMRs by selected study factors. As expected from the results of Table 3, worker type 1 and 5 were both statistically significant predictors of lung cancer mortality (global test P value = 0.039 and 0.012, respectively). Although the statistically significantly elevated lung cancer SMR for Sweden is clearly distinct from the deficits in deaths observed for the other countries, the country-specific RRs for lung cancer, using Sweden as the arbitrary baseline, did not reveal statistically significant heterogeneity. Table 7 also reveals no trends or patterns in SMRs or RRs for factors that related more directly to working experience, such as DOE and TSFE. In fact, among workers within the four categories of longer DOE and TSFE (DOE by TSFE), where an occupational association would most likely be detected, we found slight, not statistically significant excesses or deficits in lung cancer deaths. When based on long-term (1+ years) workers only, none of the factors examined in Table 7 (excluding worker type 1) was a statistically significant predictor of lung cancer mortality (data not shown).
We repeated the analysis in Table 7 for female long-term (1+ years) to evaluate further the elevated female lung cancer risks observed in Table 6 (Table 8). Only one study factor, country, revealed evidence of an association with lung cancer risk (global P value = 0.026), but interpretation is limited by the absence of lung cancer deaths in two of the five countries. Additionally, Table 8 reveals no evidence that the elevated lung cancer risks in observed in Table 6 were related to employment in the hardmetal industry. For example, the largest RR for female lung cancer by age at hire occurred among women hired at age 50 or older compared with women hired under the age of 30 (RR = 1.20, 95% CI = 0.49–2.48), and the RR for women employed 20+ years compared with women employed 1 to 4 years was 0.69 (95% CI = 0.31–1.55).
Lung Cancer Mortality in Relation to Occupational Exposure Among Long-Term Workers
Because of the marked differences in lung cancer mortality risks between short and long-term (1+ years) workers observed in our general mortality analyses (Tables 3, 6, 7, 8), we limited our exposure-response analyses for lung cancer to long-term workers. Table 9 shows univariate RR regression models relating internal cohort rates for lung cancer to unlagged duration, cumulative, and average intensity of exposure to W, Co, and Ni. We categorized exposures at approximate percentiles of the distribution for the lung cancer cases. Because Agent_AIEs tended to cluster at specific values, using a smaller number of categories for Nickel_AIE was necessary for to balance the number of observed deaths across categories adequately. To provide complementary external comparisons and to evaluate lung cancer mortality risks in the baseline categories of the RRs, we also computed SMRs by exposure category.
The global and trend test P values and continuous form slope estimates P values in Table 9 indicate that no Agent_Metric was a statistically significant predictor of lung cancer risk. Moreover, we observed elevated lung cancer risks in certain exposure Agent_Metric categories, including Cobalt_Cum4; however, none of the RRs was statistically significant. Only a few isolated SMRs were statistically significant and there was no evidence of any trends in SMRs or RRs relative to increasing exposure.
When limited to long-term workers (1+ years), none of the study factors examined in Tables 7 and 8 was a statistically significant predictor of lung cancer risk. We refitted the RR regression models in Table 9 to include country (a key factor in the pooled analysis) as a potential confounding factor. The results of the RR models adjusted for country are shown in Table 10. Similar to Table 9, none of the Agent_Metrics in the adjusted models was a statistically significant predictor of lung cancer risk, and we observed no evidence of any positive exposure-response relationships. The results of the models in Table 10 are illustrated in Figs. 3 and 4 for the AIE and CUM metrics.
We also performed parallel analyses of (1) the models in Table 9 based on a 15-year lagged version of the same Agent_Metrics; (2) the models in Table 9 based on unlagged and 15-year lagged models excluding Sweden (due to its inordinately elevated lung cancer risks even among Swedish long-term (1+ years) workers); (3) the models in Table 10 based on a 15-year lagged version of the same Agent_Metrics; and (4) the models in Table 10 limited to female long-term workers (with recategorized Agent_Metrics due to smaller number of observed deaths). Although some of these additional models revealed statistically significantly elevated SMRs or RRs for certain Agent_Metric categories, we observed no clear evidence of any exposure-response relationships, including when Sweden was omitted from the models. Most of the continuous form slope estimates were close to zero and none was statistically significant (data not shown).
