To the Editor: The study of survival associated with treatment vs observation of localized prostate cancer by Dr Wong and colleagues suggested that active treatment of elderly men with lowor moderate-risk prostate cancer was associated with a survival benefit. However, 5-year relative survival for localized or regional prostate cancer is 100%. Therefore, any true survival benefit associated with a treatment would likely begin to first manifest only after 5 years of follow-up. This effect was seen in the recent randomized controlled trial comparing radical prostatectomy with watchful waiting conducted by the Scandinavian Prostate Cancer Group. This delay in survival advantage is why it is generally recognized that patients must have a life expectancy of approximately 10 years or more to significantly benefit from any active treatments for prostate cancer. Figure 1 in Wong et al, however, demonstrated that most of the observed survival differences occurred in the first 5 years of follow-up; survival curves were relatively parallel beyond 6 years of follow-up. Such a pattern suggests that the observed difference in overall survival might be largely due to incompletely adjusted differences in underlying patient characteristics or other biases despite the extensive adjustments and modeling performed by the authors. In addition, the exclusion of perioperative deaths is not consistent with intention-to-treat principles. Although the absolute rate of these events was expected to be low (0.66% for men aged 70-79 years), the relative effect on the study’s findings could be substantial given that during the study period only 2.1% of the study participants died of prostate cancer. One approach to address these issues would be to separately compare nonprostate cancer mortality by treatment after adjustment for the propensity score. Radical prostatectomy should only affect prostate cancer-specific mortality. If men receiving radical prostectomy had significantly lower nonprostate cancer mortality rates after adjustment, residual confounding would be the most plausible explanation. We believe that until such analyses are performed and perioperative deaths are included, the results of this study should be interpreted with caution.