Gastrostomy feeding in cerebral palsy: a systematic review
- National Perinatal Epidemiology Unit, Institute of Health Sciences, Old Road, Headington, Oxford OX3 7LF, UK
- Correspondence to:
Dr G Sleigh
1 Walton Crescent, Oxford OX1 2JG, UK;
- Accepted 17 September 2003
Aims: To determine benefits and risks for gastrostomy or jejunostomy feeding compared with oral feeding for children with cerebral palsy.
Methods: Systematic review. Search strategy: electronic databases—Cochrane Library, Medline, Embase, Cinahl, Lilacs, databases of theses, grey literature. Included: relevant systematic reviews, randomised controlled trials, observational studies, case reports. Excluded: non-systematic reviews and qualitative research. Participants: children with cerebral palsy. Intervention: use of gastrostomy or jejunostomy tube to provide nutrition. Outcome: evaluated outcome measures included death, growth, gastro-oesophageal reflux, other complications, psychosocial aspects, and caregiver wellbeing.
Results: No relevant systematic reviews or randomised controlled trials were found. Two cohort studies, 15 case series, and eight case reports met the inclusion criteria. Eight studies specifically described percutaneous endoscopic gastrostomy as the intervention. Weight gain resulted from gastrostomy feeding in most cases. There was an approximately fourfold increased risk of death reported in one cohort study for the gastrostomy fed children. Many complications were reported, including potential for increased gastro-oesophageal reflux and fluid aspiration into the lungs.
Conclusions: Benefits associated with gastrostomy or jejunostomy feeding are difficult to assess from the available evidence. Risks of gastrostomy, particularly in relation to surgical complications, have been described but the size of the risk could not be quantified. The finding of a higher death rate for children fed by gastrostomy may merely reflect the greater disability of these compared with orally fed children. Lack of available evidence and the substantial risk of bias in observational studies suggests that a well conducted randomised controlled trial of sufficient size will be needed to answer these problems.
- ARP, antireflux procedure
- CP, cerebral palsy
- GF, gastrostomy fed
- GOR, gastro-oesophageal reflux
- PEG, percutaneous endoscopic gastrostomy
- OR, orally fed
- RCT, randomised controlled trial
Children with cerebral palsy (CP) often have difficulty eating and drinking. These difficulties are due to problems with oro-pharyngeal control1–3 and oesophageal motility,4–6 related gastro-oesophageal reflux (GOR),4,7 and the high risk of aspiration of food/fluid into the lungs.8–10 Among the consequent health problems of these feeding difficulties are undernutrition,11–16 oesophagitis with bleeding and pain,17 recurrent chest infections,9 and progressive lung disease.18 Children most at risk for these problems are those with spastic quadriplegic and dystonic CP.19,20 These categories account for one third of all cases of CP in developed countries.21 In other countries this proportion may be higher, for example it was 46% in one report from the Philippines.15 Assuming a prevalence of CP of 2.5–2.0/1000 live births21 we can estimate that around 6000 children in the UK22,23 will have significant feeding related health problems due to CP.
Much of this feeding difficulty can be bypassed by giving nourishment through a gastrostomy (or jejunostomy). The percutaneous endoscopic gastrostomy (PEG) has become widely used.24,25 However for families of children with CP the idea of a feeding tube is often distressing.26–29
The effects of enteral feeding for children with CP, apart from weight gain,30 have not been clearly described. There is controversy about whether gastrostomy feeding may increase the risk of death20,31 or GOR.32–35
Understanding the effects of tube feeding for children with CP is made even more difficult because gastrostomy feeding is also used for children36–39 and adults40 with a wide range of other conditions both acute and chronic. These include rehabilitation from traumatic head injury41 or cerebrovascular accidents,42 inflammatory bowel disease,43 progressive neurological or metabolic disorders,44 cancer,45,46 and AIDS.47 Studies of gastrostomy feeding often include a range of ages and underlying conditions. But the type and frequency of outcomes will be influenced by the underlying medical condition for which tube feeding was prescribed. For example, cancer or AIDS may increase the risk of infection,47 children with a progressive disorder may have an increased mortality risk,32,48 and some children recover from their physical disability32 and no longer need their gastrostomy. CP is a non-progressive disorder,49 but the neurological problems that result in the feeding difficulty persist. So the relevance, for children with CP, of the findings from studies that include a mixed age range and a variety of diagnoses is unclear.
