Britta Steffens 1,2, Freya Bachmann 3, Johannes Schropp 3, Marc Pfister 1, Gabor Szinnai 4, Gilbert Koch 1
1 Pediatric Pharmacology and Pharmacometrics, University of Basel Children's Hospital UKBB (Basel, Switzerland), 2 School of Life Sciences, University of Applied Sciences and Arts Northwestern Switzerland (Muttenz, Switzerland), 3 Department of Mathematics and Statistics, University of Konstanz (Konstanz, Germany), 4 Pediatric Endocrinology and Diabetology, University of Basel Children's Hospital UKBB (Basel, Switzerland)
Objectives: Can pharmacometric (PMX) modeling support individualized dose optimization in pediatric thyroid diseases, where inter-individual variability, developmental changes, and heterogeneous clinical practices challenge conventional dosing strategies?
This question initiated our 10-year journey to translate an unmet clinical need into a scalable digital solution for routine pediatric thyroid care through close international collaboration among clinicians, pharmacometricians, and mathematicians.
In pediatric thyroid care, timely and appropriate treatment is crucial, as hormone imbalance may impair neurodevelopment, cognition, and growth. Dosing is guided by established recommendations combined with clinicians’ expertise and patient-specific therapeutic goals, typically defined by achieving a target free thyroxine (FT4) concentration at specific time points. However, variability in disease dynamics and drug responses, particularly in the neonatal period and during puberty, makes maintaining target FT4 concentrations challenging. As a result, over- and underdosing persist despite frequent laboratory monitoring and repeated dose adjustments. Such deviations from the respective age-specific FT4 reference range may increase the risk of adverse drug effects and negatively affect patients’ quality of life.
Therefore, we pursued three milestones: (i) development of a mathematical algorithm for individualized dose optimization targeting predefined FT4 concentrations, (ii) design and implementation of a clinically applicable and scalable clinical decision support system (CDSS), and (iii) initiation of a prospective randomized controlled trial (RCT) across diverse international settings.
Methods: Two independent retrospective studies, national and international, were conducted to develop and refine PMX models describing FT4 dynamics under treatment in two pediatric thyroid diseases, congenital hypothyroidism (CH) and Graves’ disease (GD). Models were developed within the nonlinear mixed-effects (NLME) framework, yielding both population and individual parameter estimates. In parallel, concept of dose optimization algorithm OptiDose [1], based on optimal control theory, was developed to calculate optimal dosing regimens by minimizing an objective function quantifying the deviation between a predefined target and the model prediction. Refined PMX models for the two thyroid diseases were integrated into OptiDose to calculate individual doses targeting predefined FT4 concentrations. In clinical practice, dose optimization is intended to take place at each outpatient visit, with a fixed daily dose prescribed until the subsequent scheduled visit. In close collaboration with clinicians, a clinically applicable workflow and user-friendly graphical user interface (GUI) were designed to ensure feasibility and enable seamless integration into routine clinical practice.
Results: The overall cohort comprised 267 patients with CH and 141 with GD (median age at diagnosis 8 days and 12 years, respectively) with 250 and 127 FT4 measurements, respectively, available at diagnosis. PMX models incorporated weight progression and relied exclusively on routinely available covariates, including age and disease severity, making them applicable in low- and middle-income countries as well. Initial models [2,3] and the combined algorithm were iteratively refined to address heterogeneity across international centers, including differences in medications, formulations and formulation strengths, laboratory assays, visit schedules, and the clinicians’ dosing habits. The finalized algorithm, OptiThyDose, was integrated into the CDSS, whose GUI was designed to capture relevant clinical inputs and reflect center-specific variability, with a focus on clear and comprehensive visualization of dosing recommendations. Initial retrospective proof-of-concept analyses indicated fewer dose adjustments, reduced dosing variability, and lower cumulative drug exposure while maintaining target FT4 concentrations. To assess clinical performance, an international non-inferiority RCT across three continents was initiated (target sample size: 75 patients per disease), comparing the proportion of patients achieving FT4 concentrations within age-specific laboratory reference ranges with and without support from OptiThyDose.
Conclusion: The lessons learned from our 10-year translational journey underscore four principles for developing a clinically meaningful and scalable CDSS: (i) translating a clinical challenge into a robust CDSS requires iterative refinement, (ii) stakeholder engagement should begin before PMX model development to ensure clinical relevance and sustained usability, (iii) scalability demands explicit consideration of global heterogeneity in clinical workflows, laboratory assays, and medication management, and (iv) combining PMX modeling with optimal control enables individualized, target-driven dose optimization and has the potential to improve clinical outcomes in pediatric thyroid care by increasing the likelihood of achieving and maintaining therapeutic FT4 concentrations.
References:
[1] Bachmann F et al., J Optim Theory Appl. (2021) 189(1):46-65.
[2] Koch G et al., J Pharmacokin Pharmacodyn (2021) 48(5):711-723
[3] Steffens B et al., Front. Med. (2023) 10:1099470
Reference: PAGE 34 (2026) Abstr 12228 [www.page-meeting.org/?abstract=12228]
Poster: Oral: Clinical Applications