Accuracy and prognostic significance of oncologists’ estimates and scenarios for survival time in a randomised Phase 2 trial of regorafenib in advanced gastric cancer

Abstract

Background: We have proposed that best, worst and typical scenarios for survival, based on simple multiples of an individual's expected survival time (EST) estimated by their oncologist, are a useful way of formulating and explaining prognosis in advanced cancer. We aimed to determine the accuracy and prognostic significance of such estimates in a multicentre, randomized trial. Method(s): 66 oncologists estimated the EST at baseline for each of 152 participants in the INTEGRATE trial. We expected oncologists' estimates of EST to be well calibrated (~50% of patients living longer or shorter than their EST) and imprecise (<33% living within 0.67 to 1.33 times their EST), but to provide accurate scenarios for survival time (~10% dying within a quarter of their EST,~10% living longer than 3 times their EST and~50% living for half to double their EST). We also hypothesized that oncologists' estimates of EST would be independently predictive of overall survival in a multivariable Cox model including conventional prognostic factors and health related quality of life. Result(s): Oncologists' estimates of EST were well calibrated (45% shorter than observed), imprecise (29% lived within 0.67 to 1.33 times observed), and moderately discriminative (Harrell C-statistic 0.62, P=0.001). Scenarios derived from oncologists' estimates were remarkably accurate: 9%of patients died within a quarter of their EST, 12% lived longer than three times their EST and 57% lived within half to double their EST. Oncologists estimates of EST (in months) were independently significant predictors of overall survival (HR=0.89, 95% CI 0.83 to 0.95, P=0.001) in a Cox model including ECOG performance status, primary site, number of metastatic sites, neutrophil to lymphocyte ratio>= 3, treatment group, age, and the EORTC QLQC30 physical function score. Conclusion(s): Oncologists' estimates of survival time were well calibrated, moderately discriminative, and independently significant predictors of overall survival. Best, worst and typical scenarios for survival based on simple multiples of the EST were remarkably accurate and would provide a useful method for estimating and explaining prognosis in this setting.