Using genome-wide scans, we sought to identify serum miRNAs that could be used as prognostic predictors for sepsis patients.
We used microarray screens to identify differentially expressed serum miRNAs by comparing samples from 12 surviving and 12 nonsurviving sepsis patients.
These differentially expressed serum miRNAs were validated by quantitative reverse transcriptase-polymerase chain reaction assays for 118 sepsis patients.
The validated miRNAs along with sepsis patients' clinical indictors were analyzed in a multivariate logistic regression model.
Microarray analysis showed that miR-297 and miR-574-5p were differentially expressed in sepsis survivors and nonsurvivors.
Upon validation with 118 sepsis patients' samples, these two miRNA expressions were significantly different, with P < 0.001. miR-297 was more closely associated with survival from sepsis, whereas miR-574-5p was associated with death from sepsis.
Multivariable logistic regression analysis showed that a combination of sepsis stage, Sepsis-Related Organ Failure Assessment scores, and miR-574-5p was correlated with the death of sepsis patients.
The predictive capability of these three combined variables was analyzed by a receiver operating characteristic curve; the area under the curve was 0.932 (95% confidence interval, 0.887-0.977). When the cutoff point was set at 0.288, these three combined variables provided 78.13% sensitivity and 91.84% specificity.
Our results showed that serum miR-574-5p was correlated with the death of sepsis patients.
The combined predictive capability of sepsis stage, Sepsis-Related Organ Failure Assessment scores, and serum miR-574-5p for the death of sepsis patients was better than any single indicator.