Simulation-Based Optimization of Sampling Schedules for Model-Informed Precision Dosing of Once-Daily and 4-Times-Daily Busulfan in Pediatric Patients

Abstract

Background: Therapeutic drug monitoring (TDM) is crucial in optimizing the outcomes of hematopoietic stem cell transplantation by guiding busulfan (Bu) dosing. Limited sampling strategies show promise for efficiently adjusting drug doses. However, comprehensive assessments and optimization of sampling schedules for Bu TDM in pediatric patients are limited. We aimed to establish optimal sampling designs for model-informed precision dosing (MIPD) of once-daily (q24h) and 4-times-daily (q6h) Bu administration in pediatric patients.

Methods: Simulated data sets were used to evaluate the population pharmacokinetic model-based Bayesian estimation of the area under the concentration-time curve (AUC) for different limited sampling strategy designs. The evaluation was based on the mean prediction error for accuracy and root mean square error for precision. These findings were validated using patient-observed data. In addition, the MIPD protocol was implemented in the Tucuxi software, and its performance was assessed.

Results: Our Bayesian estimation approach allowed for flexible sampling times while maintaining mean prediction error within ±5% and root mean square error below 10%. Accurate and precise AUC 0-24h and cumulative AUC estimations were obtained using 2-sample and single-sample schedules for q6h and q24h dosing, respectively. TDM on 2 separate days was necessary to accurately estimate cumulative exposure, especially in patients receiving q6h Bu. Validation with observed patient data confirmed the precision of the proposed limited sampling scenarios. Implementing the MIPD protocol in Tucuxi software yielded reliable AUC estimations.

Conclusions: Our study successfully established precise limited sampling protocols for MIPD of Bu in pediatric patients. Our findings underscore the importance of TDM on at least 2 occasions to accurately achieve desired Bu exposures. The developed MIPD protocol and its implementation in Tucuxi software provide a valuable tool for routine TDM in pediatric hematopoietic stem cell transplantation.

Keywords: busulfan; hematopoietic stem cell transplantation; limited sampling strategy; model-informed precision dosing; pediatric; therapeutic drug monitoring.

FIGURE 1.

Overall methodology workflow for the study. AUC_0–24h, area under the plasma concentration curve, up to 24 hours after treatment start: cAUC_0–∞, cumulative AUC across the treatment; GUI: graphical user interface; MPE %: relative MPE; q6h: four-times-daily dosing (every 6 hours); q24h: once-daily dosing (every 24 hours); RMSE %: relative root mean square error. ¹Using our previously published model. ²Using a previously published and clinically evaluate model by Shukla et al.

FIGURE 2.

Simulation-based evaluation of limited sampling time points for accuracy (relative mean prediction error: MPE%) and precision (relative root mean square error: RMSE) of MAP estimation of AUC_0–24h and cAUC_0–∞. Schedules are ranked by ascending RMSE % of cAUC_0–∞. A and C illustrate the results of q6h (4 times daily) dosing, and B and D illustrate the results of q24h (once daily) dosing. Reference values for MPE and RMSE calculations were the full-simulated AUC values. For a priori estimation, no sampling times were input, and only the covariate model was used for calculation. The optimal 3-sample, 2-sample, and 1-sample schedules determined by the PFIM software are indicated by black arrows.

FIGURE 3.

Simulation-based selection of optimal sampling scenarios (limited sampling schedule + day of monitoring) based on accuracy (relative mean prediction error: MPE %) and precision (relative root mean squared error: RMSE) of MAP estimation of cAUC_0–∞. Scenarios are ranked by ascending RMSE % of cAUC_0–∞. Reference: full-simulated AUC values. A and C: MPE and RMSE of tested scenarios for q6h dosing. B and D: MPE and RMSE of tested scenarios for q24h dosing.

FIGURE 4.

Evaluation of optimal limited sampling scenarios with data from the validation patient cohort. All selected patients were monitored through extensive sampling on at least 2 occasions. Limited sampling schedules included selected time points on either day 1 + day 2 or day 1 + day 3, depending on data availability. Full day 1: full sampling on the first day of Bu administration. Dashed lines represent ±10% deviation from reference value. Dotted lines represent 80%–125% deviation interval. Error bars represent median with 95% CI. Red dots indicate patients with Bu dose modifications >25%; yellow, 10%–25%; green, and less than 10%. Panels A to D display evaluation of q6h scenarios, and panels E to H display evaluation of q24h scenarios.