Despite important advances in neonatal intensive care medicine over the last years, necrotising enterocolitis (NEC) remains the most common gastrointestinal emergency affecting 1–5% of all infants admitted to a NICU and approximately 7% of preterm infants <1500 g with a mortality rate of 10–30%.1 2 Infants requiring surgical treatment (20–40%) due to delayed diagnosis and/or extensive disease, carry an even higher mortality rate of up to 50%2 as well as additional risks for long-term problems such as short bowel syndrome, growth impairment and adverse neurodevelopmental outcome.3 4 Thus, early definite diagnosis and immediate onset of appropriate therapy is crucial in dealing with this devastating disorder.

The initial clinical presentation of NEC is often non-specific and indistinguishable from common gastrointestinal problems, such as feeding intolerances of prematurity and abdominal distension resulting from non-invasive ventilation or non-NEC septicaemia. Thus, the diagnosis of NEC is sometimes established late after the development of more specific and severe clinical symptoms, pathological biochemical parameters or abdominal radiographic findings, such as marked intestinal dilatation, pneumatosis intestinalis (PI), portal venous gas (PVG) or free abdominal air.

Radiological PI and PVG are considered specific diagnostic hallmarks of NEC, but recent reports have shown that they might also be present in viral enteritis and cow’s milk protein allergy.2 5 Pathophysiologically, PI and PVG are associated with bowel inflammation with or without necrosis. However, the sensitivity of radiological PI for diagnosis of NEC is low6^–8 and radiological PVG is only detectable in about 10–33% of infants suffering from NEC.9 10

Real-time ultrasound has been used to identify PI11 12 and PVG.12^–15 PI corresponds to gas bubbles along the subserosal and submucosal layer.16 These may be absorbed into the intestinal venous system, travel into the portal vein and can be localised by real-time ultrasound as flowing echogenic dots. Finally, PVG is trapped in the small branches of the portal vein inside the liver and is identifiable as dense granular echogenicities in the liver parenchyma.13

Several reports indicate a role of PVG in abdominal ultrasound (PVG-US) for the diagnosis of NEC even in early stages of the disease.12 13 However, the specific diagnostic value of PVG-US for the diagnosis of NEC has not been systematically evaluated in an unselected cohort of infants admitted to a NICU.

In this study, we prospectively screened infants without abdominal symptoms for the presence of PVG-US and hypothesised that PVG-US would be absent in these infants.

Moreover, infants developing gastrointestinal symptoms were examined to determine prospectively whether PVG-US in the presence of gastrointestinal symptoms indicates NEC with sufficient certainty.

PATIENTS AND METHODS

The subjects of this study were 352 infants consecutively admitted to the NICU from 1 January 2004 to 31 December 2005. We included all patients with written parental consent, regardless of their medical condition, gestational age, neonatal disease or congenital disorder. All were repeatedly screened for PVG-US (total of n = 844 times) either in the course of routinely performed cranial ultrasounds (n = 796) or because of developing symptoms consistent with NEC (n = 48) as defined by Walsh and Kliegman.1 All the latter screenings were combined with an evaluation of acidosis, electrolyte and glucose disturbances, leucocyte and thrombocyte count, C reactive protein (CRP) and interleukin-6 (IL-6). Abdominal radiography was performed upon decision of the attending neonatologist in 19/20 NEC stage ⩾II episodes.

We recorded clinically relevant obstetric, delivery and neonatal medical data in a structured form. Each time during PVG-US screening we further documented parameters of mechanical ventilation, non-invasive ventilation, need of catecholamines, time interval since last feeding, any central and peripheral venous, and/or arterial catheter, oral and parenteral medications, contemporary surgical procedures, any medical condition potentially associated with air in tissues or circulation during the neonatal period (for overview)17 18 as well as all clinical symptoms related to NEC.

Table 1 depicts the clinical characteristics of the infants involved in the study. Their gestational ages (GAs) ranged from 23 to 41 weeks (mean 34 weeks), and 78.1% of them were premature (GA <37 weeks). The patients weighed 400 to 5160 g (mean 1940 g), and 32% of them were very low birth weight neonates (<1500 g). The perinatal history of the patients included birth asphyxia (n = 4; 1.1%), umbilical venous access (n = 5; 1.4%), mechanical ventilation (n = 76; 21.6%) and catecholamine treatment (n = 35; 10%). Thirty-three (9%) infants presented with major congenital malformations (21 cardiac, 12 gastrointestinal); all of those underwent surgery during the study period. All clinical characteristics were present shortly before or during NEC episodes. Thirty-four infants presented with clinical symptoms suggestive for NEC, nine of whom had more than one occasion.