Lung Cancer Mortality in Relation to Pre- and Post-Sintering Operations
Table 11 shows the number of workers employed in pre- and post-sintering jobs by country. Among the total pooled cohort, 18,451 (57.0%) workers were employed only in pre- or post-sintering jobs or both types of jobs; the remaining 43% were in administrative jobs, support jobs (eg, trades, material handling) and/or production jobs that did not involve hardmetal (eg, ceramics). Austria and Germany had the largest percentages of workers employed in these jobs (69 and 66%, respectively). Tables 12 and 13 show, respectively, regional rate-based SMRs for lung cancer by overall work experience and by duration of time worked in pre- and post-sintering jobs among long-term workers (1+ years). Workers with pre-sintering jobs only or post-sintering jobs only revealed small, not statistically significant 1.15 and 1.10-fold excesses in lung cancer mortality, similar to the total pooled cohort (SMR = 1.10). Work with mixed pre-/post-sintering jobs had a 1.42-fold risk of lung cancer that was not statistically significant (Table 12). Aside from an isolated statistically significant 1.71-fold excesses lung cancer among workers employed 5 to 9 years in pre-sintering jobs (SMR = 1.71, 95% CI = 1.12–2.50), Table 13 reveals no evidence of an increasing risk of lung cancer in relation to duration of time worked in pre- or post-sintering jobs.
Lung Cancer Mortality Risk Estimates Adjusted for Potential Confounding by Smoking
Table 14 shows the results of our application of the Richardson method24,25 to indirectly adjust lung cancer RRs from the exposure-response analysis presented in Table 10 for potential confounding by smoking. Shown are observed deaths and RRs for lung cancer and ischemic heart disease categorized into “unexposed” and “exposed” groups for the three Agent_AIEs of interest. For each Agent_AIE, the unexposed category was the baseline category used in the models in Table 10 and the exposed category was the aggregate of the corresponding non-baseline categories. Mathematical details of our application of the Richardson method and associated variance estimates are provided in Appendix D.
Table 14 shows that the RRs for ischemic heart disease for Tungsten_AIE and Cobalt_AIE were larger than 1.0 (1.05 and 1.17, respectively) indicating a small degree of positive confounding by smoking, and the RR for Nickel_AIE was below 1.0 (0.83), indicating a small degree of negative confounding by smoking. Using these RRs for ischemic heart disease to adjust for confounding by smoking, the unadjusted RR for workers exposed to W and Co are reduced toward or below the null (from 1.34 to 1.28 and from 1.07 to 0.91, respectively) and the unadjusted RR for workers exposed to Ni increased from 0.88 to a value slightly above the null (1.06). Overall, the adjusted RRs in relation to Agent_AIE in Table 14 provide even less evidence of an exposure-response relationship than those shown in Table 10.
We also applied the Richardson method to female long-term (1+ years) workers to indirectly adjust lung cancer RRs from the corresponding exposure-response analysis for potential confounding by smoking. This yielded RRs for ischemic heart disease of 1.33, 1.42, and 1.15 for Tungsten_AIE, Cobalt_AIE and Nickel_AIE, respectively, indicating a moderate degree of positive confounding by smoking larger than that observed among all long-term workers for Tungsten_AIE and Cobalt_AIE (Table 14). For female long-term workers categorized as “exposed,” lung cancer RRs adjusted for confounding by smoking via the Richardson method were 0.68, 0.86, and 0.71 for Tungsten_AIE, Cobalt_AIE, and Nickel_AIE (data not shown). We repeated the above analyses using the Agent_Cum metrics and this yielded similar pattern of results (data not shown).
Our pooled cohort analysis revealed no consistent evidence of elevated lung cancer mortality risks overall or among the many demographic and exposure-based subgroups examined. Of particular note were consistent deficits in lung cancer mortality based on internal comparisons of workers with the greatest potential for risk (eg, employed more than 5 or more than 10 years, and followed for 20 or more or 30 or more years) with workers without these risk potentials. Our external mortality comparisons using regional rates also yielded statistically significant 1.39 and 1.42-fold excesses in lung cancer mortality among short-term workers compared with long-term workers who revealed slight, not statistically significant 1.02- and 1.10-fold excesses (using both 5 and 1 year cut points, respectively).
