The design of the PLCO trial has been described in detail elsewhere [11]. In short, the trial followed a randomized, two-arm design to examine the effectiveness of yearly screening on cancer-specific mortality. From 1993 to 2001, men and women were enrolled in the trial. Participants were recruited from 10 U.S. cancer screening centers. Eligible participants were 55 to 74 years; not previously diagnosed with breast, prostate, lung, colorectal or ovarian cancer; had no more than one PSA test in the 3 years before randomization (starting in 1995); and had not taken finasteride (Proscar) in the past 6 months before randomization. Men underwent yearly prostate-specific antigen (PSA) tests for 5 years and annual digital rectal exams (DREs) for the first 3 years. Participants were initially followed for up to 13 years or the end of 2009, whichever came first (questionnaires are described below). In 2011, PLCO switched to a centralized follow-up system, with participants re-consented to either active follow-up, passive follow-up or refusal of all further follow-up. Active and passive follow-up allowed for linkage with state cancer registries and the National Death Index, while only active follow-up allowed for the trial to re-contact participants for further questionnaires. Active follow-up participants were re-contacted in 2013 for another questionnaire, which also included a consent question for linkage to Medicare claims and Part D prescription drug files.

At randomization a baseline questionnaire (BQ) was administered to the participants. The BQ collected information on demographics, medical history (including history of BPH), smoking, physical activity, and gender specific information. Between 2006 and 2008, a supplemental questionnaire (SQ) was administered to the participants, which collected similar information.

To be eligible for this analysis, men had to have reconsented to active follow-up, be alive in 2013 and consented to Medicare linkage at that time (Fig. 1). In addition, men had to have had at least 12 months of Medicare Part D coverage from 2008 to 2015 and had not been diagnosed with prostate cancer through the end of 2007 (Fig. 1). The original study was approved by all 10 participating centers’: University of Colorado Health Sciences Center, Lombardi Cancer Research Center of Georgetown University, Pacific Health Research Institute, Henry Ford Health System, University of Minnesota School of Public Health/Virginia L. Piper Cancer Institute, Washington University School of Medicine, University of Pittsburgh/Pittsburgh Cancer Institute/Magee-Women’s Hospital, University of Utah School of Medicine, Marshfield (Wisconsin) Medical Research and Education Foundation, and the University of Alabama at Birmingham institutional review boards and that of the National Cancer Institute [12]. No additional approval was needed for this current study as the data were publicly available.

Fig. 1.

The study flowchart depicts which men from the original Prostate, Lung, Colorectal, Ovarian Cancer Screening Trial were eligible for the current study.

With participants’ written consent, medication prescription data was obtained from the linkage of participants with Medicare Part D (prescription drug) from 2008 to 2015. From this linkage, medication use information such as drug name, number of days the medication was used, the number of prescriptions filled, and the supply quantity (days supply) was available and compiled for analysis. In addition, for those men with Medicare fee for service coverage, chronic condition flag variables (derived from algorithms that utilize procedure claims data) were available, including a variable for BPH. Men with the BPH chronic condition variable available were denoted as the chronic condition sub-cohort; for these men, BPH status could be assessed based not only on self-report.

Logistic regression (Proc logistic, version 9.4; SAS Institute Inc, Cary, NC, USA) was utilized to examine use of finasteride and dutasteride overall for the study period (2008 to 2015) while adjusting for covariates (e.g., study years, age, race/ethnicity and BPH status). BPH status was taken from the SQ if available, otherwise from the BQ. Both univariate and multivariable models were created with the outcome of any finasteride/dutasteride use. Within the chronic condition sub-cohort, a general estimating equation (GEE) model (Proc Genmod, version 9.4; SAS Institute Inc, Cary, NC, USA) was developed to examine annual trends in the use of finasteride, dutasteride over time with covariates of race/ethnicity, age and time-varying BPH status as assessed from the CMS chronic condition variables. Specifically, for a given year, if the chronic condition variable indicated the presence of BPH at that year or a prior year, then the subject was denoted as having BPH in that year. In addition, to assess the potential confounding effect of alpha blockers (tamsulosin, terazosin, doxazosin, alfuzosin and silodosin), we added alpha blocker use (as a time-varying variable) to the above GEE model to assess any changes in the odds ratios for finasteride and dutasteride use over time when additionally controlling for alpha blocker use.

Table 1 describes the demographic characteristics of the 14,833 eligible men in 2008. Just over 88% identified as non-Hispanic white (n = 13,555); while the remaining identified as either African-American or non-Hispanic black (n = 257; 1.7%), non-Hispanic Asian/Pacific Islander (n = 769; 5.2%), or of Hispanic origin (n = 247; 1.7%). Ages ranged from 62 to 82 with a median of 71 years. Almost 42% (n = 6174) reported a BPH diagnosis (Table 1). The majority of participants were married (n = 11,189; 84.5%), while slightly less than 50% were college graduates (n = 7251; 48%) and a majority were retired (n = 9193; 73.1%).

