F025 A robust method for studying early motor function decline in Huntington’s disease (HD) using enroll-HD data

<h3>Background</h3> Understanding the development of motor function decline, particularly in the early stages, presents unique opportunities to investigate windows for early intervention. Evidence suggests that motor function may have an accelerated decline prior to motor onset. Although various methods have been developed to study these issues, they often fail to account for finer details such as distribution assumptions and outlying observations. <h3>Objective</h3> To develop a method that investigates whether an accelerated decline in motor function occurs before motor onset, without assuming normal distributions and allowing for robust performance in the presence of outliers. <h3>Methods</h3> We developed an anchoring-event based changepoint model for longitudinal measures of motor function using the quantile regression framework. The timing of motor onset, defined as the first instance of reaching a diagnostic confidence level score of 4, was used as the anchoring event. The model assumes a linear trend before and after the changepoint. <h3>Results</h3> We used the Enroll-HD dataset as the primary data source and applied the proposed method to participants with a CAG length of 42, which is the most prevalent among HD cases. The estimated changepoints were 1.18, 2.16, and 2.54 years prior to motor onset at the 10th, 50th, and 90th quantiles, respectively, compared to 2.18 years from the method that assumes normal distributions. <h3>Conclusions</h3> The quantile-model based method offers granular information for the changepoints. The estimates from the quantile-based method at median are close to those of the mean-based model that assumes normal distributions for motor function.