Validation of Risk Prediction System for Denosumab-Induced Hypocalcemia with an External Clinical Dataset

Abstract

Denosumab is used to reduce skeletal-related events such as fractures in cancer patients with bone metastasis, but may cause severe hypocalcemia. We previously developed and updated a risk prediction model for ≥ grade 2 denosumab-induced hypocalcemia from a hospital-based administrative database and clinical datasets from two facilities. The final risk-scoring system using only calcium, albumin, and alkaline phosphatase levels provided high performance. Here, we aimed to externally validate the scoring system’s performance using an independent clinical dataset from Gunma Prefectural Cancer Center. Clinical data (May 2017–November 2023) were retrospectively collected and the discriminant performance of the previously developed model (sensitivity, specificity, positive predictive value, negative predictive value and receiver operating characteristic-area under the curve (ROC-AUC)) was evaluated. For 161 cases analyzed, the model demonstrated a sensitivity of 85.7%, specificity of 72.1%, positive predictive value of 12.2%, and negative predictive value of 99.1%. ROC-AUC was 0.813. All performance parameters were comparable to those in the previous study. The results strongly support the generalizability of the scoring system. This straightforward, easily interpretable, high-performance risk prediction system is expected to enhance the safety of denosumab treatment.