Definition and diagnosis of necrotising enterocolitis

The Paediatric Cardiac Critical Care Consortium defines necrotising enterocolitis by the modified Bell’s criteria. ^{Reference Bell, Ternberg and Feigin5} Adjuncts to diagnosis not included in Bell’s criteria include serum markers for inflammation and infection (i.e. white blood cell count, c-reactive protein) and alternative imaging techniques such as abdominal ultrasound and the use of near infra-red spectroscopy. ^{Reference Gao, Yi and Mao6,Reference Howarth, Banerjee, Leung and Aladangady7 Reference Kessler, Hau and Kordasz8} The role of biomarkers in necrotising enterocolitis diagnosis, treatment guidance, and prognostication is expanding ^{Reference Kelleher, McMahon and James1} yet, to date, universally adopted standard biomarkers do not exist in routine clinical practice.

Necrotising enterocolitis pathophysiological features and risk factors in infants with CHD do not completely overlap with those of necrotising enterocolitis in premature neonates. Therefore, applying necrotising enterocolitis diagnostic criteria (either Bell’s criteria or adjunct diagnostic criteria) to the CHD population is complex. Traditional Bell’s criteria are not tailored to accommodate the unique risk factors and presentation of CHD patients, such as impaired mesenteric ischaemia from dynamic and transient episodes of hypoperfusion, and high abdominal venous pressures limiting organ perfusion.

Diagnostic imaging:

Abdominal radiography has long been a mainstay of the diagnosis of necrotising enterocolitis. While pneumatosis intestinalis and/or portal venous gas are considered pathognomonic, more subtle signs may be noted on radiographs. These include dilated loops of bowel, paucity of bowel gas, and gas-filled loops of bowel that are unchanged on follow-up examinations. ^{Reference Neu and Walker9} Because radiographs can be normal in the earliest stages and findings may be equivocal, serial examinations are often warranted.

Abdominal ultrasound can be useful when radiographs are non-diagnostic. Abdominal ultrasound can be used to differentiate intraluminal stool from pneumatosis, which can be difficult to distinguish on radiographs. ^{Reference Kallis, Roberts and Aronowitz10} Abdominal ultrasound may also have higher sensitivity for portal venous gas, pneumoperitoneum, and ascites; however, specificity is inconsistent. ^{Reference Hwang, Tierradentro-García, Dennis and Anupindi11} Though first described over 40 years ago, adoption of abdominal ultrasound for diagnosis of necrotising enterocolitis remains limited due to resource availability, technical skills, and a lack of diagnostic consensus among neonatologists, intensivists, radiologists, and surgeons. ^{Reference Kallis, Roberts and Aronowitz10,Reference Hwang, Tierradentro-García, Dennis and Anupindi11} Concern has also been raised about high false positive rates, though this has not been well documented. ^{Reference Tracy, Lazow and Castro-Aragon12}

Across the necrotising enterocolitis working group, the availability of abdominal ultrasound was variable and inconsistent, used at five of the nine sites to some extent. According to general surgeons from the group sites, abdominal ultrasound is not used as a determinant for surgical intervention. Overall, the availability and consistency of abdominal ultrasound were variable, but abdominal ultrasound can serve as a highly sensitive tool for individuals at high risk or with suggestive clinical presentation, even in the absence of radiographic findings.

Anatomy

In patients with CHD, necrotising enterocolitis is a multifactorial disease process distinct from preterm necrotising enterocolitis. ^{Reference Neu, Modi and Caplan13,Reference Neu14} CHD-associated necrotising enterocolitis has a mixture of inflammatory and vascular injury involving mesenteric hypoperfusion and/or ischaemia. Proposed mechanisms for abnormal perfusion in CHD include diastolic steal and flow reversal in the abdominal aorta. Multiple types of CHD can lead to necrotising enterocolitis, including a patent ductus arteriosus with significant left-to-right shunting leading to systemic hypoperfusion, generalised hypoxaemia in cyanotic CHD, compromised diastolic gut perfusion pressures restricting oxygenated blood flow in patients with ductal dependent CHD (i.e. coarctation of the aorta), or lesions such as complete atrioventricular canal defect with accompanying atrioventricular valve regurgitation leading to pulmonary over circulation and compromised systemic output. ^{Reference Kashif, Abuelgasim, Hussain, Luyt and Harky15}

