Abstract
Background
Hepatic portal venous gas (HPVG) is an uncommon but potentially life-threatening ultrasonographic finding in the pediatric population. Early identification of high-risk HPVG cases requiring urgent surgical intervention remains a clinical challenge. This study aimed to develop and validate an ultrasound-based nomogram for predicting high-risk HPVG in children.
Methods
A retrospective cohort of 107 pediatric patients with ultrasound-confirmed HPVG was analyzed. Participants were divided into high-risk group (n=20) and benign group (n=87) based on whether they underwent abdominal surgery or died. Clinical characteristics and ultrasound features were compared between groups. Univariate and multivariate logistic regression analyses were performed to identify independent predictors. A nomogram was subsequently constructed, with model performance assessed via receiver operating characteristic (ROC) curves, calibration curves, the Hosmer-Lemeshow goodness-of-fit test, bootstrap internal validation, and decision curve analysis (DCA).
Results
Multivariate logistic regression identified bowel wall layer [unclear vs. clear: odds ratio (OR) =25.171, 95% confidence interval (CI): 4.249–227.708, P=0.001] and intestinal gas distribution (localized vs. none: OR =6.978, 95% CI: 1.515–51.211, P=0.02; diffuse vs. none: OR =65.077, 95% CI: 10.107–663.804, P<0.001) as independent predictors of high-risk HPVG. The model demonstrated good discriminative capability [area under curve (AUC) =0.899, 95% CI: 0.843–0.956]. The Hosmer-Lemeshow test confirmed adequate model fit (P=0.051). Bootstrap internal validation with 500 resamples yielded an optimism-corrected AUC of 0.879 (95% CI: 0.755–0.965), with minimal optimism of 0.020. DCA confirmed meaningful clinical net benefit across a wide range of threshold probabilities.
Conclusions
This study developed and validated an ultrasound-based prediction model for identifying high-risk HPVG in children. Loss of bowel wall stratification and abnormal intestinal gas distribution were identified as independent predictors of adverse outcomes. This model provides a practical, point-of-care risk stratification tool that can assist physicians in distinguishing children with HPVG who may benefit from urgent surgical intervention from those who can be managed conservatively.
Keywords: Hepatic portal venous gas (HPVG), children, ultrasound, prediction model, nomogram
Highlight box.
Key findings
• This study developed and validated an ultrasound-based nomogram for predicting high-risk hepatic portal venous gas (HPVG) in children. Multivariate logistic regression identified bowel wall layer unclear and intestinal gas distribution as independent predictors of high-risk HPVG. The nomogram demonstrated good discriminative ability, with an apparent area under curve (AUC) of 0.899 (95% confidence interval: 0.843–0.956), an optimism-corrected AUC of 0.879, and minimal optimism (0.020), with decision curve analysis (DCA) confirming meaningful clinical net benefit across a wide range of threshold probabilities (0.05–0.80).
What is known and what is new?
• HPVG is an uncommon but potentially life-threatening ultrasonographic finding in children, with etiologies ranging from benign self-limiting diseases to life-threatening emergencies requiring urgent surgery. Prior studies have examined ultrasound features such as bowel wall thickening, bowel peristalsis, and ascites in specific disease entities like necrotizing enterocolitis (NEC).
• This is the first prediction model specifically designed for pediatric HPVG risk stratification using a nomogram approach that encompasses the full spectrum of HPVG etiologies rather than focusing on a single disease. The model identifies loss of bowel wall stratification and abnormal intestinal gas distribution—two routinely assessable ultrasound parameters—as independent predictors, extending prior NEC—specific observations to the broader pediatric HPVG population regardless of underlying etiology.
What is the implication, and what should change now?
• The validated nomogram transforms pediatric HPVG from a subjective radiological sign into an evidence-based clinical decision tool. Clinicians should integrate this model into emergency and intensive care workflows to rapidly triage patients, accurately distinguishing self-limiting, benign cases from those requiring immediate life-saving intervention. External validation in prospective, multicenter cohorts is warranted before widespread clinical adoption.
Introduction
HPVG is an imaging sign characterized by the presence of gas within the portal venous system and its intrahepatic branches (1,2). Initially described as an ominous finding associated with high mortality, HPVG is increasingly recognized in diverse clinical conditions spanning the spectrum from benign self-limiting diseases (such as enteritis and food allergy) to life-threatening emergencies requiring urgent surgical intervention [such as necrotizing enterocolitis (NEC) and intestinal necrosis] (3,4). In pediatric populations, the etiology of HPVG is particularly heterogeneous. High-resolution ultrasound is the preferred examination method for evaluating HPVG in children, compared with abdominal X-ray and computerized tomography (CT) examination, it provides a radiation-free, real-time display, bedside-accessible imaging that can visualize characteristic branching hyperechoic foci within the liver parenchyma (1,5,6). Ultrasound has demonstrated high sensitivity for identifying portal venous gas, often detecting it at an early stage, which is critical in acutely ill children (1,4). In addition to detecting HPVG itself, abdominal ultrasound can also comprehensively evaluate related intestinal abnormalities, including intestinal wall gas, bowel wall thickness, layering, peristalsis, and the presence of ascites and intraperitoneal free gas in the abdominal cavity (7-9). These sonographic features of intestinal injury may provide additional prognostic information beyond the mere presence of portal venous gas.
