Appropriate antibiotic dosing to ensure early and sufficient target attainment is crucial for improving clinical outcome in critically ill patients. Parametric survival analysis is a preferred modeling method to quantify time-varying antibiotic exposure - response effects, whereas bias may be introduced in hazard functions and survival functions when competing events occur. This study investigated predictors of in-hospital mortality in critically ill patients treated with meropenem by pharmacometric multistate modeling. A multistate model comprising five states (ongoing meropenem treatment, other antibiotic treatment, antibiotic treatment termination, discharge, and death) was developed to capture the transitions in a cohort of 577 critically ill patients treated with meropenem. Various factors were investigated as potential predictors of the transitions, including patient demographics, creatinine clearance calculated by Cockcroft-Gault equation (CLCRCG ), time that unbound concentrations exceed the minimum inhibitory concentration (fT>MIC ), and microbiology-related measures. The probabilities to transit to other states from ongoing meropenem treatment increased over time. A 10 mL/min decrease in CLCRCG was found to elevate the hazard of transitioning from states of ongoing meropenem treatment and antibiotic treatment termination to the death state by 18%. The attainment of 100% fT>MIC significantly increased the transition rate from ongoing meropenem treatment to antibiotic treatment termination (by 9.7%), and was associated with improved survival outcome. The multistate model prospectively assessed predictors of death and can serve as a useful tool for survival analysis in different infection scenarios, particularly when competing risks are present.
© 2023 The Authors. CPT: Pharmacometrics & Systems Pharmacology published by Wiley Periodicals LLC on behalf of American Society for Clinical Pharmacology and Therapeutics.