Fragility in cardiovascular randomized controlled trials with primary continuous outcomes (2018-2022) from...

This study aims to assess the robustness of cardiovascular disease randomized controlled trials (RCTs) with primary continuous outcomes from a clinical perspective, utilizing the concepts of continuous fragility index (CFI), reverse continuous fragility index (RCFI) and their corresponding quotients (CFQ, RCFQ).

A cross-sectional study was conducted, searching PubMed for cardiovascular RCTs published between January 1, 2018, to December 31, 2022, in eight high-impact journals. Inclusion criteria were phase III or IV trials with 1:1 randomization, reporting at least one primary continuous outcome. Data analysis involved altering each outcome until achieving the reversal of significance (ɑ = 0.05) to determine the CFI or RCFI. The fragility quotients were then calculated by dividing the CFI or RCFI by the sample size, and Spearman's correlation assessed correlation analyses.

Of 3983 records were screened, and 64 RCTs (76 outcomes) were included. The fragility index was analysed with 72 outcomes. The overall median CFI was 7, with an associated median CFQ of 0.032. Nonsignificant P values exhibited greater statistical instability (median RCFI = 5, RCFQ = 0.023) than significant P values (median CFI = 14, CFQ = 0.062). Interestingly, "fragile" values were found in 36% (9/25) of CFI or 46.7% (7/15) of RCFI. Additionally, fragility index showed a significant association with several variables.

The findings suggest that changing only a small number of interventions (median of 7) could alter outcome significance. Reporting the fragility index alongside P values is recommended to provide a clearer understanding of statistical findings' robustness.HighlightsThe continuous fragility index (CFI) represents the minimum patient count needed to modify significance by altering their intervention.Among 72 primary continuous outcomes in 64 cardiovascular RCTs, the overall median CFI was 7, with a corresponding CFQ of 0.032.CFI demonstrated moderate to strong correlations with sample size, total dropouts, and patient numbers analyzed.