Population pharmacokinetic model selection assisted by machine learning

Simulation-based evaluation of the Pharmpy Automatic Model Development tool for population pharmacokinetic modeling in early clinical drug development

The Pharmpy Automatic Model Development (AMD) tool automates the building of population pharmacokinetic (popPK) models by utilizing a systematic stepwise process. In this study, the performance of the AMD tool was assessed using simulated datasets. Ten true models mimicking classical popPK models were created. From each true model, dataset replicates were simulated assuming a typical phase I study design-single and multiple ascending doses with/without dichotomous food effect, with rich PK sampling. For every dataset replicate, the AMD tool automatically built an AMD model utilizing NONMEM for parameter estimation. The AMD models were compared to the true and reference models (true model fitted to simulated datasets) based on their model components, predicted population and individual secondary PK parameters (SP) (AUC0-24, cmax, ctrough), and model quality metrics (e.g., model convergence, parameter relative standard errors (RSEs), Bayesian Information Criterion (BIC)). The models selected by the AMD tool closely resembled the true models, particularly in terms of distribution and elimination, although differences were observed in absorption and inter-individual variability components. Bias associated with the derived SP was low. In general, discrepancies between AMD and true SP were also observed for reference models and therefore were attributed to the inherent stochasticity in simulations. In summary, the AMD tool was found to be a valuable asset in automating repetitive modeling tasks, yielding reliable PK models in the scenarios assessed. This tool has the potential to save time during early clinical drug development that can be invested in more complex modeling activities within model-informed drug development.

Dose Individualisation of Antimicrobials from a Pharmacometric Standpoint: The Current Landscape

Successful antimicrobial therapy depends on achieving optimal drug concentrations within individual patients. Inter-patient variability in pharmacokinetics (PK) and differences in pathogen susceptibility (reflected in the minimum inhibitory concentration, [MIC]) necessitate personalised approaches. Dose individualisation strategies aim to address this challenge, improving treatment outcomes and minimising the risk of toxicity and antimicrobial resistance. Therapeutic drug monitoring (TDM), with the application of population pharmacokinetic (popPK) models, enables model-informed precision dosing (MIPD). PopPK models mathematically describe drug behaviour across populations and can be combined with patient-specific TDM data to optimise dosing regimens. The integration of machine learning (ML) techniques promises to further enhance dose individualisation by identifying complex patterns within extensive datasets. Implementing these approaches involves challenges, including rigorous model selection and validation to ensure suitability for target populations. Understanding the relationship between drug exposure and clinical outcomes is crucial, as is striking a balance between model complexity and clinical usability. Additionally, regulatory compliance, outcome measurement, and practical considerations for software implementation will be addressed. Emerging technologies, such as real-time biosensors, hold the potential for revolutionising TDM by enabling continuous monitoring, immediate and frequent dose adjustments, and near patient testing. The ongoing integration of TDM, advanced modelling techniques, and ML within the evolving digital health care landscape offers a potential for enhancing antimicrobial therapy. Careful attention to model development, validation, and ethical considerations of the applied techniques is paramount for successfully optimising antimicrobial treatment for the individual patient.

Selecting the Best Pharmacokinetic Models for a Priori Model-Informed Precision Dosing with Model Ensembling

BACKGROUND AND OBJECTIVE

When utilizing population pharmacokinetic (popPK) models for a priori dosage individualization, selecting the best model is crucial to obtain adequate doses. We developed and evaluated several model-selection and ensembling methods, using external evaluation on the basis of therapeutic drug monitoring (TDM) samples to identify the best (set of) models per patient for a priori dosage individualization.

METHODS

PK data and models describing both hospitalized patients (n = 134) receiving continuous vancomycin (26 models) and patients (n = 92) receiving imatinib in an outpatient setting (12 models) are included. Target attainment of four model-selection methods was compared with standard dosing: the best model based on external validation, uninformed model ensembling, model ensembling using a weighting scheme on the basis of covariate-stratified external evaluation, and model selection using covariates in decision trees that were subsequently ensembled.

