Mortality Among Hardmetal Production Workers: German Historical Cohort Study : Journal of Occupational and Environmental Medicine

Secondary Logo

Journal Logo


Mortality Among Hardmetal Production Workers

German Historical Cohort Study

Morfeld, Peter PhD; Groß, Juliane Valérie MD; Erren, Thomas C. MD; Noll, Birgit MS; Yong, Mei PhD; Kennedy, Katrin J. PhD; Esmen, Nurtan A. PhD; Zimmerman, Sarah D. MS; Buchanich, Jeanine M. PhD; Marsh, Gary M. PhD

Author Information
Journal of Occupational and Environmental Medicine 59(12):p e288-e296, December 2017. | DOI: 10.1097/JOM.0000000000001061
  • Free


In 2006, the International Agency for Research on Cancer (IARC) classified tungsten carbide with a cobalt binder (WC+Co), also referred to as “hardmetal”, as a probable human carcinogen (2A).1 This classification was based on sufficient evidence in experimental studies but limited evidence in humans. The limited epidemiological evidence was primarily derived from occupational studies of Swedish2 and French workers.3–5

Recent reviews6,7 revealed methodologic limitations of these occupational studies, such as residual confounding, inadequate definition of unexposed groups, lacking statistical power, crude exposure assessment, or potential bias due to healthy worker effects. These studies, however, raised concerns that hardmetal dust exposures may cause excess risks for the following endpoints: all causes of death; all malignant neoplasms; cancer of the bronchus, trachea, lung; nonmalignant respiratory disease; bronchitis, emphysema, and asthma.

To address the limitations of existing epidemiological studies, the International Tungsten Industry Association (ITIA) initiated a three-phase occupational epidemiology investigation of workers employed in the hardmetal industry in the early 2000s. A multinational occupational cohort study was designed as an essential step in appraising the health implications of occupational exposure to WC+Co.8,9

The present report is on the German component that comprises three manufacturing sites in the states of North Rhine-Westphalia (plant 1) and Bavaria (plant 2, plant 3). The present report focuses on overall mortality and lung cancer mortality.


Study Population

For this retrospective cohort study, former and current employees working at three German plants were included [two Bavarian sites (start of production in 1960 and 1971, respectively) and one in North Rhine-Westphalia (start of production in 1926). A total of 7950 blue-collar workers (25% female, 75% male) who were employed in hardmetal processing for at least 6 weeks were enumerated, including 265 (170 female, 95 male) blue-collar office workers having their office inside the plant. Only 2.4% of all current workers refused to take part in the study.

A full analytical cohort was defined that included 6865 enumerated workers who survived until the start of mortality follow-up (ie, 2005 from plant 1, and until 1975 in plants 2 and 3). In addition, this cohort was censored at an age of 85 years due to difficulties of assigning causes of death in old ages. Due to data protection issues and short storage periods of data at the health offices, access to cause of death was strongly limited, in particular in a health authority in the state of North Rhine Westphalia (location of plant 1). (937 from 4427 compared with 116 from 2220 and 32 from 1303 lost workers at plant 2 and plant 3, respectively). Therefore, plant 1 cohort is a census cohort. Totally, 3490 exposed workers from plant 1, 2104 in plant 2, and 1271 in plant 3 were included as eligible population. The study base is described in Table 1. All statistical analyses in the present report are based on this cohort.

Description of Study Population and Tracing of Vital Status and Death Causes Ascertainment

Exposure Assessment

Industrial hygiene (IH) data and job history data were extracted to reconstruct a complete exposure profile for each individual. A total of 1443 (989 area related, 454 personal) IH measurements were included, during the time period between 1970 and 2012. For validation, two different job-exposure matrices (JEMs) were developed for exposure assessment, one based on Similar Exposure Groups (SEGs) and another based on Job Class Number (JCN). First, three SEGs (low, medium, and high) were categorized on the basis of expert judgment according to department, jobs, and tasks per plant. The categories of low and medium were combined together due to limited discrimination in concentration data, remaining low and high categories finally.

