Serum TRAIL and CSF PCT emerged as the strongest signals of mortality risk
A biomarker panel drawn from blood and cerebrospinal fluid identified distinct prognostic patterns in hospitalized neonates and young infants with suspected serious bacterial infection. In this Moroccan cohort, TRAIL in serum carried the strongest association with in-hospital death, while procalcitonin in CSF performed best for severe progression ending in death. The separation between those two compartments is clinically meaningful: serum reflects systemic host response, whereas CSF may capture central nervous system inflammation or bacterial burden when meningitis is part of the presenting syndrome.
The study focused on a population in which bedside uncertainty is common and consequences are immediate. Neonates and young infants can deteriorate quickly, their clinical signs are often nonspecific, and conventional tests rarely provide the kind of early risk signal needed for triage. In that setting, a biomarker that points toward heightened mortality risk is not a laboratory curiosity; it may help determine who needs closer observation, faster escalation, or a lower threshold for intensive support.
A prospective cohort, revisited through a biomarker lens
This was a prespecified secondary analysis of the UNITED-Meningitis project, conducted at the National Reference Centre for Neonatology and Nutrition in Rabat, Morocco. Recruitment ran from 17 March 2022 to 10 January 2024 and included hospitalized neonates and young infants aged 2 months or younger with suspected serious bacterial infection and a clinical indication for lumbar puncture. Eligibility rested on neurological symptoms or elevated inflammatory markers, defined as CRP at least 20 mg/L or PCT above 0.5 ng/mL. Infants with contraindications to lumbar puncture were excluded.
Residual serum and CSF from routine care were stored at -80 °C without freeze-thaw cycles and later analyzed at ISGlobal using Luminex multiplex assays. Twenty-one biomarkers spanning immune activation, endothelial dysfunction, and neuronal injury were measured head-to-head. The panel included CRP, PCT, Angpt-1, Angpt-2, IL-6, IL-8, IL-10, IL-12p70, IL-15, TNFR1, TREM-1, CHI3L1, TRAIL, ULBP-2/5/6, uPAR, VEGFR1/Flt-1, NGAL, NSE, S100B, and tau. Laboratory personnel were blinded to clinical data, and outcomes were assessed separately from biomarker measurement.
Mortality clustered around the markers of host failure
Among 179 participants with available serum and/or CSF samples, 9% died during hospitalization. The serum sample set included 136 participants with 11 deaths, and the CSF set included 148 participants with 12 deaths. Those numbers are small, but in neonatal studies they carry weight: death is uncommon enough to challenge model stability, yet common enough in this clinical context to demand a usable prognostic signal.
In serum, lower TRAIL concentrations and higher concentrations of Angpt-2, IL-8, and uPAR were associated with in-hospital death. TRAIL produced the strongest prognostic performance, with an AUROC of 0.86 and 95% confidence interval of 0.78 to 0.93. That pattern fits a plausible immunobiologic framework: reduced TRAIL may reflect impaired regulated immune signaling, while elevated Angpt-2, IL-8, and uPAR point toward endothelial activation, neutrophil recruitment, and tissue-level inflammatory stress.
The CSF findings carried a different signature. PCT was the strongest predictor of severe progression leading to in-hospital death, with an AUROC of 0.87 and 95% confidence interval of 0.80 to 0.95. In the same compartment, higher levels of IL-6, IL-12p70, TNFR1, TREM-1, Angpt-1, NGAL, and NSE were associated with higher odds of probable or laboratory-confirmed meningitis. Central nervous system injury and inflammation were therefore not represented by a single molecule, but by a cluster of signals that moved together.
A low TRAIL signal in serum may be more than a biomarker abnormality. It may be the molecular trace of a host response too weak to contain invasive infection.
Why the compartment matters in neonatal infection
Neonatal infection rarely behaves like adult sepsis. Immune responses are immature, clinical deterioration can be abrupt, and the same syndrome may involve bloodstream infection, meningitis, or both. Measuring biomarkers in both serum and CSF allowed the analysis to separate systemic inflammation from central nervous system involvement rather than forcing one matrix to stand in for the other.
That design choice matters because biomarker performance can collapse when the wrong biological compartment is sampled. A blood marker may track generalized endothelial injury, while a CSF marker may be more informative when meningeal inflammation or neurologic involvement dominates the presentation. The current analysis supports that distinction without overstating it: serum TRAIL aligned best with mortality, while CSF PCT and several inflammatory and neuronal markers aligned with meningitis-related outcomes.
What the findings can and cannot support today
The study was not built to prove clinical utility at the bedside. It was an exploratory validation exercise in a single center, and the sample size was limited by the practical constraints of obtaining enough residual blood and CSF from sick neonates. Firth’s penalized logistic regression was used to reduce small-sample bias, and false discovery rate correction was applied across the biomarker panel, but the findings still need confirmation in larger cohorts.
Two other constraints shape interpretation. First, the primary endpoint was revised to in-hospital death because timing data for critical-care interventions were not available in a form that supported the original composite outcome. Second, the study measured biomarkers at baseline only, so kinetic behavior remains unknown. In neonates, a marker that looks strong at presentation may behave very differently six or twelve hours later, especially after antibiotics, fluid resuscitation, or respiratory support.
Even with those limits, the clinical message is clear enough to matter. Biomarkers in both serum and CSF carried prognostic information beyond routine assessment, and the most informative molecules were not the same across compartments. If future multicenter work validates these signals, risk stratification for serious bacterial infections in neonates and young infants could move beyond a single inflammatory marker toward a structured panel anchored in host response, endothelial activation, and nervous system injury. For now, the findings remain exploratory, and the next step is straightforward: confirm them in larger studies, define how they change over time, and test whether they add value to the decisions that determine whether a fragile infant survives the first days of illness.