RESULTS

Overall, the use of PK models improved the proportion of patients exposed to concentrations within the therapeutic window for both cohorts. Relative improvement of proportion on target for best model, unweighted, weighted, and decision trees were - 7.0%, 2.3%, 11.4%, and 37.0% (vancomycin method-development); 23.2%, 7.9%, 15.6%, and, 77.2% (vancomycin validation); 40.7%, 50.0%, 59.5%, and 59.5% (imatinib method-development); and 19.0%, 28.5%, 38.0%, and 23.8% (imatinib validation), respectively.

CONCLUSIONS

The best (set of) models per patient for a priori dosage individualization can be identified using a relatively small set of TDM samples as external evaluation. Adequately performing popPK models were identified while also excluding poor-performing models. Dose recommendations resulted in more patients within the therapeutic range for both vancomycin and imatinib. Prospective validation is necessary before clinical implementation.

Explainable machine learning prediction of edema adverse events in patients treated with tepotinib

Tepotinib is approved for the treatment of patients with non-small-cell lung cancer harboring MET exon 14 skipping alterations. While edema is the most prevalent adverse event (AE) and a known class effect of MET inhibitors including tepotinib, there is still limited understanding about the factors contributing to its occurrence. Herein, we apply machine learning (ML)-based approaches to predict the likelihood of occurrence of edema in patients undergoing tepotinib treatment, and to identify factors influencing its development over time. Data from 612 patients receiving tepotinib in five Phase I/II studies were modeled with two ML algorithms, Random Forest, and Gradient Boosting Trees, to predict edema AE incidence and severity. Probability calibration was applied to give a realistic estimation of the likelihood of edema AE. Best model was tested on follow-up data and on data from clinical studies unused while training. Results showed high performances across all the tested settings, with F1 scores up to 0.961 when retraining the model with the most relevant covariates. The use of ML explainability methods identified serum albumin as the most informative longitudinal covariate, and higher age as associated with higher probabilities of more severe edema. The developed methodological framework enables the use of ML algorithms for analyzing clinical safety data and exploiting longitudinal information through various covariate engineering approaches. Probability calibration ensures the accurate estimation of the likelihood of the AE occurrence, while explainability tools can identify factors contributing to model predictions, hence supporting population and individual patient-level interpretation.

Utility of life stage-specific chemical risk assessments based on New Approach Methodologies (NAMs)

The safety assessments for chemicals targeted for use or expected to be exposed to specific life stages, including infancy, childhood, pregnancy and lactation, and geriatrics, need to account for extrapolation of data from healthy adults to these populations to assess their human health risk. However, often adequate and relevant toxicity or pharmacokinetic (PK) data of chemicals in specific life stages are not available. For such chemicals, New Approach Methodologies (NAMs), such as physiologically based pharmacokinetic (PBPK) modeling, biologically based dose response (BBDR) modeling, in vitro to in vivo extrapolation (IVIVE), etc. can be used to understand the variability of exposure and effects of chemicals in specific life stages and assess their associated risk. A life stage specific PBPK model incorporates the physiological and biochemical changes associated with each life stage and simulates their impact on the absorption, distribution, metabolism, and elimination (ADME) of these chemicals. In our review, we summarize the parameterization of life stage models based on New Approach Methodologies (NAMs) and discuss case studies that highlight the utility of a life stage based PBPK modeling for risk assessment. In addition, we discuss the utility of artificial intelligence (AI)/machine learning (ML) and other computational models, such as those based on in vitro data, as tools for estimation of relevant physiological or physicochemical parameters and selection of model. We also discuss existing gaps in the available toxicological datasets and current challenges that need to be overcome to expand the utility of NAMs for life stage-specific chemical risk assessment.

