Participants

Studies were included if they involved infants born with simple gastroschisis, defined as gastroschisis with an intact continuous bowel that is not compromised or breached at delivery or presentation. Infants with complex gastroschisis, defined as the presence of at least one of intestinal perforation, necrosis, or atresia were excluded from this review, as they only account for approximately 11.5% of infants with gastroschisis, are known to have a worse prognosis than the majority with simple gastroschisis [2], and thus represent a major confounder in any examination of the outcomes of different management techniques.

Interventions

Correction of gastroschisis can broadly be broken down into two parts; method of visceral reduction, and method of closure. Our intervention of interest was SR, defined as use of a silo (pre-formed, custom or improvised) for staged reduction of the abdominal viscera, followed by any form of defect closure. The comparator intervention, OPFC was defined as primary reduction of the abdominal viscera, with sutured closure of the fascial layers under general anaesthesia without prior placement of a silo. Studies were included if they compared OPFC with SR. Studies including data on only one intervention (i.e. with no comparison group) were specifically excluded.

Outcome measures

Primary outcome measures:

1. Mortality
2. Length of hospital stay
3. Time to complete enteral feeding/duration of parenteral feeding

Secondary outcome measures:

1. Duration of ventilator use
2. Infective complication
3. Perforation
4. Ischaemic bowel
5. Necrotising Enterocolitis (NEC)
6. Anastomotic stricture
7. Adhesional small bowel obstruction
8. Complications related to stoma
9. Short bowel syndrome
10. Liver disease associated with intestinal failure

Potential confounding factors

In recognition that other patient factors including prematurity, low birth weight and associated anomalies may also impact on outcome, we sought to describe the distribution of these potential confounders. We did not attempt to control for these amongst the inclusion criteria, as we felt this would be unnecessarily restrictive.

Study assessment

Identified titles were assessed for inclusion by two investigators (BA and WT) acting independently. Any conflicts were resolved by a third investigator (MK). Data from included articles were extracted independently by the same two investigators (BA and WT), and any differences resolved by discussion. Where required, unpublished data were requested from the authors of each of the included studies to allow for meta-analysis.

Statistical analysis

Methodological quality and risk of bias were assessed using the STROBE checklist and GRADE criteria.

Data were synthesized using standard methods as described in the Cochrane Reviewers Handbook [8]. The chi-squared test for heterogeneity was used to assess the extent to which the results of the studies were in agreement. Using a conservative cut-off of p<0.01, no statistical heterogeneity was detected and therefore fixed-effect methods were used throughout. Dichotomous outcomes were meta-analysed using Mantel–Haenszel fixed effects methods to produce summary odds ratios with 95% confidence intervals. Continuous variables were meta-analysed using the inverse-variance fixed-effect method to produce weighted mean differences and 95% confidence intervals. All analyses were performed using Review Manager 5.2 (Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2012.)

Characteristics of included studies

Using the stated search strategy, 3081 papers were identified. After removal of duplicates, 751 unique articles remained. From this, 8 papers were deemed to meet the inclusion criteria, all of which were cohort studies (Fig. 1). No randomised controlled trials were identified that met the inclusion criteria. All papers bar one were published in English language journals. One Spanish paper was translated. Unpublished data were obtained from two authors [2, 9].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Study flow diagram.

https://doi.org/10.1371/journal.pone.0116908.g001

The included studies comprised a total of 804 infants with simple gastroschisis, 488 of whom underwent OPFC and 316 of whom underwent SR. The largest study was Owen 2010 [2] with 290 eligible patients. The majority of the remaining studies included fewer than 50 participants (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Summary of Study Characteristics.

https://doi.org/10.1371/journal.pone.0116908.t002

Although reporting of baseline characteristics was variably done by the included studies, where it was reported, there were no statistically significant differences in birth-weight, gestational age, or associated anomalies between the OPFC and SR groups.

Characteristics of excluded studies

There were three main types of excluded studies. These were: studies which did not exclude patients with complex gastroschisis, studies which did not directly compare OPFC and SR, and studies which adulterated either the control or intervention group with other closure methods. As a result, we were unable to include several potentially useful studies, including three large retrospective cohort studies based upon data from the Canadian Paediatric Surgical Network (CAPSNet) [10–12] and one randomised controlled trial [6]. See supplementary table for summary of characteristics of excluded studies.