^{Reference Becker, Hornik and Cotten16} Newborns with truncus arteriosus and lesions with ductal-dependent pulmonary blood flow also carry significantly increased risk of developing necrotising enterocolitis. ^{Reference Maheshwari, Roychaudhuri, Grewal, Vijayashankar and Lavoie17,Reference McElhinney, Hedrick and Bush18} In premature or very low birth weight infants, the highest risk of necrotising enterocolitis development is associated with the presence of atrioventricular canal defect. ^{Reference Kashif, Abuelgasim, Hussain, Luyt and Harky15,Reference Becker, Hornik and Cotten16} Despite this, caution should be made when comparing incidences of necrotising enterocolitis among specific cardiac lesions. In a systemic review and meta-analysis of cardiogenic necrotising enterocolitis in infants with CHD by Asztalos et al, evidence for the incidence of cardiogenic necrotising enterocolitis in specific populations of cardiac anomalies was extremely limited and subject to bias. ^{Reference Asztalos, Hill, Nash, Schachtner and Palm2}

Within the necrotising enterocolitis working group, pre-operative CHD lesions that were deemed high risk of developing necrotising enterocolitis included ductal dependent systemic blood flow lesions as well as truncus arteriosus, independent of primary anatomy. This was based on institutional experience despite literature reviews on the topic. In addition, the working group’s consensus is that patients who have ductal stents and pulmonary arterial bands, populations with little data as it relates to necrotising enterocolitis, tend to have high rates of necrotising enterocolitis at their sites.

Vasoactive infusions

There is limited literature regarding how vasoactive support in infants with CHD impacts the development of cardiac necrotising enterocolitis. Hemodynamic abnormalities and stressors of gastrointestinal health are common occurrences for infants hospitalised for cardiac surgery. Poor gut perfusion and an imbalance between vasodilatory and vasoconstrictor signalling in the mesenteric microvasculature are factors that can be influenced by vasoactive medications and further exacerbate an at-risk mesentery vascular bed. Episodes stressing the mesentery, including feeds during hypoperfusion states and episodes of diastolic runoff, then add to the risk of developing necrotising enterocolitis. For example, early post-operative hypoperfusion, independent of a low cardiac output state, makes an infant more susceptible to developing necrotising enterocolitis days later in their course. ^{Reference McElhinney, Hedrick and Bush18,Reference Casals and Spaeder19} There is limited literature on vasoactive infusions and necrotising enterocolitis in patients with CHD.

Within the necrotising enterocolitis working group, most centres feed patients while on a vasopressin infusion but at low doses (Table 1). Most of the centres do not feed patients while on norepinephrine infusions but all centres do feed patients while on low-dose epinephrine infusions. With regards to dopamine, three of the nine centres involved in the working group use dopamine and each of those centres included centres with high and low rates of necrotising enterocolitis, while four of those five centres fed patients while on dopamine infusion. Instead of attempting to reach a goal nutritional feed rate while the patient is on vasopressor support, all centres with low rates of necrotising enterocolitis hold feeds at a lower-than-goal rate.

Table 1 Number of centres that feed while on vasoactive drips

Transfusions

There is conflicting literature assessing the association between red blood cell transfusions and necrotising enterocolitis in the premature infant population. ^{Reference Garg, De Plaen and Christensen20} The exact mechanism of transfusion-associated necrotising enterocolitis or acute gut injury is unknown and may result from an alteration in intestinal circulation [e.g. dynamic balance between vasoconstrictive and vasodilatory mechanisms (nitric oxide)] with subsequent tissue hypoxia and reoxygenation that may stimulate gut inflammation (predominately macrophages) and mucosal barrier damage. ^{Reference Maheshwari, Patel and Christensen21,Reference Salem and Patel22} Given this data, approaches to necrotising enterocolitis mitigation have included preventing intestinal circulatory stress during times of transfusions (holding feeds).