Despite advances in imaging detection, significant uncertainties persist in the clinical management of pediatric HPVG. First, it is impossible to determine whether a child needs emergency surgery or can be treated conservatively through portal gas ultrasound identification alone (1,3). Second, several studies have attempted to identify associated ultrasonographic features that may aid in prognostication, such as bowel wall thickening, bowel wall layer stratification, bowel peristalsis, and ascites characteristics (3,10). However, these studies have been limited in scope: most were small, descriptive, single-disease studies focusing predominantly on NEC, did not encompass the full spectrum of HPVG etiologies in children, and none provided a validated, quantitative tool for individualized risk prediction (10,11). Therefore, there is still a lack of a comprehensive and quantifiable risk stratification tool that can predict high-risk pediatric patients with HPVG who require emergency surgery. To address this issue, we have developed and validated a novel ultrasound-based nomogram prediction model for predicting high-risk HPVG in children. The model is intended for a specific and common clinical scenario: a child in whom HPVG has been detected on an abdominal ultrasound performed for the evaluation of abdominal symptoms or critical illness, and in whom the need for urgent surgical intervention remains uncertain after initial clinical assessment. In this setting, the model aims to assist-rather than replace-clinical judgment by providing an objective, quantitative risk estimate, thereby helping clinicians rapidly and objectively stratify children with HPVG, optimizing clinical decision-making and improving patient treatment outcomes. We present this article in accordance with the TRIPOD reporting checklist (available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0541/rc).
Methods
Study participants
The inclusion criteria were as follows: (I) patient age under 18 years; (II) total abdominal ultrasound examination and confirmation of HPVG; (III) complete ultrasound and clinical data. The exclusion criteria were as follows: (I) no follow-up data; (II) poor ultrasound image quality, inability to confirm certain ultrasound features.
This retrospective cohort study was conducted at Tianjin Children’s Hospital. During the study period from January 2020 to December 2025, 107 pediatric patients with ultrasound-confirmed HPVG were enrolled. Participants were divided into a high-risk group (n=20) and a benign group (n=87) based on whether they underwent abdominal surgery or died within two weeks of the ultrasound. The study workflow is detailed in Figure 1.
Figure 1.
Study flowchart. AUC, area under curve; DCA, decision curve analysis; HPVG, hepatic portal venous gas; ROC, receiver operating characteristic.
In our institution, abdominal ultrasound is routinely used as the first-line imaging examination for children presenting with acute abdominal symptoms or signs (such as vomiting, abdominal distension, feeding intolerance, or bloody stools) and for critically ill neonates with suspected abdominal pathology. In all enrolled patients, HPVG was detected during these clinically indicated ultrasound examinations—in most cases at the initial diagnostic evaluation, before any decision regarding surgical intervention had been made.
The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of Tianjin Children’s Hospital (No. W-2026-030). Informed consent was waived for this retrospective study.
Data collection
All data were extracted from the electronic medical record system and included demographic information, confirmed diagnoses, treatment data, and ultrasound data. The original ultrasound examinations were performed by attending pediatric sonographers with at least 5 years of experience in pediatric abdominal ultrasound. Throughout the study period, our institution followed a standardized abdominal ultrasound scanning protocol in routine clinical practice: examinations were performed using Mindray Resona 7 (China), GE Logiq E9 (USA), or Philips Epiq 5 (the Netherlands) ultrasound systems, with high-frequency linear array transducers (5–12 MHz) for scanning, supplemented by convex transducers (2–5 MHz) to allow assessment of both superficial and deep organs and tissues. The patient was placed in the supine position with the abdomen fully exposed, multi-section scanning of the entire abdominal cavity (including abdominal organs such as the liver and intestines) was performed, and the ultrasound images were collected and stored. All ultrasound images were subsequently reviewed retrospectively by two pediatric sonographers from our institution, each with more than 10 years of experience in pediatric abdominal ultrasound, who were unaware of the clinical outcomes. Ultrasound features on which they agreed were retained, and those on which they disagreed were resolved by discussion and consensus. The following ultrasound features were systematically evaluated (Figure 2):
Figure 2.
Gray-scale ultrasound shows different sonographic features. (A) and (B) are ultrasound images of a 3-day-old male patient diagnosed with NEC: (A) shows a wide range of hyperechoic spots within the liver parenchyma, suggesting diffuse HPVG; (B) shows bowel wall thickening, unclear bowel wall layer, multiple intestinal wall gas (thick arrows) and poor echo of ascites (thin arrow). (C) and (D) are ultrasound images of a 10-month-old male patient diagnosed with enteritis: (C) shows scattered punctate hyperechoic within the parenchyma of the right lobe of the liver, suggesting localized HPVG (thick arrows); (D) shows bowel wall thickening, clear bowel wall layer, and intestinal wall gas (thick arrows). (E) is an ultrasound image of a 136-month-old female patient diagnosed with intestinal obstruction, which shows good echo of ascites (thick arrows). (F) is an ultrasound image of a 14-month-old male patient diagnosed with closed duodenal injury, it shows intra-abdominal hyperechoic (thick arrows), suggesting free peritoneal air. HPVG, hepatic portal venous gas; NEC, necrotizing enterocolitis.
