Estimates of cumulative coal mine dust exposure for each mine were previously developed as the summation of the products of each job-specific dust concentration and the duration of time worked at that job.²² For this study, we constructed estimates of respirable quartz dust exposure by using Mine Safety and Health Administration compliance data from 1982 to 2002. Data were obtained from Mine Safety and Health Administration under the Freedom of Information Act and estimates were compiled with technical assistance from the National Institute for Occupational Safety and Health Division of Respiratory Disease Studies. We determined the mean per cent silica within each of the five regions for each Lainhart job grouping.²³ Lainhart groupings include the following job categories: underground face, transportation, maintenance and miscellaneous activities; and, surface transportation, maintenance, tipple, strip mining and miscellaneous activities. We calculated job-specific silica estimates by multiplying the percentage of respirable silica for each region and Lainhart job category by the mean coal mine dust concentration estimate for each job within the region. We then assigned an individual cumulative silica exposure estimate to each study participant by summing the product of the number of years at each job by the corresponding region-specific and job group-specific silica exposure estimate.

We determined the vital status of this cohort through 31 December 2007. We submitted records of the 5636 participants alive when the previous follow-up concluded (31 December 1993) to the Social Security Administration to determine vital status. We then submitted the 3081 records identified as deceased or unknown to the National Death Index to confirm vital status and obtain cause of death.

We sought termination date for employment as a coal miner for cohort members from the United Mine Workers of America Health and Retirement Funds (UMWA Funds) and the Department of Labor (DOL), Office of Workers’ Compensation Programs. When conflicting dates were provided, we considered the DOL date more reliable. We coded any termination date greater than 1 year after death as missing.

We conducted modified Life Table Analysis using the National Institute for Occupational Safety and Health Life Table Analysis System (LTAS) mortality programme.²⁴ The LTAS calculates person-days for follow-up time from the date of enrolment in the study until date of death or end of follow-up. Expected deaths were calculated by applying the age-specific, race-specific and calendar year-specific mortality rates for all US men to the person-years in each corresponding group of the study population. SMRs and 95% CIs were calculated as the ratio of the observed to the expected numbers of deaths for all deaths due to pneumoconiosis, COPD and cancers of the trachea, bronchus and lung (henceforth, ‘lung cancer’). Underlying cause of death was assigned by LTAS using the International Classification of Diseases revisions 8 through 10, according to the year of death (for specific codes, see online supplemental table I). We also calculated SMRs for these outcomes stratified by coal-rank region, race, chest X-ray status at enrolment and calendar year.

Of the 9033 participants for whom vital status was determined at the previous follow-up, we included 8829 (97.7%) in the current study. We excluded 137 (1.5%) participants due to missing or invalid data and 67 (0.7%) because they were lost to follow-up.

The mean age at enrolment was 45 years (range 17–69). Most participants were either current (54%) or former (26%) smokers (table 1). The average tenure in coal mining was 21 years, with an average of 9 years working at the coal face. The mean cumulative exposure was 64.6 mg/m³-years (range 0.1–346.9 mg/m³ years) for coal mine dust and 2.6 mg/m³-years (range 0.5–14.2 mg/m³-years) for respirable silica. The correlation between coal mine dust and silica exposures was significant but only moderately strong (Pearson’s correlation coefficient=0.29, p<0.01).

By the end of follow-up, 5907 (67%) participants had died. There were 399 deaths attributable to CWP pneumoconiosis and 4 to silicosis (analysed as 403 deaths due to pneumoconiosis); 309 to COPD; and 568 to lung cancer as the underlying cause of death (table 2).

We observed excess mortality from COPD, (SMR=1.11; 95% CI 0.99 to 1.24) and pneumoconiosis-related mortality (SMR=79.70; 95% CI 72.1 to 87.67; table 2). SMRs increased monotonically by calendar year of death for pneumoconiosis and lung cancer and for all outcomes were positive and significant only in the more recent follow-up periods. Concerning regional distribution, for COPD significant excess mortality was seen only in the Midwest (SMR=1.44; 95% CI 1.07 to 1.88). In contrast, SMRs for pneumoconiosis were lowest in the Midwest and dramatically higher in the highest coal-rank region, eastern Pennsylvania, which also had the highest mean cumulative coal mine dust values.

