Baseline 25-hydroxyvitamin D and progression-free survival in de novo metastatic hormone-sensitive prostate cancer: prognostic model development and external validation

Authors:
Open Access
Publisher website
Google Scholar
Abstract
Background: Prognostic biomarkers for risk stratification remain limited in de novo metastatic hormone-sensitive prostate cancer (mHSPC) treated with androgen receptor pathway inhibitors (ARPIs). We evaluated baseline serum 25-hydroxyvitamin D [25(OH)D] and developed and externally validated a progression-free survival (PFS) model. Methods: This multicenter retrospective study included 273 patients initiating ARPI-based therapy between January 2021 and April 2026. Etlik City Hospital formed the training cohort (n = 192; 85 PFS events), and Gazi University and Kütahya City Hospital formed the external validation cohort (n = 81; 27 PFS events). LASSO-penalized Cox regression retained age, ISUP grade group, CHAARTED disease volume, log-transformed PSA, and hemoglobin. Model B added continuous 25(OH)D. Internal validation used 1000 bootstrap resamples. Performance was assessed by the Harrell C-index, time-dependent AUC, and calibration at 1 and 2 years. PFS was the primary endpoint; overall survival analyses were exploratory. The study was not prospectively registered. Results: Higher baseline 25(OH)D was associated with longer PFS in Model B (HR per 10-ng/mL increase, 0.64; 95% CI, 0.49-0.83; P < .001). Adding 25(OH)D improved model fit (likelihood-ratio χ² = 12.71; P < .001) and increased apparent C-index from 0.712 to 0.740. In external validation, C-indices were 0.71 (95% CI, 0.64-0.78) and 0.76 (95% CI, 0.69-0.83), respectively (Δ = 0.05; 95% CI, 0.01-0.10). Conclusions: Baseline 25(OH)D added modest prognostic information beyond clinical factors, with supportive external validation. These findings do not establish treatment-predictive utility or benefit from vitamin D supplementation. Prospective validation is needed before clinical implementation.

This publication has 28 references indexed in Scilit: