Background: Vancomycin dosing is commonly adjusted using therapeutic drug monitoring (TDM), traditionally guided by trough concentrations. An Auckland before-after cohort study evaluated a local change from empirical trough-guided TDM to NextDose-informed Bayesian 24-hour area under the concentration-time curve (AUC24)-guided target concentration intervention (TCI). In routine care, dosing performance is shaped by pharmacokinetics and ward execution, including initial dosing, sampling timing, delayed or omitted doses, and attrition. A clinical trial simulation (CTS) should therefore reflect both exposure patterns and study conduct before wider scenario evaluation.
Objective: To develop an execution-aware vancomycin CTS for the Auckland cohort setting and benchmark simulated covariates, AUC24 exposure and execution events against selected anchors from the Auckland before-after study.
Methods: Virtual patients were resampled from local vancomycin treatment data after filtering eligible records from Auckland City Hospital and Starship Children’s Hospital. The primary simulation included 200 enrolled virtual patients, matching the AUC-guided cohort size. The CTS was implemented in R, NONMEM and Wings for NONMEM. Initial dosing followed local practice, with a 25 mg/kg loading policy operationalized through local weight bands and maintenance dosing based on a Cockcroft-Gault creatinine-clearance nomogram before concentration feedback. The first concentration was scheduled before the fourth dose: within 0–2 h before dose for the trough-comparator pathway, and not restricted to trough sampling for Bayesian TCI. Bayesian TCI began after implemented dose history and at least one concentration observation were available, targeting AUC24 = 500 mg·h/L. Delayed doses, missed or cancelled doses, missing samples and dropout were simulated using prespecified execution-event probabilities, with additional clinical-response triggers for high exposure or simulated nephrotoxicity risk.
Results: Among 200 enrolled virtual patients, the CTS closely reflected key features of the AUC-guided clinical cohort, including body weight, serum creatinine, sex distribution and pediatric/adult structure, although the simulated cohort was younger. Median age, weight and serum creatinine were 49 years, 77 kg and 71 µmol/L, compared with 55 years, 79 kg and 73 µmol/L in the reference cohort. During the 7-day simulated treatment course, AUC24 was broad after empirical initial dosing: median 570 mg·h/L, 5th–95th percentile 354–870, with 49.2% within 400–600 mg·h/L. After Bayesian feedback became available, AUC24 moved toward target and the final review remained close to target: median 491 mg·h/L, 5th–95th percentile 409–590, with 94.1% of evaluable patients within 400–600 mg·h/L. Over the same treatment-course horizon, delayed dosing occurred in 14.7% of dose-event opportunities, missed or cancelled dosing in 12.6%, and missing samples in 4.2%. Final cumulative attrition was 23/200 enrolled patients.
Conclusions: This execution-aware CTS closely reflected key features of the Auckland before-after study: broad exposure after empirical initial dosing, convergence toward the AUC24 target after Bayesian feedback, and recorded dosing, sampling and attrition events. These auditable outputs support further calibration against the final clinical dataset before broader scenario or economic evaluation.