Blood Markers KLF2, YKL-40 and Vitamin A Track Severity and Predict Death Risk in Newborn Lung Disease

Three substances measurable in a single blood sample — the transcription factor Krüppel-like factor 2 (KLF2), the inflammatory protein chitinase-3-like protein 1 (YKL-40), and vitamin A — closely tracked both the severity of neonatal respiratory distress syndrome and the risk that an affected infant would die, according to a new study published in Frontiers in Pediatrics. When the three markers were combined in a statistical prediction model, the researchers report an area under the receiver operating characteristic curve of 0.918 (95% CI: 0.863–0.973), with a sensitivity of 0.905 and a specificity of 0.859 for identifying infants who did not survive their hospital stay.

Neonatal respiratory distress syndrome (NRDS) is primarily caused by insufficient pulmonary surfactant production, which leads to diffuse alveolar collapse, impaired gas exchange and severe hypoxemia. It remains one of the leading causes of neonatal morbidity and mortality worldwide. Although surfactant replacement therapy and modern respiratory support have substantially improved survival, severe disease continues to carry substantial mortality and long-term respiratory complications, particularly in resource-limited settings. Early identification of infants at high risk of progression or poor outcome is therefore a central goal of neonatal care, and blood-based biomarkers that could support such stratification have attracted growing interest.

The study, led by Na Li of Shanxi Children’s Hospital & Shanxi Women and Children’s Healthcare Hospital in Taiyuan, China, together with Feifei Yuan of Guangzhou Medical University, enrolled 120 neonates diagnosed with NRDS at their institution between January 2022 and December 2023, along with 60 neonates without NRDS born during the same period as controls. Diagnosis was established through imaging examinations and clinical signs, with chest x-rays showing diffuse bilateral pulmonary opacities and echocardiography excluding left atrial hypertension. Infants with major congenital malformations, inherited disorders, severe extrapulmonary infection, chromosomal abnormalities, transient tachypnea of the newborn, aspiration pneumonia or incomplete data were excluded. The authors describe the work as a single-center, retrospective study, and this design shapes how the findings should be interpreted.

For severity assessment, the NRDS infants were classified by initial chest radiographic findings into mild (grade I), moderate (grades II–III) and severe (grade IV) groups of 40 infants each, following the grading criteria described in Practical Pediatric Radiology Diagnostics.

Within 24 hours of diagnosis, peripheral venous blood was drawn from the NRDS infants, while control samples were obtained within 24 hours after birth. Serum KLF2 and YKL-40 were measured with commercially available enzyme-linked immunosorbent assays, and vitamin A was quantified by high-performance liquid chromatography with retinol standards. Across the severity groups, the biomarkers moved in opposite directions: relative to controls, infants with NRDS showed reduced KLF2 and vitamin A concentrations accompanied by elevated YKL-40 levels. Mean KLF2 fell from 2.07 ± 0.11 pg/mL in controls to 1.71 ± 0.25 pg/mL in mild, 1.66 ± 0.26 pg/mL in moderate and 1.43 ± 0.32 pg/mL in severe disease, while YKL-40 rose from 49.98 ± 5.68 mg/L in controls to 64.49 ± 12.04, 73.25 ± 10.95 and 80.43 ± 5.63 mg/L across the mild, moderate and severe groups. Vitamin A declined from 0.82 ± 0.23 μmol/L in controls to 0.60 ± 0.27, 0.59 ± 0.25 and 0.49 ± 0.18 μmol/L in the mild, moderate and severe groups. All of these comparisons reached statistical significance.

Outcome status divided the cohort into 99 survivors and 21 non-survivors, a mortality figure consistent with the approximately 21% rate the authors used in their sample size calculation. Demographic and perinatal variables, including maternal age, gravidity, gestational hypertension, gestational diabetes mellitus, gestational age at birth, neonatal sex, birth weight, postnatal age, term status, premature rupture of membranes and delivery mode, were similarly distributed between survivors and non-survivors. The biomarker profiles, however, differed markedly: non-survivors had lower KLF2 (1.34 ± 0.30 vs 1.66 ± 0.27 pg/mL) and vitamin A (0.42 ± 0.12 vs 0.59 ± 0.25 μmol/L) levels, and higher YKL-40 concentrations (84.47 ± 7.84 vs 70.23 ± 11.03 mg/L) than infants who survived.

When prognosis was modeled with logistic regression, using survival status as the dependent variable, all three serum markers emerged as independent predictors. In the multivariate analysis, KLF2 carried an odds ratio of 0.099 (95% CI: 0.012–0.807), YKL-40 an odds ratio of 1.190 (95% CI: 1.075–1.317), and vitamin A an odds ratio of 0.020 (95% CI: 0.001–0.726). The authors note that a multicollinearity check was performed before the multivariate analysis and showed no severe collinearity among the variables.

