Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2018 May 15;7(2):129-135.
doi: 10.1093/jpids/pix021.

Validation of the Sepsis MetaScore for Diagnosis of Neonatal Sepsis

Free PMC article

Validation of the Sepsis MetaScore for Diagnosis of Neonatal Sepsis

Timothy E Sweeney et al. J Pediatric Infect Dis Soc. .
Free PMC article


What’s known on this subject: Neonates are at increased risk for developing sepsis, but this population often exhibits ambiguous clinical signs that complicate the diagnosis of infection. No biomarker has yet shown enough diagnostic accuracy to rule out sepsis at the time of clinical suspicion.

What this study adds: We show that a gene-expression-based signature is an accurate objective measure of the risk of sepsis in a neonate or preterm infant, and it substantially improves diagnostic accuracy over that of commonly used laboratory-based testing. Implementation might decrease inappropriate antibiotic use.

Background: Neonatal sepsis can have devastating consequences, but accurate diagnosis is difficult. As a result, up to 200 neonates with suspected sepsis are treated with empiric antibiotics for every 1 case of microbiologically confirmed sepsis. These unnecessary antibiotics enhance bacterial antibiotic resistance, increase economic costs, and alter gut microbiota composition. We recently reported an 11-gene diagnostic test for sepsis (Sepsis MetaScore) based on host whole-blood gene expression in children and adults, but this test has not been evaluated in neonates.

Methods: We identified existing gene expression microarray-based cohorts of neonates with sepsis. We then tested the accuracy of the Sepsis MetaScore both alone and in combination with standard diagnostic laboratory tests in diagnosing sepsis.

Results: We found 3 cohorts with a total of 213 samples from control neonates and neonates with sepsis. The Sepsis MetaScore had an area under the receiver operating characteristic curve of 0.92-0.93 in all 3 cohorts. We also found that, as a diagnostic test for sepsis, it outperformed standard laboratory measurements alone and, when used in combination with another test(s), resulted in a significant net reclassification index (0.3-0.69) in 5 of 6 comparisons. The mean point estimates for sensitivity and specificity were 95% and 60%, respectively, which, if confirmed prospectively and applied in a high-risk cohort, could reduce inappropriate antibiotic usage substantially.

Conclusions: The Sepsis MetaScore had excellent diagnostic accuracy across 3 separate cohorts of neonates from 3 different countries. Further prospective targeted study will be needed before clinical application.


Figure 1.
Figure 1.
Receiver operating characteristic (ROC) curves of the Sepsis MetaScore comparing noninfected controls with neonates with sepsis (as defined by the original authors of each cohort). Abbreviations: AUC, area under the receiver operating characteristic curve; CI, confidence interval.
Figure 2.
Figure 2.
Receiver operating characteristic (ROC) curves for logistic models of the Sepsis MetaScore (SMS) alone (replotted to account for patients with missing laboratory data), plus clinical laboratory results in binary or raw form, with or without the SMS, for each cohort: (A) Cernada et al [29] (n = 36); (B) Smith et al [27] (n = 68) (no CRP data were available for this cohort); and (C) Wynn et al [23] (n = 83) (all 4 subgroups are combined because no substantial differences were found in subgroup analysis for individual laboratory results; thus, the same controls were counted twice in their respective comparison groups). Abbreviations: AUC, area under the receiver operating characteristic curve; CI, confidence interval; CRP, C-reactive protein.

Similar articles

See all similar articles

Cited by 3 articles

Publication types