A risk factor-based predictive model for linezolid-induced anaemia: A 7-year retrospective study

Abstract

What is known and objective: The primary adverse reaction of linezolid is haematological toxicity, leading to thrombocytopenia and anaemia. This study aimed to investigate the risk factors of linezolid-induced anaemia (LI-AN) and establish a predictive model by multivariate logistic regression model analysis to predict LI-AN risks in Chinese adult patients.

Methods: Demographic and clinical data of patients who underwent linezolid therapy for more than three days between January 2014 and December 2020 in Zhongshan Hospital, Fudan University, were retrieved from the hospital's electronic medical record for analysis. Multivariate logistic regression analysis was employed to establish a predictive model, whose predictability was further evaluated by the area under the receiver operating characteristic (ROC) curve.

Results and discussion: The study comprised 298 patients among the 2322 patients who underwent linezolid treatment between 2014 and 2020. Among the 298 patients, 32 (10.7%) developed anaemia with an average of 11.4 (SD 6.2) days after the initiation of linezolid therapy. Multivariate logistic analysis revealed that age ≥60 years (odds ratio [OR] 2.815, 95% confidence interval [CI] 1.242-6.379), higher total bilirubin (TBi) (OR 1.031, 95% CI 1.011-1.051), eGFR < 60 ml/(min·1.73 m² ) (OR 2.537, 95% CI 1.054-6.106), duration of linezolid therapy (DLT) (OR 1.091, 95% CI 1.023-1.163) and intensive care unit (ICU) admittance (OR 2.664, 95% CI 1.150-6.174) were the independent risk factors for anaemia occurrence among patients receiving linezolid therapy. A logistic regression equation based on the five risk factors was subsequently established and transformed to obtain the calculation formula of the combined predictor: Y_{(Combined predictor)} = X_TBi + 34.5 × X_Age≥60 + 31.1 × X_eGFR<60 + 32.7 × X_ICU + 2.9 × X_DLT , (where Age ≥60 years, yes = 1, no = 0; eGFR < 60 ml/(min·1.73 m² ), yes = 1, no = 0; ICU admittance, yes = 1, no = 0). The area under the ROC curve of the combined predictors equation was 0.773 with an optimal cut-off point value of 92.4, corresponding to a 75.0% sensitivity and 76.7% specificity.

What is new and conclusion: LI-AN is associated with age (≥60 years), higher TBi, eGFR < 60 ml/(min·1.73 m² ), DLT and ICU admittance. Physicians should thus calculate the combined predictor value at the beginning of linezolid treatment to predict and evaluate the risk of LI-AN. An optimal cut-off value larger than 92.4 indicates that the patient has a higher LI-AN risk. As such, Hb levels should be monitored regularly, and dosage regimens adjusted accordingly to prevent anaemia occurrence. This study provides an evidence-based logistic model that reduces LI-AN incidences and promotes the safe clinical use of linezolid.

Keywords: anaemia; linezolid; predictive model; receiver operating characteristic curve; risk factors.