|Home | About | Journals | Submit | Contact Us | Français|
To investigate possible associations between miscarriage and occupational exposures in the Shanghai Textile Industry.
We conducted a retrospective cohort study of miscarriages among 1,752 women in the Shanghai textile industry. Reproductive history was self-reported by women and occupational work histories were collected from factory personnel records. Occupational exposures were assigned by linking work history information to an industry-specific job-exposure matrix informed by factory-specific textile process information and industrial hygiene assessments. Estimates of cotton dust and endotoxin exposure were also assigned. Odds ratios (OR) and 95% confidence intervals (CI) were estimated by multivariate logistic regression, with adjustment for age at pregnancy, education level, smoking status of woman and spouse, use of alcohol, and woman’s year of birth.
An elevation in risk of a spontaneously aborted first pregnancy was associated with exposure to synthetic fibers (1.89, 95% CI: 1.20–3.00) and mixed synthetic and natural fibers (3.31, 95% CI: 1.30–8.42). No increased risks were observed for women working with solvents, nor were significant associations observed with quantitative cotton dust or endotoxin exposures. Associations were robust and similar when all pregnancies in a woman’s reproductive history were considered.
Occupational exposure to synthetic fibers may cause miscarriages, and this possibility should be the subject of further investigation.
As an increasing number of women work outside the home during pregnancy, the role of occupational risk factors in adverse pregnancy and birth outcomes is of increasing importance. Work in the textile industry subjects employees to an array of exposures during the spinning, dyeing, weaving, sewing, and finishing of garments. In a study from Finland 1, textile spinning was associated with an increased risk of miscarriage (odds ratio [OR]=1.3, 95% confidence interval [CI]: 1.0–1.9), as has weaving (OR=1.4, 95% CI: 1.1–1.9), fabric inspection (OR=1.5, 95% CI: 1.0–2.4), and other textile work (OR=1.4, 95% CI: 1.0–2.0). In a separate Finnish study, women working at a textile plant had an increased rate of miscarriage (miscarriage rates of 16.7% vs. 11.4%) 2, but no significant increase in risk was found in a U.S. study of maternal occupation in the textile industry overall 3.
Textile workers may be exposed to cotton dust during fiber processing and spinning, solvents in the dyeing, weaving, and finishing of textile goods, and to bleaches and dyes during dyeing and weaving processes. Elevated miscarriage risks have been found in several occupational groups of women exposed to solvents, including factory workers (toluene) 4, semiconductor industry workers (glycol ethers) 5, and laboratory workers (toluene and xylene) 6. Carbon disulfide is used in the manufacture of viscose rayon fibers in the textile industry, and increased miscarriage risk has been identified for women viscose rayon workers 7, 8. Fiber dust exposure is widespread in the textile industry, with frequent exposure to cotton dust in the Shanghai, China textile industry. Cotton textile workers are commonly exposed to endotoxin, a lipopolysaccharide (LPS) component of the cell wall of gram-negative bacteria present in these and other organic dusts. Endotoxin is an established respiratory toxicant, and has been shown to induce miscarriage in animal studies 9–12. Chemical and dust exposures of textile workers also are present in other industries; thus, findings from this study could have implications for exposures in other occupational settings.
A prior study of cancer risks in the Shanghai, China textile industry provided the opportunity to conduct a historical cohort study to examine associations between occupational exposures and self-reported history of miscarriage in a group of women textile workers.
Study subjects for the current study were selected from participants previously enrolled in a large randomized trial (n=267,400) of breast self-examination conducted among workers from 526 factories affiliated with the Shanghai Textile Industry Bureau (STIB) in Shanghai, China. Enrollment in the trial occurred during 1988–91 13, 14. Women eligible for the randomized trial were Shanghai residents, born between 1925 and 1958 (aged 30–64 at entry into the study), who were current or retired workers in any of the factories. Participation in the study was high, with 98% of eligible women taking part in the study.
As part of two case-control studies of breast diseases nested within the randomized trial, 1814 control women (women without a diagnosis of breast cancer or benign breast disease in 1994) were administered a questionnaire that inquired about the woman’s pregnancy history 15–17. These women form the cohort of women for the current study. This reproductive questionnaire elicited detailed information for each pregnancy, including end date of pregnancy, outcome of each pregnancy, and length of gestation via in-person interview by trained Chinese physicians and nurses. Also recorded was information on personal and spousal smoking, alcohol use habits, demographic data, and gynecologic history (age at menarche, contraceptive use). The overall response rate was 81%. All procedures were approved by the relevant Institutional Review Boards of the Fred Hutchinson Cancer Research Center (Seattle, USA) and the Shanghai Textile Industry Bureau (Shanghai, China).
