Gastroenteritis is an illness characterized by the acute onset of diarrhea, which may or may not be accompanied by nausea, vomiting, fever, and abdominal pain (AAP 1996). It can be caused by a variety of infectious agents including bacteria and viruses (Armon 2000). Acute diarrhea refers to the passage of loose or watery stools, usually at least three times per 24 hours, and lasting less than 14 days; the consistency of stools being more important than the frequency.
Mild cases of gastroenteritis are usually self-limiting and may cause mild dehydration, which can be treated or prevented by continued feeding and drinking more fluids. Children who lose a large volume of liquid stool may develop moderate or severe dehydration; in the most severe cases this can lead to death. These children should be given rehydration therapy in order to restore the lost fluids and electrolytes.
Worldwide, 12% of deaths among children less than five years are due to diarrhea (WHO 2000). In low-income and middle-income countries, an estimated 1.8 million children below the age of five years die of diarrhea each year (Bern 1992). Almost 50% of these deaths are due to dehydration and most affect children less than one year of age (WHO 1996). Children in high-income countries are also affected by diarrhea. In the USA, for example, each year there are roughly 21.5 to 38 million episodes of diarrhea among the 16.5 million children under the age of five years (Glass 1991). Diarrhea accounts for an estimated 2.1 to 3.7 million physician visits per year (Glass 1991) and 9% to 10% of all hospital admissions of children under the age of 5 years (Glass 1991; Gangarosa 1992). Approximately one per 15,000 children born in the USA or one per 500 children hospitalized with gastrointestinal illness will die of their illness (Glass 1991).
Widespread use of oral rehydration salt (ORS) solutions began in the 1970s as an effective and inexpensive method of treating diarrhea in low-income and middle-income countries (Duggan 1992). The basis for their use lies in the knowledge that glucose enhances sodium and water absorption in the bowel, even during diarrhea (Mackenzie 1988; Duggan 1992). There are a number of ORS solutions that vary in terms of their electrolyte and carbohydrate concentrations (Santosham 1991). The World Health Organization recommends a specific formulation of ORS solution for both rehydration and maintenance of hydration (WHO 2002). It has an osmolarity similar to that of plasma and contains citrate to correct metabolic acidosis and a glucose concentration that allows maximum absorption of sodium and water.
Despite its success and proven efficacy (Gavin 1996), and recommendations for use by the American Academy of Pediatrics and the Centers for Disease Control (Duggan 1992; AAP 1996), oral rehydration therapy (ORT) continues to be less frequently used by family physicians and pediatricians in North America, where intravenous therapy (IVT) is more in vogue (Snyder 1991; Ozuah 2002).
IVT, usually with normal saline or Ringer's lactate (AAP 1996), is familiar to physicians and is rapid and effective in promptly reversing cases of hypovolemic (low blood volume) shock. Since it must be administered in an outpatient or inpatient setting by specially trained staff, it is expensive in terms of money and human resources. In addition, IVT is a traumatic experience for most children and is known to have complications related to rapid over correction of electrolyte imbalances (WHO 1995), leaking of solutions into surrounding tissues (Garland 1992), and infection or inflammation (Garland 1992).
ORT is colloquially used in the literature as a substitute for the perhaps more appropriate term enteral rehydration therapy. It is important to note that rehydration can be provided enterally in two manners: orally or via nasogastric tube. Although ORT is not widely popular in developed countries because it is thought to take extra time and effort (Goepp 1993), it has many advantages. If administered by mouth, it is less traumatic to the child and can be administered by caregivers in a variety of settings including the home (Mackenzie 1988; AAP 1996). Research has shown ORT to be less expensive than IVT and to be associated with lower hospital admission rates and shorter lengths of stay (Listernick 1986; Gremse 1995). ORT is not recommended if the child has paralytic ileus or glucose malabsorption (Duggan 1992; WHO 1995), which are rare events. In both these clinical scenarios, fluid remains in the gut lumen rather than being absorbed into the intravascular space where the body can use it. Further delivery of fluid then just causes abdominal distention.
In 1996, Gavin and colleagues published the results of a meta-analysis that evaluated the efficacy of glucose-based ORT among well-nourished children in developed countries (Gavin 1996). The review included six studies that compared ORT with IVT and seven studies that compared ORS solutions with different sodium contents. They found that failure of ORT (defined as the need to revert to IVT) varied among trials, ranging from 0% to 18.8% with an overall failure rate of 3.6% (95% confidence interval 1.4 to 5.8). They found no significant difference in failure between different ORS solutions. They also found no higher risk of iatrogenic hypernatremia or hyponatremia with ORT compared with IVT and no significant differences in failure rates between inpatients and outpatients. The authors suggested that ORT may, in fact, be associated with more favourable outcomes such as increased weight gain and shorter duration of diarrhea.
