Mundt, Kenneth A PhD; Birk, Thomas Dipl rer soc; Parsons, William MS; Borsch-Galetke, Elisabeth PhD, MD; Siegmund, Klaus MD; Heavner, Karyn PhD; Guldner, Karlheinz PhD
The causal association is well established between occupational respirable crystalline silica exposure and silicosis—a specific type of pneumoconiosis. This conclusion is supported by many epidemiological studies of workers historically heavily exposed to respirable crystalline silica in several industries including mining, quarrying, and potteries.1 In addition to quantity of exposure, risks appear to be greater with freshly fractionated quartz and specific crystalline polymorphs such as crystoballite.2 Silicosis appears not to occur among workers exposed only to ambient or low concentrations of respirable crystalline silica; however, it remains unknown at which specific concentrations and durations of occupational exposure to respirable crystalline silica risk of silicosis is increased.
Occupational exposure to respirable crystalline silica also has been associated with increased risk of lung cancers,3 especially among individuals with silicosis.4 In 1997, the International Agency for Research on Cancer (IARC)5 found “sufficient evidence” in both humans and in animal studies to classify occupational exposure to respirable crystalline silica in the form of quartz or crystoballite as a known (ie, a Group 1) human carcinogen. This finding was recently reiterated in IARC's Monograph 100, Part C review.6 However, as noted by IARC, respirable crystalline silica's carcinogenicity is not evident under all industrial exposure circumstances. That lung cancer risk is mostly observed among individuals previously diagnosed with silicosis raises a basic question of whether respirable crystalline silica causes lung cancer in the absence of silicosis7 or apart from the heavy exposures most strongly associated with silicosis. The scientific debate continues8–10 and newer research reports and reviews have generated mixed conclusions.11–19
Consequently, a key regulatory question remains as to whether silicosis and lung cancer are prevented at current occupational exposure limits (OELs), which vary by country. In Germany, the OEL had been 0.15 mg/m3 until a change in regulations suspended this OEL in favor of a health-based OEL (to be determined). In the US and the UK, the OEL is 0.10 mg/m3 and 0.05 mg/m3 in Denmark, Spain, and Sweden.20
The goal of this article is to quantitatively evaluate risks of silicosis morbidity and lung cancer mortality by individually estimated cumulative and average respirable crystalline silica exposures among a cohort of nearly 18,000 German porcelain manufacturing workers previously followed for mortality and evaluated using basic standardized mortality ratio (SMR) analysis.21 Using German national lung cancer mortality rates as the reference, no excess lung cancer risk was seen: SMR = 0.71 (95% confidence interval [CI], 0.56 to 0.89) based on 74 observed lung cancer deaths among men; and SMR = 0.72 (95% CI, 0.44 to 1.12) based on 20 observed lung cancer deaths among women.21 For the subcohort from Bavaria—the region where most of the porcelain factories were located—SMRs were closer to unity: SMR = 0.98 (95% CI, 0.75 to 1.27) based on 59 lung cancer deaths among men; and SMR = 0.91 (95% CI, 0.52 to 1.48) based on 16 lung cancer deaths among women. These results provided no evidence of any excess risk of lung cancer among the German porcelain workers. On the other hand, silicosis mortality was significantly elevated (SMR = 11.37; 95% CI, 3.66 to 26.53) but was based on only 5 silicosis deaths, all of which were among men.
These preliminary results, however, were not based on respirable crystalline silica exposures quantified at the individual level. Extensive industrial hygiene exposure data for this cohort have been obtained and used to derive quantitative exposure estimates for each cohort member.22 Furthermore, evaluation of a large archive of radiographic examinations obtained as part of a medical surveillance program for German porcelain workers including a substantial proportion of this cohort facilitated radiographic determination of silicosis morbidity. Combining the mortality data, the silicosis morbidity data, and the quantitative exposure assessment allows quantitative evaluation of potential relationships between estimates of respirable crystalline silica exposure and silicosis morbidity and lung cancer mortality. This occupational setting, that is, the Post–World War II German porcelain industry, presents an unusual opportunity to evaluate exposure–response relationships for silicosis and lung cancer at exposure levels ranging from near to well above current OELs. In addition, this study allows evaluation of these risks among women, who comprise half of this cohort.
The study population and cohort definition have been described previously.21 Briefly, workers employed by any of more than 100 porcelain-manufacturing plants in the western states of Germany (ie, the former West Germany) and participating in an industry-wide preventive medical screening program for silicosis between January 1, 1985, and December 31, 1987, were eligible. Medical screening results data were maintained electronically or on paper by the Berufsgenossenschaft der keramischen und Glas-Industrie (BGGK), whereas only paper work history records were maintained by the BGGK. On the basis of available records, we defined the eligible cohort as those employees for whom detailed work histories were available (n = 18,000). Exclusion of individuals employed for a total of less than 6 months in the porcelain industry resulted in a final analytical cohort of 17,644 employees, which we followed for mortality and silicosis morbidity through 2005.
Cause of Death Determination
Vital status and cause of death were determined using information from various sources: the BGGK; health insurance and pension fund records; written enquiries to community registration offices for residential histories and last known residence; and the central population registry for Bavaria. Vital status ultimately was determined for 94% of the cohort, with 1610 (9%) reported to be deceased as of the end of follow-up. Subjects with unknown vital status at the end of the follow-up period were censored, that is, they contributed person time only up to the date last known to be alive.