Among women who answered a detailed reproductive health questionnaire, 1752 (96.5%) had ever been pregnant and were included in the present study. The pregnancies as reported on the questionnaire occurred between 1942 and 2001. Most (92.6%) first pregnancies occurred between 1950 and 1989, and only 24 (1.4%) first pregnancies occurred between 1990 and 1996. A woman was considered to have had a miscarriage as a pregnancy outcome if she reported a miscarriage of less than 20 weeks gestation. Live births, termination of pregnancy (induced abortions), still births, and ectopic pregnancies were also reported pregnancy outcomes.
The STIB was an industry-wide governing body for all textile factories in Shanghai. Work in the factories included all textile processes from processing raw materials to completing the finished products. The STIB also employed women in jobs such as teaching, nursing and medical clinic work, cafeteria work, and daycare, resulting in a wide spectrum of occupational exposures.
Exposures for each woman were determined by collecting information on her work history and job assignments. Each woman’s data on work history and job assignments were abstracted in 2005–2006 from factory personnel records using a standardized form and protocol by trained personnel who were all former medical workers employed at STIB factories. The factory in which a woman worked, the textile processes and fibers with which she worked, job tasks of the woman, and the dates of employment were collected beginning with the woman’s first job until her last STIB job. For 89% of the job assignments held by the women, information was ascertained from factory records. In cases where factory personnel records could not be identified, supervisors were identified and completed the form (2.6 % of job assignments), or the woman was contacted and interviewed (7.3 % of job assignments).
Factory-level information was collected for each of the textile factories in the STIB using a standardized form. The data collection instrument was modeled after the exposure measurement forms routinely collected by the Shanghai Municipal Centers for Disease Control, district offices, or STIB industrial hygienists as part of their routine monitoring. Experienced industrial hygienists, trained in China, collected information from factories regarding factory demographics (date operation began, factory size currently and at peak, and fiber used in production) and a figure describing the factory process. The industrial hygienists also recorded every industrial workshop and process within the factory, excluding administrative and educational services. Within each textile process, the industrial hygienists detailed dates of operation, fiber, worker types, exposure to up to 5 hazardous agents, and any changes in the process (e.g., introducing synthetic fiber into a cotton factory). The industrial hygienists assessed hazardous exposure based on their experience, monitoring data, and investigations by the STIB. Textile processes were classified using a combination of 18 major process codes, which usually represented the workshop name (e.g., material handling, fiber processing, spinning, dyeing), and 149 specific process codes which identified one component of textile manufacturing within the major process (e.g., carding, drawing, combing, roving). A list of potentially hazardous exposures within each textile process was also recorded by the industrial hygienists and included 14 dusts (e.g., cotton, wool, synthetic fibers) and 239 chemicals (e.g., ammonia, formaldehyde, xylene). Infrequent exposures (<2 exposed women) are not presented in the tables.
A job-exposure matrix (JEM) specific to the textile industry was developed using a combination of information recorded by the Shanghai industrial hygienists and a priori assessment by a team of experts from the University of Washington and the STIB 18. Exposure was assigned if the experts agreed that workers in the process would be exposed during normal operation, or if more than 30% of the textile processes reviewed above reported exposure on the standardized factory data collection forms 18. Woman-level information and factory-level information were then linked by fiber type, major textile process type, specific textile process type, type of work performed (e.g. operator, floor supervisor), and year. Exposure to potentially hazardous agents was determined by linkage to the JEM. Further extensive details on exposure assessment, factory data collection and job-exposure matrix development are detailed in Wernli et al 18.
Quantitative exposure information was also available to determine cotton dust and endotoxin exposures. For cotton dust, historical industrial hygiene records from 56 factories between 1975 and 1999 were transcribed for use in this study to estimate exposure for each specific textile process. Endotoxin data for this industry from a study of respiratory health in three cotton products factories by Christiani et al. 19–21 and additional contemporary measurements of cotton dusts collected for this study 22, 23 were used in the development of predictive models for endotoxin exposure. For cotton dust and endotoxin, both exposures during the pregnancy alone and cumulative exposures from the beginning of her textile employment were of interest. Quantitative cotton dust and endotoxin estimates were combined with each woman’s job-duration information from the work history to determine cumulative exposures.