The NHS Centre for Reviews and Dissemination (CRD) at the University of York, England critically appraised the Gavin 1996 review. While there was little detail in the paper on the individual studies reviewed and some aspects of the methods used for the review, the appraisal concluded that sufficient information was presented to suggest that the findings were likely reliable (DARE 2002). In addition, we evaluated the review by applying Oxman and Guyatt's index of the scientific quality of research overviews (Oxman 1991). The weaknesses identified by the Oxman and Guyatt index included the limited search (MEDLINE up to 1993 and contact with organizations and content experts; English language articles only) and the lack of assessment and consideration of the validity of the included studies. The purpose of the present review is to update and build on the work started by Gavin and colleagues by increasing the scope (countries of all income levels, method of administration of ORT) and comprehensiveness (broader search strategy, inclusion not limited by language of publication or publication status), and by assessing the risk of bias in the included studies.
To compare oral with intravenous therapy for treating dehydration due to acute gastroenteritis in children.
Criteria for considering studies for this review
Types of studies
Randomized and quasi-randomized controlled trials.
Types of participants
Children up to 18 years of age with dehydration secondary to acute gastroenteritis.
We included hospital inpatients and outpatients.
Acute gastroenteritis was defined as the rapid onset of diarrhea (lasting less than 14 days) with or without nausea, vomiting, fever, or abdominal pain (AAP 1996). Diarrhea is the passage of loose or watery stools, usually with increased frequency and volume (Baldassano 1991). Dehydration due to diarrhea is a "deficiency of water and salt" (Santosham 1991), and is most often assessed in terms of the percentage of weight lost during the dehydrating episode (Armon 2000). The severity of dehydration can be classified as mild (3% to 5%), moderate (6% to 9%), or severe (10% or greater) (Duggan 1992; AAP 1996).
Types of interventions
Oral rehydration therapy administered orally or via nasogastric tube.
Types of outcome measures
- Failure of rehydration or failure to maintain hydration after initial rehydration (as defined in the trials).
- Weight gain.
- Length of hospital stay for inpatients.
- Hypernatremia (excessive concentration of sodium in the blood).
- Hyponatremia (reduced concentration of sodium in the blood).
- Duration of diarrhea.
- Total fluid intake.
- Sodium intake and sodium levels.
- Complications and adverse events.
Search methods for identification of studies
We attempted to identify all relevant trials regardless of language or publication status (published, unpublished, in press, and in progress).
We searched the following databases using the search terms and strategy described in Appendix 1: Cochrane Infectious Diseases Group Specialized Register (March 2006);Cochrane Central Register of Controlled Trials (CENTRAL), published in The Cochrane Library (2006, Issue 1); MEDLINE (1966 to March 2006); EMBASE (1974 to March 2006); and LILACS (1982 to March 2006).
Researchers, organizations, and pharmaceutical companies
We contacted the World Health Organization Division of Child and Adolescent Health, Technologies for Primary Health Care Project (sponsored by US Agency for International Development), International Child Health Foundation, International Children's Centre, International Center for Diarrhoeal Diseases Research, and individual researchers working in the field for unpublished data, confidential reports, and raw data of published trials. We also contacted the manufacturers of Pedialyte (Ross Products Division, Abbott Laboratories) and Gastrolyte (Aventis Pharma Inc.) for any unpublished information or studies they may have possessed information pertaining to the efficacy of ORT.
Data collection and analysis
Selection of studies
Two authors (Steven Bellemare (SB) and Kelly Russell (KR)) independently screened the results of the literature search. The full text of all potentially relevant articles was retrieved. Two authors (KR, Don McConnell (DM), or Lisa Hartling (LH)) independently assessed the trials for inclusion in the review using predetermined inclusion criteria. We resolved any disagreements through discussion or by consulting a third party.
Data extraction and management
One author (SB or LH) extracted data from the included trials, and a second author (KR) checked the data for completeness and accuracy. We requested unpublished data from authors where necessary. We used a standard data extraction form to extract data on trial characteristics including methods, participants, interventions, and outcomes. We also collected information on source of funding and intention-to-treat (whether an intention-to-treat analysis was planned and whether an intention-to-treat analysis was done). We resolved any disagreements by referring to the trial report and through discussion.
