Therapeutic drug monitoring and receiver operating characteristic curve prediction may reduce the development of linezolid-associated thrombocytopenia in critically ill patients

To investigate the risk factors associated with the development of thrombocytopenia, and define the thresholds of efficacy and safety in critically ill patients who received linezolid therapy. A retrospective study was performed in critically ill patients treated with linezolid. Risk factors associated with thrombocytopenia were identified via medical records and trough levels (C_min) measured during linezolid treatment. By establishing a logistic model, the risks were predicted by the receiver operating characteristic (ROC) curve and the thresholds of efficacy and safety were identified in the patients. Logistic analysis showed that, weight (OR = 0.906; 95 % CI, 0.839–0.978; P = 0.011), baseline platelet count (OR = 0.989; 95 % CI, 0.977–1.000; P = 0.049), C_min (OR = 1.545; 95 % CI, 1.203–1.983; P = 0.001), and APACHE II score (OR = 1.130; 95 % CI, 1.003–1.273; P = 0.044) were significant factors for linezolid-associated thrombocytopenia. The area under the ROC curve of the combined predictor was larger based on the above factors. When the Youden index was the maximum, the best optimal cut-off point was 205.6 on the ROC curve; when C_min ≥ 2 mg/L, the probability of bacterial eradication was more than 80 %; when C_min ≥ 6.3 mg/L, the probability of thrombocytopenia was more than 50 %. In clinical practice, when the calculating results of the combined predictor ≤205.6, the risk of the development of thrombocytopenia may be higher. Furthermore, maintenance of C_min between 2 and 6.3 mg/L over time may be helpful in retaining appropriate efficacy and reducing the associated thrombocytopenia.