The underlying cause of death for each decedent was obtained from the official death certificate, which in Germany usually is stored at the community health department in the town or city where the death occurred. However, minimum record retention times differ from state to state in Germany, ranging, for example, from 30 years in Bavaria to only a few years in other states such as Rhineland-Palatinate. Therefore, cause of death determination based on death certificates was slightly more complete for the Bavarian subcohort (95%) compared with the overall cohort (93%). Underlying cause of death was coded by a professional nosologist according to the 10th revision of the International Classification of Diseases (ICD-10).
Chest radiograph (x-ray) is the typical method for detecting silicosis. The International Labour Organization (ILO) Classification System for interpreting and classifying radiographs23,24 is commonly used for epidemiological research. More than 120,000 radiographs since the early 1950s were available on the study population from the preventive medical surveillance program. Medical data have been stored electronically since the mid-1980s, and earlier paper records have also been entered into the electronic database. Most of the follow-up examinations were conducted by the BGGK using mobile radiographic units, and films were read by physicians specially trained in reading these smaller radiographs. Most of original films were available from the BGGK archive, with a mixture of small- to full-sized formats.
The original radiographs had been read over many years by different physician readers, and therefore interreader variability was expected to be large. Known problems resulting from large inter- and intrareader variability include poor sensitivity and specificity of silicosis determination, especially when the signs of silicosis are low-grade.9 Furthermore, because the original radiographic readings were intended for prevention (ie, early detection of radiographic changes consistent with pneumoconiosis and silicosis specifically) as well as administrative purposes (ie, workers' compensation)—and not for etiological research, per se—we performed a rigorous two-stage standardized rereading exercise.
All radiographic rereadings were performed by specially trained teams of two radiologists certified in the classification of chest radiographs for pneumoconiosis according to ILO 2000, comparable to the B-reader designation in the United States. In addition, readers were required to have experience reading small format radiographs generated by the mobile radiographic units. Rereading was performed blinded to the original BGGK physicians' ratings as well as to the second rereader's interpretation of the same films. Where rereadings disagreed, the two readers were asked to discuss and obtain a consensus reading, invoking a third reader if needed to adjudicate any remaining differences.
The first stage of rereading was based on a representative stratified sample (stratified by gender, birth cohort and silicosis status) of 400 cohort members, resulting in the rereading of more than 1600 individual chest radiographs. Overall, the two independent rereadings were in close agreement, and consensus was easily reached for all films. To classify each cohort member as normal or by degree of silicosis using the ILO categories, each rereader examined all available chest radiographs for each subject using the side-by-side method, consistent with the approach most likely used by the original readers. This approach further reduced variability (partly due to film size and quality) leading to a more stable interpretation against which original BGGK classifications could be compared.
Consistent differences were identified between the original BGGK classification and the consensus results, with considerably more films having been classified as “positive” (ie, 1/1 or greater) by the original BG readers than by the study rereaders. On the contrary, no films read by the original readers as clearly negative (ie, 0/0 or 0/1) were classified as positive (ie, 1/1 or greater) by the consensus rereading, indicating a very low false negative rate. Therefore, we were reasonably confident that original readings of 0/0 or 0/1 were unlikely to represent true positive results (on the basis of our rereading criteria). Because a majority of films fell into this category, we were able to avoid rereading these films for which the yield of true positive (1/1 or higher) readings likely would have been very low.
For the second phase of rereading two of us (E.B.G. and K.S.)—both experienced radiographic readers—independently reread all available x-ray films for all 606 cohort members whose most recent radiograph was originally classified by the BGGK as 1/0 or above, using the same methods as in the first phase. For some radiographs, an original BGGK assessment was unavailable, and for some individuals radiographs were unavailable possibly because they have been lost, misfiled, or were in use by the BGGK at the time of our request, leaving 552 (91% of the original sample) for rereadings. In addition, a random sample of radiographs from the first phase was included to verify that the new readers would replicate the results of the first consensus reading (ie, especially no false negative results). Ultimately, more than 4700 radiographs were reread by both rereaders. For this study, we defined silicosis as radiographic evidence of small rounded opacities with a profusion score equal to or greater than 1/1 (according to ILO, 2000) on the basis of the consensus rereading.
Details of the historical exposure assessment have been reported elsewhere.22 In short, more than 8000 combined static (stationary area) and personal total, respirable and silica dust industrial hygiene measurements were available for about 100 discrete production area or job task code combinations beginning in 1954 and covering all years through the end of folow-up. Original sampling and analysis protocols were available for these measurements. Because gravimetric measurements were not available before 1959, we performed laboratory and field exercises to derive conversion factors for particle count measurement results to mass values, and for gravimetric results using obsolete devices to values comparable with post-1975 gravimetric techniques.22
Exposure data were summarized into similar exposure groups or SEGs for production areas such as preparation, forming, drying, firing preparation, firing and finishing, and by calendar year to form a job exposure matrix (JEM). The cells of the JEM were populated by smoothing longitudinal industrial hygiene measurement data for each SEG using LOESS regression (SAS v9.1, SAS Institute, Cary, NC). Estimates for the earliest years with no exposure data (ie, 1938 to 1953) were derived by extrapolating the values backward from the summary exposure measurement data from 1954 through the 1960s.
