#434 - [Journal Club] - 📌 Can Ultrasound Help Us Better Identify Surgical NEC?

In this Journal Club episode, Ben takes the lead and reviews a prospective cohort pilot study from the Archives of Disease in Childhood examining whether combining abdominal ultrasound with plain radiography can improve surgical risk stratification in neonates with suspected NEC. With mortality remaining as high as 20–40% and diagnosis still heavily reliant on clinical judgment, the stakes couldn't be higher. Ben walks through the study's unsupervised clustering approach, explaining how adding ultrasound data to X-ray findings produced a more than six-fold difference in the odds of surgery between risk groups — something X-ray alone simply couldn't achieve. Tune in to hear why dynamic ultrasound features like peristalsis, ascites, and bowel perfusion may be the missing piece in your NEC diagnostic toolkit!

----

Combining abdominal ultrasound and radiography for surgical risk stratification in necrotising enterocolitis: a prospective cohort pilot study. Priyadarshi A, Angiti R, Chabra S, McAdams R, Webb A, Badawi N, Hinder MK, Tracy MB.Arch Dis Child Fetal Neonatal Ed. 2026 Mar 5:fetalneonatal-2025-329960. doi: 10.1136/archdischild-2025-329960. Online ahead of print.

Transcript

Ben Courchia (00:00.47) Hello, everybody. Welcome back to The Incubator Podcast. We are back for another episode of Journal Club. Daphna, good morning. How are you?

Daphna Yasova Barbeau (00:07.009) Good morning. Doing well.

Ben Courchia (00:00.47) Yeah, we had a great episode yesterday — a very long one on the recent AAP report on HIE. Please go check that out. Today it's my turn, and I think I'll take a slightly lighter route and give people a bit of a break. I found this very interesting paper in the Archives of Disease in Childhood, Fetal and Neonatal Edition, called "Combining Abdominal Ultrasound and Radiography for Surgical Risk Stratification in Necrotizing Enterocolitis: A Prospective Cohort Pilot Study." This is a paper I've been waiting for.

Daphna Yasova Barbeau (00:48.225) Love it.

Ben Courchia (00:56.91) The first author is Archana Priyadarshi with last author Mark B. Tracy. This comes to us from Australia.

The background sets up the problem well. Mortality from necrotizing enterocolitis remains high at 20 to 30%, sometimes up to 40%. Survivors often require surgical intervention and can face short bowel syndrome and adverse neurodevelopmental outcomes. Plain abdominal radiography is the first-line investigation for suspected NEC. The presence of pneumatosis intestinalis or portal venous gas on radiographs has traditionally defined the diagnosis, although the diagnostic exclusivity of these findings has recently been questioned. In the absence of a reliable biomarker, diagnosis really depends on clinical judgment. If you see pneumatosis or portal venous gas, the diagnosis is largely made for you — but many times you don't see that, and then the question becomes: what am I looking at?

Because the disease can progress rapidly and the consequences of misdiagnosis are severe, NEC is more frequently suspected than confirmed. The authors therefore bring up the role of ultrasound, which provides a complementary, radiation-free modality that allows bedside assessment of bowel wall thickness, vascularity, and peristalsis — features critical for assessing the presence and progression of NEC. In Australia and New Zealand, clinicians routinely use point-of-care ultrasound to evaluate the heart, brain, and lungs, and there is growing interest in extending that capability to abdominal imaging for early detection and monitoring of NEC.

The objective of the study: can combining abdominal ultrasound and abdominal X-ray — a multimodal imaging approach — improve diagnostic precision and surgical risk stratification in neonates with suspected NEC?

Daphna Yasova Barbeau (02:27.413) Mm-hmm.

Ben Courchia (02:50.644) This was a prospective pilot observational cohort study of 67 neonates admitted between 2021 and 2025 to either a tertiary perinatal or surgical neonatal intensive care unit with a clinical diagnosis of suspected NEC. Following consent, all neonates underwent abdominal X-ray — standard of care — followed by POCUS within four hours of the initial X-ray. Data collected included gestational age, birth weight, type of treatment (medical or surgical), and outcome at discharge.

Case definitions were determined prior to the study beginning. At study completion, each case was categorized as definitely NEC, suspected NEC, or not NEC — three categories — based on final clinical diagnosis, operative findings where applicable, radiographic evidence of pneumatosis or portal venous gas, and overall clinical course. Neonates treated medically with nonspecific radiographic findings were designated as "suspected NEC." That category captures where most of us live when it comes to this diagnosis. Based on final outcome, cases were then classified as either surgical or nonsurgical.

In terms of the statistical approach — I know we often skip this, but I think in this case we shouldn't. The core question is straightforward: can imaging findings automatically sort patients into higher and lower surgical risk groups, without the investigators deciding in advance who belongs where?

To do this, the authors used an approach called unsupervised clustering. Rather than testing a pre-specified hypothesis, the algorithm looks at all imaging variables across all patients and asks whether the data naturally separates into two groups — crucially, without knowing who actually went to surgery. It takes all the initial ultrasounds and X-rays and makes its predictions. Once the two groups are formed, you then look back and ask: did the patients in one group go to surgery more than the other?