The management of eating problems for children with CP is difficult and the potential effects of gastrostomy tube feeding on the health of these children is controversial. Families with affected children need better information when making the difficult decision about whether to accept or request a gastrostomy for their child. We therefore undertook a systematic review of the available literature in order to answer the question “What is the balance of benefits and risks to children with CP from gastrostomy or jejunostomy tube feeding?”.
Types of studies
The review included any systematic reviews, randomised controlled trials (RCTs), whether published or unpublished, and any published observational studies which addressed the effect of gastrostomy feeding for children with eating difficulties due to CP. Neither the language nor country of origin were reasons for exclusion of studies.
Qualitative studies, non-systematic reviews, personal practice papers, and annotations were excluded.
Types of participants
Children with cerebral palsy (as defined by the study authors) and feeding difficulties.
Studies that included a majority (greater than 50%) of children with feeding difficulties due to causes other than cerebral palsy or studies that included a majority of adults (over 16 years of age) were excluded.
Types of interventions
Delivery of nutrition via a gastrostomy or jejunostomy tube. Studies in which some children in the intervention group also received some nutrition by mouth, had surgical antireflux procedures performed, or were taking antireflux or other medication were eligible for inclusion. The comparison group were children receiving nutrition solely by mouth.
Studies that had a majority of children who were fed by nasogastric or nasojejunal route either as the intervention or prior to the intervention were excluded.
Types of outcome measures
Death, growth, development, psychosocial effects, and other measures of health, such as complications of surgery, changes in symptoms of GOR and respiratory disease for the child, and measures of physical health and psychosocial effects for their carers.
Search strategy for identification of studies
This included searches in electronic databases (the Cochrane Library, Medline, Cinahl, Latin American and Caribbean Center on Health Sciences Information (Lilacs), ASLIB, and Dissertation Abstracts), hand searching relevant journals, grey literature, and contacting authors and manufacturers who are concerned with gastrostomy feeding to ask if they knew of any relevant unpublished RCTs.
The following search strategy was used in Medline:
#1 explode “Child-”/all subheadings in MIME, MJME
#2 infant* or baby or babies or child* or teen*
#3 young person* or young people or youth or adolescen*
#4 girl* or boy* or preschool*
#5 #1 or #2 or #3 or #4
#6 explode “Enteral-Nutrition”/all topical subheadings in MIME, MJME
#7 (tube* near feed*) or (enteral near feed*) or (enteral near nutrition)
#8 gastrostom* or jejunostom* or gastro-jejunostom* or gastrojejunostom*
#9 #6 or #7 or #8
#10 explode “Central-Nervous-System-Diseases”/all topical subheadings in MIME, MJME
#11 explode “Cerebral-Palsy”/all topical subheadings in MIME, MJME
#12 cerebral palsy or Little* disease or (spastic near diplegia*) or (spastic near quadriplegia*)
#13 nervous system disorder* or nervous system disease*
#14 (cerebral near palsy) or (neuro* near disab*) or (neuro* near impair*)
#15 #10 or #11 or #12 or #13 or #14
#16 #5 and #9 and #15
The search strategy, with minor modifications as required, was also used in Embase, Cinahl, and the Cochrane Controlled Trials Register. A simpler free text search was used in Lilacs, Aslib, and Dissertation abstracts. The search histories used in each of the electronic databases are available from the authors on request.
Methods of the review
The titles and abstracts of all studies found through electronic searches were scrutinised by one reviewer (GS). Search for randomised controlled trials was conducted independently by two researchers (GS and JA). Full copies of potentially relevant studies found through the complete search were obtained and, for those that met the inclusion criteria, data were extracted onto a specifically designed form. These studies were assessed independently by the two reviewers (GS and PB). Any disagreement was resolved by discussion. In some cases the author of the paper was contacted for further information. If the data remained unavailable the study was excluded. The reason for exclusion of studies was recorded. Data were summarised in tables. Published guidelines for assessment of study quality50–56 were used.
A priori, synthesis of data for this review (meta-analysis) was planned for RCTs only. As meta-analysis of data published in observational studies can only combine crude data, without adjustment for potential confounding factors, it was not considered appropriate to perform any data synthesis for these studies. Where possible adjusted estimates of effect (risk ratios, odds ratios, or hazards ratios) are presented separately for each study.