A detailed employment history record was reconstructed for all cohort members on the basis of official employment records and in part on questionnaire information collected by medical personnel during the industry-wide medical screening program. By linking the work histories with the JEM, we calculated cumulative respirable crystalline silica exposure, average exposure, and 10-year lagged cumulative exposure for each cohort member. Lagging of cumulative exposure was time-dependent and was achieved by disregarding the previous 10-years exposure as of each time increment of follow-up. The employment records also provided necessary information on sex, smoking status as of the date of the medical screening examination, date of birth, date of hire and separation, and documentation of silica exposure outside of the porcelain industry (eg, in a previous job). Potential prior silica exposure was coded as probable, possible, unlikely or unknown by an expert (K.G.) in occupational exposures and familiar with the historical BGGK records.
We used Cox proportional hazards models to evaluate the time-dependent relationships between cumulative and average exposure estimates and silicosis morbidity and lung cancer mortality, allowing control for other time-dependent variables such as duration of employment.25 Because follow-up time differs for mortality and silicosis morbidity, separate estimates of person-time were generated. The date of the first radiograph meeting the silicosis definition (ie, consensus reading of 1/1 or higher on the ILO scale) was recorded as the date of diagnosis. Person-time for each cohort member was accumulated until the later of the date of the last available radiograph or the date of silicosis diagnosis. We used age at end of follow-up as the time scale variable in all Cox models, and additional variables considered included sex, smoking status, age at hire, and duration of employment. Covariates producing 10% or more change in the estimate compared to the bivariate model were retained.26 Analyses were generated both including and excluding cohort members with evidence of prior occupational silica exposure, with and without lagging exposures 10 years and limited to those hired since 1960.
Exposure cutpoints were arbitrarily selected in 0.5 mg/m3-years increments up to 1.5 mg/m3-years, and cumulative exposures above 1.5 mg/m3-years were divided using 3.0 mg/m3-years as an additional cutpoint.22 These values provided adequately large numbers in each group to allow further stratification within categories. More than 40% of the cohort failed to exceed a cumulative exposure of 0.5 mg/m3-years, and just more than a third of the cohort members accumulated exposures greater than 1 mg/m3-years. Roughly, 2000 cohort members fell into each of the highest three exposure groups, with similar numbers of men and women, except in the highest category where there were considerably more men. However, for both silicosis morbidity and lung cancer mortality, the largest number of cases fell into the highest exposure category, and there was no indication of increased risk across the lower exposure categories. We therefore combined the lowest four original categories as the new referent category and subdivided the highest category into the following categories: ≤3.0; >3.0 to 4.0; >4.0 to 5.0; >5.0 to 6.0; and >6.0 mg/m3-years. Average annual exposure (cumulative divided by duration of employment on a time-dependent basis) was categorized into the following groups: ≤0.05; >0.05 to 0.10; >0.10 to 0.15; >0.15 to 0.20; and >0.20 mg/m3. Because an ancillary goal of this analysis was to evaluate the possibility of a threshold effect for silicosis and lung cancer, cumulative exposure was also modeled as a continuous variable using polynomial models of the second, third, and forth degrees. A formal threshold analysis using more sophisticated techniques is beyond the scope of this article.
More than 40% of the cohort accumulated less than 0.5 mg/m3-years, and nearly 70% of the cohort had average exposures (over all working years) less than 0.05 mg/m3 respirable crystalline silica. On the contrary, over a third of the cohort (approximately, 4700 cohort members) accumulated more than 1.5 mg/m3-years, and nearly 10% of the cohort (approximately, 1600 cohort members) had average exposures (over all working years) of more than 0.15 mg/m3 respirable crystalline silica. Similarly, half of those ever working in the materials preparation area accumulated over 3 mg/m3-years, whereas only 12% of those never working in this area accumulated exposure this high.
A total of 1595 deaths (9.2% of the cohort) occurred, 535 (33% of decedents) due to cancers and 94 (17% of all cancers) due to lung cancer. Only 5 deaths were attributed to silicosis, and only 2 of the 10 decedents we determined to have silicosis were identified by death certificate to have died of this disease. A total of 40 cases of silicosis were identified on the basis of radiographic evidence. None of the individuals identified as having silicosis subsequently died of lung cancer, precluding analyses to detect elevated lung cancer risk among silicotics.
Descriptive statistics for the cohort, overall and by lung cancer mortality and silicosis morbidity status, are presented in Table 1. Men represented 79% and 85% of lung cancer deaths and silicosis cases, respectively, compared with 47% of the overall cohort. Risks of both diseases were also higher among those hired before 1960, employed longest and among smokers (much stronger for lung cancer). Lung cancer mortality and silicosis morbidity also were associated with higher categories of cumulative and average exposure, as well as annual exposure ever 0.15 mg/m3 or more, especially for those with silicosis (85% compared with 25.5% of the remainder of the cohort).
Respirable Crystalline Silica Exposure, Lung Cancer, and Other Mortality
Lung cancer mortality among men (n = 74) was not associated with cumulative exposure in adjusted Cox proportional hazards analyses using either the lower or higher cumulative exposure classification scheme (Table 2). Nearly half of the lung cancer deaths occurred among those with more than 3 mg/m3-years and one quarter among those with more than 6 mg/m3-years cumulative exposures; yet, hazard ratios (HRs) for both these categories were essentially unity. Statistical analyses modeling silica exposure as a continuous term (with and without second to fourth order terms) showed no relationship with lung cancer risk (results not shown). Increasing levels of average exposure were not consistently associated with increasing risks of lung cancer, although some of the HRs were inconsistently elevated (Table 2). Excluding individuals with probable prior silica exposure had no clear impact on results (not shown). Risks were clearly high, however, among men known to be smokers at the time of their medical surveillance examination and those with unknown smoking status compared with known never smokers (Table 2).
