The fragility index reveals classification stability that p values conceal, but neither measure whether an observed result is geometrically separated from no effect, the point of therapeutic neutrality. Recent editorials in major journals have cataloged this limitation with unusual clarity, identifying sample-size dependence, p value correlation, restriction to dichotomous outcomes, and inability to measure effect strength as fundamental deficiencies. This commentary argues that these deficiencies point to a structural gap requiring an orthogonal robustness dimension — one that quantifies distance from therapeutic neutrality, where therapies have no effect on outcomes, independent of significance classification. The p–fr–nb framework addresses this gap by defining complete statistical evidence as a triplet: significance (p), fragility (fr), and robustness (nb). Empirical validation across 100 pharmaceutical trials demonstrates that half of all trials show discordance between p value classification and complete evidence assessment, confirming that reporting p-values alone yields systematically incomplete evidence.
Thomas F Heston (Wed,) studied this question.