Our comprehensive quantitative exposure assessment enabled us to rigorously evaluate lung cancer mortality risks in relation to the duration, average intensity and cumulative exposure to W, Co, and Ni. Our exposure-response analyses of these Agent_Metrics, based primarily on RR regression modeling, was limited to long-term workers and included the co-factor country. The unadjusted univariate models and adjusted multivariable models indicated that none of the agents examined was a risk factor for lung cancer mortality. These findings were maintained in models with 15-year lagged Agent_Metrics exposure and/or models that included or excluded the Swedish cohort. Our exposure-response findings are consistent with the observation that the median AIE to each agent (calculated across all workers) was well below the 2016 TLVs for W (5.0 mg/m3 total aerosol), Co (0.02 mg/m3 total aerosol), and Ni (1.5 mg/mg3 inhalable fraction) and indicate that workers exposed to these levels of W, Co, and Ni are not at an increased risk for lung cancer.27
W is not considered to be a known or suspected carcinogen by IARC.1 Ni compounds and metallic nickel have been deemed a Group 1 (carcinogenic in humans) and Group 2B carcinogen (possibly carcinogenic to humans), respectively.28 However, the Ni exposures in this study are considerably lower than those found in cohort studies of Ni refinery workers.29–33 IARC classified Co metal with WC, or hardmetal (WCCo), as Group 2A (probably carcinogenic), and Co metal without WC as Group 2B (possibly carcinogenic).1 There was some support among IARC working group members to designate hardmetal as Group 1 (carcinogenic) based on the earlier epidemiologic studies. The IARC evaluation identified the highest Co exposures in weighing, grinding and finishing.
In this study, Co exposures were highest in pre-sintering and powder production operations. Overall, Co exposure levels here were equal to or lower than those reported for hardmetal operations in other studies of workers in the industry.8 Our analysis of long-term workers classified into pre-sintering and post-sintering jobs revealed no important excesses or differences in lung cancer risk. We were able to evaluate qualitative WCCo exposure in one plant (Plant 3) in the US study7; however, due to the small numbers of deaths it was not possible to evaluate exposure-response relationships among these workers.8
We also observed higher lung cancer risks among female versus male long-term workers in Germany, Sweden, and the US. An analysis of these study factors among female long-term workers revealed no evidence that the elevated lung cancer risks observed in Table 6 were occupationally related. For example, the largest RR for female lung cancer by age at hire occurred among women hired at age 50 years or older compared with women hired under the age of 30 (RR = 1.20, 95% CI = 0.49–2.48), and the RR for women employed 20+ years compared with women employed 1 to 4 years was 0.69 (95% CI = 0.31–1.55) (data not shown).
These findings in short-term and female workers are unlikely due to occupational factors in the hardmetal industry but rather to differences in lifestyle and behavior (eg, higher rates of smoking) that could impact lung cancer risk,34 or to exposures received before or after employment in the hardmetal industry. The elevated RR for women hired at or after age 50 in particular indicates that there was considerable time for previous employment in other industries. Our application of the Richardson method24,25 to female long-term workers indicated that positive confounding by smoking accounted for a major portion of the lung cancer excesses noted among female workers. Finally, with the exception of Sweden, we observed deficits in lung cancer deaths among all long-term workers in all countries. The elevated lung cancer risks in Sweden among short- and long-term workers were found by the authors of the country-specific study not to be related to occupational factors due to a lack of an exposure-response relationship.12
Comparison of Results With Earlier Studies in France and Sweden
Figure 5 compares lung cancer SMRs from the pooled cohort analysis (total pooled cohort and long-term workers with and without Sweden included) to those observed in the country-specific analyses and the earlier French and Swedish studies. Our pooled and country-specific results are generally not consistent with the results from the earlier studies. The current studies revealed elevated and mostly statistically significant overall SMRs for lung cancer compared with deficits in deaths in all countries except Sweden. For example, overall lung cancer SMRs in the studies of French workers on which the 2006 IARC evaluation of Co was primarily based4–6 ranged from a borderline statistically significant 1.30 (95% CI = 1.00–0.66)5 to a statistically significant 2.13 (95% CI = 1.02–3.93).4 The earlier Swedish study3 observed an overall SMR of 1.34 that was not statistically significant. Using Poisson regression of lung cancer SMRs, we confirmed the heterogeneity of the five country-specific SMRs in our international study (P < 0.001) and our pooled SMR (without Sweden) versus the earlier French and Swedish studies (P = 0.0005), and confirmed the homogeneity of the Austrian, German, UK, and US SMRs (P = 0.503) and the SMRs in the earlier French and Swedish studies (P = 0.350).