In parallel, 29 JCNs were verified by the company's Environmental Health and Safety (EHS) experts. Those jobs performed in rotation were handled with the varying time fractions held at the plant in specific years. Concentration data for the five agents were allocated to the 29 JCNs and their associated SEG.

Quantitative occupational exposure estimates were generated through a process of modeling and validation using IH data from the three study plants and information on jobs held during measurements. Methods were developed to extrapolate the estimates backwards in time where no measurement data were available taking manufacturing process changes into account (eg, moving from manual to automated methods). The resulting time-dependent concentration estimates by SEG or JCN formed two JEMs, following the methods outlined in the study by Ignacio and Bullock.10

Log-linear regression models were fitted to the adjusted measurement data (cobalt, nickel, tungsten, respirable and inhalable dust) using the two SEGs (high vs low exposure), plants and calendar time as explanatory variables. This yielded agent-specific SEG-JEMs for the time periods with IH measurements available. Analogously, a time-dependent agent and plant-specific JCN-JEM was generated.

To allocate recent measurements to historical exposure, backward extrapolation methods11 were applied, taking the degree of uncertainty into account.12 Briefly, two scenarios for backward extrapolation were applied:

  • constant backward extrapolation per plant and exposure agent based on the concentration estimate at the date of the first available measurement;
  • increasing backward extrapolation per plant and exposure agent based on the concentration trend estimated for the first 5 years of available cobalt measurements and refined after discussions with plant experts.

Finally, the individual exposure data set comprised the time-dependent JCN, department code, SEG and time-dependent duration of exposure in days (for each day under study, ie, year–month–day combination) and the time-dependent cumulated exposure concentrations (time-dependent cumulative exposures) for each agent in mg/m3·days in the following four variations:

  • constant backward extrapolation and expert-adjusted increasing backward extrapolation of the concentrations based on the SEG-JEM;
  • constant backward extrapolation and expert-adjusted increasing backward extrapolation of the concentrations based on the JCN-JEM.

Urinary Data and Data on Smoking Habits

Regarding smoking behavior, urinary data and medical data on smoking behavior were extracted from paper files at the two Bavarian plants, supplemented with additional urinary data collected from a former study at one plant,13 while data on smoking behavior were available for only 16% of the study participants at plant 1. Due to the scarce urinary data, no further detailed analyses could be carried out.

Vital Status Tracing and Cause-of-Death Determination

The information to identify the vital status, time, and place of death was provided by registry offices, while information of causes of death and underlying diseases was provided by health offices. Certificates of support from the Ministers for Internal Affairs and Ministers of the Health of North Rhine-Westphalia and Bavaria were obtained to facilitate our enquiries at the registry and health offices.

Vital Status

The study participant list of the three German plants was exported from a specific software portal (Health Study Application, HSA), which was conceived to collect personal data for epidemiological studies respecting data protection rules. Responsible registry offices were identified through the study individuals’ last addresses. The registry offices were contacted via a letter, including information about the study, its approval by Cologne University's Ethics Committee, written support of the ministers, and the individual study participant identification data (name, date of birth, address). If a study participant had moved, multiple enquiries with different registry offices were necessary, as most of these offices just store information about the “next” place the person moved to. Very few offices offered electronical access to recent information.

The registry offices provided information whether a person was alive, dead, or moved to another place. For those individuals who were deceased, the date and place of death were provided. In case the information was incomplete, alternative strategies (eg, searches via the internet) were applied to enable us to identify the place the study participant moved to. For persons who were lost to follow-up before the end of our observation period, the last date when the individual could be classified as alive was assessed and documented.

Cause of Death

The responsible health offices, identified by the city where a person deceased, were contacted via a letter, including comprehensive information about the study and the individual study participants’ name, date of birth, date and place of death, and identification number for pseudonymization. The health offices provided pseudonymized copies of the death certificates (including causes of death and diseases). At the University of Cologne, an experienced nosologist identified the underlying cause of disease and transferred all diagnoses listed on the death certificate in ICD9 and ICD10 codes. Finally, the data were entered into the specific HSA software to warrant data security. The completeness of the data with regard to death causes is displayed in Table 1.

