Requirement: data set with a dichotomous outcome and a p value < 0.05
Derivation of the fragility index:
(1) Redetermine the p value after changing one "positive" result to a "negative" one.
(2) Repeat until the p value is > 0.05.
(3) The fragility index is the number of results that were changed.
Interpretation:
• A small fragility index says that only a few results needed to be different to get a different p value.
Limitations and observations:
(1) The fragility index depends on the p value, which can have limited significance.
(2) In some studies the underlying problem is poor experimental design with the studies being underpowered.
(3) The number of results lost to follow-up become significant if the fragility index is small, since the missing results could have impacted the p value if they had been included.