We conducted a case-control study of prostate cancer and hormones and alcohol intake (the PROMEN STUDY) in Erie and Niagara Counties, NY, USA, between December 1998 and April 2001. The methods for this study have been previously described in detail . Participants provided informed consent; the Institutional Review Board of the University at Buffalo, School of Medicine and Biomedical Science, and each of the participating hospitals approved the procedures for the protection of human subjects recruited for the study.
Cases were men aged 45 to 85 years with incident, primary, histologically confirmed prostate cancer. Men with a previous history of cancer (except non-melanoma skin cancer), or already on hormonal or chemotherapy treatment (current or in the 6 months prior to diagnosis), as well as those affected by chronic or acute liver diseases, were excluded. Cases aged 35–65 years were also required to have a driver's license, because we used driver's license records to identify age matched controls.
During the study period, 504 men were identified with incident prostate cancer. Of these, 336 men did not meet the eligibility criteria; we invited the remaining 163 patients to participate in the PROMEN study. After being contacted, 50 men refused to participate resulting in a participation rate of 70%. Ninety-six patients had complete data for the variables of interest.
Controls aged between 35 and 65 years were selected from a list of individuals holding a New York State driver's license and residing in Erie and Niagara Counties. Those aged 65 and over were selected from the rolls of the Health Care Financial Administration. As with cases, men on hormonal treatment (current or in the 6 months prior the diagnosis), or diagnosed with metabolic or endocrine disease were excluded, as well as participants with a previous story of cancer other than non-melanoma skin cancer. Since it is well known that latent prostatic carcinoma has a high prevalence in men over 50 [22, 23], we evaluated prostate specific antigen (PSA) in the blood samples obtained from controls. Controls found to have a PSA value higher than 4 ng/ml were excluded from the control group, in accordance with the criterion established by the American Cancer Society Prostate Cancer Detection Project  until the completion of further diagnostic procedures to clarify their true case-control status.
During the study period, 1373 potential controls were contacted. One hundred and seventy nine of these individuals were deceased or were too ill to participate, 293 did not meet the eligibility criteria and we were not able to contact 272 persons. We identified eight prostate cancer cases as a result of PSA determination in subjects who initially were recruited as controls. Three hundred and seventeen of the remaining 513 subjects (60%) were enrolled and interviewed: 304 had complete data for analysis.
Extensive data on demographics, smoking history, alcohol consumption, and other study variables were collected by trained interviewers during in-person computer-assisted interviews  and with self-administered questionnaires. Height, weight, waist and hip circumferences were measured by trained technicians using a standardized protocol. Body mass index (BMI) was calculated as weight in kilograms divided by square of the height in meters (kg/m2). Waist to hip ratio (WHR) was calculated as waist circumference divided by hip circumference.
Detailed information on alcohol consumption throughout the lifetime was collected using the Cognitive Lifetime Drinking History [26, 27]. Prior to the interview, participants completed a lifetime events calendar on which they recorded the date and their age when significant events in their life occurred. The calendar was used during the interview to help them remember what they were doing during specified periods of their lives and whether drinking alcohol was involved. Participants reported the age when they started drinking alcohol regularly (at least once a month for six months) and when their drinking changed over the years. When changes were reported, participants were asked whether they continued regular drinking; if not, they were asked if they ever resumed regular drinking. Using this information, we defined intervals during each participant's life when drinking patterns were relatively homogeneous and computed the total number of drinking years and the total number of abstinent years. Lists of alcoholic beverages, beer, wine, wine coolers, and liquor, and models of glasses and bottles were used to help respondents recall what beverages they drank over their lifetimes; their usual drink size of each beverage; and whether drink size changed over their lifetimes. This provides information used to: (1) calculate absolute alcohol intakes and (2) tailor the computer-assisted interview to the each respondent's drinking history. Patterns of drinking were ascertained for intervals during which respondents drank weekly or more often by asking how often respondents drank on Fridays, Saturdays, Sundays, and weekdays, and how many drinks they usually had on each. For intervals during which respondents drank less often than weekly, they were asked standard quantity and frequency questions. Quantity and frequency for times when they drank more than usual were assessed for all intervals, as was the frequency of intoxication; the proportion of drinks they had with meals/snack/without eating; and the proportion of drinks from beer, wine, wine coolers, and liquor.
Drinks per interval was estimated by multiplying quantity by frequency for days of the week and more than usual and adding. Drinks per interval was translated into ounces of ethanol per interval based on the proportion of drinks represented by specific beverages, respondents' beverage-specific drink size in ounces, and factors representing the average percent per ounce of absolute alcohol for a given beverage to estimates of drinks per interval. Factors used were 0.048, 0.12, 0.04 and 0.40, for beer, wine, wine cooler and hard liquor, respectively. These estimates were summed across drinking intervals to yield lifetime totals.
We considered several variables in these analyses: total number of years alcohol was consumed, number of drinks per day during the drinking years (total number of drinks/total number of days in drinking years), number of drinks per drinking day (total number of drinks/total number of days on which alcohol was consumed in drinking years), total lifetime ounces of ethanol and beverage-specific total lifetime ounces of ethanol. Because few participants consumed wine coolers, wine and wine coolers were combined. A drink was defined as 12 ounces of beer, 5 ounces of wine, and 1.5 ounces of liquor.
Statistical analyses were conducted using SPSS for Windows version 11.0. Differences between cases and controls in demographic characteristics and alcohol consumption were assessed using t-tests for continuous variables and χ2 for categorical variables. Lifetime abstainers, defined as those subjects who had less than 12 drinks in any one year over their lifetimes, were excluded from our analyses. The biological and social differences between lifetime abstainers and both former and current drinkers [28, 29] and the very low number of these subjects in our sample (5 cases and 11 controls) represent the reasons for their exclusion from our analyses. Our final sample size for analysis included 88 cases and 272 controls.
In analyses of risk associated with lifetime alcohol intake, tertiles of total and beverage specific ounces and total drinking years were computed based on the distribution in the controls. For the beverage specific analyses, non-drinkers were those respondents not consuming that particular alcoholic beverage. For risk associated with drinks per day and drinks per drinking day, we categorized consumption as two drinks or less per day and greater than two drinks per day. Odds ratios (OR) and 95% confidence intervals (CI) for risk of prostate cancer associated with alcohol consumption were computed using unconditional logistic regression adjusting for age, cigarette smoking status, education, body mass index (BMI), and waist to hip ratio (WHRATIO). The beverage specific analyses were further mutually adjusted for the other beverages.