Purpose: To coherently integrate multiple sources of available evidence to project lifetime outcomes for newly diagnosed men considering immediate treatment or active surveillance (AS).
Method: Lifetime estimates of time from treatment to progression (T-P) and time from progression to mortality (P-M) were estimated for the 11,347 men diagnosed in 2004-2006 in the SEER cancer registry with low-risk disease (≤ grade 6 and ≤ stage T2a) who were treated immediately with surgery. Over 38% of all patients diagnosed during these years had low-risk disease. Outcomes under an alternative scenario of active surveillance were estimated for this cohort. Estimates for this scenario integrated a model for diagnosis to delayed treatment (D-T) including parameters for grade and PSA progression, with the same T-P and P-M models. Estimates of the potential harm of surveillance were based on advanced disease characteristics at time of delayed treatment. Large cohorts from CaPSURE, Johns Hopkins, and Mayo Clinic were used to inform the models.
Result: With immediate surgery, 26% of low-risk patients will experience biochemical failure and 1.8% will die from prostate cancer. The sampled surgery included only men who were low grade and only 37% had a PSA ≥6. The surveillance scenario resulted in 58% of patients going on to be treated, with 34% upgraded at time of treatment and 59% having a PSA ≥6. Although disease characteristics were more advanced at the time of delayed treatment, there were no additional deaths due to prostate cancer with surveillance and only a total of 18% of men experienced biochemical failure. Mean life expectancy for low-risk men treated with surgery between 2004-2006 is projected to be 19.5 years. Under the surveillance scenario, had this cohort selected surveillance they would have experienced only 11.3 treated-person years.
Conclusion: Although active surveillance is associated with more advanced disease characteristics for some low-risk men who go on to be treated, projections of mortality based on these upgraded disease states did not result in any additional deaths. Surveillance offers a substantial reduction in the number of treated-person years. Models can help integrate multiple sources of data to help overcome the extremely long time required to observe outcomes in prospective studies between diagnosis and prostate cancer mortality.