Statistical Analyses

For external comparisons, standardized mortality ratios (SMRs) were computed according to age, sex, and calendar year specific death rate. This was done in reference to respective mortality data of the German population and state populations of Bavaria and North Rhine-Westphalia from the Federal Statistical Office (Destatis) for the specified period 1980 to 2012. Person years were censored at age 85 years. SMRs and 95% confidence intervals (CIs) based on a Poisson distribution were stratified time-dependently by duration of exposure (defined as time difference between end of last job phase and start of first job phase at the plants), time since first exposure (defined as period from hire), and time since cessation of exposure (defined as period since leaving employment) in 5-year intervals. Start of work and end of work were lagged by 0, 5, 10, 15, and 20 years. Poisson models were fitted to SMRs and we determined global P values and trend-P values for the effect of the covariables.14

For internal comparisons, time-dependent Cox proportional hazards regression models (“Cox analyses”) were fitted.15,16 To set-up these analyses, the individual time-dependent exposure data sets for the five agents and four estimation procedures were merged with the data set, including fixed individual information on date of birth, sex, nationality, age at hire, smoking status, vital status, date of death and cause of death, and loss to follow-up data. In internal Cox analyses, we applied a binary indicator to represent nationality (German: yes or no), and used Plant 2 as reference. Smoking status was categorized into smokers (active smokers) versus non-smokers (ex-smoker or never smoker). We performed the time-dependent set-up across attained age as the basic time variable with staggered entry using the survival time set-up (stset) features of Stata.17 Entry into the study was set to the start of the first job phase but no earlier than January 1, 1975, in plants 2 and 3 and no earlier than January 1, 2005, in plant 1. Exit from the study was defined by the date of death, last date known to alive (if lost to follow-up), or end of study date (December 31, 2012, if alive at the end of study). All analyses were censored at an attained age of 85 years. Exposures were lagged by 0, 5, 10, 15, and 20 years, considering all-cause mortality as an outcome of interest. Multiple testing was not performed.18 Cumulative exposures and time variables were used in internal analyses with or without taking logs.

All calculations were done with the Stata Version 13 statistics package.17 A value of 5% was applied throughout as statistical significance level.


The full analytical cohort of 6865 study participants consists of 5212 (75.9%) men and 1653 (24.1%) women. Descriptive characteristics are given in Table 2. At plant 1, 3490 former and current blue-collar workers fulfilled the criteria of the analytical cohort comprising 2576 male and 914 female workers. At plant 2 (plant 3), the total number of study participants were 2104 (1271) comprising 1585 (1051) male and 519 (220) female study participants. At plant 1, follow-up started on average at the age of 48.8 years showing that the participants were about 20 years older than that in plant 2 (26.8 years) and plant 3 (28.1 years). Women were on average 5 years older at the start of follow-up than men. Age in years at the end of follow-up was at plant 1 on average of 56.2 years, similar to the average age at end of follow-up in plant 3 (54.1 years). At plant 2, follow-up ended at an average age of 51.7 years. Women were about 5 years older at the end of follow-up than men with the exception of plant 3 where mean age at the end of follow-up was about 10 years higher for women on average than for men. Most of the study participants are German.

Characteristics of Study Population of 6,865 Exposed Workers

Cumulative Exposures and Long-Term Average Concentrations

The average duration of exposure was 11.0 years (median = 5.8 years) and the 95-percentile was at 35.3 years. Cumulative exposures and long-term average concentrations for five agents are summarized in Table 3, respectively, with low backward extrapolation and high backward extrapolation based on SEG-JEM and JCN-JEM for the analytical cohort. There was no substantial difference between unlagged cumulative exposure estimates (inhalable fraction) based on low or high extrapolations in the SEG-JEM, except for tungsten: 0.98 mg/m3·years (median, low backward extrapolation) contrasts with 2.87 mg/m3·years (median, high backward extrapolation). Nickel exposures were low (median less than 0.05 mg/m3·years) in comparison to cobalt and tungsten exposures. Cumulative cobalt exposures were estimated as 0.16 mg/m3·years (median, low backward extrapolation) or 0.23 mg/m3·years (median, high backward extrapolation).

