Effect of Preoperative Treatment on the Performance of Predictive Nomograms in Primary Retroperitoneal Sarcoma.

Abstract

Retroperitoneal sarcoma (RPS)-specific nomograms provide estimates of survival and recurrence risk following resection in the individual patient. The effect of preoperative treatment on nomogram performance has not been previously examined. Our aim was to evaluate the predictive accuracy of existing RPS-specific nomograms in patients managed at our center, where the majority of patients received preoperative radiation. All patients who underwent curative treatment for primary RPS at Mount Sinai Hospital/Princess Margaret Hospital between 1996 and 2016 were identified. The performance of four previously published nomograms was assessed by measuring the agreement between nomogram-predicted and observed outcomes using Harrell's C-Index and level of calibration. Outcomes included in each of the nomograms [overall survival (OS), disease-free survival (DFS), disease-specific death (DSD), local recurrence (LR), distant recurrence (DR)] at each of the specified post-resection timepoints were examined. In total, 253 patients were included. When observed outcomes were compared with those predicted by each of the four nomograms, the C-Index ranged from 0.60 to 0.81, representing a wide range of predictive accuracy. The lowest C-Index was for prediction of LR. Calibration plots revealed that the Memorial Sloan Kettering Cancer Center (MSKCC) nomogram predicted a 5-year LR of 45%, whereas the observed LR was 24%. Overprediction of LR was detected in patients who had undergone preoperative radiotherapy, but not in patients treated with surgery alone. Preoperative radiotherapy appeared to preclude the use of the LR component of existing nomograms for primary RPS. Updated nomograms should be created to reflect this variable, particularly in light of the recently published STRASS trial results.