3137 - Association of Tumor Volume and Radiographic T-Stage for Patients with Localized Prostate Cancer Treated with Radiation or Radical Prostatectomy

C. Belant¹, L. K. Lee², G. Pratt³, M. Rowan³, J. Parisi¹, J. E. Leeman⁴, K. N. Lee⁵, D. D. Yang⁵, P. L. Nguyen⁶, P. F. Orio III⁷, A. V. DAmico⁶, and M. T. King⁶; ¹Dana-Farber Cancer Institute/Brigham and Womens Hospital, Boston, MA, ²Department of Radiology, Brigham and Womens Hospital, Boston, MA, ³Brigham and Womens Hospital/Dana-Farber Cancer Institute, Boston, MA, ⁴Department of Radiation Oncology, Dana-Farber Cancer Institute/Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, ⁵Harvard Radiation Oncology Program, Boston, MA, ⁶Department of Radiation Oncology, Brigham and Women’s Hospital/Dana-Farber Cancer Institute, Boston, MA, ⁷Department of Radiation Oncology, Dana-Farber Cancer Institute, Boston, MA

Purpose/Objective(s): Tumor volume (V_AI) delineated from a deep learning segmentation algorithm has shown to be prognostic for metastasis for patients with localized prostate cancer treated with radiation therapy (RT) and radical prostatectomy (RP). The purpose of this study is to evaluate the interaction between V_AI and radiographic T-stage for biochemical failure (BF) and metastasis in patients treated for localized prostate cancer. Materials/

Methods: Using retrospective data from a single institution, 764 patients with PIRADS scores were split into RT (438) and RP (326) cohorts. V_AI was obtained from deep-learning autosegmentation (nnUNet) which was trained on 288 patients treated with RT through 5-fold cross-validation and applied to the remaining patients. Radiographic T-stage (rT1-2: organ-confined PIRADS 0-5, rT3a: radiographic EPE, rT3b: radiographic seminal vesicle invasion) was determined from radiology reports. For each cohort, we analyzed the association of rT-stage with V_AI using linear regression analysis. We then evaluated if rT-stage, V_AI, and their interactions were predictive of BF and metastasis using multivariable Cox regression.
Results: For RT, rT3a (2.41 mL (1.68 to 3.13, p<0.001)) and rT3b (7.73 mL (6.67 to 8.79, p<0.001)) tumors were larger than rT1-2 tumors. There were 61 BFs and 28 metastases at a median follow up of 7.9 years). V_AI (adjusted hazard ratio (AHR): 1.37 (1.16-1.63, p<0.001)), radiographic T-stage (T3a: 3.20 (1.53-6.71, p=0.002), rT3b: 3.74 (1.35-10.35, p=0.011)), and their respective interactions (V_AI/rT3a: 0.80 (0.66-0.96, p=0.018), V_AI/rT3b: 0.79 (0.66-0.94, p=0.009)) were associated with BF. Respective AHRs of V_AI for the rT1-T2, rT3a, and rT3b subgroups were 1.39 [1.16, 1.67] (p = <0.001), 1.07 [0.98, 1.18] (p = 0.12), and 1.09 [1.01, 1.16] (p = 0.02). Significant interactions were also observed for metastasis (V_AI/rT3a: 0.80 (0.61-1.05, p=0.12), V_AI/rT3b: 0.72 (0.55-0.94, p=0.016)). For RP, rT3a (3.27 mL (2.09 to 4.45, p<0.001)) and rT3b (8.55 mL (6.59 to 10.52, p<0.001)) tumors were larger than rT1-2 tumors. There were 83 BFs and 23 metastases at a median follow up of 5.5 years). V_AI (AHR: 1.16 (1.06-1.28, p=0.002)), radiographic T-stage (T3a: 1.47 (0.75-2.89, p=0.263), rT3b: 4.91 (1.89-12.72, p=0.001)), and their respective interactions (V_AI/rT3a: 0.98 (0.87-1.10, p=0.728), V_AI/rT3b: 0.89 (0.79-0.99, p=0.036) were associated with BF. Respective AHRs of V_AI for the rT1-T2, rT3a, and rT3b subgroups were 1.16 [1.06, 1.28] (p = 0.003), 1.12 [1.05, 1.20] (p = 0.001), and 1.03 [0.97, 1.09] (p = 0.36). A near-significant interaction was observed for metastasis (V_AI/rT3a: 0.91 (0.73-1.14, p=0.424), V_AI/rT3b: 0.80 (0.64-1.00, p=0.055)).
Conclusion: Tumor volume was most prognostic for patients with radiographically organ-confined prostate cancer treated with RT or RP.