Limited Sampling Strategies Supporting Individualized Dose Adjustment of Intravenous Busulfan in Children and Young Adults

A limited sampling strategy for estimating busulfan exposure in pediatric hematopoietic stem cell transplantation

Background

Busulfan (Bu) is the foundation of conditioning regimens for pediatric hematopoietic stem cell transplantation (HSCT). Evidence indicates that the efficacy and side effects of Bu are intimately tied to the area under its concentration-time curve (AUC). Given its cytotoxic nature and a small therapeutic index, coupled with marked inter-individual pharmacokinetic variability, Bu requires therapeutic drug monitoring to facilitate individualized therapy. However, research investigating the relationship between Bu exposure and clinical outcomes among the Chinese population remains scarce. This study aimed to develop a limited sampling strategy (LSS) for estimating Bu exposure in pediatric HSCT recipients using multiple linear regression (MLR) analysis to predict the AUC0-360.

Methods

We enrolled 26 pediatric patients who underwent Bu-based conditioning for HSCT. Blood samples were collected at 11 time points after Bu infusion. Pharmacokinetic parameters were calculated using non-compartmental methods. MLR models were developed using 1-4 sampling points to predict the AUC0-360. Model accuracy was assessed using the Jackknife and Bootstrap methods, with consistency evaluated via intraclass correlation coefficient (ICC) and Bland-Altman (BA) analyses.

Results

The mean ± standard deviation (SD) for AUC0-t, mean residence time 0-t, clearance, and volume of distribution were 845.54 ± 111.03 μmol min/L, 181.37 ± 10.55 min, 0.23 ± 0.04 L/h/kg, and 0.73 ± 0.15 L/kg, respectively. Models with 2-4 sampling points showed improved prediction accuracy compared to single-point models. The four-point model (60, 135, 240 and 360 min) demonstrated the highest accuracy with an adjusted r 2 of 0.965. Internal validation confirmed the models' stability and accuracy, with the four-point model exhibiting the best performance. External validation using three additional cases supported the predictive accuracy of the model.

Conclusion

The LSS model developed in this study accurately predicts the Bu AUC0-360 with 2-4 sampling points, offering a practical and clinically valuable tool for therapeutic drug monitoring in pediatric HSCT recipients. The four-point model was found to be the most accurate and is recommended for clinical applications.

Leveraging machine learning in limited sampling strategies for efficient estimation of the area under the curve in pharmacokinetic analysis: a review

OBJECTIVE

Limited sampling strategies are widely employed in clinical practice to minimize the number of blood samples required for the accurate area under the curve calculations, as obtaining these samples can be costly and challenging. Traditionally, the maximum a posteriori Bayesian estimation has been the standard method for the area under the curve estimation based on limited samples. However, machine learning is emerging as a promising alternative for this purpose. Here, we review studies that utilize machine learning approaches to develop limited sampling strategies and compare the strengths and weaknesses of these machine learning methods.

METHODS

We searched the literature for studies that used machine learning to estimate the area under the curve using a limited sampling strategy approach.

RESULTS

We identified ten studies that developed machine learning models to estimate the area under the curve for six different drugs. Several of these models demonstrated good accuracy and precision in area under the curve estimation in reference to the traditional Bayesian approach, highlighting the potential of machine learning models in precision dosing.

CONCLUSIONS

Despite these promising early results, the development of machine learning for limited sampling strategies is still in its early stages. Further research might be needed to validate machine learning models with larger, high-quality clinical datasets to ensure their reliability and applicability in clinical settings.

Limited Sampling Strategies for Estimating Busulfan Area Under the Concentration-Time Curve: Based on Peak and Trough Concentrations in Saliva

