Associated factors for unfavorable prognosis in neonates with septic shock

Abstract

Background

Neonatal septic shock represents a life-threatening complication of neonatal sepsis associated with substantial in-hospital mortality. Conventional diagnostic criteria only detect shock at the decompensated stage, which inevitably delays optimal treatment. This study aimed to identify independent predictors associated with adverse outcomes in neonates with septic shock and evaluate the predictive performance of combined early laboratory biomarkers and point-of-care cardiac ultrasound parameters.

Methods

A single-center retrospective cohort study was performed on 98 neonates diagnosed with septic shock admitted to the Neonatal Intensive Care Unit of Chengdu Women’s and Children’s Central Hospital from January 2024 to December 2025. Patients were stratified into a favorable prognosis group (n=67) and an unfavorable prognosis group (n=31) based on discharge outcomes. Perinatal baseline data, initial laboratory tests and bedside critical cardiac ultrasound measurements were compared between groups. Univariate analyses followed by backward stepwise binary logistic regression were used to screen independent prognostic predictors, and receiver operating characteristic (ROC) curves were constructed to quantify predictive efficacy.

Results

Compared with patients with favorable outcomes, neonates with poor prognosis exhibited lower gestational age, birth weight and Apgar scores, elevated serum interleukin-6 (IL-6), myoglobin (MYO) and high-sensitivity cardiac troponin I; reduced serum albumin and calcium, as well as increased serum potassium. Cardiac ultrasound revealed significantly decreased velocity-time integral (VTI), cardiac output (CO) and tricuspid annular plane systolic excursion (TAPSE) in the unfavorable prognosis group (all P<0.05). Multivariate logistic regression identified gestational age, VTI and TAPSE as independent protective factors, while elevated IL-6 served as an independent predictor for adverse prognosis (P<0.05). ROC analysis demonstrated that TAPSE yielded the highest area under the curve (AUC =0.857), followed by VTI (AUC =0.785), gestational age (AUC =0.739) and IL-6 (AUC =0.716). The combined four-marker model achieved an AUC of 0.905, which outperformed all individual indicators.

Conclusions

Lower gestational age, elevated IL-6, and reduced VTI and TAPSE independently associated with poor prognosis in neonates with septic shock. Integrated assessment combining perinatal characteristics, admission inflammatory biomarkers and bedside cardiac ultrasound enables identification of high-risk infants, and may facilitate risk stratification for neonatal septic shock.

Highlight box.

Key findings

• Four independent predictors of poor prognosis were identified in neonates with septic shock: lower gestational age, elevated interleukin-6 (IL-6), reduced velocity-time integral (VTI) and tricuspid annular plane systolic excursion (TAPSE). TAPSE showed the strongest predictive power, and the combined four-indicator model achieved excellent discriminatory performance. Multiple abnormal laboratory markers (hypoalbuminemia, hyperkalemia, hypocalcemia, elevated myocardial injury biomarkers) and impaired ventricular systolic function were more prevalent in infants with adverse outcomes.

What is known and what is new?

• Prematurity and inflammatory activation worsen neonatal septic shock outcomes, and conventional inflammatory markers C‑reactive protein (CRP) and procalcitonin (PCT) widely used for risk evaluation.

• In this cohort, admission IL-6 was identified as an independent prognostic marker, while CRP and PCT were not significantly associated with outcomes on univariate analysis. VTI and TAPSE showed stronger predictive performance than left ventricular ejection fraction (LVEF) for detecting cardiac dysfunction. Lung ultrasound (LUS) scores were elevated in both groups with no significant intergroup difference, possibly reflecting a ceiling effect in this population.

What is the implication, and what should change now?

• Clinicians should integrate gestational age, initial IL-6 and bedside cardiac ultrasound (VTI, TAPSE) to stratify high-risk neonates at admission. These findings suggest that IL-6 and TAPSE may serve as prognostic markers at admission, and that LVEF alone may be insufficient for assessing cardiac function in septic shock. The potential value of serial monitoring of IL-6 and right ventricular systolic function to guide interventions warrants further investigation in future prospective studies.

Introduction

Sepsis refers to a severe inflammatory response syndrome induced by invasive pathogens, which is often accompanied by organ dysfunction. The neonatal incidence is approximately 2,202 per 100,000 live births (1,2). As a life-threatening progressive complication of sepsis, septic shock is pathologically featured by a vicious cycle of overactivated systemic inflammation and microcirculatory disturbance. This pathological cascade rapidly leads to multiple organ damage or failure, ranking among the top leading causes of neonatal death. Cohort epidemiological data from China and other countries indicate that neonatal septic shock accounts for approximately 1.3% of all Neonatal Intensive Care Unit (NICU) admissions, with an in-hospital mortality rate reaching up to 20% (3,4).

