Article Text

Original article
Mortality in a cohort of Staffordshire pottery workers: follow-up to December 2008
  1. Nicola Cherry1,
  2. Jessica Harris2,
  3. Corbett McDonald2,
  4. Susan Turner3,
  5. Tony Newman Taylor2,
  6. Paul Cullinan2
  1. 1Division of Preventive Medicine, University of Alberta, Edmonton, Alberta, Canada
  2. 2National Heart and Lung Institute, Imperial College, London, UK
  3. 3Centre for Occupational and Environmental Health, University of Manchester, Manchester, UK
  1. Correspondence to Dr Nicola Cherry, Division of Preventive Medicine, University of Alberta, 5-30 University Terrace, 8303-112St, Edmonton, Alberta T6G 2T4, Canada; nicola.cherry{at}ualberta.ca

Abstract

Objectives To examine mortality from lung cancer, chronic obstructive pulmonary disease (COPD) and chronic non-malignant renal disease (cNMRD) in pottery workers exposed to silica.

Methods A cohort of Stoke-on-Trent pottery workers (N=5115), previously followed to 1992, was traced for vital status and cause of death to December 2008. Standardised mortality ratio (SMR) analyses, comparing deaths to England and Wales and Stoke-on-Trent, examined underlying cause in 1985–1992 and 1993–2008 and mentioned cause for 1993–2008. Survival analysis considered exposure duration and concentration of respirable silica for lung cancer, COPD and cNMRD, using Cox regression.

Results Excess risks of lung cancer, COPD and cNMRD were seen against both England and Wales and Stoke-on-Trent for 1985–2008. SMRs for lung cancer and COPD were lower in 1993–2008 and non-significant for lung cancer against Stoke-on-Trent in that period (SMR 1.07 95% CI 0.92 to 1.25). Exposure concentration, estimated for 1943 subjects, was related to lung cancer in smokers for early but not later deaths with mean silica concentration >200 µg/m3 among deaths to June 1992 (HR 2.80 95% CI 1.21 to 6.50). For COPD an increasing trend with duration and (non-significantly) with mean concentration was seen for early but not later deaths in smokers. No relation was observed between estimated exposures and cNMRD.

Conclusions Excess rates of death from COPD and lung cancer were more marked in the period of the first follow-up (1985–1992) than in the second, with any relation to estimated exposure being limited to the earlier period. Conclusions about COPD and exposure were limited by an early selective destruction of files.

Statistics from Altmetric.com

What this paper adds

  • Although silica is recognised to be a human carcinogen, and to cause lung fibrosis, its relation to other non-malignant respiratory disease is less certain.

  • Non-malignant renal disease and autoimmune connective tissue disease have been reported in some, but not all, occupational cohorts.

  • This study demonstrates an excess of both malignant and non-malignant lung and renal disease, but with greater risk in men dying in the early years of the cohort.

  • Early selective destruction of files, following administrative guidelines, greatly limited the power of the study to draw conclusions about COPD and intensity of exposure, but the results suggested greater risk of COPD with longer duration of exposure.

  • The earlier observation of an increased risk of lung cancer in those with greater exposure concentration was again demonstrated, using an alternative statistical approach.

