For over two decades, imatinib has transformed chronic myeloid leukemia from a death sentence into a manageable chronic condition, yet its full potential remains stubbornly out of reach for many patients. The drug exhibits enormous inter-individual pharmacokinetic variability, meaning that a standard dose can produce wildly different blood concentrations from one person to the next. Plasma levels that fall too low risk therapeutic failure and the emergence of drug-resistant clones, while concentrations that climb too high can trigger debilitating toxicities. Therapeutic drug monitoring, the practice of measuring drug levels and adjusting doses to keep them within a proven target window, has long been proposed as a way to navigate this variability, but the logistical reality has kept it from becoming routine clinical practice. Conventional monitoring demands multiple venous blood draws at precisely timed intervals, followed by batch analysis in centralized laboratories, a cumbersome and slow workflow that is fundamentally incompatible with the fast decision-making needed at the point of care.
A new study published in BMC Pharmacology and Toxicology by a Swiss research team takes a decisive step toward dismantling these barriers. Pétermann and colleagues tackled the core problem that has stymied point-of-care imatinib monitoring: how to extract clinically actionable exposure metrics from the fewest possible blood samples, collected at flexible times that suit a busy outpatient clinic, without sacrificing accuracy. Their work leverages a population pharmacokinetic modeling approach, a mathematical framework that describes how a drug is absorbed, distributed, and eliminated across a patient population while quantifying both the average behavior and the random variability between individuals. From this, they set out to identify and validate optimal limited sampling strategies—pared-down blood draw schedules that can reliably estimate a patient’s total drug exposure, usually expressed as the area under the concentration–time curve over a dosing interval.
The technical heart of the investigation lies in Bayesian estimation. The team first constructed a robust population model using dense pharmacokinetic data from patients on long-term imatinib therapy, characterizing the drug’s absorption lag, first-order elimination, and the significant between-occasion variability that can cause the same patient to have different concentration profiles on different days. They then computationally simulated thousands of candidate sampling schedules, each comprising just two to three time points scattered across a dosing interval, and evaluated how well a Bayesian estimator—primed with the population model as a prior—could forecast the true area under the curve when fed only those sparse measurements. The critical innovation was to search not only for schedules with the best predictive performance but also for those that remain robust when sample collection times deviate by up to thirty minutes from the nominal plan, a nod to the unpredictability of real-world clinical workflows.
What emerged from these simulations is a set of streamlined sampling schemes that are both pragmatic and remarkably precise. A two-sample strategy taking blood at two and six hours post-dose, or a three-point approach adding a trough level just before the next dose, provided estimates of imatinib exposure with a bias of less than five percent and an imprecision below fifteen percent compared to full twelve-hour profiles. Importantly, the models corrected for the drug’s complex absorption kinetics, including the occasional double-peak phenomenon linked to enterohepatic recirculation, and accounted for the impact of covariates such as body weight and the co-administration of proton pump inhibitors that can alter gastric pH and drug dissolution. The result is an algorithmic backbone that can be embedded directly into a mobile device or a benchtop analyzer, instantly translating a pair of finger-prick concentrations into a personalized dosing recommendation.
The researchers validated their limited sampling strategies on a separate cohort of patient data, confirming that the Bayesian estimators retained their accuracy when applied to individuals who were not part of the original model-building set. This external validation is crucial because it mimics the genuine clinical scenario where a new patient walks into the clinic with no prior pharmacokinetic information. The system only needs the population prior, the patient’s covariate values, and the sparse samples to compute a posterior estimate of that individual’s drug exposure, effectively turning a generic dosing table into a personalized prediction engine.
This work does not just refine a laboratory method; it directly enables a model-informed precision dosing pipeline that could operate at the point of care. Imagine a patient with chronic myeloid leukemia arriving for their routine follow-up, providing a capillary blood sample via a painless finger prick at the beginning and end of their consultation, and leaving the clinic minutes later with a mathematically optimized dose adjustment already communicated to their pharmacy. No venipuncture, no cold-chain shipping, no days-long wait for results. By reducing the sampling burden to the absolute minimum and proving that the resulting estimates are safe and reliable, the study clears the path for integrating therapeutic drug monitoring into the compact, microfluidic testing platforms currently being developed for decentralized settings.
Beyond imatinib, the methodological blueprint carries implications for the broader class of oral anticancer agents plagued by similar pharmacokinetic unpredictability, including other tyrosine kinase inhibitors like nilotinib and dasatinib. The marriage of population modeling, Bayesian statistics, and sparse sampling optimization creates a template that could be adapted to virtually any chronically administered drug where exposure correlates with outcome. As the oncology community inches toward truly personalized care, the ability to squeeze maximal diagnostic information from minimal biological material will be a defining capability. The study by Pétermann and colleagues does not claim to have solved all the regulatory and implementation hurdles, but it provides the pharmacokinetic evidence base that regulators and device manufacturers need to move forward. The next logical step—a prospective clinical trial demonstrating that point-of-care therapeutic drug monitoring with these sampling strategies actually improves patient outcomes—now has a solid quantitative foundation on which to build.
Subject of Research: Optimization and validation of limited sampling strategies for imatinib therapeutic drug monitoring using a population pharmacokinetic model and Bayesian estimation to enable point-of-care model-informed precision dosing.
Article Title: Optimizing imatinib sampling strategies through a population approach: a crucial step to model-informed precision dosing validation for point-of-care therapeutic drug monitoring
Article References:
Pétermann, Y.J., Chen, J., Alberi, L. et al. Optimizing imatinib sampling strategies through a population approach: a crucial step to model-informed precision dosing validation for point-of-care therapeutic drug monitoring.
BMC Pharmacol Toxicol (2026). https://doi.org/10.1186/s40360-026-01181-5
Image Credits: AI Generated
DOI: 10.1186/s40360-026-01181-5
Keywords: imatinib, therapeutic drug monitoring, population pharmacokinetics, model-informed precision dosing, limited sampling strategy, Bayesian estimation, point-of-care testing, chronic myeloid leukemia, area under the curve, personalized medicine