Among the 14,833 men, 6152 (41.5%) could be evaluated for BPH based on the CMS chronic conditions variables (chronic condition sub-cohort). This subset was similar demographically to the entire cohort (e.g., (age 70–79, 60.7%) race [non-Hispanic white, 85.5%]). Based on the chronic condition variables, 3474 men (56.5%) in the sub-cohort had a history of BPH at some time during 2008 to 2015. Prevalence of BPH in the sub-cohort increased over time, from 38.6% in 2008 to 58.96%) in 2015.

Overall, finasteride use at any time during 2008 to 2015 was low (n = 2023; 13.6%) (Table 2). Having a self-reported diagnosis of BPH was associated with an elevated univariate odds ratio (OR) of having used finasteride of 2.8 (95% CI: 2.5–3.1); usage rates were 21.0% and 8.4% for men with and without BPH, respectively. Univariate analysis showed that in addition to BPH status, age was associated with finasteride use. Compared to men 60–69 in 2008, men 70–79 and men 80+ were significantly more likely to have used finasteride during the period (OR = 1.55; 95% CI: (1.4–1.7) and OR = 2.0, 95% CI: (1.7–2.0), respectively) (Table 2). Results from the multivariable model indicated that age, BPH status and race were associated with finasteride use. Compared to men 60–69 in 2008, older men were more likely to have used finasteride (70–79 years, OR = 1.40; CI = 1.3–1.5); $\ge$80 years (OR = 1.63; CI = (1.4–1.9)) (Table 2). In addition, when compared to non-Hispanic white men, Hispanic men (OR = 1.46; 95% CI: (1.0–2.1)) and non-Hispanic Asian/Pacific Islander men (OR = 1.52; 95% CI: (1.3–1.8)) were more likely to use finasteride. For the chronic condition sub-cohort, use of finasteride at any time was similar as the entire analysis cohort, 12.4%. By BPH status (based on the chronic condition variables), among those with no history of BPH during 2008 to 2015, 1.5% used finasteride at any time, compared to 20.8% of men with a history of BPH during the same period.

Overall, dutasteride use was very low (n = 777; 5.2%) in the study population (Table 3). Men with BPH were significantly more likely to have used dutasteride than men without BPH (OR = 2.86; 95% CI: (2.5–3.3)). Men 70+ years and men who identified as Asian/Pacific Islanders were also more likely to use dutasteride when compared to their younger or non-Hispanic white counterparts. Similar results were obtained from the multivariable model; men 70–70 years, 80+ years, Asian/Pacific Islander and men with BPH were more likely to have taken dutasteride (OR = 1.26; 95% CI: (1.1–1.5), OR = 1.38; 95% CI: (1.1–1.8); OR = 1.73; 95% CI: (1.3–2.3); OR = 2.82; 95% CI: (2.4–3.3), respectively). For the chronic condition sub-cohort, use of dutasteride at any time was also similar to the entire analysis cohort, 5.8%. Among those with no history of BPH (based on the chronic condition variables) during 2008–2015, 0.8% used dutasteride at any time, compared to 9.6% of men with a history of BPH during the same period.

Fig. 2A–D shows the use over time of finasteride, dutasteride, either drug, and alpha blockers, respectively. Use of finasteride increased steadily over time for all men and for both men with and without BPH. Finasteride use increased overall from 3.6% in 2008 to 6.8% in 2012 and to 9.7% in 2015 (Fig. 2A). For men with BPH, finasteride use increased from 6.7% in 2008 to 11.7% in 2012 and to 14.7% in 2015, while for men without BPH, use of finasteride increased from 1.2% in 2008 to 3.3% in 2012 and to 6.1% in 2015. For the sub-cohort, the GEE model controlling for yearly BPH status (and other covariates) showed increasing finasteride use over time, with an OR of 1.10 (95% CI: 1.07–1.13) per additional year. When controlling for alpha blocker use in this model, the OR was similar (OR = 1.11).

Fig. 2.

Trends in drug use. (A) Finasteride use. (B) Dutasteride use. (C) Either Finasteride or dutasteride use. (D) Alpha blockers use. All Men, All eligible men; Men with BPH, Men diagnosed with benign prostatic hyperplasia; Men without BPH, Men without a diagnosis of benign prostatic hyperplasia.

In contrast, for all men, dutasteride use decreased over time from 2.8 % in 2008 to 1.9% in 2015 (Fig. 2B). For the sub-cohort, the GEE model controlling for yearly BPH status (and other covariates) showed a decrease in dutasteride use over time, with an OR of 0.89 (95% CI: 0.86–0.92) per additional year. When controlling for alpha blocker use in this model, the OR remained at 0.89.

Fig. 2C shows trends in the use of either drug (finasteride or dutasteride) over time. The GEE model in the sub-cohort, controlling for yearly BPH status, gave an OR for use of either drug of 1.04 (95% CI: 1.01–1.06) per additional year. The OR was unchanged when use of alpha blockers was added to the model.

Fig. 2D shows use of alpha blockers over time, Overall use increased from 15.5% in 2008 to 26.4% in 2015, while among men with BPH, use increased from 24.3% in 2008 to 33.9% in 2015.