Among women (n = 20), only six lung cancer deaths occurred among those with the highest cumulative exposure category from the original classification (ie, >3 mg/m3-years), and only one in the lowest category (ie, ≤ 0.5 mg/m3-years), resulting in elevated but highly unstable HRs (CIs spanning from 0.2 to 63). However, no pattern with exposure category was seen (Table 2). Excluding individuals with probable prior exposure to silica did not meaningfully change results (not shown). Hazard ratios were only weakly, and not statistically significantly, elevated for smoking, suggesting that the observed results for women may be unreliable because of uncontrolled confounding by smoking but mainly because of very small numbers. As with men, average exposure was not clearly associated with lung cancer risk among women.
Mortality due to all cancers combined, pancreatic cancer, liver cancer, kidney cancer, cardiovascular disease, digestive diseases, diabetes, renal disease,* and ill-defined conditions were not statistically significantly associated with cumulative silica exposure at any level, stratified by sex and adjusting for age, smoking history, and duration of employment (see Table 1 and Supplemental Digital Content Table A at http://links.lww.com/JOM/A47, Table B at http://links.lww.com/JOM/A48, and Table C at http://links.lww.com/JOM/A49). Smoking was statistically significantly associated with many of these causes of death, as was years of employment, with lower HRs seen among those employed longest.
Respirable Crystalline Silica Exposure and Silicosis
For this study, silicosis status was determined on the basis of rereading of radiographs by two independent readers according to a standard rereading protocol. Exact agreement between radiograph rereaders was reached for approximately 90% of all initial film rereadings. Most disagreements were within one ILO scoring category with similar proportions higher and lower, suggesting only expected random error. After consensus reading, 92.7% of all radiographs were classified below 1/1, and ultimately 89.3% of cohort members were classified as not having silicosis. Compared to the original BGGK classifications, the consensus rereading agreed exactly for 58% of all radiographs, but disagreement was systematically different: for 3% of radiographs, rereaders assigned a higher score, but for 39% the rereading generated a lower score. Among radiographs with BGGK and rereader consensus disagreement, 57% differed by one scoring category and for 43% the discrepancy was two or more categories. A total of 61 cohort members were therefore classified as having radiographic evidence of silicosis on the basis of a reading of 1/1 or higher. We excluded 21 silicotics from the analysis regarding silicosis morbidity, because their diagnosis date fell before (n = 20) or after (n = 1) the study follow-up period; however, their person-time during the study follow-up period (and cause of death, if deceased) were retained for the mortality analyses. Only 6 of the 40 silicosis cases were identified among women; however, four of these cases fell into the two highest cumulative exposure categories (ie, >5 mg/m3-years). Therefore, all analyses combined men and women and statistically controlled for sex.
Silicosis risk was strongly associated with both average and cumulative exposure (Table 3). Adjusted HRs for groups with average exposures greater than 0.15 tended to be large (ranging from about 13 to 23) and statistically significant and exclusion of individuals with probable prior occupational exposure to silica only moderately reduced HRs.
Increased risk of silicosis was observed only among the group with cumulative exposure above 3.0 mg/m3-years on the basis of the initial cutpoints (Table 3). Combining all categories below 3.0 mg/m3-years as the referent, adjusted HRs tended to be consistently elevated for all exposure categories above 4 mg/m3-years. Similar results were obtained when analyses were restricted to cohort members hired only after 1960; however, this reduced the number of exposed cases to eight and eight in the referent group. Lagging exposure by 10 years for the whole cohort produced larger but less precise relative risk estimates; however, all HRs were statistically significant for all categories above 4.0 mg/m3-years. Exploratory models using cumulative exposure as a continuous variable and sequentially incorporating polynomial terms of the second, third, and fourth order (not shown) generated considerable differences in the Akaike's information criterion, or AIC score (605.9; 595.0; 592.4 and 578.1, respectively), suggesting that the relationship between cumulative respirable silica exposure and silica is not linear (all of the models with higher-order terms generated lower AIC scores, indicating a better fit than the model with only a first-order exposure term). Formal threshold analyses may be warranted; however, available numbers of silicosis cases may limit their precision.
This study is the largest to date that evaluates cancer and silicosis risks among porcelain workers, and one of only a few with substantial industrial hygiene data on which quantitative exposure–response analyses may be based. Other studies have considered respirable crystalline silica exposures in the ceramics industry, but none focus on the porcelain sector. We focused on the porcelain sector because (1) exposures to respirable crystalline silica were common and relatively well-documented; (2) the health surveillance program, including periodic chest radiographs, was nearly comprehensive; and (3) technological developments and work processes are highly comparable across plants over time.22
The majority of results using age, sex, and smoking-adjusted Cox proportional hazards models demonstrated no association between any major category of cause of death and cumulative respirable crystalline silica exposure levels. These results are consistent with previous findings of no important excesses of these causes of death based on SMR analyses.21 One exception noted earlier was renal disease mortality among men, based on four observed deaths and only two deaths in the referent category. While probably due to chance (there were no other statistically significant results among the nearly 100 stratum-specific HRs calculated), other investigators have reported associations between respirable crystalline silica exposure and renal disease morbidity and mortality.27–33
Consistent with the previously reported SMR results,21 lung cancer mortality results by cumulative and average respirable crystalline silica exposure provided no support for an association with lung cancer risk at exposure levels prevalent in the German porcelain industry during the study period. The only exception was a statistically significantly elevated HR for the subgroups with average exposure of less than 0.05 to 0.1 and less than 0.15 to 0.2; however, there was no apparent trend with increasing exposure. Whether this finding reflects some actual risk, residual confounding or random measurement error is not clear. With a study size of approximately 18,000 and 132 expected lung cancer deaths (from the SMR analyses), the statistical power was more than 99% to detect (at the α = 0.05 level) a relative risk for lung cancer of at least 1.5, and 80% power to detect a relative risk as small as 1.2. In addition, we were able to adjust for smoking on about 70% of the cohort, for which there was evidence of confounding by smoking. Among those with known smoking status, approximately two-thirds were smokers. The significantly increased lung cancer risk among those with unknown smoking status indicates that some proportion of these workers were smokers. Finally, the strong and statistically significant association seen between respirable crystalline silica exposure and silicosis underscores that the lack of a positive finding for lung cancer cannot be explained by an invalid exposure assessment.
