Table 1.

Examples of how published (mechanistic) models rooted in the clinical development space (model informed drug development, MIDD) could address uncertainties in new medicinal product assessment reports.

Uncertainty not completely addressed in competent authority assessment report	Example use of MIDD relevant to address uncertainty potentially also during HTA
What is the optimal dosage in the clinical context?	Physiologically based pharmacokinetic models can investigate dosing-regimens relevant for regulatory review and product labels (9) and can also mimic real-life adherence to prescribed treatment regimens (see also below) or pharmacology-relevant characteristics of special populations as well as drug-drug interactions.
What is the duration of the effectiveness, especially with chronic use of a treatment?	Mechanistic models can predict the long-term disease progression by extrapolation of shorter-term findings under the constraints of how the components of the system function (and these constraints convey biological plausibility by design). An example is the use of a mechanism-based disease progression model for comparison of long-term effects of pioglitazone, metformin, and gliclazide on disease processes underlying Type 2 Diabetes Mellitus (10). Another example is prediction of long-term outcomes by short-term marker data as demonstrated by a semi-mechanistic approach in context of osteoporosis treatment (11).
What is the efficacy for relevant clinical outcomes?	Mechanistic models combined with pharmacometric approaches can translate findings for one outcome to a range of other outcomes. An example of survival modeling on the back of a mechanistic description is the modeling framework for CD19-Specific CAR-T cell immunotherapy using a quantitative systems pharmacology model (12).
What is the size of the clinical effect dependent on patient characteristics and extrinsic factors?	Data-driven modeling techniques can capture correlation within clinical data. Describing the clinical effect of a drug can also be based on mechanistic considerations. Such models either (a) link disease phenotypes to increasingly granular mathematical representations of pathophysiologic processes (top-down approach) or (b) derive functional, computable cellular networks from the molecular building blocks of genes and proteins to elucidate the impact of pathologic or therapeutic alterations on network operating states and hence clinical phenotype (bottom-up) [see (13)]. In this way, functional relationships can explain the found correlations and can be used for quantitative analysis of the effect size and the causality dependent on intrinsic and extrinsic factors.
What is the difference in effect when compared head-to-head to other comparators?	Mechanistic modeling is a commonly used tool to explore treatment combinations in immuno-oncology [see for example (14)] which can enable head-to-head comparisons. A mechanistic approach with clinical trial simulation can provide model-based meta-analysis which can ameliorate indirect comparison of clinical data (15).
What is the efficacy compared to placebo or the standard of care, when controlled studies are hard to conduct?	For comparative effectiveness research, data from a control arm is needed. When such control arm is unfeasible (for example because of ethical reasons), external or synthetic control data may be an avenue to put uncontrolled clinical data into a controlled setting, but mitigation of the risk of bias needs adjustment techniques. Mechanistic modeling can quantitatively predict the effect of an intervention on a clinical outcome as a function of patient characteristics and extrinsic factors, on a single patient level. These features render mechanistic models promising to set up unbiased synthetic control arms [SCA, see (16)].
What is the effect of real-life compliance on efficacy?	Explicit simulation of administration adherence can be coupled with pharmacokinetic models. One example is the simulation of adherence patterns using Markov Chains for trial design (17, 18).
What is the distribution of responders in the target population?	Predicting individual response to treatments needs the convergence of large-scale mechanistic models [e.g., in cancer pathways (19)], appropriate responder profiling framework and cost-effectiveness analysis [for example the Effect Model approach, see (20, 21)]
What is the size of the benefit at the population level?	Mechanistic models providing clinical outcome estimates can be used on the entire population level to predict effectiveness, given that adapted metrics are used (22)
What is the long-term safety and what impact does the occurrence of rare side effects have over long-term use?	The combination of quantitative systems toxicity (23) with organ (e.g., cardiac, and renal) impairment (24) in frame of disease progression modeling (25) can be used to simulate long term safety aspects of a treatment from a mechanistic point of view

Emphasis is put on mechanistic models.