Key Outcomes

All three primary outcome measures appeared to favour the OPFC group over the SR group. Both length of stay and number of post-operative days to complete enteral feeding were clinically and statistically significantly longer in the SR group, with a mean difference of 8.97 days, (95% CI 2.14–15.80 days), and a mean difference of 7.19 days (95%CI 2.01–12.36 days) respectively (Fig. 2). Although the odds of death were raised in the SR group (OR 1.96, 95%CI 0.71–5.35), this difference was not statistically significant (Table 3 and Fig. 2). Results for secondary outcome measures are displayed in Table 3 and Fig. 3.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Forest plots showing the effect of SR on primary outcome measures.

https://doi.org/10.1371/journal.pone.0116908.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Summary Outcome Measures for Silo Repair in Neonates with Simple Gastroschisis.

https://doi.org/10.1371/journal.pone.0116908.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Forest plots showing the effect of SR on key secondary outcome measures.

https://doi.org/10.1371/journal.pone.0116908.g003

Study limitations

This study has highlighted three key themes that recur throughout research into paediatric surgery, and which help account for the lack of a robust evidence base for clinical practice. These are; limitations of study type, limitations of methodology, and inconsistency in outcomes and definitions.

As is representative of most research in paediatric surgery [14], our meta-analysis was limited by the fact that the majority of studies available for inclusion were small, retrospective case series or cohort studies. Due to the impact of chance and confounding on the results of these studies, it is impossible to generate robust guidelines based upon them. To produce evidence-based guidelines, the type of study used for assessment of any given intervention should progress as described by the IDEAL recommendations [15] through the use of large cohort studies, to the gold standard randomised controlled trial. Doing so not only increases the weight of the conclusions of the primary research, but also of any meta-analysis including them.

Even with the most robust meta-analysis, it is impossible to remove elements of bias introduced by methodological limitations within the primary research. Often, surgeons who would preferably perform OPFC will instead perform SR if pre-operatively they identify significant viscero-abdominal disproportion. Such surgeon selection means that the SR group may be more likely to contain infants who are sicker or at greater risk of complications. Given that none of the studies included in this review provided a meaningful comparison of the demographics and severity of illness in the OPFC and SR groups, we have been unable to control for any disparity between these groups when drawing our conclusions, introducing a potential bias against SR. Further bias may have been introduced as in many of our included studies, the comparison between OPFC and SR was made when an institution switched from performing OPFC to SR. This may have introduced a degree of bias in favour of OPFC, as institutions had greater experience in this technique than they did for SR, therefore potentially further artificially improving the outcomes in the OPFC group.

Inconsistency in intervention definition between the included studies led to significant clinical heterogeneity within our studied treatment groups, whilst inconsistency in choice of outcome measures and methods of reporting further hampered efforts at meta-analysing the little data that were available. Payne 2009 [9], Rodriguez 2009 [16] and Kandasamy 2009 [17] only performed SR once OPFC had been attempted and failed, whilst Tsai 2010 and Bonnard 2008 included this same group of patients in the OPFC analysis on an intention to treat stratagem. Heterogeneity also existed in the method of skin closure used. Evidence exists to suggest that this as well as the reduction method may affect outcome [18]. Rodriguez 2009, Kandasamy 2010 and Tsai 2010 [16, 17, 19] presented time dependent variables continuously, whilst Bradnock and Owen presented the same variables as categorical data [2, 7]. Outcome measures were also variably reported, with means reported without standard deviations [17], and data presented in such a way as to make it impossible to determine whether means or medians were being reported [20]. Such incomplete data prevents meta-analysis, whilst the clinical heterogeneity reduces the validity of conclusions drawn from any meta-analysis that could be performed.

The Way Forward

A large majority of the problems described stem from the low case numbers and heterogeneous population encountered by any single institution. We therefore believe that the way to promote evidence-based paediatric surgery is by enhancing collaboration between centres. With only 28 paediatric surgical centres in the United Kingdom and Northern Ireland, achieving co-ordination of research at each of these centres should not be unattainable. A centralised organisation, such as the British Association of Paediatric Surgeons, outside of the influence of any individual surgical centre would provide the ideal starting point for creation of a National Paediatric Surgical Trials Unit. Inviting study proposals and principal investigator applications from each of the contributory units, combined with the necessity for consultants to show contribution to national audit and research in order to attain revalidation could be used as incentives for ensuring the participation of individual centres. With national research co-ordinated through one centre, the possibility of International co-operation is also increased. Developing links with units such as the Canadian Paediatric Surgery Network (CAPSNet)[12] would create the potential for a truly global network of researchers to be developed, with access to a large population base, and the ability to provide truly robust evidence on which to base the clinical practice of paediatric surgery.

A National Paediatric Surgical Trials Unit could act both as a centre to facilitate co-operation between individual surgical centres, and as a research unit in its own right. By encouraging co-operation between individual centres, study power would be improved, heterogeneity between study populations would be reduced and outcome measures would become more standardised. With buy-in and co-operation from each of the individual surgical centres, a National Paediatric Surgical Trials Unit would have the ability to develop an overarching research programme that would allow generation of sufficient quality data to not only evaluate existing practice, but to take new surgical ideas through from initial evaluation to assessment with national randomised controlled trials, and on to large scale review of established practice, as outlined by the IDEAL recommendations [15]. A similar national approach could be used in other smaller surgical specialties.