Abstract
OBJECTIVES: Sepsis is a highly heterogeneous syndrome. This study aims to identify potential phenotypes of sepsis based on commonly used clinical inflammatory parameters and to investigate the impact of vitamin C infusion therapy on survival outcomes in patients with different phenotypes. METHODS: Patients with sepsis in the Intensive Care Medical Information Database IV were used as the study population, and the data were randomly divided into two groups of 7:3, which were used for derivation and validation, respectively. In the derivation cohort, the K-means clustering was employed to identify potential phenotypes based on three key feature parameters: white blood cell count (WBC), neutrophil-to-lymphocyte ratio (NLR), and temperature, and then assess the effect of infused vitamin C treatment on survival outcomes of each phenotype by Kaplan-Meier survival analysis and multifactorial Cox proportional risk model. The same analyses were used in the validation cohort to verify the clustering effect and the heterogeneity of vitamin C efficacy on each phenotype. RESULTS: A total of 3 771 patients were included in the final analysis, 2 639 in the derivation cohort and 1 132 in the validation cohort. In the derivation cohort, patients with sepsis were categorized into three phenotypes: cluster A (n=471, 17.8%) was characterized by moderately elevated white blood cells (WBC), neutrophil-to-lymphocyte ratio (NLR), and hyperthermia; cluster B (n=1 812, 68.7%) was characterized by mildly elevated WBC, NLR, and essentially normal body temperature, and patients with this type had the relatively best clinical status and survival prognosis, and cluster C (n=356, 13.5%) was characterized by extreme elevation of WBC and NLR, and insignificant elevation of body temperature, and patients with this type had the most severe organ dysfunction and high mortality. Significant differences in age, vital signs, history of comorbidities, laboratory tests, pathophysiological scores and clinical prognosis were found between phenotypes and showed different responsiveness to vitamin C treatment (all P<0.05). A Kaplan-Meiersurvival analysis showed that intravenous vitamin C infusion was associated with a lower risk of death at 28 days only in patients with cluster A [hazard ratio (HR)=0.388, 95% confidence interval (CI) 0.166 to 0.906, P=0.023], and this protective effect remained statistically significant after correction for confounders by multivariate Cox regression (HR=0.353, 95% CI 0.165 to 0.755, P=0.007), whereas a beneficial effect of vitamin C on survival outcomes was not detected in the other phenotypes. This heterogeneous response yielded similar results in the validation cohort. CONCLUSIONS: This study identified three distinct sepsis phenotypes with different clinical characteristics based on three key parameters: WBC, NLR, and temperature. Each phenotype exhibits varying responsiveness to vitamin C therapy, with patients with type A sepsis benefiting from intravenous vitamin C treatment.