Comparing lung cancer mortality risks across country-specific studies within specific demographic, work history or exposure-related subgroups was made difficult by the variability in the analytical or presentation methods used and/or the small number of observed deaths found in many subcategories. The earliest French study,4 based on small numbers of observed lung cancer deaths, found not statistically significantly elevated SMRs for lung cancer among workers with the longest DOE (20+ years, one observed death, SMR = 2.52, 95% CI = 0.06–14.02) and longest TSFE (20+ years, three observed deaths, SMR = 2.17, 95% CI = 0.45–6.34), and a statistically significant SMR for workers with the highest degree of exposure to WCCo determined qualitatively (six observed deaths, SMR = 5.03, 95% CI = 1.85–10.95). We were unable to evaluate mortality risks in relation to WCCo specifically in the pooled analysis, but found no evidence of increased risk of lung cancer with longest DOE or TSFE.
Moulin et al5 performed a cohort study of 10 French hardmetal plants that included a nested case-control study on lung cancer. A semi-quantitative exposure matrix was developed, and measured Co concentrations were used to validate exposures. At the cohort level, there was little evidence that lung cancer risk increased with TSFE with the largest, not statistically significant, SMR occurring among workers with TSFE 20–29 years (SMR = 1.42, 95% CI = 0.93–2.06). The nested case-control study found higher but not statistically significant risks in workers involved in “before sintering” jobs (OR = 1.69, 95% CI = 0.88–3.27) compared with “after-sintering” jobs (OR = 1.26, 95% CI = 0.66–2.40) and a statistically significant trend in lung cancer risk with increasing DOE and frequency weighted cumulative exposure, especially in pre-sintering operations. Adjustment for smoking or other potential confounders did not change the risk estimates.
Wild et al6 did not quantify exposures, but instead used a qualitative level of WCCo exposure and an evaluation of work with exposure to pre-sintered versus sintered product. They found a statistically significant lung cancer SMR for pre-sintering of 2.42 (nine deaths, 95% CI = 1.10–4.59) compared with a not statistically significant lung cancer SMR for sintered product workers of 1.28 (five deaths, 95% CI = 0.41–2.98). Our pooled analysis of long-term workers yielded a lung cancer SMR for pre-sintering of 1.15 (95% CI = 0.91–1.44) based on 78 deaths, and SMR for post sintering of 1.10 (95% CI = 0.84–1.42) based on 58 deaths.
The earlier Swedish cohort study,3 using a qualitative categorization of Co exposure, found not statistically significant lung cancer SMRs similar to that observed in the total cohort for workers with “low” exposure (SMR = 1.31, 95% CI = 0.77–2.13) and “high” exposure (SMR = 1.39, 95% CI = 0.51–3.04). The authors also reported a statistically significant increased risk of lung cancer among workers with DOE 10+ years and TSFE 20+ years (SMR = 2.78, 95% CI = 1.11–5.72).3 We found a statistically significant excess in the third quartile of cumulative Co exposure, but no evidence of a monotonically increasing relationship with exposure levels to Co, or to W or Ni.
A recent cancer incidence study of 995 men employed in a Finnish Co plant found no evidence of an overall risk of cancer (standardized incidence ratio [SIR] = 1.00, 95% CI = 0.81–1.22), no evidence of excess lung cancer risk (SIR = 0.50, 95% CI = 0.18–1.08) and no dose-response effect.35 Mean levels of Co exposure in the Finnish cohort study (0.02 mg/m3) are slightly higher than the Cobalt_AIE in the pooled study (0.013 mg/m3).
As noted earlier, we observed decreases during the study time period in the average exposure levels of Co, Ni, and W that may be due to the incorporation of more efficient risk management measures at the study sites over time (eg, local exhaust ventilation, process enclosures and personal protection equipment) compared with the measures used during the earlier time frames of the French and Swedish studies.
