Introduction

The optimum surgical treatment of infants born with gastroschisis remains unclear. Following Watkins’s first report of primary closure of the abdominal wall in 1943 [1], the principle of surgical treatment has remained returning the eviscerated abdominal organs to the abdomen as soon as possible whilst avoiding the potential complications of viscero-abdominal disproportion (i.e. abdominal compartment syndrome and/or need for prolonged ventilation). Traditionally, surgeons have aimed to achieve primary abdominal wall closure whenever safe and, if not, to fashion a surgical ‘silo’ to allow gradual visceral reduction prior to definitive abdominal wall closure.

Over the past 20 years a pre-formed silo (PFS) comprising a transparent silastic bag fitted with a spring loaded ring has been introduced and its use adopted widely. This has led to a change of practice such that many units now use placement of a PFS on the neonatal intensive care unit without general anaesthesia (GA) as an alternate to attempted emergency abdominal wall closure under GA. This is followed by gradual reduction of the abdominal contents and delayed abdominal wall closure in a semi-elective setting. Recent national and international surveys confirm the uptake of this technique [2, 3]. Outcomes of infants treated with a PFS have been reported by a number of institutions and in some instances compared with those of infants treated with primary closure. Several reports describe the benefits of using a PFS for both patient and surgeon and propose their routine use [4, 5]. However, other authors have highlighted potential pitfalls with the use of the PFS [6, 7]. As a result, the precise role for the PFS in the treatment of infants with gastroschisis remains unclear.

The aim of this study was to evaluate current evidence comparing the use of a PFS with alternate treatment strategies in infants with gastroschisis. We performed a systematic review of the existing literature. We also aimed to perform a meta-analysis of available data by applying strict eligibility criteria to ensure comparability.

Methods

We searched Medline, Embase and the Cochrane controlled trials register from inception to July 2014 using the terms ‘gastroschisis’, ‘gastroschisis and silo’, ‘preformed silo’ and ‘silastic silo’. Abstracts of the unfiltered literature were reviewed and full text versions of selected publications were assessed for inclusion. Reference lists of these publications were also checked to identify additional relevant reports. The literature search, assessment for inclusion and data extraction were performed independently by three reviewers and disagreements resolved by consensus.

Studies were selected for inclusion in the review if they reported comparative outcomes between infants treated with a PFS and infants treated with an alternate treatment strategy. Studies were excluded if they reported a cohort of infants treated only with a PFS with no comparative group, reported a cohort of infants treated with a PFS grouped with infants treated with another treatment strategy such as a hand-sewn silo, originated from a non-developed country or were not in English. For infants treated with a PFS no selection was made on the basis of final technique of abdominal wall closure (i.e. all were included).

All outcomes were selected a priori. The primary outcome measure was time on a ventilator. Secondary outcomes were number of infants never ventilated, time on parenteral nutrition (PN), time to achieve full enteral feeds, length of hospital stay, incidence of necrotising enterocolitis, number of unplanned re-operations, infectious complications, mortality and occurrence of ventral hernia following repair.

Where studies reported outcomes of more than two treatment types (e.g. PFS, primary closure and staged closure using a hand-sewn silo), we selected the group undergoing primary closure as the comparator group and excluded the other group(s) from the analysis.

Data were extracted and entered into Review Manager (v5.1, The Cochrane Collaboration) and meta-analysis was performed using a random effects model due to variation in study design and reporting. Summary statistics for continuous variables are reported as mean difference (MD) with 95 % confidence interval (95 % CI) and dichotomous variables as risk difference (RD) with 95 % CI in order to allow inclusion of studies with zero events [8]. An I 2 statistic for heterogeneity was calculated for each pooled dataset. For the purposes of meta-analysis continuous data are required to be in the format of mean and standard deviation (SD). For studies that only reported median and range, mean and SD were estimated using validated formulae specifically developed for this purpose [9].

We anticipated a large proportion of the data available would arise from retrospective cohort studies. Our a priori intention was therefore to perform subgroup analysis that included only studies reporting pre-planned management strategies for gastroschisis. Only studies that used an intended policy to treat all infants routinely with a PFS unless clinically contra-indicated were included in this subgroup analysis. If it was unclear whether such a policy existed then it was assumed that none did and the study was excluded from this subgroup analysis. The subgroup analysis therefore allows a comparison of outcomes based on an ‘intention to treat’ all infants in the PFS group with a PFS unless clinically contra-indicated.

