Background: Active surveillance (AS) for prostate cancer includes follow-up with serial prostate biopsies. The optimal biopsy frequency during follow-up has not been determined. The goal of this investigation was to use longitudinal AS biopsy data to assess whether the frequency of biopsy could be reduced without substantially prolonging the time to the detection of disease with a Gleason score ≥ 7.
Methods: With data from 1375 men with low-risk prostate cancer enrolled in AS at Johns Hopkins, a hidden Markov model was developed to estimate the probability of undersampling at diagnosis, the annual probability of grade progression, and the 10-year cumulative probability of reclassification or progression to Gleason score ≥ 7. It simulated 1024 potential AS biopsy strategies for the 10 years after diagnosis. For each of these strategies, the model predicted the mean delay in the detection of disease with a Gleason score ≥ 7.
Results: The model estimated the 10-year cumulative probability of reclassification from a Gleason score of 6 to a Gleason score ≥ 7 to be 40.0%. The probability of undersampling at diagnosis was 9.8%, and the annual progression probability for men with a Gleason score of 6 was 4.0%. On the basis of these estimates, a simulation of an annual biopsy strategy estimated the mean time to the detection of disease with a Gleason score ≥ 7 to be 14.1 months; however, several strategies eliminated biopsies with only small delays (<12 months) in detecting grade progression.
Conclusions: Although annual biopsy for low-risk men on AS is associated with the shortest time to the detection of disease with a Gleason score ≥ 7, several alternative strategies may allow less frequent biopsying without sizable delays in detecting grade progression. Cancer 2018;124:698-705. © 2017 American Cancer Society.
Keywords: Markov model; active surveillance; biopsy; prostate cancer; reclassification.
© 2017 American Cancer Society.