Results of the literature search
The search in Medline, Embase, and CINAHL retrieved 418 studies. Only seven additional studies were found from the rest of the search strategy. A total of 120 papers that appeared to be relevant were obtained and read. Twenty five studies, two cohort,16,31 15 case series,57,19,20,33,58–68 and eight case reports18,69–75 met the inclusion criteria and were reviewed. One research project in progress was found.76 This was not included in the review. Ninety five papers were excluded. Thirty three of these did not include a majority of children with CP.6,44,47,48,77–91,35,92–96,34,97–103 A further seven studies did not specify the number of children with CP. Since it was not possible to obtain the data from the authors they were excluded.104–110 Sixteen studies involved mainly or all adults40–42,111–123 and 22 were not research, for example, annotations or reports of personal practice.1,24,25,124–142 Sixteen were excluded for other reasons, including use of naso-gastric, or naso-jejunal tube feeds as the intervention,30,143 or prior to gastrostomy/jejunostomy;32,144–146 surgical procedures;147–149 and qualitative studies.26–29,150,151 For one further study it was not possible to tell the proportion of either adults or children with CP among the gastrostomy fed group.152
These 120 potentially relevant studies originated from the following countries: USA, UK, Canada, Australia, New Zealand, Sweden, France, Italy, Spain, Germany, Switzerland, Austria, Brazil, Chile, Taiwan, and Japan.
Characteristics of the studies
All 25 included studies were observational; eight studies reported PEG as the intervention.
No relevant systematic reviews of randomised controlled trials or of observational studies were found and no relevant randomised controlled trials were found.
Tables 1–3 summarise the characteristics and results of the studies according to methodology.
Only two studies had a concurrent control group, both cohort studies (table 1). In the first of these31 data were gathered retrospectively from the client development evaluation report. This report is maintained for each individual who accesses the services for people with developmental disability in the State of California. The study compared the risk of death for the 1060 children with CP who were fed via a gastrostomy feeding tube with the risk of death for children with CP who were orally fed. The orally fed group comprised 5980 children who had at least some self-feeding skills and 5670 children who were totally dependent on someone else to feed them.
The second study16 involved children already enrolled in the North American Growth in Cerebral Palsy Project (NAGCPP) that includes all known children with CP in several geographically defined areas in the USA and Canada. All participants in this study16 were children with moderate or severe motor impairment, scoring III to V on the Gross Motor Function Classification System (GMFCS).153 The study’s aim was to document, using a simple parental questionnaire, the prevalence of feeding dysfunction in children with moderate to severe physical disability due to CP. Within this study group there was a subgroup of 49 children who were fed by gastrostomy tube. These were compared with a subgroup of 70 orally fed children who scored V, the most severe category, on the GMFCS. These two subgroups comprised the cohort study. Outcome measures for the cohort study comprised administration of the Child Health Questionnaire (CHQ),154 the NAGCPP questionnaire (NAGQ)16 and measures of growth and nutrition that had been recorded already from the NAGCPP.
The 15 case series studies (table 2) describe children with CP, severe physical disability, and associated feeding problems referred to a specialist clinic because of professional concern about poor nourishment. This had resulted in placement of a gastrostomy or jejunostomy feeding tube. One of the studies66 described children who in addition to the above characteristics had osteopenia diagnosed radiologically. Many of the children in these 15 studies also had severe learning difficulties and epilepsy. Aspiration and GOR were additional to poor nutrition as indications for gastrostomy or jejunostomy in some studies.64,65 Five studies were retrospective, five prospective. Five were mixed retrospective and prospective; these studies used retrospective chart review to identify the children and provide some data, with prospective assessments of, for example, growth,19 caregiver satisfaction,19,20,59 and diet.66
Mean reported follow up ranged from 8.4 months to 3.5 years.
A variety of outcomes were assessed in the 15 studies: GOR (10), growth (8), survival (7), major complications (8), other complications (7), caregiver satisfaction (6), nutritional assessment (4) restoration of full oral feeding (2), and other benefits for the child, such as state of alertness and improvement in mood (5).