The early clinical signs of neonatal septic shock are usually subtle and non-specific. Traditional diagnostic criteria based on hypotension or obvious hypoperfusion can only recognize shock at the decompensated stage, which inevitably delays treatment and forfeits the optimal therapeutic window. This study retrospectively analyzed clinical data of neonates with septic shock to identify factors associated with unfavorable prognosis based on initial assessment at admission, with the aim of informing risk stratification for this population. We present this article in accordance with the STROBE reporting checklist (available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0665/rc).

Methods

Subjects and study design

This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Institutional Review Board of Chengdu Women’s and Children’s Central Hospital, School of Medicine, University of Electronic Science and Technology of China [No. 2025 (81)], and individual consent for this retrospective analysis was waived. Neonates diagnosed with septic shock hospitalized in the NICU of Chengdu Women’s and Children’s Central Hospital between January 2024 and December 2025 were enrolled as study subjects. All enrolled infants met the following inclusion criteria: (I) diagnosed with neonatal septic shock by the attending NICU physician team based on comprehensive clinical assessment, including the presence of infection, hemodynamic instability (hypotension or requirement for vasoactive support), and signs of hypoperfusion; (II) postnatal age ≤ 28 days at the onset of the disease. Neonates were excluded based on the following criteria: (I) patients presenting with hemodynamic instability at admission; (II) previous use of vasoactive agents prior to hospitalization; (III) severe congenital heart disease (patent ductus arteriosus and patent foramen ovale without hemodynamic significance were excluded); (IV) incomplete clinical and laboratory data. A total of 116 neonates were initially diagnosed with septic shock. After applying the exclusion criteria, 18 cases were excluded, including 5 cases due to incomplete clinical and laboratory data, 9 cases due to overt septic shock at admission, and 4 cases due to previous use of vasoactive agents prior to hospitalization, leaving 98 cases for final analysis. All enrolled neonates were divided into the favorable prognosis group and the unfavorable prognosis group according to their discharge outcomes. Neonates who were improved or cured at discharge were classified into the favorable prognosis group. The unfavorable prognosis group was defined as a composite endpoint, including neonates who died during hospitalization or whose families abandoned treatment due to critical conditions caused by severe complications secondary to septic shock.

Study setting

This study was conducted at the NICU of Chengdu Women’s and Children’s Central Hospital, a tertiary‑level A (Grade III, Class A) hospital and a municipal-level referral center for critically ill neonates in Chengdu, China. The unit has 211 beds and admits approximately 7,000 neonates annually. The medical team consists of 66 neonatologists and 157 nurses, with a nurse‑to‑patient ratio that meets the standards for a tertiary NICU. The unit is fully equipped with invasive and non‑invasive ventilators, high‑frequency oscillatory ventilation, inhaled nitric oxide, therapeutic hypothermia, continuous renal replacement therapy (CRRT), and extracorporeal membrane oxygenation (ECMO). As a tertiary referral center, the unit receives transferred patients from primary and secondary hospitals across the region. Transfer decisions are made when the attending physicians at the referring hospitals determine that their current facilities and expertise are insufficient to manage the neonatal condition. This information is provided to help readers assess the generalizability of our findings to their own clinical settings.

Data collection

Clinical data of the patients were retrospectively extracted from the hospital electronic medical record system, which mainly included the following information. (I) General information: maternal obstetric history, birth setting (in-hospital vs. out-of-hospital), mode of delivery, gender, gestational age, birth weight, Apgar score and postnatal age. (II) Admission information: onset time of symptoms, chief manifestations and pre-admission medication history. (III) Auxiliary examinations: routine blood test, C-reactive protein (CRP), interleukin-6 (IL-6), serum procalcitonin (PCT), blood gas analysis, blood biochemistry, myocardial injury markers, and bedside critical care ultrasound (primarily cardiac ultrasound). To standardize the timing of measurements in this retrospective analysis, only the results of the first tests obtained after admission were adopted. However, the time interval from symptom onset to measurement was not controlled for, which limits any claim of early prediction. (IV) Treatment and prognosis: type and duration of antibiotic therapy, duration of invasive mechanical ventilation, category and administration course of vasoactive agents, and concomitant complications.

All point-of-care ultrasound examinations were performed using a color Doppler ultrasound diagnostic system (Wisonic, Shenzhen, China) uniformly provided by the department. The examinations were conducted by two NICU physicians with specialized certification, using the same device, and all procedures strictly adhered to the principles of critical care ultrasound application and quality control. The vasoactive agents used in this study included epinephrine, norepinephrine, dopamine, dobutamine and milrinone. All patients received their first-dose antibiotic within 2 hours after admission. Fluid resuscitation was delivered according to standard institutional protocols. Vasoactive-agent dosing was weight-based in accordance with clinical practice guidelines. In addition, all patients were maintained on invasive mechanical ventilation and received sedation or neuromuscular blockade at the time of echocardiography.

