Luke Kosinski1, Lauren Quinlen1, Nathan Cunicelli1, Terina Martinez1, Christopher Mezias1, Yi Zhang1
1Critical Path Institute
Introduction/Objectives: Huntington’s disease (HD) is a rare, fatal, and progressive neurodegenerative disorder that has been challenging to develop disease modifying therapies for. Disease progression modeling of HD has potential to help inform new trials and de-risk drug development, but disease progression in HD is typically assessed using clinical rating scales which are integer in nature. Modeling strategies that treat endpoints as continuous violate the integer nature of the data, while ordered categorical approaches require a large number of parameters for larger scales. We apply the bounded integer model [1, 2] to the total motor score (TMS) subsection of the composite Unified HD Rating Scale (cUHDRS) [3] and compare its performance to logistic and beta regression model structures. The bounded integer model divides a normal distribution into equal-area bins corresponding to each integer possibility of a given rating scale, preserving the integer nature of the data, and uses a latent variable to estimate the probability of a rating, keeping the number of parameters low. Note that we do not test ordered categorical models as TMS has over 100 possible values, requiring excessive numbers of parameters for categorical approaches. Methods: Aggregated data from four clinical trials – 2CARE [4], CREST-E [5], AMARYLLIS [6], and First-HD [7] – and four observational studies – Enroll-HD [8], Predict-HD [9], TRACK-HD [10], and TRACK-ON [11] – comprising over 16,500 HD patients from the Critical Path Institute database was leveraged for modeling. The data selection criteria are: A) complete cases analysis for TMS , B) at least 40 CAG repeats, the cutoff for 100% likelihood of developing the inherited disease [12], and C) a baseline prognostic score index [13] of at least zero. The analysis set post-data selection contains over 12,500 patients. Four base models of TMS over time were compared: a generalized logistic model, a generalized logistic model with model-fit bounds, a beta regression model with a logit link function, and the bounded integer model. Patient-level random effects were included for both intercept and slope terms. Models were fit in NONMEM version 7.5, PsN 5.4, and Pirana 24.9.2, and compared using mean squared error (MSE) and mean absolute error (MAE) from a 5-fold cross validation. Results: All models demonstrated reasonable performance based on diagnostic plots and visual predictive checks (VPCs). VPCs indicated slightly improved performance for the bounded integer model, with good performance at the mid-range and improved performance at the upper 95 % and lower 5% intervals relative to the other models. The 5-fold cross validation results for the generalized logistic, the generalized logistic with model-fit bounds, the beta with a logit link, and the bounded integer model, were 90.57, 93.35, 92.27, and 88.27, respectively, for MSE and 15.58, 15.37, 15.49, and 14.82, respectively, for MAE. These results indicate consistently better performance for the bounded integer model compared to the other models. Conclusions: The bounded integer model better models TMS in our HD data than the logistic and beta regression approaches explored here. These results suggest that disease progression models in HD, which are useful in their ability to help understand the disease and plan clinical trials, may benefit from modeling TMS or other cUHDRS subsections using a bounded integer approach.
[1] Wellhagen, Gustaf J., Maria C. Kjellsson, and Mats O. Karlsson. “A bounded integer model for rating and composite scale data.” The AAPS journal 21 (2019): 1-8. [2] Ueckert, Sebastian, and Mats O. Karlsson. “Improved numerical stability for the bounded integer model.” Journal of Pharmacokinetics and Pharmacodynamics 48 (2021): 241-251. [3] Mestre, Tiago A., Maria João Forjaz, Philipp Mahlknecht, Francisco Cardoso, Joaquim J. Ferreira, Ralf Reilmann, Cristina Sampaio et al. “Rating scales for motor symptoms and signs in Huntington’s disease: critique and recommendations.” Movement disorders clinical practice 5, no. 2 (2018): 111-117. [4] McGarry, Andrew, Michael McDermott, Karl Kieburtz, Elisabeth A. de Blieck, Flint Beal, Karen Marder, Christopher Ross et al. “A randomized, double-blind, placebo-controlled trial of coenzyme Q10 in Huntington disease.” Neurology 88, no. 2 (2017): 152-159. [5] Hersch, Steven M., Giovanni Schifitto, David Oakes, Amy-Lee Bredlau, Catherine M. Meyers, Richard Nahin, Herminia Diana Rosas, and Huntington Study Group CREST-E Investigators and Coordinators. “The CREST-E study of creatine for Huntington disease: A randomized controlled trial.” Neurology 89, no. 6 (2017): 594-601. [6] Delnomdedieu, Marielle, Ye Tan, Adam Ogden, Zdenek Berger, and Ralf Reilmann. “J06 A randomized, double-blind, placebo-controlled phase ii efficacy and safety study of the PDE10A inhibitor PF-02545920 in Huntington disease (amaryllis).” (2018): A99-A100. [7] Frank, Samuel, Claudia M. Testa, David Stamler, Elise Kayson, Charles Davis, Mary C. Edmondson, Shari Kinel et al. “Effect of deutetrabenazine on chorea among patients with Huntington disease: a randomized clinical trial.” Jama 316, no. 1 (2016): 40-50. [8] Sathe, Swati, Jen Ware, Jamie Levey, Eileen Neacy, Robi Blumenstein, Simon Noble, Alzbeta Mühlbäck, Anne Rosser, G. Bernhard Landwehrmeyer, and Cristina Sampaio. “Enroll-HD: an integrated clinical research platform and worldwide observational study for Huntington’s disease.” Frontiers in Neurology 12 (2021): 667420. [9] Paulsen, Jane S., Douglas R. Langbehn, Julie C. Stout, Elizabeth Aylward, Christopher A. Ross, Martha Nance, Mark Guttman et al. “Detection of Huntington’s disease decades before diagnosis: the Predict-HD study.” Journal of Neurology, Neurosurgery & Psychiatry 79, no. 8 (2008): 874-880. [10] Tabrizi, Sarah J., Douglas R. Langbehn, Blair R. Leavitt, Raymund AC Roos, Alexandra Durr, David Craufurd, Christopher Kennard et al. “Biological and clinical manifestations of Huntington’s disease in the longitudinal TRACK-HD study: cross-sectional analysis of baseline data.” The Lancet Neurology 8, no. 9 (2009): 791-801. [11] Klöppel, Stefan, Sarah Gregory, Elisa Scheller, Lora Minkova, Adeel Razi, Alexandra Durr, Raymund AC Roos et al. “Compensation in preclinical Huntington’s disease: evidence from the track-on HD study.” EBioMedicine 2, no. 10 (2015): 1420-1429. [12] Tabrizi, Sarah J., Scott Schobel, Emily C. Gantman, Alexandra Mansbach, Beth Borowsky, Pavlina Konstantinova, Tiago A. Mestre et al. “A biological classification of Huntington’s disease: the Integrated Staging System.” The Lancet Neurology 21, no. 7 (2022): 632-644. [13] Long, Jeffrey D., Douglas R. Langbehn, Sarah J. Tabrizi, Bernhard G. Landwehrmeyer, Jane S. Paulsen, John Warner, and Cristina Sampaio. “Validation of a prognostic index for Huntington’s disease.” Movement Disorders 32, no. 2 (2017): 256-263.
Reference: PAGE 33 (2025) Abstr 11512 [www.page-meeting.org/?abstract=11512]
Poster: Methodology - Estimation Methods