Mortality From Nonmalignant Respiratory Disease
Another health endpoint of interest in our study was NMRD, especially the occupational pneumoconiosis hardmetal disease, which can be caused by dust or fume exposures from hardmetal objects containing Co. We found a statistically significant excess in NMRD (SMR-R = 1.20, 95% CI 1.10–1.31) that was primarily due to an excess of emphysema (SMR-R = 1.44, 95% CI = 1.10–1.86). However, upon comparing short- and long-term workers, we see that the NMRD risk was primarily among short-term rather than long-term workers (long-term worker type 1 [SMR-R = 0.93, 95% CI = 0.82–1.05]). There was no evidence of an occupationally related risk for the NMRD subcategory Other NMRD, in which hardmetal disease and pneumoconioses are contained. The Other NMRD SMR-R for long-term workers was 0.99 (95% CI = 0.76–1.26). This is consistent with the country-specific analyses, which found small numbers of deaths in the Other NMRD category and no evidence of an exposure-response relationship. These findings indicate that the results are most likely due to behavioral factors among the short-term workers (eg, higher smoking rates) or occupational factors outside of employment in the hardmetal industry.
Evaluation of Possible Confounding by Smoking
Negative confounding of lung cancer risk estimates by smoking does not appear to be a likely explanation for the generally decreased lung cancer risk estimates observed in our international study of hardmetal workers. Our indirect adjustment of potential confounding by smoking among all long-term workers in the pooled cohort using the Richardson method24,25 indicated some positive confounding by smoking among workers exposed to W and Co and some negative confounding among workers exposed to Ni. However, for each agent the adjusted RRs for lung cancer among exposed workers were close to the null value suggesting even less evidence of an association with exposure than revealed in the unadjusted analysis. Our application of the Richardson method to female long-term workers indicated a moderate degree of positive confounding by smoking for each of the three Agent_Metrics that was greater than that observed among all long-term workers.
Moulin et al5 adjusted lung cancer ORs for smoking and found little evidence of confounding. The indirect external adjustment for confounding by smoking in the US study7 produced uniformly lower risk estimates, indicating the possibility of positive confounding by smoking. Although based on limited data, the UK study reported that lung cancer risk estimates were not confounded by smoking,9 and the German study found no evidence of confounding by smoking on the risk of all cancers combined.10
Limitations and Strengths of Our Pooled Analysis and the International Study
A challenging aspect of our pooled analysis of data from the international cohort study of hardmetal workers was properly accounting for the known marked elevations in total and certain cause-specific mortality risks observed in the Swedish12 and German10 studies, compared with mostly deficits in deaths observed in the Austrian,11 US,7 and UK9 studies. This challenge was heightened by the fact that Sweden comprised 45.7% of the total pooled cohort of 32,354 workers and 69.5% of the 7187 total observed deaths. To a lesser but important extent, Germany represented 21.1% of the pooled cohort and 10.4% of the total deaths.
Another challenging aspect of our pooled analysis was its inclusion of a large number of short-term workers in the overall cohort (9848 or 30.4% of cohort worked less than 1 year and 17,762 or 55% worked less than 5 years) and a disproportionate percentage of short-term workers in the Swedish and German cohorts (40.7 and 22.2%, respectively, vs 10.2 and 6.6% for Austria and UK). The typically unhealthy behavioral and lifestyle characteristics associated with short-term workers often lead to less favorable mortality outcomes compared with longer-term workers.17,18,36
Thus, our pooled analysis of the international data was influenced heavily by marked elevations in mortality risks in two of the larger country-specific cohorts and by a large percentage of short-term workers that was disproportionately higher in these two larger, higher risk cohorts. We accounted for these cohort features by stratifying our pooled mortality analyses by worker type 1 (<1, 1+ years of employment), by adjusting exposure-response models for the factors worker type 1 and/or country, and by performing certain analyses with and without the inclusion of the Swedish cohort data.
