Bowel ultrasound for predicting surgical management of necrotizing enterocolitis: a systematic review and meta-analysis

Necrotizing enterocolitis (NEC) represents a significant cause of neonatal morbidity and mortality especially in preterm infants. It is estimated that about 5% to 14% of infants weighing <1500 g at birth develop NEC and about 25% to 40% die because of it [1, 2, 3]. Because NEC can rapidly progress to intestinal perforation, peritonitis and shock, timely diagnosis and treatment are essential for better outcomes [4]. Diagnosis of NEC is made based on clinical presentation, laboratory tests and imaging. The role of imaging is important to confirm the diagnosis and follow disease progression, but interpretation can also be challenging as normal or nonspecific findings do not rule out NEC entirely [5]. Initial treatment – which involves bowel rest, intravenous fluids, broad-spectrum antibiotics and other supportive measures – can be ineffective in controlling disease progression in some infants, and surgical intervention with either peritoneal drainage or bowel resection may be required [6, 7]. The transition from medical to surgical care represents a critical decision-making point in NEC management that can have important implications for clinical outcomes.

Traditionally, plain abdominal radiography is the imaging modality of choice for evaluating infants for NEC [8]. However, abdominal radiographs are challenging to interpret due to wide variability and significant overlap between radiographic signs of NEC and other intestinal pathology [9]. Abdominal radiographs are also limited in characterizing disease progression; free air, which radiographs can detect reasonably well, is already a late finding that indicates intestinal perforation. There is also lack of evidence regarding optimal frequency, use of one versus two views, and frequency and duration of radiographic imaging for monitoring disease progression in NEC.

Bowel ultrasound (US) has been reported as a useful adjunct to radiography for evaluating NEC [9, 10]. Advantages of bowel US include real-time assessment of peristalsis, vascular perfusion, bowel wall thickening and abdominal fluid [11]. Small, observational studies suggest that additional details provided by bowel US may be helpful for earlier identification of high-risk infants with NEC who may benefit from more aggressive treatment, including surgery. It is unclear from the literature, however, which bowel US findings provide the most useful prognostic information to help guide clinical decision-making. This systematic review and meta-analysis summarize evidence from existing literature and identify bowel US findings associated with surgical management or death in infants with NEC.

This systematic review was reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statements and an ethical review was not required. The protocol for this systematic review was registered on PROSPERO (doi: CRD42017068467) and is available in full on its website (www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42017068467).

In this meta-analysis, we identified several bowel US characteristics associated with poor outcomes in infants with NEC. Bowel US features most strongly associated with surgery or death were free air, absent peristalsis, complex ascites and focal fluid collection. Other features with more moderate association with surgery or death included bowel wall thinning, increased bowel wall echogenicity, bowel wall thickening, absent perfusion and dilated bowel. We also identified bowel US findings not associated with surgical NEC or death, including portal venous gas, increased bowel perfusion and simple ascites. Together, these findings suggest that the addition of bowel US in evaluating infants with NEC may help identify infants at high-risk for poor outcomes.

The current standard for imaging evaluation of infants with suspected NEC is plain abdominal radiography. However, the use of US as an imaging adjunct to radiography has increased in recent years. Initial case reports from the 1980s described the ability of US to detect pneumatosis and portal venous gas in infants with NEC [24, 25, 26]. Since then, use of US in NEC has extended to include evaluation of bowel wall, bowel perfusion and abdominal fluid [9]. Bowel US has been used as an adjunctive imaging tool for diagnosis in infants with clinical suspicion for NEC but whose plain abdominal radiographs show nonspecific findings [27]. Bowel US has also been used for follow-up of infants already diagnosed with NEC to monitor disease progression and guide clinicians regarding management, including surgery [9, 28]. The majority of these studies, however, were small single-center studies. In this systematic review, we combined data from these different studies in a meta-analysis and identified bowel US findings associated with surgery or death. These bowel US findings may provide valuable prognostic information and can be used to support more aggressive management including surgery.

Of the 13 bowel US findings included in this review, only 3 – free air, pneumatosis and portal venous gas – can also be accurately assessed by plain radiography. Although other findings – including ascites, bowel wall thickening and nonviable bowel – can be insinuated from serial radiographs, bowel US is more capable of characterizing these findings in greater detail. We found that US evidence of free air, but not of pneumatosis or portal venous gas, was associated with increased risk of surgery or death. Pneumatosis was also associated with surgery, while portal venous gas was not – although confidence interval values indicated a trend toward positive association. Some studies have suggested that portal venous gas was associated with worse outcomes including need for surgery [29, 30], although a more recent prospective study found that clinical severity is more important than portal venous gas in predicting poor outcomes [31]. Taken together, these results are consistent with current clinical practice regarding radiographs in NEC and likely reflect similar interpretation given to portal venous gas, pneumatosis and free air whether demonstrated on bowel US or radiographs.