To define NEC and NEC stage, we used clinical symptoms, radiographic findings and blood examinations based on the definition by Walsh and Kliegman (table 2).1 We considered clinical abdominal symptoms suggestive for NEC whenever one or more of the following symptoms were present: increased gastric residuals before feeding (>20% of feeding volume),19 marked abdominal distension, absent bowel sounds, abdominal tenderness with or without discolouration and bloody stool without evidence of a rectal fissure. Mandatory for the diagnosis of definite NEC (stage ⩾II) was radiological evidence for NEC such as marked intestinal dilatation and/or PI and/or PVG and/or pneumoperitoneum. One episode in an infant with severe abdominal symptoms, PVG-US and elevated CRP/IL-6 values was considered to have NEC stage ⩾II by the treating neonatologist without radiographic evaluation and was included in the analysis.

Ultrasound

For the whole study population, scanning of the portal vessels was performed during routine cranial ultrasound at days 1, 3, 7, 14, 28 and monthly thereafter depending on GA, clinical course and patient availability. Most of the infants were screened between three and six times during their clinical course — the more immature and/or instable the infants were, the longer they stayed and the more PVG-US screenings were performed. One-third of the infants were examined only once during their first week of life. These infants were comparatively mature and clinically stable and, therefore, discharged home or transferred to a non-intensive care unit. Ultrasound was performed with a real time scanner (Toshiba Nemio SSA-550A, Zoetermeer, the Netherlands) using a 7.5 MHz linear transducer. We scanned the liver and the portal vein continuously longitudinally and/or transversely in real-time modus for a minimum of 3 minutes (mean 4.5 minutes). This time frame, although chosen arbitrarily, appeared to be sufficient. It was based on our clinical experience with examination for PVG-US for many years and its clinical value has been reported recently.20 It also accounted for the need of keeping time-consuming examinations to a minimum for instable preterm infants. We regarded continuous streaming echogenic dots in the bloodstream and scattered hepatic echogenic spots as PVG (see fig 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Appearance of portal venous gas in abdominal ultrasound as continuous streaming echogenic dots in the bloodstream of the portal vein and scattered hepatic echogenic spots.

Statistical analysis

Cross-tables were generated for NEC stage ⩾II according to our definition (see above) and PVG-US. Sensitivity, specificity, positive and negative predictive values were calculated from the cross-tables and determined on the basis of the infants’ initial clinical NEC diagnosis. All analyses were performed with SPSS 14.0 (SPSS Inc., Chicago, Illinois, USA).

RESULTS

PVG-US was a rare phenomenon in our NICU. None of the 796 PVG-US screenings in the course of routinely performed cranial ultrasounds showed PVG-US. Apart from NEC or suspected NEC, no neonatal disease or any other specific clinical condition was associated with PVG-US.

A diagnosis of NEC (any stage) was made on 48 occasions in 34 neonates, with NEC stage ⩾II comprising 20 of these occasions. Eight infants with NEC stage ⩾II underwent surgery due to clinical deterioration (n = 5) or pneumoperitoneum (n = 3). Seven out of those had no NEC by intraoperative finding (volvulus: 1, septicaemia: 3, focal intestinal perforation: 2, pneumoperitoneum due to pulmonary interstitial emphysema: 1) and did not demonstrate PVG-US either (table 3). In the one infant with intraoperatively proven NEC, PVG-US was detected, but no PI or PVG on abdominal radiograph.

Table 4 depicts the calculations for sensitivity, specificity and positive/negative predictive values. PVG-US was present in 13/48 occasions when a clinical diagnosis of NEC was made, with four occurrences diagnosing NEC stage I. These four infants presented with abdominal symptoms, PVG-US and dilated intestinal loops (but no PI) on abdominal radiograph, but developed neither disturbance of vital signs nor elevation of inflammatory parameters. Serum concentrations of inflammatory mediators were increased in five of the 13 cases where PVG-US was demonstrated; all five were diagnosed as NEC stage ⩾II. Four infants classified as NEC stage ⩾II remained negative for inflammatory serum parameters.

