Subjects
The methods of the study have been published elsewhere [12–14]. In brief, we conducted an incident case-control study at the University Hospital of Essen, Germany. Patients diagnosed with incident primary uveal melanoma in the ophthalmic clinic at the University Hospital of Essen, Germany, between September 2002 and September 2004, aged 20-74 years at diagnosis, living in Germany, and proficient in the German language, were eligible. We recruited three different control groups. Population-based control subjects were selected from the census of the local districts and were matched to case patients by age (5-year age groups: 20-24, 25-29,...,70-74 years), sex, and region of residence. Two controls were aspired for each case. Sibling controls that were within 10 years of the age of the cases were recruited after the case interviews (aspired matching ratio 1:1). Ophthalmology controls were recruited from practices of the same ophthalmologists who had referred uveal melanoma cases to the ophthalmic clinic at the University Hospital of Essen and had to have a newly diagnosed benign disease of the eye. Here we matched controls on age (20-24, 25-29,...,70-74 years), and sex to aspire a case-control matching ratio 1:2. However, recruitment of ophthalmology controls became difficult because of lack of support from the ophthalmologists. We therefore stopped recruiting ophthalmologists' controls for incident cases during the second half of the recruitment period. Details of the inclusion and exclusion criteria of three different control groups of the RIFA study have been published elsewhere [14]. Eligible cases and controls were approached and interviewed between September 2002 and March 2005. Identical procedures were followed for cases and controls. A letter of invitation was sent and in case of nonresponse up to 10 telephone contacts were undertaken. Controls who could not be contacted because the phone number was unavailable received a reminder letter of invitation.
Exposure assessment
The exposure assessment began with a self-administered postal questionnaire. Subjects were asked about each job held for at least six months in their lifetime and additionally on their job tasks and industrial branches. Thereafter subjects underwent a standardized computer-assisted telephone interview (CATI), which took 35 minutes on average. Trained study personnel conducted standardized computer-assisted telephone interviews. The interviewers were unaware of study hypotheses and case-control status of study participants as they used/applied the structured questionnaires.
The interview topics included medical history, phenotypic characteristics, occupational sun exposure, artificial UV radiation, use of sunglasses or hats to protect against intensive sun light, UV related keratitis, smoking status, and social class. To study selected work tasks we used job-specific supplementary questionnaires to obtain details of the job tasks (e.g. cooking and food processing) and materials that were used. Each job period was coded according to the International Standard Classification of Occupations (ISCO) [15]. We classified people as exposed to an occupational category if they ever worked within this category for at least six months or more. According to ISCO, the job tasks of cooks (ISCO: 5-31.20-5-31.90) include planning, supervising, organizing, preparing and cooking foodstuffs in hotels, restaurants and other public eating places, on board ship, on passenger trains and in private households. Thus, people who are assigned to the ISCO codes of cooks do not necessarily cook. Therefore we used the information from the job-specific supplementary questionnaire to classify people as "cooked actually". We classified subjects as exposed if they had ever worked within the job task of cooks for at least six months or more according to ISCO (called ISCO-based cooks). We thereafter distinguished between ISCO-based cooks who actually cooked and ISCO-based cooks who did not cook.
This study was conducted in accordance with the German guidelines of Good Epidemiological Practice [16]. The study was approved by the ethics committee of the Medical Faculty in Essen. Informed consent was obtained from all patients and controls.
Statistical methods
The reference date for cases and matched controls was the date of diagnosis of the cases. Exposures after the reference date were not considered. We calculated response proportions by case-control status according to the definition of Slattery et al. [17]. The study sizes (cases : controls) for the three matched evaluations were: 455 cases : 827 population-based controls, 133 cases : 180 ophthalmologist controls, and 187 cases : 187 sibling controls.
We used conditional logistic regression models to estimate ORs and 95% CI taking the matching factors into account. All analyses were performed using SAS 9.1 (SAS for windows. Cary, NC: SAS Institute; 2002).