However, authors tended to use different criteria to assess the same outcomes. For example, deaths and medical complications were only counted as complications of the gastrostomy if they occurred in the early postoperative period in two studies,62,64 while other studies counted complications that occurred throughout the follow up period (the length of which differed between studies); and presence of GOR was variously assessed by radiology,19,57,61 24 hour pH monitoring,33,62,64 and oesophagitis diagnosed histologically.61,62 In four studies the method was not stated. Caregiver satisfaction was mainly assessed by questionnaires that had been devised by the authors for the study. Only one study67 used prospective completion of the questionnaires pre-gastrostomy and at follow up. The development and piloting of the questionnaire for this study had been described previously.14 However assessment for validity, reliability, appropriateness, and acceptability were not reported for any of the author designed questionnaires in the reviewed studies.
The case reports (table 3) describe children with CP, all of whom had severe physical disability and were fed via a gastrostomy feeding tube. Six of the eight reports are about complications.
Thus the included studies showed considerable heterogeneity in study design that included different methods of recruitment, outcomes assessed, criteria for the inclusion of the outcomes, tools used to assess similar outcomes, and length of follow up. The studies were open to systematic and random bias, only two had a control group.
Findings of the included studies
Findings are summarised in tables 1–3 individually for all included studies.
The first of the two cohort studies31 suggests that type of feeding has a major impact on survival. When children with CP fed by gastrostomy tube were compared with orally fed children who had at least some self-feeding skills the hazard ratio was 23.65. However when other factors, such as level of physical disability were held constant the hazard ratio reduced considerably (3.85, 95% CI 2.88 to 5.14) when the gastrostomy was placed between 2 and 3 years of age.
The second cohort study16 found similarities and differences between the gastrostomy (GF) and orally fed (OF) groups. There was no significant difference in functional communication, nor in hospital stays nor time in bed and school missed due to illness nor in measures of arm muscle mass. Gastrostomy fed children scored worse than controls on the global health z score (OF mean = −0.46 (SD 1.24), GF mean = −1.84 (SD 1.04), p<0.001) and physical summary z score (OF mean = 38.1 (SD 15.6), GF mean = 23.6 (SD 17.3), p<0.001) of the CHQ; they were more likely to be incontinent (OF 47/70 and GF 46/49 children, p<0.001) and families of gastrostomy fed children reported (CHQ) a greater impact on their time (z score OF mean = −0.91 (SD 1.80), GF mean = −1.38 (SD 1.70), p = 0.1) and greater emotional impact, which meant more worry about their child’s general health (z score OF mean = −0.07 (SD (1.20), GF mean = −0.80 (SD 1.40), p = 0.004). However, gastrostomy fed children were reported to have less respiratory illness during the previous year (54/70 and 28/49 children, p = 0.03). Growth measures, weight z score (OF mean = −2.77 (SD 2.56), GF mean = −2.15 (SD 2.19), p = 0.082), height z score (OF mean = −3.20 (SD 1.63), GF −2.55 (SD 1.26), p = 0.014) and triceps skinfold thickness z score (OF mean −0.94 (SD 0.99), GF mean = −0.15 (SD 1.31), p = 0.001), all favoured the gastrostomy fed children.
For the case series (table 2), GOR was most the most frequently reported outcome. All 10 of the case series studies that assessed GOR19,20,33,57–59,61,62,64,65 reported that one or more children had new, continued, or worse GOR following gastrostomy tube feeding. Sometimes this occurred in spite of a surgical antireflux procedure (ARP) concurrent with the gastrostomy.19,20,58,59 Most children’s symptoms improved when treated with antacids and prokinetic medication, but in seven of the studies some children underwent new or further ARPs because of continuing severe symptoms of GOR. Two studies suggested improvement in GOR following PEG. In one61 7/14 children had less vomiting but, although in two GOR was worse, neither underwent an ARP. The second study33 reported that 6/10 children improved and remained off antireflux medication during the follow up period of 8–18 months. However four of the 10 could not be weaned off the antireflux medication and two of these underwent ARPs. In spite of this the author’s conclusion was that “nutritional rehabilitation resulted in marked improvement of GER in these patients”.33
All eight studies that assessed growth19,33,57,58,60,64,65,68 found that most, though not all, children showed improved weight gain after gastrostomy tube feeding. Change in rate of length growth, reported in four studies,19,57,58,65 appeared to be less predictable and occurred only in a minority of children.
Of the seven case series reporting death,19,58,59,62–65 no deaths occurred in two studies.62,64 The lack of a control group, varying length of follow up, and varying numbers of study participants makes it impossible to gain any impression of the risk of death posed by gastrostomy feeding.