None of the individuals identified as having silicosis subsequently died of lung cancer (with approximately 1.7 expected on the basis of the overall rate in this cohort), precluding any statistical analyses of lung cancer risk among silicotics. However, most (75%) cohort members with silicosis were still alive as of the end of follow-up. Conversely, the number of lung cancer decedents expected to have silicosis (on the basis of the overall rate in this cohort) is less than one. Because of the modest number of silicotics in this study, the fact that 75% of those with silicosis were still alive at the end of follow-up, and the relative rarity of lung cancer (17% of all deaths in this cohort), no conclusion can be made regarding lung cancer risk among silicotics. Some,34–38 but not all,7 studies have advanced the hypothesis that lung cancer risk may be increased only in association with silicosis.
The study that is the most similar to ours is that published by Cherry et al39 on the British potteries, which included porcelain workers among ceramics workers. Lung cancer (SMR = 1.91; 95% CI, 1.48 to 2.42) was significantly elevated compared with national rates (England and Wales) but not when regional reference rates were used (SMR = 1.28; 95% CI, 0.99 to 1.62). A JEM was developed for 12 job groups by 10-year period beginning in 1930.40 Exposure estimates were based on 390 area air samples (1950s-1960s) evaluated as particle counts, and 1000 personal air samples evaluated as gravimetric silica mass (late 1960s and later). Smoking information was extracted from medical records from the employment period. Average concentration of respirable crystalline silica (0.1 mg/m3) was significantly associated with lung cancer risk after adjustment for smoking (OR = 1.67; 95% CI, 1.13 to 2.47) but not with cumulative exposure (OR = 1.01; 95% CI, 0.85 to 1.19).39
Silicosis in the British potteries study was defined as radiographic evidence of small rounded opacities compatible with an ILO (1980) score equal to or greater than 1/0 (vs 1/1 in our study). Radiographs were obtained approximately every 4 years, but apparently assessed by only one reader. No rereading of radiographs was performed. Sixty-four cases were statistically evaluated using a nested case-control approach, and a dose–response relationship for cumulative (OR = 1.38 per 1 mg/m3-year; 95% CI, 1.24 to 1.53) and average silica exposure (OR = 2.69 per 0.1 mg/m3; 95% CI, 1.96 to 3.70) were reported. Exposure estimates were not time-dependent, but reflected values at the end of follow-up. Because of these and other differences between the British potteries study and the German porcelain workers study, direct comparison of specific findings may not be valid.
In a pooled analysis of 10 cohorts with 66,000 workers (two-thirds from mining populations and two-thirds from the Chinese study) exposed to respirable crystalline silica, a significant positive exposure–response for lung cancer was reported on the basis of more than 1000 lung cancer cases.14 Lung cancer rate ratios (RR) were moderately increased in cumulative exposure categories above 2 mg/m3-years (RRs of 1.3, 1.5, and 1.6 for 2.0 to 5.4, 5.4 to 12.8, and more than 12.8 mg/m3-years, respectively). Smoking and silicosis status information was not available. The overall SMR of 1.2 (95% CI, 1.1 to 1.3) and the increased RRs in the categorical analysis led the authors conclude that respirable crystalline silica is likely a rather weak carcinogen.14
A meta-analysis of 10 studies examining silica exposure and lung cancer risks reported a dose–response relationship.17 Considerable heterogeneity across studies was noted, with two studies presenting sharp increases in lung cancer risk at comparatively low cumulative exposure levels (ie, <2 mg/m3-years),18,41 in contrast to the remaining eight studies, some of which demonstrated no increase in risk with higher cumulative exposures.13,42–44 Furthermore, only 2 of the 10 studies43,44 appear to have intentionally excluded individuals with silicosis. Other methodological criticisms of this meta-analysis have been raised.45 Erren et al19 evaluated lung cancer risk among nonsilicotics and discussed methodological limitations of both epidemiological and toxicological approaches. Nevertheless, this meta-analysis identified three studies that controlled for smoking, estimating a meta-relative risk of 1.0 (95% CI, 0.8 to 1.3); however, for eight remaining studies, the meta-relative risk was slighted but statistically significantly elevated (RR = 1.2; 95% CI, 1.1 to 1.4) with significant heterogeneity noted.19
In contrast with the lack of support for an association between respirable crystalline silica exposure and lung cancer risk, our study demonstrated clearly and statistically significantly increased risk of radiographic evidence of silicosis (ILO 1/1 or higher) in the highest categories of both average (generally above 0.15 mg/m3) and cumulative (generally >4.0 mg/m3-years) exposure, controlling for age, sex, smoking, and duration of employment. Although no formal threshold analyses were performed, this study generated considerable support for a threshold.