Assessment of risk of bias in included studies
Two authors (SB, William Craig (WC), or LH) evaluated each trial using the Jadad 5-point scale to assess randomization (0 to 2 points), double blinding (0 to 2 points), and withdrawals and dropouts (0 to 1 point) (Jadad 1996). SB, WC, or LH also assessed concealment of allocation as being adequate, inadequate, or unclear (Schulz 1995). We resolved any differences through consensus or by consulting a third party. We provided overall quality scores according to the Jadad scale. We also described and displayed the quality information by individual components − generation of random sequence, blinding, loss to follow up, and allocation concealment − which we classified as adequate, inadequate, or unclear.
Since we had frequent zero event rates (per group, per trial), we expressed dichotomous data (eg failure of treatment) as a risk difference (RD) rather than using a relative measure that would force the trial data to be omitted or approximated (adding ½ to each 2 x 2 table cell) (Higgins 2005a). We also calculated the number needed to treat (NNT) to help clarify the degree of benefit for the baseline intravenous risk. We calculated baseline risks using the same weights calculated from the risk difference analyses. We converted continuous data to a mean difference and then calculated an overall mean difference. We analyzed the results in Review Manager 4.2 using a random-effects model and presented all estimates with 95% confidence intervals (CI). We analyzed the longitudinal outcome of total fluid intake for two time points: six and 24 hours. One trial assessed total fluid intake at four hours (Spandorfer 2005), while two trials assessed total fluid intake at eight hours (Santosham 1982i; Santosham 1982ii); these trials were included in the analysis for total fluid intake at six hours.
We quantified statistical heterogeneity using the I
The relationship between the osmolarity of the ORT solution and failure to rehydrate was explored using a chi-square subgroup test (Deeks 2001). We calculated osmolarity based on the reported constituent concentrations in the solutions (Appendix 2). We also examined the choice of model for sensitivity to the results (eg Mantel-Haenszel fixed-effect model). We performed intention-to-treat analyses for our primary outcome measure (failure to rehydrate). After viewing the trials, we created a more homogeneous failure definition and applied it to each article in order to check the robustness of the primary results (ie using failure to rehydrate as defined in the trials). The decision to examine persistent vomiting as an exclusion criterion was selected post hoc for subgroup analysis; all other subgroups were specified a priori. We identified and explored all statistical outliers.
We tested for publication bias using the funnel plot visually and quantitatively using the rank correlation test (Begg 1994), weighted regression (Egger 1997), and the trim-and-fill method (Duval 2001) in Stata 7.0.
When standard deviations were not reported, we performed a sensitivity analysis by substituting both the minimum and maximum standard deviations from the other outcome-specific included studies to gauge whether these omitted studies might significantly alter our results.
Description of studies
We identified 466 unique references through the electronic databases and obtained the full text of 28 potentially relevant articles. We identified 11 additional studies via recommendations from authors and experts in the field and references. Seventeen of these met the inclusion criteria (see 'Characteristics of included studies'). One trial was conducted concurrently and slightly differently in two countries; we have analyzed this as two separate studies (Santosham 1982i; Santosham 1982ii). The reasons for excluding studies are noted in the 'Characteristics of excluded studies'.
All trials compared an IVT arm with one or more ORT arms (oral or nasogastric). The trials varied widely in methodology and quality. They were published from 1982 to 2005 and though most were randomized, two trials were quasi-randomized (Singh 1982; Mackenzie 1991).
The trials varied in their countries of origin. Nine were conducted in high-income countries: six in the USA (Tamer 1985; Listernick 1986; Gremse 1995; Atherly-John 2002; Nager 2002; Spandorfer 2005); one in Canada (Issenman 1993); one in Australia (Mackenzie 1991); and one in Finland (Vesikari 1987). The other trials were from lower income countries such as Puerto Rico (de Pumarejo 1990), Egypt (el-Mougi 1994), Mexico (Gonzalez 1988), Iran (Sharifi 1985), Afghanistan (Singh 1982), Colombia (Hernandez 1987), and Peru (Brown 1988). One trial was conducted in the USA and Panama simultaneously (Santosham 1982i; Santosham 1982ii).