Machine learning guided prediction of warfarin blood levels for personalized medicine based on clinical longitudinal data from cardiac surgery patients: a prospective observational study

BACKGROUND

Warfarin is a common oral anticoagulant, and its effects vary widely among individuals. Numerous dose-prediction algorithms have been reported based on cross-sectional data generated via multiple linear regression or machine learning. This study aimed to construct an information fusion perturbation theory and machine learning prediction model of warfarin blood levels based on clinical longitudinal data from cardiac surgery patients.

METHODS AND MATERIAL

The data of 246 patients were obtained from electronic medical records. Continuous variables were processed by calculating the distance of the raw data with the moving average (MA ∆vki(sj)), and categorical variables in different attribute groups were processed using Euclidean distance (ED ǁ∆vk(sj)ǁ). Regression and classification analyses were performed on the raw data, MA ∆vki(sj), and ED ǁ∆vk(sj)ǁ. Different machine-learning algorithms were chosen for the STATISTICA and WEKA software.

RESULTS

The random forest (RF) algorithm was the best for predicting continuous outputs using the raw data. The correlation coefficients of the RF algorithm were 0.978 and 0.595 for the training and validation sets, respectively, and the mean absolute errors were 0.135 and 0.362 for the training and validation sets, respectively. The proportion of ideal predictions of the RF algorithm was 59.0%. General discriminant analysis (GDA) was the best algorithm for predicting the categorical outputs using the MA ∆vki(sj) data. The GDA algorithm's total true positive rate (TPR) was 95.4% and 95.6% for the training and validation sets, respectively, with MA ∆vki(sj) data.

CONCLUSIONS

An information fusion perturbation theory and machine learning model for predicting warfarin blood levels was established. A model based on the RF algorithm could be used to predict the target international normalized ratio (INR), and a model based on the GDA algorithm could be used to predict the probability of being within the target INR range under different clinical scenarios.

Machine-Learning Assisted Screening of Correlated Covariates: Application to Clinical Data of Desipramine

Stepwise covariate modeling (SCM) has a high computational burden and can select the wrong covariates. Machine learning (ML) has been proposed as a screening tool to improve the efficiency of covariate selection, but little is known about how to apply ML on actual clinical data. First, we simulated datasets based on clinical data to compare the performance of various ML and traditional pharmacometrics (PMX) techniques with and without accounting for highly-correlated covariates. This simulation step identified the ML algorithm and the number of top covariates to select when using the actual clinical data. A previously developed desipramine population-pharmacokinetic model was used to simulate virtual subjects. Fifteen covariates were considered with four having an effect included. Based on the F1 score (an accuracy measure), ridge regression was the most accurate ML technique on 200 simulated datasets (F1 score = 0.475 ± 0.231), a performance which almost doubled when highly-correlated covariates were accounted for (F1 score = 0.860 ± 0.158). These performances were better than forwards selection with SCM (F1 score = 0.251 ± 0.274 and 0.499 ± 0.381 without/with correlations respectively). In terms of computational cost, ridge regression (0.42 ± 0.07 seconds/simulated dataset, 1 thread) was ~20,000 times faster than SCM (2.30 ± 2.29 hours, 15 threads). On the clinical dataset, prescreening with the selected ML algorithm reduced SCM runtime by 42.86% (from 1.75 to 1.00 days) and produced the same final model as SCM only. In conclusion, we have demonstrated that accounting for highly-correlated covariates improves ML prescreening accuracy. The choice of ML method and the proportion of important covariates (unknown a priori) can be guided by simulations.

Improving prediction of tacrolimus concentration using a combination of population pharmacokinetic modeling and machine learning in chinese renal transplant recipients

Aims

The population pharmacokinetic (PPK) model-based machine learning (ML) approach offers a novel perspective on individual concentration prediction. This study aimed to establish a PPK-based ML model for predicting tacrolimus (TAC) concentrations in Chinese renal transplant recipients.

