Local variation in primary treatment of localized prostate cancer

Abstract

5126 Background: We aimed to characterize and quantify variation in the primary management of localized prostate cancer at the level of clinical practice sites. Methods: Data were abstracted from patients accrued to the CaPSURE national prostate cancer registry. Patients were accrued from the 36 clinical practice sites which contributed at least 30 patients to the registry, and represented all those diagnosed since 1990 with localized disease who received radical prostatectomy (RP), external beam radiation therapy (EBRT), brachytherapy, active surveillance / watchful waiting (WW), or primary androgen deprivation therapy (PADT) were included. Descriptive analyses were performed, and a random effects logit hierarchical model was constructed, controlling for year of diagnosis, age, comorbidity, PSA, Gleason score, clinical T stage, and percent of biopsy cores positive, to estimate the proportion of variation in primary treatment selection explicable by practice site. Analyses were conducted for all patients and for low-risk patients (Gleason score ≤6, PSA ≤10 ng/ml, clinical stage ≤T2a). Results: 10,080 men were analyzed. The distribution among primary treatments at each clinical practice site varied widely: use of RP, for example, ranged from 12% to 95% of enrolled patients. Patterns of treatment are not reliably explained by patient risk distribution at each site. The proportion of variation attributable to clinical practice sites was 10% for PADT, 19% for WW, 21% for EBRT, 28% for RP, 37% for brachytherapy, and 75% for cryotherapy. For low-risk patients only, this proportion was higher for all treatment types except brachytherapy and cryotherapy. Only a small amount of the variation attributable to practice site can be explained by measured sociodemographic factors such as ethnicity, income, education, and geographic region. There are significant trends in treatments over time, including more use of PADT for intermediate- and high-risk patients, and more use of RP and WW for low-risk patients. Conclusions: These data do not represent a random sampling of the United States population. However, the significant variation in practice patterns across individual clinical sites suggests that factors other than patient clinical and sociodemographic factors may be driving selection of primary treatment. [Table: see text]