Development of a simplified multivariable model to predict neutropenic complications in cancer patients undergoing...

Neutropenic complications remain the major dose-limiting toxicities of cancer chemotherapy. The aim of this study was to develop and internally validate a comprehensive and easily measurable scoring system for prediction of severe or febrile neutropenia in the first chemotherapy cycle of patients with solid tumors or lymphoma.

This prospective cohort study included consecutive patients at a tertiary referral hospital. Many clinical and laboratory-independent variables were measured at baseline. A multivariable logistic regression analysis was applied after unadjusted analysis, and the multivariable model was transformed into a simplified risk score based on 6 bootstrapped regression coefficients. The simplified scoring system was internally validated using cross-validation. All statistical tests were two-sided.

A total of 305 patients were enrolled and followed during 1732 chemotherapy cycles. Of these, 259 were eligible for analysis. The multivariable model revealed 6 predictive factors for severe or febrile neutropenia (scores in parentheses): high-risk regimen without colony-stimulating factor (4 points), intermediate-risk regimen without colony-stimulating factor (3 points), age > 65 years and elevated ferritin (3 points), body mass index < 23 kg/m² and body surface area < 2 m² (2 points), estimated glomerular filtration rate < 60 mL/min/1.73m² (2 points), and elevated C-reactive protein (1 point). The receiver operating characteristic curve was 0.832 (95% confidence interval [Cl], 0.767-0.897) for the simplified model and 0.816 (95% Cl, 0.771-0.860) for the cross-validation.

We developed and internally validated a user-friendly prediction model to guide personalized decision-making using available clinical data and few cost-effective laboratory tests. External validation in other centers with different patients is required.