Methods

Conventional TAC monitoring data from 127 Chinese renal transplant patients were divided into training (80%) and testing (20%) datasets. A PPK model was developed using the training group data. ML models were then established based on individual pharmacokinetic data derived from the PPK basic model. The prediction performances of the PPK-based ML model and Bayesian forecasting approach were compared using data from the test group.

Results

The final PPK model, incorporating hematocrit and CYP3A5 genotypes as covariates, was successfully established. Individual predictions of TAC using the PPK basic model, postoperative date, CYP3A5 genotype, and hematocrit showed improved rankings in ML model construction. XGBoost, based on the TAC PPK, exhibited the best prediction performance.

Conclusion

The PPK-based machine learning approach emerges as a superior option for predicting TAC concentrations in Chinese renal transplant recipients.

Learning pharmacometric covariate model structures with symbolic regression networks

Efficiently finding covariate model structures that minimize the need for random effects to describe pharmacological data is challenging. The standard approach focuses on identification of relevant covariates, and present methodology lacks tools for automatic identification of covariate model structures. Although neural networks could potentially be used to approximate covariate-parameter relationships, such approximations are not human-readable and come at the risk of poor generalizability due to high model complexity.In the present study, a novel methodology for the simultaneous selection of covariate model structure and optimization of its parameters is proposed. It is based on symbolic regression, posed as an optimization problem with a smooth loss function. This enables training of the model through back-propagation using efficient gradient computations.Feasibility and effectiveness are demonstrated by application to a clinical pharmacokinetic data set for propofol, containing infusion and blood sample time series from 1031 individuals. The resulting model is compared to a published state-of-the-art model for the same data set. Our methodology finds a covariate model structure and corresponding parameter values with a slightly better fit, while relying on notably fewer covariates than the state-of-the-art model. Unlike contemporary practice, finding the covariate model structure is achieved without an iterative procedure involving manual interactions.

Artificial Intelligence for Quantitative Modeling in Drug Discovery and Development: An Innovation and Quality Consortium Perspective on Use Cases and Best Practices

Recent breakthroughs in artificial intelligence (AI) and machine learning (ML) have ushered in a new era of possibilities across various scientific domains. One area where these advancements hold significant promise is model-informed drug discovery and development (MID3). To foster a wider adoption and acceptance of these advanced algorithms, the Innovation and Quality (IQ) Consortium initiated the AI/ML working group in 2021 with the aim of promoting their acceptance among the broader scientific community as well as by regulatory agencies. By drawing insights from workshops organized by the working group and attended by key stakeholders across the biopharma industry, academia, and regulatory agencies, this white paper provides a perspective from the IQ Consortium. The range of applications covered in this white paper encompass the following thematic topics: (i) AI/ML-enabled Analytics for Pharmacometrics and Quantitative Systems Pharmacology (QSP) Workflows; (ii) Explainable Artificial Intelligence and its Applications in Disease Progression Modeling; (iii) Natural Language Processing (NLP) in Quantitative Pharmacology Modeling; and (iv) AI/ML Utilization in Drug Discovery. Additionally, the paper offers a set of best practices to ensure an effective and responsible use of AI, including considering the context of use, explainability and generalizability of models, and having human-in-the-loop. We believe that embracing the transformative power of AI in quantitative modeling while adopting a set of good practices can unlock new opportunities for innovation, increase efficiency, and ultimately bring benefits to patients.

Go beyond the limits of genetic algorithm in daily covariate selection practice

Covariate identification is an important step in the development of a population pharmacokinetic/pharmacodynamic model. Among the different available approaches, the stepwise covariate model (SCM) is the most used. However, SCM is based on a local search strategy, in which the model-building process iteratively tests the addition or elimination of a single covariate at a time given all the others. This introduces a heuristic to limit the searching space and then the computational complexity, but, at the same time, can lead to a suboptimal solution. The application of genetic algorithms (GAs) for covariate selection has been proposed as a possible solution to overcome these limitations. However, their actual use during model building is limited by the extremely high computational costs and convergence issues, both related to the number of models being tested. In this paper, we proposed a new GA for covariate selection to address these challenges. The GA was first developed on a simulated case study where the heuristics introduced to overcome the limitations affecting currently available GA approaches resulted able to limit the selection of redundant covariates, increase replicability of results and reduce convergence times. Then, we tested the proposed GA on a real-world problem related to remifentanil. It obtained good results both in terms of selected covariates and fitness optimization, outperforming the SCM.

