A lower than expected lung cancer incidence among workers in cotton mills (textile workers) was first described more than 70 years ago,1 and the concept of environments with lower than expected risks for lung cancer was reviewed in 1990.2 In 2002, a meta-analysis of epidemiological studies of textile industry workers showed that consistent risk estimates for lung cancer risk were reported in studies published through the 1970s and 1980s: the pooled relative risk (PRR) was significantly lower than 1.0 in workers exposed to cotton (PRR = 0.77; 95% confidence interval, CI: 0.69 to 0.86) and wool dust (0.71; CI: 0.50 to 0.92), as well as in carders and fibre preparers (0.73; CI: 0.54 to 0.91), weavers (0.71; CI: 0.56 to 0.85) and spinners and weavers (0.78; CI: 0.66 to 0.91).3 In studies published after 1990, there was a general tendency to move towards unity for all the cancer risk estimates (and an increasing heterogeneity among studies) which is most likely due to better exposure control in this industry.3

Endotoxin is a contaminant of cotton dust and the exposure is higher in cotton mills as compared to other industries with organic dusts such as tea packing or tobacco factories.4^–6 Although measurements of endotoxin levels in cotton mills have been performed in many studies, none of the earlier investigations reporting a reduced risk for lung cancer had endotoxin exposure data in their study design. By contrast, a more detailed analysis of exposure conditions was presented in several recent studies7^–12 that examined cancer incidence in a large cohort of cotton textile workers (267 400 active and retired female workers from 526 textile factories operated by the Shanghai Textile Industry Bureau). In this cohort, cumulative exposure to endotoxin was inversely related to risks of lung cancer (ptrend = 0.002),7 oesophageal cancer (ptrend = 0.01) and stomach cancer (ptrend = 0.001),8 rectum cancer (ptrend = 0.08),9 liver cancer (ptrend = 0.02),10 pancreatic cancer (ptrend = 0.001),11 and breast cancer (ptrend = 0.001),12 when exposures were lagged by 20 years. Endotoxin has not been previously reported to be a protective factor for cancer other than lung cancer and therefore the findings from the Shanghai cohort warrant further studies.

In cotton mills, the amount of endotoxin per unit of airborne dust is generally higher in the card room than in other areas.4 6 A proxy for endotoxin exposure is thus work location/job category in the mills. In order to evaluate a possible reduced mortality from lung and other cancer sites in cotton mill workers, a study was undertaken on cotton mill workers in the province of Vicenza, Italy. Mortality was investigated using plant personnel registers and was related to different work locations and duration of employment and compared to that of the general population.

MATERIAL AND METHODS

The study subjects (n = 4365) were collected from personnel registers in four textile industries in Vicenza and defined as those who were actively working some time during the years 1946 to 1985. The registers contained individual information on the subjects, dates of starting and ceasing work, and work site/location within the mill.

Vital status was ascertained through the registry office of the municipalities of residence from 1 January 1970 to 31 December 1994. Some cases were excluded: 171 were lost during follow-up, and 233 had moved out of the Veneto region. The subjects lost for the study had a slightly longer duration of employment and were slightly older as could be expected. The final cohort comprised 3961 workers.

Within the observation period 612 persons had died and the cause of death was acquired from death certificates, and coded according to the IX Revision of the ICD Classification. The man–year mass was calculated using the program PYRS,13 taking the commencement of work as the entry date and death or end of the follow-up period as the exit date. The cause-specific mortality of the general population of the Veneto region was used as the reference, taking into account age (stratified in 5-year groups), and calendar year (5-year periods: 1970–4, 1975–9, 1980–4, 1985–9, 1990–4). The standardised mortality ratio (SMR) was calculated together with the 95% CI, estimated on the basis of a Poisson distribution.

In addition, using the LogXact statistical package, a Poisson regression analysis was fitted to the data of the sub-cohort of carders, which was broken down into three age classes (<49, 50–59, ⩾60 years) and three periods of calendar years (1970–9, 1980–9, 1990–4), coded as factor variables. In the regression analysis, the trend in lung cancer risk across tertiles of length of employment (6.5–70, 71–147, 148–476 months) was investigated through modelling an interval variable.

The number of individual tumours was rather low and some forms of cancer were combined for the analysis. According to the 2002 IARC Monograph on tobacco smoking,14 cancer of the oral cavity, pharynx, nose, larynx, oesophagus, stomach, pancreas, liver, bladder, other urinary, and bone marrow (myeloid leukaemia) were grouped together in a single nosologic category: “all smoking-related cancer except lung”. In order to allow comparison with the data from the Shanghai cohort,8^–12 a single SMR was calculated, based on observed and expected figures for tumours of the oesophagus, stomach, rectum, liver, pancreas and breast.