As expected based on the country-specific findings, our pooled analysis of total and cause-specific mortality risks including lung cancer, via external comparisons using regional rates, revealed marked increases in mortality risks among short-term (<1 year) workers compared with mostly deficits in deaths or slight, not statistically significant excesses among long-term (1+ year) workers. Pooled cohort analyses excluding the Swedish data also generally revealed lower total and cause-specific mortality risks.
Our current findings for long-term workers in the total pooled cohort, and the previous findings in Austria, the UK, and the US of no elevated overall mortality risks for all causes, all heart disease and NMRD, are consistent with those found in other occupational groups.34 These favorable mortality patterns, especially for the long-term chronic disease categories, are probably influenced in part by the “healthy worker effect,” the relative absence of deleterious employment-related health risks, the positive health effects of continuing employment, and better health care access. Although all risk estimates were standardized for age, the study populations in most countries were young, contributing a relatively small percentage of total deaths before the end of the observation period (eg, in Austrian, German, UK, and US cohorts, only 9.0, 11.0, 11.9, and 14.9%, respectively, of workers had died by the end of follow-up). Consequently, many members did not reach the older age groups for which mortality rates for many cause of death categories increase dramatically in the general population. Continued follow-up of the country-specific cohorts may reveal different patterns and rates of mortality for some of the cause of death categories associated with older age and/or longer latency periods. Thus, our study may not be fully representative of the long-term mortality experience of workers from this industry, arguing for continued follow-up of this cohort. The extensive exposure assessment provided quantitative estimates for W, Ni, and Co, but, aside from the qualitative exposures for one site in the US study, we were unable to assess exposure to WC or WCCo (the primary exposure of interest) either quantitatively or qualitatively for the total pooled cohort. It was not possible to generate independent exposure estimates for WC or WCCo, or for carbon black, because there are no analytical methods specific to these combined agents (ie, usually reported as total aerosol) and therefore no IH measurements available.
Our pooled analysis of US and EU hardmetal workers has many methodological strengths. It included 32,354 workers from three companies and 17 manufacturing sites in five countries (eight US, three German, three Swedish, two UK, and one Austrian), each independently conducted under the direction of country-specific occupational epidemiology experts. Our pooled analysis included a comprehensive, quantitative study member-level assessment of exposure to W, Co, and Ni that enabled robust RR regression modeling of exposure-response with control for potential confounding factors. Each country conducted a systematic verification of cohort completeness that led to the exclusion or modification of data from the original enumerated cohorts. Overall, we attained excellent mortality follow-up and cause of death ascertainment rates that covered a long 63-year (1952–2014) observation period. The focus of our study was to determine whether hardmetal production workers are at an increased risk for lung cancer. We observed a total of 459 lung cancer deaths, including 287 among long-term (1+ yr) workers, that yielded excellent statistical power to detect important excesses in deaths overall and in most of the subcategories examined.
Our pooled analysis of country-specific cohort data from our international study of hardmetal production workers provided no consistent evidence that work in this industry is associated with an increased risk of lung cancer, as suggested in the earlier French and Swedish epidemiologic studies. We found no evidence that duration, average intensity or cumulative exposure to tungsten, cobalt or nickel, at levels experienced by the workers examined, increases lung cancer mortality risks. We also found no evidence that work in the US or EU hardmetal industry increases mortality risks from any other cause of death. Our results, which were consistent with the country-specific study findings, should help guide risk management efforts for workers exposed to hardmetal. The relatively young age distribution of the international study cohort warrants continued mortality follow-up.
This pooled analysis was sponsored by a grant from the International Tungsten Industry Association (ITIA). The design, conduct, analysis and conclusions of the study are exclusively those of the authors. We would like to acknowledge the cooperation and assistance of the investigators from the country-specific studies in Europe and of representatives from the ITIA and its member companies.
Appendix A. Procedures Used to Optimize Data Standardization in the Development of the Pooled Cohort Data and Standard Mortality Rate Files
Overall Pooled Cohort
- As shown in the table below, time periods associated with the ICD revisions varied across countries. For consistency, we used the US ICD time periods as the standard time periods. Because of this variation in time periods, for a given EU country we needed to recode any deaths that occurred in years outside of the US time periods into the corresponding revision in effect at time of death to match the US standard periods. The number of recoded deaths per country are shown below in the country-specific sections.