HPVG distribution: classified as diffuse when gas echo extended to both hepatic lobes, or localized when confined to a single lobe or segment.
Free peritoneal gas: presence or absence of gas echo in the peritoneal cavity.
Bowel wall thickening: measured perpendicular to the intestinal wall (from the interface between the serosal layer and the muscularis propria to the interface between the mucosal layer and the intestinal lumen), an intestinal wall thickness >2.5 millimeters is considered thickened (7).
Bowel wall layer stratification: categorized as clear when the normal five-layer architecture was preserved, or unclear when layer definition was obscured or obliterated.
Bowel peristalsis: assessed in real time over a minimum 1-minute observation period in at least three bowel segments, including any segment with abnormal wall morphology, and classified as normal (visible peristaltic waves) or lacking (absent or markedly diminished motility). When motility differed among segments, the classification was based on the most abnormal segment: absent or markedly diminished motility in any assessed segment led to classification as lacking.
Ascites: characterized as none, good echo (anechoic fluid), or poor echo (complex or echogenic fluid).
Intestinal gas distribution: evaluated across the entire accessible bowel and classified as diffuse (abnormally increased gas echoes involving multiple bowel segments in more than one abdominal quadrant), localized (abnormally increased gas confined to a single bowel segment or region), or none (no abnormally increased intraluminal gas). Localized and diffuse patterns were collectively defined as abnormal intestinal gas distribution. This classification inevitably involves a degree of subjective judgment, which is acknowledged in the Limitations.
Statistical analysis
Statistical analysis was performed using SPSS software (version 25.0) and R software (version 4.5.1). Data distribution was assessed using the Shapiro-Wilk normality test. Nonnormally distributed continuous variables were expressed as median [interquartile range (IQR)] and compared using the Mann-Whitney U test. Categorical variables are presented as counts (%) and compared using the χ2 test or Fisher’s exact test, as appropriate. Candidate variables that demonstrated an association with high-risk HPVG at the P value <0.05 level in univariate logistic regression analysis were advanced to multivariable logistic regression. A forward stepwise selection procedure was used, specifying the first ordinal category of each multilevel variable as the reference level. Given the limited number of outcome events (n=20), we deliberately restricted the number of candidate predictors entering the multivariable model: only variables significant in univariate analysis were considered, and forward stepwise selection was used to retain the minimum set of independent predictors, thereby limiting the model’s degrees of freedom and reducing the risk of overfitting.
After identifying the independent predictor variables, a visual nomogram was constructed using the R software package. Model discrimination was quantified by area under curve (AUC) of the ROC. Calibration performance was evaluated by calibration plots and the Hosmer-Lemeshow goodness-of-fit test, with P value >0.05 indicating adequate fit. To address potential overoptimism inherent in single-cohort model development, internal validation was performed using a bootstrap resampling approach with 500 iterations, producing optimism-corrected AUC estimates. The model’s clinical applicability and net benefit were further tested by decision curve analysis (DCA). A two-sided P value <0.05 was considered statistically significant.
Results
Patient characteristics
A total of 107 children with HPVG were enrolled, including 87 cases (81.3%) in the benign group and 20 cases (18.7%) in the high-risk group. The median age was 1 month (IQR, 0–23 months) in the benign group and 0 months (IQR, 0–42 months) in the high-risk group. The very low median ages reflect that the cohort was dominated by neonates and young infants, the population in which pediatric HPVG most frequently occurs. In the benign HPVG group, enteritis was the predominant confirmed diagnosis, accounting for 48 cases (55.2%), followed by food allergy in 9 cases (10.3%), pneumonia in 6 cases (6.9%), NEC in 4 cases (4.6%), and intestinal obstruction in 4 cases (4.6%). The remaining cases comprised a variety of less frequent diagnoses, each accounting for fewer than 4% of the benign group. In the high-risk HPVG group, NEC was the most prevalent diagnosis, representing 7 cases (35.0%). Intestinal atresia was the second most common diagnosis with 4 cases (20.0%), followed by intestinal necrosis in 3 cases (15.0%) and gastric muscular defect in 3 cases (15.0%). The remaining diagnoses included intestinal volvulus (1 case, 5.0%), closed duodenal injury (1 case, 5.0%), and intestinal obstruction (1 case, 5.0%). Among the 20 cases in the high-risk HPVG group, 19 cases underwent abdominal surgery, including 1 case of NEC who died after operation, and the rest survived after operation. The remaining 1 case died without surgical treatment and was diagnosed as NEC. Both deaths occurred in children with NEC and were directly attributable to the HPVG-associated gastrointestinal disease process (intestinal necrosis and its complications).
