Comments on “How Fragile Are the Results of a Trial? The Fragility Index” by Dettori and Norvell

To the Editor,

It is laudable that Dettori and Norvell ¹ reviewed the “fragility index” in their recent article. This novel metric is rapidly gaining popularity as an intuitive measure of the robustness of statistically significant results from clinical trials that use dichotomous outcomes. Their discussion is both insightful and educational, and their attention to underlying concepts is commendable.

Nonetheless, it should be noted that calculation of the fragility index as presented in the authors’ Table 1 deviates slightly from the original description reported by Walsh et al, ² which may confuse some readers. Rather than subtracting events from the arm with the greatest number of events until the first time the calculated P value becomes equal to or greater than .05, one should iteratively add events (ie, “change from a nonevent to an event” ^2,3 ) to the arm with the smaller number of events while subtracting nonevents from the same arm to keep the total number of patients constant. In the described example, this would manifest simply as increasing the event rate and reducing the nonevent rate in the iliac crest bone graft arm from 151 and 18, respectively, to 152 and 17, then 153 and 16, and so on, while maintaining 186 and 8 in the rhBMP-2 arm.

This clarification of course yields the same result as was presented, but consistent descriptions and methodology may aid the broad group of spine clinicians, researchers, and other evidence users who become interested in the fragility index as awareness increases.