Initial empiric carbapenem versus non-carbapenem β-lactam/β-lactamase inhibitor therapy in sepsis across simplified SENECA phenotypes: a retrospective study of two independent cohorts.

PURPOSE: Whether initial empiric carbapenem therapy is associated with better outcomes than non-carbapenem β-lactam/β-lactamase inhibitor (BL/BLI) therapy in sepsis remains uncertain. We examined this association overall and across simplified Sepsis Endotyping in Emergency Care (SENECA) phenotypes.

METHODS

This retrospective study included 4,091 patients from MIMIC-IV and 405 patients from a single-center hospital cohort. Initial therapy was defined by the first eligible carbapenem or BL/BLI regimen initiated within 24 h of admission. The primary outcome was 28-day mortality. Secondary outcomes included selected clinical and in-hospital microbiological detection outcomes. Each cohort was analyzed separately using inverse probability of treatment weighting and weighted logistic regression.

RESULTS

After weighting, no overall association with lower 28-day mortality was observed in MIMIC-IV (OR 0.91, 95% CI 0.71-1.17) or the hospital cohort (OR 0.74, 95% CI 0.48-1.16). However, evidence of treatment-by-phenotype interaction was observed in both cohorts (P for interaction = 0.039 and 0.019, respectively), with lower odds of 28-day mortality among patients with the γ phenotype in MIMIC-IV and the hospital cohort (OR 0.66, 95% CI 0.44-0.97, and OR 0.13, 95% CI 0.04-0.46, respectively). A similar direction was observed among mechanically ventilated patients. In MIMIC-IV, carbapenem therapy was associated with longer hospitalization and more frequent detection of fungal organisms and Clostridioides difficile; these associations were not observed in the hospital cohort.

CONCLUSION

Initial empiric carbapenem therapy was not significantly associated with lower 28-day mortality overall. Consistent findings in the γ phenotype suggest possible treatment heterogeneity. These findings should be considered alongside patient characteristics and potential clinical and microbiological trade-offs.