Text

Objectives and questions: Observational studies are crucial in orthopedic trauma surgery when randomized controlled trials (RCTs) are not feasible. However, they are susceptible to confounding by indication, as treatment decisions are often influenced by patient characteristics that also affect outcomes. While statistical adjustments are frequently used, they cannot account for unmeasured confounders. Clinical equipoise—defined as genuine uncertainty among experts regarding the optimal treatment—has been proposed as an inclusion criterion to improve baseline comparability in observational studies. This study investigates whether higher levels of expert disagreement, as a proxy for clinical equipoise, are associated with improved baseline comparability between treatment groups in patients with proximal humerus fractures.

Material and methods: We used data from the LADON Humerus study, a multicenter, prospective cohort including patients treated operatively or non-operatively in five hospitals in the Netherlands and Switzerland. An international expert panel of trauma surgeons assessed anonymized cases, blinded to the actual treatment, and recommended a preferred approach. Disagreement between the treatment given and the expert recommendations was categorized into four levels (A: complete disagreement, B: partial, C: minimal, D: complete agreement). Standardized differences (StDiff) for age, sex, ASA score, and AO classification were used to assess comparability between treatment groups across disagreement levels.

Results: A total of 745 patients were evaluated. Comparability between treatment groups improved as the level of disagreement increased. At complete disagreement (Level A), the groups were well balanced in terms of age, sex, and ASA score (StDiff ≤ 0.1). AO classification showed less improvement. Narrower inclusion criteria led to better balance but smaller sample sizes.

Discussion and conclusions: Using expert panel disagreement as a clinical equipoise criterion can enhance group comparability in observational research. This approach offers a feasible alternative to randomization and may reduce confounding when RCTs are not possible.