Materials and Methods
Study Population
Data were collected from 3 large prospective cohort studies: the Nurses' Health Study (NHS), the Nurses' Health Study 2 (NHS2), and the Health Professionals Follow-Up Study (HPFS). The NHS began in 1976 with 121,701 registered nurses aged 30–55 years, and the NHS2 began in 1989 with a cohort of 116,686 registered nurses aged 25–42 years. Data were initially limited to participants from 11 and 14 states, respectively, but now include participants in every US state. The HPFS began in 1986 and is an all-male cohort consisting of 51,529 US men aged 40–75 years working in various health-related professions. In all 3 cohorts, information on lifestyle habits and disease history is collected via biennial questionnaires.
For this study, follow-up began with collection of data on skin cancer risk and phenotype in the respective cohorts with 28, 20, and 22 years of follow-up, respectively (NHS, 1980–2008; NHS2, 1989–2009; and HPFS, 1986–2008). Data on state of residence at birth, at age 15 years, and at age 30 years were collected in 1992. Those who died during the follow-up period were excluded as were those with a diagnosis of invasive melanoma, squamous cell carcinoma, basal cell carcinoma, or any other cancer at study inception or during follow-up. Those with a history of MMIS prior to enrollment were also excluded. Because melanoma is rare in nonwhite populations and the cohorts are each approximately 97% Caucasian (reflecting the ethnic background of registered nurses and male health professionals nationally at the time of cohort inception), nonwhite participants were also excluded from analysis. Appropriate research approval for institutional human studies was obtained at Brigham and Women's Hospital (Boston, Massachusetts).
Case Ascertainment
Incident cases of MMIS were self-reported by participants via the biennial questionnaires during each 2-year cycle. All cases were then confirmed by study physicians through acquisition and review of patient medical records and primary pathology reports of the lesions. Cases were subsequently categorized as lentigo maligna and nonlentigo maligna (superficial spreading) type MMIS. Only pathologically confirmed cases were included for analysis. For mortality data, primary death certificates were obtained and reviewed by physicians for confirmation of melanoma-related death.
Exposure Assessment
All data on risk factors, states of residence, and exposures were collected via the biennial questionnaires. Questionnaires are mailed to each participant, and for each cycle, average follow-up has been more than 90%. Across all 3 cohorts, the following data were collected: 1) family history of malignant melanoma in first-degree relatives; 2) the number of nevi measuring ≥3 mm on an extremity; 3) natural hair color at age 21 years; 4) skin burning reaction after ≥2 hours of bright sun exposure during childhood/adolescence; and 5) the number of lifetime severe or blistering sunburns. Family and personal disease history are updated with each questionnaire cycle. For the nevus count on an extremity, the left arm (shoulder to wrist) was used in the NHS, the bilateral lower legs (knee to ankle) were used in the NHS2, and the bilateral forearms (wrist to elbow) were used in the HPFS. In the NHS cohort only, tanning ability was assessed by asking what kind of tan developed after repeated sun exposures (e.g., a 2-week vacation outdoors) during childhood or adolescence. This information was combined with participants' responses regarding susceptibility to burn to determine the Fitzpatrick skin type of NHS participants.
Fifty US states and Washington, DC, were stratified according to the average ultraviolet index in the month of August as determined by the National Oceanic and Atmospheric Administration. By using residential data from the study participants at birth, at age 15 years, and at age 30 years, we categorized states into low, medium, and high indices of ≤5, 6, and ≥7, respectively.
Statistical Analysis
Each participant contributed person-time from the date of the collection of phenotypic data to the date of incident MMIS or the end of the follow-up period, whichever came first. Categories representing the lowest perceived risk of MMIS (i.e., no burning reaction, no family history, lowest ultraviolet index) were used as referents except in the case of hair color, in which light brown was used because it is the most common hair color. For both age-adjusted (5-year categories) and multivariate regression models, variables were modeled as dichotomous or categorical. Family history of melanoma was a dichotomous variable (yes/no), and categorical variables were based on the questions and answers from the original questionnaires.
Age-adjusted relative risk, multivariate relative risk, and 95% confidence intervals were calculated by using Cox proportional hazards regression to adjust for age and other covariates including family history, number of nevi, number of sunburns, skin reaction to sunlight, and natural hair color at age 21 years for each cohort. Models were adjusted for calendar year because the time parameter and person-time for each participant were calculated from the date of return of the questionnaires (1980 for the NHS, 1989 for the NHS2, and 1986 for the HPFS) to the first endpoint (incident MMIS, death, or the end of the follow-up period). All multivariate analyses were rerun by controlling for tanning response in the NHS cohort with no appreciable changes in relative risks (data not shown). A meta-analysis was then conducted by using a random effects model and P values for trend, and Q statistics for heterogeneity were calculated. We used SAS, version 9.2, software for all analyses (SAS Institute, Inc., Cary, North Carolina).