Total Cumulative Exposure in mg/m3.years, Low and High Backward Extrapolation, respectively, Based on SEG-JEM and JCN-JEM, for Full Analytical Cohort (N = 6,865) with 0 Lag Year

Findings based on the JCN-JEM differ slightly from those derived from the SEG-JEM, while no divergent pattern can be found when comparing the results based on either JEM.

The distributions of the long-term arithmetic mean concentrations within the analytical cohort and based on the SEG-JEM (no lag applied) are summarized in Table 4, respectively, with low backward extrapolation and high backward extrapolation. There is no noteworthy difference between both extrapolation scenarios documented. Long-term average inhalable nickel concentrations were low at about 0.01 mg/m3 (median) and 0.02 mg/m3 (mean). For inhalable cobalt, we found 0.04 mg/m3 (median) and 0.07 mg/m3 (mean), respectively. Inhalable tungsten concentrations were at 0.2 mg/m3 (median) and 0.4 mg/m3 (mean). The long-term average concentration of the respirable dust was estimated as about 0.25 mg/m3 (median) and 0.35 mg/m3 (mean); for the inhalable dust fraction, we got approximately 1.3 mg/m3 (median) and 1.7 mg/m3 (mean).

Arithmetic Mean of Concentration in mg/m3, Low and High Backward Extrapolation, Respectively, Based on SEG-JEM and JCN-JEM, for Full Analytical Cohort (N = 6,865) with 0 Lag Year

Quantitative estimates of exposures based on JCN-JEM differ slightly with no divergent pattern for an analysis based on the SEG-JEM.

Standardized Mortality Ratios (SMRs)

Table 5 summarizes SMRs for specific endpoints based on West-German reference rates in the full analytical cohort contributing to 109,202 person-years of follow-up. A total of 753 deaths occurred, leading to a significantly elevated SMR of 1.16 (95% CI: 1.08 to 1.24) for all-cause mortality. We note that this significant excess was also observed in both subcohorts of male and female workers. All cancers showed a small but non-statistically significant elevation in all three plants. The SMR for lung cancer, based on 45 observed deaths, was slightly decreased in the full analytical cohort, but nonsignificantly (SMR = 0.93, 95% CI: 0.68 to 1.25). The decrease reflects a deficit among men (36 observed, SMR = 0.87, 95% CI: 0.61 to 1.20), whereas women showed a nonsignificant increase based on nine cases (SMR = 1.29, 95% CI: 0.59 to 2.45). Nonmalignant respiratory diseases (58 deaths) were in excess: SMR = 1.56 (95% CI: 1.18 to 2.02), in men (47 deaths, SMR = 1.62; 95% CI: 1.19 to 2.16) and in women (11 deaths, SMR = 1.34; 95% CI: 0.67 to 2.39). Eight deaths occurred from nonmalignant respiratory diseases such as bronchitis, emphysema, or asthma, leading to a decreased but imprecise SMR = 0.92 (95% CI: 0.40 to 1.82).

Standardized Mortality Ratios According to Selected Causes of Deaths for Full Analytic Cohort (N = 6,865)

Among the SMRs that are listed because of their post-hoc selection were cancers of the buccal cavity and pharynx with almost a doubling of risk (significant excess based on 14 deaths: SMR = 1.87, 95% CI: 1.02 to 3.13), and cancers of the large intestine that showed some elevation (observed = 23, SMR = 1.53, 95% CI: 0.97 to 2.29). Noteworthy, SMRs for all heart diseases and ischemic heart diseases were definitely elevated. The SMR for all heart diseases was 1.56 (95% CI: 1.35 to 1.81) based on 187 cases, and clearly in excess among men (133 deaths, SMR = 1.45, 95% CI: 1.22 to 1.72) and even more in the female subcohort (54 deaths, SMR = 1.92, 95% CI: 1.44 to 2.51). Ischemic heart diseases showed similar excesses, also statistically significant despite the lower number of cases. The category “all other causes” comprised 142 observed deaths, while 105.06 were expected (SMR = 1.35, significant excess). An elevation was observed among men and women.

