Objectives: Patients included in randomized controlled trials (RCT) are poorly representative of the general population. We compared outcomes of patients excluded from the PIRATE trial, a point-of-care RCT evaluating antibiotic durations for Gram-negative bacteraemia, with those of enrolled patients.

Methods: A prospective observational cohort study, 'EPCO' (Excluded Patients' Clinical Outcomes) included patients excluded from the PIRATE trial. As in PIRATE, whose patients were randomized to 7-day, 14-day, or C-reactive-protein (CRP)-guided antibiotic durations, EPCO's primary outcome was occurrence of clinical success at 30 days. We also compared baseline characteristics, outcome rates and treatment-effect estimates.

Results: In all, 405 patients were included in EPCO and compared with the 503 PIRATE patients. Reasons for exclusion were mainly medical (317/405; 78%), the most frequent being complicated infection. Excluded patients had more co-morbidities (Charlson median 3 versus 1, p < 0.001). Bacteraemia was more often health-care-associated (26% versus 9%, p < 0.001). The 30-day success rate was significantly lower among EPCO patients (299/396; 76% versus 469/493; 95%, p < 0.001), but the success rate was not significantly different for those excluded for non-medical reasons (68/75; 91%, p 0.09). There was no significant difference in failure rates of EPCO patients according to their treatment duration (difference 7 days versus 14 days: p 0.75; 7 days versus CRP-correspondent: p 1.00; 14 days versus CRP-correspondent: p 1.00).

Conclusion: Although point-of-care-randomized trials are more inclusive and representative than traditional RCTs, they are still likely to select patients with lower failure risk. Shorter antibiotic durations were not associated with failure in either included or excluded patients, supporting the generalizability of the PIRATE trial's findings.