Table 4 lists the other major complications that were reported in relation to gastrostomy or jejunostomy tube feeding in the reviewed studies. This table is derived from the studies in tables 2 and 3 and, as there is no clear denominator, the incidence of these complications cannot be estimated. The complications involve issues about the surgical technique, the intra-abdominal equipment, and the artificial “feeds” that are used to provide nourishment. Relatively minor complications are frequently mentioned in the case series (table 2) and included site infections, granulations, leakage round the tube, tube migration, pneumoperitoneum, blocked tube, vomiting, retching, dumping syndrome, diarrhoea, and constipation. The proportion of children affected by these minor complications can be as high as 95%.20 Most of the caregivers polled in the case series appeared satisfied with the gastrostomy for their child.19,20,59,61 Benefits included: ease of feeding; improvement in child’s disposition and nutrition;19 53/57 were pleased with the gastrostomy and child’s comfort and abilities enhanced;59 17/19 reported less stress; and 19 would recommend PEG to other families.61 In one study20 32/40 believed gastrostomy had a positive impact on the life of their child and the rest of the family. However, 11 of these 40 found problems with family functioning and stress. Problems with family functioning and stress included difficulty getting respite care because of lack of adequately trained caregivers, restriction of mobility especially if a pump was used, finding a changed relationship with their child, and their child missing the taste of food. Restoration of total oral feeding as an outcome was the subject of two case reports70,74 (table 3); it was mentioned in only two of the case series. In one of these59 four children were being fully orally fed at follow up, but in the other58 both patients had died following resumption of full oral feeding.
Although the inclusion criteria for this review were specific, it was not possible to draw firm conclusions about the risks and benefits of gastrostomy tube feeding for children with cerebral palsy because of the severe methodological weaknesses of most of the included studies.
The results of this review show that gastrostomy feeding for children with physical difficulty eating due to cerebral palsy is practiced in many countries. There is a general assumption that it is a necessary, safe, and effective treatment.61,64 However, this review suggests that there is no firm evidence for this assumption and the outcomes for the two cohort studies appear less in favour of gastrostomy than conclusions drawn from the case series studies.
The main weakness of both cohort studies is that the control group children were unlikely to have been as severely disabled as the gastrostomy group. In the first, gastrostomy fed children were compared with orally fed children who had “some self-feeding skills” rather than those who were “fed by others, no feed tube”.31 The children with some self-feeding skills are likely to be the least disabled of these three groups and those with gastrostomy the most disabled. This may explain the excess of deaths in the gastrostomy group. This explanation is supported by the finding31 (table 1) that the association between feeding and death was substantially reduced when other confounding factors were controlled for. This raises the possibility that this adjusted relation is still biased by residual or uncontrolled confounding. This study also found a strong correlation between survival time and degree of physical disability. For example, 50% of gastrostomy fed children with CP unable to lift their heads in prone lying survived beyond the age of 7 years. But this was extended to 50% survival beyond 12 years of age if they could lift their heads. This suggests that quite a subtle difference in motor control has a major impact on additional years survived.
In the second cohort study, the investigators chose the control group from those orally fed children who were most severely disabled (GMFCS V). However, within category V there are gradations of disability.153 For instance both categories of head control mentioned above,31 that had markedly different survival outcomes, would be included in category V. Moreover this study16 found that the tube fed children were significantly more likely to be incontinent than their orally fed controls. The gastrostomy fed children then, may have been more severely disabled and/or had poorer general health than the orally fed children. If this was the case it could explain why the scores for global health and physical summary domains were worse for the gastrostomy fed children than controls, and why parents of gastrostomy fed children had more worries about their child’s health.
These concerns about confounding mean that neither of the cohort studies answers the crucial question as to whether the apparently less favourable aspects for the gastrostomy fed children are due to the gastrostomy tube feeding or to the child’s disability and general health.
Aspiration of food or fluid into the lungs is almost certainly a major risk factor for ill health9 and thus decreased length of survival in children with CP; respiratory problems are not infrequently mentioned as the cause of death.20,32,155
Direct aspiration of orally ingested material and saliva may be a reason for recommending gastrostomy tube feeding.64,65 But GOR is also thought to contribute to aspiration indirectly; for example, refluxed gastric juices were seen to reach the pharynx during videofluoroscopy.9 It is important then to know whether gastrostomy tube feeding for children with CP results in new or worse GOR. As has been shown this review did not allow any firm conclusion to be drawn about whether GOR was increased or decreased with gastrostomy tube feeding.
