(AK) An Innovative Way to Generate Blinded Data in Single Arm Clinical Trials

Publication Number: 1035

Background: Clinical trials of gene therapies are often unrandomized due to limited sample size, lack of effective treatment and ethical considerations for placebo treatment in often debilitating diseases. External control arms (ECAs) can be used to provide context on disease progression. Critical to applying an ECA strategy is comparability of the population, outcomes, procedures and assessment timepoints between the trial and ECA data source, but in most situations, analyses of the trial and ECA cannot be blinded.

Objectives: To review development and pre-testing of methods to minimize rater bias and generate blinded data for single-arm trials, despite the lack of an internal comparator.

Methods: We developed an innovative process to demonstrate effectiveness of a gene therapy in a trial conducted in children with significant development delays and mortality with disease progression. The primary endpoint is assessed by clinicians in the trial and will be assessed similarly in a natural history study (NHS); the assessment could not be blinded. However, interviews with experts and caregivers/parents (CG/P) as well as review of existing measures identified observable activities that were meaningful in child development as well as to CG/Ps and could be video recorded. Following development of standardized materials to instruct CG/Ps to record their child attempting specific activities (e.g., communication, eye movement, head control, limb movement, walking, block play), a centralized 3-expert panel reviewed videos from a pre-test of materials and video recording procedures prior to implementation in the two studies.

Results: CG/Ps showed good understanding of pre-test of materials and procedures, having minor suggestions. Testing the expert review process with pre-test videos showed mostly good agreement among the three experts (agreement: 80% to 100%) for gross motor skills and lower agreement for more subtle aspects and reactive communications. For the clinical trial and NHS, the panel will review videos in random order, blinded to timepoint, treatment (i.e., study) and dose cohort, with quality control videos randomly intermixed to monitor longitudinal drift within raters.

Conclusions: This novel approach reflects an innovative way to minimize rater bias and generate blinded, more rigorous data in single-arm trials to show improvement and progression in treated patients in a gene therapy trial and untreated NHS patients.