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

Abstract

Drug approval is based on exposure, response, and variability of studied populations, typically excluding comorbidities/medications and very ill patients, thus not representing real-world populations. This results in wide variability in therapeutic outcome for individual patients. Model-informed precision dosing (MIPD) can characterize/quantify this variability, support optimal dose selection, and enable individualized therapy. The aim of this perspective is to raise awareness for MIPD, identify challenges hindering its implementation in clinical practice, provide recommendations, and highlight opportunities. MIPD aims at tailoring doses to patients’ needs, and therefore presents a promising tool to increase treatment success.1 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. With the increase in available computing power,3 further methodological advances enabling a comprehensive uncertainty quantification,2 numerous publications demonstrating the clinical benefits of MIPD,4 and also the user-friendliness of few already existing MIPD tools,5 the question arises as to why the implementation of MIPD in clinical practice—with the exception of local initiatives at academic hospital centers6—still largely fails. In the following, we summarize selected key challenges that need to be addressed, and further perspectives beyond (see also Figure 1, Table 1). We propose an alignment for terminology and across scientific disciplines, thus enabling collaborative work. Furthermore, we provide a comprehensive literature overview of current applications, review articles, innovative methodology, initiatives, and already available software tools (Table S1). MIPD is a rapidly evolving research area in which multiple scientific disciplines/communities and other stakeholders meet. Due to the diverse origins of MIPD research projects or MIPD tools, various terms exist in the literature for these approaches and methodologies (Table 1). As a result, terms and labels are used interchangeably, and although there are many common features, various MIPD initiatives exist in parallel without touch points. Harmonization of the definitions across different therapeutic areas and scientific disciplines is therefore crucial.7 Following the successful example of “model-informed drug discovery and development” a joint effort and unified appearance under the consensus term “MIPD” will ensure greater visibility and extension of the target audience to facilitate and accelerate implementation of MIPD in drug development and clinical practice. Complementary to this, MIPD must be recognized as a collaborative effort between different scientific disciplines, which ought to be leveraged in research projects investigating and developing MIPD approaches/tools. Only through collaborative efforts, MIPD can be realized and will become a clinical reality. Moreover, the term “precision medicine” also comprises other approaches (e.g., pharmacogenomics), which are often seen as unconnected with MIPD, at times even competing concepts in scientific discourse and clinical practice, whereas they should rather be understood as complementary approaches to take full advantage of them. MIPD offers the potential to serve as a platform simultaneously integrating various approaches for therapy individualization and optimization.4 A major obstacle is still constituted by the fact that current curricula for physicians, clinical pharmacists, and other healthcare professionals are lacking in-depth training in quantitative pharmacology, which is needed to enable understanding, application, and evaluation of MIPD concepts and tools. This needs to be addressed via increasing awareness (e.g., by publishing good examples and “best practices” in the right journals,7 and presenting at conferences and workshops) and education (offer more training opportunities to acquire knowledge and expertise in model-informed/quantitative approaches). User-friendly software and integration into clinical workflow are ultimately needed to remove the remaining barriers and unleash the full power of MIPD tools. Healthcare professionals involved in drug treatment need to understand that “one-dose-fits-all” must be replaced by MIPD for the individual patient to improve therapy outcome. An MIPD tool will offer decision support, but the decision about the individual therapy still rests with the treating physician or clinical pharmacist, who integrates the overall status of the patients. Additionally, current regulatory (e.g., registration of MIPD tools as medical devices) and healthcare system (e.g., reimbursement framework) level barriers further discourage the application of MIPD in clinical routine. Joint efforts among the scientific community promoting MIPD, healthcare professionals, pharmaceutical companies, and stakeholders in the regulatory/healthcare system environment are required. MIPD provides opportunities for various approved and investigational drugs, however, not all drug therapies might benefit. Drugs with high associated treatment costs, potential severe adverse drug reactions or a narrow therapeutic index, and/or associated high interindividual variability, are particularly qualified for precision dosing. However, there must be a reliable correlation between the drug or surrogate biomarker concentration (preferably in a readily available matrix) and the clinical effect. In addition, robust therapeutic PK, pharmacodynamic, or PK/pharmacodynamic targets, utility functions, or target ranges must be established and thoroughly evaluated, which warrant large-scale clinical trials and observational studies in “real-world” populations, if not yet established during drug development. This can be particularly challenging for drugs with delayed (un-)desired effect (e.g., chemotherapy), for which drug tolerance is often developed (e.g., depression) or active metabolites. Next, if a drug is identified as a suitable candidate for MIPD, there is often more than one model or methodology applicable. Pooling of available data (or even models) for the same drug-disease system and continuous updating with new incoming data allows to capture most realistic patient population scenarios into a single model most appropriate for use in MIPD. In the future, as more frequent sampling (biosensors, wearables, point-of-care, or home-sampling devices) and advanced data-analysis methods (e.g., data assimilation or machine learning), become available for MIPD, better informed and automated precision dosing could be achieved. Nevertheless, also in the future using big data and machine learning, the quality of the available data and the mechanistic understanding of the underlying processes are crucial. Despite readily developed MIPD tools emerging from academic/clinical research collaborations, implementation