Of the trials that mentioned the source of funding, five received funding or sponsorship from the pharmaceutical industry (Ross or Abbott Laboratories) (Santosham 1982i; Listernick 1986; de Pumarejo 1990; Issenman 1993; Gremse 1995), one trial received funding from other external sources (Santosham 1982i), one trial received funding from Nestle, the World Health Organization, and the US government (Brown 1988), and two trials used Pedialyte in their protocol but did not specifically acknowledge Ross Laboratories (the manufacturer of Pedialyte) as a funding source (Nager 2002; Spandorfer 2005).
Inclusion and exclusion criteria
Most trials were similar in their inclusion and exclusion criteria except in the area of treatment of children with persistent vomiting and severe dehydration and shock. All trials excluded children in shock except for two (Santosham 1982i; Santosham 1982ii; Sharifi 1985), and three trials did not mention whether they included or excluded these children (Hernandez 1987; Vesikari 1987; Brown 1988). All children in the Santosham trials presenting in either severe dehydration or shock were first treated immediately with IVT to reverse the condition before being randomized to a treatment group.
Three trials enrolled children with only dehydration secondary to acute diarrhea and made no mention of vomiting in the inclusion criteria (Brown 1988; Vesikari 1987; el-Mougi 1994). Of the remaining trials, five excluded children with persistent or protracted vomiting (Singh 1982; de Pumarejo 1990; Issenman 1993; Atherly-John 2002; Nager 2002), while four trials included these children (Spandorfer 2005; Sharifi 1985; Listernick 1986; Gremse 1995). The remaining authors did not elaborate on the inclusion or exclusion of children with persistent vomiting (Santosham 1982i; Santosham 1982ii; Tamer 1985; Hernandez 1987; Gonzalez 1988; Mackenzie 1991).
The populations studied were similar: 1015 (56%) people were randomized to the ORT group and 796 (44%) people to the IVT group. Overall more people were randomized to ORT as some trials included more than one ORT group (Santosham 1982i; Santosham 1982ii; Hernandez 1987; Brown 1988; el-Mougi 1994). Most trials included children from three months to five years of age. The other trials included children from eight weeks to three years (Spandorfer 2005), children up to the age of 17 years (Atherly-John 2002), or neonates aged less than 14 days (Vesikari 1987; Gonzalez 1988; de Pumarejo 1990). One trial did not specify the age range (Singh 1982). Finally, Hernandez 1987 enrolled children less than five years but did not specify a lower age limit (15.3% of their population was ≤ 3 months).
ORS solutions: description and administration
All the trials used ORS solutions containing glucose (75 to 144 mEq/L) or dextrose (70 to 139 mEq/L), as well as sodium (45 to 90 mEq/L), potassium (13 to 30 mEq/L), and chloride (35 to 80 mEq/L) (Appendix 2). One trial used a combination of glucose and fructose as the carbohydrate component (Listernick 1986). All the trials used either citrate or bicarbonate in their ORS. Eight trials reported the osmolarity of the ORS solutions (Santosham 1982i; Santosham 1982ii; Singh 1982; Sharifi 1985; Hernandez 1987; de Pumarejo 1990; Issenman 1993; el-Mougi 1994; Nager 2002), which ranged from 210 to 390 mmol/L. One trial did not report on the makeup or osmolarity of the ORS used (Atherly-John 2002).
The route of administration of ORT differed. Some trials administered by mouth only (Santosham 1982i; Santosham 1982ii; Singh 1982; el-Moughi 1983; Tamer 1985; Listernick 1986; Brown 1988; de Pumarejo 1990; Issenman 1993; Atherly-John 2002; Spandorfer 2005), while others administered it by mouth, using an nasogastric tube only when required (Hernandez 1987; Vesikari 1987; Gonzalez 1988; Mackenzie 1991). Children randomized to the ORT arm received ORS exclusively via nasogastric tube in one trial (Gremse 1995); but before study enrolment children in both arms had failed a prior uncontrolled trial of ORS administered by mouth. One trial administered ORS exclusively via nasogastric tube (Nager 2002), while another gave ORS via nasogastric tube in the rehydration phase of the trial and by mouth in the maintenance phase (Sharifi 1985).
All trials reported on the primary outcome measure of interest: failure to rehydrate using ORT. However, the definition of failure of ORT varied. While some trials counted children taking ORS by mouth with persistent vomiting as treatment failures, others inserted nasogastric tubes in these children, thus giving ORT more chances of success in the process (Hernandez 1987; Vesikari 1987; Gonzalez 1988; Mackenzie 1991). Only three trials reported data on deaths (Singh 1982; el-Mougi 1994; Sharifi 1985), while additional data were obtained from a fourth (Mackenzie 1991). Secondary outcome measures, when not determined a priori, were tabulated as they were reported in individual trials.