Assessed individually, each marker showed some discriminatory ability for mortality, but none was sufficient on its own. The reported area under the ROC curve was 0.779 for KLF2 (cut-off 1.020 pg/mL), 0.848 for YKL-40 (cut-off 76.860 mg/L) and 0.731 for vitamin A (cut-off 0.630 μmol/L). Combining the three markers improved performance to an AUC of 0.918 with sensitivity of 0.905 and specificity of 0.859. The authors suggest that this gain reflects the distinct biological processes each marker represents: KLF2 relates to pulmonary endothelial and epithelial homeostasis, YKL-40 reflects inflammatory activity, and vitamin A participates in lung maturation and repair, so simultaneous assessment may capture the disease state better than any single marker.

From the regression results the team constructed a nomogram assigning each of the three variables a score, with total scores of 0–160 points corresponding to predicted probabilities of poor prognosis ranging from 0.16 to 0.80; higher scores indicate higher predicted risk. Validation relied on internal methods only. ROC analysis of the nomogram reproduced the AUC of 0.918 (95% CI: 0.863–0.973). The calibration curve showed regions of both overestimation and underestimation of risk: for predicted values between 15% and 20%, or 27% and 32%, the curve fell below the diagonal, indicating that true positive probability was lower than predicted, while in other ranges risk was generally underestimated. A Hosmer-Lemeshow goodness-of-fit test yielded a chi-square of 8.210 with P = 0.143. For internal validation by cross-validation, with a training set proportion of 0.8 and a naive Bayes model with Gaussian feature distribution, the final model achieved an accuracy of 87.50%, a precision of 87.27%, a recall of 87.50% and an f1-score of 0.87 on the test set. Decision curve analysis indicated a positive net benefit across threshold probabilities from 0 to 0.88.

The discussion places these findings in a broader biological context, much of it drawn from experimental rather than clinical work. KLF2 is a zinc-finger transcription factor involved in endothelial homeostasis, anti-inflammatory regulation and vascular protection, and experimental studies have linked reduced KLF2 expression to aggravated lung injury and endothelial dysfunction. In neonatal rats with hyperoxia-induced lung injury, restoration of KLF2 expression reportedly alleviated alveolar damage and attenuated inflammation through inhibition of NF-κB signaling, and KLF2 downregulation has been described in neonates with persistent pulmonary hypertension of the newborn, a condition that frequently coexists with NRDS. YKL-40, an inflammatory glycoprotein secreted primarily by activated macrophages and neutrophils, has been implicated in inflammation, tissue remodeling and fibrotic responses, and elevated levels have previously been reported in bronchopulmonary dysplasia, acute respiratory distress syndrome and NRDS itself. Vitamin A is essential for lung morphogenesis, alveolar maturation and epithelial integrity; its active metabolite retinoic acid upregulates surfactant protein B expression via the retinoic acid receptor, and animal experiments have shown that vitamin A deficiency reduces total pulmonary surfactant phospholipids and surfactant protein B mRNA levels in fetal rats. A recent meta-analysis also found that vitamin A supplementation reduced the incidence of bronchopulmonary dysplasia in very preterm infants.

The authors are explicit that these mechanistic threads remain unproven in their own data. No mechanistic experiments were conducted, and the biological pathways linking the observed biomarker changes to adverse outcomes remain incompletely understood. They also acknowledge that the study did not collect maternal glycemic control data, placental pathology or genetic sequencing, so proposed links involving gestational diabetes, oxidative stress and surfactant gene mutations await validation through basic and genetic studies.

The limitations section is similarly candid. The study was single-center and retrospective, drawing on one hospital’s records, so selection bias cannot be excluded and generalizability may be restricted. Several variables that could affect mortality, including gestational age, birth weight, prenatal glucocorticoid exposure and disease severity, were not included in the multivariate model, creating the possibility of residual confounding, and no subgroup analyses were performed to test whether marker performance is consistent across patient subgroups. The biomarkers were measured only once, within 24 hours of diagnosis, so temporal trends and their relationship to prognostic evolution could not be assessed. Finally, the combined model has not been validated in an external independent cohort.

On this basis, the authors frame the nomogram as a potential tool for early identification of high-risk infants rather than a validated clinical instrument, noting that its value for clinical dissemination requires further prospective study. They propose multicenter prospective cohorts with fuller confounder adjustment and subgroup analyses, repeated sampling to define the optimal detection window, basic experiments on mechanism, and external validation through bootstrap resampling, cross-validation and independent cohorts as the next steps. They also emphasize that, because the design is observational, the direction of causality between low vitamin A levels and mortality remains unclear, and that any suggestion of vitamin A supplementation as an intervention would need prospective interventional trials before it could be considered established.

Subject of Research: Serum biomarkers for predicting severity and mortality in neonatal respiratory distress syndrome

Article Title: Serum KLF2, YKL-40, and vitamin A levels and their prognostic value in neonatal respiratory distress syndrome

Article References: Li, N., & Yuan, F. (2026). Serum KLF2, YKL-40, and vitamin A levels and their prognostic value in neonatal respiratory distress syndrome. Frontiers in Pediatrics, 14, Article 1898953. https://doi.org/10.3389/fped.2026.1898953

Image Credits: AI Generated

DOI: 10.3389/fped.2026.1898953

Keywords: neonatal respiratory distress syndrome, KLF2, YKL-40, vitamin A, biomarkers, prognosis, newborns, nomogram, logistic regression, pediatric pulmonology, mortality prediction, surfactant deficiency