Because most symptomatic children respond to prokinetic and antacid medication, and ARPs are not always successful for children with CP,59 most surgeons do not advise routine ARP concurrent with gastrostomy unless a child’s symptoms are severe despite appropriate medication.95,156 But both GOR and aspiration can occur without obvious symptoms (silent)8 and the investigations that are often used to diagnose GOR appear to be unreliable predictors, in children with CP, so that GOR will become worse after gastrostomy tube feeding.32,61,95,105,156 From this review then, regarding GOR, several issues remain unsolved. These include: not understanding the relative contributions of direct and indirect aspiration to chronic lung disease in children with CP; not having reliable ways of diagnosing GOR and aspiration in children with CP; not being able to predict whether and to what extent gastrostomy tube feeding is likely to significantly increase GOR (silent or symptomatic) and, therefore, being unable to predict for individual children whether gastrostomy tube feeding is likely itself to cause lung damage.
Most children with CP appear to get fatter as a result of enteral tube feeding as evidenced by increased weight, triceps skinfold thickness, and altered body fat composition (table 219,33,58,60,64,65,68 and table 116). The weight gain reported for most children with CP who receive enteral feeds has advantages, such as looking healthier and feeling warmer, but also disadvantages, such as being more awkward to lift and requiring larger and more obtrusive equipment at home.29
From the review it was not possible to tell whether the children’s overall health or survival was better or worse than it would have been with oral feeding alone, because the majority of studies did not have a control group. Many of the minor complications are unpleasant for the children and/or their carers and may significantly affect their quality of life.74
There are also concerns about replacing ordinary food with commercially prepared “feeds”. Dietary balance may be more easily disrupted,66,72,75,120 giving rise to micronutrient and vitamin deficiencies,66,75,120 and problems with refeeding may occur72 if, for some reason, too little feed is given.
The potential negative consequences of gastrostomy feeding are especially important for children with CP since for them tube feeding tends to be a long term solution for their feeding difficulties. Few instances of successful reintroduction of all nutrition by mouth have been reported for this group; it requires skill and patience.70,124 Generally the gastrostomy tube can be removed without complication but gastrocutaneous fistula may result and require operative closure.81,87 In spite of the reported adverse effects caregiver satisfaction tended to be high; this may simply reflect the likely high level of bias for these findings.
A weakness of this systematic review is that only one author did the search for observational studies; some could have been missed. It is very unlikely that relevant RCTs or systematic reviews were missed, since two researchers searched independently and none of the authors we contacted knew of any published or unpublished RCTs.
This systematic review has shown that there is little robust evidence about the effect of gastrostomy (or jejunostomy) tube feeding for children with eating difficulty due to CP. Moreover, serious issues are raised about a potential increased risk of death, the necessity for further surgical procedures, and some life threatening complications. In addition there is some evidence that gastrostomy feeding has a negative impact for families. It is not possible from this systematic review to draw any firm conclusions about whether placing a gastrostomy or jejunostomy for children with CP who have difficulty eating and drinking gives overall benefit or harm. These issues could be settled by carrying out a well conducted randomised controlled trial of sufficient size to address some of these important outcomes. For example, the sample size required to exclude a doubling of the risk of death with gastrostomy feeding,31 assuming that over a five year period 10% of the orally fed children will die, would be 438 children (with 80% power and 95% confidence). For outcomes such as GOR or quality of life measure, substantially fewer children would need to be recruited.
GS was funded by the NHS Executive, South East Region, Research and Knowledge Management Directorate; however any views or opinions expressed do not necessarily reflect those of NHS Executive or the Department of Health. We wish to thank Nicola Bexon (Information Services Manager at the Institute of Health Sciences, Oxford) for help with the searching and obtaining papers; Jo Abbott (Trials Search Co-ordinator, Cochrane Developmental, Psychosocial and Learning Problems Group) for help with the search strategy and searching the Cochrane Library and her colleague Celia Almeida for searching the Portuguese version of Lilacs and translating the abstracts found. Thanks are also extended to Rachel Rowe (Researcher) and Simon Gates (Trials Researcher/Statistician) at the National Perinatal Epidemiology Unit, Oxford, for their helpful criticism of the draft manuscript, and Andreas Hassman and Catriona Ferguson for help with translating Swedish and Italian.