Only work periods in STIB factories were included as the extensive exposure characterization was conducted solely for STIB factories. Three hundred twenty-three women were excluded from the analyses because their work history during their pregnancies could not be ascertained, leaving 1429 women included in the analyses.
For this historical cohort, quantitative analyses were first conducted evaluating associations between established risk factors (e.g., age at pregnancy, smoking status, and educational attainment), and pregnancy outcome to allow comparison with other epidemiologic studies. Then, analyses were conducted examining associations between pregnancy outcome(s) and textile industry exposures. Two separate analyses were conducted, the first utilizing all pregnancies, and the second using only each woman’s first pregnancy. Analyses limited to the first pregnancy only were conducted to exclude biases from a woman’s pregnancy history, including prior adverse outcomes and within-woman correlations among pregnancies, while avoiding bias associated with choosing the most recent pregnancy 24. The most meaningful result to present from our data was a summary measure of the average risk ratio, hence the use of logistic regression analysis methods. In most analyses, we considered exposure only during pregnancy (“current exposure”) to evaluate the effects of acute exposures during pregnancy. A woman was considered to have current exposure during the pregnancy if she was exposed between the estimated date of conception and the earlier of either her 20th week of pregnancy or the end date of the pregnancy. Prior mechanistic data supported a model of chronic exposure 25–28 potentially leading to miscarriage for endotoxin, thus cumulative exposure up to the first pregnancy was also analyzed for cotton dust and endotoxin exposure. Cumulative occupational exposure was accumulated from the beginning of a woman’s first employment in the STIB up until the earlier of either the 20th week of the pregnancy or the end date of the pregnancy, excluding periods of nonexposure if the woman changed jobs. Analyses were conducted evaluating associations with specific potentially hazardous agents of interest as assigned using the JEM and with employment in a textile manufacturing process as a surrogate of potential exposure. Quantitative cotton dust and endotoxin exposures were analyzed by quartile of exposure among exposed women.
Analyses considering all pregnancies in a woman’s reproductive history were conducted. In order to consider the within-woman correlations between pregnancies, analyses considering all pregnancies utilized generalized estimating equations (GEE) to estimate the odds ratio (OR) and 95% confidence interval(CI) 29. Terminated pregnancy (induced abortions, n=1122), ectopic pregnancy (n=22), or other unspecified outcomes (n=4) were excluded. Our analyses thus compare pregnancies that end in miscarriage with the comparison group of pregnancies that could have ended in miscarriages, still births and live births. Analyses were adjusted for gravidity, age at pregnancy (in 5-year age groups), woman’s year of birth, highest level of education completed (never went, elementary school, middle and high school, college or more), woman’s smoking status during pregnancy (yes/no), spouse’s smoking status during the pregnancy (yes/no), and woman’s alcohol use (never, ≤1/month, ≤1/week, >1/week).
For secondary analyses considering the first pregnancy only, women whose first pregnancy resulted in a termination of pregnancy (n=46) or other pregnancy outcome (ectopic pregnancy (n=5) and other unspecified outcome (n=3)) were excluded from the comparison group. Analyses examining the outcome of the first pregnancy were conducted using logistic regression to estimate the odds ratio (OR) and 95% confidence interval (CI). As before, all analyses were adjusted for age at pregnancy, woman’s year of birth, level of education, smoking status of the woman and the spouse, and alcohol use. All analyses were conducted in STATA version 9.0 (Stata Corporation, College Station, Texas).
There were a total of 4258 pregnancies experienced by the 1429 women in the study. Table 1 shows the numbers of pregnancies contributed by the women. Women had a median of 2 pregnancies, ranging from 1 to 10 pregnancies per woman. Approximately six percent of first pregnancies were miscarriages, with an additional 0.9% still birth, 0.4% ectopic pregnancy, and 89.4% live births. An additional 3.2% of first pregnancies were terminations of pregnancy. When considering all pregnancies, 4.4% were miscarriages and 59% were live births, but a much higher percent of pregnancy outcomes were terminations of pregnancy. The gestational durations differed by outcome of pregnancy, with live births having a mean of 40 weeks, miscarriage with a mean of 10.1 weeks, and termination of pregnancy occurred at a mean of 6.7 gestational weeks.