Cox Regression Analyses

Table 6 reports on a Cox regression model on mortality from all causes that take the cumulative exposures to cobalt, nickel, tungsten, respirable, and inhalable dust simultaneously into account (full analytical cohort, low backward extrapolation, SEG-JEM, no lagging). No exposure effects were identified. After taking logs of cumulative exposures, the only noteworthy effect is a significant downward trend with cumulative exposure to respirable dust. The global P value was significant (P = 0.045). The model confirms a significantly higher overall mortality for male workers, for German workers, and a nonsignificantly decreasing mortality with birth date, whereas no substantial influence of age at hire was found.

Cox Regression Model: Hazard Ratios for All Causes of Death (ICD-9:001-999) and 5 Cumulative Exposure in mg/m3·years

Table 7 reports on a Cox model analyzing overall mortality that simultaneously took account of the five long-term average concentration variables (cobalt, nickel, tungsten, respirable dust, inhalable dust), applying the low backward extrapolation (SEG-JEM). No statistically significant effect was described on the basis of the low extrapolation scenario.

Cox Regression Model: Hazard Ratios for All Causes of Death (ICD-9:001-999) and 5 Mean Concentrations in mg/m3

Estimated effects for overall mortality that are based on exposures calculated with the JCN-JEM are similar to those reported as for the SEG-JEM (data not shown). Analyses lagged by 20 years returned no substantially different results (data not shown).

Table 8 reports on Cox regression model on lung cancer mortality that takes the cumulative exposures to cobalt, nickel, tungsten, respirable, and inhalable dust simultaneously into account applying low backward extrapolation based on SEG-JEM with no lagging of exposures. No exposure effects were identified. The global P value (joint test of cumulative exposures) was 0.37. Analyses on cumulative exposures based on the JCN-JEM but applying the low backwards extrapolation returned very similar results (data not shown).

Cox Regression Model: Hazard Ratios for Lung Cancer (ICD-9:162) and 5 Cumulative Exposure in mg/m3· years

Table 9 reports on Cox model analyzing lung cancer mortality that simultaneously took account of the five long-term average concentration variables (cobalt, nickel, tungsten, respirable dust, inhalable dust), applying the low backward extrapolation based on SEG-JEM with no lagging of exposure. No effects of long-term mean exposures were described and global P value was 0.89.

Cox Regression Model: Hazard Ratios for Lung Cancer (ICD-9:162) and 5 Mean Concentration in mg/m3

The estimated effects that are based on exposure calculation with JCN-JEM are similar to those reported as for the SEG-JEM; occasionally, a significant upward trend was shown (eg, for inhalable dust), but this could not be substantiated in other analyses.


The present cohort study including former and current workers in three German plants investigated the exposure to hardmetal (based on tungsten carbide and cobalt, WC+Co) and its potential impact on overall and cause-specific mortality. We found elevated SMRs for all-cause mortality (SMR = 1.16; 95% CI: 1.08 to 1.24), for heart diseases (SMR = 1.56; 95% CI: 1.35 to 1.81) and nonmalignant respiratory diseases (SMR = 1.56; 95% CI: 1.18 to 2.02), but no excess for lung cancer (SMR = 0.93; 95% CI: 0.68 to 1.25). Internal comparisons by means of Cox regression models did not show increased risks for any endpoint in relation to cumulative exposure or long-term average concentration when evaluating the five exposure agents separately or combined (cobalt, nickel, tungsten, respirable, and inhalable dust).

No increased risks for lung cancer were found in both external and internal comparisons. This finding was robust and seen in almost all subgroup analyses. We note that lung cancer SMRs showed a deficit in the Bavarian cohort on 15 deaths, more pronounced than in the full cohort but with higher imprecision (SMR = 0.61; 95% CI: 0.34 to 1.01).

