Alzahra Hamdan1, Andreas Traschütz2,3, Lukas Beichert2,3, Xiaomei Chen1, Rebecca Schüle2,4, Andrew C. Hooker1, EVIDENCE-RND consortium, PROSPAX consortium, Matthis Synofzik2,3, Mats O. Karlsson1
1Pharmacometrics Research Group, Department of Pharmacy, Uppsala University, , 2Department of Neurodegenerative Diseases, Center for Neurology and Hertie Institute for Clinical Brain Research, University of Tübingen, 3German Center for Neurodegenerative Diseases (DZNE) Tübingen, 4Division of Neurodegenerative Diseases, Department of Neurology, Heidelberg University Hospital
Background: Robust and sensitive outcomes are needed to enable precise estimation of disease severity, longitudinal changes, and treatment effects in rare degenerative ataxias. Degenerative ataxias are a group of heterogeneous diseases that affect the cerebellum and its associated tracts, leading to gait and limb coordination disturbances. Among other outcomes, digital-motor outcomes (DMOs) are used to assess the severity and progression of ataxia. DMOs are sensor-based performance outcomes that quantitatively assess the patient’s gait and limb coordination using multiple measures/metrics (1,2). In previous studies, DMOs displayed high sensitivity even in early and pre-ataxic stages (2–4) compared to other ataxia outcomes, including the Scale for the Assessment and Rating of Ataxia (SARA) (5). SARA is, so far, the most widely used outcome for ataxia; however, due to the slow-progressive nature of degenerative ataxias, and despite the good item performance of SARA (6), its sensitivity to longitudinal changes was shown to be insufficient to support small trials with short durations (7). To better inform decision-making in clinical trials, a model-based comparison of SARA and DMOs as outcomes of ataxia is needed. In this work, we develop Item Response Theory (IRT) models of DMOs that describe the individual DMO measures data using a common underlying variable that reflects ataxia severity. IRT is a powerful statistical method used to evaluate the performance of clinical outcome assessments and facilitates more powerful analyses through their efficient use of composite data (8). In contrast to the traditional use of IRT limited to discrete data, we present here an IRT application for continuous data from multiple measures. Methods: Data were obtained from the prospective natural history study for spastic ataxias; PROSPAX (NCT04297891) (9). SARA and DMOs data were collected from 199 patients with a total of 517 visits. The DMOs include 11 digital gait measures, and 18 digital limb coordination measures captured using body-worn (APDM) and stationary sensors (Q-Motor), respectively. Separate IRT models were built for digital gait and digital limb coordination measures. Linear and non-linear functions were tested to describe data from each measure as a function of a common latent variable (i.e., ataxia severity) underlying the DMO measures. Non-linear functions include 4-parameter logistic functions with lower and upper asymptotes, slope, and inflection point parameters. 3-parameter functions were tested when needed by fixing lower/upper asymptotes. The model selection was based on the Akaike Information Criterion as a measure of goodness-of-fit and diagnostic evaluations (e.g., residuals-based diagnostics). The uncertainty of estimated functions was determined by sampling from the uncertainty distribution of model parameters and calculating the 95% confidence interval of model predictions. Correlations between residual errors were estimated to account for any residual correlations between measures that are not described by the latent variable. Model estimation was performed using NONMEM version 7.5.1 and Perl-Speaks-NONMEM (10). Results: Nonlinear functions were selected for most DMO measures indicating their differential informativeness at different ataxia levels including identifying ataxia levels where a certain measure is no longer sensitive. IRT models were reduced by selecting the top-ranking measures with additive residual error variance (s²) = 0.50, and acceptable model certainty, as the most informative measures. Final IRT models included 4 digital gait measures (s²=0.05-0.48), all with non-linear functions, and 7 digital limb coordination measures (s²=0.18-0.60), two of them are with linear functions. Selected gait measures are from 3 gait domains (pace and spatial/temporal variability). Digital limb coordination measures are from 4 different tasks that capture 4 different movement features (speed, efficiency, smoothness, and variability). Two from each type were also among the best three measures chosen in the previous work with an integrated IRT model of SARA and digital-motor outcomes (11). Conclusions: The developed IRT models allow the description of individual ataxia severity underlying multiple digital-motor measures while incorporating their differential informativeness, which in future work can be extended to assess longitudinal changes and support model-based comparisons with other ataxia outcomes.
1. Hermle D, Schubert R, Barallon P, Ilg W, Schüle R, Reilmann R, et al. Multifeature quantitative motor assessment of upper limb ataxia including drawing and reaching. Annals of Clinical and Translational Neurology. 2024;11(5):1097–109. 2. Ilg W, Milne S, Schmitz-Hübsch T, Alcock L, Beichert L, Bertini E, et al. Quantitative Gait and Balance Outcomes for Ataxia Trials: Consensus Recommendations by the Ataxia Global Initiative Working Group on Digital-Motor Biomarkers. Cerebellum [Internet]. 2023 Nov 13 [cited 2023 Nov 16]; Available from: https://doi.org/10.1007/s12311-023-01625-2 3. Ilg W, Müller B, Faber J, van Gaalen J, Hengel H, Vogt IR, et al. Digital Gait Biomarkers Allow to Capture 1-Year Longitudinal Change in Spinocerebellar Ataxia Type 3. Mov Disord. 2022;37(11):2295–301. 4. Rochester L, Galna B, Lord S, Mhiripiri D, Eglon G, Chinnery PF. Gait impairment precedes clinical symptoms in spinocerebellar ataxia type 6. Mov Disord. 2014 Feb;29(2):252–5. 5. Schmitz-Hübsch T, du Montcel ST, Baliko L, Berciano J, Boesch S, Depondt C, et al. Scale for the assessment and rating of ataxia: development of a new clinical scale. Neurology. 2006;66(11):1717–20. 6. Hamdan A, Hooker AC, Chen X, Traschütz A, Schüle R, Group AS, et al. Item performance of the scale for the assessment and rating of ataxia in rare and ultra-rare genetic ataxias. CPT: Pharmacometrics & Systems Pharmacology. 2024;13(8):1327–40. 7. Hamdan A, Hendrickx N, Hooker AC, Chen X, Comets E, Traschütz A, et al. Longitudinal Analysis of Natural History Progression of Rare and Ultra-Rare Cerebellar Ataxias Using Item Response Theory. Clinical Pharmacology & Therapeutics. 2024;116(6):1593–605. 8. Ueckert S. Modeling Composite Assessment Data Using Item Response Theory. CPT Pharmacometrics Syst Pharmacol. 2018;7(4):205–18. 9. PROSPAX [Internet]. [cited 2024 Mar 13]. PROSPAX – an integrated multimodal progression chart in spastic ataxias. Available from: https://www.prospax.net/ 10. Karlsson MO, Nordgren R. Perl-speaks-NONMEM: User documentation [Internet]. [cited 2025 Feb 4]. Available from: https://uupharmacometrics.github.io/PsN/docs.html 11. Hamdan A, Traschütz A, Beichert L, Chen X, Schüle R, Hooker AC, et al. Integrated modeling of digital-motor outcomes and clinical outcome assessments using IRT: a framework for developing better outcomes for clinical trials in rare neurological diseases. In 2024. Available from: https://www.page-meeting.org/default.asp?abstract=11161
Reference: PAGE 33 (2025) Abstr 11463 [www.page-meeting.org/?abstract=11463]
Poster: Drug/Disease Modelling - CNS