Abstract
CONCLUSIONS
We developed a prognostic model and identified a subgroup of patients in whom third-line LPD treatment has no meaningful benefit. Our results need to be confirmed by prospective clinical trials.
RESULTS AND LIMITATIONS
Of 1011 mCRPC patients progressive on second-line LPD, 602 (60%) received third-line LPD. Patients receiving third-line LPD had a more favorable prognostic profile at baseline and longer median OS than patients with best supportive care (10.4 vs 2.4 mo, p < 0.001). Eastern Cooperative Oncology Group performance status 1 and ≥2 (hazard ratio [HR] 1.51, p < 0.007 and HR 3.08, p < 0.001, respectively), opioid use (HR 1.55, p = 0.019), visceral metastases (HR 2.09, p < 0.001), hemoglobin <7 mmol/l (HR 1.44, p < 0.002), prostate-specific antigen ≥130 μg/l (HR 1.48, p = 0.001), alkaline phosphatase ≥170 U/l (HR 1.52, p < 0.001), and lactate dehydrogenase ≥250 U/l (HR 1.44; p = 0.015) were associated with shorter survival. Harrell's C-index was 0.74. The median OS values for low-, low-intermediate-, high-intermediate-, and high-risk groups were 14, 7.7, 4.7, and 1.8 mo, respectively. Limitations include the retrospective design.
BACKGROUND
Evidence concerning third-line life-prolonging drugs (LPDs) in the treatment of metastatic castration-resistant prostate cancer (mCRPC) patients is incomplete.
DESIGN, SETTING, AND PARTICIPANTS
Patients with mCRPC who were progressive on second-line LPD before July 1, 2017 were retrospectively identified from the Dutch Castration-resistant Prostate Cancer Registry (CAPRI) and followed until December 31, 2017.
PATIENT SUMMARY
We reported outcomes from third-line life-prolonging drugs in metastatic prostate cancer patients and developed a prognostic model that could be used to guide treatment decisions.
OBJECTIVE
To evaluate third-line LPD outcomes in a real-world cohort of mCRPC patients, identify variables associated with overall survival (OS), and establish a prognostic model.
OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS
Association of potential risk factors with OS was tested by Cox proportional hazard models after multiple imputation of missing baseline characteristics. A predictive score was computed from the regression coefficient and used to classify patients into risk groups.