[Association between fluid balance and clinical outcomes in sepsis patients with identified subtypes]

Objective: To identify clinical subphenotypes of sepsis by clinical variables, and the association between fluid balance (FB) and outcomes in sepsis patients with different subphenotypes. Methods: Data were derived from China Critical Care Sepsis Trial. A total of 2 075 patients were included in the study, with a mean age of (62.5±17.4) years, 1 362 (65.6%) were male, 713 (34.3%) were female. In-hospital mortality was the primary outcome. Clinical variables were entered into the latent profile analysis (LPA) model which was used to identify the subphenotypes of sepsis. Multivariable logistic regression model was used to explore the association of FB on the first 2 days admitted to ICU and in-hospital mortality. Results: Total of 687 (33.1%) patients died in the hospital. Four subphenotypes of sepsis were identified by the PLA model: 1 421 (68.5%) patients were identified as profile 1, which was characterized by the lowest in-hospital mortality (28.3%, 402/1 421) and was considered as the baseline subphenotype; profile 2 was characterized by multiple organ dysfunction(232, 11.2%); profile 3 was characterized by respiratory dysfunction (196, 9.4%) and profile 4 was characterized by kidney injury/failure (226, 10.9%). Profile 2 showed the highest in-hospital mortality (54.3%, 126/232), followed by profile 4 (39.4%, 89/226), profile 3 (35.7%, 70/196). Profile 2 had the highest FB of 68.5(34.7, 89.4)ml/kg on day 1 and decreased rapidly to 29.4(13.6, 49.8) ml/kg on day 2. Multivariable analysis showed that for every 10 ml/kg of FB increased on the first 48 hours admitted to ICU, the risk of in-hospital mortality decreased 14% (OR=0.86, 95%CI: 0.81-0.97) in profile 2, and increased 4% (OR=1.04, 95%CI:1.02-1.07) in profile 4. However, there were no significant association between FB on the fist 2 days admitted to ICU and the risk of in-hospital mortality in profile 1 (OR=1.06, 95%CI: 0.94-1.19)and in profile 3 (OR=0.99, 95%CI: 0.94-1.05). Conclusions: Four subphenotypes of sepsis were identified according to the clinical variables, which showed different characteristics, distributions, outcomes and responses to fluid balance.