Among infants with symptoms suggestive for NEC, PVG-US had a positive predictive value of a diagnosis of NEC stage ⩾II of 69% (clinical diagnosis of NEC stage ⩾II in 9/13 episodes with PVG-US). The negative predictive value of PVG-US was 69% as well (no NEC stage ⩾II in 24/35 episodes without PVG-US). The specificity of PVG-US of a diagnosis of NEC stage ⩾II was 86% (no PVG-US in 24/28 episodes without NEC stage ⩾II) and the sensitivity of PVG-US of a diagnosis of NEC stage ⩾II was 45% (PVG-US detectable in 9/20 episodes with NEC stage ⩾II).

The calculation of sensitivity was confounded by the fact that seven out of 20 of the clinical NEC stage ⩾II cases showed an intraoperative finding other than NEC. Three additional, conservatively treated infants were classified as NEC stage ⩾II because of clinical symptoms and unspecific radiological intestinal dilatation. None of these 10 infants had PVG-US detectable. Four out of eight infants who underwent surgical treatment were operated on because of free abdominal air (NEC stage III), but in three out of those four infants the clinical–radiological diagnosis of NEC stage III was wrong (table 3). Thus, with retrospective analysis by taking into account the intraoperative findings, the sensitivity of PVG-US increased to 69%. Finally, with application of specific radiological signs according to a recently proposed new NEC classification5 the sensitivity of PVG-US increased to 90%.

DISCUSSION

No study yet has prospectively examined sensitivity, specificity, negative and positive predictive values of PVG-US in an unselected neonatal population treated within a NICU. Our results indicate that PVG-US in a prospective clinical setting has an acceptable specificity (86%) and a high sensitivity (90%) for definite NEC diagnosis, when the diagnostic criteria are based on a recently proposed new NEC classification.5 However, with application of the commonly used Walsh criteria,1 PVG-US features a comparatively low sensitivity (45%) for definite NEC diagnosis.

Few studies have systematically examined the diagnostic significance of PVG-US for NEC diagnosis.11 15 21 In contrast to Faingold and colleagues, who report only 10% of infants with NEC presenting with PVG-US,21 others demonstrated a higher rate between 66% and 100%.13^–15 Merritt and colleagues13 found sonographic air in the portal vein and/or the liver parenchyma in all of 12 infants with NEC and no sonographic abnormalities in a control group of 232 infants weighing less than 2000 g, accounting for a high specificity and sensitivity. However, none of these studies analysed its findings strictly according to a standard NEC definition or commented on disease severity with regard to the presence of PVG-US. Thus, comparison between different studies is hampered. Nonetheless, radiological studies reported incoherent relationships between the presence of radiographic PI and PVG on one hand and NEC on the other.22 9 23 In addition, abdominal radiography for diagnosis of PVG has been reported to be a comparatively insensitive method13 14 24 with a high interobserver variability.7 8

In this study, only one out of 10 infants with definite NEC (Gordon criteria5) would have been missed by screening for PVG-US. What may be the limitations of PVG-US in the diagnosis of a definite NEC? During NEC the presence of gas in the portal vein is not persistent and may be detectable for a short time interval only. PVG-US may become undulated, being associated with intestinal peristalsis, which is reduced in immature infants and particularly with NEC. In addition, NEC progression is more fulminant in very immature infants,10 thus skipping or shortening the phase of intramural bacterial growth and, therefore, the phase with PI and PVG. PVG-US may also be missed, because necrotic bowel tissue is poorly perfused, thus, reducing the transport of intramural air to the portal vein. Also, gas formation by bacteria has been described to depend on the presence of substrate such as milk.16 Thus, unfed or cautiously fed extremely low birth weight infants may be less likely to present with PVG-US in case of NEC.9 Finally, PVG-US may occur as a rather late finding since the traceability of PVG could be seen as a disease extension with comparatively large amounts of intramural gas. In fact, some smaller studies reported a poorer prognosis in patients with NEC demonstrating PVG in abdominal radiography.22 25 26 However, our results and a large study evaluating 194 infants with NEC9 did not confirm this finding. In addition, studies comparing the diagnostic value of PVG-US and radiological PVG clearly indicated a comparatively high sensitivity of PVG-US even in early stages of the disease.13 14 24

In this study, the high specificity of PVG-US for the diagnosis of NEC of any stage was demonstrated by the lack of observing this abnormality during 796 examinations in an unselected cohort of infants without significant gastrointestinal symptoms, even in medical conditions potentially associated with air in tissue or circulation.17 18 However, the specificity for the diagnosis of NEC stage ⩾II among infants with gastrointestinal symptoms suggestive for NEC was lower (85%). PVG-US was detectable in NEC episodes of any stage and was not associated with disease severity.