Introduction

Occupational exposure to silica has long been recognised as a threat to workers’ health through fibrotic effects on the lung (silicosis), but in the last two decades there has been increasing recognition that silica exposure may be responsible for other conditions. The International Agency for Research on Cancer found silica to be a human carcinogen in 1997 and confirmed this decision in 2009.1 In the recent review five industrial settings were considered most relevant for lung cancer, including ceramics, diatomaceous earth, ore mining, quarries and sand and gravel, with potential for confounding in studies of ceramics workers and ore mining. There has been continuing interest in the exposures associated with lung cancer risk, with pooled assessments2 and meta-analysis3 showing increased risk, but with some more recent dissenting reviews4 and studies interpreted as negative.5 ,6 Other outcomes of interest and debate include chronic non-malignant renal disease (cNMRD), with an excess reported in several studies7–10 and findings of dose-effect in some8 ,11 but not all.10 While lung fibrosis (silicosis) is unequivocally associated with silica exposure,12 the contribution of silica to other non-malignant pulmonary disease is less certain, with some reviews13 ,14 suggesting an increased risk of chronic obstructive pulmonary disease (COPD) even in the absence of radiological signs of fibrosis. In the UK chronic bronchitis/emphysema in coal miners was recognised (prescribed) as occupationally related in 1999 by the Industrial Injuries Advisory Council with loss of forced expiratory volume in one second (FEV1) used to document obstruction and years of underground work used to define exposure.15 The Industrial Injuries Advisory Council again reviewed COPD as an industrial disease in 2007 and concluded that there was insufficient evidence to prescribe COPD in silica workers but recommended further research.16 An established cohort of pottery workers in Stoke-on-Trent,17 with an excess of both lung cancer and non-malignant respiratory disease at first follow-up to June 1992, was followed further to 2008. This provided an opportunity to examine the imputation of deaths to COPD in relation to silica exposure, to determine the persistence of the observed excess of lung cancer and to examine, for the first time in this cohort, the relation of pottery work to cNMRD.

Methods

The proposal for this follow-up of the pottery cohort was approved by the Brompton, Harefield and National Heart and Lung Institute ethics committee.

The cohort, already described,17 was of men registered 1931–1992 with the Silicosis Medical Panel (and its successors) as working in the pottery refractory or sandstone industries in Stoke-on-Trent, and who had been born 1916–1945. During employment in prescribed occupations medical forms including information on previous jobs and smoking were completed by physicians at the Medical Boarding Centre every 2 years, and chest radiograms every 4 years. From an identified cohort of 7064 registered, a cohort of 5115 was retained, including only those living in the Stoke-on-Trent area at first registration and who were not known, from the medical record, to have worked with asbestos or in foundries or for a year or more in mining or other dust exposed industries. At the 1992 follow-up vital status was ascertained from the Department of Social Security (DSS) and cause of death from the Office of National Statistics (ONS). An intermediate follow-up had been completed (but not analysed) with ONS in 2004 and a further follow-up, reported here, to December 2008 through the National Health Service Information Centre. Inspection of the cohort suggested that registration cards had been routinely destroyed for those who had died prior to 1985, and the analysis of deaths was thus restricted to those dying in 1985 or later. Although some medical files had been destroyed after this date, as discussed below, the registration cards had been retained.

From 1985 to 2000 cause of death had been coded by ONS to the International Classification of Diseases version 9 (ICD-9) and from 2001 to 2008 to ICD-10 (appendix 1). Underlying cause had been coded for all years, but ‘any mention’ (ie, medical conditions listed as coexisting or contributory, in addition to underlying) was coded routinely only from 1993. For the internal analysis, all mentioned causes were coded to ICD-10 by ONS for deaths 1985–2008.

Date of last contact was taken as date of death or 31 December 2008 for those known to be alive at that date. For those lost to contact the data were censored at 30 June 1992, if traced at the first follow-up but not since, and as 27 October 2003 (the last recorded date of death at the 2004 follow-up) if traced alive at that point but untraced in 2008. Subjects never traced were excluded from all analysis.

Only limited data were available on which to assess exposure for an internal (dose–response) analysis as access to individual files was not granted for the follow-up for 1992–2008. Duration of pottery work June 1992 was estimated for all subjects using the time between first and last medicals carried out at the Boarding Centre (or date of last chest radiograph if later) with the addition of the number of years worked in the industry before the first medical. Type of work within the pottery industry, needed to estimate exposure concentration of respirable crystalline silica, was only available for three groups (1) the cohort used to develop and validate the job exposure matrix,18 ,19 comprising men who had started work before 1960 and been employed in the pottery industry for at least 10 years: (2) men who had been included as either case or referent in two earlier studies and for whom full exposure histories had been extracted17 ,20 and (3) men appearing in record books that had been kept by the Boarding Centre and made available for the period 7 January 1981–17 March 1992, for whom previous exposure was extrapolated. For members of each of these groups an estimate was made of cumulative exposure (the sum of duration × exposure in each period), mean exposure (cumulative exposure/duration) and maximum exposure (the highest concentration estimated in any time period). Where a subject appeared in more than one subcohort, the best available exposure estimates were used. More detail of this exposure assessment is given in appendix 2.