Statistical analysis

All statistical analyses were performed using SPSS 26.0 (IBM Corp, Armonk, NY, USA) and R 4.3.0. After normality testing, normally distributed continuous data were presented as mean ± standard deviation and compared between groups using the independent-samples t-test. Non-normally distributed data were expressed as median (interquartile range), and the Mann-Whitney U test was applied for intergroup comparisons. Categorical data were described as frequency (percentage) and analyzed with the chi-square test or Fisher’s exact test. Univariate and multivariate Logistic regression analyses were conducted to identify early independent prognostic predictors for neonates with septic shock. Continuous variables were analyzed as continuous measures in the logistic regression model. Receiver operating characteristic (ROC) curves were subsequently constructed to quantify the predictive discriminatory efficacy of single indicators, with the area under the curve (AUC) calculated for performance evaluation. A P value <0.05 was considered statistically significant.

Results

A total of 98 neonates with septic shock were enrolled in this study, including 31 cases (31.6%) with unfavorable prognosis and 67 cases (68.4%) with favorable prognosis. Among the 31 neonates with unfavorable prognosis, 3 died during hospitalization and 28 were abandoned after physician-family consensus that complications were irreversible. We further examined the time from symptom onset to admission. Among the 98 enrolled neonates, 54 (80.6%) in the favorable prognosis group and 22 (71.0%) in the unfavorable prognosis group were admitted within 24 hours of symptom onset, with no significant intergroup difference (P=0.29). Owing to well-developed local transportation networks, all neonates were transferred to our hospital within 8 hours.

Comparison of baseline characteristics between the two groups

There were no significant differences between the favorable prognosis group and the unfavorable prognosis group in terms of postnatal age, gender, mode of delivery, birth setting, premature rupture of membranes, intrauterine distress, chorioamnionitis, prenatal antibiotic use, prenatal corticosteroid use, maternal gestational hypertension, or maternal gestational diabetes mellitus (P>0.05). However, neonates in the unfavorable prognosis group had significantly lower gestational age (P<0.001), a higher proportion of preterm birth (P=0.007), and lower birth weight (P<0.001) compared with those in the favorable prognosis group. In addition, the 1-, 5-, and 10-minute Apgar scores in the unfavorable prognosis group were all significantly lower than those in the favorable prognosis group (P<0.05), as shown in Table 1.

Table 1. Comparison of demographic and perinatal characteristics between different prognosis groups.

Parameters	Favorable prognosis group	Unfavorable prognosis group	χ2/Z	P
Age (days)	1 (1, 13)	1 (1, 9)	−0.975	0.33
Gender			0.191	0.66
Male	42 (62.7)	18 (58.1)
Female	25 (37.3)	13 (41.9)
Gestational age (weeks)	37.4 (35.1, 39.1)	32.4 (28.6, 37.1)	−3.794	<0.001
Preterm birth	30 (44.8)	23 (74.2)	7.386	0.007
Birth weight (g)	2,890 (2,420, 3,180)	2,000 (1,250, 2,740)	−3.866	<0.001
Mode of delivery			1.058	0.37
Vaginal delivery	23 (34.3)	14 (45.2)
Cesarean section	44 (65.7)	17 (54.8)
Place of birth			0.363	0.63
Our hospital	19 (28.4)	7 (26.5)
External hospital	48 (71.6)	24 (73.5)
Apgar score
1 min	9 (7, 9)	7 (4, 9)	−2.596	0.009
5 min	10 (8, 10)	8 (8, 9)	−3.219	0.001
10 min	10 (9, 10)	9 (8, 9)	−3.483	<0.001
Premature rupture of membranes	14 (20.9)	9 (29.0)	0.781	0.44
Intrauterine infection	10 (14.9)	5 (16.1)	0.024	0.88
Chorioamnionitis	1 (1.5)	2 (6.5)	1.756	0.23
Antenatal antibiotic use	2 (3.0)	2 (6.5)	0.651	0.59
Antenatal steroid use	9 (13.4)	7 (22.6)	1.298	0.26
Maternal pregnancy-induced hypertension	5 (7.5)	2 (6.5)	0.033	0.86
Maternal gestational diabetes mellitus	15 (22.4)	8 (25.8)	0.138	0.71

Continuous variables are presented as median (interquartile range). Categorical variables are presented as n (%).

