External validation and comparison of clinical prediction models for cisplatin-associated acute kidney injury: a single-centre retrospective study.

Background: Cisplatin-associated acute kidney injury (C-AKI) is a major complication of cisplatin therapy. Although two clinical prediction models have been developed for the US population, their external validity in the Japanese population remains unclear. This study aimed to evaluate the external validity of these models and compare their predictive performances in a Japanese cohort.

Methods: We assessed the performance of two C-AKI prediction models developed by Motwani et al. and Gupta et al. in a retrospective cohort of 1,684 patients treated with cisplatin at Iwate Medical University Hospital. C-AKI was defined as a ≥ 0.3 mg/dL increase in serum creatinine or a ≥ 1.5-fold rise from baseline. Severe C-AKI was defined as a ≥ 2.0-fold increase or renal replacement therapy initiation. Model performance was evaluated using discrimination (area under the receiver operating characteristic curve [AUROC]), calibration, and decision curve analysis (DCA). Logistic recalibration was applied to adapt the model to the local population.

Results: The discriminatory performance for C-AKI was similar between the Gupta and Motwani models (AUROC, 0.616 vs. 0.613; p = 0.84). However, the Gupta model showed better discrimination of severe C-AKI (AUROC, 0.674 vs. 0.594; p = 0.02). Both models exhibited poor initial calibrations, which improved after recalibration. The recalibrated models yielded a greater net benefit in the DCA, with the Gupta model demonstrating the highest clinical utility in severe C-AKI.

Conclusions: Both models demonstrated discriminatory ability, with the Gupta model showing particular utility in predicting severe C-AKI. Given the observed miscalibration, recalibration is essential before applying these models in Japanese clinical practice.