Therapeutic drug monitoring for busulfan is currently performed by multiple plasma sampling. Saliva is considered a noninvasive therapeutic drug monitoring matrix. This study aimed to investigate intravenous busulfan pharmacokinetics (PK) in plasma and saliva, and establish a limited sampling strategy (LSS) for predicting the area under the concentration-time curve from time zero to infinity in plasma (AUC0-∞,p) by using saliva samples. Therefore, the PK of busulfan was studied in 37 Chinese patients. Pearson correlation analysis was used to evaluate the correlation between the AUC of busulfan in plasma and saliva. LSS models were established by the multiple linear regression analysis. The prediction error, the mean prediction error, and the root mean square error were used to evaluate the predictive accuracy. The agreement between the predicted and observed AUC0-∞ in saliva was investigated by the intraclass correlation coefficient and Bland-Altman analysis. The accuracy and robustness of the models were evaluated by using the bootstrap procedure. The result of PK analysis 62.2% of patients (23/37) was within the target range of AUC0-∞,p . A good correlation between saliva and plasma busulfan AUC0-∞ was observed (r = 0.63, p < .01). The bias and precision of the models 7 and 13 were less than 15%. The intraclass correlation coefficient exceeded 0.9, and the limits of agreement were within ±15%. The 2-point LSS model in saliva is a convenient and desirable approach to predict the AUC0-∞ of 4 times daily intravenous busulfan in plasma, which can be used to design personalized dosing for busulfan.

Limited Sampling Strategy Using End of Infusion and Six-Hour Concentrations Overestimates Intravenous Busulfan Clearance Compared With Standard Six-Point Sampling in Hematopoietic Stem Cell Transplant Patients

BACKGROUND

Therapeutic drug monitoring for busulfan (Bu) is important to improve outcomes of hematopoietic stem cell transplantation. However, standard therapeutic drug monitoring requires multiple samples and is inconvenient, labor-intensive, and costly. Accordingly, a limited sampling strategy (LSS) was evaluated, using 2-point sampling at end of infusion and at 6 hours, and the area-under-the-curve and Bu clearances (CLs) were compared with the results obtained from the standard sampling strategy (SSS) using 5-6 samples.

METHOD

The analysis was based on retrospective clinical data from 202 patients receiving intravenous Bu before hematopoietic stem cell transplantation for malignant or nonmalignant conditions. Bu plasma concentrations were measured via liquid chromatography tandem-mass spectrometry, and pharmacokinetic parameters were calculated using the PKCNA package in R program.

RESULT

A total of 502 doses were analyzed by applying SSS and LSS. Using the modified Bland-Altman plot, the mean percentage difference in CL between the SSS and LSS estimates of Bu 6-hourly regimen was -41% (Limits: -53% and -30%). In the once daily regimen, the mean difference in CL between the 2 strategies on the modified Bland-Altman plot was -22% (Limits: -66% and +22%).

CONCLUSIONS

The Bu CL values estimated based on the BU concentration at end of infusion and at 6 hours postinfusion were significantly higher than the values obtained via the SSS.

Predicting busulfan exposure in patients undergoing hematopoietic stem cell transplantation using machine learning techniques

PURPOSE

This study aimed to establish an optimal model to predict the busulfan (BU) area under the curve at steady state (AUCss) by using machine learning (ML).

PATIENTS AND METHODS

Seventy-nine adult patients (age ≥18 years) who received BU intravenously and underwent therapeutic drug monitoring from 2013 to 2021 at Fujian Medical University Union Hospital were enrolled in this retrospective study. The whole dataset was divided into a training group and test group at the ratio of 8:2. BU AUCss were considered as the target variable. Nine different ML algorithms and one population pharmacokinetic (pop PK) model were developed and validated, and their predictive performance was compared.

RESULTS

All ML models were superior to the pop PK model (R2 = 0.751, MSE = 0.722, 14 and RMSE = 0.830) in model fitting and had better predictive accuracy. The ML model of BU AUCss established through support vector regression (SVR) and gradient boosted regression trees (GBRT) had the best predictive ability (R2 = 0.953 and 0.953, MSE = 0.323 and 0.326, and RMSE = 0.423 and 0.425).

CONCLUSION

All the ML models can potentially be used to estimate BU AUCss with the aim of facilitating rational use of BU on the individualized level, especially models built by SVR and GBRT algorithms.

