A novel prognostic model to optimize the timing of docetaxel (Doc) following an androgen receptor pathway inhibitor (ARPi) in metastasic castration-resistant prostate cancer (mCRPC).

5063 Background: ARPi has become a standard of care for advanced prostate cancer due to its activity and safety profile. DoC is a therapeutic alternative as first-line in unselected mCRPC patients (pts) after a first ARPi. Best timing to initiate Doc following an ARPi remains uncertain, with most predictive models based on baseline characteristics and without variables related to the disease evolution. We have developed and validated a tool to predict progression-free survival (PFS) and overall survival (OS) with Doc to capture individual tumor behavior and guide the optimal timing of Doc. Methods: mCRPC pts from PROREPAIR-B (NCT03075735) treated with an ARPi followed by Doc were selected for a derivation cohort (n=111). Those pts from PROSTAC (NCT02362620), PROSABI (NCT02787837) and/or PROSENZA (NCT02922218) not included in PROREPAIR-B treated with an ARPi followed by Doc were included in the validation cohort (n=243). Baseline characteristics at ARPi initiation, type and time of response/progression to ARPi and prognostic variables at initiation of Doc were collected. All variables underwent univariate Cox regression for PFS and OS from Doc in the derivation set. Only significant variables underwent the least absolute shrinkage and selection operator (LASSO) method to select the most relevant. Individual risk scores for PFS and OS were generated using Cox to classify patients into low- and high-risk groups and to construct a nomogram. Finally, an external validation was conducted in our second independent cohort. Predictive accuracy was assessed using Harrell's C-index. Results: The Lasso method identified 6 variables associated to PFS and OS: age at diagnosis, time to (t)mCRPC, development of new nodal and/or new visceral metastasis during ARPi, elevated LDH and ECOG at Doc initiation (Table). After bootstrap correction for internal validation the PFS and OS model showed a C-index of 0.71 (CI95% 0.66–0.77) and 0.74 (0.68–0.80), respectively. The model was validated in our second cohort with c-index of 0.62 (0.6-0.7) and 0.64 (0.6-0.7) for PFS and OS, respectively. Conclusions: Our validated model provides a pragmatic and widely accessible tool for physicians to estimate the potential benefit of Doc after an ARPi according to pts age and time to mCRPC, particularly in settings where targeted therapies at the time of progression to ARPi are not available. Variable PFS risk coefficient (rc) PFS (HR CI95%) OS rc OS (HR CI95%) Age (<60, 60-70, >70) .35 1.5 (1.1-1.9) .33 1.6 (1.2-2.3) tmCRPC (>98, 98-23, <23) -.17 0.7 (0.5-0.9) -.49 0.4 (0.2-0.7) Visceral MTS (never, persistent, new) .48 1.2 (0.9-1.6) .54 1.9 (0.9-3.9) Nodal MTS (never, persistent, new) .58 1.4 (1.0-1.9) .66 2.0 (1.0-4.2) LDH (xULN) .10 1.2 (1-1.2) .8 1.9 (1.0-2.4) ECOG at Doc .63 2.1 (1.4-2.9) .74 4.1 (2.1-8.2) Total Score (Low vs high) 1.9 (1.2-3.1) 2.13 (1.3- 3.5)