No significant differences in age, gender and bowel wall thickening (P>0.05) between groups. Significant differences between the groups in HPVG distribution, free peritoneal gas, bowel wall layer, bowel peristalsis, ascites and intestinal gas distribution (P<0.05). The comparison of clinical and ultrasound characteristics between groups is presented in Table 1.
Table 1. Comparison of clinical and ultrasound characteristics between benign and high-risk HPVG groups.
| Characteristics | Benign HPVG group (n=87) | High-risk HPVG group (n=20) | χ2/Z value | P value |
|---|---|---|---|---|
| Age (months) | 1 [0–23] | 0 [0–42] | −1.906 | 0.057 |
| Gender | 1.694 | 0.19 | ||
| Female | 47 (54.0) | 14 (70.0) | ||
| Male | 40 (46.0) | 6 (30.0) | ||
| HPVG distribution | 4.168 | 0.041* | ||
| Localized | 34 (39.1) | 3 (15.0) | ||
| Diffuse | 53 (60.9) | 17 (85.0) | ||
| Free peritoneal gas | 8.866 | 0.03* | ||
| Absent | 87 (100.0) | 18 (90.0) | ||
| Present | 0 (0.0) | 2 (10.0) | ||
| Bowel wall thickening | 1.501 | 0.22 | ||
| Absent | 48 (55.2) | 8 (40.0) | ||
| Present | 39 (44.8) | 12 (60.0) | ||
| Bowel wall layer | 23.555 | <0.001* | ||
| Clear | 84 (96.6) | 12 (60.0) | ||
| Unclear | 3 (3.4) | 8 (40.0) | ||
| Bowel peristalsis | 8.452 | 0.004* | ||
| Normal | 53 (60.9) | 5 (25.0) | ||
| Lacking | 34 (39.1) | 15 (75.0) | ||
| Ascites | 21.883 | <0.001* | ||
| None | 63 (72.4) | 5 (25.0) | ||
| Good echo | 22 (25.3) | 10 (50.0) | ||
| Poor echo | 2 (2.3) | 5 (25.0) | ||
| Intestinal gas distribution | 31.937 | <0.001* | ||
| None | 54 (62.1) | 3 (15.0) | ||
| Localized | 30 (34.5) | 8 (40.0) | ||
| Diffuse | 3 (3.4) | 9 (45.0) |
Data are presented as median [interquartile range] or n (%); *, P<0.05 is considered statistically significant. HPVG, hepatic portal venous gas.
Univariate and multivariate logistic regression
In univariate analysis, bowel wall layer [unclear vs. clear: odds ratio (OR) =18.667, P<0.001], bowel peristalsis (lacking vs. normal: OR =4.676, P=0.006), ascites (good echo vs. none: OR =5.727, P<0.001; poor echo vs. none: OR =31.500, P<0.001), and intestinal gas distribution (localized vs. none: OR =4.800, P=0.002; diffuse vs. none: OR =54.000, P<0.001) were significantly associated with high-risk HPVG. Multivariate analysis identified bowel wall layer [unclear vs. clear: OR =25.171, 95% confidence interval (CI): 4.249–227.708, P=0.001] and intestinal gas distribution (localized vs. none: OR =6.978, 95% CI: 1.515–51.211, P=0.02; diffuse vs. none: OR =65.077, 95% CI: 10.107–663.804, P<0.001) as independent predictors. Bowel peristalsis (P=0.68) and ascites (P=0.053/0.17) were not significant in the multivariate model. Detailed data are shown in Table 2. Free peritoneal gas was detected in only 2 of 107 patients (1.9%), both in the high-risk group (P=0.03). Because a variable with only two positive observations cannot be reliably modeled-it would cause complete or quasi-complete separation and produce unstable, uninterpretable coefficient estimates-free peritoneal gas was excluded from regression analysis and model construction; this exclusion is acknowledged as a modeling limitation (see Discussion).
Table 2. Univariate and multivariate logistic regression analysis for predictors of high-risk HPVG.
| Variables | Univariate analysis | Multivariate analysis | |||
|---|---|---|---|---|---|
| OR (95% CI) | P value | OR (95% CI) | P value | ||
| Age (months) | 1.001 (0.986–1.012) | 0.93 | |||
| Gender | |||||
| Female (reference) | |||||
| Male | 0.504 (0.165–1.381) | 0.20 | |||
| HPVG distribution | |||||
| Localized (reference) | |||||
| Diffuse | 3.635 (0.990–16.409) | 0.052 | |||
| Bowel wall thickening | |||||
| Absent (reference) | |||||
| Present | 1.846 (0.694–5.140) | 0.23 | |||
| Bowel wall layer | |||||
| Clear (reference) | |||||
| Unclear | 18.667 (4.714–94.820) | <0.001* | 25.171 (4.249–227.708) | 0.001* | |
| Bowel peristalsis | |||||
| Normal (reference) | |||||
| Lacking | 4.676 (1.646–15.465) | 0.006* | 1.397 (0.299–7.325) | 0.68 | |
| Ascites | |||||
| None (reference) | |||||
| Good echo | 5.727 (1.830–20.149) | <0.001* | 4.184 (0.996–19.383) | 0.053 | |
| Poor echo | 31.500 (5.408–266.024) | <0.001* | 6.235 (0.455–104.791) | 0.17 | |
| Intestinal gas distribution | |||||
| None (reference) | |||||
| Localized | 4.800 (1.282–23.180) | 0.002* | 6.978 (1.515–51.211) | 0.02* | |
| Diffuse | 54.000 (10.770–381.271) | <0.001* | 65.077 (10.107–663.804) | <0.001* | |
*, P<0.05 is considered statistically significant. CI, confidence interval; HPVG, hepatic portal venous gas; OR, odds ratio.
