Sepsis Antibiotic Trial Meets Four of Five Feasibility Criteria

Evidence

The EMPRESS sepsis antibiotic trial met four of five prespecified feasibility criteria—practical checks on whether the study could run as planned. Investigators judged that the trial could continue without modifications, although outcome-data timeliness fell short of its overall target.

The investigator-initiated trial randomized 200 of 284 screened patients across 10 intensive care units in Denmark. Participants were critically ill adults with sepsis, a life-threatening response to infection. The study compared meropenem with piperacillin/tazobactam as empirical antibiotics: treatment selected before the infection’s cause is definitively established. It was open-label, meaning treatment assignments were not masked, and adaptive, meaning accumulating data could guide trial decisions.

Investigators evaluated five feasibility criteria after participants completed 30-day follow-up. Primary outcome data were available within 45 days for 85.5% of participants, below the prespecified threshold of at least 95.0%. That measure improved to 95.8% in the final month of the feasibility phase.

Across the full cohort, mortality from any cause at 30 days was 30.5%, and specified serious adverse reactions occurred in 4.0% of participants. These are overall figures, not comparisons between antibiotics. The EMPRESS sepsis trial feasibility report was published October 1, 2026.

Analysis

Interpretation: The strongest finding concerns trial operations, not treatment effectiveness. Recruitment, random assignment and outcome collection were implemented in this multicentre setting, and four satisfied criteria supported the investigators’ decision to proceed.

The final-month improvement matters because adaptive trial decisions depend on timely data. But one month above the threshold does not erase the overall shortfall or establish that improvement will persist. Likewise, successful trial delivery is distinct from demonstrating a clinically meaningful benefit for either antibiotic.

Limitations

This is a single-study feasibility report, not corroboration across independent studies. It does not establish superiority or equivalence between the antibiotics. The reported findings provide neither between-group treatment-effect estimates nor numerical confidence intervals for those effects.

The open-label design leaves less protection against bias related to knowledge of treatment assignment. Experience in 10 Danish intensive care units also does not establish that recruitment and data collection would perform equally well elsewhere. Overall mortality and adverse-reaction rates cannot identify which regimen produced better outcomes.

What to watch

For treatment effectiveness, watch for randomized, between-group outcome estimates with uncertainty, rather than pooled event rates. For operational feasibility, confidence would increase if the proportion of primary outcome data available within 45 days remained at or above 95.0% over subsequent months, rather than only the final feasibility month.