The SMRs for all heart diseases were statistically significantly elevated by about 50% in the full analytical cohort; the excess was observed among men as well as in the female subcohort (with almost a doubling of risk). Ischemic heart diseases showed similar excesses, also statistically significant despite a lower number of cases. This excess in heart diseases mortality is the major reason for the significantly elevated all-cause mortality in external comparisons. Poisson regression analyses showed upward trends of SMRs for all-cause mortality across time since first exposure and time since cessation of exposure, often statistically significant. These trends are probably due to selection effects and not directly linked to exposure because analyses of SMRs across duration of exposure found no positive trends or they returned trend-P values far off from significance, as a healthy worker selection bias is generally likely in comparing occupational to general population. An underestimation of SMRs cannot be ruled out.

Internal analyses (Cox regression modeling) showed no link between hardmetal related exposures and death from heart diseases. No adverse effects of cumulative exposures on total mortality were demonstrated. The only significant effects were seen for duration of exposure and cumulative exposure to respirable dust, but both variables were negatively related to total mortality. These downward trends in mortality are probably a result of longitudinal selection, which is referred to a healthy worker survivor effect.19

Associations of dust exposures and cardiac effects were judged to be “causal” by leading scientific societies based on environmental epidemiological studies and animal experiments.20 Interestingly, the evidence of a link between dust exposures and cardiac diseases is rather scarce in the occupational field21 and exposures to pure carbon dusts seem not to be related to mortality excesses.20

Note that the data on smoking behavior in the Bavarian subcohort, albeit limited, do not indicate any confounding of mortality in comparison with the general population. In internal analyses with respect to lung cancer, however, the smoking behavior could not be adjusted for because no smoking data were available for the observed lung cancer cases. Cox models for all-cause mortality and cancer mortality, and smoking showed a nonsignificant increase of mortality rates due to smoking but did not indicate confounding.

A major limitation of this study is the incomplete mortality follow-up that resulted from data restriction in Germany. Despite of a comprehensive strategy to protect individual information, data from Health Offices, mainly in North-Rhine Westphalia were not obtainable. The study's results on mortality follow-up are compatible with the numbers presented by Zeeb et al22, who reported to have cleared 40% to 61% of the causes of death by contacting health offices in North Rhine-Westphalia in 2003. We expected to achieve higher numbers by including additional data from the Regional Statistical Offices—a supplemental strategy that Zeeb et al22 described as highly successful at the time. However, this approach was no longer feasible because of the data protection law that does not allow transferring this kind of data (“Bundesstatistikgesetz1”).

Due to incomplete cause-of-death determination, we had to focus our analyses on an analytical subcohort of all enumerated workers. Mortality follow-up was started in this analytical subcohort on January 1, 2005, at plant 1 and January 1, 1975, in plants 2 and 3. This restriction introduced a potential selection bias because we could only focus on survivors up to the start of mortality follow-up. In addition, person-years were lost so that the statistical power to detect excess risks was reduced. Fortunately, in the analytical cohort, the standard criteria of follow-up completeness of 95% in vital status tracing and cause-of-death determination were met.19 In the event, we observed a significant excess of deaths in the category “all other causes” (142 observed vs 105.06 expected), which may point to a less successful ascertainment of causes of death in our cohort than usual for the general population (“all other causes” is an unspecific category covering unexplained deaths). We note that this excess was observed among both men and women. In addition, internal modeling returned hazard ratios for German versus non-German workers of about two for total mortality and about four for lung cancer, after adjustment for covariates: this may indicate that the incomplete follow-up may be rather pronounced among non-German workers, who are the minor part of the participants.

However, it is possible that for specific causes of death, the reported number of observed deaths is too small. If we assume that 6% of the deaths are lung cancer deaths (as in the full analytical cohort), the missed number of lung cancers can be estimated as 6%*(142–105.06) = 2.2 cases. We observed 45 lung cancer deaths and had 48.34 expected. Thus, the corrected number of observed lung cancer deaths is 45+2.2 = 47.2 and the corrected lung cancer SMR is still unexceptionable: 47.2/48.34 = 0.98. Thus, it is improbable that the potential underascertainment of causes of death in the cohort masked a true elevation of the lung cancer SMR.