Results

Search results

One-thousand five-hundred and sixty-seven articles were identified using the specified search criteria and their abstracts reviewed. Following application against our inclusion criteria, 1440 were excluded on the basis of their abstract alone and the full text of 127 publications was scrutinised. One hundred and seven of these were subsequently excluded as they did not report a comparative group, did not use a PFS, or did not report outcomes from a group all treated with a PFS. Two further reports were excluded to avoid duplication of patients; the study by Bonnard et al. [10] was excluded as it reported patients included in another report during an overlapping time period from the same centre [11] and the study by Allotey et al. [4] was excluded as all patients included were subsequently included in a larger report from the same centre 7 years later by Charlesworth et al. [12]. The remaining 18 publications were included in this systematic review. Characteristics of included studies are shown in Table 1. Of these 18, four met the criteria for inclusion in the subgroup analysis (Fig. 1).

Table 1 Description of included studies
Full size table
Fig. 1
figure 1

Flowchart of article selection

Full size image

There was one prospective randomised controlled trial (RCT) comparing use of a PFS with primary fascial closure [11]. This was a multi-centre study which recruited 54 infants over a 5½ year period and was terminated prior to full recruitment due to low accrual rate. Seventeen of the 18 included studies were cohort studies including three reporting data collected from multiple centres during national cohort studies on gastroschisis [13–15]. Two of these report separate outcomes from the same dataset obtained from a national cohort study in the United Kingdom [13, 14]. To avoid duplication, we extracted data exclusively from one or other of these reports for each outcome measure. The single RCT and three other studies met the criteria for inclusion in the intention to treat analysis; all infants in these reports who received a PFS did so as part of a pre-planned management strategy (either as part of an institutional policy [5, 12, 16] or as part of a RCT [11]) to treat all infants reported as receiving a PFS with a PFS unless clinically contra-indicated.

Outcome measures reported by individual studies varied. The outcomes most frequently reported were number of days of ventilation, time to achieve full enteral feeds and length of stay. It was possible to retrieve data on the primary outcome measure (number of days of ventilation) from 8 of the 18 included papers. Other clinically important outcomes including incidence of necrotising enterocolitis, need for further surgery and mortality were variably reported.

Patients and treatment received

In total 1516 patients are included in this review of whom 666 were treated with a PFS. The intention to treat subgroup analysis includes data from 318 patients of whom 133 were treated with a PFS. There were no significant differences between infants treated with a PFS and those treated with an alternate treatment strategy in gestational age [MD 0.05 weeks (−0.3, 0.41); p = 0.77], birth weight [MD 0.01 kg (−0.09, 0.1); p = 0.94] or gender distribution [difference in proportion of males 2 % (−9, 13 %); p = 0.67].

All infants included in the PFS group were treated with a PFS. The comparator group consists of infants treated with an alternate strategy including predominantly primary closure with GA, but also primary closure at the cotside without GA, and attempted primary closure with a ‘hand-sewn silo’ or PFS if primary closure could not be achieved (Table 1). In the intention-to-treat subgroup analysis, the comparator group was attempted primary closure under general anaesthesia with formation of either a hand-sewn silo [5, 12, 16] or insertion of a PFS [11] if primary closure was not possible.

Infants with complex gastroschisis as defined by the criteria of Molik [17] were excluded in six studies whilst others included both simple and complex cases (Table 1). In the intention-to-treat subgroup analysis 3 studies included all infants with simple and complex gastroschisis [5, 11, 16] and 1 study specifically excluded infants with atresia, gut infarction or short bowel syndrome [12]. It was not possible to perform an additional subgroup analysis for simple and complex gastroschisis as outcome data were not reported by severity of disease in the majority of studies.

Quantitative analysis of outcomes

The primary outcome of number of days of ventilation was reported in 8 studies (641 patients). In the overall analysis there was no statistically significant difference in number of days of ventilation between infants treated with a PFS and alternate techniques (Fig. 2a). However, in the intention-to-treat analysis, use of PFS was associated with a significant reduction in days of ventilation of 2.2 days (0.47, 3.93); p = 0.01 (Fig. 2b). There was significant heterogeneity between studies in the overall analysis (I 2 = 70 %, p = 0.002) but not in the subgroup analysis (I 2 = 35 %, p = 0.2). Five studies reported the proportion of infants in each group who were never mechanically ventilated. None of these were in the intention-to-treat subgroup analysis. There was no difference in the proportion of infants never ventilated in each group and significant heterogeneity between studies likely as a result of differing indications for use of the PFS in different settings (Fig. 2c).

Fig. 2
figure 2

Forest plots of duration of ventilation (days) for the overall analysis (a) and intention-to-treat subgroup analysis (b). Panel c Forest plot for need for ventilation at any stage

Full size image

Overall duration of PN was significantly longer in infants treated with a PFS [MD 6.38 days (1.3, 11.46); p = 0.01], whereas time to reach full enteral feeds and length of stay were similar between groups (Fig. 3). In the intention-to-treat subgroup analysis, duration of PN, time to reach full enteral feeds and length of stay were all similar between infants treated with a PFS and those with an alternate strategy (Fig. 4). There was again significant heterogeneity between results of individual studies in both the overall and intention-to-treat subgroup analyses.