The distribution of the study subjects by age at first pregnancy is shown in table 2. Women whose first pregnancy occurred in their 20’s represented 79% of the women in the study, and only 2% of women had a first pregnancy at age 35 years or later. Table 2 also shows that 76% of the women completed at least a middle school education (completed 8 years of education), all women had been married, smoking was infrequent (but spousal smoking was much more frequent), and alcohol use was also infrequent. Risk of a miscarriage was increased in teen pregnancies, and in women who smoked, and was lower than expected in women married to smokers. Women were more likely to have been born after the second world war, with 53% born between 1950 and 1958.
No statistically significant associations between miscarriages and employment in a specific textile process during pregnancy were identified when considering all pregnancies or when analyses were restricted to the first pregnancy (table 3).
An increased risk of miscarriages was found for women with exposure to synthetic fibers and mixed synthetic and natural fibers, both when analyses were limited to the first pregnancy, and when all pregnancies were considered (table 4). Among those few women who were exposed to synthetic or a mixture of natural and synthetic fibers, but not during their pregnancies, no elevation in risk was observed (for synthetic fibers, n=3 cases, OR=1.16, 95% CI: 0.35–3.88; and for mixed synthetic and natural fibers, n=1 case, OR=0.90, 95% CI: 0.12–6.97). Results for synthetic fiber exposures for first or all pregnancies were not appreciably changed when adjustments were made for exposures to other fiber types (OR=1.93, 95% CI: 1.21–3.08).
No statistically significant elevations in risk were found for women exposed to solvents, or any of the other exposures considered. No statistically significant elevations in risk of a miscarriage were observed with any of the measures of exposure to endotoxin or cotton dust that were utilized (Table 5).
An elevated risk of miscarriage was found in women with exposure to synthetic fibers and a mix of synthetic and natural fibers. No elevations in risk were identified with exposure to cotton dust, endotoxin, or solvents.
Synthetics fibers used in the Shanghai textile industry to which the women were exposed include polyester, lastrile, nylon, and rayon. The specific hazardous agents potentially responsible for the observed associations with synthetic fiber exposures are not known. Carbon disulfide, a major ingredient in rayon manufacturing, has been previously found to have reproductive effects in female viscose rayon workers 7, 8, 30. Although carbon disulfide has been detected in historical measurement data in this industry, all measurements were conducted in synthetic fiber manufacturing factories, and no women in the current study were employed during pregnancy in factories manufacturing synthetic fibers. Only fifteen out of all five hundred STIB factories produced synthetic fibers and men may have been preferentially employed in manufacturing processes. Women working in manufacturing processes were not preferentially selected for inclusion in this study. In this study, women with exposure to synthetic fiber dusts encountered these dusts in non-manufacturing factories which used synthetic fibers to make finished products, and carbon disulfide exposure would not be expected in these processes. Textile processes associated with synthetic fiber dust exposure outside of manufacturing included applying coatings, spooling, weaving, knitting, printing, laminating, cutting, sewing, and packing of synthetic fiber products. It is also possible that the elevated risks associated with synthetic fiber exposure are surrogates of an exposure we were not able to fully characterize using our semi-quantitative job-exposure matrices.
Exposure to other solvents has also been associated with elevations in risk of a miscarriage. In a cohort of 3,105 non-smoking, 20 to 24 year old, female petrochemical workers in Beijing, the relative risk of a first pregnancy ending in a miscarriage was 2.7 (95% CI: 1.8–3.9) for exposure to any petrochemical and 2.5 (95% CI: 1.7–3.7) for exposure to benzene 31. In a population based cohort of 15 to 29 year old pregnant women in the San Jose, CA area, elevated risks of miscarriage were found in relation to self reported jobs likely involving solvent exposures (OR=4.4, 95% CI: 1.5–12.7) and self-reported solvent exposures (OR=2.8, 95% CI: 1.1–7.0) 32. In this study, women classified as having exposure to solvents were likely exposed to a mixture of solvents, and we were not able to identify exposures to specific solvents. No elevation in risk with exposure to solvents as a class was found. In the Shanghai textile industry, commonly used solvents included carbon disulfide, toluene, benzene, and xylene. Although historical industrial hygiene data were available, inconsistent data quality and lack of consistent monitoring did not allow for a quantitative job-exposure matrix of solvent exposure.
