Statistical Fragility of Studies Comparing Revision Rates After Manual Versus Robotic-Assisted or Computer-Navigated Total Hip Arthroplasty: A Systematic Review of Randomized Controlled Trials.

BACKGROUND

Despite the growing use of computer-navigated and robotic-assisted total hip arthroplasty (THA), uncertainty remains regarding their outcomes compared with manual THA. Therefore, we implemented the concept of reverse fragility index (RFI) to assess the statistical robustness of available randomized controlled trials (RCTs) designed to study comparative revision rates between manual and technology-assisted THA.

METHODS

The RCTs reporting revision rates for manual and robotic-assisted/computer-navigated THA were included in our systematic review. Our search included all RCTs published until February 2, 2025. There were nine RCTs with 894 patients analyzed. The RFI, defined as the minimum number of event reversals necessary to change a nonsignificant result to statistically significant, was calculated for each study using a threshold of P < 0.05. The reverse fragility quotient (RFQ), which adjusts the RFI relative to the study sample, was additionally calculated.

RESULTS

The median RFI was five (interquartile range [IQR], 4 to 5) for both computer-navigated and robotic-assisted THA subgroups, meaning that a median of five events would be needed to change the results from nonsignificant to significant. The median RFQ was 0.045 (IQR, 0.036 to 0.083), indicating that an event reversal in 4.5% of patients would be sufficient to make the results significant. The median number of patients lost to follow-up was six (IQR, 0 to 41). Of the nine studies, seven (78%) had a loss to follow-up greater than or equal to its RFI.

CONCLUSIONS

We observed significant statistical fragility in all included RCTs that compared revision rates of manual versus technology-assisted THA. We recommend routine reporting of RFI and RFQ metrics in future RCTs with nonsignificant findings as well as careful interpretation of P values.

LEVEL OF EVIDENCE

II.