Fig. 3
figure 3

Forest plots of duration of parenteral nutrition (top panel a), time taken to achieve full enteral feeds (middle panel b) and length of hospital stay (bottom panel c) for the overall analysis

Full size image
Fig. 4
figure 4

Forest plots for the intention-to-treat subgroup analyses showing duration of parenteral nutrition (top panel a), time taken to achieve full enteral feeds (middle panel b) and length of hospital stay (bottom panel c)

Full size image

All other outcomes were similar between infants treated with a PFS and those with an alternate strategy in the main analysis and subgroup analysis (Table 2) with the exception of need for unplanned re-operation and occurrence of ventral hernia. In the intention-to-treat analysis fewer infants treated with a PFS required an unplanned operation and in the overall analysis fewer infants treated with a PFS developed a ventral hernia following repair. No study in the subgroup analysis reported incidence of ventral hernia.

Table 2 Results of other outcomes included in this meta-analysis
Full size table

Discussion

This systematic review aimed to determine comparative outcomes for infants with gastroschisis treated with either a PFS or alternate treatment strategy. Overall the quality of evidence currently available in this field is poor; to date only one randomised controlled trial has been reported [11]. Whilst meta-analysis is typically aimed as synthesising evidence obtained in RCTs, where such studies are lacking, data obtained from studies using alternate methodology may be valuable. The majority of studies included were retrospective cohort studies which compared outcomes between PFS and alternate strategies either contemporaneously or in some cases between separate defined time periods. In several studies the indication for treatment with a PFS or alternate strategy was poorly defined, if at all. As a result there is significant potential for both treatment and selection biases to influence the findings of this analysis.

The influence of bias is one of the difficulties encountered when combining data from multiple retrospective series. The key source of bias in this review is indication for treatment received resulting in significant selection bias. For instance, it is possible that infants with the least abdomino-visceral discrepancy were treated with primary closure and those with greater discrepancy received a PFS. Alternatively, in some studies, infants who received a PFS were allocated to that treatment on the basis that they had failed treatment by an alternate (preferred) method. This bias likely has an effect on the findings of this meta-analysis. We considered from the outset the value of including all comparative series in this meta-analysis compared to focussing only on those reports using a pre-planned management strategy. In the interest of greater data transparency we included all reports and additionally performed a pre-planned subgroup analysis. Studies were selected for inclusion in the subgroup analysis if they reported data from series where there was a pre-planned strategy, either at an institutional level or as part of a RCT, to use a PFS for all infants reported in the PFS group unless clinically contra-indicated.

Whilst this approach has the advantage of reducing the impact of selection bias from the group treated with a PFS, it should be noted that it also introduces selection bias in some cases to the comparator group as some of those infants were by definition not suitable for treatment with a PFS. The sole RCT [11] reports two groups without selection bias although it could be claimed that selection bias may exist at the level of trial recruitment as only a small proportion of eligible patients were recruited. It may be that surgeons recruiting to the trial maintained a degree of bias and that this influenced their decision of whether to offer the trial to their patients or not.

The quantitative findings of this review suggest that clinically important outcomes are similar between infants who received a PFS and those treated with an alternate treatment strategy. We selected days of ventilation as the primary outcome measure as a surrogate marker of increased abdominal pressure, a phenomenon that may be avoided when using a PFS. Whilst factors other than abdominal pressure may influence need for ventilation, the absence of a sudden increase in intra-abdominal pressure that can be achieved with a PFS when compared to primary closure in particular has been cited as a potential advantage of the PFS [16]. Further, we believe that days of ventilation is an important outcome measure in a critical care setting. Overall there was no statistically significant difference in days of ventilation between groups. The proportion of infants never ventilated, time to achieve full enteral feeds and total length of stay were also similar between groups, although duration of parenteral nutrition (PN) was significantly shorter in infants treated with an alternate treatment strategy than with a PFS. Longer duration of PN in infants treated with a PFS has been reported by a number of the individual studies contributing to this review [18–20]. This may be explained by the inherent added time taken to achieve abdominal wall closure in infants treated with a PFS which contributes to a delay in starting enteral feeds and overall a longer requirement for PN.

In contrast to the overall findings, in the intention-to-treat subgroup analysis use of a PFS was associated with a shorter duration of ventilation. This demonstrates that when a PFS is used as a preferred treatment strategy or in a randomised trial, duration of ventilation can be reduced. Indeed many infants included in the included studies never required ventilatory support. Unfortunately, none of the studies eligible for inclusion in the subgroup analysis formally reported this figure, thereby precluding quantitative meta-analysis of this outcome. Additionally, the association between PFS use and increased duration of PN seen in the overall analysis is not seen in the subgroup analysis. We believe that selection bias is the most likely cause of these two differences and justifies our use of a pre-planned subgroup analysis.

