Prediction of paclitaxel-induced myelosuppression: Validation of a regression model using creatinine clearance and other risk factors obtained retrospectively in hospitalized cancer patients.

OBJECTIVE To develop a predictive model for paclitaxel (PTX)-induced moderate-to-severe myelosuppression in hospitalized cancer patients and to evaluate its accuracy through cross-validation.

MATERIALS AND METHODS A total of 348 cancer patients treated with PTX at Yingshang County People's Hospital between January 2022 and August 2025 were retrospectively reviewed. Clinical data from 278 patients were used to develop the model, including sex, age, underlying diseases, and baseline laboratory parameters at admission. Multivariate logistic regression was used to identify independent risk factors for moderate-to-severe myelosuppression and to construct the predictive model. The model was evaluated using receiver operating characteristic (ROC) curve analysis. An additional 70 patients were used for cross-validation to assess the model's accuracy.

RESULTS The incidence of PTX-induced moderate-to-severe myelosuppression was 32.37% in the 278-patient cohort. Multivariate analysis identified the following independent predictors: creatinine clearance rate (OR = 3.81, 95% CI: 1.853 - 7.836, p < 0.001), red blood cell count (OR = 2.69, 95% CI: 1.263 - 5.726, p = 0.01), hemoglobin (OR = 6.49, 95% CI: 2.431 - 17.328, p < 0.001), neutrophil count (OR = 3.196, 95% CI: 1.077 - 9.485, p = 0.036), and nutritional status (OR = 2.373, 95% CI: 1.008 - 5.586, p = 0.048). The predictive model incorporating these five factors demonstrated strong performance, achieving an AUC of 0.845, a sensitivity of 75.56%, and a specificity of 78.19%. Furthermore, cross-validation on an independent cohort of 70 patients confirmed a prediction accuracy of 84.29%.

CONCLUSION The developed predictive model demonstrates strong performance in assessing the risk of PTX-induced moderate-to-severe myelosuppression and may serve as a useful tool for early identification of high-risk patients, supporting clinical decision-making and personalized treatment strategies.