Abstract
BACKGROUND: Linezolid (LZD) is used to treat infectious diseases caused by Gram-positive bacteria, but thrombocytopenia is one of the main adverse reactions to LZD administration. Early prediction of linezolid-induced thrombocytopenia (LI-TP) is of great importance to improve the clinical outcomes and prognoses. The aim of this study was to develop and validate a prediction model for LI-TP. METHODS: A retrospective cohort of hospitalized adults receiving LZD therapy (January 2014-June 2022) was analyzed. Independent risk factors for LI-TP were identified via logistic regression in the training set (n = 757). A nomogram model for LI-TP were developed based on independent risk factors, and verified in validation set (n = 123). RESULTS: The incidence of LI-TP was 13.5% (102/757). A logistic regression model was developed based on the seven independent risk factors, including age (≥ 60 y), duration of LZD therapy (> 11 d), bPLT (< 308 × 10(9)/L), ALT (> 100 IU/L), Ccr (< 67.5 mL/min), and concomitant use with VPA or Tac (p < 0.01) and transformed into a quantifiable nomogram. The nomogram demonstrated strong discrimination with AUCs of 0.760 in training (95% CI: 0.709-0.812, P < 0.001) and 0.767 in validation (95% CI: 0.635-0.899, P < 0.001). The calibration curves and Hosmer-Lemeshow tests confirmed good reliability and specificity of the nomogram model. CONCLUSION: This nomogram provides a practical tool for stratifying LI-TP risk, which provide an important reference for enabling timely clinical interventions to enhance LZD safety.