Information was available for the whole (male) study population on age and smoking, which was recorded (except in the earliest years) on each medical form. This was coded as never smoked, smoker (at last medical), ex-smoker or, for those whose employment was confined to the early years, as unknown. The result of the most recent radiograph was also extracted from the file where present, and coded as positive if the recorded International Labour Organization (ILO) reading was ≥1/0.

Statistical methods

Standardised mortality ratios (SMR) were calculated against rates for England and Wales and separately for Stoke-on-Trent, using male population data on age, date of death and cause supplied by National Health Service Information Centre. Expected frequencies were calculated by applying the age and period-specific deaths in males to person years at risk for each year from 1985 to 2008. Stratified analyses were carried out for the years 1985–1992 and 1993–2008. SMRs for underlying cause were calculated, for both reference groups, for all causes, all malignancies, all non-malignant respiratory diseases, all pneumoconiosis, COPD and cNMRD, as well as the specific conditions comprising these larger groups (ICD codes are given in appendix 1). SMRs for ‘any mention’ were also examined and reported for COPD and cNMRD. The SMR analyses were conducted using Stata software.21

For internal analyses ‘any mention’ of ICD-10 codes for lung cancer COPD or cNMRD were considered as outcomes. For each outcome, survival analyses were conducted using Cox regression in SPSS 18,22 with age at last contact as the time variable. These analyses considered first the whole cohort, using duration of exposure, then the subcohort with estimates of concentration. The analyses were repeated, stratifying by follow-up period (deaths to June 1992 or subsequently). Because of the very small number of non-smokers with outcomes of greatest interest (lung cancer, COPD) these survival analyses excluded known non-smokers. In order to assess risk in those without radiographic changes, analyses were repeated restricting the population to those whose last chest radiograph was read as normal (ie, an ILO reading of <1/0).

Results

Of the 5115 in the original cohort 27 were found to be duplicates, 28 were never traced, 1 had an incorrect death certificate and 258 had died before 1985, leaving a potential total for analysis of 4801. Of these, 14 were reported by the DSS to have died but were traced as alive by ONS in 2008 and were included in the internal analysis (but excluded from the SMR). In all 1904 had died between 1 January 1985 and 31 December 2008, all but one with known cause of death. The number of early deaths differed slightly from the earlier report17 because of additional early deaths identified at further follow-up, and with deaths in 1992 being considered to December (rather than June).

Compared to England and Wales reference data, deaths were significantly elevated for all-cause mortality in both the early and later follow-up. All-cause SMR against England and Wales was 1.46 (95% CI 1.33 to 1.59) in 1985–1992 and 1.28 (95% CI 1.22 to 1.35) in 1993–2008 and was elevated for each of the outcomes of particular interest (lung cancer, COPD and cNMRD) except for cNMRD (based on three cases) in 1985–1992 where the SMR of 2.56 had a 95% CI of 0.53 to 7.50. Details of the SMR results are given in appendix 3 with a summary of the results of the SMR analysis using Stoke-on-Trent as the reference shown in table 1. Against Stoke-on-Trent all-cause mortality, lung cancer, COPD and cNMRD were elevated for 1985–2008, as was pulmonary tuberculosis. When the period of follow-up was considered all-cause, lung cancer and renal cancer were significantly elevated in 1985–1992 but not 1993–2008, cNMRD was significantly elevated in 1993–2008 but not (based on three cases) in 1985–1992. COPD was elevated in both periods but with an SMR (1.32) in 1993–2008 that was lower than in 1985–1992 (2.20). Deaths from heart disease were close to the expected in both periods.

