Toward Optimized Platinum Therapy: A Risk Prediction Model for Toxicity Based on Platinum Accumulation and Its Impact on Treatment Outcomes

Background: Platinum-based chemotherapeutic agents are fundamental in the treatment of various malignancies, yet their long-term efficacy is often compromised by cumulative toxicity. This study investigates the association between platinum accumulation and length of hospital stay (LOS), and develops a predictive tool to optimize individualized platinum therapy. Methods: A total of 300 cancer patients receiving platinum-based chemotherapy were enrolled. Serum platinum concentration was measured on day 21 ± 3 after drug administration. The LOS and adverse events were recorded. Mediation analysis explored underlying mechanisms, and a classification and regression tree (CART) model was constructed for risk prediction. Results: Elevated serum platinum concentration significantly prolonged LOS, primarily mediated through reduced red blood cell count (mediation effect accounted for 32.3%). The CART model identified serum platinum concentration > 534.40 μg/L as a key threshold and major predictor of adverse reactions, alongside age > 60.5 years, serum creatinine > 94.50 μmol/L, and use of cisplatin or carboplatin. The model demonstrated an AUC of 0.782 and accuracy of 75.7%. Conclusion: A serum platinum concentration > 534.40 μg/L serves as a clinically meaningful cutoff for identifying patients at high risk of platinum‑related toxicity. The proposed decision‑tree model offers a clinically actionable framework for the early recognition of these high‑risk individuals, facilitating timely dose management and treatment optimization in clinical practice.