The Modified Frailty Index Predicts Failure-to-Rescue Mortality After Pancreatoduodenectomy: A Comparison of the 11-Item and 5-Item Versions.

Background

Frailty may influence outcomes after pancreatoduodenectomy (PD). We compared the prognostic performance of the 11-item modified Frailty Index (mFI-11) and the simplified 5-item version (mFI-5) and explored their link with post-pancreatectomy mortality (PPM).

Methods

We retrospectively analyzed 1926 adults undergoing PD for periampullary lesions (2015-2023). Frailty was classified as non-frail, mildly frail, or severely frail. Co-primary outcomes were major complications, defined as Clavien-Dindo grade ≥ III, and 90-day mortality. Separate adjusted logistic models were fitted for each index; Firth regression was used for mortality sensitivity analyses. Discrimination was assessed by AUC and compared using DeLong tests. Deaths were independently adjudicated as PPM1, PPM2, or PPM3.

Results

Major complications occurred in 350 patients (18.2%), and 90-day mortality was 52 (2.7%). Both indices showed a stepwise increase in event rates across frailty strata. In adjusted models, severe frailty was independently associated with mortality (mFI-11: OR 4.1, 95% CI 1.6-10.3, p < 0.001; mFI-5: OR 2.8, 95% CI 1.4-5.6, p = 0.015 by Firth regression). Discrimination was modest and comparable between indices for major complications (AUC 0.629 vs. 0.630; DeLong p = 0.68) and good and comparable for mortality (AUC 0.761 vs. 0.757; DeLong p = 0.48). Severe frailty was independently associated with PPM2 deaths for both indices but not with PPM1 or PPM3, although the latter subgroups were underpowered.

Conclusion

The mFI-5 performs similarly to the mFI-11 after PD and is preferable for routine use due to its simplicity. Frailty is more strongly linked to mortality, especially pancreas-specific failure-to-rescue trajectories, than to major morbidity alone.