We saw a borderline significant excess mortality due to lung cancer (SMR=1.08; 95% CI 1.00 to 1.18, p=0.06 table 2). Lung cancer SMRs varied by region, with higher risk in western Appalachia and the Midwest and lower risk in the West. These findings may be partially confounded by smoking, the single strongest known population risk factor for lung cancer, as the largest proportion of ever smokers was in the Midwest (86.7%) and a lower proportion was in the West (81.7%). However, the percentage of ever smokers in eastern Appalachia (82.3%), where we saw no excess lung cancer mortality, was equivalent to that in western Appalachia (82.4%), where we did see an excess. We observed a significant excess of lung cancer mortality (SMR=1.13; 95% CI 1.04 to 1.24) among workers who had no evidence of pneumoconiosis on their radiograph at baseline and an apparent decreasing trend in SMRs for lung cancer with increasing severity of pneumoconiosis (from chest X-ray) at the start of follow-up.

We observed positive exposure-response relationships between cumulative coal mine dust exposure and COPD among never and former smokers but not among current smokers, which is consistent with the latest findings from the PFR and a previous study of pulmonary function in the NSCWP.⁶³⁰ Our finding of a relationship between coal mine dust and COPD mortality among former and never smokers may be due in part to selection bias, that is, smokers with more severe lung deficits may have stopped smoking or may have left coal mining earlier in their work life than non-smoking workers of the same age, and hence introduce negative bias. We observed a clear healthy worker effect in the SMR analysis for all outcomes. For COPD and lung cancer related mortality we saw a deficit in overall mortality in the first two decades of follow-up, which disappeared in subsequent decades. It has been proposed that there may be a negative interaction between coal mine dust exposure and COPD among smokers attributable to a number of theories including that tobacco smoke induced mucous hypersecretion protecting the distal lung from dust deposition.³¹ However there is little supporting evidence for this theory.

The strong positive exposure-response relationship between coal mine dust and pneumoconiosis-related mortality that we observed is consistent with previous studies from this and other coal miner cohort studies.⁶⁷¹¹ The strength of that relationship varied by region, generally declining with coal rank, that is, from east to west in the US coal-rank region has previously been observed as a strong risk factor for pneumoconiosis-related mortality in this cohort.⁷

Our finding that silica exposure is significantly associated with mortality from COPD in single exposure models dust (HR 1.08; 95% CI 1.00 to 1.17) is consistent with those from the latest mortality follow-up from the PFR study.⁶ These findings are generally supportive of the accumulating evidence of the importance of silica exposure in the excess of obstructive lung disease among coal miners, and in the increasing prevalence and severity of pneumoconiosis in US miners.³²³³ Chronic exposure to silica dust has been shown to cause chronic bronchitis and emphysema even in the absence of radiological silicosis.³⁴ However autopsy studies of US and Australian coal miners found quantity of coal mine dust, but not silica, significantly predicted emphysema severity after accounting for cigarette smoking and age.³¹³⁵ In our study the effect of silica exposure on COPD was weakened when we controlled for coal mine (HR 1.04; 95% CI 0.94 to 1.94).

Our silica estimates were based on data from 1982 to 2002, and more removed in time from exposure than the data used to develop our coal mine dust estimates. There is therefore more uncertainly and potential for non-differential misclassification bias in the estimates of disease occurrence associated with the silica exposure data compared with those associated with the coal mine dust exposure. Hence the impact of silica exposure may be underestimated in our study.

The negative coefficients for TSLE, which we observed in all models, reflect an increased mortality rate in the years immediately after leaving employment. The tendency for mortality rates to rise in the time period following leaving work has been described by Steenland et al,³⁶ who proposed that such a trend likely reflects aspects of selection due to the healthy worker survivor effect. Controlling for TSLE significantly improved the model fit and increased the estimate of the exposure-response associations for nearly all models. While controlling for TSLE likely reduced some bias from the healthy worker survivor effect,²⁰³⁶ traditional epidemiological methods, as used here, may not fully control for the healthy worker survivor effect.³⁷

Many earlier studies have reported a deficit in lung cancer mortality among coal miners.^9–111416 In sharp contrast to other studies, we observed an overall excess of lung cancer mortality (SMR=1.08; 95% CI 1.00 to 1.18) and a significant association with cumulative coal mine dust exposure and lung cancer. This association with coal mine dust was present in models with and without silica exposure.

We did not see an overall significant association between silica exposure and lung cancer mortality, which was surprising as inhaled crystalline silica is classified as a Group 1 lung carcinogen by The International Agency for Research on Cancer.³⁸ In the most recent follow-up of the British cohort, an overall significant exposure-response relationship was observed for lung cancer mortality with silica but not coal mine dust exposure.⁶ The PFR had more robust exposure estimates and a longer follow-up time than our study (1959–2005 and 1969–2007, respectively). In the most recent follow-up period, 2000–2007, we saw a positive and significant association between lung cancer mortality and coal mine dust exposure and a positive association with silica exposure. This may be attributable to the long latency of lung cancer and/or the mitigation of the healthy worker effect over time.

