Clinical risk prediction tools for prostate cancer: Tnm to capra–should risk be redefined?

Abstract

Variation in prostate cancer biology and outcomes warrants highly calibrated assessment instruments that allow for reliable risk estimation at relevant decision points. In this chapter, we provide a critical overview of the principles by which risk stratification tools are evaluated as well as a longitudinal contextualization of clinical prognostication. Risk groupings including the D’Amico, AUA, and NCCN classification systems derived from categorical clinical characteristics including PSA, Gleason score, and clinical stage offer a parsimonious framework that can be recalled with relative ease, yet demonstrate suboptimal discrimination due to very wide heterogeneity within groups. Nomograms derived from multivariate models offer improvements in spectrum bias and discrimination by incorporating continuous variables accounting for a wide range of values, rather than dichotomized groupings, and however face limitations in the requirement of multi-step drawing or online calculator. The UCSF-CAPRA score, a multivariate risk prediction estimate expressed on a 10-point scale incorporating pretreatment clinical variables, has been widely validated in academic, community, and international cohorts. Such methodology has also been applied in the postsurgical setting (CAPRA-S) and for patients with advanced disease treated with primary androgen deprivation therapy (J-CAPRA). Further refinement of prostate cancer risk prediction will likely require incorporation of advanced imaging, novel biomarkers, and tissue-based gene expression assays.