Comparison of laboratory parameters between the two groups

The comparison of laboratory parameters between the two groups is presented in Table 2. For arterial blood gas analysis on admission, although lactate and base excess (BE) showed an upward trend in the unfavorable prognosis group, the differences did not reach statistical significance. No significant intergroup differences were observed in pH, bicarbonate, or blood glucose levels (P>0.05). Among inflammatory markers, the IL-6 level was significantly higher in the unfavorable prognosis group than in the favorable prognosis group. No significant differences were found in PCT or C-reactive protein (CRP) levels between the two groups (P>0.05). For routine blood parameters, white blood cell count, neutrophil count, neutrophil-to-lymphocyte ratio (NLR), hemoglobin (Hb), and platelet count were comparable between the groups (P>0.05). Regarding blood biochemical parameters, the unfavorable prognosis group had significantly lower albumin levels (P=0.02), higher potassium levels (P=0.009), and lower calcium levels (P=0.02) compared with the favorable prognosis group. No significant differences were observed in alanine aminotransferase (ALT), aspartate aminotransferase (AST), total bilirubin, creatinine, prealbumin, total protein, or sodium levels between the two groups (P>0.05). For myocardial injury markers, the myoglobin (MYO) level was significantly higher in the unfavorable prognosis group (P=0.01), and the high-sensitivity cardiac troponin I (hs-cTnI) level was also significantly elevated (P=0.002). In contrast, no significant differences were found in brain natriuretic peptide (BNP) or creatine kinase-MB (CK-MB) levels between the two groups (P>0.05).

Table 2. Comparison of laboratory parameters between different prognosis groups.

Parameters	Favorable prognosis group	Unfavorable prognosis group	χ2/Z	P
Admission blood gas analysis
pH	7.305 (7.258, 7.371)	7.277 (7.204, 7.351)	−1.222	0.22
Lactate (mmol/L)	2.8 (1.9, 4.4)	3.6 (2.0, 8.7)	−1.723	0.09
Bicarbonate (mmol/L)	21.0 (17.7, 22.4)	19.1 (15.3, 22.0)	−1.383	0.17
BE (mmol/L)	−5.2 (−8.0, −3.0)	−7.2 (−9.9, −3.8)	−1.688	0.09
Blood glucose (mmol/L)	5.1 (3.8, 7.4)	5.5 (4.1, 7.6)	−0.815	0.42
Inflammatory biomarkers
PCT (ng/mL)	14.02 (4.98, 33.09)	13.22 (1.73, 25.94)	−0.527	0.60
IL-6 (pg/mL)	83.1 (51.6, 456.2)	1,580.0 (78.1, 5,000.0)	−3.433	0.001
C-reactive protein (mg/L)	4.9 (1.0, 15.3)	1.6 (0.8, 8.3)	−0.995	0.32
Routine blood Test
White blood cell (×109/L)	12.6 (7.1, 19.3)	13.3 (7.7, 20.2)	−0.098	0.92
Neutrophil count	8.7 (4.2, 14.7)	7.7 (4.8, 11.5)	−0.425	0.67
Neutrophil percentage (%)	74.2 (56.5, 83.2)	65.0 (50.7, 79.6)	−0.921	0.36
NLR	3.63 (2.33, 6.27)	3.36 (1.54, 4.36)	−1.085	0.28
Hb (g/L)	155.0±34.1	140.0±41.3	1.876	0.06
PLT (×109/L)	209.3±66.3	177.8±88.7	1.737	0.09
Serum biochemistry
ALT (U/L)	8.4 (5.8, 13.9)	7.3 (4.1, 47.0)	−0.221	0.83
AST (U/L)	71.9 (43.4, 91.4)	74.9 (41.1, 210.3)	−0.833	0.41
Total bilirubin (μmol/L)	78.8 (52.5, 102.4)	83.0 (39.8, 124.5)	−0.027	0.98
Creatinine (μmol/L)	66.2 (52.6, 77.2)	65.7 (47.1, 89.0)	−1.036	0.30
Prealbumin (g/L)	67.2 (58.7, 89.3)	66.4 (44.6, 85.4)	−0.824	0.41
Albumin (g/L)	27.5±4.2	25.1±4.5	2.369	0.02
Serum potassium (mmol/L)	3.9±0.6	4.5±1.1	−2.776	0.009
Serum sodium (mmol/L)	137.1 (134.3, 139.0)	136.5 (133.8, 140.3)	−0.317	0.75
Serum calcium (mmol/L)	1.36 (1.16, 1.81)	1.19 (1.03, 1.46)	−2.252	0.02
BNP (pg/mL)	4,492 (1,659, 9,653)	5,000 (400, 17,677)	−0.353	0.72
CK-MB (ng/mL)	7.66 (3.22, 15.82)	10.3 (5.5, 22.2)	−1.383	0.17
MYO (ng/mL)	36.8 (13.8, 83.1)	113.4 (56.3, 511.1)	−2.468	0.01
hs-cTnI (ng/mL)	0.02 (0.01, 0.10)	0.22 (0.03, 0.94)	−3.073	0.002