Table 1

Underlying cause analysis: observed and expected deaths compared to rates for Stoke-on-Trent

When ‘any mention’ on the death certificate was considered for 1993–2008 the number of lung cancer deaths increased only slightly (from 167 to 178), those for COPD more sharply (from 119 to 208) and for cNMRD nearly sixfold (from 11 to 65). However the associated SMRs tended to be somewhat lower than for the underlying cause. For COPD the ‘any mention’ SMR was based on expected number of deaths (using England and Wales reference data) of 103.83 with SMR 2.00 (95% CI 1.74 to 2.29). The expected number of deaths using reference data from Stoke-on-Trent was 158.60; SMR 1.31 (1.14, 1.50). For cNMRD the expected number of deaths for ‘any mention’ using reference data from England and Wales was 46.24; SMR 1.41 (1.08, 1.79) and for Stoke-on-Trent the expected number was 56.42 with SMR 1.15 (0.89, 1.47).

For the internal analysis there were potentially 4801 subjects, but the medical files could not be found for 367. Among those who had died and for whom a medical file could not be found (N=299) there was a disproportionately high number with mention of pneumoconiosis (29/299:9.7%) and/or COPD (79/299: 26.4%) on the death certificate. This compared with 30/1604 (1.9%) and 205/1604 (12.8%) in deaths where the file was found. The proportion of missing files was greater in the period of the first follow-up of the cohort (to June 1992) (181/468 deaths: 38.7%) than in the period to 2008 (118/1436: 8.2%): files could not be found for 74% (54/73) of those with COPD on death certificates to June 1992.

Those with no employment data (ie, with missing files) were excluded from further analysis, leaving 4434 cohort members. Of these, 1605 (36.2%) had died during the period of interest (1 January 1985–31 December 2008) The analysis of all deaths was thus based on 4434, with 1605 deaths including, as underlying cause, 200 deaths from lung cancer, 122 from COPD and 12 for cNMRD. For ‘any mention’ the numbers were 212 for lung cancer, 205 for COPD and 62 for cNMRD. Exposure concentration could be estimated for 1943/4434 (43.8%).

Smoking history was strongly related to mortality with only 4 deaths from lung cancer and 13 from COPD (as ‘any mention’ cause) occurring in subjects recorded in the medical file as never having smoked. Excluding known non-smokers reduced the total cohort to 3506 of whom 1396 (39.8%) died between 1985 and 2008. In the exposure concentration cohort there were 447 known non-smokers and their exclusion reduced the cohort with exposure to 1496 with 691 (46.3%) deaths. Results of the last chest radiograph were present for 3100 (88.4%) of the total cohort of smokers and 1454 (97.2%) of smokers in the exposure concentration cohort, with 82 (2.3%) and 65 (4.3%) respectively having an ILO reading of 1/0 or greater.

Exposure estimates are shown in table 2 for the cohorts excluding non-smokers. The duration of employment for the whole cohort (mean 13.8 years) was very much less than for the cohort with exposure data (24.3 years) and it was necessary to use different cut points for the two groups. The table first shows duration, dividing the whole cohort approximately into quartiles. For the cohort with estimates of concentration, cut points for duration, mean and maximum concentration and for cumulative exposure were chosen to give reasonable power to examine higher exposures.

Table 2

Distribution of whole cohort and cohort with concentration estimates (respirable silica) by duration, concentration and cumulative exposure

Chest radiograph results were strongly associated with duration of exposure in the whole smoking cohort. More than a quarter (27.3%) of those with duration <2.5 years had no chest radiograph result recorded. Among those with an abnormal chest radiograph all but 5 (93.9%) had been exposed for at least 10 years. In the subcohort of smokers with estimates of concentration and the result of a chest radiograph, an abnormal result was strongly related to mean and maximum concentration and to cumulative exposure. However both a positive chest radiograph and all measures of exposure were strongly related to age, with those born earlier having longer duration, higher exposure and more likely to have an abnormal radiograph (data not shown).

