Abstract:

Objective To investigate the determinants of linezolid-associated neurological adverse reactions in patients with multidrug-resistant tuberculosis and develop a risk prediction model for such adverse events. Methods A prospective cohort study design was employed to select 120 patients with drug-resistant pulmonary tuberculosis who received a chemotherapy regimen containing linezolid at Guangzhou Chest Hospital from April 2023 to January 2024 as the study population. Clinical data， adverse reactions， and plasma concentration of linezolid were collected during fasting and at 2 hours post-medication. Univariate analysis and multivariate logistic regression were conducted to identify factors influencing linezolid-related neurological adverse reactions. Furthermore， a prediction model for such adverse reactions was developed， and its predictive efficacy and calibration ability were evaluated using ROC analysis. Results Re-treatment（OR = 2.540，P =0.028）， coexistence of cavities （OR = 4.092，P =0.021）， anemia （OR = 10.921，P = 0.005）， and C_min ≥ 0.7665 mg/L （OR = 6.813，P < 0.001） are independent risk factors for the occurrence of linezolid-related neurological adverse reactions. The prediction model， based on these four factors， exhibits an AUC of 0.851 （95% CI：0.774 ~ 0.929）， accompanied by a Youden index of 0.590， a sensitivity of 66.7%， and a specificity of 92.3%. Moreover， the prediction model demonstrates excellent calibration ability. （Hosmer-lemeshow χ² = 8.719， P = 0.273）. Conclusion In MDR/RR-TB patients， the presence of cavitation， retreatment， and anemia may confer a heightened risk of linezolid-related neurological adverse reactions. A risk prediction model incorporating these four indicators demonstrates significant predictive value for the occurrence of such adverse events.

Key words: linezolid, multidrug-resistant, pulmonary tuberculosis, nervous systemadverse reactions, prediction model