A number of authors have commented on the proposed benefits or disadvantages of the PFS compared to other strategies [5, 12, 16, 18]. Despite collating the largest body of comparative data to date, all remaining outcomes were similar between the groups in both the overall analysis and subgroup analysis with the exception of the number of unplanned re-operations and development of ventral hernia (Table 2). Fewer unplanned operations were reported in infants treated with a PFS and this difference was statistically significant in the intention-to-treat analysis. Within the intention-to-treat analysis these operations were for necrotising enterocolitis, intestinal perforation, stricture or obstruction [5, 12, 16]. Particular concern has been raised that use of a PFS may be associated with bowel necrosis within the PFS [7]. Bowel necrosis may also be due to high abdominal pressure after primary closure. Both the overall and the intention-to-treat analysis did not reveal any significant difference in bowel necrosis with PFS or alternate strategy. Regarding the development of ventral hernia following repair, infants treated with a PFS had a lower incidence of ventral hernia in the overall analysis. It should be noted there was significant heterogeneity in individual study outcomes in this analysis which likely reflects different abdominal wall closure (as opposed to reduction) techniques between studies. We believe the final technique of abdominal wall closure is likely a better predictor of later herniation than the technique of visceral reduction.

The results of this review and meta-analysis require careful interpretation in terms of implications for clinical practice. The association between PFS use and increased duration of PN in the overall analysis suggests that infants treated with a PFS may be being subjected to an unnecessarily long duration of PN dependency. We speculate that a delay in commencing enteral feeds whilst waiting for abdominal wall closure may be responsible for this difference. Whether such a delay is strictly necessary has never been formally tested to our knowledge. Conversely, a treatment strategy of primary fascial closure may achieve a shorter duration of PN at the expense of longer duration of ventilation with its associated risks and cost implications. Which of these detriments is preferable is not known.

Whilst we selected days of ventilation as the primary outcome for this review, it is unclear which of the outcome measures reported is of greatest importance to clinicians and/or parents. For instance, whilst some may view a reduction in days of ventilation of greatest importance others may place greater emphasis on duration of PN or total length of stay. No doubt the range of outcomes reviewed here encompasses the majority of those felt to be important to clinicians. Whether parents and indeed other stakeholder groups consider these to be important is unknown. The importance of this is in selecting the right outcomes by which to measure success or otherwise of any treatment intervention known as a core outcome set. To our knowledge, no such core outcome set exists for infants with gastroschisis.

This study highlights the lack of high-quality evidence regarding the optimum treatment strategy for infants with gastroschisis. Whilst it would be easy to propose a RCT as the optimal solution to this controversy, the difficulties in designing a RCT that is acceptable to both surgeons and parents should not be underestimated as Pastor and colleagues’ experience demonstrates [11]. In particular, surgeons would have to overcome the prejudice formed during their previous accumulated clinical experience of treating infants with gastroschisis. The widespread adoption of the PFS with its clear logistical advantages over emergent primary fascial closure is another obstacle. The PFS may be placed by a suitably trained trainee at any time of day, thereby avoiding out-of-hours emergency operating by a consultant/attending surgeon. It is likely therefore that the PFS is here to stay. However, by casting aside such prejudices we do believe it possible to perform a robust investigation of the available treatment strategies for these infants. We therefore favour further study in the form of prospective, multicentre, protocol-driven collaboration.

Recently Kunz and colleagues have presented the findings of a similar piece of work comparing outcomes between primary closure and staged closure using any type of silo (including both PFS and hand-sewn silo). There are several important differences between Kunz’s review and ours which are relevant. Firstly, Kunz and colleagues only included reviews that compared staged closure with primary fascial closure. Any other closure technique such as immediate bedside reduction followed by ‘sutureless’ or ‘plastic’ closure was not included in their review. Secondly, they grouped infants treated with a PFS together with infants treated with a traditional hand-sewn silo. It is our observation from the literature and our communication with paediatric surgeons worldwide that the PFS is now used in preference to a hand-sewn silo almost universally (where available and affordable) and that many surgeons perform immediate reduction and plastic closure if the clinical situation permits. Our review focuses entirely on the controversy surrounding the use of the PFS versus all of the alternate strategies which we believe accurately reflects the treatment options employed by the modern paediatric surgeon.

In summary, this review demonstrates that use of a PFS results in largely equivalent outcomes compared to alternate treatment approaches for infants with gastroschisis. Although planned use of a PFS is associated with fewer days on a ventilator such a strategy may expose some infants to unnecessary prolonged durations of PN. Further investigation is required to identify not only the optimal treatment pathway for infants with gastroschisis which will likely include a case by case decision algorithm, but also the tools by which to measure success.