Lifetime occupational history and risk of endometriosis

Objectives Endometriosis is the presence of functioning endometrial glands and stroma outside the uterine cavity, most often in the pelvic peritoneal cavity. Women with endometriosis commonly have dysmenorrhea, dyspareunia, pain, menorrhagia, and/or metrorrhagia. Disease complications can include adhesions, chronic pain, and infertility. In this exploratory case-control study, we investigated the relationship between lifetime occupational history and surgically confirmed endometriosis in a population-based sample. Methods We conducted interviews with participants, all reproductive-aged female members of a large health-maintenance organization who were first diagnosed with surgically confirmed endometriosis between April 1, 1996 and March 31, 2001. Interviews were also conducted with randomly selected controls, reproductive-aged female enrollees of the same organization from the same time period. Each reported job was coded using US Census Occupations and Industries codes, and classified into categories. We used unconditional logistic regression to compare having worked in a given job class with never having done so. Results Our study found that an increased risk of endometriosis was associated with having worked as a flight attendant, service station attendant, or health worker, particularly as a nurse or health aide (flight atten -dant: odds ratio (OR) 9.80, 95% CI 1.08–89.02; service station attendant: OR 5.77, 95% CI 1.03–32.43; health worker: OR 1.49, 95% CI 1.03–2.15). Income and education did not make a difference in the OR estimates for the occupations examined. This exploratory study suggested that there might be an associated risk of endometriosis for those women who have worked as a flight attendant, service station attendant, or health worker, particularly a nurse.

Endometriosis, which affects 5-10% of women of reproductive age in the United States, is the presence of functioning endometrial glands and stroma outside the uterine cavity, most often in the pelvic peritoneal cavity (1). It is a symptomatic disease that commonly features dysmenorrhea, dyspareunia, pain, menorrhagia, and/or metrorrhagia; complications can include adhesions, chronic pain, and infertility. Endometriosis is associated with an increased overall cancer risk (2), with particular elevation in risk of ovarian cancer (3)(4)(5)(6).
Endometriosis has been studied as a potential outcome of occupational exposure to dioxins and included as a secondary outcome in general studies of reduced fertility among workers in certain industries and occupations. These studies have found an increased risk of endometriosis-associated infertility among workers exposed to formaldehyde (7), video display terminals (8), chemical dusts (8), or organic solvents (7,8), and among workers in agricultural industries and occupations, in particular farm workers (9). To our knowledge there has been no examination to date of the full spectrum of occupational risk of endometriotic disease. In the United States, 72% of women age 25-64 years work in the civilian labor force, and the majority of these work full-time all year (10). If occupational risks for endometriosis exist, such exposures will affect many women. This analysis explored the relationship between occupation and endometriosis in a large population-based case-control study.

Study design
As described elsewhere (11), we opened participation in our study to all female members of Group Health Cooperative, a large health-maintenance organization in Washington State, USA, age 18-49 years and first diagnosed with surgically confirmed endometriosis between April 1, 1996 andMarch 31, 2001 (N=467). The reference dates of potential cases were the month and year of the first Group Health Cooperative visit for symptoms ultimately leading to the diagnosis of endometriosis. Potential controls (N=1016) were randomly selected from a list of women enrolled in Group Health Cooperative during the same time period, frequencymatched on a five-year age group, and assigned reference dates corresponding to the distribution of case reference dates.
A total of 341 (73.0%) subjects and 742 (73.0%) controls participated in the study. A face-to-face or telephone interview using a structured questionnaire collected information concerning exposures prior to the reference date, as well as demographics, prior medical conditions, reproductive history, and other characteristics. Participants' in-and out-patient medical records were reviewed and abstracted for type and severity of symptoms (eg, dysmenorrhea, menorrhagia), and endometriosis characteristics (eg, dimensions, presence of adhesions). Case definitions were further refined and participants whose disease did not meet the Holt-Weiss standards for definite or probable disease were excluded from this analysis (12). Briefly, under these criteria, ovarian endometriomas, pelvic endometriotic lesions over 5 mm deep, and pelvic endometriotic lesions with adhesions not attributable to other causes are classed as "definite disease", while other endometriotic implants with at least one major endometriosis symptom (eg, infertility, moderate to severe dysmenorrhea, moderate to severe dyspareunia, or moderate to severe pelvic pain) are classed as "probable disease". Therefore, cases without surgical evidence of disease and asymptomatic cases with superficial or ambiguous pelvic lesions (N=12) were excluded from the study. Additionally, three cases of extrapelvic scar endometriosis were excluded.
Subjects (N=13) and controls (N=14) who reported a previous surgically confirmed endometriosis diagnosis were excluded from this analysis as well. After all exclusions, 313 cases and 727 controls remained for this analysis. This study was approved by the institutional review boards of Fred Hutchinson Cancer Research Center and Group Health Cooperative. Each participant gave written informed consent to participate in the study and was compensated $20 for her time.

