Bayesian Data Assimilation to Support Informed Decision Making in Individualized Chemotherapy

Abstract

An essential component of therapeutic drug/biomarker monitoring (TDM) is to combine patient data with prior knowledge for model‐based predictions of therapy outcomes. Current Bayesian forecasting tools typically rely only on the most probable model parameters (maximum a posteriori (MAP) estimate). This MAP‐based approach, however, does neither necessarily predict the most probable outcome nor does it quantify the risks of treatment inefficacy or toxicity. Bayesian data assimilation (DA) methods overcome these limitations by providing a comprehensive uncertainty quantification. We compare DA methods with MAP‐based approaches and show how probabilistic statements about key markers related to chemotherapy‐induced neutropenia can be leveraged for more informative decision support in individualized chemotherapy. Sequential Bayesian DA proved to be most computationally efficient for handling interoccasion variability and integrating TDM data. For new digital monitoring devices enabling more frequent data collection, these features will be of critical importance to improve patient care decisions in various therapeutic areas.

Optimizing Predictive Performance of Bayesian Forecasting for Vancomycin Concentration in Intensive Care Patients

“…MAP estimation does not address the uncertainty of the posterior distribution and hence is not able to quantify the potential risk of the relevant clinical outcome. Recently, MAP-based approaches have been questioned in certain circumstances, such as when trough concentration is of interest because even a subtle discrepancy could be damaging for concentrations at a low level (18). Although a full Bayesian approach is superior to MAP-based approaches in handling mentioned issues, the benefit from such an approach could be limited or perhaps clinically irrelevant, due to the relatively large noise in the routine TDM data from ICU patients which influences the results as well.…”

Reinforcement learning and Bayesian data assimilation for model‐informed precision dosing in oncology

“…MIPD builds on prior knowledge from NLME analyses of clinical studies 21 . The structural and observational models are generally given as:with state vector (e.g., neutrophil concentration), parameter values (e.g., mean transition time), and rates of change for given doses .…”

“…For neutrophil‐guided dosing, a reward function was suggested that maps (MAP‐based) nadir concentrations to a continuous score 19 or penalizes the deviation from a target nadir concentration () 17 ; in this study, we used a utility function but also provide a comparison of the results with the suggested target concentration, see also Section S8.5 in Appendix and Figure . The individualized uncertainties quantified via DA allow to consider the probability of being within/outside the target range in the reward function, 21 which is more closely related to clinical reality. For the patient state of Equation used in RL, we also designed the reward function to account for efficacy and toxicity.…”

“…The definition of a target concentration or utility function is, however, difficult because in many therapies rather subtherapeutic or toxic ranges are known. For therapeutic ranges, MAP‐guided dosing is not readily suited, 20 because only a (potentially biased) point estimate is used, neglecting associated uncertainties 21 . A post hoc uncertainty quantification for MAP‐based predictions often relies on a normal approximation located at the MAP estimate, which was previously shown to not necessarily transform accurately into quantities of interest for nonlinear models (e.g., to the nadir concentration 21 ).…”

Perspectives on Model‐Informed Precision Dosing in the Digital Health Era: Challenges, Opportunities, and Recommendations

“…Within a Bayesian framework, typically prior knowledge about drug pharmacokinetics (PK) and exposure‐response relationships are individualized based on individual patient characteristics (“covariates,” e.g., age, weight, sex, disease characteristics, or comedication) and PK or biomarker data to obtain individual model parameters (maximum a‐ posteriori estimates). Recently, Bayesian data assimilation methods have come into focus, overcoming major limitations of maximum a posteriori ‐based approaches by enabling accurate uncertainty quantification and propagation 2 . In contrast to traditional and well‐established therapeutic drug/biomarker monitoring (TDM), MIPD provides quantitative decision support to healthcare professionals for real‐world patient populations integrating multi‐level data.…”