In the textile industry, exposure to cotton and other dusts is widespread, and exposures occur at exceptionally high levels in certain processes. Endotoxin is a lipopolysaccharide (LPS) component of the cell wall of gram-negative bacteria and is commonly found in cotton dusts. In animal studies, injections of endotoxin have led to miscarriages 9–11, 33, possibly as a result of a systemic inflammatory response which has been fairly consistently documented between studies and has been observed following either injection or inhalation exposure routes 34. Occupational exposure to dusts has been previously associated with reproductive effects 1, 35. Self-reported exposure to dusts was associated with a 2.7 (95% CI: 1.2–6.1) fold increase in risk of infertility, specifically ovulatory causes (3.0, 95% CI: 1.2–7.5) and tubal causes (2.9, 95% CI: 1.1–7.9) 35. In contrast, another epidemiologic study found no association when using job titles to identify dust exposure (OR=1.1, 95% CI: 0.9–1.2) 1. A strength of our current study was the ability to test the association of miscarriage with quantitative measurements of cotton dust and endotoxin exposure 23. However, no significant elevations in risk of miscarriages with either endotoxin or cotton dust exposure were identified in the current study when examining the outcome of all pregnancies, when assessing cumulative exposures, or with concurrent exposure during the first pregnancy.
This study has several limitations that should be acknowledged. One is the potential for exposure misclassification that may have reduced our ability to detect true associations with miscarriage. Exposures were assigned based on linking employment history with a job-exposure matrix which was based on fiber type, major textile process type, specific process type, type of work, and year. The JEM was supplemented by historical industrial hygiene monitoring data, some of which did not span full years of the cohort’s employment and were sparse in other years. The cotton dust and endotoxin data were based on considerably larger numbers of quantitative samples than were available for other agents.
Another potential limitation is the accuracy of pregnancy outcome data. All pregnancy outcomes of women in this study were determined via self-report a mean of 26 years (range 3–55 years) after occurrence. Self-reported miscarriages have been shown to be imperfectly recalled and the recall accuracy to decrease with time 36. However, in this study 36, recall of miscarriages was over 80% within 10 years of the event, and 72% even after 30 years. Recall in our study may have decreased with time. Although validation of miscarriage is not possible with our available data, comparisons can be made to live births, which are likely recalled with a high degree of accuracy. The percent of first pregnancies ending in miscarriage remained between 6.0–7.7% for 3–29 years of recall decreasing to 5% between 30–49 years of recall. However, 11.5% of first pregnancies were miscarriages after 50 years of recall. Sociocultural factors in this population and time period supporting accurate recall of reproductive history include rates of miscarriage in this study similar to those in another textile industry population of women in China that utilized daily biomarkers to identify pregnancies and losses 37 although our rates of miscarriage are lower due to lack of biomarker pregnancy identification; under the one-child family planning policy and prior family planning policies, pregnancy occurrence was closely monitored and therefore care-seeking patterns did not play a role in the identification of pregnancies or pregnancy loss; and because all women in the study had routine access to medical care within their factory 38, inadequate access to care is not a likely factor leading to under-reporting of pregnancy outcomes. Furthermore, it is unlikely that any under-reporting of outcome of pregnancy would be related to any of the occupational exposures of interest, so any effect of under ascertainment of miscarriage would only obscure true associations, but would not lead to false relationships. It is possible that higher industrial exposures and more frequent underreporting of miscarriage may have occurred in the earlier study period years. Exposures were not found to vary significantly by decade of time, and thus time-dependent exposures were not included in the job-exposure matrices 18. Underascertainment of miscarriage is possible due to the self-reported nature of these data. The current study found a miscarriage rate of 6% for the first pregnancy. A much lower miscarriage rate in subsequent pregnancies is due to the high rate of termination of pregnancy, in which some of those pregnancies would have gone on to become miscarriages. Although information on timing and dates was collected, self-reported pregnancy outcome data did not allow for a main analysis using a survival time approach. When sensitivity analyses were conducted using survival analysis techniques, no differences in conclusions were found.