This confounding by age was controlled in the survival analyses (Cox regression) with age at last contact (or death) as the time variable. Table 3 shows the analysis for the whole smoking cohort, with duration of employment as the exposure surrogate and smoking (at last medical), ex-smoker or unknown as a covariate, for all deaths (any mention) from lung cancer, COPD or cNMRD and also for early and later deaths separately. Overall there is little indication of increased risk with longer duration of exposure although there is some suggestion, consistent over time periods, that those employed for 10 years or more were more likely to have COPD mentioned on the death certificate. When this analysis was repeated with duration as a continuous variable the risk of death from COPD increased significantly with each 10 years of duration in the earlier period (HR=1.43, 95% CI 1.04 to 1.96) and overall (HR=1.12, 95% CI 1.01 to 1.25) but not (significantly) in the later period alone (HR=1.10, 95% CI 0.98  to 1.23) Risk of lung cancer did not increase with greater duration except in the early period in which there were 43 lung cancer deaths. For this period a very high and statistically significant lung cancer risk was seen when durations of 2.5 years or greater were compared with durations less than this: this arises from an apparent deficit in such deaths in those with short exposure (with only 3 lung cancer deaths being recorded in the 979 workers with such short exposure). When the analyses were repeated including last radiograph as a covariate, an abnormal radiograph was related to outcome only for COPD deaths in the early period (1985–92), where 4 of the 19 deaths had an abnormal radiograph (HR=6.13, 95% CI 1.86 to 20.18). The results in table 3 were essentially unchanged when the analysis was restricted to those with a last radiograph read as normal, although the size of the (non-significant) increase in COPD risk with at least 10 years duration was reduced.

Table 3

Survival analysis by duration of employment: adjusted for current smoking, ex-smoking or unknown smoking with non-smokers excluded (HR)

In the cohort with estimates of exposure concentration there was some evidence that increasing mean concentration was related to both lung cancer and to COPD, but only in the early death analyses (table 4). Those with mean concentration ≥200 µg/m3 were at greater risk for lung cancer (HR=2.80, 95% CI 1.21 to 6.50) compared to those exposed to mean concentration of <100 µg/m3. In the later period no such effect was seen and indeed the risk was significantly lower at higher exposures. For COPD as a mentioned cause there appeared to be an increasing risk with increasing mean concentration, at least in the earlier years, with HRs (compared to <100 µg/m3) of 1.28 (100<150 µg/m3), 1.53 (150<200 µg/m3) and 3.37 (≥200 µg/m3) but with wide CIs: when this analysis, based on only 11 deaths, was repeated with mean exposure to respirable silica as a continuous variable, the HR associated with each 100 µg/m2 increase in concentration of respirable silica was 1.68 (95% CI 0.93 to 3.018, p<0.09). No relation was seen, in the early or late time periods, between mean concentration and cNMRD; this result was unchanged when the analyses were repeated including non-smokers. In additional analyses (details not shown) cumulative exposure was unrelated to any of the outcomes of interest in either period. Maximum exposure (≥400 µg/m3) was related to lung cancer in the early period (HR=2.42, 95% CI 1.12% to 5.23%) but not in the later period or to other outcomes. When the analyses were repeated including last radiograph as a covariate, an increased risk of death from lung cancer was seen in those with missing radiograph but the increased risk with mean exposure ≥200 µg/m3 remained (HR=2.81, 95% CI 1.19 to 6.65). Similarly, although there was an increased risk of COPD deaths in 1985–92 for those with an abnormal chest radiograph, the (non-significant) increasing risk with exposure remained in evidence. The pattern of results in table 4 was essentially unchanged when the analyses were restricted to those with last radiograph read as normal, although the size of the increased risk with mean exposure ≥200 µg/m3 was reduced (HR=1.97, 95% CI 0.76 to 5.09).