Risk of bias in included studies
The quality of the studies ranged from zero to three on the Jadad scale (median two). Since none of the trials could be double blind due to the nature of the intervention, the maximum Jadad score was three rather than the conventional five. Two trials scored a three (Atherly-John 2002; Spandorfer 2005). One trial scored a zero on the Jadad scale as it was described as randomized (one point) but used an inappropriate method of randomization (point deducted) (Singh 1982). Six trials scored a one on the Jadad score (Sharifi 1985; Hernandez 1987; Gonzalez 1988; de Pumarejo 1990; el-Mougi 1994; Gremse 1995). All received a point for randomization. The seven remaining trials scored a two: one point for randomization and one point for reporting withdrawals. The quality components for each trial are detailed in Appendix 3.
Two trials had adequately concealed allocation (Atherly-John 2002; Spandorfer 2005); it was unclear in the remaining trials. Although four trials had incomplete follow up and one other (Issenman 1993) counted a withdrawal as a failure (Tamer 1985; Brown 1988; Mackenzie 1991; Nager 2002), only two trials reported doing an intention-to-treat analysis (Atherly-John 2002; Spandorfer 2005).
Effects of interventions
Failure to rehydrate
There was a statistically significant difference in failure to rehydrate between treatment groups (RD 4%, 95% CI 1 to 7; 1811 participants, 18 trials, Analysis 1.1). The NNT was 25 (95% CI 14 to 100), and the failure risks were 4.9% for ORT and 1.3% for IVT. Failure definitions varied by trial and are discussed later. The results for failure to rehydrate were not sensitive to the choice of model, as the fixed-effect model also favoured IVT (RD 4%, 95% CI 2 to 6; NNT 25, 95% CI 17 to 50).
Gonzalez 1988 was a statistical outlier in this analysis because its risk difference was large given its sample size (RD 13%, 95% CI 6 to 20; 200 participants). An influence plot and the Galbraith plot show evidence to this effect (Figure 1 and Figure 2). Removing this trial shifted the overall risk difference closer to the null, but the result was still statistically significant (RD 2%, 95% CI 0.08 to 5; NNT 50, 95% CI 20 to 1250) and reduced the heterogeneity (from I
|Figure 1. Influence plot: meta-analysis using random-effects model (linear form)|
|Figure 2. Galbraith plot|
Gremse 1995 and Nager 2002 met all the inclusion criteria but randomized the participants to the nasogastric or intravenous route; this occurred after the participants had already failed an uncontrolled trial of ORT in Gremse 1995. These trials were neither outliers nor influential.
Three trials reported deaths, and a fourth author provided supplemental data (Mackenzie 1991). Singh 1982 reported that all children were successfully rehydrated; however one participant succumbed to a "severe pyrogenic reaction". el-Mougi 1994 reported one death in the IVT group due to pneumonia and ileus. Sharifi 1985 reported seven deaths: two in the ORT group and five in the IVT group. The cause of death was not reported although four of the seven deaths occurred in participants below the third percentile weight class. Mackenzie 1991 reported that no deaths occurred. All reported deaths occurred in low-middle income countries (UN Statistics).
Weight gain at discharge
There was no statistically significant difference in weight gain between treatment groups (WMD -26.33 g, 95% CI -206.92 to 154.26; 369 participants, 6 trials, Analysis 1.2), but there was substantial heterogeneity (I
Length of hospital stay for inpatients
Children treated with ORT spent less time in hospital (WMD -1.20 days, 95% CI -2.38 to -0.02, I
The combined estimate of the two trials reporting on this did not show a significant difference (RD 1%, 95% CI -13 to 15, I
The number of cases of hypernatremia was not statistically different between treatments (RD 0%, 95% CI -1 to 1, I
Duration of diarrhea
The mean length of diarrhea was not statistically different between groups (WMD -5.90 h, 95% CI -12.70 to 0.89, I
Total fluid intake
Total fluid intake did not differ significantly between the two groups at six hours after starting treatment (WMD 32.09 mL/kg, 95% CI -26.69 to 90.88, I
Complications and adverse events
There were statistically significantly more children with paralytic ileus in the ORT group when analyzed using the fixed-effect model (RD 3%, 95% CI 1 to 5, IVT risk 0%; I
Sodium intake and sodium levels
Sodium intake at six hours was not statistically significantly different between the ORT and IVT groups (WMD 5.80 mmol/kg, 95% CI -1.48 to 13.07, I
Subgroup and sensitivity analyses
We explored participant status (inpatient versus outpatient), state of nourishment (well nourished versus some malnourished), country's income (low-middle income versus high-income; UN Statistics), funding source (funded versus not reported), allocation concealment (adequate versus unclear), and Jadad scores (0, 1, 2) in a meta-regression using failure to rehydrate as the dependent variable. None were found to be statistically significant (Appendix 4). The remaining a priori subgroup comparisons were not reported by subgroup (age and extent of dehydration) and could not be analyzed. Although two trials reported that more than 20% of participants were severely dehydrated (Singh 1982; Tamer 1985), neither was considered an outlier.
