Sepsis remains a major public health and critical care challenge, and accurate early biomarkers could help differentiate sepsis from other causes of acute illness at intensive care unit admission. Neutrophil gelatinase-associated lipocalin (NGAL) has been proposed as a sepsis biomarker because of its role in innate immune responses and its release during inflammatory and renal injury processes. This study assessed whether plasma NGAL measured at ICU admission discriminates patients with sepsis from those without and whether adding NGAL to C‑reactive protein (CRP) improves diagnostic performance.
This was a retrospective, multicentre observational analysis of adult patients admitted to four intensive care units between 2015 and 2018. Plasma NGAL and CRP were analysed in a cohort of 2,950 patients. All patients were retrospectively screened for sepsis using Sepsis‑3 criteria at the time of ICU admission.
Diagnostic discrimination of NGAL and CRP for sepsis was evaluated using receiver operating characteristic (ROC) analysis and area under the curve (AUC) estimates. NGAL values were adjusted for age and the presence of chronic kidney disease (CKD). The cohort was further stratified according to renal function, including subgroups with no kidney injury, acute kidney injury (AKI) only, CKD only, and CKD with AKI, to explore the influence of renal status on NGAL diagnostic performance.
The analysis compared AUCs for NGAL alone, CRP alone, and the combination of NGAL plus CRP. Reported statistical details in the source include p values and 95% confidence intervals for AUC estimates and increments; additional analytical specifics (for example, exact NGAL cutoffs, assay methods, or adjustment model parameters) were not reported in the source abstract.
Plasma NGAL concentrations were significantly higher in patients meeting Sepsis‑3 criteria compared with non‑septic patients (p < 0.001). Despite this difference in median/mean levels, the discriminative performance of NGAL for sepsis was modest when evaluated across the whole cohort.
Key ROC results reported in the source:
These results indicate that while NGAL levels are higher in septic patients, NGAL by itself delivered only modest discrimination (AUC well below levels generally considered strong diagnostic performance), and CRP alone performed better than NGAL alone in this cohort.
Because NGAL is affected by renal status and was adjusted for CKD and age in the analyses, the investigators stratified diagnostic performance by renal function categories. The source reports differential results across these strata:
These stratified findings suggest that NGAL may add diagnostic information in patients without chronic renal impairment and in those with new AKI, but the benefit is not evident in patients with established CKD.
In this large multicentre ICU cohort (n = 2,950), plasma NGAL measured at ICU admission showed only modest discrimination for diagnosing sepsis (AUC 0.66). CRP performed better as a single marker (AUC 0.71), and combining NGAL with CRP yielded only a small nominal improvement in overall AUC (increment 0.018).
Stratified analyses indicate that the incremental value of NGAL may be confined to patients without chronic kidney disease and to those with AKI only. However, in patients with preexisting CKD or combined CKD and AKI, adding NGAL did not improve diagnostic discrimination over CRP alone. These results underline the influence of renal function on NGAL interpretation and limit the generalisability of NGAL as a broadly useful sepsis diagnostic marker in heterogeneous ICU populations.
Overall, the authors conclude that available data from this observational study do not support a meaningful role for plasma NGAL in improving sepsis diagnosis in the ICU when added to CRP across an unselected ICU population.
The authors declared no competing interests. The study is published as an open‑access article in PLoS One. The source abstract reports the cohort, key ROC metrics, adjustment for CKD and age, and stratified renal analyses; additional methodological or granular data (such as assay platforms, cutoffs, or full multivariable model specifications) were not reported in the abstract and would require consultation of the full article for further detail.