Advancing Precision Medicine: A Review of Innovative In Silico Approaches for Drug Development, Clinical Pharmacology and Personalized Healthcare

The landscape of medical treatments is undergoing a transformative shift. Precision medicine has ushered in a revolutionary era in healthcare by individualizing diagnostics and treatments according to each patient's uniquely evolving health status. This groundbreaking method of tailoring disease prevention and treatment considers individual variations in genes, environments, and lifestyles. The goal of precision medicine is to target the "five rights": the right patient, the right drug, the right time, the right dose, and the right route. In this pursuit, in silico techniques have emerged as an anchor, driving precision medicine forward and making this a realistic and promising avenue for personalized therapies. With the advancements in high-throughput DNA sequencing technologies, genomic data, including genetic variants and their interactions with each other and the environment, can be incorporated into clinical decision-making. Pharmacometrics, gathering pharmacokinetic (PK) and pharmacodynamic (PD) data, and mathematical models further contribute to drug optimization, drug behavior prediction, and drug-drug interaction identification. Digital health, wearables, and computational tools offer continuous monitoring and real-time data collection, enabling treatment adjustments. Furthermore, the incorporation of extensive datasets in computational tools, such as electronic health records (EHRs) and omics data, is also another pathway to acquire meaningful information in this field. Although they are fairly new, machine learning (ML) algorithms and artificial intelligence (AI) techniques are also resources researchers use to analyze big data and develop predictive models. This review explores the interplay of these multiple in silico approaches in advancing precision medicine and fostering individual healthcare. Despite intrinsic challenges, such as ethical considerations, data protection, and the need for more comprehensive research, this marks a new era of patient-centered healthcare. Innovative in silico techniques hold the potential to reshape the future of medicine for generations to come.

Predicting quetiapine dose in patients with depression using machine learning techniques based on real-world evidence

BACKGROUND

Being one of the most widespread, pervasive, and troublesome illnesses in the world, depression causes dysfunction in various spheres of individual and social life. Regrettably, despite obtaining evidence-based antidepressant medication, up to 70% of people are going to continue to experience troublesome symptoms. Quetiapine, as one of the most commonly prescribed antipsychotic medication worldwide, has been reported as an effective augmentation strategy to antidepressants. The right quetiapine dose and personalized quetiapine treatment are frequently challenging for clinicians. This study aimed to identify important influencing variables for quetiapine dose by maximizing the use of data from real world, and develop a predictive model of quetiapine dose through machine learning techniques to support selections for treatment regimens.

METHODS

The study comprised 308 depressed patients who were medicated with quetiapine and hospitalized in the First Hospital of Hebei Medical University, from November 1, 2019, to August 31, 2022. To identify the important variables influencing the dose of quetiapine, a univariate analysis was applied. The prediction abilities of nine machine learning models (XGBoost, LightGBM, RF, GBDT, SVM, LR, ANN, DT) were compared. Algorithm with the optimal model performance was chosen to develop the prediction model.

RESULTS

Four predictors were selected from 38 variables by the univariate analysis (p < 0.05), including quetiapine TDM value, age, mean corpuscular hemoglobin concentration, and total bile acid. Ultimately, the XGBoost algorithm was used to create a prediction model for quetiapine dose that had the greatest predictive performance (accuracy = 0.69) out of nine models. In the testing cohort (62 cases), a total of 43 cases were correctly predicted of the quetiapine dose regimen. In dose subgroup analysis, AUROC for patients with daily dose of 100 mg, 200 mg, 300 mg and 400 mg were 0.99, 0.75, 0.93 and 0.86, respectively.

