Affiliations: [a] Department of Statistics, Texas A&M University, College Station, TX, USA
| [b] Department of Biostatistics, Columbia University, New York, NY, USA
| [c] Department of Neurology, Georgetown University, Washington, DC, USA
| [d] Department of Psychological and Brain Sciences, The University of Iowa, Iowa City, IA, USA
| [e] Department of Neurology, Columbia University Medical Center, New York, NY, USA
Correspondence to: Tanya P. Garcia, PhD, Department of Statistics, Texas A&M University, 3143 TAMU, College Station, TX 77843, USA. Tel.: +1 979 845 3141; E-mail: [email protected].
Abstract: Background:Critical to discovering targeted therapies for Huntington disease (HD) are validated methods that more precisely predict when clinical outcomes occur for different patient profiles. Objective:To more precisely predict the probability of when motor diagnosis (diagnostic confidence level 4) on the Unified Huntington’s Disease Rating Scale (UHDRS), cognitive impairment (two or more neuropsychological scores on the UHDRS were 1.5 standard deviations below normative means) and Stage II Total Functional Capacity (TFC) first occur by accounting for dependencies between these outcomes. Methods:Adult premanifest participants with ≥36 CAG repeats were selected from multi-center, longitudinal, observational studies: Prospective Huntington At Risk Observational Study (PHAROS, n = 346), Neurobiological Predictors of Huntington Disease (PREDICT, n = 909); and Cooperative Huntington Observational Research Trial (COHORT, n = 430). Probabilities were estimated for each study, and pooled using the Joint Progression of Risk Assessment Tool (JPRAT) which accounts for dependencies between outcomes. Results:All studies had similar probabilities of when motor diagnosis, cognitive impairment, and Stage II TFC first occurred. Probability estimates from JPRAT were 43% less variable than from models that ignored dependencies between outcomes. The probability of experiencing motor-diagnosis, cognitive impairment, and Stage II TFC within 5 years was 10%, 18%, and 7%, respectively for 45-year-olds with 42 CAG repeats, and was 4%, 10% and 5%, respectively, for 40 year olds with 42 CAG repeats. Conclusions:Improved predictions from JPRAT may benefit treatment studies of rare diseases and is an alternative to composite outcomes when the objective is interpreting individual outcomes within the same model.