We observed an inverse trend with lung cancer mortality and radiographic CWP status at study enrolment. Similar trends have been reported by some other studies.⁹³⁹ While CWP may be considered a surrogate marker for coal mine dust exposure, in the current study the findings for CWP are likely attributable to selection biases. Selection bias may have occurred in this cohort because it initially included actively employed miners, many of whom were well into their work experience at the start of this study (average 21 years). Negative bias may have occurred if miners who had higher dust exposures and were at increased risk of lung cancer were more likely to leave work before other men their age, and thus would not be eligible for enrolment in the study. Another possible explanation is that the latency period for lung cancer is longer than that for CWP, so miners may die with pneumoconiosis before lung cancer could be diagnosed.

The analysis of TSLE and lung cancer mortality provided some insights into how mortality rates vary by work status. The risk of death from lung cancer associated with coal mine dust declined with time after leaving work. The same pattern was seen for the association between coal mine dust and gastrointestinal cancer mortality (data not shown). Hence, the observations may reflect the tendency for mortality rates to rise in the time period following leaving work.

Early studies have previously observed a deficit of lung cancer mortality among coal miners.^9–1114 These observations led to a theory that clay particles may partially detoxify silica by coating it, potentially rendering it less carcinogenic. Our findings, and those of the British coal miners’ cohort, suggest that the deficit might be simply explained by a strong healthy worker effect that has diminished with extended follow-up. It also might be attributable to the prohibition of smoking in underground mines which may also diminish with time as miners leave work,⁶ although there is no empirical evidence for or against this hypothesis.

Smoking is the strongest risk factor for lung cancer in the general population. US male smokers are about 23 times more likely to develop lung cancer than non-smokers.⁴⁰ In our models all measures of smoking at enrolment were strong predictors of lung cancer mortality. Compared with the general US male population in 1970, the miners in our study were more likely to be current smokers, but on average smoked fewer cigarettes per day (see online supplemental table IV), consistent with the prohibition on smoking in the mines. Whatever other factors may contribute to lung cancer incidence and mortality in this population, interventions to reduce smoking and smoking initiation among miners and their communities should be investigated.

There were a number of limitations to our study. Perhaps of greatest concern was the lack of job histories on which to base exposure estimates after study enrolment. For previous studies of this cohort, it was assumed that most exposure was accumulated before enrolment.⁷ Our sensitivity analysis found no meaningful differences between the exposure-response relationships for death from NMRD using the original and extended estimates for either coal mine dust or silica exposure, lending empirical support to our assumptions. A limitation of the sensitivity analysis was the assumption that miners continued to work in the same job and the same level of exposure after study enrolment until termination of employment. We have no empirical way to test this assumption. However, our estimates were likely exaggerated as dust levels in US mines declined dramatically after the passage of the 1969 Mine Health and Safety Act. Thus it is unlikely that the lack of work histories in this study had a large impact on our findings.

The electronic data archive from this cohort did not include the start and end dates for each job. As with previous mortality studies from of this cohort, this prevented the evaluation of exposure lags and construction of internal time-varying variables (through enrolment) in the Cox analyses.

Cause of death misclassification is a source of potential bias in occupational studies.⁴¹ Of particular concern for our study, using underlying cause of death from death certificates may preferentially underestimate COPD.⁴²⁴³ Among coal miners is it likely that NMRD may be preferentially classified as CWP, hence COPD would be under-reported. In the general US population the accuracy of reporting lung cancer on death certificates is high (94%).⁴⁴ However, it is possible that lung cancer mortality in coal miners may be underdiagnosed in the presence of other radiographic findings.⁴⁵⁴⁶

In addition to silica, coal miners may be exposed to other substances that are recognised to cause respiratory cancer in humans including diesel exhaust⁴⁷⁴⁸ and radon gas.⁴⁹ Diesel exposure is unlikely to have contributed significant to lung cancer related mortality in this cohort as many of the miners worked much of their mining tenure in coal mines with little or no diesel use. Subsequently some of these miners were used as non-exposure controls for a study of diesel exhaust exposure and respiratory health outcomes among coal miners.⁵⁰ There is no information regarding radon exposure in the study mines and it is unclear to what extent, if any, radon exposure may vary with coal or silica exposure.