Risk Stratification in Hormone-sensitive Metastatic Prostate Cancer: More Questions than Answers

Abstract

The approach to metastatic prostate cancer in 2014 has changed little from that detailed by Huggins and Hodges in 1941. Although themainstay of therapy for newly diagnosed noncastrate metastatic prostate cancer (NCMPC) remains androgen deprivation therapy (ADT), the landscape of options available in 2014 has evolved considerably beyond surgical castration [1]. Despite this proliferation of therapeutic options, our understanding of prognostic factors relating to overall survival and optimal response to novel pharmacologic strategies remains relatively rudimentary. Current options for men with NCMPC extend beyond classic medical castration, and include combined androgen blockade, peripheral androgen blockade, and, as recently reported, up-front cytotoxic chemotherapy [2]. Furthermore, the merits of intermittent versus continuous ADT continue to be debated [3], and the role of locally directed therapy has entered the arena as a potential strategy to postpone palliative ADT in those with oligometastatic prostate cancer [4–6]. The net effect of these advances has been a reduction in the mortality rate for patients with NCMPC [7]. With increasing therapeutic options and treatment strategies, however, the ability to risk-stratify a patient with NCMPC becomes ever more important. In this issue of European Urology, Gravis et al [8] sought to contribute to the sparse literature addressing prognostic factors for overall survival in NCMPC, building on themodel developed by Glass and colleagues in 2003 [9] in which good, intermediate, and poor risk groups were identified from 1076 patients accrued in the SWOG 8894 trial on the basis of appendicular versus axial skeletal bone metastases, performance status 1, prostate-specific antigen (PSA) 65 ng/ml, and Gleason score 8. Gravis et al tested the Glass model in a post hoc analysis of the GETUG-15 cohort, in which 385 men with NCMPC were randomized 1:1 to ADT with or without docetaxel. The model performed disappointingly in this contemporary cohort (c-index 0.59), which was attributed to improvements in NCMPC management, as well as lead-time bias due to widespread PSA screening. Patients enrolled in GETUG-15 additionally differed from those in SWOG 8894 in that they were younger, they had better performance status, and a smaller proportion had poor-risk disease. Furthermore, in the GETUG-15 trial, variable subsequent therapies were used in the majority of both arms following progression, which potentially confounded the study outcomes. Following the methodology of Glass et al, Gravis and colleagues developed and validated new prognostic risk groups. Alkaline phosphatase (ALP), Gleason score, and pain intensity provided the greatest degree of discrimination, with ALP alone outperforming the Glass model (c-index 0.64), although the confidence intervals for both indices overlapped. The authors assert that ALP may be used alone as an inexpensive and readily available prognostic factor in patients with NCMPC. However, in clinical practice, the utility of a single measure, even if easily obtained and costeffective, that predicts all-cause mortality only slightly better than the flip of a coin is arguable. Furthermore, in the GETUG-15 study, abnormal ALP was not independently associated with response to up-front docetaxel with ADT (hazard ratio 0.98, 95% confidence interval 0.66–1.45, p = 0.94) [10], further drawing into question its clinical utility if it is ineffective in predicting response to treatment. Candidate predictors that have emerged as potentially useful include both the location and volume of the disease, E U RO P E AN URO LOG Y 6 8 ( 2 0 1 5 ) 2 0 5 – 2 0 6