Purpose: To identify external beam radiation therapy (EBRT)-related and diagnostic factors that predict biochemical failure-free survival (BFFS) for prostate cancer patients.
Methods: Patients diagnosed with localized prostate adenocarcinoma (N0, M0) between January 2005 and December 2016 and treated with curative EBRT using a Varian linear accelerator at one of the four cancer centers in Alberta, Canada, were retrospectively included. Biochemical failure-free survival (Astro-Phoenix definition) was calculated using Kaplan-Meier estimator. Patient demographic and diagnostic factors, as well as EBRT treatment planning and delivery parameters, were collected. Cox proportional hazard model with LASSO regularization and random survival forest were used to identify the predictive factors.
Results: This study included a cohort of 2827 patients with a median follow-up of 6.4 years (IQR: 4.0 - 9.1 years). The BFFS for this patient cohort was 85% at 5 years (CI: 83.5%-86.4%). The statistically significant factors in the Cox model with LASSO regularization were: Gleason score, prostate-specific antigen (PSA) at diagnosis, clinical T stage (increased risk of biochemical failure with worse diagnosis), PTV V98% (lower risk with better coverage), and irradiation technique (2-phase pelvic irradiation had a lower risk than seminal vesicles having lower dose than prostate). The five most predictive variables in random survival forest were, in order of decreasing importance: Gleason score, PSA at diagnosis, PTV V98%, clinical T stage, and CTV D99%. Variables including dose calculation algorithm, bladder volume, hip implant, PTV D99%, hormone treatment, PTV volume, age at diagnosis, image-guidance type, RT modality, number of malignancies, equivalent PTV margin, and fiducial marker usage were not predictive in either method.
Conclusion: Using a large retrospective cohort of patients from multiple centers, we identified that dosimetric parameters including PTV V98% and CTV D99% were predictive of BFFS in addition to diagnostic factors including Gleason score, PSA at diagnosis, and clinical T stage.
Funding Support, Disclosures, and Conflict of Interest: This study is supported by Alberta Innovates Graduate Studentship in Health Solution.