Prediction model and nomogram
Based on the multivariate analysis results, a prediction model was constructed using the two independent predictors: bowel wall layer and intestinal gas distribution. The model equation was: logit(P) = −3.6170 + 3.2257 × Bowel wall layer Unclear + 1.9427 × Intestinal gas distribution Localized + 4.1756 ×Intestinal gas distribution Diffuse.
Where P represents the probability of high-risk HPVG. A nomogram for visual risk assessment was further developed (Figure 3). Each variable is assigned points on the upper scale, and the total points correspond to the predicted probability of high-risk HPVG on the lower scale.
Figure 3.
Nomogram for predicting high-risk HPVG in children. The nomogram includes two independent predictors: bowel wall layer (clear vs. unclear) and intestinal gas distribution (none vs. localized; none vs. diffuse). Reference categories (clear, none) receive the lowest points, while adverse categories (unclear, diffuse) receive higher points. HPVG, hepatic portal venous gas.
Diagnostic performance
The ROC curve analysis demonstrated that the prediction model had good discriminative ability, with an AUC of 0.899 (95% CI: 0.843–0.956) (Figure 4). This indicates that for a randomly selected pair of high-risk and benign HPVG cases, the model correctly assigns a higher predicted probability to the high-risk case 89.9% of the time.
Figure 4.
ROC curve of the prediction model. AUC, area under curve; CI, confidence interval; ROC, receiver operating characteristic.
Model calibration
The calibration curve showed good agreement between predicted probabilities and observed outcomes (Figure 5). The Hosmer-Lemeshow goodness-of-fit test yielded a chi-squared statistic of 18.2146 with 10 degrees of freedom (P=0.051), indicating adequate model fit. The non-significant result confirms that the model’s predictions are statistically consistent with observed outcomes across the entire range of predicted probabilities.
Figure 5.
Calibration curve of the prediction model. The solid blue line represents apparent calibration (LOESS smoothing), and red dots with error bars indicate bootstrap-validated calibration with 95% confidence intervals (500 resamples), and the diagonal dashed line represents good calibration. LOESS, locally estimated scatterplot smoothing.
Bootstrap internal validation
Bootstrap internal validation with 500 resamples was performed to assess model robustness (Figure 6). The apparent AUC was 0.899 and the estimated optimism was 0.020, yielding an optimism-corrected AUC of 0.879 (95% CI: 0.755–0.965). The small optimism indicates that the model’s apparent performance is only marginally inflated; however, the wide CI reflects the limited number of outcome events, and the stability of the performance estimates should be interpreted with caution.
Figure 6.
Bootstrap internal validation of the ROC curve. The bold blue line represents the apparent ROC curve derived from the original dataset, and the gray lines depict individual ROC curves from each bootstrap iteration using out-of-bag samples. AUC, area under curve; CI, confidence interval; ROC, receiver operating characteristic.
DCA
DCA was performed to evaluate the clinical utility of the prediction model (Figure 7). The prediction model demonstrated superior net benefit across a wide range of threshold probabilities (approximately 0.05–0.80) compared to both the “treat all” and “treat none” strategies. This indicates that using the model to guide clinical decisions would result in better clinical outcomes than either treating all or treating no patients.
Figure 7.
DCA of the prediction model. The blue solid line represents the prediction model, and the red dashed line represents the “treat all” strategy, and the horizontal dotted line at zero represents the “treat none” strategy. DCA, decision curve analysis.
Discussion
Due to its advantages such as non-radiation, real-time imaging and bedside examination, ultrasound has become the preferred examination method for pediatric patients (1,5,12). Ultrasound easily identifies HPVG, and most pediatric patients with HPVG are discovered through abdominal ultrasound examination. However, after clinicians identify HPVG, how to differentiate between patients (high-risk) who require emergency surgery and those (benign) who only need conservative treatment remains an unresolved issue. To provide some assistance in solving the aforementioned problem, we conducted a retrospective study. In our study, we developed and validated an ultrasound-based nomogram for predicting high-risk HPVG in children. Our model identified bowel wall layer and intestinal gas distribution as independent predictors, achieving an apparent AUC of 0.899 and an optimism-corrected AUC of 0.879 (optimism =0.020). To our knowledge, this is the first prediction model specifically designed for pediatric HPVG risk stratification using a nomogram approach.
