Use of imaging biomarkers and ambulatory functional endpoints in Duchenne muscular dystrophy clinical trials: Systematic review and machine learning-driven trend analysis

Abstract

Duchenne muscular dystrophy (DMD) is a rare X-linked genetic muscle disorder affecting primarily pediatric males and leading to limited life expectancy. This systematic review of 85 DMD trials and non-interventional studies (2010-2022) evaluated how magnetic resonance imaging biomarkers-particularly fat fraction and T2 relaxation time-are currently being used to quantitatively track disease progression and how their use compares to traditional mobility-based functional endpoints. Imaging biomarker studies lasted on average 4.50 years, approximately 11 months longer than those using only ambulatory functional endpoints. While 93% of biologic intervention trials (n = 28) included ambulatory functional endpoints, only 13.3% (n = 4) incorporated imaging biomarkers. Small molecule trials and natural history studies were the predominant contributors to imaging biomarker use, each comprising 30.4% of such studies. Small molecule trials used imaging biomarkers more frequently than biologic trials, likely because biologics often target dystrophin, an established surrogate biomarker, while small molecules lack regulatory-approved biomarkers. Notably, following the 2018 FDA guidance finalization, we observed a significant decrease in new trials using imaging biomarkers despite earlier regulatory encouragement. This analysis demonstrates that while imaging biomarkers are increasingly used in natural history studies, their integration into interventional trials remains limited. From XGBoost machine learning analysis, trial duration and start year were the strongest predictors of biomarker usage, with a decline observed following the 2018 FDA guidance. Despite their potential to objectively track disease progression, imaging biomarkers have not yet been widely adopted as primary endpoints in therapeutic trials, likely due to regulatory and logistical challenges. Future research should examine whether standardizing imaging protocols or integrating hybrid endpoint models could bridge the regulatory gap currently limiting biomarker adoption in therapeutic trials.

Keywords: Duchenne muscular dystrophy; and machine learning; clinical trials; functional endpoints; imaging biomarkers; rare diseases; systematic review; trend analysis.

Figure 1.

Data exclusion. Flowchart depicting the selection process for eligible trials retrieved from ClinicalTrials.gov. A total of 108 trials were initially identified, with 13 trials excluded based on criteria such as focus on Becker Muscular Dystrophy, use of pulmonary or cardiac biomarkers as sole imaging markers. After primary screening, 95 trials remained, and an additional 10 trials were excluded for lacking MRI biomarkers or ambulatory functional endpoints, or for including only dystrophin expression without other biomarkers or ambulatory functional endpoints. The final eligible trial count was 85.

Figure 2.

(A) Comparison of endpoint usage across intervention types. Trials were categorized into four groups: non-interventional trials and three types of interventional trials (small molecule drugs, biologic drugs, and non-pharmacological interventions). Imaging biomarkers were relatively low compared to ambulatory functional endpoints within each intervention group. Despite similar absolute numbers of trials using imaging biomarkers across all groups, they constituted a greater percentage in non-pharmacological and non-interventional studies due to fewer total trials in those categories. Some trials studied both imaging biomarkers and ambulatory functional endpoints, leading to the sum of green and blue bars exceeding the white bars on the x-axis. (B) Ongoing studies by year. For each year in our study period, the sum of all trials that had started on that year or before and ended on that year or after were counted here. The total counts each trial individually and does not duplicate data for trials including both imaging biomarkers and ambulatory functional endpoints. (C) Studies by start year. The non-stacked line graph displays higher amounts of trials using ambulatory functional endpoints each year, except for the year 2015, in which use of both measures was even. In 2016, both lines reached their local maxima, with imaging biomarkers declining in the following years to a total of 0 trials in 2018. Trials that included both measures were counted in both categories for their respective start year. (D) Trial durations stratified by intervention type and primary outcome measure. This box-and-whisker plot illustrates the distribution of trial durations, in years, across eight groups categorized by intervention type (Biologic, Small Molecule, Non-Interventional, or Non-Pharmacological) and inclusion or exclusion of imaging biomarkers. Each box represents the interquartile range (IQR), with the horizontal line denoting the median and the “X” marking the mean. Whiskers indicate the range within 1.5 times the IQR; outliers beyond this range are plotted individually. Trials involving biologic interventions paired with imaging outcomes exhibited the widest duration range. Among non-interventional trials, those using imaging outcomes had longer durations than those using functional endpoints, both in terms of median and overall distribution. This visualization highlights variation in trial length based on both therapeutic strategy and endpoint selection.

Figure 3.

A representative decision tree and important features. Decision trees were trained using XGBoost to classify previous clinical trial characteristics —a set including intervention type, start year, phase, trial status, trial duration, and study design— were organized into one of three outcome categories: functional endpoints, imaging biomarkers, or both. By predicting the outcome category based on six input components, this approach aids in guiding the design of future clinical trials. (A) Representative decision tree. Each tree learns decision rules, identifying thresholds for numerical components and splitting criteria for categorical variables. The final prediction is made by aggregating the outputs of all trees, weighted according to their trained importance. (B) Feature importance of clinical trial components derived from trained decision tree.. The relative importance of six clinical trial characteristics was assessed using F-scores, which reflect each feature’s contribution to the predictive performance of the model. The three most influential variables were trials classified as “Non-interventional” in the Phase category, “Non-Pharmacological” in Intervention Type, and “Non-interventional” in Study Design.