- We removed anyone who was hired and terminated before their 18th birthday. If they were hired before their 18th birthday and continued working after, their person-year start date was set to their 18th birthday and earlier jobs were removed from the cohort.
- Person-years were stopped at date of death, the end of study observation period or for persons lost to follow-up, at the time of last known vital status or termination date from employment.
- Less than 20 unique ICD codes needed to be changed from 9th to 10th revision due to ICD time period differences.
- No one died after 12/31/2014 so no modifications were made.
- Truncated person years for those to left Austria.
- Truncated person years at last observed date.
- 13 people were removed from the cohort due to being hired/terminated before their 18th birthday.
- All deaths in the German cohort analysis were coded to the 9th revision. However, they provided 10th revision codes. In the pooled analysis, 9th and 10th revision codes were used.
- No one died after December 31, 2012 so no modifications were made.
- People who died after their 85th birthday (n = 89) were changed to alive.
- Truncated person years at 85th birthday.
- Truncated person years at last observed date.
- 55 people were removed from the cohort due to being hired/terminated before their 18th birthday.
- Sweden had the most deaths that needed to be recoded due to ICD time period differences. See list below:
- Less than 20 people were duplicated in the cohort because they worked at more than one plant.
- 120 people with multiple plants were assigned to the plant of longest work.
- 15 people with no work history were excluded.
- People who died after 12/31/2012 (n = 366) were changed to alive.
- Person years were not truncated at any time.
- Less than 20 unique ICD codes were recoded due to differences in the ICD time period.
- 11 people with no work history were excluded.
- People who died after December 31, 2012 (n = 17) were changed to alive.
- Person years were not truncated at any time.
- Seven people were removed from the cohort due to being hired/terminated before their 18th birthday.
- No one died after December 31, 2012 so no modifications were made.
- Person years were not truncated at any time
Appendix B. Regional Groups Used for the Construction of Regional Comparison Mortality Rates, by Country and Plant
Appendix C. Causes of Death Categories and Revision-Specific International Classification of Disease Codes Used in Mortality Analysis
Based on the Richardson's method,24,25 the estimated adjusted RR for lung cancer can be calculated as the exponentiated difference between the log estimated RR for lung cancer and log estimated RR for ischemic heart disease (Equation 1):
Thus, the estimated variance of
can be calculated using the Delta method (Equation 2):
We calculated the variance estimate for
approximately as the sum of the two variance estimates for
without considering the covariance between the beta estimates (which can only be estimated in the joint modeling of the two outcomes). We calculated the conservative 95% confidence interval for the adjusted RR as:
1. IARC Working Group on the Evaluation of Carcinogenic Risks to Humans. Cobalt in hardmetals and cobalt sulfate, gallium arsenide, indium phosphide and vanadium pentoxide. IARC Monogr Eval Carcinog Risks Hum
2. Ward EM, Schulte PA, Straif K, et al. Research recommendations for selected IARC-classified agents. Environ Health Perspect
3. Hogstedt C, Alexandersson R. Dodsorsaker hos Hardmetallarbetare. Arbete och Halsa
4. Lasfargues G, Wild P, Moulin JJ, et al. Lung cancer mortality in a french cohort of hard-metal workers. Am J Industr Med
5. Moulin JJ, Wild P, Romazini S, et al. Lung cancer risk in hard-metal workers. Am J Epidemiol
6. Wild P, Perdrix A, Romazini S, Moulin JJ, Pellet F. Lung cancer mortality in a site producing hardmetals. Occup Environ Med
7. Marsh GM, Buchanich JM, Zimmerman SD, et al. Mortality among hardmetal production workers: US cohort and nested case-control studies. J Occup Environ Med
8. Kennedy KJ, Esmen NA, Buchanich JM, Zimmerman S, Sleeuwenhoek AJ, Marsh GM. Mortality among hardmetal production workers: occupational exposures. J Occup Env Med