It is important to define the intended use of this model. Our cohort was etiologically heterogeneous, and we acknowledge that in several high-risk diagnoses (such as intestinal atresia, volvulus, or intestinal perforation), management decisions are often driven by the clinical presentation and conventional imaging findings irrespective of HPVG. The nomogram is therefore not intended to diagnose surgical disease or to replace standard clinical and radiographic assessment. Rather, it addresses the common and unresolved situation in which HPVG is detected on ultrasound and the treating team must decide whether the child requires urgent surgical evaluation or can be safely observed. In this scenario, the model provides an objective risk estimate derived from two routinely assessable ultrasound features, which can be integrated with the clinical manifestations and conventional imaging to support and expedite decisions regarding surgical consultation, monitoring intensity, and transfer to centers with pediatric surgical capability.
Our finding that loss of bowel wall stratification is a strong independent predictor is consistent with the pathophysiology of intestinal ischemia. Normal bowel wall exhibits a characteristic five-layer architecture on high-frequency ultrasound, corresponding to the mucosa, muscularis mucosae, submucosa, muscularis propria, and serosa (13,14). Preservation of this layered pattern generally indicates viable bowel, whereas loss of layer definition suggests transmural ischemia or necrosis (14,15). In the context of NEC, the most common cause of HPVG in neonates, previous studies have demonstrated that loss of bowel wall stratification is associated with transmural necrosis and the need for surgical intervention (16,17). Cuna et al. conducted a systematic review and meta-analysis showing that bowel ultrasound findings, including loss of wall stratification, have significant predictive value for surgical NEC (18). Our study extends these observations to the broader pediatric HPVG population, demonstrating that this ultrasound feature maintains its prognostic significance regardless of the underlying etiology. The magnitude of the OR (25.171) underscores the critical importance of this finding in clinical decision-making, as children with unclear bowel wall layering face a dramatically elevated risk of adverse outcomes.
Intestinal gas distribution, encompassing both localized and diffuse patterns, emerged as the second independent predictor of high-risk HPVG. Compared to the absence of intestinal gas, diffuse gas distribution conferred a 65-fold increased risk (OR =65.077), while localized distribution increased risk approximately 7-fold (OR =6.978). HPVG is caused by the entry of gas into the portal venous system through a damaged intestinal mucosal barrier or via gas-producing bacteria in the portal venous circulation (3,19). When intestinal gas distribution is abnormal-whether localized to specific segments or diffused throughout multiple bowel segments-it suggests more extensive mucosal injury and impaired intestinal motility (20,21). Hosokawa et al. previously reported that the degree and distribution of portal venous gas on ultrasound correlate with clinical outcomes in children, with more extensive gas distribution associated with poorer prognosis Although our study shares with theirs a pediatric, ultrasound-based approach to the prognostication of portal venous gas, the two studies differ substantially: their analysis focused on the extent of the portal venous gas itself in a smaller cohort and did not incorporate bowel ultrasound features or a quantitative prediction model, whereas our study covered a broader etiological spectrum of HPVG, systematically assessed bowel wall and intraluminal gas findings, and constructed an internally validated nomogram. These differences in population, predictors, and methodology should be considered when comparing the results of the two studies (1). Similarly, Chen et al. demonstrated that the extent of pneumatosis intestinalis, a related finding, correlates with disease severity in NEC (10). Our findings align with and extend these observations, showing that the pattern of intestinal gas distribution serves as an independent predictor when combined with bowel wall assessment. The clinical implication is that sonographers and clinicians should systematically evaluate not only the presence of HPVG but also the overall distribution pattern of intestinal gas as part of a comprehensive risk assessment.
Although bowel peristalsis and ascites were significant in univariate analysis, they were not significant in the multivariate model. One possible explanation is that these variables are correlated with bowel wall layer and intestinal gas distribution and share common underlying pathophysiological mechanisms-for example, loss of peristalsis may be a downstream consequence of bowel wall ischemia, and ascites may reflect inflammatory responses secondary to intestinal injury-so that their prognostic information was largely captured by the retained predictors (22,23). However, this explanation remains speculative, and our data do not directly demonstrate these mechanisms; alternatively, the limited number of outcome events may have reduced the statistical power to detect their independent effects.
The nomogram we developed offers several advantages. First, we systematically evaluated multiple ultrasound parameters and clinical variables, used appropriate statistical methods for model development and validation, and employed bootstrap resampling to assess overfitting, the inclusion of DCA provides evidence of clinical utility beyond statistical performance. Second, the management of HPVG heavily relies on subjective clinical judgment and radiological interpretation now, with significant variations among different doctors and institutions. Our nomogram transforms this subjective assessment into an evidence-based quantitative risk prediction tool. For children identified as high-risk by the nomogram, clinicians can prioritize emergency surgical consultations and enhanced monitoring. For low-risk children, conservative treatment through close observation can be implemented. This approach is expected to reduce unnecessary surgical interventions in benign HPVG cases, while ensuring timely intervention for high-risk patients. Third, the model requires only two ultrasound parameters that are routinely assessed during abdominal ultrasound examinations, its simplicity makes it easy to apply in various clinical settings, including community hospitals that may not immediately provide pediatric surgical expertise.
