Mohamed Gewily1, Elodie L. Plan1, Elham Yousefi2, Franz König2, Martin Posch2, Franziska Hopfner3, Günter Höglinger3,4,5,6, Mats O. Karlsson1
1Department of Pharmacy, Uppsala University, Uppsala, Sweden 2Center for Medical Data Science, Medical University of Vienna, Vienna, Austria 3Department of Neurology, LMU University Hospital, Ludwig-Maximilians-Universität (LMU) München, Munich, Germany. 4German Center for Neurodegenerative Diseases (DZNE), Munich, Germany. 5Munich Cluster for Systems Neurology (SyNergy), Munich, Germany. 6Department of Neurology, Hannover Medical School, Hanover, Germany
Background:
Progressive Supranuclear Palsy (PSP) is a neurodegenerative late-onset disease that is challenging in terms of assessment. The Progressive Supranuclear Palsy Rating Scale (PSPRS), a 28-item clinician reported questionnaire, is the established clinical outcome assessment method to assess the severity of PSP. Recently, the U.S. Food and Drug Administration (FDA) has proposed a subscale of 10 items (PSPRS-10) as an alternative to PSPRS. The FDA also recommended the rescoring of some of those items into coarser response categories (rPSPRS-10).
Objectives:
Our objective was to quantitatively evaluate and compare the properties of PSPRS, PSPRS-10, and rPSPRS-10 using Item Response Theory (IRT) models. We also aimed to develop a progression model of the disease and assess relative merits of study designs and analysis options.
Methods:
Data of 974 patients from 4 interventional trials and 2 registries (1–5) were available for analysis. Our workflow was divided into: i) fitting IRT models to the three scales and assessing if correlations in item variability can be explained by a single latent variable (a.k.a. unidimensionality) ii) estimating informativeness of the items of the relevant scales iii) estimating disease progression rate iv) comparing the scales, trial designs, and analysis options with respect to power to detect a clinically relevant disease-modifying treatment effect.
Results:
A single latent variable could not account for the correlation in item variability of PSPRS and three of the items showed no correlation to the remaining items. The analysis indicated that the 10 FDA-selected subscale items were the most informative of the original PSPRS items and that they contained almost 75% of the total information content of the original PSPRS 28 items. There was a loss of item information after following the FDA rescoring recommendation, ranging from 3% to 36%.
The disease severity of patients from the interventional trials was predicted to be higher at baseline and 38% more rapidly progressing compared to registry data. Based on trial simulations, IRT-based longitudinal modelling of the FDA subscale required markedly lower number of subjects to detect treatment effects in two-armed parallel designs compared to end-of-treatment analysis of PSPRS. For example, longitudinal IRT-based analysis of PSPRS-10 and rPSPRS-10 predicted a considerably smaller sample size (38 and 44 patients, respectively) needed to detect a 50% change in progression with 80% power compared to an estimate (102 patients (6)) based on a PSPRS total score end-of-treatment analysis
Conclusions:
The IRT analysis revealed that the PSPRS has some suboptimal properties, since some of its items were found to not be contributing to the estimation of the same latent variable as others. The FDA-selected items coincided with those estimated to be most informative in the IRT analysis.
Based on PSPRS-10, a longitudinal model to predict individual patient progression was developed. The model was used to evaluate different trial designs in terms of power to detect a treatment effect. The IRT analysis was more powerful in terms of detecting a disease-modifying treatment effect compared to classic total score linear models among other alternative methods.
Not performing the FDA rescoring recommendation had a positive influence on power. The analysis in general indicated that the power of a trial is positively influenced by analysis methods making use of item level data, and incorporating all studied time points.
This work also demonstrates how IRT can be used to quantitatively evaluate and compare different scale versions.
References:
- Dam T, Boxer AL, Golbe LI, Höglinger GU, Morris HR, Litvan I, et al. Safety and efficacy of anti-tau monoclonal antibody gosuranemab in progressive supranuclear palsy: a phase 2, randomized, placebo-controlled trial. Nat Med. 2021 Aug;27(8):1451–7.
- Tolosa E, Litvan I, Höglinger GU, Burn D, Lees A, Andrés MV, et al. A phase 2 trial of the GSK-3 inhibitor tideglusib in progressive supranuclear palsy. Mov Disord Off J Mov Disord Soc. 2014 Apr;29(4):470–8.
- Nuebling G, Hensler M, Paul S, Zwergal A, Crispin A, Lorenzl S. PROSPERA: a randomized, controlled trial evaluating rasagiline in progressive supranuclear palsy. J Neurol. 2016 Aug;263(8):1565–74.
- Respondek G, Höglinger GU. DescribePSP and ProPSP: German Multicenter Networks for Standardized Prospective Collection of Clinical Data, Imaging Data, and Biomaterials of Patients With Progressive Supranuclear Palsy. Front Neurol [Internet]. 2021 [cited 2023 Nov 7];12. Available from: https://www.frontiersin.org/articles/10.3389/fneur.2021.644064
- Höglinger GU, Litvan I, Mendonca N, Wang D, Zheng H, Rendenbach-Mueller B, et al. Safety and efficacy of tilavonemab in progressive supranuclear palsy: a phase 2, randomised, placebo-controlled trial. Lancet Neurol. 2021 Mar 1;20(3):182–92.
- Stamelou M, Schöpe J, Wagenpfeil S, Del Ser T, Bang J, Lobach IY, et al. Power calculations and placebo effect for future clinical trials in progressive supranuclear palsy. Mov Disord Off J Mov Disord Soc. 2016 May;31(5):742–7.
Reference: PAGE 32 (2024) Abstr 10952 [www.page-meeting.org/?abstract=10952]
Poster: Drug/Disease Modelling - CNS