into routine clinical care often fails. Translation of research findings into easy-to-use software tools is a crucial and challenging part of MIPD implementation, and expertise beyond quantitative pharmacology/pharmacometrics is needed. Implementation research is required: there are various implementation strategies (training, education, adaptability, flexibility, and changed clinician behavior) that need to be explored and evaluated to increase the uptake of MIPD into routine clinical practice. Following successful examples for MIPD tools5 (and also other healthcare applications, Table S1), the establishment of guidelines and “best practices” should be fostered. Furthermore, in collaboration with medical societies, guidelines how to implement and successfully realize MIPD for specific drug-disease systems should be established, as they are currently still lacking. Next to disease areas in which traditional TDM is well-established (e.g., infectious diseases, immunology, or transplantation medicine), new fields of application can be identified for MIPD. Besides scientific questions, there are many practical challenges that need to be addressed for implementation of MIPD in patients’ drug therapy. First, the clinical infrastructure needs to be adapted to be ready for MIPD use at the bedside, but also for providing optimal ambulant or home-based therapies using digital healthcare devices (e.g., wearable biosensors or point-of-care devices) that allow patients to measure and report online individual drug/biomarker concentrations. In the hospital, integration of available patient data might be difficult and time-consuming if they are still paper-based or multiple software tools are used for analysis, reporting, and communication of clinical samples. As electronic health record systems and point-of-care devices become widely available, integration and interfaces with MIPD tools should be explored and expanded to further support and accelerate this digital trend. With today’s technical progress, this should soon no longer pose a challenge at least for high-income countries with good healthcare systems. Implementation of model-informed and optimal design approaches can further contribute to overcome “classical” TDM problems, such as inappropriate timing, quality, and quantity of PK or biomarker samples. Often highlighted major concerns in traditional TDM still comprise long bioanalytical turnaround times of samples (several hours to weeks), lack of standardization in workflows, and high instrumentation costs with complex sample preparation.8 Raising awareness for benefits of optimal (often earlier and less) sampling timepoints, streamlining internal processes to shorten turnaround times, and introducing new concepts (e.g., point-of-care/bedside analytics, biosensors/wearables, and home-monitoring systems9), using not only plasma, but also, for example, saliva, interstitial fluid, or capillary blood, will offer practical solutions to the challenges listed above. Nowadays, pharmacometric approaches are widely used within drug research and development and model-informed approaches (MID3) have been well accepted by regulatory agencies. Therefore, further exploiting their potential in postapproval phases, particularly to investigate populations that have not been well-studied within clinical development (e.g., pregnant women and obese patients) should be encouraged. For future drug development, it should be acknowledged that accepting more complex dosing will trigger higher response rates and fewer adverse drug reactions, thus facilitating drug approval and reimbursement.10 Recently, regulatory agencies and stakeholders in the healthcare systems have increasingly acknowledged the value and opportunities of MIPD and are open for input and discourse. The regulatory and reimbursement frameworks will need to adapt to, but at the same time also trigger changing trial designs, methods of analysis, and more flexible and complex dosing recommendations. Of course, this does not apply for all investigational new drugs, but the right candidates should be identified early (e.g., based on projected low therapeutic indices or high treatment costs), and incorporation of precision dosing should become a key consideration for approval, reimbursement, or therapeutic use.10 First initiatives with regard to precision dosing were kicked-off by regulatory agencies11 and the potential of MIPD should also be evaluated for other existing initiatives, such as the development of so-called “companion diagnostics.” Global awareness of the emerging need for precision dosing instead of the historic “one-dose-fits-all-approach” for established but also newly approved and investigational drugs is rising. New mobile healthcare devices gathering data from various sources become available, and the overall complexity of treatment decision making increases. Keeping pace in the era of digital health is only possible through advances in the field of MIPD. This includes user-friendly, evaluated, and scalable decision-support tools integrated in the clinical workflow, improved training for physicians and clinical pharmacists, establishment of “MIPD good practices,” and eventually initiatives investigating the benefits of such MIPD tools over current practices. For the future of MIPD, we envision that MIPD will not only become an integral part of drug development, intrinsically motivated by initiatives within pharmaceutical companies, and as a requirement by regulatory agencies, but also that through education, training, and widespread availability, its application goes beyond academic hospital centers in high-income countries and becomes readily available for more patients in need for optimized and individualized therapy. Multistakeholder collaborations ranging from drug development to real-world and bedside application will be crucial to validate, implement, and demonstrate the value of MIPD.11 Open access funding enabled and organized by Projekt DEAL. No funding was received for this work. C.K. and W.H. report research grants from an industry consortium (AbbVie Deutschland GmbH & Co. K.G., AstraZeneca, Boehringer Ingelheim Pharma GmbH & Co. K.G., Grünenthal GmbH, F. Hoffmann-La Roche Ltd., Merck KGaA, and SANOFI) for the PharMetrX program. In addition, C.K. reports research grants from the Innovative Medicines Initiative-Joint Undertaking (“DDMoRe”) and Diurnal Ltd. C.K. report grants from the Federal Ministry of Education and Research within the Joint Programming Initiative on Antimicrobial Resistance Initiative (JPIAMR). All funding was outside the submitted work. F.K. and L.K.-S. are current employees of Boehringer Ingelheim Pharma GmbH & Co. K.G. and Merck KGaA, respectively. All other authors declared no competing interests for this work. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Topics & Concepts