Toward Inclusive Trial Protocols in Heterogeneous Neurological Disorders: Prediction-Based Stratification of...

Area of Science:

• Neuroscience
• Clinical Trials
• Biostatistics

Background:

• Novel therapies for spinal cord injury (SCI) are emerging, necessitating efficient clinical trials.
• Heterogeneity in incomplete SCI (iSCI) recovery complicates participant enrollment and stratification.
• Optimizing trial design is crucial for evaluating new SCI treatments effectively.

Purpose of the Study:

• To enhance SCI trial design by maximizing eligible participant enrollment.
• To develop a method for stratifying heterogeneous iSCI populations into homogeneous cohorts.
• To create prediction-based decision rules for iSCI participant inclusion in trials.

Main Methods:

• Retrospective, longitudinal analysis of prospectively collected data from the European Multicenter study about Spinal Cord Injury (EMSCI).
• Application of conditional inference trees for prediction-based stratification.
• Validation of the algorithm internally and externally using clinical endpoints.

Main Results:

• Conditional inference trees successfully partitioned iSCI participants into more homogeneous groups based on baseline data and clinical endpoints.
• The upper extremity motor score was used as an illustrative endpoint for stratification.
• The developed algorithm demonstrated stable and generalizable results upon validation.

Conclusions:

• Conditional inference trees offer a feasible approach for stratifying iSCI participants in clinical trials.
• The algorithm provides easily implementable, prediction-based decision rules for trial inclusion and stratification.
• This method can be adapted to model various trial endpoints and outcome thresholds for SCI research.