Data are presented as median (P₂₅, P₇₅) or mean ± SD. ALT, alanine aminotransferase; AST, aspartate aminotransferase; BE, base excess; BNP, brain natriuretic peptide; CK-MB, creatine kinase-MB isoenzyme; Hb, hemoglobin; hs-cTnI, high-sensitivity cardiac troponin I; IL-6, interleukin-6; MYO, myoglobin; NLR, neutrophil-to-lymphocyte ratio; PCT, procalcitonin; PLT, platelet count; SD, standard deviation.

Comparison of critical ultrasound parameters between the two groups

Compared with the prognosis group, the unfavorable prognosis group had a significantly lower velocity-time integral (VTI) (P<0.001) and cardiac output (CO) (P=0.005). Tricuspid annular plane systolic excursion (TAPSE) was also markedly decreased in the unfavorable prognosis group (P<0.001). No significant differences were detected in other cardiac function parameters, including left ventricular end-diastolic diameter, mitral annular plane systolic excursion (MAPSE), left ventricular ejection fraction (LVEF) and left ventricular output (LVO). There was no significant difference in lung ultrasound (LUS) score between the two groups (P>0.05), as shown in Table 3.

Table 3. Comparison of critical care ultrasound parameters between different prognosis groups.

Parameters	Favorable prognosis group	Unfavorable prognosis group	χ2/Z	P
Left ventricular end-diastolic diameter (mm)	15.8±2.8	14.9±3.2	1.164	0.25
MAPSE (cm)	0.50±0.15	0.43±0.15	1.502	0.14
LVEF (%)	64 (57, 69)	62 (54, 70)	−0.599	0.55
VTI (cm)	9.76 (7.8, 11.0)	6.22 (5.21, 7.82)	−4.515	<0.001
CO (mL/min)	270 (220, 360)	180 (130, 330)	−2.817	0.005
LVO (mL/min/kg)	107 (75, 140)	117 (75, 146)	−0.390	0.70
TAPSE (cm)	0.66 (0.55, 0.74)	0.51 (0.45, 0.54)	−5.669	<0.001
LUS score	22.0±8.5	23.0±7.9	−0.309	0.76

Data are presented as median (P₂₅, P₇₅) or mean ± SD. CO, cardiac output; LUS, lung ultrasound; LVEF, left ventricular ejection fraction; LVO, left ventricular output; MAPSE, mitral annular plane systolic excursion; SD, standard deviation; TAPSE, tricuspid annular plane systolic excursion; VTI, Velocity-time integral.

Independent predictors of unfavorable prognosis in neonatal septic shock

Linear regression analysis was performed to assess multicollinearity among all variables that exhibited statistical significance in the univariate analysis. Based on clinical relevance and statistical significance (P<0.05), four variables—gestational age, IL-6, VTI, and TAPSE—were selected as candidate predictors. The variance inflation factors (VIFs) for these variables ranged from 1.073 to 1.503 (all <2.0), indicating no significant multicollinearity, which made them eligible for inclusion in the binary logistic regression model. Given the high correlation between gestational age and birth weight (Spearman’s ρ=0.844), we retained only gestational age in the multivariable model and did not include birth weight to avoid collinearity. After Bonferroni correction, the four core predictors in the final multivariable model remained significant.

Binary logistic regression analysis with backward stepwise selection was performed to screen independent predictors of unfavorable prognosis in neonatal septic shock. All eight candidate variables were initially incorporated into the regression model, and non-significant variables were eliminated stepwise. The final model retained all four variables as significant independent predictors (P<0.05), with good calibration (Hosmer-Lemeshow test: χ²=4.862, P=0.77) and acceptable explanatory power (Nagelkerke R²=0.599). Gestational age, VTI, and TAPSE served as protective factors, indicating that increased gestational age and improved cardiac systolic function were correlated with decreased risk of unfavorable prognosis. In contrast, IL-6 acted as an independent risk factor, whereby elevated inflammatory levels increased the likelihood of adverse clinical outcomes, as shown in Table 4.

Table 4. Binary logistic regression analysis for unfavorable prognosis in neonatal septic shock.