Although smoking is believed not to contribute to silicosis risk, our study generated a nonstatistically significant sex-adjusted HR of 1.7, for smoking. On the contrary, those with unknown smoking status—believed to include some proportion of smokers on the basis of the lung cancer results—generated an HR less than unity. Therefore, the elevated HR likely reflects overestimation of silicosis risk among smokers. Duration of employment was inversely associated with silicosis risk. This might have been expected, given the cross-sectional definition of the cohort, that is, included were those actively employed and participating in the medical surveillance program during 1985 to 1987, regardless of when they were first employed in the porcelain industry. Those first employed in the 1940s and 1950s had to remain employed (or returned to employment) and survived until 1985 to be included. Thus, there is evidence of a survivor bias, but this is potentially addressed by controlling for duration of employment. Analyses limited to cohort members first employed since 1960 resulted in the loss of 60% (n = 24) silicosis cases. Further follow-up of the 14,026 employees hired since 1960 will be important, as there may not have been sufficient follow-up time to observe all cases of diseases with long latencies including lung cancers and silicosis.
Study Strengths and Weaknesses
Apart from this study's large size and inclusion of nearly equal proportions of men and women, one of its main strengths was the extent and quality of the available exposure data, allowing quantitative exposure–response evaluation of cause-specific mortality and silicosis morbidity. Workers' exposures were estimated on the basis of a JEM created from approximately 8000 personal and area industrial hygiene measurements.22 To include older data obtained using obsolete measurement devices, we conducted special exercises using historic equipment side by side with modern industrial hygiene devices. Series of measurements were made in both controlled dust tunnel and actual porcelain material preparation settings and were compared to derive conversion factors for data originally based on the historical measures. Because no exposure measurement data were available before the mid-1950s, we considered backward extrapolations on the basis of actual data trends apparent through the 1960s and of flat backward extrapolations from the earliest years for which data were available. Because there was a relatively small quantity of person-time accrued over these early decades, the potential impact of any misclassification is expected to have been modest. Nevertheless, estimated exposures for these early years were similar regardless of extrapolation approach. Therefore, for these analyses, we used those estimates extrapolated from actual data trends observed during the late 1950s and through the 1960s, a period over which significant advances in dust control and worker safety were implemented, including ventilation systems and replacement of wood floors with less porous and easier-to-clean materials.
Despite having access to information on prior employment on much of the study cohort, misclassification of respirable crystalline silica exposure before employment in the porcelain industry may have occurred. Although excluding those with known prior exposure resulted in some minor differences, it did not change the fact that several silicosis cases occurred among cohort members estimated to have the lowest exposures. Available work history records for these subjects indicate that prior employment was in jobs tracked by the BG, and most were classified into the “unlikely” outside exposure group. If these subjects actually had meaningful prior exposure, they would have inappropriately been included in the referent group. Similarly some false positive interpretations of radiographs likely occurred, even among those with low exposure. On the contrary, a total lack of (or too few) silicosis cases in the referent group would have prevented or limited the statistical analyses.
Although results of analyses lagging cumulative exposure by 10 years are reported, it is not clear whether this is entirely valid, as cases of silicosis can occur after only a few years of high-intensity exposure.46 Furthermore, it is not known whether crystalline silica exposures sustained after the silica-related fibrotic process begins accelerates the disease process and probability of radiographic changes consistent with silicosis. Therefore, we believe that lagging may introduce additional exposure misclassification and that results should be interpreted with caution.
Another strength of this study derives from the comprehensive records systems we were able to access at the BGGK, including remarkably detailed work history information and the large collection of historical radiographs. This not only allowed independent verification of silicosis status according to a standard protocol but also confirmation of no signs of silicosis on most cohort members. By rereading all radiographs that originally were scored as abnormal (ie, ILO-equivalent scores of 1/0 or higher) with modern standard criteria and blind to occupational history and exposure status, we enhanced the quality and consistency of the silicosis determination. Rereading of all available films for each employee with radiographic evidence of silicosis allowed assignment of a date associated with the first radiograph interpreted as category 1/1 or higher, which in turn allowed more precise estimation of follow-up time for each employee.
Because of exhaustive retrieval of vital status information from multiple sources, no significant loss to follow-up occurred in this study. Nevertheless, among the 5% of decedents for which we were unable to ascertain cause of death, there could be specific causes of death that, if known, would have changed our results. For the causes of death of primary interest, however, this is unlikely, as dust-exposed porcelain workers were under active surveillance for these respiratory diseases. Furthermore, participation in the surveillance program—including periodic radiographic evaluation—preserved the employees' eligibility for compensation should they contract a work-related disease, at least until retirement.
Additional evaluation of the 94 lung cancer cases using additional exposure metrics—or using a nested case-control approach—is unlikely to be informative because of the clear lack of any detectable excess of lung cancer deaths and lack of association between respirable crystalline silica exposure and lung cancer. However, additional evaluation of the silicosis cases by various additional exposure attributes (or combinations thereof) is warranted. For example, a formal exposure–response threshold modeling may elucidate where the threshold of effect likely falls.47,48 Additional follow-up of this cohort may be valuable, as only 9.2% of the cohort was deceased as of end of follow-up. Additional follow-up would also be helpful for silicosis evaluation, as the median age at silicosis determination was only 56. The quantitative exposure assessment covering the past 75 years is unlikely to be further improved, unless additional historical exposure data are discovered or more innovative methods for using them are derived. Modern exposures are uniformly low and will make a relatively minor additional contribution to the estimated exposures.