CONCLUSIONS

In this work, machine learning techniques are used for the first time to estimate the dose of quetiapine for patients with depression, which is valuable for the clinical drug recommendations.

Tumor growth inhibition modeling in patients with second line biliary tract cancer and first line non-small cell lung cancer based on bintrafusp alfa trials

This analysis aimed to quantify tumor dynamics in patients receiving either bintrafusp alfa (BA) or pembrolizumab, by population pharmacokinetic (PK)-pharmacodynamic modeling, and investigate clinical and molecular covariates describing the variability in tumor dynamics by pharmacometric and machine-learning (ML) approaches. Data originated from two clinical trials in patients with biliary tract cancer (BTC; NCT03833661) receiving BA and non-small cell lung cancer (NSCLC; NCT03631706) receiving BA or pembrolizumab. Individual drug exposure was estimated from previously developed population PK models. Population tumor dynamics models were developed for each drug-indication combination, and covariate evaluations performed using nonlinear mixed-effects modeling (NLME) and ML (elastic net and random forest models) approaches. The three tumor dynamics' model structures all included linear tumor growth components and exponential tumor shrinkage. The final BTC model included the effect of drug exposure (area under the curve) and several covariates (demographics, disease-related, and genetic mutations). Drug exposure was not significant in either of the NSCLC models, which included two, disease-related, covariates in the BA arm, and none in the pembrolizumab arm. The covariates identified by univariable NLME and ML highly overlapped in BTC but showed less agreement in NSCLC analyses. Hyperprogression could be identified by higher tumor growth and lower tumor kill rates and could not be related to BA exposure. Tumor size over time was quantitatively characterized in two tumor types and under two treatments. Factors potentially related to tumor dynamics were assessed using NLME and ML approaches; however, their net impact on tumor size was considered as not clinically relevant.

Bridging the Worlds of Pharmacometrics and Machine Learning

Precision medicine requires individualized modeling of disease and drug dynamics, with machine learning-based computational techniques gaining increasing popularity. The complexity of either field, however, makes current pharmacological problems opaque to machine learning practitioners, and state-of-the-art machine learning methods inaccessible to pharmacometricians. To help bridge the two worlds, we provide an introduction to current problems and techniques in pharmacometrics that ranges from pharmacokinetic and pharmacodynamic modeling to pharmacometric simulations, model-informed precision dosing, and systems pharmacology, and review some of the machine learning approaches to address them. We hope this would facilitate collaboration between experts, with complementary strengths of principled pharmacometric modeling and flexibility of machine learning leading to synergistic effects in pharmacological applications.

Population Pharmacokinetic Modeling Combined With Machine Learning Approach Improved Tacrolimus Trough Concentration Prediction in Chinese Adult Liver Transplant Recipients

This study aimed to develop and evaluate a population pharmacokinetic (PPK) combined machine learning approach to predict tacrolimus trough concentrations for Chinese adult liver transplant recipients in the early posttransplant period. Tacrolimus trough concentrations were retrospectively collected from routine monitoring records of liver transplant recipients and divided into the training data set (1287 concentrations in 145 recipients) and the test data set (296 concentrations in 36 recipients). A PPK model was first established using NONMEM. Then a machine learning model of Xgboost was adapted to fit the estimated individual pharmacokinetic parameters obtained from the PPK model with Bayesian forecasting. The performance of the final PPK model and Xgboost model was compared in the test data set. In the final PPK model, tacrolimus daily dose, postoperative days, hematocrit, aspartate aminotransferase, and concomitant voriconazole, were identified to significantly influence the clearance. The postoperative days along with hematocrit significantly influence the volume of distribution. In the Xgboost model, the first 5 predictors for predicting the clearance were concomitant with voriconazole, sex, single nucleotide polymorphisms of CYP3A4*1G and CYP3A5*3 in recipients, and tacrolimus daily dose, for the volume of distribution were postoperative days, age, weight, total bilirubin and graft : recipient weight ratio. In the test data set, the Xgboost model showed the minimum median prediction error of tacrolimus concentrations, less than the PPK model with or without Bayesian forecasting. In conclusion, a PPK combined machine learning approach could improve the prediction of tacrolimus concentrations for Chinese adult liver transplant recipients in the early posttransplant period.

