Men who did and did not use tetracycline were compared by calculating age-standardized means and proportions of covariates of interest by tetracycline use. Covariates of interest included (i) factors previously observed to be associated with prostate cancer incidence or progression in the HPFS cohort (race/ethnicity, family history of prostate cancer, height, cigarette smoking in the past 10 years, intakes of total energy, alcohol, tomato sauce, red meat, fish, calcium, fructose, α-linolenic acid and vitamin E, zinc supplementation, vigorous physical activity, having had a vasectomy and a history of diabetes)13, 14, 15, 16, 17, 18, 19, 20, 21; (ii) factors or other medical conditions potentially associated with long-term tetracycline use (regular nonsteroidal antiinflammatory drug (NSAID) use, and histories of periodontal disease and prostatitis), (iii) purported surrogate markers of androgenicity (vertex baldness, and body habitus, vigorous physical activity and ejaculation frequency in adolescence and early adulthood), (iv) other factors hypothesized to be associated with acne development, aggravation or treatment (intakes of milk, hard cheese, butter, milk shakes, ice cream, cookies and fried potatoes in high school, residence in a Southern state or territory in adolescence or early adulthood, amount of time spent outdoors in a swimsuit in the summer as a teenager, and lifetime number of blistering facial sunburns) and (v) other early life factors (body mass index, cigarette smoking and alcohol consumption in adolescence and early adulthood and histories of gonorrhea or syphilis). Food and sun exposure as a teenager was explored because certain foods (carbohydrates, such as bread, potatoes, pastries, candy and sugar, fatty or greasy foods, dairy and pork products, shellfish, nuts, spicy foods and chocolate) were commonly believed to cause or exacerbate acne, and natural sunlight was believed to improve acne.11, 12, 22 Age-adjusted (5-year age intervals) associations between tetracycline use and prostate cancer were explored using Mantel-Haenszel rate ratios. Person-time was calculated from the month of return of the 1992 questionnaire to the month of prostate cancer diagnosis, death or end of the analysis period on January 31, 2002. Multivariable-adjusted associations between tetracycline use and prostate cancer were investigated using Cox proportional hazards regression. All models included a term for missing duration of tetracycline use. To investigate the influence of this missing information, sensitivity analyses were performed by including all participants with missing duration of use in either the group of men who reported less than 4-years of tetracycline use or the group of men who reported 4 or more years of tetracycline use. Similar inferences were observed as in analyses that used a separate missing term. Age (1-month intervals) and calendar time (2-year intervals) were controlled for as stratification variables in all regression models. Confounding by the aforementioned covariates was explored by adding each covariate individually and in combination to the regression model, and comparing to univariable results. As the number of exposed cases was low and none of the explored covariates altered the magnitude or significance of the main findings, only known risk factors for prostate cancer (age, race/ethnicity and family history of prostate cancer) were retained in the final multivariable model.
Possible detection bias was explored by restricting the analyses to men who reported prostate cancer screening, either a routine prostate specific antigen test or digital rectal examination. The specificity of long-term tetracycline use for acne was explored in 3 different ways. First, we explored the age distribution of men who reported 4 or more years of tetracycline use to determine whether this distribution was consistent with the availability of tetracycline (introduced in 1948 and more widely used for acne in the early 1950s9), and the maximum age of acne persistence in men (usually 25-years of age, although 7% of cases may still have acne up to age 45 Ref.23). Second, we performed analyses restricted to men 25 years of age or less in the early 1950s who would have been more likely than older men to receive tetracycline for their acne, and men with no reported histories of periodontal disease or prostatitis, as these might have also been reasons for long-term tetracycline use. Finally, we tabulated reasons for tetracycline use by duration of use in the sample of participants who completed the supplemental questionnaire in 2006 regarding indications for tetracycline use.