Two additional points deserve emphasis regarding clinical application. First, although free peritoneal gas was excluded from the model because of its very low frequency, it remains a critical warning sign: the detection of free intraperitoneal gas on ultrasound should prompt immediate surgical evaluation irrespective of the nomogram score, and the nomogram should never be used to override this finding. Second, surgical decisions in our institution were made by pediatric surgeons on the basis of a comprehensive assessment of clinical presentation, laboratory findings, and conventional imaging rather than any single ultrasound feature; nevertheless, because bowel ultrasound findings may have contributed to surgical decision-making, the possibility of incorporation bias-whereby the predictors are partly incorporated into the outcome definition-cannot be excluded, and the model’s performance may be correspondingly overestimated.
Several limitations of this study should be acknowledged. First, the retrospective design inherently introduces potential selection and information biases. Selection bias may have enriched our cohort for children who were ill enough to undergo complete abdominal ultrasound, which could inflate the observed associations between ultrasound abnormalities and adverse outcomes; information bias arising from retrospective image review and feature classification may have led to misclassification, which would generally bias the estimated associations toward the null and reduce apparent model performance. Second, this was a single-center study conducted at a tertiary children’s hospital, which may limit generalizability to other clinical settings and populations, and no a priori sample size or power calculation was performed because all consecutive eligible cases over the study period were included. Third, the relatively small number of high-risk cases (n=20) limited the statistical power to identify additional predictors, rendered the estimated ORs and performance measures unstable—as reflected by the wide CIs—and made the model vulnerable to overfitting despite bootstrap internal validation; the regression coefficients should therefore be interpreted with caution. Fourth, the two predictors retained in the model-bowel wall layer and intestinal gas distribution-involve subjective interpretation, and formal assessment of interobserver agreement was not performed; although all images were reviewed by two experienced pediatric sonographers with consensus adjudication, the reproducibility of these findings across operators and institutions remains to be established. Fifth, because bowel ultrasound findings may have contributed to surgical decision-making, incorporation bias cannot be excluded, and model performance may be overestimated. Sixth, free peritoneal gas, while clinically important, was excluded from the model due to extreme data imbalance (2/107 cases); the model therefore does not capture this high-risk feature, and its presence should prompt surgical evaluation regardless of the nomogram score. Seventh, only children who underwent complete abdominal ultrasound were included; HPVG can also be identified on limited or dedicated examinations (such as right upper quadrant or pyloric ultrasound) (24), so our findings may not apply to cases detected on such focused scans. Finally, the model was only internally validated using bootstrap resampling; external validation in independent cohorts from different institutions and geographic regions is essential before widespread clinical adoption. Future prospective, multicenter studies with larger sample sizes, prespecified sample size estimation, formal interobserver reliability assessment, and external validation cohorts are needed to confirm our findings.
Conclusions
Loss of bowel wall stratification and abnormal intestinal gas distribution are independent predictors of high-risk HPVG in children. The ultrasound-based nomogram incorporating these two parameters demonstrates good discriminative ability (apparent AUC =0.899; optimism-corrected AUC =0.879) and provides a practical, rapid tool for risk stratification. This model may assist clinicians in identifying children with HPVG who require urgent surgical intervention. External validation in prospective, multicenter cohorts is warranted before widespread clinical adoption. From a practical standpoint, whenever HPVG is detected on ultrasound, clinicians should routinely assess bowel wall stratification and intestinal gas distribution: loss of bowel wall stratification or an abnormal (localized or diffuse) intestinal gas pattern should prompt urgent surgical consultation and intensified monitoring, whereas preserved wall layering without abnormal intestinal gas supports conservative management with close observation.
Supplementary
The article’s supplementary files as
Acknowledgments
None.
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. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of Tianjin Children’s Hospital (No. W-2026-030). Informed consent was waived for this retrospective study.
Footnotes
Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0541/rc
Funding: This work was funded by Tianjin Key Medical Discipline Construction Project (Grant No. TJYXZDXK-3-003B).
Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0541/coif). The authors have no conflicts of interest to declare.
