Aims The efficacy and toxicity of polymyxin B (PB) are closely related to its pharmacokinetic/pharmacodynamic (PK/PD) index area under the concentration-time curve (AUC) to minimum inhibitory concentration (MIC) ratio. The purpose of this study was to obtain PK data for PB in Chinese severe pneumonia patients and establish appropriate blood sampling time points for the PB therapeutic drug monitoring (TDM). Subject and method After treatment with at least four doses of PB (50 IU, q12h), the blood samples were collected immediately after the end of infusion (C-0) and 1.5, 2, 4, 6, 8, and 12 h (C-1.5, C-2, C-4, C-6, C-8, C-12) after PB administration. The PB blood plasma concentrations were determined using an ultra-performance liquid chromatography-tandem mass spectrometer (UPLC-MS/MS). All 42 patients were randomly divided into modeling (n= 24) and validation (n= 18) groups. The relationship between AUC(ss,24h)and PB plasma concentration at each time point in modeling group was analyzed using limited sampling strategy and a PK method based on one-compartment with correction model. Results C(6)scheme was found to provide the most accurate prediction of AUC(ss,24h)values (r(2)= 0.984) with the target value of 1.9-4.2 mu g/ml at steady state to reach the 50-100 mu g h/ml criteria of AUC(ss,24h). C(0)with target value of 1.0-2.8 mu g/ml can be considered an alternative sampling scheme (r(2)= 0.900) but prediction deviation may exist. C(0)and C(max)sampling scheme also demonstrated good predicting ability of AUC values using PK model. Conclusion This study provides a clear plan for the implementation of TDM of PB, which is useful for optimizing the dosing regimen and individualizing treatment in severe pneumonia patients.