Measuring the Comparative Effectiveness of Prostate Cancer Treatment Options

Treatment of localized prostate cancer is well established and consists primarily of expectant management (active surveillance) and definitive treatment (radical prostatectomy or radiation). Yet, the evidence to date of their comparative effectiveness relies on randomized controlled trials (RCTs). Small sample sizes and self-selection of trial participants limit the generalizability of RCTs to common care settings and the general population.

The paper presented in this ZEW Research Seminar uses records from the Surveillance, Epidemiology, and End Results (SEER)-Medicare linked database of prostate cancer patients aged 65 to 84 years who were diagnosed between 2004 and 2017 to estimate the causal effect of definitive treatment versus expectant management on prostate cancer outcomes. The authors‘ regression-discontinuity design exploits the fact that prostate cancer risk classification is discrete, while prostate cancer progression is arguably a continuous, biological process that does not change discontinuously at risk-classification thresholds. They estimate that patients with PSA = 20 were 5-12 percentage points more likely to choose definitive treatment within six months of diagnosis than observably identical patients whose PSA value was 19.9. As 70.7% of patients with PSA = 18 or 19 chose definitive treatment, this increase represents a 7-17% jump that is attributable to the cancer's higher risk classification. The authors plan to present estimated causal effect of choosing definitive treatment on measures of health care quality, use, and cost, treatment complications, disease recurrence, and survival.