Can physicians accurately predict survival time in patients with metastatic cancer? Analysis of RTOG 97-14

Abstract

Purpose: To determine if physician prediction of survival duration (PSD) is accurate for patients with metastatic breast or prostate cancer.

Methods: Radiation Therapy Oncology Group 9714 (RTOG 9714) was a randomized comparison of radiotherapy schedules for treatment of bone metastases. The treating physician assigned a baseline Karnofsky Performance Score (KPS) and predicted survival duration at study entry. Patients completed the Functional Assessment of Cancer Therapy (FACT). These three were compared to actual survival time.

Results: Eight hundred ninety-eight patients were eligible and analyzable. Actual median survival was 9.3 months. The median PSD was 12 months. PSD, KPS, and FACT were all moderately correlated with actual survival. Patients with higher KPS had a longer survival time (882 patients, Spearman's rho = 0.259, p < 0.0001). The median survival of the 618 expired patients is 6.5 months (PSD was 12 months). The PSD was within 1 month of actual survival in 61 (10%), with 177 (29%) patients surviving more than 1 month longer than predicted and 375 (61%) surviving more than 1 month less than predicted. A univariate analysis of actual overall survival was performed, dividing the PSD into 4 groups. For predicted survivals of 6 months or less, less than 6 to less than 12 months, 12 months, and more than 12 months, median actual survivals were 7.0, 7.2, 9.7. and 13.5 months (p < 0.0001).

Conclusions: KPS, FACT scores, and PSD all are correlated with actual survival. Physicians on this study were able to predict which patients would have longer survival times, although prediction of survival was optimistic compared to actual survival by an average of 3 months.