Parameters	B	SE	Wald	P	OR	95% CI
Gestational age	−0.192	0.076	6.312	0.02	0.715	0.540–0.946
IL-6	0.066	0.020	11.201	0.001	1.068	1.028–1.109
VTI	−0.336	0.143	5.527	0.02	0.727	0.555–0.953
TAPSE	−5.809	2.562	5.142	0.02	0.003	0.001–0.455
Constant	11.072	3.058	13.105	0.001	64,356.7	–

B, regression coefficient; CI, confidence interval; IL-6, interleukin-6; OR, odds ratio; SE, standard error; TAPSE, tricuspid annular plane systolic excursion; VTI, velocity-time integral.

ROC curves were plotted to assess the predictive efficacy of the four independent predictors (gestational age, IL-6, VTI, TAPSE) for unfavorable prognosis in neonates with septic shock (Figure 1). Among all single indicators, TAPSE yielded the largest AUC of 0.857, followed by VTI (AUC =0.785), gestational age (AUC =0.739), and IL-6 (AUC =0.716). The combined multi-parameter model achieved an AUC of 0.905 [95% confidence interval (CI): 0.843–0.966, P<0.001], which was significantly higher than the AUC of any single parameter. This finding indicated that integrating multiple indicators could substantially improve the predictive performance for adverse outcomes. Internal validation using 2,000 bootstrap resamples yielded a bias-corrected C-index of 0.888 (optimism: 0.017), indicating that the model did not suffer from substantial overfitting and retained good discriminative ability after correction. To facilitate clinical interpretation, optimal cutoffs for the four independent predictors were determined using the Youden index. The optimal cutoff values, along with their corresponding sensitivity and specificity, are presented in Table 5.

Figure 1.

ROC curve analysis of independent predictors for predicting unfavorable prognosis in neonates with septic shock. GA, gestational age; IL-6, interleukin-6; ROC, receiver operating characteristic; TAPSE, tricuspid annular plane systolic excursion; VTI, velocity-time integral.

Table 5. Optimal cutoffs, sensitivity, specificity, and AUC of individual predictors.

Parameters	Cut-off	Sensitivity	Specificity	AUC (95% CI)	P
Gestational age	35.7	0.710	0.731	0.739 (0.629–0.849)	<0.001
IL-6	1073.1	0.613	0.910	0.716 (0.590–0.843)	0.001
VTI	7.83	0.774	0.746	0.785 (0.685–0.884)	<0.001
TAPSE	0.58	0.999	0.716	0.857 (0.783–0.931)	<0.001

AUC, area under the curve; CI, confidence interval; IL-6, interleukin-6; TAPSE, tricuspid annular plane systolic excursion; VTI, velocity-time integral.

Discussion

Septic shock is a severe and prevalent neonatal critical condition originating from multisite infections and remains a leading cause of neonatal mortality and adverse long-term outcomes. Early recognition of high-risk critically ill neonates is crucial for enabling timely and targeted interventions to optimize clinical prognosis. The present study indicates that neonates with septic shock who have lower gestational age, decreased VTI and TAPSE, and elevated IL-6 levels are at significantly higher risk of unfavorable outcomes, which warrants close clinical vigilance.

Gestational age and birth weight are fundamental prognostic factors for neonatal septic shock. Neonates in the unfavorable prognosis group exhibited significantly lower gestational age, lower birth weight, and a higher proportion of preterm births relative to those in the favorable prognosis group. Previous studies have confirmed that small-for-gestational-age and low-birth-weight neonates are at increased risk of sepsis and subsequent adverse neonatal morbidity (5,6). Preterm infants possess immature immune function, impaired mucosal and cutaneous barrier integrity, and insufficient organ reserve, which greatly increases their susceptibility to severe infection and irreversible septic shock (7). A 5-minute Apgar score below 7 has also been recognized as an independent predictive indicator for neonatal sepsis (5). Consistent with previous evidence, the present study found that neonates in the unfavorable prognosis group exhibited lower Apgar scores, indicating a close correlation between perinatal asphyxia and unfavorable outcomes in neonatal septic shock. This association may be explained by hypoxia–ischemia-mediated multiple organ damage secondary to perinatal hypoxemia.

IL-6 levels were significantly elevated in the unfavorable prognosis group compared with the favorable prognosis group, while no significant intergroup differences were found in conventional inflammatory markers including CRP and PCT. IL-6 is an upstream regulator of the inflammatory cascade. In the context of sepsis, IL-6 rises rapidly after infection and typically peaks earlier than CRP and PCT, which explains its sensitivity as an early-stage inflammatory marker. However, because the time interval from symptom onset to measurement was not controlled in this retrospective study, our findings should be interpreted as associations between admission IL-6 levels and outcomes rather than evidence of early prediction before clinical deterioration. Mechanistically, IL-6 triggers the JAK/STAT signaling pathway, promotes acute-phase protein synthesis, and facilitates T-cell and B-cell activation, thereby amplifying systemic inflammatory responses (8). Sustained excessive release of IL-6 further induces vascular endothelial damage, increases capillary permeability, disturbs microcirculatory perfusion, and ultimately contributes to multiple organ dysfunction in critically ill neonates (9).