Several bowel US features used to evaluate NEC included assessment of the bowel wall. The seminal study by Faingold et al. [28], in which bowel US was performed in 30 neonates with proven or suspected NEC and 30 normal neonates as controls, provided reference values for normal bowel wall thickness and echogenicity that have been widely used in other studies. Faingold et al. [28] also performed color Doppler interrogation of the bowel wall and assessed bowel perfusion as normal, increased or absent perfusion based on subjective assessment of flow patterns and objective measurement of color Doppler signals or “dots” per square centimeter. In this review, bowel wall thickening and thinning, increased bowel wall echogenicity, absent perfusion and absent peristalsis were all associated with surgery or death of infants with NEC. These bowel wall findings are thought to represent a continuum of changes that reflects worsening severity of NEC [9]. In the initial stages, bowel wall thickness, echogenicity and perfusion are increased due to ongoing intestinal inflammation and mucosal edema. As NEC progresses, bowel wall thinning, absent perfusion and absent peristalsis become more predominant and are highly suggestive of impending bowel perforation. The dynamic changes in bowel wall findings, which cannot be adequately evaluated by radiography, provide a distinct advantage for bowel US in characterizing disease progression of NEC before perforation occurs.

US is also superior to radiography in characterizing fluid in the abdomen. In this review, US findings of complex ascites and focal fluid collections were associated with surgery or death, while simple (anechoic) ascites was not. This is consistent with other studies that found that free, anechoic intraperitoneal fluid can be a normal finding in neonates [28], whereas ascitic fluid with echoes or septations is more suggestive of pus or intestinal contents that indicate perforation [32, 33].

It is important to note that careful consideration of imaging together with clinical and laboratory findings is important in assessing infants for NEC. For example, although positive findings in imaging such as portal venous gas or pneumatosis are strongly suggestive of NEC, other possibilities such as inadvertent injection of air through an umbilical venous catheter [34] or cow milk protein allergy [35] should also be considered, especially when clinical and laboratory findings are otherwise reassuring. Likewise, a normal or equivocal imaging finding by itself cannot rule out NEC entirely, especially when clinical and laboratory findings are strongly suggestive of it.

This meta-analysis has several limitations. The majority of included studies were retrospective observational cohorts, which may overestimate the strength of association between US findings and outcomes in NEC. The wide confidence intervals in some point estimates also suggest small sample size and lack of homogeneity across studies. Publication bias, where only positive studies are reported in the literature, could also have influenced the results of this review, although we sought to systematically search the gray literature for unpublished studies. There was also considerable variation among the different studies. As a result, we were often only able to analyze data from at most seven studies at once, despite identifying 11 studies of interest. Some studies reported on bowel US performed in infants with suspected NEC, while others included only infants with definite NEC (Modified Bell’s stage 2–3). In some studies, bowel US was part of routine evaluation for NEC, while in others bowel US was used selectively per clinical discretion. Although variation among the different studies is an important limitation and potential source of bias, such differences also closely reflect clinical practice. Indications for surgical intervention were also not clearly delineated and likely varied between studies. A potential bias exists in that US results may have contributed to decisions to perform surgery. Comparison to objective surgical or pathological findings was not possible because these findings were not reported in the source studies. Studies were also conducted over the course of 30 years during which US capability has significantly changed, with more recent machines capable of higher spatial and contrast resolution and cine captures. Included studies also analyzed the association of each bowel US finding with surgery or death separately. This does not reflect clinical practice where US findings are analyzed and interpreted together. Thus, the prognostic value of one bowel US finding by itself, or of two or more bowel US findings combined, cannot be determined.

This meta-analysis identified several bowel US findings associated with surgery or death in necrotizing enterocolitis including free air, absent peristalsis, absent perfusion, complex ascites, focal fluid collection, bowel wall thinning and thickening, increased bowel wall echogenicity and dilated bowel. We also identified bowel US findings not associated with surgical necrotizing enterocolitis or death, such as portal venous gas, increased bowel perfusion and simple ascites. We conclude that bowel US may be used for early identification of high-risk infants with necrotizing enterocolitis who may benefit from more aggressive treatment, including surgery. Future studies are needed to determine (1) the optimal timing and frequency of bowel US that would be most useful in the diagnostic pathway for necrotizing enterocolitis and (2) whether the addition of bowel US in necrotizing enterocolitis evaluation would improve outcomes.