Machine Learning for Pharmacokinetic/Pharmacodynamic Modeling

A variety of new recurrent neural networks (RNNs) including the ODE-LSTM, Phased LSTM, CTGRU and GRU-D, were evaluated on modeling irregularly sampled PK/PD data with 6 or 12 time points/day and predicting PD data of unseen dosing regimens and dosing levels. The one-compartment absorption PK model and the Indirect PK/PD model I was used to simulate the PK/PD with inter-individual variabilities in volume of distribution and residual errors in PD measurement. The four RNNs were able to successfully model daily dose (QD) PK/PD and extrapolate to twice daily (BID) dose PD based on BID PK. The RNNs not only captured the additional fluctuations in the BID regimen but also the return phase to the baseline PD. However, extrapolating to unseen dose levels outside of the dose range for training proved to be challenging for all the RNNs tested. Only the GRUD demonstrated reasonable prediction results when extrapolating to unseen doses that were 3 or 10-fold outside the training doses. Overall, these new RNNs were able to overcome some limitations of previous RNNs evaluated and showed promise of integrating neural networks in PK/PD.

An interpretable stacking ensemble learning framework based on multi-dimensional data for real-time prediction of drug concentration: The example of olanzapine

Background and Aim: Therapeutic drug monitoring (TDM) has evolved over the years as an important tool for personalized medicine. Nevertheless, some limitations are associated with traditional TDM. Emerging data-driven model forecasting [e.g., through machine learning (ML)-based approaches] has been used for individualized therapy. This study proposes an interpretable stacking-based ML framework to predict concentrations in real time after olanzapine (OLZ) treatment.

Methods: The TDM-OLZ dataset, consisting of 2,142 OLZ measurements and 472 features, was formed by collecting electronic health records during the TDM of 927 patients who had received OLZ treatment. We compared the performance of ML algorithms by using 10-fold cross-validation and the mean absolute error (MAE). The optimal subset of features was analyzed by a random forest-based sequential forward feature selection method in the context of the top five heterogeneous regressors as base models to develop a stacked ensemble regressor, which was then optimized via the grid search method. Its predictions were explained by using local interpretable model-agnostic explanations (LIME) and partial dependence plots (PDPs).

Results: A state-of-the-art stacking ensemble learning framework that integrates optimized extra trees, XGBoost, random forest, bagging, and gradient-boosting regressors was developed for nine selected features [i.e., daily dose (OLZ), gender_male, age, valproic acid_yes, ALT, K, BW, MONO#, and time of blood sampling after first administration]. It outperformed other base regressors that were considered, with an MAE of 0.064, R-square value of 0.5355, mean squared error of 0.0089, mean relative error of 13%, and ideal rate (the percentages of predicted TDM within ± 30% of actual TDM) of 63.40%. Predictions at the individual level were illustrated by LIME plots, whereas the global interpretation of associations between features and outcomes was illustrated by PDPs.

Conclusion: This study highlights the feasibility of the real-time estimation of drug concentrations by using stacking-based ML strategies without losing interpretability, thus facilitating model-informed precision dosing.