In addition, the study is characterized by insufficient latency time due to a young analytical cohort. The average age at end of follow-up was 54.1 years. Accordingly, the percentage of deceased subjects was only 12.7% and just 6% of the deceased died of lung cancer. Thus, the German component is relatively young to study long-term effects of exposures, in particular on cancer. This problem was particularly pronounced for the young Bavarian cohort. An extended follow-up of this study, in particular of the Bavarian subcohort, is indicated for the future. Furthermore, the precision of the risk estimates is rather low because of small numbers of events of interest. As expected, the German component alone cannot justify reliable conclusion about exposure effects. A pooled analysis within the multinational occupational cohort study was planned in an a priori fashion.

However, there are strengths of this study that are worthwhile to highlight. The assessment of exposure was comprehensive in comparison to former studies. Ascertaining the exposure to hardmetal and other agents was based on a large IH data base and detailed job-histories. Multiple JEMs based on SEGs (SEG-JEM) and JCN (JCN-JEM) were developed to allocate the measurements to individual exposure. In addition to the imputation technique for approximately 28% of hygiene measurements below detection limits and adaptation from respirable to inhalable particle size fractions, multiple exposure metrics were used to estimate personal exposures to the agents of interest. Validation of exposure data was discussed extensively in a former study,23 in which a couple of different exposure assessment scenarios were developed to tackle inherent uncertainties, in particular unclear historical levels of exposures.

In effect, it is encouraging that the different scenarios (SEG-JEM vs JCN-JEM, low or high backward extrapolation) did not lead to much variation in estimates of cumulative exposures and long-term average concentrations. In addition, no substantially different findings in internal analyses on overall mortality and lung cancer mortality were found when applying the varying exposure scenarios. Finally, analyses of available smoking data did not indicate confounding of results by smoking.

Furthermore, we performed both external and internal analyses and explored many model specifications with lagging of exposures up to 20 years, which represents an advantage over former studies.

Even though data restriction due to data protection laws in Germany is limiting the access to important data, it is worth noting that we extensively inquired vital status and causes of death. By combining different approaches to collect data and tracking study participants’ changes of residence with utmost care, we are convinced that we collected as much information as possible.


From the German component of the multinational occupational cohort, lung cancer SMRs were unexceptionable, but heart diseases showed excesses in SMR analyses. No associations could be found in relation to cumulative exposures or long-term mean concentrations (cobalt, nickel, tungsten, respirable, and inhalable dust). This study has limitations due to an incomplete mortality follow-up, insufficient latency time, and a potential survivor selection bias. This limits conclusions about exposure effects. A future follow-up in Germany is indicated.

The conclusions from the present report can only apply to the German component of the industry, whereas a pooled analysis including all components will enable a broad generalization of results.


We like to thank the workers’ council, the occupational health service, the data protection officer, and the company experts who supported this study with their engagement.

The IFA-Institute (Institut für Arbeitsschutz der Deutschen Gesetzlichen Unfallversicherung) would like to thank for their expertise on measurement systems and analyses. For supporting the study with measurement data and urinary data of plant 3, we like to thank Professor Thomas Kraus, TH Aachen. We thank Prof Maria Blettner, Mainz University, for helpful discussions on techniques of vital status tracing and cause-of-death determination in Germany. We like to thank the staff at the German registry and health offices who supported the study with diligent work on ascertaining information on our study individuals. In addition, we like to thank the ITIA for sponsoring this study.

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 representatives from the ITIA and its member companies. In particular, we thank all the employees who provided assistance throughout the study.