In conclusion, this study evaluated quantitative respirable crystalline silica exposure in the German porcelain industry and silicosis morbidity and mortality due to several selected causes of death including lung cancer. Our preliminary finding of no increased risk of lung cancer based on our SMR analyses was corroborated by exposure–response analyses demonstrating no consistent associations. Increased risk of silicosis morbidity was clear, with evidence of possible thresholds at or above roughly 4 mg/m3-years cumulative and at or above roughly 0.15 average respirable crystalline silica exposure. More formal threshold analyses might elucidate where such thresholds might lie on the exposure distribution scales; however, accurate determination may require a larger number of silicosis cases.
The authors thank the members of the Scientific Advisory Group for this study—Drs Lesley Rushton (chair), Peter Morfeld, Dirk Taeger, Frank Bochmann, and Hans Kromhout—all of whom provided helpful suggestions on the study methods, analysis and on drafts of this report. They extend their gratitude to the members of the Occupational Medical Service of the Deutsche Steinkohle AG, who provided assistance with rereading and interpreting radiographs to determine silicosis status of the cohort members. They also appreciate the astute comments provided by the anonymous journal reviewers.
This study was sponsored by the Berufsgenossenschaft der keramischen und Glas-Industrie (BGGK, now VBG), the Steinbruchs-Berufsgenossenschaft (StBG, now BG RCI), and by EUROSIL, the European Association of Industrial Silica Producers, with additional support from other trade associations and individual companies.
* The only exception was one very high HR (31.0; 95% CI, 2.5 to 387) for renal disease, based on four deaths among men in one of the lowest cumulative exposure categories (>1.0 to 1.5 compared with those with ≤0.05 mg/m3-years) Cited Here...
1. Hughes JM. Radiographic evidence of silicosis in relation to silica exposure. Appl Occup Environ Hyg. 1995;10:1064–1069.
2. Cocco P. The long and winding road from silica exposure to silicosis and other health effects. Occup Environ Med. 2003;60:157–158.
3. Goldsmith DF, Winn DM, Shy CM. Silica, Silicosis, and Cancer: Controversy in Occupational Medicine. New York, NY: Praeger Publishers; 1986.
4. Smith AH, Lopipero PA, Barroga VR. Meta-analysis of studies of lung cancer among silicotics. Epidemiology. 1995;6:617–624.
5. International Agency for Research on Cancer. Silica, Some Silicates, Coal Dust and Para-Aramid Fibrils. Lyon, France: IARC Press; 1997:506.
6. Straif K, brahim-Tallaa L, Baan R, et al.. A review of human carcinogens—Part C: metals, arsenic, dusts, and fibres. Lancet Oncol. 2009;10:453–454.
7. Checkoway H, Hughes JM, Weill H, Seixas NS, Demers PA. Crystalline silica exposure, radiological silicosis, and lung cancer mortality in diatomaceous earth industry workers. Thorax. 1999;54:56–59.
8. Hessel PA, Gamble JF, Gee JB, et al.. Silica, silicosis, and lung cancer: a response to a recent working group report. J Occup Environ Med. 2000;42:704–720.
9. Checkoway H, Franzblau A. Is silicosis required for silica-associated lung cancer? Am J Ind Med. 2000;37:252–259.
10. Soutar CA, Robertson A, Miller BG, Searl A, Bignon J. Epidemiological evidence on the carcinogenicity of silica: factors in scientific judgement. Ann Occup Hyg. 2000;44:3–14.
11. Attfield MD, Costello J. Quantitative exposure-response for silica dust and lung cancer in Vermont granite workers. Am J Ind Med. 2004;45:129–138.
12. Graham WG, Costello J, Vacek PM. Vermont granite mortality study: an update with an emphasis on lung cancer. J Occup Environ Med. 2004;46:459–466.
13. Brown TP, Rushton L. Mortality in the UK industrial silica sand industry: 2. A retrospective cohort study. Occup Environ Med. 2005;62:446–452.
14. Steenland K, Mannetje A, Boffetta P, et al.. Pooled exposure-response analyses and risk assessment for lung cancer in 10 cohorts of silica-exposed workers: an IARC multi-centric study. Cancer Causes Control. 2001;12:773–784.
15. Cassidy A, Mannetje A, van Tongeren M, et al.. Occupational exposure to crystalline silica and risk of lung cancer: a multicenter case-control study in Europe. Epidemiology. 2007;18:36–43.
16. Pelucchi C, Pira E, Piolatto G, Coggiola M, Carta P, La Vecchia C. Occupational silica exposure and lung cancer risk: a review of epidemiological studies 1996–2005. Ann Oncol. 2006;17:1039–1050.
17. Lacasse Y, Martin S, Gagne D, Lakhal L. Dose-response meta-analysis of silica and lung cancer. Cancer Causes Control. 2009;20:925–933.
18. Westberg HB, Bellander T. Epidemiological adaptation of quartz exposure modeling in Swedish aluminum foundries: nested case-control study on lung cancer. Appl Occup Environ Hyg. 2003;18:1006–1013.
19. Erren TC, Glende CB, Morfeld P, Piekarski C. Is exposure to silica associated with lung cancer in the absence of silicosis? A meta-analytical approach to an important public health question. Int Arch Occup Environ Health. 2009;82:997–1004.
