Â鶹´«Ã½Ó³»­

Â鶹´«Ã½Ó³»­

Explore the latest content from across our publications
Join The Â鶹´«Ã½Ó³»­

Log In

Forgot Password?
Create New Account

Loading... please wait

Abstract Details

Development of a Multivariable Proxy Model for Six-Minute Walk Distance (6MWD) in Duchenne Muscular Dystrophy (DMD) Using Machine Learning Methods
Movement Disorders
P11 - Poster Session 11 (8:00 AM-9:00 AM)
3-007
The study objective was to identify a 6MWD proxy using variables commonly available in clinical practice.
6MWD is a common functional test in DMD clinical trials, but is rarely utilized by clinicians. 
A retrospective cohort study was conducted in DMD patients with exon-51 skip-amenable mutations from the placebo arm of DEMAND III, a 48-week phase-3 drisapersen trial. The outcome was 6MWD (in meters) measured at 2-5 visits per patient. Candidate predictors included demographics, genetic markers, functional outcomes, lab values, vitals, body composition (DEXA) measures, and pulmonary and heart function measurements with <25% missing values. Remaining missing values were imputed using single mean-value imputation (continuous predictors) and most-frequent value imputation (categorical predictors). Imputation flags were included, and continuous predictors were standardized. Predictors were selected using regression trees, optimal regression trees (ORT), random forests, and LASSO regression. Ten-fold cross-validation (CV) was used for hyperparameter tuning, model training, and testing. Top predictors were validated using cross-validated linear regression. Mean coefficient of determination (CV-R2) and root mean squared error (CV-RMSE) indicated model performance.
Analyses included 59 patients with data from 298 visits. Mean age was 7.8 years and 75.0% of patients were White/European. Mean 6MWD was 321.7m (median, 336.5m; SD, 125.6m). ORT had the best model fit (CV-R2, 0.85 [SD, 0.06]; CV-RMSE, 45.4m [SD, 6.2m]), followed by random forests (CV-R2, 0.83 [SD, 0.05]; CV-RMSE, 49.7m [SD, 8.5m]). The most important predictors were 10-meter walk/run, 4-stair climb, rise from floor, and North Star Ambulatory Assessment total score. A linear model with functional timed tests, demographics, and disease and treatment characteristics performed nearly as well (CV-R2, 0.79 [SD, 0.07]; CV-RMSE, 54.7m [SD, 8.8m]).
A 6MWD proxy model was developed using machine learning. Functional timed tests and ambulatory assessments were the strongest 6MWD proxies and could be used to assess outcomes in clinical practice.
Authors/Disclosures
Nicolae Done, PhD (Analysis Group)
PRESENTER 		Dr. Done has received personal compensation for serving as an employee of Analysis Group, Inc.
No disclosure on file
No disclosure on file
No disclosure on file
Erik K. Henricson, MPH (University of California Davis) No disclosure on file
No disclosure on file
Craig McDonald, MD (UC Davis Dept. of PM&R) Dr. McDonald has received personal compensation in the range of $50,000-$99,999 for serving as a Consultant for Sarepta Therapeutics. Dr. McDonald has received personal compensation in the range of $10,000-$49,999 for serving as a Consultant for PTC Therapeutics. Dr. McDonald has received personal compensation in the range of $10,000-$49,999 for serving as a Consultant for Solid Biosciences. Dr. McDonald has received personal compensation in the range of $500-$4,999 for serving as a Consultant for Roche. Dr. McDonald has received personal compensation in the range of $10,000-$49,999 for serving on a Scientific Advisory or Data Safety Monitoring board for Sarepta Therapeutics. Dr. McDonald has received personal compensation in the range of $10,000-$49,999 for serving on a Scientific Advisory or Data Safety Monitoring board for Solid Biosciences. Dr. McDonald has received personal compensation in the range of $10,000-$49,999 for serving on a Scientific Advisory or Data Safety Monitoring board for Edgewise Therapeutics. The institution of Dr. McDonald has received research support from Sarepta Therapeutics. The institution of Dr. McDonald has received research support from PTC Therapeutics. The institution of Dr. McDonald has received research support from Edgewise Therapeutics. The institution of Dr. McDonald has received research support from Capricor Therapeutics. The institution of Dr. McDonald has received research support from Italfarmaco. Dr. McDonald has received research support from NS Pharma. The institution of Dr. McDonald has received research support from NIH (NINDS). The institution of Dr. McDonald has received research support from Parent Project Muscular Dystrophy. The institution of Dr. McDonald has received research support from Muscular Dystrophy Association. Dr. McDonald has received personal compensation in the range of $500-$4,999 for serving as a Member National Advisory Board for Medical Rehabilitation Research with NIH.