Reframing the fragility index as bias analysis: sensitivity analysis, P-values, parameterizations, and confidence...

The fragility index (FI) is generally presented as a metric to assess how sensitive the passing of a particular P-value threshold is to the number of changes in event status. In many cases, the "statistical significance" of a result may be quite fragile. An alternative interpretation of the FI is proposed that reframes it as a sensitivity analysis or bias analysis for particular forms of misclassification. The reframing also makes clear the comparative relevance of different analogues of the FI for survival data, such as that proposed by Xing, et al. (Am J Epidemiol. 0000;000(00):0000-0000) and others, and further moves away from more dichotomous interpretations concerning P-value thresholds.