In the cotton mills where the study was performed, no measurements of exposure had been conducted. To obtain an exposure gradient the subjects were divided into those working with carding (with high exposure), and with spinning or weaving (with lower exposure to endotoxin). The whole group (carders and others) was then divided into tertiles of duration of employment. The trend in SMRs across the tertiles was analysed using the χ² test for linear trend.15

RESULTS

There were 96 734 person-years in the cohort. Table 1 shows the overall analysis of mortality in the 3961 cotton textile workers followed up from 1970 to 1994.

A significantly increased mortality was observed for malignant tumours, mainly due to cancer of bladder and ill-defined sites. No cases were found for melanoma of skin, other tumour of skin, tumour of connective tissue, testis, other genital, and thyroid gland (data not shown).

Table 2 shows the mortality for lung cancer in relation to duration of employment for carders and other jobs.

The overall lung cancer SMR was 0.93 (CI: 0.45 to 1.72; n = 10) and 1.07 (0.70 to 1.57; n = 26) for carders and non-carders, respectively. When different lengths of employment were compared, the risk among carders decreased linearly with increasing work duration, with a statistically significant (p<0.05) trend.

Table 3 shows the results of the Poisson regression, which confirms the inverse relationship between lung cancer risk and duration of employment as carders (p = 0.04).

Table 4 reports the mortality for all smoking-related cancer except lung. SMR was higher in carders than in other workers: 1.73 (1.20 to 2.42; n = 34) against 1.16 (0.86 to 1.54; n = 49). There were no indications of a lower risk with increasing duration of employment.

Table 5 shows the mortality for tumours at the sites (oesophagus, stomach, rectum, liver, pancreas and breast) where a reduction in mortality was reported in the Shanghai cohort.8^–12 SMR was 1.14 (0.74 to 1.68; n = 25) and 1.07 (0.80 to 1.41; n = 51) for carders and non-carders, respectively. There were neither any reduced risks in the work categories, nor any significant trends for SMRs across the different classes of employment.

DISCUSSION

The major finding in the study was a decreased risk of lung cancer with duration of employment among workers in carding. For other tumours no differences were found compared to the reference population.

Some methodological shortcomings of this study should be considered. The size of the population studied was relatively small, which resulted in a low number of cases; thus, several cancer sites had to be combined into a single category (see Tables 4 and 5), leading to decreased specificity of any observed difference in SMR for individual diagnosis. The study represents a limited setting, but exposures in Italian cotton mills were probably typical of those in plants in other parts of the world,6 thus contributing to the possibility to generalise from the data reported here.

An important confounder for all studies on lung cancer is smoking, and smoking habits could not be determined in this cohort. However, smoking could not be the major determinant of decreasing lung cancer mortality with increasing duration of work for those in the carding department, since this effect was not observed for other smoking-related cancers. Supporting evidence comes from a meta-analysis of epidemiological studies among textile industry workers, where adjustment for smoking made little difference to the findings.3 Likewise, data from the Shanghai cohort suggested that the lower lung cancer incidence at long exposure duration was present among non-smokers as well as among smokers.7 Furthermore, studies in India16 and the USA17 have suggested that the smoking rate in cotton textile workers was slightly higher than in the general population. All the above indicates that smoking is not a confounder for the lower rates of lung cancer, although it cannot be excluded that subjects with a slight inflammation in the airways, induced by the high exposure to endotoxin, could have a shallower inhalation of the smoke than controls, thus leading to a reduced pulmonary dose of smoke. This possibility could be examined in future studies.

Based upon a review of the literature18 and animal models,19 it has been suggested that the agent responsible for the lower lung cancer risk in the studies reviewed above is bacterial endotoxin. This agent is a powerful inducer of inflammation through activation of immune cells, which could suppress and inhibit tumour growth.20

In studies published through the 1970s and 1980s, PRR was significantly lower than 1.0 in carders and fibre preparers (0.73; CI: 0.54 to 0.91), as well as in weavers (0.71; CI: 0.56 to 0.85), and spinners and weavers (0.78; CI: 0.66 to 0.91).3 In the present study, the lung cancer risk was reduced in carders with more than 12 years of employment. Likewise, lung cancer risk in 14 650 textile cotton workers in Lithuania was significantly lower than unity after 10 years of employment and at the highest level of cumulative exposure to textile dust.21

Like in the cotton mills, exposure to endotoxin-containing dusts is ubiquitous in agriculture, with high air endotoxin concentrations in dairy farming.22 23 In previous studies from our group, an inverse exposure–response relationship was found between number of cattle or duration of dairy farming and lung cancer risk.24 25 Moreover, from occupation and industry-coded US death certificates collected from 26 US states for the period 1984–93, lung cancer mortality risk was decreased among livestock with respect to crop farmers, endotoxin exposure being lower in the latter than in the former.26 All the above supports the validity of work location and duration as a proxy for endotoxin exposure.

In conclusion, the results from this study support findings from several earlier reports that cotton workers exposed to high levels of endotoxin-containing dust for prolonged periods of time have a lower risk for lung cancer. No relation was found for other forms of cancer.