Stop-It Randomized Clinical Trial (RCT) for Intra-Abdominal Infection (IAI) Revisited: Multivariate Analyses To...

The STOP-IT randomized clinical trial (RCT) pioneered limiting antibiotic agents in intra-abdominal infection (IAI) with adequate surgical source control, but NIH funding ended before an adequate power sample size was enrolled to determine equivalence between STOP-IT study regimens: four days of antibiotic agents (4-days) after source control versus antibiotic agents until resolution of signs and symptoms of IAI plus two days (standard of care [SOC]). The objective of this investigation was to identify possible significant treatment effects 4-days versus SOC, and independent variables defining and predicting outcomes. De-identified data from 518 STOP-IT subjects were analyzed retrospectively in two groups: 4-days (n = 258) and SOC (n = 260), and separately as one group (n = 518). Statistics: multivariate regression analysis, chi-squared, and simple Cohen kappa coefficient. No pre-randomization variable predicted protocol FAILURE (surgical site infection, recurrent IAI, or death at 30 d) in 4-day subjects. APACHE II predicted SOC FAILURE, but no cut point determined treatment effect (AUC = 0.608). Both observations implied that FAILURE may not reflect patient outcomes. Additionally, Cohen kappa for FAILURE and hospitalization at 7, 14, and 21 days was weak (0.1154, 0.2084, and 0.1969, respectively) with high numbers of discordant values. Pre-randomization variables associated with hospitalization/discharge at days 7, 14, and 21: extra-abdominal infection 1 (p < 0.0001), APACHE II score (p < 0.0001), age (p = 0.006), and WBC maximum (p < 0.05). However, all of these pre-randomization variables did not predict FAILURE, except APACHE II. Poor Cohen kappa coefficients indicate STOP-IT FAILURE agreed only weakly with hospital/discharge at 7, 14, or 21 days, and is not a valid reliable endpoint in IAI or for determining success or failure of any treatment. Pre-randomization extra-abdominal infection, APACHE II score, age, and WBC maximum strongly predicted hospitalization, but only APACHE II predicted failure. The study should use the appropriate sample size calculation when doing an equivalence on the basis of the Two One-Sided Test design. RCTs in IAI need prospectively validated clinically reliable endpoints that align with known patient outcomes that determine success.