1. World Health Organization International Agency for Research on Cancer. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans. Cobalt in Hard-metals and Cobalt Sulfate, Gallium Arsenide, Indium Phosphide and Vanadium Pentoxide. Vol. 86. Lyon: IARC; 2006. Available at: Accessed January 9, 2017.
2. Hogstedt C, Alexandersson R. Mortality among hard metal workers. Arbete Hälsa 1990; 21:1–26.
3. Lasfargues G, Wild P, Moulin JJ, et al. Lung cancer mortality in a French cohort of hard-metal workers. Am J Ind Med 1994; 26:585–595.
4. Moulin JJ, Wild P, Romazini S, et al. Lung cancer risk in hard-metal workers. Am J Epidemiol 1998; 148:241–248.
5. Wild P, Perdrix A, Romazini S, Moulin J-J, Pellet F. Lung cancer mortality in a site producing hard metals. Occup Environ Med 2000; 57:568–573.
6. National Toxicology Program. Report on Carcinogens Monograph on Cobalt and Cobalt Compounds That Release Cobalt Ions In Vivo. Office of the Report on Carcinogens, Division of the National Toxicology Program. Research Triangle Park, North Carolina: National Institute of Environmental Health Sciences, U.S. Department of Health and Human Services; 2016.
7. Wild P, Bourgkard E, Paris C. Lung cancer and exposure to metals: the epidemiological evidence. Methods Mol Biol 2009; 472:139–167.
8. Marsh G, Buchanich J, Zimmerman S, et al. An international historical cohort study of workers in the hard-metal industry: mid-study epidemiology update. Occup Environ Med 2014; 71 (suppl 1):A96–A97.
9. Kennedy K, Esmen N, Marsh G, et al. An international historical cohort study of workers in the hard-metal industry: exposure assessment. Occup Environ Med 2014; 71 (suppl 1):A65.
10. Ignacio JS, Bullock WH. A Strategy for Assessing and Managing Occupational Exposures. 3rd ed.Association IHA: AIHA, Fairfax; 2006.
11. Creely KS, Cowie H, Van Tongeren M, Kromhout H, Tickner J, Cherrie JW. Trends in inhalation exposure: a review of the data in the published scientific literature. Ann Occup Hygiene 2007; 51:665–678.
12. Morfeld P, Taeger D, Mitura H, et al. Cross-sectional study on respiratory morbidity in workers after exposure to synthetic amorphous silica at five German production plants: exposure assessment and exposure estimates. J Occup Environ Med 2014; 56:72–78.
13. Kraus T, Schramel P, Schaller KH, Zobelein P, Weber A, Angerer J. Exposure assessment in the hard metal manufacturing industry with special regard to tungsten and its compounds. Occup Environ Med 2001; 58:631–634.
14. Breslow NE, Day NE. Statistical Methods in Cancer Research. Volume II-The Design and Analysis of Cohort Studies. IARC Scientific Publications. No. 82 ed. Lyon: International Agency for Research on Cancer; 1987. Available at: Accessed January 9, 2017.
15. Cox D. Regression models and life tables (with discussion). J Royal Stat Soc 1972; 34:187–220.
16. Cox D. Partial likelihood. Biometrika 1975; 62:269–276.
17. StataCorp. Release 13. Statistical Software. College Station, TX: StataCorp LP; 2014.
18. Rothman KJ. No adjustments are needed for multiple comparisons. Epidemiology 1990; 1:43–46.
19. Checkoway H, Pearce N, Kriebel D. Research Methods in Occupational Epidemiology. 2nd ed. Vol. 34. Oxford: Oxford University Press; 2004.
20. Morfeld P, Mundt K, Dell L, Sorahan T, McCunney R. Meta-analysis of cardiac mortality in three cohorts of carbon black production workers. Int J Environ Res Public Health 2016; 13:302.
21. McElvenny DM, Miller BG, MacCalman LA, et al. Mortality of a cohort of workers in Great Britain with blood lead measurements. Occup Environ Med 2015; 72:625–632.
22. Zeeb H, Klug SJ, Ahrens W, Straif K. Follow-up study in Germany: Further experience and novel development in inquiring causes of death. Informatics, Biometry and Epidemiology in Medicine and Biology 2003; 34:612–618.
23. Morfeld P, Mundt KA, Taeger D, Guldner K, Steinig O, Miller BG. Threshold value estimation for respirable quartz dust exposure and silicosis incidence among workers in the German porcelain industry. J Occup Environ Med 2013; 55:1027–1034.

1 Reference:, accessed September, 16, 2016.
Cited Here

Copyright © 2017 by the American College of Occupational and Environmental Medicine