Revisiting the FAITH Trial: A Secondary Analysis Yielding Novel Insights with the Win Ratio

Abstract

Background: Many orthopaedic trials use any unplanned reoperation as the primary outcome, but this overlooks how patients experience those outcomes. Using a high-quality hip fracture trial, we demonstrate how the relative importance of multiple patient-important outcomes can be effectively incorporated into data analysis, providing a more comprehensive understanding of treatment impact.

Methods: This secondary analysis of the Fixation using Alternative Implants for the Treatment of Hip Fracture (FAITH) trial included 1,079 patients aged 50 years or older with a low-energy femoral neck fracture who were randomly assigned to treatment with a sliding hip screw or cancellous screws. The original trial used unplanned revision surgery as the primary outcome. Our primary analysis instead used a composite outcome of all-cause mortality at 4 months, ambulation status at 10 weeks (measured by the EuroQol-5 Dimension [EQ-5D] mobility dimension), and days at home within 4 months. We assessed outcomes hierarchically using the win ratio method, comparing each patient with every other patient in the alternative treatment group in a pairwise manner. We conducted sensitivity analyses at 6 and 12 months, and subgroup analyses to explore smoking status and fracture displacement as potential effect modifiers.

Results: Of the 1,079 participants, 741 had EQ-5D data available for the primary analysis at 4 months, yielding 137,114 pairwise comparisons. A sliding hip screw was superior to cancellous screws in 65,158 (47.5%) comparisons, inferior to cancellous screws in 63,378 (46.2%) comparisons, and tied in 8,578 (6.3%), leading to a win ratio of 1.03 (95% confidence interval [CI] 0.86-1.23), but this difference was not statistically significant (p = 0.76). The sensitivity analysis results were similar at 6 and 12 months. In the subgroup analysis, a sliding hip screw was superior to cancellous screws in current smokers, with a win ratio of 1.65 (95% CI 1.02-2.65) at 6 months (p = 0.007).

Conclusion: This analysis approach should be considered for future orthopaedic trials as it was consistent with the FAITH primary analysis findings but yielded a more nuanced interpretation of the patients' experience and offers deeper insights into intervention effectiveness. The bounds of the 95% CI for the primary outcome were within many standard definitions of equivalence, suggesting surgeons can assume similar patient-important outcomes with either treatment.

Fig. 1

Primary analysis—win ratio results at 4 months.

Fig. 2

Sensitivity analysis—win ratio results at 6 months.

Fig. 3

Sensitivity analysis—win ratio results at 12 months.

Fig. 4

Subgroup analyses.