Within the working group, two of the nine centres withheld enteral feeds for the duration of the red blood cell transfusion in infants deemed to be at high risk of developing necrotising enterocolitis; both were centres with low incidence of necrotising enterocolitis. The effectiveness of withholding enteral feeding periprandial should be further evaluated in neonates with CHD during red blood cell transfusion. The role of red blood cell transfusion in necrotising enterocolitis aetiology for infants with complex CHD is unclear, but a conservative treatment approach may include withholding enteral feeds during transfusion. ^{Reference Martini, Beghetti and Annunziata23}

Anticoagulation

Data are limited linking anticoagulation strategy with necrotising enterocolitis in CHD, either as a risk or protective factor. No consensus regarding the use of systemic anticoagulation as a risk factor for necrotising enterocolitis was reached by the working group, owing to the multifactorial pathophysiology of necrotising enterocolitis in CHD. Several centres employ treatment guidelines that seek to lower the risk of systemic thromboembolism in general, which in theory ameliorates risk of mesenteric thrombosis and ischaemic gut injury.

The protocolized use of anticoagulation with therapeutic-dosed enoxaparin to prevent systemic thromboembolic disease in patients with indwelling venous central lines is employed at five of the nine centres. While this protocol is for thrombus prophylaxis, and independent of necrotising enterocolitis assessment, four of these five centres were high performing for necrotising enterocolitis. One centre holds anticoagulants when a patient has hematochezia or true necrotising enterocolitis given concern that it increases risk for worsened bleeding after an ischaemic bowel injury.

While all centres in our necrotising enterocolitis working group used laboratory values (e.g. low-molecular weight heparin assays) to assess the effectiveness of their anticoagulation; these levels are not used to guide feeding strategies. In all participating centres, routine anticoagulation was achieved with either heparin infusion or subcutaneous enoxaparin.

Nil per os time and antibiotics

Despite necrotising enterocolitis in infants with cardiac disease having different pathophysiology and outcomes compared to necrotising enterocolitis in premature infants, there are currently no evidence-based treatment guidelines specific to this population and management plans are often extrapolated from the neonatal literature. “Suspected necrotizing enterocolitis” is as frequent as confirmed necrotising enterocolitis ^{Reference Schuchardt, Kaufman, Lucas, Tiernan, Lujan and Barrett24} and most centres will treat with broad-spectrum antibiotics with coverage for enteric Gram-negative and anaerobic organisms, as well as keeping the patients nil per os for 24–48 hours. For medical necrotising enterocolitis, treatment courses of both nil per os time and antibiotics are most often 7–10 days, which is consistent with neonatal guidelines, though more recent studies in premature infants suggest earlier refeeding may be beneficial. ^{Reference Patel, Wilson, Brennan, Lesher and Ryan25}

Surgical intervention

The proportion of patients with CHD and necrotising enterocolitis that progresses to surgical intervention for necrotising enterocolitis is estimated to be 11% among term infants and 21% among preterm infants. ^{Reference Asztalos, Hill, Nash, Schachtner and Palm2} Compared to premature patients, those with CHD-associated necrotising enterocolitis had lower incidences of perforation, need for bowel operation, strictures, need for a stoma, sepsis, and short bowel syndrome. ^{Reference Pickard, Feinstein, Popat, Huang and Dutta26} Conflicting data suggests that intestinal necrosis was present with greater frequency intraoperatively in infants with CHD and necrotising enterocolitis in comparison to infants without CHD. ^{Reference Diez, Tielesch, Weiss, Halbfass, Müller and Besendörfer27} This offers insight for earlier consideration for surgical intervention for patients with CHD. ^{Reference Carr and Gadepalli28}

Surgical treatments for necrotising enterocolitis include bedside peritoneal drain placement or standard laparotomy. The primary goals of surgical intervention in necrotising enterocolitis are to control enteric spillage and/or resect necrotic intestine while maximising the length of viable intestine. Ideally, the optimal time for intervention would be after the onset of severe ischaemia but before intestinal perforation and/or progression of physiologic derangement. ^{Reference Robinson, Rellinger and Hatch29}

Determining optimal treatment strategy (surgical intervention vs. medical management) based on patient comorbidities remains controversial. Currently, the absolute criterion for operative intervention is evidence of pneumoperitoneum on radiography. ^{Reference Robinson, Rellinger and Hatch29} Clinical deterioration despite maximum medical therapy tends to guide surgical intervention.

