Dose selection is perhaps the most fundamental aspect that drug developers need to get right for a drug to be successful. As the Swiss physician and alchemist Paracelsus already stated in the 16th century, “Only the dose makes the poison.” Beyond its toxicological relevance, the right dose must also influence human biology in the intended way, ultimately leading to therapeutic benefit.
This challenge is already evident in early clinical development. In first-in-human studies, dose ranges should ideally span from minimally active to supratherapeutic exposures to characterise both efficacy and safety boundaries. Results derived from these studies must subsequently inform Phase II dose-finding trials in relevant patient populations, where the goal is to identify optimal therapeutic doses to be tested in late-stage clinical trials. As such, early dose selection decisions have far-reaching consequences, shaping each subsequent phase of development and underscoring the importance of getting them right early on. However, defining such “relevant” exposure ranges is far from straightforward. Preclinical data, often derived from multiple species, are inherently variable and may not directly translate to humans. Moreover, as clinical data emerge from Phase I and II studies, biomarker responses must also be carefully evaluated and interpreted, adding another layer of complexity to the decision-making process.
Pharmacometrics has emerged as an indispensable method to address these challenges by integrating diverse data sources into a coherent quantitative framework. Through mathematical modelling of dose–exposure–response relationships, pharmacometric approaches link administered dose to pharmacokinetics (PK) and downstream pharmacodynamic (PD) effects. This allows not only a structured interpretation of observed data, but also the identification of key sources of variability between individuals, species, and study settings.
Importantly, pharmacometric models often incorporate mechanistic or semi-mechanistic representations of the underlying biology, enabling translation across contexts. For example, allometric scaling and physiologically based pharmacokinetic (PBPK) models can be used to predict human PK from animal data, while accounting for differences in absorption, distribution, and metabolism. Similarly, differences in target affinity, potency, and turnover rates between species, as well as between healthy and diseased populations, can be incorporated into PD models, supporting translation from preclinical systems to humans or from healthy participants to patients. Once established, these models also allow simulation of untested scenarios, such as alternative dosing regimens, patient subpopulations, or combination therapies.
In practice, pharmacometric approaches support decision-making in early development in several important ways.
First, they enable more informed dose selection for Phase I and II studies. By integrating preclinical PK, PD, and safety data across species, pharmacometric models can predict human drug exposures and their associated levels of pharmacological activity and toxicity. For instance, in the development of a monoclonal antibody targeting an inflammatory pathway, preclinical models may describe target-mediated drug disposition (PK) and cytokine suppression (PD). Translating these models to humans can help select dose levels expected to achieve various levels of target engagement, ensuring that early clinical studies adequately explore the full exposure–response curve.
Second, pharmacometrics facilitates translation from early signals to expected clinical effects. Biomarker responses observed in Phase I studies are often the first indication of drug activity in humans. By linking biomarker changes to drug exposure and incorporating knowledge of disease biology, pharmacometric models can be used to project potential clinical outcomes. For example, in patients with Parkinson’s disease who carry the pathogenic leucine-rich repeat kinase 2 (LRRK2) variant, increased kinase activity of LRRK2 is thought to underlie disease pathology. Preclinical studies have shown that reducing its activity leads to disease modification by restoring lysosomal function. Target engagement can be measured in animals and humans using phosphorylated serine 935 LRRK2 in whole blood and phosphorylated threonine Rab10, a direct substrate of LRRK2, in peripheral blood mononuclear cells.1 Quantitatively linking changes in these biomarkers to plasma drug concentrations helps demonstrate target engagement, increasing confidence in disease modulation by the drug. Such analyses can guide go/no-go decisions and support the design of Phase II trials.
Third, pharmacometric modelling improves the interpretation of biomarker data. Biomarkers are often subject to substantial variability, delayed responses, and indirect relationships with drug concentrations. (Semi-)mechanistic PK/PD models can disentangle these dynamics, distinguishing drug effects from baseline variability or disease progression. For example, in acromegaly, a pituitary adenoma disrupts the highly regulated mechanisms that control the stimulation and inhibition of growth hormone (GH), leading to severe hypersecretion. As GH is regulated via a circadian rhythm associated with high intra- and inter-individual variability in pulsatility, assessment of drug effects is difficult and the ability to simulate new dosing regimens is limited. Comprehensive population PK/PD modelling has been able to describe the concentration–effect relationship of a somatostatin receptor agonist on reducing GH secretion while maintaining an individual’s pulsatile profile, enabling simulations to estimate probability of success and inform the design of new trials.2 This case study elegantly shows how modelling these dynamics allows a more accurate characterisation of drug potency and onset of action, preventing misinterpretation of noisy data.
Ultimately, early integration of pharmacometrics into drug development enables a more quantitative and mechanistic understanding of dose–exposure–response relationships. By leveraging all available data and providing a framework for the simulation and prediction of new scenarios, pharmacometric approaches support more confident decision-making, reduce uncertainty, and increase the likelihood of selecting the right dose for later-stage trials.
References
Jennings D, Huntwork-Rodriguez S, Vissers MFJM, et al. LRRK2 Inhibition by BIIB122 in Healthy Participants and Patients with Parkinson’s Disease. Mov Disord. 2023;38(3):386-398. doi:10.1002/MDS.29297
van Esdonk MJ, Burggraaf J, Dehez M, van der Graaf PH, Stevens J. Quantification of the endogenous growth hormone and prolactin lowering effects of a somatostatin-dopamine chimera using population PK/PD modeling. J Pharmacokinet Pharmacodyn. 2020;47(3):229-239. doi:10.1007/S10928-020-09683-3
