Answers are grounded in this paper and cite their sources.
Purpose We aimed to evaluate the a priori predictive performance of four model-informed precision dosing (MIPD) software programs and identify factors associated with prediction error in inpatients receiving vancomycin. Methods We retrospectively included patients who received intravenous vancomycin and underwent therapeutic drug monitoring (TDM). C min , C max , and AUC 0−24 were predicted using SmartDose, PharmVAN, NextDose, and IOA. Performance was evaluated by prediction error (PE), absolute prediction error (APE), F30 (proportion of APE ≤ 30%), root mean square error (RMSE), and classification agreement rate (CAR). For the best-performing program, multivariable generalized linear modeling (GLM) identified factors associated with AUC 0−24 APE. Results IOA demonstrated the most favorable overall performance. Its median APE and F30 were 29.8% and 51.0% for C min ; 15.3% and 77.6% for C max ; and 20.9% and 70.9% for AUC 0−24 , respectively. IOA also had the lowest RMSEs (7.74 mg/L, 10.86 mg/L, and 202.68 mg·h/L) and the largest relative radar-chart area (58.9%). Acute kidney injury (AKI; Exp( B ) = 2.103, P = 0.024) and low daily dose (< 2 g/day; Exp( B ) = 1.498, P = 0.030) were independently associated with greater AUC 0−24 APE. Patients with AKI had lower F30 (20.0% vs 73.7%) and higher RMSE (507.16 vs 171.64 mg·h/L) than those without. Conclusions IOA showed relatively favorable a priori performance, but errors persisted, particularly in patients with AKI. A priori predictions should be interpreted cautiously in unstable renal function, and early TDM incorporation is recommended for Bayesian updating.
No takes yet. Share an insight, caveat, or question.
Wang et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: