Abstract
Abstract
Background
Hepatic failure carries a high mortality and most patients in the intensive care unit (ICU) require continuous sedation. The commonly used agents differ in their pharmacokinetics in liver dysfunction, but the optimal sedative strategy is undefined. We compared midazolam and dexmedetomidine with propofol for 28-day mortality in critically ill patients with hepatic failure.
Methods
MIMIC-IV v3.1 (2008–2022) was used to identify adult ICU patients with hepatic failure (ICD-9 570 / ICD-10 K72) who received a continuous infusion of midazolam, propofol or dexmedetomidine for at least 24 h; the agent with the longest cumulative infusion was the primary sedative. The primary outcome was 28-day all-cause mortality, analysed with Kaplan-Meier estimation, log-rank tests, multivariable logistic regression (crude; Model 1, age and sex; Model 2, Model 1 plus ventilation hours, vasopressor use, sepsis, acute kidney injury, hepatic coma and MELD), Cox models and inverse-probability-of-treatment weighting (IPTW). Secondary outcomes were ventilation hours, ICU and hospital length of stay and hepatic encephalopathy with coma. Findings were validated in eICU-CRD.
Results
We analysed 2358 ICU stays (2038 patients; 2110 admissions): midazolam 593, propofol 1485, dexmedetomidine 280. Overall 28-day mortality was 51.5%. Compared with propofol, the Cox-adjusted hazard ratio (HR) was 1.48 (95% CI 1.30–1.69; P < 0.001) for midazolam and 0.56 (0.45–0.69; P < 0.001) for dexmedetomidine (log-rank P < 0.001; C-index 0.69). Logistic regression was concordant: odds ratios (OR) 1.55 and 0.65 adjusted for age and sex, 1.52 and 0.60 after full adjustment and 1.53 and 0.62 with IPTW. Dexmedetomidine recipients had longer ICU (8.1 vs 5.6 days) and hospital (19.9 vs 16.0 days) stays and longer ventilation (87.2 vs 72.3 and 63.2 h; P = 0.0002). In eICU-CRD (257 stays) the midazolam association was replicated (adjusted OR 1.62, 1.06–2.42; P = 0.040) whereas that for dexmedetomidine was not (OR 0.90, 0.53–1.43; P = 0.647). Findings were consistent across subgroups and sensitivity analyses and were not mediated by mechanical-ventilation duration.
Conclusions
After multivariable adjustment and inverse-probability weighting, midazolam was associated with higher 28-day mortality than propofol in hepatic failure, and dexmedetomidine with lower mortality, with a concordant midazolam association externally. Randomised trials are warranted.
Trial registration
Not applicable (retrospective analysis of deidentified public databases no intervention).