A Method for Evaluating Robustness of Limited Sampling Strategies—Exemplified by Serum Iohexol Clearance for Determination of Measured Glomerular Filtration Rate

In combination with Bayesian estimates based on a population pharmacokinetic model, limited sampling strategies (LSS) may reduce the number of samples required for individual pharmacokinetic parameter estimations. Such strategies reduce the burden when assessing the area under the concentration versus time curves (AUC) in therapeutic drug monitoring. However, it is not uncommon for the actual sample time to deviate from the optimal one. In this work, we evaluate the robustness of parameter estimations to such deviations in an LSS. A previously developed 4-point LSS for estimation of serum iohexol clearance (i.e., dose/AUC) was used to exemplify the effect of sample time deviations. Two parallel strategies were used: (a) shifting the exact sampling time by an empirical amount of time for each of the four individual sample points, and (b) introducing a random error across all sample points. The investigated iohexol LSS appeared robust to deviations from optimal sample times, both across individual and multiple sample points. The proportion of individuals with a relative error greater than 15% (P15) was 5.3% in the reference run with optimally timed sampling, which increased to a maximum of 8.3% following the introduction of random error in sample time across all four time points. We propose to apply the present method for the validation of LSS developed for clinical use.

External Evaluation of Population Pharmacokinetic Models of Busulfan in Chinese Adult Hematopoietic Stem Cell Transplantation Recipients

Objective: Busulfan (BU) is a bi-functional DNA-alkylating agent used in patients undergoing hematopoietic stem cell transplantation (HSCT). Over the last decades, several population pharmacokinetic (pop PK) models of BU have been established, but external evaluation has not been performed for almost all models. The purpose of the study was to evaluate the predictive performance of published pop PK models of intravenous BU in adults using an independent dataset from Chinese HSCT patients, and to identify the best model to guide personalized dosing.

Methods: The external evaluation methods included prediction-based diagnostics, simulation-based diagnostics, and Bayesian forecasting. In prediction-based diagnostics, the relative prediction error (PE%) was calculated by comparing the population predicted concentration (PRED) with the observations. Simulation-based diagnostics included the prediction- and variability-corrected visual predictive check (pvcVPC) and the normalized prediction distribution error (NPDE). Bayesian forecasting was executed by giving prior one to four observations. The factors influencing the model predictability, including the impact of structural models, were assessed.

Results: A total of 440 concentrations (110 patients) were obtained for analysis. Based on prediction-based diagnostics and Bayesian forecasting, preferable predictive performance was observed in the model developed by Huang et al. The median PE% was -1.44% which was closest to 0, and the maximum F₂₀ of 57.27% and F₃₀ of 72.73% were achieved. Bayesian forecasting demonstrated that prior concentrations remarkably improved the prediction precision and accuracy of all models, even with only one prior concentration.

Conclusion: This is the first study to comprehensively evaluate published pop PK models of BU. The model built by Huang et al. had satisfactory predictive performance, which can be used to guide individualized dosage adjustment of BU in Chinese patients.

Optimization of Busulfan Dosing Regimen in Pediatric Patients Using a Population Pharmacokinetic Model Incorporating GST Mutations

Purpose

The aim of this study was to develop a novel busulfan dosing regimen, based on a population pharmacokinetic (PPK) model in Chinese children, and to achieve better area under the concentration-time curve (AUC) targeting.

Patients and Methods

We collected busulfan concentration-time samples from 69 children who received intravenous busulfan prior to allogeneic hematopoietic stem cell transplantation (allo-HSCT). A population pharmacokinetic model for busulfan was developed by nonlinear mixed effect modelling and was validated by an external dataset (n=14). A novel busulfan dosing regimen was developed through simulated patients, and has been verified on real patients. Limited sampling strategy (LSS) was established by Bayesian forecasting. Mean absolute prediction error (MAPE) and relative root mean Squared error (rRMSE) were calculated to evaluate predictive accuracy.

Results

A one-compartment model with first-order elimination best described the data. GSTA1 genotypes, body surface area (BSA) and aspartate aminotransferase (AST) were found to be significant covariates of Bu clearance, and BSA had significant impact of the volume of distribution. Moreover, two equations were obtained for recommended dose regimens: dose (mg)=34.14×BSA (m²)+3.75 (for GSTA1 *A/*A), Dose (mg)=30.99×BSA (m²)+3.21 (for GSTA1 *A/*B). We also presented a piecewise dosage based on BSA categories for each GSTA1 mutation. A two-point LSS, two hours and four hours after dosing, behaved well with acceptable prediction precision (rRMSE=1.026%, MAPE=6.55%).

Conclusion

We recommend a GSTA1-BSA and BSA-based dosing (Q6 h) based on a PPK model for personalizing busulfan therapy in pediatric population. Additionally, an optimal LSS (C_2h and C_4h) provides convenience for therapeutic drug monitoring (TDM) in the future.