A Prediction Model for Chemotherapy-Induced Thrombocytopenia Based on Real-World Data and a Close Relationship...

Chemotherapy-induced thrombocytopenia (CIT) can lead to chemotherapy dose delay or reduction, and even serious bleeding. This study aimed to develop a CIT-predicting model based on the laboratory indices of cancer patients undergoing chemotherapy.

From Jun 1, 2017 to Dec 30, 2021, a total of 2043 patients who had received 7676 cycles of chemotherapy were retrospectively enrolled. A logistic regression analysis was performed to identify predictive factors, on the basis of which a nomogram model for predicting CIT was established. A bootstrapping technique was applied for internal validation. A generalized additive mixed model (GAMM) was constructed to analyze the trends in the changes of aspartate aminotransferase (AST), ratio of AST to alanine transaminase (ALT) (AST/ALT ratio), and platelet (PLT) count in patients with solid tumors. P values ≤0.05 were considered statistically significant.

The patient-based incidence of CIT was 20.51% and the cycle-based incidence was 10.01%. The multivariate analysis showed that AST level, AST/ALT ratio, and total bilirubin (Tbil), white blood cell (WBC), platelet (PLT), hemoglobin (Hb) levels were significantly associated with the risk of CIT. The GAMM analysis showed that PLT level was inversely associated with AST/ALT ratio and AST level, more significantly with AST/ALT ratio. And both exhibited statistically predictive abilities for CIT. The model achieved an area under the receiver operating characteristic curve (AUC) of 0.793, a sensitivity of 0.543 and a specificity of 0.930.

The AST/ALT ratio was inversely associated with the CIT risk in cancer patients. The GAMM model based on laboratory indices presented a high accuracy in predicting the risk of CIT, and a potential to be translated into clinical management.