An endpoint adjudication committee for the assessment of computed tomography scans in fracture healing.

Introduction: Endpoint Adjudication Committees (EACs) benefit the quality of randomized control trials (RCTs) where outcomes depend on subjective interpretations. However, assembling a committee to adjudicate large datasets is cumbersome. In a recent RCT, the primary outcome was time to union following operative fixation of scaphoid non-union, with real or placebo adjunctive ultrasound treatment. Union status was determined with computed tomography (CT) scans interpreted by treating surgeons and radiologists. An EAC was established to deliberate discrepancies between radiologists' and surgeons' interpretations of union status.

Methods: Three hundred sixty-four CT scans from 142 participants were collected in the RCT. The treating surgeon and an MSK radiologist categorized images by percent-union (0 %, 1-24 %, 25-49 %, 50-74 %, 75-99 %, 100 %). Union was defined as at least 50 % trabecular bridging. The EAC adjudicated those images that were deemed major discrepancies. The committee was composed of three members assembled by the committee chair, an MSK radiologist. A charter was established to guide the adjudication process. Ten minutes were allotted to each scan, including 2-3 min of an independent adjudicator's review, followed by 5-7 min of committee discussion to reach a diagnosis.

Results: Adjudicators spent an average of seven minutes on each scan. The EAC assessed 101 CT scans from 69 patients collected across five study sites: four scans from the agreed upon group as practice interpretations, 75 major discrepancies, and 22 missing interpretations from either the initial MSK radiologist, the treating orthopaedic surgeon, or both. These were adjudicated for final union status. Twenty-eight of the images with major discrepancies were adjudicated to union, and 47 to non-union. Adjudication changed the primary outcome of time to union in 40/142 (28 %) of study participants.

Conclusion: This adjudication process provides a valuable research tool for reference by other clinical investigators whose RCTs' outcomes are dependent on interpretation of radiographic images.