Exposure classification
All participants were asked about each paid job they had held, part-time or full-time -for six months or longer -between 18 years of age and the reference date. Information collected included the job title, type of business or industry, average hours worked per week, and number of years worked at each job. Jobs were coded using the 1980 Census -Occupations and Industries codes (13,14). These codes were grouped into 56 categories using the Schnitzer et al method (15). To improve statistical power, multiple Schnitzer et al categories were merged to form two broader classes: "wood-and paper-related jobs" and "health workers". The Schnitzer et al categories "forestry and logging workers", "sawmill workers", "paper workers" and "carpenters/woodworkers", which respectively had 7, 4, 3, and 6 participants, were combined into a new category "wood-and paperrelated jobs" (N=19). [Numbers do not sum because a respondent could work more than one job class during her history.] A new category, "health workers", consisted of the Schnitzer et al categories "dentists/dental technicians" (N=13, all of whom were dental technicians), "physicians/surgeons" (N=3), and "nurses/health technicians" (N=149). Within the "health worker" category, subcategories were created for (i) physicians/surgeons (job code 084), (ii) registered nurses/licensed practical nurses (job codes 095 and 207), (iii) radiological technician (job code 206), (iv) nursing/health aides (job codes 446 and 447) and (v) all other health workers. The Schnitzer et al category "aircraft operators" (occupation codes 226 and 465), was renamed "flight attendants" after examination of the individual jobs revealed that all of the respondents in this category listed their jobs as flight attendants (occupation code 465). These changes reflect the adaptation of the Schnitzer et al tool to an exclusively female population of workers, rather than the mixed-gender group originally used by Schnitzer et al, or the all-male population of the Canadian system from which the instrument emerged.

Statistical analysis
We used unconditional logistic regression to calculate odds ratios (OR) and 95% confidence intervals (95% CI) for the association between endometriosis and occupation using STATA 8.2 (Stata Corp LP, College Station, TX, USA). Those who had ever worked in an occupation in a Schnitzer et al category were compared with all those who had never worked in that category. Using a low-stringency standard of a<0.15, occupational groups that were significantly associated with endometriosis were examined more closely, and divided or further aggregated to obtain reasonably sized and meaningful categories for further analysis.

Marino et al
Potential confounding factors considered in the creation of logistic regression models were race or ethnicity, household income, education, marital status, gravidity, parity, and use of alcohol, cigarettes, and oral contraceptives. Each potential confounder, treated categorically, was evaluated individually in a model containing each re-aggregated occupational group and the frequency-matching variables (age and reference year), to determine which, if any, changed the estimate of the association between the occupation and endometriosis by 10% or more. Those confounders were then included in a multivariate model to determine a final, conservatively adjusted estimate of association.
As women who seek treatment for infertility may have endometriosis discovered incidentally as part of the diagnostic process, rather than the result of symptomatic disease, we also considered separately cases who reported seeking care only for reasons other than infertility (N=260). Presentation of endometriosis was determined by the case participants' answers to the question, "on or before [the reference date], what was the reason you first visited a doctor or other health care practitioner that led to your recent diagnosis of endometriosis?" and their sexual intercourse histories. Those who listed infertility as one reason (N=36) or the only reason (N=17) for their initial visit were classified as having infertilityassociated endometriosis. Pre-menopausal women who reported 12 or more consecutive months of unprotected sexual intercourse with a male partner without a pregnancy immediately before their reference date, and who had no other reason for their initial visit, were also classed as having infertility-associated endometriosis.

Results
Demographic characteristics are shown in table 1. Cases and controls were similarly distributed in terms of age, race or ethnicity, marital status, education, income, and use of oral contraceptives. Cases had fewer pregnancies and fewer births than controls. Cases were more likely than controls to use tobacco or alcohol at the reference date.
The distribution of Schnitzer et al's occupational categories among cases and controls is reported in table 2. Under the non-stringent criterion of a<=0.15, the categories "biological scientists", "teachers, librarians", "aircraft operators", and "service station attendants" were significantly associated with endometriosis case status after adjustment for matching variables age and reference year. These categories were further examined in multivariate models, as were the categories that were reconstructed to add meaning and power (table 3). Potential confounders that made a 10% or greater difference in any "ever/never" risk estimate were alcohol use, oral contraceptive use, gravidity, and nulliparity. All "ever/never" models were thus adjusted for these variables as well as for age and reference year. In these multivariate models, health workers, service station attendants, and flight attendants were Lifetime occupational history and risk of endometriosis   . Working in an occupation for five years or longer did not appear to strengthen negative or positive associations with that occupation.
The associations between selected occupations and case status by presentation are shown in table 5. In general, associations between occupation and case status were more pronounced among those presenting for reasons other than infertility than among all cases. Those whose primary disease manifestation was not infertility were more likely to have worked as service station attendants (OR 5.98, 95% CI 1.07-33.45), flight attendants (OR 10.18, 95% CI 1.12-92.14), or health workers (OR 1.53, 95% CI 1.06-2.22), including registered nurses/licensed practical nurses (OR 1.80, 95% CI 0.97-3.34), and nursing/health aides (OR 1.60, 95% CI 1.00-2.55).

