Updating Predictive Model for Cisplatin-Induced Acute Kidney Injury: Incorporating Concomitant Medications and the Standard Supportive Care.

Background and Objective: Cisplatin-induced acute kidney injury (C-AKI) is detrimental to adequate cancer treatment. While scoring systems to predict C-AKI are available, they do not account for the impact of concomitant medications. This study aimed to enhance the predictive model by incorporating concomitant medications as a new predictor. Methods: We included data from 1785 patients who received cisplatin at Iwate Medical University Hospital between April 2014 and March 2023. Initially, we assessed the performance of the existing model in our cohort. We then explored additional predictors to improve their discriminatory ability guided by the Akaike information criterion. Candidates for new predictors were concomitant anticancer and supportive care medications that were previously unexamined. Finally, we assessed the statistical usefulness of the updated model using the C-statistic and its clinical usefulness using net reclassification improvement (NRI) and decision curve analysis (DCA). Results: The discriminatory power of the existing model was poor, with a C-statistics of 0.621 (95% confidence interval [CI]: 0.582-0.660). Incorporating magnesium supplementation as a novel predictor significantly improved the model's performance, increasing the C-statistic to 0.695 (95% CI: 0.660-0.731). The updated model demonstrated a superior NRI of 0.143 (95% CI: 0.043-0.243). In the DCA, the updated model yielded higher net benefits for most threshold probabilities. Conclusion: The existing model did not demonstrate satisfactory clinical performance in our cohort. While incorporating magnesium supplementation significantly improved model discrimination, its status as standard care limits its utility as a predictive variable. These findings underscore the necessity of developing C-AKI prediction models within cohorts receiving uniform, contemporary supportive care regimens.