The authors ran this process twice — once feeding in only X-ray findings, and once feeding in X-ray and ultrasound findings combined. That head-to-head comparison is the central test of the paper. Clinical concordance with actual surgical outcome was then quantified using logistic regression, reported as odds ratios, and by directly comparing surgical rates between the two groups. To identify which specific imaging features were driving the separation, the authors applied principal component analysis to the combined imaging dataset — essentially asking: out of all the variables fed into the model, which ones are doing the heavy lifting? The result is a ranked list of imaging features by their contribution to group separation, which tells you not just that the combined model worked better, but why.

Daphna Yasova Barbeau (06:45.719) That was excellent — very helpful.

Ben Courchia (07:03.702) Cohort characteristics: 67 neonates with clinically suspected NEC were enrolled. Mean gestational age was 29.5 weeks, mean birth weight 1,400 grams — always a sobering reminder that bigger babies do get NEC.

Daphna Yasova Barbeau (07:06.077) Mm-hmm. And that's the worst — when they're doing fine and then they get into trouble.

Ben Courchia (07:13.582) 88% were born preterm and 18% were growth restricted. 49% underwent surgery and 10% died before discharge.

At discharge, 23 of the 67 — 34% — were diagnosed with definite NEC, of whom 16 had operative confirmation. 9 babies, 13%, were classified as suspected NEC. 35 babies, 52%, were categorized as non-NEC, including those considered to have feeding intolerance. So a nice mix: 34% definite NEC, 13% suspected, 52% non-NEC.

In terms of imaging findings: ultrasound scans were categorized as reassuring in 40% of the cohort and non-reassuring in 60%. X-rays were considered reassuring in only 4 infants — 6% — and non-reassuring in 63 infants, 94%. You can see the difference: ultrasound was quite discriminatory, really separating the cohort 40/60. When you're watching peristalsis in real time, that matters. The fact that 94% of X-rays were non-reassuring across the entire cohort — including patients who ultimately just had feeding intolerance — speaks directly to the well-known low specificity of plain radiography.

Among infants with non-NEC feeding intolerance, ultrasound findings were reassuring in 100% of cases, whereas 83% of X-rays in that same group were non-reassuring. In contrast, both imaging modalities were non-reassuring in all infants with definite NEC.

Now for the cluster analysis. In the X-ray-only model, the proportion requiring surgery did not differ significantly between clusters: 58.8% versus 39.4%, p-value 0.11. In the combined X-ray and ultrasound model, by contrast, two distinct groups emerged with markedly different surgical rates: 78.3% versus 34%, corresponding to an odds ratio of approximately seven. That contrast is the central finding of the paper. X-ray data alone could not generate clusters that meaningfully tracked surgical outcome. Adding ultrasound data produced a more than six-fold difference in the odds of surgery between the two clusters — a clinically substantial separation the X-ray model simply could not achieve.

Looking at the principal component analysis of the combined imaging dataset: the first two principal components together explained 70% of the total variance, meaning most of the discriminating signal is captured in a manageable number of dimensions. The biplot reveals a clear separation between neonates who underwent surgery and those managed medically. Complex ascites on ultrasound, absent peristalsis, abnormal bowel perfusion, and abnormal bowel gas pattern were the features most strongly associated with the surgical cluster. These are features many of us already weigh heavily at the bedside — the PCA formalizes that clinical intuition and confirms that it is these ultrasound-derived dynamic findings — peristalsis, ascites, perfusion — rather than the X-ray, that carry most of the discriminating weight.

The conclusion: abdominal ultrasound combined with radiography improved the identification of neonates at high surgical risk from NEC. Ultrasound provided valuable dynamic functional information that complemented static radiograph findings and enhanced diagnostic precision. Broader adoption of multimodal imaging and standardized reporting may support more accurate diagnosis, targeted management, and improved outcomes for neonates with NEC. For me personally — I often do ultrasound on these babies when I'm worried, but I think this paper makes the case for doing it every time there's a question.

Daphna Yasova Barbeau (12:26.016) Yeah, very hopeful. We need something. Even when we get the diagnosis right, the outcomes are poor. But if we don't get it right, the outcomes are catastrophic. I wonder — for now, is using ultrasound to rule babies out the better use versus using it to rule babies in?

Ben Courchia (13:01.012) I think both. But the biggest practical challenge I see is that peristalsis is easy to assess on POCUS — you put the probe on the belly, wait a bit, and watch. But looking at gut perfusion, assessing the vessels, identifying ascites — that's a bit more involved. I worry that people will say, I don't know how to do that, and just assess peristalsis and call it a day. But that's not really what this study is showing you. The good news is those additional skills are not hard to learn — it just takes some effort.

Daphna Yasova Barbeau (13:45.804) And I think peristalsis is actually the least sensitive finding — we see absent peristalsis in ileus too, for example, not just NEC.

Ben Courchia (13:56.674) Exactly — and there's a difference between any peristalsis and normal peristalsis. Ascites is always very telling and something you can easily miss on X-ray. Surgeons often report opening the abdomen and finding significant ascites. So just learn to look for those fluid pockets.

Daphna Yasova Barbeau (14:18.424) All right, buddy. Thanks for that.

Ben Courchia (14:24.608) We'll see you guys tomorrow. Thank you, Daphna.