9. McElvenny DM, MacCalman LA, Sleeuwenhoek A, et al. Mortality among hardmetal production workers: UK cohort and nested case-control studies. J Occup Environ Med
10. Morfeld P, Groß JV, Erren TC, et al. Mortality among hardmetal production workers: German historical cohort study. J Occup Environ Med
11. Wallner P, Kundi M, Moshammer H, Zimmerman SD, Buchanich JM, Marsh GM. Mortality among hardmetal production workers: A retrospective cohort study in the Austrian hardmetal industry. J Occup Environ Med
12. Westberg H, Bryngelsson I-L, Marsh G, et al. Mortality among hardmetal production workers: the Swedish cohort. J Occup Environ Med
13. Westberg H, Bryngelsson I-L, Marsh G, et al. Mortality among hardmetal production workers: Swedish measurement data and exposure assessment. J Occup Environ Med
14. Youk AO, Marsh GM, Stone RA, Buchanich JM, Smith TJ. Historical cohort study of US man-made vitreous fiber production workers: III. Analysis of exposure-weighted measures of respirable fibers and formaldehyde in nested case-control study of respiratory system cancer. J Occup Environ Med
15. Hauptmann M, Lubin JH, Stewart PA, Hayes RB, Blair A. Mortality from solid cancers among workers in formaldehyde industries. Am J Epidemiol
16. Marsh GM, Youk AO, Stone RA, Sefcik S, Alcorn C. OCMAP-PLUS: a program for the comprehensive analysis of occupational cohort data. J Occup Environ Med
17. Gilbert ES. Some confounding factors in the study of mortality and occupational exposures. Am J Epidemiol
18. Kolstad HA, Olsen J. Why do short term workers have high mortality? Am J Epidemiol
19. Breslow NE, Day NE. The Design and Analysis of Cohort Studies. Statistical Methods in Cancer Research. Lyon, France: Oxford University Press; 1987.
20. Cox DR. Regression models and life-tables. J R Stat Soc Ser B (Methodological)
21. Cox DR. Partial likelihood. Biometrika
22. StataCorp. Stata Statistical Software: Release 14. College Station, TX: StataCorp LP; 2015.
23. SAS Institute Inc. Base SAS®
9.4 Procedures Guide: Statistical Procedures. Cary, NC: SAS Institute Inc; 2013.
24. Richardson DB. Occupational exposures and lung cancer: adjustment for unmeasured confounding by smoking. Epidemiology
25. Richardson DB, Laurier D, Schubauer-Berigan MK, Tchetgen ET, Cole SR. Assessment and indirect adjustment for confounding by smoking in cohort studies using relative hazards models. Am J Epidemiol
26. American Conference of Governmental Industrial Hygienists. Documentation of the Threshold Limit Values and Biological Exposure Indices. Cincinnati, OH: American Conference of Governmental Industrial Hygienists; 1991.
27. American Conference of Governmental Industrial Hygienists. Threshold Limit Values for Chemical Substances and Physical Agents and Biological Exposure Indices. Cincinnati, OH: American Conference of Governmental Industrial Hygienists; 1986–2016.
28. Chromium, Nickel and Welding. IARC Monogr Eval Carcinog Risks Hum
29. Grimsrud TK, Berge SR, Haldorsen T, Andersen A. Exposure to different forms of nickel and risk of lung cancer. Am J Epidemiol
30. Grimsrud TK, Berge SR, Haldorsen T, Andersen A. Can lung cancer risk among nickel refinery workers be explained by occupational exposures other than nickel? Epidemiology
31. Grimsrud TK, Berge SR, Martinsen JI, Andersen A. Lung cancer incidence among Norwegian nickel-refinery workers 1953–2000. J Environ Monit
32. Grimsrud TK, Berge SR, Resmann F, Norseth T, Andersen A. Assessment of historical exposures in a nickel refinery in Norway. Scand J Work Environ Health
33. Grimsrud TK, Peto J. Persisting risk of nickel related lung cancer and nasal cancer among Clydach refiners. Occup Environ Med
34. Checkoway H, Pearce N, Kriebel D. Research Methods in Occupational Epidemiology. New York, NY: Oxford University Press; 2004.
35. Sauni R, Oksa P, Uitti J, Linna A, Kerttula R, Pukkala E. Cancer incidence among Finnish male cobalt production workers in 1969–2013: a cohort study. BMC Cancer
Copyright © 2017 by the American College of Occupational and Environmental Medicine
36. Boffetta P, Sali D, Kolstad H, et al. Mortality of short-term workers in two international cohorts. J Occup Environ Med