Estimating patient health in prostate cancer treatment counseling

We assessed the concordance among urologists' judgment of health quartiles for patients with localized prostate cancer, and compared the life expectancy (LE) and ensuing treatment recommendations when following National Comprehensive Cancer Network (NCCN) guidelines based on actuarial life tables versus the Kent model, a validated LE prediction model.

NCCN suggests using actuarial life tables and relying on surgeon assessment of patient health to increase (for the best quartile) or decrease (for the worst quartile) LE by 50%. Eleven urologic surgeons allocated quartile of health and recommended treatments for ten patient vignettes. The 10-year survival probability was calculated using the Kent model and compared to the life-table estimate based on health quartile by surgeon consensus.

Surgeon assessment agreed with the presumed true quartile of health based on a validated model in 41% of cases. For no case did three-quarters of surgeons assign health quartile correctly; in half of cases, <50% of surgeons assigned the correct quartile. The NCCN comorbidity-adjusted LE estimates underestimated risk of death in the best health quartile and overestimated risk of death in the worst health quartile, compared to the Kent model. Patients with LE > 10 years on NCCN estimation were recommended more frequently for surgery (81%) and those with ≤10 years estimated LE were more commonly recommended for radiation (57%) or observation (29%).

A method based on physician-assessed health quartiles for LE estimation, as suggested by the NCCN guidelines, appears too crude to be used in the treatment counseling of men with localized prostate cancer, as compared to a validated prediction model, such as the Kent model.