Data Sharing Statement
Available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0541/dss
References
- 1.Hosokawa T, Yamada Y, Tanami Y, et al. Predictors of prognosis in children with portal venous gas detected by ultrasound. Med Ultrason 2019;21:30-6. 10.11152/mu-1727 [DOI] [PubMed] [Google Scholar]
- 2.Liebman PR, Patten MT, Manny J, et al. Hepatic--portal venous gas in adults: etiology, pathophysiology and clinical significance. Ann Surg 1978;187:281-7. 10.1097/00000658-197803000-00012 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Abboud B, El Hachem J, Yazbeck T, et al. Hepatic portal venous gas: physiopathology, etiology, prognosis and treatment. World J Gastroenterol 2009;15:3585-90. 10.3748/wjg.15.3585 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Lim YJ, Jung HK. Portal Vein Gas in Neonates: Predicting Clinical Treatment Outcomes From Ultrasound Findings. J Ultrasound Med 2022;41:2557-66. 10.1002/jum.15945 [DOI] [PubMed] [Google Scholar]
- 5.Furuta T, Fujiwara M, Motonaga T, et al. Ultrasound and computed tomography findings of hepatic portal venous gas associated with acute appendicitis in a paediatric patient: A case report. Ultrasound 2024;32:67-70. 10.1177/1742271X231195752 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Hosokawa T, Yamada Y, Tanami Y, et al. Difference in Detection Rate of Portal Venous Gas via Computed Tomography and Ultrasonography in Pediatric Patients. Ultrasound Q 2019;35:181-5. 10.1097/RUQ.0000000000000390 [DOI] [PubMed] [Google Scholar]
- 7.Boccatonda A, D'Ardes D, Tallarico V, et al. Gastrointestinal Ultrasound in Emergency Setting. J Clin Med 2023;12:799. 10.3390/jcm12030799 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Muchantef K, Epelman M, Darge K, et al. Sonographic and radiographic imaging features of the neonate with necrotizing enterocolitis: correlating findings with outcomes. Pediatr Radiol 2013;43:1444-52. 10.1007/s00247-013-2725-y [DOI] [PubMed] [Google Scholar]
- 9.Inarejos Clemente EJ, Barber I, Navallas Irujo M, et al. US for Evaluation of Acute Abdominal Conditions in Neonates. Radiographics 2023;43:e220110. 10.1148/rg.220110 [DOI] [PubMed] [Google Scholar]
- 10.Chen S, Hu Y, Liu Q, et al. Comparison of abdominal radiographs and sonography in prognostic prediction of infants with necrotizing enterocolitis. Pediatr Surg Int 2018;34:535-41. 10.1007/s00383-018-4256-y [DOI] [PubMed] [Google Scholar]
- 11.Hosokawa T. Discrepancy between patient prognosis and degree of portal venous gas detected using ultrasound. J Med Ultrason (2001) 2023;50:267-8. 10.1007/s10396-023-01294-y [DOI] [PubMed] [Google Scholar]
- 12.Ruscica A, Chen C, Ng L. Updates in pediatric ultrasound. Curr Opin Pediatr 2023;35:324-30. 10.1097/MOP.0000000000001244 [DOI] [PubMed] [Google Scholar]
- 13.Lin PC, Lin CY, Chang HY. Sonographic diagnosis of portal venous gas in necrotizing enterocolitis. Pediatr Neonatol 2022;63:93-4. 10.1016/j.pedneo.2021.06.019 [DOI] [PubMed] [Google Scholar]
- 14.Wang MQ, Lee MY, El Teo H. Ultrasound in the evaluation of necrotic bowel in children: A pictorial essay. Ultrasound 2019;27:207-16. 10.1177/1742271X18814864 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Yang L, Xu W, Li YW, et al. Value of abdominal ultrasound in the diagnosis of neonatal necrotizing enterocolitis and evaluation of disease severity. Zhongguo Dang Dai Er Ke Za Zhi 2016;18:108-12. 10.7499/j.issn.1008-8830.2016.02.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Garbi-Goutel A, Brévaut-Malaty V, Panuel M, et al. Prognostic value of abdominal sonography in necrotizing enterocolitis of premature infants born before 33 weeks gestational age. J Pediatr Surg 2014;49:508-13. 10.1016/j.jpedsurg.2013.11.057 [DOI] [PubMed] [Google Scholar]
- 17.He Y, Zhong Y, Yu J, et al. Ultrasonography and radiography findings predicted the need for surgery in patients with necrotising enterocolitis without pneumoperitoneum. Acta Paediatr 2016;105:e151-5. 10.1111/apa.13315 [DOI] [PubMed] [Google Scholar]
- 18.Cuna AC, Reddy N, Robinson AL, et al. Bowel ultrasound for predicting surgical management of necrotizing enterocolitis: a systematic review and meta-analysis. Pediatr Radiol 2018;48:658-66. 10.1007/s00247-017-4056-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Nelson AL, Millington TM, Sahani D, et al. Hepatic portal venous gas: the ABCs of management. Arch Surg 2009;144:575-81; discussion 581. 10.1001/archsurg.2009.88 [DOI] [PubMed] [Google Scholar]
- 20.Chen J, Mu F, Gao K, et al. Value of abdominal ultrasonography in predicting intestinal resection for premature infants with necrotizing enterocolitis. BMC Gastroenterol 2022;22:524. 10.1186/s12876-022-02607-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Hwang M, Tierradentro-García LO, Dennis RA, et al. The role of ultrasound in necrotizing enterocolitis. Pediatr Radiol 2022;52:702-15. 10.1007/s00247-021-05187-5 [DOI] [PubMed] [Google Scholar]
- 22.Kastenberg ZJ, Sylvester KG. The surgical management of necrotizing enterocolitis. Clin Perinatol 2013;40:135-48. 10.1016/j.clp.2012.12.011 [DOI] [PubMed] [Google Scholar]
- 23.Robinson JR, Rellinger EJ, Hatch LD, et al. Surgical necrotizing enterocolitis. Semin Perinatol 2017;41:70-9. 10.1053/j.semperi.2016.09.020 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Butler K, Patton T. Portal venous gas: A benign finding in pyloric stenosis? JPGN Rep 2025;6:423-6. 10.1002/jpr3.70051 [DOI] [PMC free article] [PubMed] [Google Scholar]