Abstract Introduction Benzodiazepines and α2 adrenoceptor agonists exert opposing effects on innate immunity and mortality in animal models of infection. We hypothesized that sedation with dexmedetomidine (an α2 adrenoceptor agonist), as compared with lorazepam (a benzodiazepine), would provide greater improvements in clinical outcomes among septic patients than among non-septic patients. Methods In this a priori-determined subgroup analysis of septic vs non-septic patients from the MENDS double-blind randomized controlled trial, adult medical/surgical mechanically ventilated patients were randomized to receive dexmedetomidine-based or lorazepam-based sedation for up to 5 days. Delirium and other clinical outcomes were analyzed comparing sedation groups, adjusting for clinically relevant covariates as well as assessing interactions between sedation group and sepsis. Results Of the 103 patients randomized, 63 (31 dexmedetomidine; 32 lorazepam) were admitted with sepsis and 40 (21 dexmedetomidine; 19 lorazepam) without sepsis. Baseline characteristics were similar between treatment groups for both septic and non-septic patients. Compared with septic patients who received lorazepam, the dexmedetomidine septic patients had 3.2 more delirium/coma-free days (DCFD) on average (95% CI for difference, 1.1 to 4.9), 1.5 (-0.1, 2.8) more delirium-free days (DFD) and 6 (0.3, 11.1) more ventilator-free days (VFD). The beneficial effects of dexmedetomidine were more pronounced in septic patients than in non-septic patients for both DCFDs and VFDs (P-value for interaction = 0.09 and 0.02 respectively). Additionally, sedation with dexmedetomidine, compared with lorazepam, reduced the daily risk of delirium [OR, CI 0.3 (0.1, 0.7)] in both septic and non-septic patients (P-value for interaction = 0.94). Risk of dying at 28 days was reduced by 70% [hazard ratio 0.3 (0.1, 0.9)] in dexmedetomidine patients with sepsis as compared to the lorazepam patients; this reduction in death was not seen in non-septic patients (P-value for interaction = 0.11). Conclusions In this subgroup analysis, septic patients receiving dexmedetomidine had more days free of brain dysfunction and mechanical ventilation and were less likely to die than those that received a lorazepam-based sedation regimen. These results were more pronounced in septic patients than in non-septic patients. Prospective clinical studies and further preclinical mechanistic studies are needed to confirm these results. Trial Registration NCT00095251.