Within the working group, the decision for surgical intervention is most often based on pneumatosis on abdominal X-ray or continued bleeding per rectum. Among our necrotising enterocolitis working group, with input from general surgeons, the greatest determinant of surgical versus medical necrotising enterocolitis, aside from clinical examination, is serial laboratories. Cohort studies have identified severe thrombocytopenia as an indicator of underlying intestinal necrosis ^{Reference Kenton, O’donovan and Cass30} and proposed criteria for metabolic derangements including acidosis and low bicarbonate may predict timely operative intervention. ^{Reference Tepas, Sharma, Leaphart, Celso, Pieper and Esquivia-Lee31} However, this has been limited in its generalizability, due to the lack of evidence from prospective or randomised trials. ^{Reference Robinson, Rellinger and Hatch29} Abdominal ultrasound was not suggested as a determinant for operating.

Formula/Nutrition

There is a lack of consensus about optimal nutrition progression for post-cardiac surgery patients post-hematochezia or necrotising enterocolitis. Limited data exist to guide how best to advance nutrition support for high-risk infants in the post-operative period, which results in wide variability between centres and possibly detrimental delays in nutrition advancement. For those at lower risk for necrotising enterocolitis, the decision to withhold feeds for cases of hematochezia must be balanced by considering the other aetiologies of hematochezia, such as cow’s milk allergy.

After an occurrence of necrotising enterocolitis, some nutrition strategies employed are nil per os with parenteral nutrition, a switch in feed type (breastmilk/formula), selection of a different formula or fortifier (standard, semi-elemental, elemental), altered feeding tube placement (gastric/transpyloric), adjusted caloric density (at feed initiation and goal), change in feeding regimen (bolus/continuous), or modified goal for enteral volume (mL/kg). These modifiable factors are common practice despite no clear relationship between enteral feeding patterns (day of initiation of feeds, feeding velocity, caloric density, formula type) and necrotising enterocolitis incidence in the post-operative period. ^{Reference Iannucci, Oster and Mahle32}

Without the evidence of how best to advance feeds after necrotising enterocolitis, an institution-based feeding protocol targeting calories with standard monitoring is a reasonable approach to reduce feeding complications and incidence of necrotising enterocolitis. ^{Reference Petrillo-Albarano, Pettignano, Asfaw and Easley33,Reference Braudis, Curley and Beaupre34} Balanced against a conservative approach of limiting feeds is the risk of excessive nutrition delay, as this may contribute to faltering growth. Utilising a nils per os approach requires incremental advancement of energy delivery (both parenteral and enteral nutrition), which can take several days, contributing to the accumulation of caloric deficits. Having a standardised feeding algorithm is one way to facilitate safe feeding advancement.

There is some literature to suggest that breastmilk can be a protective factor against necrotising enterocolitis among patients with CHD. ^{Reference Davis and Spatz35,Reference Elgersma, McKechnie and Schorr36} Within the necrotising enterocolitis working group, breastmilk was the preferred feeding. In addition, two centres with low incidence of necrotising enterocolitis have a post-op goal of only unfortified breastmilk. The use of elemental formulas was reported in two centres as a nutritional strategy after necrotising enterocolitis. ^{Reference Lapillonne, Matar, Adleff, Chbihi, Kermorvant-Duchemin and Campeotto37} The osmotic loads of amino-acid-based formulas are significantly higher, which can contribute to feeding intolerance. The pros and cons of using these formulas for patients with CHD after necrotising enterocolitis need further investigation.

Deciding whether to remove mother’s milk or modify her diet in the setting of suspected cow’s milk intolerance or allergy remains challenging. There is no consistent guideline on how to handle this recurring issue and when to consider developing a feeding protocol.