Table 4

Survival analysis by mean exposure concentration (respirable silica): adjusted for smoking with non-smokers excluded (HR)

Discussion

The SMR analysis has shown clearly that men working in the pottery and related industries in Stoke-on-Trent had more deaths than expected for the diagnoses of interest, using comparison data from England and Wales and, more demandingly, from the Stoke-on-Trent region, where environmental and lifestyle factors of the population will have been similar to those of the pottery workers. Excess deaths were effectively confined to malignant and non-malignant respiratory and renal disease, with a clear excess of cNMRD overall and of renal cancer in the period to 1993. Among respiratory diseases, deaths from lung cancer and COPD were in excess over the whole time period, with rates for both being lower in the more recent past than in the first follow-up. All-cause mortality was very little in excess, particularly in 1993–2008 and the observed and expected rates for heart disease were closely similar when compared with Stoke-on-Trent, suggesting that higher tobacco use by pottery workers (if that were indeed the case) would not sufficiently explain the respiratory excess. The expected deaths for Stoke-on-Trent remain appreciably higher than for England and Wales as a whole, particularly for COPD. It is unclear whether this poor mortality in the general Stoke-on-Trent population mainly reflected lifestyle, with high rates of smoking, large numbers with employment in pottery work or, perhaps, industrial pollution.

It had been hoped that the internal analysis, with allowance for smoking, would help determine whether, within pottery workers, higher silica exposure had contributed to lung cancer, cNMRD or, particularly, rates of lung disease recorded as COPD on death certificates. This analysis was however hampered by lack of power (exposure concentrations could only be estimated for 40% of the cohort) and low precision in both the outcome and exposure. Because of small numbers the outcome was taken as ‘any mention’ on the death certificate, rather than the underlying cause shown to be associated with a larger SMR. Estimation of exposure through a job exposure matrix inevitably results in misclassification, tending to bias effect estimates towards the null: in this instance the rather crude estimates of concentration for those without detailed job histories will have resulted in considerable error. The smoking data used was imprecise, taking ‘ever smoked’ rather than pack years. Among those included in the ‘smokers’ were those with missing smoking data. These were largely found among those with short early exposure but some, perhaps few, will have been non-smokers.

The early policy of destroying medical files was a grave handicap in these analyses: the DSS guideline was that a medical file need only be preserved for 2 years after a known death or 10 years after last employment or when the person reached the age of 70 years. In the absence of the medical file it was not possible to extract information on radiographic changes, but it seems likely that files were destroyed mainly following claims for death benefits (funeral costs) for workers previously found to have fibrotic changes on radiographs: 83% (20/24) of those with pneumoconiosis on the death certificate prior to 1993 had no extant file. This destruction reduced markedly the number of cases of early COPD death available for analysis: in the records remaining fibrosis was seen to be associated with death from COPD in the early years, and the suggestive (though not significant) trend with mean concentration of respirable silica might well have been supported if these early destroyed records had been available. Moreover, deaths for which files were not destroyed in the early years may well have been biased towards lower exposures (or at least fewer radiographic changes), underestimating the true risk. In contrast, those included in the exposure concentration subcohort had longer duration of exposure and were somewhat more likely to have an abnormal radiograph than the whole cohort and so may have been at greater risk: restriction of analyses of concentration to this non-random subcohort may also have produced some unknown bias. It is worth noting that the unexplained high risks of lung cancer with duration >2.5 years in 1992 and earlier was based on the whole cohort, and this aberrant result cannot be explained by the non-random inclusion of subjects in the exposure concentration subcohort. The consistent, though not always significant, increase in risk of COPD with duration was based on the whole cohort with extant records, and should not be subject to selection bias.

