Overall survival (OS) with docetaxel (D) vs novel hormonal therapy (NHT) with abiraterone (A) or enzalutamide (E) after a prior NHT in patients (Pts) with metastatic prostate cancer (mPC): Results from a real-world dataset.
5537 Background: NHT (A and E) are approved first-line (1L) treatment (Rx) for mPC. After progression on NHT, Rx include either alternate NHT or D. However, OS from a randomized trial comparing NHT vs D after progression on 1L NHT has not been reported. Methods: Pts data were extracted from the Flatiron Health EHR-derived de-identified database. Inclusion: diagnosis of mPC; 1L Rx with single agent A or E only, single-agent Rx with alternate NHT (E or A) or D in second line (2L). Exclusion: > 180 days between date of diagnosis of mPC and date of next visit to ensure Pts were actively engaged in care at data-providing site; Rx with NHT in non-metastatic setting, any prior exposure to D. OS was compared using Cox proportional hazards model stratified by Rx propensity score. Each Pts’ probability of receiving D (rather than NHT) was modeled via a random forest based on Pts and disease characteristics which may drive treatment selection. These included pre-2L Rx ECOG scores, PSA, LDH, ALPH, Hb, age, ICD codes for liver metastasis, diabetes, neuropathy, and heart failure; insurance payer, year of start of 2L Rx, time on 1 L NHT, Gleason score, PSA at the original diagnosis of mPC. Subgroup analyses included 1L Rx duration < 12 mos. Results: 1165 Pts between 2/5/2013 to 9/27/2019 were eligible. Median follow up 8 mos (range 0.1-64.5). Median OS after 1L A was higher with E as compared to D (15.7 vs. 9.4 mos). Median OS after 1L E was higher with A as compared to D (13.3 vs. 9.7 mos) (table). Propensity distributions were overlapping among Rx arms and showed only modest imbalance. In 2L, D had a worse adjusted hazard ratio of 1.29 and 1.35 as compared to E and A respectively (p < 0.05). Similar results were seen with 1L Rx duration of < 12 mos (p < 0.05). Conclusions: These hypothesis-generating data provide real-world OS estimates with 2L D & NHT in mPC. In propensity-stratified analyses, mPC Pts who progressed on NHT had a worse OS with 2L D as compared to alternate NHT. Results were consistent in unadjusted analysis & subgroup analyses of 1L Rx < 12 mos. Results are subject to residual confounding and missingness. After prospective validation these data may aid in Rx sequencing, Pts counselling, and design of future clinical trials in this setting. [Table: see text]