A measure of how fragile a trial’s result is.
It describes the number of non-events that need to become events in order to render a trial result non-significant (i.e. p>0.05). P-values are calculated with Fisher’s Exact Test.
For example, in a study with 100 patients in the placebo arm of which 20 die, and 100 in the intervention arm of which 12 die, we ask what might be the new p-value if the death rates are changed to 20/100 and 13/100. If the p-value remains significant (<0.05), we repeat the process adding one death with each iteration until p>0.05. The number added is reported as the Fragility Index.
In other words, a lower Fragility Index indicates less statistically robust results.
FI can be applied to any trial reporting a binary outcome of any kind. This can be death, MI, intubation or anything for which the question can be asked “did this event occur in my patient?” and a yes or no answer be given.
A trial of this type will often report the number of outcomes (events) in the control/placebo group, and then compare this to the number of outcomes (events) in the intervention group.
Enter the number of events that occurred in your trial’s control group, the number of events that occurred in your trial’s intervention group, and the total number of patients in each arm of the trial. Hit submit, and the calculator will give you the trial’s FI.
For some trials, simply changing from Chi-squared analysis to Fisher’s Exact Test will change the p-value to greater than 0.05. This is equivalent to FI=0.