Given these limitations it is reassuring that the earlier finding of increased risk of lung cancer at higher mean concentrations was confirmed (with cases restricted to deaths from 1985 to June 1992) using a statistical approach different from the earlier case-referent analysis, and with a somewhat different set of cases (excluding deaths prior to 1985 and including new cases found in later follow-up as well as cases where the lung cancer was not recorded as the underlying cause). As before, lung cancer was unrelated to duration or cumulative exposure.17

The cohort excluded those known to have had asbestos exposure prior to starting work in the potteries; this was not known for the 367 for whom the medical file could not be found but who were included in the cohort. In the analysis of causes of death, 10 mesothelioma deaths were found, together with 2 deaths in which asbestosis was an underlying cause and a third in which it was a mentioned cause. The available information was examined for these 13 cases, to determine how likely it was that their exposure to asbestos had been in the course of their work in the potteries: mesothelioma cases in the pottery industry have previously been reported from men in firing and kiln repair where asbestos may have been used widely. For 4 of the 13, all born in 1922 or earlier, there was no extant medical file and thus no information on previous (or later) work with asbestos outside the potteries. None of the three mesothelioma cases in this group had pottery work on the death certificate, but one asbestosis case did. There were a further four mesothelioma cases with known but short and early pottery exposure starting at age <20 years. None of these had pottery work on the death certificate and it is unknown whether they had asbestos exposure in the long period of potential employment after leaving pottery work. The remaining three deaths from mesothelioma did have pottery work on the death certificate. One had entered pottery work in his mid-fifties and died 16 years later, a fairly short latency if the exposure was indeed in this job. For the remaining two, with long latency (33 and 54 years), long years of exposure (19 and 29 years) and pottery work on the death certificate, it must be assumed that asbestos exposure had been in the job. Neither of these had a history of work in firing or other kiln work. For the two remaining asbestosis deaths the last radiograph result available had been read as normal and it seems unlikely that there had been substantial asbestos exposure in pottery work since that time. Although the data available are very limited they suggest that some workers had been exposed to asbestos in pottery work and that this may have been a confounder in the lung cancer analysis.

The lack of dose–response for lung cancer or COPD in more recent periods, or for cNMRD at any period, requires some consideration. First, this may be a reflection of the somewhat ad hoc exposure estimation, with exposure misclassification being sufficiently large to mask any suggestion of increasing risk with increasing exposure: exposure estimates for early lung cancer deaths would have been less subject to extrapolation. Second, the job-exposure matrix may have importantly underestimated early exposures that were truly massive (an estimate of exposure of, say, 200 µg/m3 in early years might imply greater exposure than 200 µg/m3 in later years). Third there may have been some unmeasured confounder more generally present in the early years. Finally, the biological mechanism for cNMRD is not fully understood, but insofar as it reflects an autoimmune response, a classic dose–response relationship may not apply. It is of interest that in the pottery cohort considered here there was only one case of autoimmune connective tissue disease listed on a death certificate as the underlying cause, failing to support earlier suggestions of excess risk of this condition8 ,23 but consistent with an earlier negative finding on rheumatoid arthritis in this cohort.20

In summary, evidence from the SMR analyses suggests that the risk of lung cancer and COPD in the pottery industry was lower in the more recent follow-up than in the period to 1992, though an increased risk of COPD was still present. The relationship of high exposure concentrations of silica to lung cancer was again supported for early deaths. The internal analysis of exposure concentration and COPD did not have the power to confirm or refute a relationship between silica exposure and COPD in smokers (with or without radiographic change) and could not comment at all on risks in non-smokers.

Acknowledgments

Magda Wheatley played a major role in the collation and checking of death certificates and in the extraction and coding of exposure data. Jessica Harris received salary support from the Colt Foundation.

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:

Footnotes

  • Contributors NC, CMcD and TNT conceived the study and with JH determined the scope of the SMR analysis, which was conducted by JH. ST and NC prepared the data for the internal analysis, which was conducted by NC. PC helped to refine the questions on COPD. All the authors commented critically on the initial draft and have agreed with the final text. NC and JH act as guarantors.

  • Funding The work was funded from research funds held at the National Heart and Lung Institute, London. The funding source had no role in the study design, collection, analysis or interpretation, in the writing of the report or in the decision to submit.

  • Competing interests None.

  • Ethics approval Brompton, Harefield and NHLI REC.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement Additional unpublished data are available to other researchers from the corresponding author following agreement to the proposed use.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.