Objective: To test the accuracy of predicting life-expectancy (LE) among 19 raters, as the accurate prediction of LE in candidates for definitive therapy for localized prostate cancer is crucial, and little is known of the ability of clinicians to predict LE.
Subjects and methods: We randomly selected the case-vignettes of 50 patients treated with either radical prostatectomy (RP, 25) or external beam radiotherapy (EBRT, 25) for prostate cancer, and who either survived for > 10 years or died earlier with no evidence of disease relapse. The median age at treatment was 67 years and the median Charlson Comorbidity Index (CCI) was 2. The raters consisted of urology staff (six), urology residents (10) and medical students (three). The case-vignettes included patient age, comorbidities and CCI score, and raters were asked to predict the survival at 10 years (yes vs no), assuming no disease relapse.
Results: Of the 50 cases, 20 (40%) did not survive for > 10 years; clinicians estimated a mean (range) of 23 (10-35) deaths before 10 years. The mean (95% confidence interval) overall predictive accuracy (0.5 = chance, 1.0 = perfect prediction) of LE predictions was 0.68 (0.64-0.71). Individual accuracy ranged from 0.52 (staff) to 0.78 (staff). There were no important differences among the rater groups (residents 0.69 vs staff 0.67 vs medical students 0.67).
Conclusions: Clinicians are relatively poor at predicting LE; tools to predict LE might be able to improve clinicians' performance in this important part of decision-making about prostate cancer treatment. It remains to be determined whether this limitation exclusively applies to prostate cancer or also to other malignancies.