Development of User-Friendly Bayesian Predictive Platform for Blood Boron-10 Pharmacokinetics following Intravenous Infusion of [10B] L-4-BORONOPHENYLALANINE
Woo Hyoung Kim1, Choi Seung Chan2, Jungyu Yi3,4, Sang Min Lee2, Kyung Ran Mun2, Hyeong-Seok Lim2
: (1) Division of Pharmaceuticals and Clinical Development, DawonMedax Co., Ltd., Seoul, South Korea, (2) Department of Clinical Pharmacology and Therapeutics, Asan Medical Center, Ulsan University College of Medicine, Pungnap-2-dong, Seoul, Republic of Korea, (3) Division of Medical Device, DawonMedax Co, Ltd., Seoul, South Korea, (4) Department of Nuclear Engineering, Hanyang University, Seoul, South Korea.
Introduction: Boron neutron capture therapy (BNCT) allows high-precision radiotherapy against tumor using boron-10 (10B) with tumor-localizing characteristics and has a strong tendency to capture thermal neutrons. As the radiation dose delivered by BNCT is affected by two factors which are neutron flux and boron concentration of each organ during the neutron irradiation, precise monitoring of each component is crucial in the management of accurate radiation dose delivery. Thus, it is important to accurately predict the blood 10B concentration during the neutron irradiation to deliver the prescribed dose as planned. This study was performed to develop user-friendly Bayesian predictive platform for pharmacokinetics (PK) of 10B which is clinically applicable in BNCT.
Objectives: The major purposes of this study were followed as:
1) A user-friendly Bayesian estimation platform has been developed for subjects undergoing Boron neutron capture therapy (BNCT).
2) The platform is designed to improve the accuracy of estimation for such subjects.
3) The platform specifically targets BPA (boronophenylalanine) for boron neutron capture therapy (BNCT).
Methods: Population PK model for 10B was constructed in NONMEM® 7.4 (ICON Development Solutions, USA) using blood 10B concentration over time data following intravenous infusion of boronophenylalanine (BPA) which were digitized from previous study results. The PK model was used as a prior distribution model for the Bayesian prediction. The predictive model was executed using R (version 4.03) with user-friendly interface provided by Shiny package. Individual blood 10B concentration over time was predicted, which were graphically displayed with relevant numerical information Sensitivity analyses were conducted to evaluate the predictive performance of the platform and identify optimal PK sampling time for blood 10B. In modeling analysis, 10/208.21 was applied as a conversion factor in consideration of the molecular weight of boron (10B) and 10B -BPA. The relationship between the representative value of the group and the individual value of the pharmacokinetic parameter was described in the following equation. ???? = ?????? × ??????(η?? ) (PTV is the representative value of the group, ηi is a random number randomly extracted from a normal distribution with a mean of 0 and a specific variance value)
Results: Elimination rate constant from central compartment value ( ) was estimated as 0.006(1/min). Volume of distribution of central compartment( ) was estimated as 0.252(L). A population pharmacokinetic (PK) model for 10B, accurately predict observed concentrations. A sensitivity analysis was conducted to determine the optimal sampling time and number of blood samples for B concentrations for individual PK prediction. The predictive platform was stable and had high accuracy and precision in predicting 10B concentration. This platform proved to be a valuable tool in clinical settings, allowing easy, and rapid prediction of individual 10B PK for accurate and precise dosing regimens in patients undergoing boron neutron capture therapy. The individual visual predictive check (VPC) indicates that the model's ability to predict boron concentration is strong. This is because the actual measurements of boron concentration tend to fall within the model's 95% prediction range, and when they fall outside this range, it is typically at the upper or lower limit.
Conclusions: Overall, the study showed that this predictive platform encompassing PK model and Shiny package can be used in the clinical trial for BNCT treatment. The study has developed a platform that is capable of predicting the therapeutic effect of BNCT more accurately, thereby improving the therapeutic outcome for patients undergoing this type of radiation therapy. The current platform has been implemented in an ongoing clinical trial (NCT05737212), and following completion of BPA administration, the platform instantly predicted the blood boron concentration over an hour with less than a 5% error margin. This predicted platform has the potential to play a crucial role in clinical trials aimed at evaluating the effectiveness of BNCT, as it provides a user-friendly, reliable tool for assessing the treatment's efficacy. By enabling a more precise estimation of the therapeutic effect, the platform can help clinicians tailor treatment plans to individual patients, ultimately improving the overall quality of care in BNCT.
References:
[1] Palmer, M. R., Goorley, J. T., Kiger, W. S., Busse, P. M., Riley, K. J., Harling, O. K., & Zamenhof, R. G. (2002). Treatment planning and dosimetry for the Harvard-MIT Phase I clinical trial of cranial neutron capture therapy. International Journal of Radiation Oncology*Biology*Physics, 53(5), 1361–1379.
[2] Chanana, Capala, Chadha, et al. Boron neutron capture therapy for glioblastoma multiforme: interim results from the phase I/II dose-escalation studies. Neurosurgery. 1999;44(6):1182-1192; discussion1192-3.
