(P077) Evaluating the Face of Failure: A Model to Predict Prostate Cancer Recurrences Using PSA Trends Following Brachytherapy Implant

Andrew J. Bishop, MD, Lawrence B. Levy, MS, Michelle H. Braccioforte, MPH, Brian J. Moran, MD, Juanita M. Crook, MD, Marco van Vulpen, MD, PhD, Peter Grimm, DO, David A. Swanson, MD, Usama Mahmood, MD, Thomas J. Pugh, MD, Rajat J. Kudchadker, PhD, Steven J. Frank, MD; UT MD Anderson Cancer Center; Prostate Cancer Foundation; Cancer Center for the Southern Interior; University Medical Center Utrecht; Prostate Cancer Center of Seattle

Purpose: Following a brachytherapy implant for the treatment of prostate cancer, serial prostate-specific antigen (PSA) measurements are an important indicator of treatment response and disease control. When PSA rises during follow-up, clinicians must formulate appropriate workup to evaluate for recurrences and avoid unnecessary management if possible. We sought to develop a model for predicting prostate cancer recurrence by evaluating post-treatment PSA kinetics in patients with clinical failures.

Materials and Methods: Our current analysis includes 1,816 patients treated with brachytherapy for prostate cancer from three contributing institutions. Disease recurrence was strictly defined as a recurrence confirmed by rising PSA and either a biopsy or radiographic confirmation of local or systemic disease. PSA trajectories were analyzed based on regression fit using a multilevel model among three arms: (1) none, (2) local, and (3) distant recurrences. Furthermore, three components of PSA trajectory (rate of decline, nadir, and rate of rise) were compared among the three groups. F-test, t-test, and analysis of variance (ANOVA) calculations were performed.

Results: The median follow-up for these 1,816 patients was 4.75 years, with a median pretreatment PSA of 6.6 ng/mL. Among this large cohort, there were only 37 clinically proven recurrence events after excluding biochemical failures alone. Using a quadratic fit model, the overall PSA trajectory analysis showed clear divergence of the PSA values for local failures compared with distant. However, the divergence was not statistically significant until 2 years postimplant (Year 1 P = .888; Year 2 P =.026; Year 3 P =.0012; Year 4 P =.001). Interestingly, prior to clinical failure, PSA doubling time for local failures remained relatively constant, while for distant failures, doubling time decreased with time. To further evaluate PSA trajectory, several key components were examined. The rate of PSA decline was analyzed using an ANOVA test and revealed a significant difference between the three groups at the 6-month (P = .037), 12-month (P = .003), and 18-month (P = .007) time points. Additionally, PSA nadir was evaluated for nonrecurrent patients compared with those who recurred (locally or distantly); nadir was not significantly different until the 12-month time point (P = .04). A cutpoint analysis between no failures and local failures identified that a nadir below 1.8 ng/mL at 24 months was predictive of no local recurrences (P = .003) and below 0.6 ng/mL at 42 months was predictive (P < .001). Finally, the rate of PSA increase was also evaluated at 6-month time intervals; once again, it did not reach significance among the three groups until the 12-month interval postimplant (P < .001) and remained significant throughout all subsequent time intervals (18 mo P = .01; 24 mo P < .001; 30 mo P < .001; 36 mo P < .001; 42 mo P < .001; 48 mo P < .001).

Conclusions: This preliminary analysis suggests that following permanent brachytherapy implant for prostate cancer, close observation of PSA trends may predict patients who will ultimately recur and that differentiating between local and distant recurrences may be possible. Currently, additional patients are being added to the model for a more robust analysis of PSA trends. Ultimately, this large, multi-institutional analysis may produce a model that can predict postbrachytherapy failures by incorporating PSA kinetics and evaluating the PSA trajectory.