In this study, the unfavorable prognosis group presented more severe metabolic abnormalities, including decreased albumin, elevated serum potassium, and reduced serum calcium levels. Hypoalbuminemia originates from impaired hepatic protein synthesis and massive capillary protein leakage, and has been well recognized as a reliable marker of poor clinical outcomes in critically ill populations (10). Hyperkalemia commonly occurs secondary to metabolic acidosis, tissue catabolism, and renal dysfunction during critical illness (11). Hypocalcemia is induced by inflammatory cytokine‑mediated parathyroid suppression and disordered vitamin D metabolism. Importantly, hypocalcemia can further aggravate myocardial depression and reduce vascular tension, ultimately exacerbating hemodynamic instability and maintaining the vicious cycle of septic shock (12,13).

Serum MYO and hs-cTnI levels were significantly higher in the unfavorable prognosis group, while BNP and CK-MB levels did not differ significantly between the two groups. As a highly specific biomarker of myocardial necrosis, elevated hs-cTnI reliably indicates the presence of myocardial injury and is closely correlated with sepsis-induced myocardial depression (14). In contrast, MYO is a non-specific marker for myocardial damage and primarily reflects systemic hypoxic injury to striated muscle tissues, including the myocardium (15). The pathogenesis of sepsis-induced myocardial depression is multifactorial, involving elevated circulating myocardial depressant factors [e.g., tumor necrosis factor-alpha (TNF-α) and interleukin-1 beta (IL-1β)], mitochondrial dysfunction, decreased cardiac β-receptor sensitivity, and intracellular calcium homeostasis disorder. Collectively, these pathological alterations lead to varying degrees of ventricular systolic and diastolic dysfunction, which serves as a critical contributor to refractory septic shock in neonates (16,17).

In the present study, cardiac functional parameters, including VTI, CO, and TAPSE, were significantly lower in the unfavorable prognosis group. Differences in treatment intensity between the two groups may have influenced the cardiac function measurements. Patients with more severe illness may have received more aggressive fluid resuscitation, which could affect VTI and CO measurements. In addition, variations in vasoactive agent selection and dosing, as well as differences in mechanical ventilation settings at the time of echocardiography, may have affected TAPSE and other echocardiographic parameters. These treatment-related factors were not controlled for in this retrospective analysis and should be considered when interpreting the cardiac function findings. Reduced VTI indicates decreased left ventricular stroke volume resulting from impaired myocardial contractility, insufficient preload, or elevated afterload, and serves as a reliable marker of left CO. Kharrat et al. (18) demonstrated significant impairment of left ventricular systolic function in preterm neonates with septic shock, and infants with a left CO below 150 mL/min/kg exhibited a higher risk of respiratory failure. Notably, LVEF showed no significant intergroup difference in this cohort, which does not rule out the presence of underlying cardiac dysfunction. This finding should be interpreted with caution. LVEF was measured at a single time point, and the sample size may have been insufficient to detect differences that might exist. LVEF may still be useful when measured serially or in specific clinical contexts. Therefore, our results suggest that LVEF may be less sensitive than VTI or TAPSE in this acute setting, rather than indicating that LVEF has no clinical value. Consistent with the consensus from the European Society for Pediatric and Neonatal Intensive Care (19), CO may decline during the compensated stage of septic shock despite preserved blood pressure. As an ejection-phase parameter, LVEF can remain within the normal range due to compensatory reductions in left ventricular end-diastolic volume, even when stroke volume is already decreased, rendering LVEF insensitive to early myocardial dysfunction. In contrast, VTI directly reflects real-time stroke volume and is highly responsive to subtle changes in CO. Reduced CO further compromises tissue perfusion and induces subsequent organ dysfunction.

