The WHI recruited women who met the following eligibility criteria: between 50 and 79 years of age, postmenopausal (no menstrual period for at least 6 months if aged 55+ years and 12 months if aged 50–54 years), and intending to reside in the area for at least 3 years. Women were recruited from 40 US clinical centers (in 24 states and the District of Columbia) between 1993 and 1998 through a mass mailing of a recruitment brochure. The study included a “Clinical Trial” cohort (n = 68,132) with subjects enrolled in four different studies: Estrogen-alone trial, Estrogen-plus-Progestin trial, Dietary Modification trial, and Calcium and Vitamin D trial. Each randomized controlled trial has its own exclusionary criteria involving safety, adherence, and retention concerns. Women ineligible or unwilling to join the clinical trials were invited to join the “Observational Study” cohort (n = 93,676). Detailed eligibility criteria and recruitment methods have been previously published [14]. Human subject review committees at all participating sites approved WHI protocols. Participants provided written informed consent. A total of 118,097 women were linked to Medicare data. Women with psoriasis prior to WHI enrollment and those who were not followed long enough for a 2-year look-back period were excluded for a remainder of 112,184 women. Those missing smoking or alcohol habit data (n = 5340) were excluded for a final analytical cohort of 106,844 postmenopausal women.

Self-administered questionnaires at baseline were used to obtain information on demographics, medical history, and lifestyle behaviors such as smoking status, age (of smoking onset and cessation), cigarette pack-years, and the frequency and serving size of alcohol consumption (information derived from “Form 34—Personal Habits Questionnaire” and “Form 60—Food Questionnaire”). The observational study arm had smoking status on years 3–8 of follow-up and alcohol status on years 3 and 6 of follow-up.

Psoriasis was identified using fee-for-service Medicare claims containing the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis codes 696.0 (psoriatic arthropathy) and 696.1 (other psoriasis) as both skin and joint involvement are manifestations of the same disease [15, 16]. International Classification of Diseases, Ninth Revision, Clinical Modification codes not entered by either a dermatologist or rheumatologist were excluded. A study of a Northern California population from 1996 to 2009 showed that psoriasis ICD-9-CM codes reported at least once by a dermatologist have a positive predictive value of 89% (95% confidence interval [CI] 79–95) [17]. A 2-year look-back period or washout period was also implemented to prevent the misclassification of prevalent cases as incidence [18, 19]. Participants entered the risk set upon completion of the 2-year look-back period. A left truncation (delayed entry) was applied to remove the prevalent cases of subjects (all subjects who had psoriasis/psoriatic arthritis at the start of the study), allowing us to fulfill the assumption that none of the subjects had psoriasis at the start of the study. Event times were censored at the date of psoriasis diagnosis, the date of no longer enrolled in fee-for-service Part A+B, or 31 December, 2014, whichever came first.

Cox proportional hazards regression models were used to estimate hazard ratios (HRs) and 95% CIs for the association of alcohol consumption with psoriasis over the cumulative 21-year follow-up. The initial model was based on common practice when using WHI data [18, 19]. The initial model was stratified by age group (< 60, 60–64, 65–69, 70–74, 75–79, ≥ 80 years), by WHI trial groups (Observational Study vs Clinical Trial cohorts), and by enrollment into each of the four individual trials (i.e., Estrogen-alone trial, Estrogen-plus-Progestin trial, Dietary Modification trial, and Calcium and Vitamin D trial). The second model additionally included demographic factors such as ethnicity (White, Asian or Pacific Islander, Black or African-American, Hispanic or Latino, Other), income in US dollars (< 20k, 20–35k, 35–50k, 50–75k, 75–100k, > 100k), body mass index (normal, overweight, obese), and history of non-melanoma skin cancer (no, yes). These factors have been associated with both alcohol habits and psoriasis incidence. The third and final “full” model adjusted for cigarette smoking (never smoker, < 5 pack-years, ≥ 5 and < 20 pack-years, ≥ 20 pack-years), which was a major confounder associated with both alcohol and psoriasis. Additional sensitivity analyses were performed to ensure findings were repeated and consistent by limiting the cohort to Caucasian women (n = 88,937, the majority with 84.8%), excluding women who had stopped drinking at baseline because of health reasons (n = 3764), excluding women younger than 65 years of age (n = 4107; most of whom qualified for Medicare through disability), and excluding women whose alcohol habits had changed at years 3 or 6 of follow-up (10,404 out of 43,502 = 23.9%, observational study only).

The proportional hazards assumption was tested with Schoenfeld residuals, and no violation of the proportionality assumption was found. The risk of psoriasis by subcategories of alcohol (wine, beer, liquor: drink frequency versus past drinkers and never drinkers) was analyzed among “Never Smokers” stratification (n = 54,751, n-psoriasis = 1263). This strategy provides insight to the different subcategories of alcohol while isolating their effect on psoriasis from cigarette smoking (as well as any potential interaction effects). Binning servings of alcohol per week into categories leads to information loss; therefore, we also tested alcohol consumption as a continuous variable (servings/week). We tested for linear associations without applying any transformation. Then, we also tested for nonlinear associations using penalized splines (a nonparametric method used to fit smooth curves along data points to test for nonlinear associations) with a standard 4 degrees of freedom.