Adoption of Machine Learning in Pharmacometrics: An Overview of Recent Implementations and Their Considerations

Pharmacometrics is a multidisciplinary field utilizing mathematical models of physiology, pharmacology, and disease to describe and quantify the interactions between medication and patient. As these models become more and more advanced, the need for advanced data analysis tools grows. Recently, there has been much interest in the adoption of machine learning (ML) algorithms. These algorithms offer strong function approximation capabilities and might reduce the time spent on model development. However, ML tools are not yet an integral part of the pharmacometrics workflow. The goal of this work is to discuss how ML algorithms have been applied in four stages of the pharmacometrics pipeline: data preparation, hypothesis generation, predictive modelling, and model validation. We will also discuss considerations before the use of ML algorithms with respect to each topic. We conclude by summarizing applications that hold potential for adoption by pharmacometricians.

A Hybrid Model Associating Population Pharmacokinetics with Machine Learning: A Case Study with Iohexol Clearance Estimation

BACKGROUND

Maximum a posteriori Bayesian estimation (MAP-BE) based on a limited sampling strategy and a population pharmacokinetic model is frequently used to estimate pharmacokinetic parameters in individuals, however with some uncertainty (bias). Recent works have shown that the performance in individual estimation or pharmacokinetic parameters can be improved by combining population pharmacokinetic and machine learning algorithms.

OBJECTIVE

 The objective of this work was to investigate the use of a hybrid machine learning/population pharmacokinetic approach to improve individual iohexol clearance estimation.

METHODS

The reference iohexol clearance values were derived from 500 simulated profiles (samples collected between 0.1 and 24.7 h) using a population pharmacokinetic model we recently developed in Monolix and obtained using all the concentration timepoints available. Xgboost and glmnet algorithms able to predict the error of MAP-BE clearance estimates based on a limited sampling strategy (0.1 h, 1 h, and 9 h) versus reference values were developed in a training subset (75%) and were evaluated in a testing subset (25%) and in 36 real patients.

RESULTS

The MAP-BE limited sampling strategy estimated clearance was corrected by the machine learning predicted error leading to a decrease in root mean squared error by 29% and 24%, and in the percentage of profiles with the mean prediction error out of the ± 20% bias by 60% and 40% in the external validation dataset for the glmnet and Xgboost machine learning algorithms, respectively. These results were attributable to a decrease in the eta-shrinkage (shrinkage for a MAP-BE limited sampling strategy = 32.4%, glmnet = 18.2%, and Xgboost = 19.4% in the external dataset).

CONCLUSIONS

 In conclusion, this hybrid algorithm represents a significant improvement in comparison to MAP-BE estimation alone.

Construction and Interpretation of Prediction Model of Teicoplanin Trough Concentration via Machine Learning

Objective

To establish an optimal model to predict the teicoplanin trough concentrations by machine learning, and explain the feature importance in the prediction model using the SHapley Additive exPlanation (SHAP) method.

Methods

A retrospective study was performed on 279 therapeutic drug monitoring (TDM) measurements obtained from 192 patients who were treated with teicoplanin intravenously at the First Affiliated Hospital of Army Medical University from November 2017 to July 2021. This study included 27 variables, and the teicoplanin trough concentrations were considered as the target variable. The whole dataset was divided into a training group and testing group at the ratio of 8:2, and predictive performance was compared among six different algorithms. Algorithms with higher model performance (top 3) were selected to establish the ensemble prediction model and SHAP was employed to interpret the model.

Results

Three algorithms (SVR, GBRT, and RF) with high R² scores (0.676, 0.670, and 0.656, respectively) were selected to construct the ensemble model at the ratio of 6:3:1. The model with R² = 0.720, MAE = 3.628, MSE = 22.571, absolute accuracy of 83.93%, and relative accuracy of 60.71% was obtained, which performed better in model fitting and had better prediction accuracy than any single algorithm. The feature importance and direction of each variable were visually demonstrated by SHAP values, in which teicoplanin administration and renal function were the most important factors.

Conclusion

We firstly adopted a machine learning approach to predict the teicoplanin trough concentration, and interpreted the prediction model by the SHAP method, which is of great significance and value for the clinical medication guidance.