Right ventricular function is equally indispensable for hemodynamic evaluation in neonatal septic shock. The neonatal right ventricle features a thin wall and high compliance, which confers good tolerance to volume overload but high susceptibility to pressure overload, such as pulmonary hypertension. Right ventricular dysfunction has been well documented in a substantial proportion of neonates with septic shock (20,21). As a core indicator of right ventricular systolic performance, decreased TAPSE reflects elevated pulmonary artery pressure and increased right ventricular afterload. In the present study, TAPSE remained a statistically significant independent predictor. After re‑scaling to millimeters for clinical interpretability, each 1‑mm increase in TAPSE corresponded to a 17.5% reduction in the risk of unfavorable prognosis [odds ratio (OR) =0.825, 95% CI: 0.710–0.958, P=0.01], supporting its clinical utility as a prognostic marker in neonatal septic shock. Resultant right ventricular dysfunction impairs systemic venous return and further deteriorates systemic circulatory failure (22). Although our findings identified TAPSE and IL-6 as independent prognostic markers at admission, whether serial monitoring of these parameters and adjusting interventions accordingly can improve survival outcomes remains unknown. This represents an important hypothesis to be tested in future prospective studies. Previous literature has indicated that the LUS score acts as an independent predictive marker of unfavorable prognosis in very low birth weight preterm neonates (23,24). In the current cohort, respiratory tract infection constituted the most common primary infectious focus, accompanied by a considerable incidence of acute respiratory distress syndrome (ARDS) at roughly 22.5%, and overall elevated LUS scores were detected across all subjects. This lack of significant intergroup difference may reflect a ceiling effect, as both groups exhibited severe lung pathology, rather than indicating that LUS is generally unhelpful. Additionally, the absence of difference could be attributable to insufficient statistical power for this specific comparison. Therefore, the prognostic value of LUS in neonates with septic shock warrants further investigation.

The present study identified gestational age, IL‑6, VTI, and TAPSE as independent predictors of unfavorable prognosis in neonatal septic shock. These indicators reflect disease severity from multiple dimensions, including baseline neonatal physical condition, systemic inflammatory burden, and cardiac systolic function, thereby offering promising clinical predictive value. In this cohort, 28 of the 31 unfavorable outcomes (90.3%) were due to treatment abandonment, with only 3 in‑hospital deaths. It is important to recognize that treatment abandonment in the local context is not merely a reflection of biological disease severity, but is also shaped by family decision‑making, healthcare costs, and cultural factors surrounding critical illness in neonates. In our center, abandonment decisions were made only after thorough physician‑family communication and consensus that the complications secondary to septic shock were irreversible, and that continued intensive care offered limited expected benefit. Nevertheless, the predominance of abandonment in the unfavorable outcome group means that the predictors identified in this study—gestational age, IL‑6, VTI, and TAPSE—should be interpreted primarily as factors associated with treatment abandonment rather than as independent predictors of biological mortality. Whether these factors truly predict in-hospital death remains uncertain. Future prospective studies with separate analyses of mortality and treatment abandonment are needed to disentangle these two distinct outcomes.

The optimal cutoffs identified in this study may assist clinicians in risk stratification at admission. For example, an IL-6 level exceeding 1,073.1 pg/mL identified neonates at higher risk of unfavorable outcomes, with a specificity of 91.0% but a relatively lower sensitivity of 61.3%, suggesting that this cutoff is more suitable for confirming high risk rather than screening. In contrast, a TAPSE cutoff of ≤0.58 cm yielded a sensitivity of 99.9% with a specificity of 71.6%, indicating its utility as a screening tool to identify nearly all neonates at risk, albeit with some false positives. The choice of cutoff should consider the clinical context: higher sensitivity may be preferred for screening purposes to avoid missing high-risk cases, whereas higher specificity may be preferred for confirming risk to guide intensive interventions. These cutoffs should be validated in prospective cohorts before widespread clinical implementation.

Several limitations of the present study should be acknowledged. As a single‑center retrospective study with a modest sample size, it is subject to selection and information bias owing to reliance on complete medical records. Only single‑time‑point admission laboratory and echocardiographic data were available, without serial follow‑up assessments. Our findings reflect population‑level admission‑based associations rather than individualized or true early‑phase prediction, since the interval from symptom onset to sample collection was uncontrolled. In addition, the multivariable model was built using stepwise selection with a limited number of outcome events, which may have introduced overfitting and selection instability. Consequently, clinically relevant variables may have been excluded after adjustment. External validation and well‑standardized prospective studies are therefore warranted before clinical translation.

Conclusions

Very and extremely preterm infants, as well as very and extremely low birth weight neonates, constitute a high-risk population for unfavorable prognosis in neonatal septic shock. Gestational age, IL-6, and critical care ultrasound parameters (VTI and TAPSE) were identified as independent prognostic factors with favorable predictive efficacy in the present cohort. These findings provide a novel strategy for the early identification of high-risk neonates with septic shock. Further multicenter prospective studies are required to validate and generalize these results, and to systematically evaluate the positive and negative predictive values of the combined model for individualized risk prediction.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Institutional Review Board of Chengdu Women’s and Children’s Central Hospital, School of Medicine, University of Electronic Science and Technology of China [No. 2025 (81)], and individual consent for this retrospective analysis was waived.

Data Sharing Statement

Available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0665/dss

tp-15-9-373-dss.pdf ^{(74.4KB, pdf)}

DOI: 10.21037/tp-2026-0665