1040 - Development of Nomograms to Predict Polymetastatic Progression Free Survival and Overall Survival in Patients Treated with Stereotactic Ablative Radiotherapy for Oligometastatic or Oligoprogressive Ca

S. Das¹, W. Liu², L. Lechner³, B. Mou⁴, W. Jiang¹, M. Liu², D. Schellenberg¹, T. Berrang⁵, A. S. Alexander⁵, C. K. Ho¹, B. Valev⁵, H. Carolan², S. Atrchian⁴, A. Bergman², N. Chng⁶, Q. Matthews³, G. Arbour³, S. Tyldesley², R. A. Olson⁷, and S. Baker¹; ¹BC Cancer - Surrey, Surrey, BC, Canada, ²BC Cancer - Vancouver, Vancouver, BC, Canada, ³University of British Columbia, Vancouver, BC, Canada, ⁴BC Cancer - Kelowna, Kelowna, BC, Canada, ⁵BC Cancer - Victoria, Victoria, BC, Canada, ⁶BC Cancer - Prince George, Prince George, BC, Canada, ⁷BCCA - Prince George, Prince George, BC, Canada

Purpose/Objective(s): While estimation of prognosis for patients with oligometastatic cancer is important to aid in treatment decision making, at present, prognostication tools are lacking. The purpose of this study was to develop predictive models for polymetastatic progression free survival (PPFS) and overall survival (OS), based on patient outcomes on the SABR-5 clinical trial. Materials/

Methods: In the SABR-5 trial (a multi-center, single arm, phase II clinical trial), 381 patients with 1-5 oligometastases or oligoprogressing lesions were treated with stereotactic ablative radiotherapy (SABR). Prostate cancer was the most common histology (32%), followed by colorectal (16%), breast (11%), lung (9%), and renal (9%). Most patients (91%) were treated for 1-2 metastases. PPFS was defined as time from SABR until polymetastatic failure (= 6 metastases, or malignant pleural effusion/malignant ascites) or death. We trained separate Cox models for PPFS and OS. An elastic net penalty was used to select from candidate covariates. The model was internally validated using 10-fold cross validation with three repetitions and evaluated using partial likelihood. The proportional hazards and linearity assumptions were tested by examination of scaled Schoenfold Residuals and Martingale Residuals, respectively.
Results: After a median follow-up time of 28.0 months (interquartile range [IQR]19.1 – 39.1), median PPFS was 34.3 months (95% confidence interval [CI] 28.8 – 39.9), and median OS was 50.5 months (95% CI 45.2 – 58.9). The resulting PPFS model contained four covariates: primary tumor type (prostate, colorectal, lung, breast, renal, other), ECOG (1-2 vs 0), oligoprogression (yes vs no), number of organs with metastases (single vs multiple). The cross-validated C-Index was 0.66 [95% CI 0.64 – 0.68]. The OS model had five covariates: age (continuous), primary tumor type (prostate, colorectal, lung, breast, renal, other), ECOG (1-2 vs 0), number of organs with metastases (single vs multiple), presence of brain metastasis (yes vs no) and had a corresponding C-Index of 0.67 [95% CI 0.64 – 0.700].
Conclusion: These nomograms can be a useful guiding tool for clinicians in predicting PPFS and OS in patients treated with SABR in oligometastatic or oligoprogressive setting. These models should be externally validated.