Discussion

It has been widely recognised that exposure to cotton dust leads to byssinotic symptoms. The acute inflammatory response, manifested as Monday morning chest tightness (byssinosis), and chronic inflammation may be different pathogenic phenomena. Although there is evidence that repeated episodes of byssinosis are associated with accelerated loss of lung function,10 17 clear distinctions between the relative contributions of acute, reversible inflammation and chronic inflammation to COPD are not established. Unfortunately, we do not have data on byssinosis among cohort members to address this question. Several studies have demonstrated a reduced risk or, at least, no overall increase in mortality rates from chronic bronchitis, emphysema or lung cancer among cotton workers.18–20 One study in USA reviewed 2895 consecutive autopsies during an 18-year period, and found no significant difference in the prevalence of emphysema or other chronic lung diseases among 282 employees of a cotton textile mill when compared with the non-textile worker population.21 However, it is difficult for these studies to fully compensate for the ‘healthy worker’ effect, resulting in an underestimation of the true risk.

In our cohort, excess COPD mortality was observed among cotton workers when compared with the remainder of the cohort. This finding corresponds to recent reports of a higher risk of COPD and chronic respiratory symptoms among cotton workers.22 23 However, this increased mortality was not seen when the general Nanjing female population was used as a reference. This discrepancy might be due to the manifestation of a ‘healthy worker effect’ that would lead to underestimated effects in the workforce. In contrast, silk workers had significantly higher COPD mortality rates compared with both the remainder of the cohort and the Nanjing female population. An increase in COPD mortality among silk workers was unexpected. In the series of lung function and symptom studies conducted by Christiani et al, cotton textile workers had more chronic cough and chronic bronchitis and greater decline in FEV₁ than silk workers.6 8 Silk workers in these studies were regarded as the non-exposed reference group, largely because silk dust had only background levels of endotoxin, compared with the relatively high levels found in cotton dust.24 A recently published report from this study found both cotton and silk workers have improved FEV₁ and reduced prevalence of chronic respiratory symptoms after work cessation, suggesting that there may be respiratory hazards in silk factories as well.25

The main components of silk dust are organic compounds, mostly protein, fats and carbohydrates. Contaminants of silk, such as silkworm larval proteins and constituents of fungal cell walls, could induce hypersensitivity pneumonitis and occupational asthma,26 which in turn could lead to COPD if exposure occurred over a prolonged period of time. Significantly, spirometric abnormalities (FVC, FEV₁, MMF, V₅₀ and V₂₅) were observed among 158 non-smoking female tussah (‘wild’) silk processing workers in Liaoning Province, China, compared with 20 non-smoking referent workers from an appliance assembling factory.27 Other studies in the Chinese literature reported an association between mulberry dust and cough, expectoration, and shortness of breath.28 29 Animal fibres, free silica, fungi and bacteria are all possible causes of these respiratory effects. Depending on the quality of the raw silk, pesticides and heavy metals such as lead may also be present in this dust. Other possible hazardous exposures include chemicals involved in processing, such as dyes and highly oxidative bleaching compounds. Previous studies have shown that humidifiers contaminated with microorganisms could cause byssinosis-like symptoms, and that this could be the mechanism by which the risk of COPD is increased among silk workers. A more extensive assessment of silk factory exposures is therefore suggested.

This study has several strengths. First, COPD is rarely seen in non-smokers. Patients diagnosed with COPD are expected to have smoked the equivalent of about 20 cigarettes per day for more than 20 years.30 The prevalence of smoking is less than 3% in our cohort and none of the silk workers who developed COPD were smokers. Although the confounding effects of secondhand smoke outside the workplace cannot be completely discounted, it is unlikely that this exposure would differ among women working in other sectors of the textile industry or the general female population in Nanjing. The textile workers in this study were relatively homogeneous with regard to residence, religion, race, marital status and social-economic status. Also, the vast majority of the women in this study had held only one job and worked in the same factory for their entire career. During the follow-up period, only 1.14% (n=3056) of the entire cohort transferred to a different sector. Among textile workers, 0.03% (n=30) transferred from cotton to silk and 0.04% (n=5) transferred from silk to cotton. None of the women who died from COPD had a factory transfer. Therefore, it is unlikely that changes in exposures due to job transfers influenced our results.

One limitation of this study is the possible misclassification of COPD using death certificate data. We included asthma as one of the underlying or contributing causes of death from COPD for several reasons. COPD and asthma share many clinical characteristics. In order to distinguish severe asthma from COPD, we would require more detailed clinical data than was available. It has been noted that chronic illnesses, COPD in particular, are prone to misdiagnosis and misclassification on death certificates in China. In an independent review of death certificates in China, 178 (73.5%) of 242 actual deaths due to COPD were correctly coded as such and 50 (20.1%) of 228 deaths coded as COPD were actually due to other causes.31 Under the reasonable assumption that misclassification of cause of death in our study was not associated with any of the possible risk factors considered, misclassification of this order of magnitude would have biased our OR estimates towards unity, but would not have resulted in spurious associations. Indeed, both conditions might be present simultaneously in an individual patient to variable degrees. In our study, 33 of the 184 COPD deaths had COPD listed as a secondary cause of death. The first causes listed were primarily heart failure, chronic cor pulmonale (pulmonary heart disease) or other respiratory illnesses. Those deaths were kept in the final analysis as COPD deaths because we believe COPD was an important contributing cause. Similar results were found when those deaths were excluded. We did not have complete employment history record and exposure for each individual worker. Complete work history information was available only for a subset (n=3199) of the cohort who were recruited for a series of cancer case-cohort studies,32 33 and there were six COPD deaths in the subcohort. Future work will assess complete work history and exposures of COPD cases compared with controls from this cohort.

Already a common cause of death in developed countries, COPD is an increasingly important cause of death and disability in the developing world.34 Much of this increased prevalence is likely due to cigarette smoking, but occupational exposures, particularly among non-smokers, are also driving this increase. Recent studies from Nicaragua,35 Turkey,36 India,37 Pakistan,38 Nigeria39 and Ethiopia all report on byssinosis in textile workers and several authors also describe chronic and irreversible changes in lung function in these workers.40 These chronic changes may go by the term ‘chronic bronchitis,’ ‘chronic declines in FEV₁’ or ‘chronic respiratory symptoms,’ but all are likely equivalent terms for what we term COPD. Epidemiologic assessments of COPD in developing countries should therefore be a high priority to identify preventable causes of disease.