PVG is usually associated with some kind of bowel inflammation with or without necrosis. However, in the three infants with volvulus and focal intestinal perforation and in 12 infants with surgically corrected gastrointestinal malformations or 21 neonates with congenital heart defects, PVG-US was not detectable. In neonatology, other causes for PVG except for NEC are rarely described in the literature. One case report showed portal venous gas in an infant with hypertrophic pyloric stenosis.27 Others demonstrated radiological PVG in infants with “benign neonatal pneumatosis coli”, a disease which is not well defined yet, but might represent cows milk allergy and in part viral enteritis.5 These infants present with symptoms suggestive for NEC such as gross blood in stools and radiological pneumatosis but usually have little or even absent local and systemic signs.28^–32 In the present study, PVG-US could not differentiate between “benign neonatal pneumatosis coli” and definite NEC. Eight (NEC stage I: 4; NEC stage ⩾II: 4) of the 13 infants with PVG-US might have presented with “benign neonatal pneumatosis coli”, because they neither developed elevated serum values for CRP and IL-6 nor disturbance of vital signs. Thus, according to our data and to our experience with evaluation of PVG-US for many years, PVG-US detection does not indicate a more severe NEC or the need for surgical treatment.

What are the limitations of our study? First, the major limitation might be the lack of a robust “gold standard” for NEC diagnosis in general. Because there is such a broad spectrum of presentations and severity of NEC, some researchers have hypothesised that the commonly diagnosed NEC might be neither a uniform nor a well-defined disease entity.1 A recently published review on NEC diagnosis suggested that it might be “time to abandon Bell’s criteria” and proposed a new guideline for the diagnosis of acquired neonatal intestinal disease based on more specific diagnostic criteria.5 Our results support this notion as 7/20 NEC stage ⩾II cases showed an intraoperative finding other than NEC.

Taking into account this new classification of acquired neonatal intestinal diseases5 which would have staged 10/20 infants as non-NEC classified as definite NEC in our study, because they did not present with PI, only one NEC stage ⩾II episode would have been missed by screening for PVG-US (sensitivity 90%). In addition, this result may be supported by the intraoperative findings in our study, because 7/8 patients, according to Walsh classified as definite NEC, actually had no NEC.

Second, another limitation might be the small number of neonates with clinical NEC. However, the incidence of clinical NEC stage II in our NICU was within the reported ranges.2 The incidence of advanced NEC stage ⩾III and mortality has been low in our NICU. During the study period no infant died from NEC or its consequences or suffered from significant NEC-related long-term morbidity. We have no compelling explanation for this finding. However, this might be attributed to a relatively early diagnosis through an already established PVG-US screening in our NICU for many years.20 Thus, further studies are indicated.

Third, we did not evaluate inter- and intraobserver variability of PVG-US and we are not aware of any study addressing these aspects. However, whereas radiographic abnormalities are often very subtle and carry a high interobserver variability,7 8 PVG-US — if present — can hardly be missed or mistaken even in small quantities, because of the large acoustic impedance of air. Thus, PVG-US could improve the diagnosis of definitive NEC especially in the absence of clear radiological evidence for PI.

Fourth, a methodological limitation might be the handling of independent and dependent measurements. In our analysis we used the number of measurements as reference parameter, although half of the measurements were repeated within the same individuals. However, since ultrasound scans were performed in certain time intervals, we consider this a valid approach.

In summary, PVG-US offers a high specificity for the diagnosis of suspected or definite NEC. Its absence does not exclude the disease. Its presence indicates NEC but does not differentiate between NEC stages according to the standard NEC definition. With application of strict radiological criteria, the diagnostic sensitivity of PVG-US for definite NEC is high. Thus, as compared with the standard radiological evaluation, screening for PVG-US features an acceptable efficacy for NEC diagnosis and offers a very high practicability. Evaluation for PVG-US is readily available and an easy, time-saving procedure, which can be performed repeatedly at any time even in the most instable infants. Screening for PVG-US could become a helpful diagnostic adjunct to abdominal films; should be carried out in every infant suspicious for NEC; and should be evaluated during future research in the diagnosis and management of suspected/definitive NEC. Finally, the value of the Walsh criteria as a gold standard for NEC diagnosis might be questioned by the results of this study.