21. Birk T, Mundt KA, Guldner K, Parsons W, Luippold RS. Mortality in the German porcelain industry 1985–2005: first results of an epidemiological cohort study. J Occup Environ Med. 2009;51:373–385.
22. Birk T, Guldner K, Mundt KA, Dahmann D, Adams RC, Parsons W. Quantitative crystalline silica exposure assessment for an historical cohort epidemiological study in the German porcelain industry. J Occup Environ Hyg. 2010;7:516–528.
23. International Labour Organization. Guidelines for the Use of ILO International Classification of Radiographs of Pneumoconiosis. Geneva, Switzerland: International Labour Organization; 1980.
24. International Labour Organization. Guidelines for the Use of ILO International Classification of Radiographs of Pneumoconiosis. Geneva, Switzerland: International Labour Organization; 2000.
25. Callas PW, Pastides H, Hosmer DW. Empirical comparisons of proportional hazards, poisson, and logistic regression modeling of occupational cohort data. Am J Ind Med. 1998;33:33–47.
26. Rothman K, Greenland S, Lash TL. Modern Epidemiology. Philadelphia, PA: Lippincott Williams & Wilkins; 2008:1–758.
27. Steenland K. One agent, many diseases: exposure-response data and comparative risks of different outcomes following silica exposure. Am J Ind Med. 2005;48:16–23.
28. McDonald JC, McDonald AD, Hughes JM, Rando RJ, Weill H. Mortality from lung and kidney disease in a cohort of North American Industrial Sand Workers: an update. Ann Occup Hyg. 2005;49:367–373.
29. Calvert GM, Steenland K, Palu S. End-stage renal disease among silica-exposed gold miners: a new method for assessing incidence among epidemiologic cohorts. JAMA. 1997;277:1219–1223.
30. Rapiti E, Sperati A, Miceli M, et al.. End-stage renal disease among ceramic workers exposed to silica. Occup Environ Med. 1999;56:559–561.
31. Steenland K. Ten-year update on mortality among mild-steel welders. Scand J Work Environ Health. 2002;28:163–167.
32. McDonald AD, McDonald JC, Rando RJ, Hughes JM, Weill H. Cohort mortality study of North American industrial sand workers. I. Mortality from lung cancer, silicosis and other causes. Ann Occup Hyg. 2001;45:193–199.
33. Steenland K, Sanderson W, Calvert GM. Kidney disease and arthritis in a cohort study of workers exposed to silica. Epidemiology. 2001;12:405–412.
34. Forastiere F, Lagorio S, Michelozzi P, et al.. Silica, silicosis and lung cancer among ceramic workers: a case-referent study. Am J Ind Med. 1986;10:363–370.
35. McLaughlin JK, Chen JQ, Dosemeci M, et al.. A nested case-control study of lung cancer among silica exposed workers in China. Br J Ind Med. 1992;49:167–171.
36. Tornling G, Hogstedt C, Westerholm P. Lung cancer incidence among Swedish ceramic workers with silicosis. IARC Sci Publ. 1990;97:113–119.
37. Winter PD, Gardner MJ, Fletcher AC, Jones RD. A mortality follow-up study of pottery workers: preliminary findings on lung cancer. IARC Sci Publ. 1990;97:83–94.
38. Meijers JM, Swaen GM, Slangen JJ. Mortality and lung cancer in ceramic workers in The Netherlands: preliminary results. Am J Ind Med. 1996;30:26–30.
39. Cherry NM, Burgess GL, Turner S, McDonald JC. Crystalline silica and risk of lung cancer in the potteries. Occup Environ Med. 1998;55:779–785.
40. Burgess GL, Turner S, McDonald JC, Cherry NM. Cherry N, Ogden T. Cohort mortality study of Staffordshire pottery workers: (I) radiographic validation of an exposure matrix for respirable crystalline silica. Inhaled Particles VIII. Proceedings of an International Symposium on Inhaled Particles organised by the British Occupational Society 26–30 August 1996. Oxford, UK: Elsevier Science Ltd; 1997:403–407.
41. Steenland K, Sanderson W. Lung cancer among industrial sand workers exposed to crystalline silica. Am J Epidemiol. 2001;153:695–703.
42. Pukkala E, Guo J, Kyyronen P, Lindbohm ML, Sallmen M, Kauppinen T. National job-exposure matrix in analyses of census-based estimates of occupational cancer risk. Scand J Work Environ Health. 2005;31:97–107.
43. Cocco P. Multifactorial aetiology of lung cancer among silica-exposed workers. Ann Acad Med Singapore. 2001;30:468–474.
44. Ulm K, Waschulzik B, Ehnes H, et al.. Silica dust and lung cancer in the German stone, quarrying, and ceramics industries: results of a case-control study. Thorax. 1999;54:347–351.
45. Morfeld P. Comment on: Lacasse Y, Martin S, Gagne D, Lakhal L (2009). Dose-response meta-analysis of silica and lung cancer. Cancer Causes Control. 2009;20:1535–1536.
46. National Institute for Occupational Safety and Health. Health Effects of Occupational Exposure to Respirable Crystalline Silica. Cincinnati, OH: Department of Health and Human Services; 2002.
47. Morfeld P, Albrecht C, Drommer W, Borm P. Dose-Response and threshold analysis of tumor prevalence after intratracheal instillation of six types of low- and high-surface-area particles in a chronic rat experiment. Inhal Toxicol. 2006;18:215–225.
48. Ulm K. A statistical method for assessing a threshold in epidemiological studies. Stat Med. 1991;10:341–349.