Discussion
Our exploratory study suggested that having ever worked as a flight attendant, service station attendant, or health worker, particularly as a nurse, may be associated with increased endometriosis risk. There was some indication that having worked as a teacher or librarian is associated with a decreased risk, although the association was not statistically significant. Some potential exposures that were found among one or more of the vulnerable jobs, but uncommon in school or library settings, included (i) radiation (health workers and flight attendants), (ii) fuel and exhaust components (service station and flight attendants), (iii) night work (all), and (iv) circadian/sleep disruption (all). Another study of this population suggested that endometriosis may be related to night and evening shift work (11), but small numbers prevented an examination of risk by job title in that analysis.
Measuring exposures is preferable to using job titles in occupational health research (16). However, since little is known about occupation and endometriosis, and the present study did not allow the measurement of a spectrum of exposures, using job categories provided an alternative exploratory approach to generate initial occupational hypotheses. Occupations include a broad array of exposures and may also capture some aspects of socioeconomic standing. In this study, income and education did not affect the OR estimates for the occupations examined, suggesting that physical exposures are more likely than sociodemography to be responsible for any associations seen. We also found that an occupational duration of ≥5 years did not appear to strengthen the associations within the occupational classes studied. The absence of a duration dose response might suggest that a shared chemical exposure is not responsible for the association, but larger studies are required to address this hypothesis more definitively. We did not find that women with endometriosis were less likely to work than women without the disease, but if women with severe symptoms were less likely to belong to the health-maintenance organization under study (perhaps they were not working at all), associations between disease risk and occupations would be attenuated.
The most commonly studied compounds likely to be occupationally related to endometriosis are dioxins and other organochlorine compounds. Associations between endometriosis and these substances have been controversial, with most studies involving simultaneous measurement of body burden and disease status (17)(18)(19), or measurements after industrial disasters (20), rather than routine occupational exposures. Previous occupational studies of infertility in relation to certain industries and occupations have evaluated endometriosis as a secondary outcome. Taskinen et al found substantially elevated endometriosis risks among woodworkers exposed to formaldehyde and organic solvents (formaldehyde OR 4.5, 95% CI 1.0-20.0; organic solvents OR 14.7, 95% CI 3.1-70.0), but the presence of endometriosis was ascertained from self-report in that study, and temporal relationships between exposure and outcome could not be determined (7). The suggestive but non-significant association we found between endometriosis and wood/paper-related jobs may support these findings.
In contrast to our results, Fuortes et al found strong associations between endometriosis and working in agricultural industries (OR 6.9, 95% CI 1.6-30.4) and occupations (OR 8.1, 95% CI 1.3-52.5) when comparing women with medically confirmed infertility caused by endometriosis to postpartum women (9). The present work compared women with surgically confirmed endometriosis (with and without infertility) to a general population of female members of a health-maintenance organization. The different control populations affect the interpretation of the OR -one reflects the relative risk of exposure among those who are infertile due to endometriosis versus those who are fertile, while the other reflects exposure among those with and without endometriosis irrespective of fertility. One possible explanation of the difference in findings is that endometriosis might be equally likely to occur among women who did or did not work in agricultural occupations, but more likely to be diagnosed as the cause of infertility, or to cause infertility, among those who had agricultural occupations.
This was the first systematic survey of an occupational risk of endometriosis. This study had a number of important strengths. The study design allowed nearly complete case ascertainment. Instead of the potentially biased samples obtained by recruiting patients from gynecology or infertility clinics, this design enabled the full spectrum of disease diagnosed within a large health-maintenance organization to be represented. For this analysis, we excluded cases without surgical confirmation, in contrast to previous occupational studies that have relied on self-report. We also excluded all those who reported prior surgically confirmed endometriosis to capture the first diagnosis and look at exposures in the correct timeframe. Experts evaluated disease features and its course directly from medical records rather than relying on self-report, and we used a well-defined set of criteria to evaluate the certainty of diagnosis, enhancing the reproducibility of our findings. We did rely, however, on self-reported occupational histories. If inaccurate, these reports could produce a biased risk estimate, particularly if cases and controls recalled occupations differently, but such histories have been found to have high validity and reliability (21), and the interview covered many topics, so occupation was