High rates of termination of pregnancy were identified in this cohort of women. Termination of pregnancy is common among primiparous women, particularly after the implementation of family planning policies in 1979. Termination of pregnancy occurred at a mean of 6.7 gestational weeks, while miscarriage occurred at a mean of 10.1 weeks. The analysis considering all reproductive events may have some bias due to the high rates of termination of pregnancy in primiparous women. However the analyses considering the first pregnancy alone are not limited by this bias yet have similar conclusions, supporting the conclusions from our primary analyses considering all pregnancies.
Our results associated with spousal smoking are unexpected. Spousal (male) smoking status was determined retrospectively by asking the woman and thus may be under-reported. Study participants and their spouses were considered smokers if they had ever had more than 100 cigarettes in their lifetime. As expected, prevalence of spousal smoking is lowest for the earlier study years and increased over time.
Women in this current study were control subjects from prior nested case-control studies of breast disease within the larger cohort of 267,400 women textile workers. The response rates for both this study and the parent study were high, reducing the potential for biased participation by infertility status. The high participation rate was due to the extensive partnership with the textile industry thus allowing the researchers access to women and records, cultural norms supporting participation, and hiring of retired nurses and physician assistants who formerly worked in these factories as study recruiters. These control subjects completed a detailed reproductive history questionnaire for each pregnancy; this level of detail was not available on the baseline questionnaire for all randomized trial participants. However, using available information from the baseline questionnaire on outcome of the first pregnancy and total number of miscarriage experienced, no increased risk of breast cancer was identified with miscarriage history among these 267,400 women (OR 1.08, 95% CI: 0.87–1.34) compared to women who never had a miscarriage, nor was a trend with increasing number of miscarriages identified (p=0.90) 17. Thus bias in the results examining associations between miscarriage outcome and textile industry exposures due to the selection method of this cohort is not expected.
This study also had some notable strengths, including availability of detailed reproductive history data for each woman and complete work history data in the textile industry, and data on potential confounders. Exposures were assigned by using written STIB personnel records for each woman, and very little misclassification is expected to have occurred because work history was found in STIB factories for 99% of women. Women also tended to be stable in a job position, with 19% of women ever holding only one job and 86% holding three or fewer jobs over her entire career. Confidence in the validity of the observed elevated risks in relation to synthetic fibers is enhanced by the lack of an elevated risk in women exposed to the same substances only while not pregnant. Use of personal protective equipment has also been uncommon historically, thus personal exposures did not vary significantly from the process area sampling used to inform the job-exposure matrices.
One of the most important features of this group of women is the virtual lack of potential bias due to more fertile workers preferentially leaving the workplace (infertile worker bias). Because of the government’s universal employment policy, all pregnant women worked until shortly before the due date, then returned to the same job position after maternity leave. Thus, the cohort enabled analyses of workplace exposures and miscarriage risks free of potential biases in most other occupational settings.
This study of occupational miscarriage risks in the textile industry identified elevations in risk with exposure to synthetic fibers and a mix of synthetic and natural fibers. This study offered an opportunity to study exposures that are relatively widespread worldwide and which occur with particularly high frequency and intensity in this industry. In conjunction with extensive health and exposure information, this study was able to quantitatively evaluate associations between miscarriage and occupational exposures and enable inference to other occupationally-exposed women. Specific causative agents were not readily identifiable, and should be the subject of additional studies.
This research was supported by the grant R01 CA80180 from the US National Cancer Institute, R01OH008149 from the US National Institute for Occupational Safety and Health, and by the training grant T32ES07262 from the US National Institute of Environmental Health Sciences. The authors would like to thank Drs. Fan Liang Chen, Yong Wei Hu, Guan Lin Zhao, and Lei Da Pan for their ongoing support. The authors thank Wen Wan Wang for project management; Drs. He Lian Dai, Zhu Ming Wang, A Zhen Qi, Zia Ming Wang, Wei Ping Xiang, and Yu Fang Li for their diligence in collection of factory information; the 30 field workers for the collection of women’s personnel records; and Georgia Green, Shirley Zhang, Richard Gandolfo, and Ted Grichuhin for technical and administrative support.
The Corresponding Author has the right to grant on behalf of all authors and does grant on behalf of all authors, an exclusive licence (or non exclusive for government employees) on a worldwide basis to the BMJ Publishing Group Ltd and its Licensees to permit this article to be published in Occupational and Environmental Medicine editions and any other BMJPGL products to exploit all subsidiary rights, as set out in our licence (http://oembmjjournals.com/ifora/licence.pdf).