The definition of "failure" varied by study. We evaluated the sensitivity of a more homogeneous definition in which we limited failures to children with persistent vomiting, having some level of dehydration persisting, and experiencing shock or seizures. (We excluded children with paralytic ileus, intussusception, cerebral palsy, septicemia, urinary tract infection, and duodenal ulcer from this analysis.) This post hoc failure definition was statistically significant and favoured IVT for the fixed-effect model (RD 2%, 95% CI +0 to 4) but not for the random-effects model (RD 2%, 95% CI -0 to 4) ( Analysis 2.1). The heterogeneity was also reduced using our homogeneous definition (from 70% to 37%). Subsequently participants who had withdrawn or dropped out (only ORT participants) were reclassified as failures, as in a worst-case intention-to-treat analysis. With this analysis, there were statistically significant differences in failure rate between treatment groups that favored the IVT group when analyzed using the fixed-effect model (RD 3%, 95% CI 1 to 5) and the random-effects model (RD 3%, 95% CI 0 to 5) ( Analysis 2.2). Heterogeneity was also reduced using this intention-to-treat analysis (from 70% to 48.1%).
Since one of the assumptions for performing a meta-regression was not met (ie the constant variance assumption), we did not explore the osmolarity subgroups with this method. Instead, we divided the trials into low osmolarity (range 208 to 270 mOsmol/L) and high osmolarity (range 299 to 331 mOsmol/L) subgroups; these cut-offs were defined post hoc based on those used in another review (Hahn 2001). The difference found by the chi-square subgroup test (Deeks 2001) was statistically significant (P < 0.0001). The risk difference for the low osmolarity group was 1% (95% CI -1 to 2) and it was homogeneous (I
We used meta-regression to examine further subgroups: inclusion and exclusion criteria for participants with persistent vomiting (post-hoc) ( Analysis 2.4) as well as the route of ORT administration (nasogastric versus oral versus a combination) ( Analysis 2.5). Neither analysis resulted in statistically significant differences, although the analysis stratified by whether or not the trial excluded participants with persistent vomiting suggests that there may be differences given sufficient power.
The rank correlation test did not indicate any publication bias (r = 23, P = 0.41). The weighted regression analysis did show a significant indication of funnel plot asymmetry (bias = 0.9, P = 0.02). The trim-and-fill method indicated five missing studies; the adjustment to overall effect size rendered the new estimate non-significant by moving it closer to the null (RD 2%, 95% CI -1 to 4). The funnel plot appears somewhat asymmetrical (Figure 3); publication bias may be present suggesting that missing studies are more likely to favour ORT.
|Figure 3. Funnel plot for primary outcome (failure to rehydrate) based on fixed-effect model|
The most conservative model showed no important clinical differences in failure to rehydrate between ORT and IVT in terms of efficacy (RD 4%, fixed-effect model). For every 25 children treated with ORT, one would fail and require IVT (ie NNT = 25, CI 14 to 100). These overall findings are similar to those found in an earlier review (Gavin 1996). The results were consistent among different populations, such as children with different states of nourishment. Moreover the results for low osmolarity solutions, the currently recommended treatment by the World Health Organization, showed a lower failure rate for ORT that was not significantly different from the failure rate of IVT (RD 1%, NNT = 100). These results support existing practice guidelines recommending ORT with a low osmolarity solution as the first course of treatment in children with dehydration secondary to gastroenteritis. A cumulative meta-graph, which adds studies by ascending year, showed that the overall estimate is unlikely to change substantially with further trials (Figure 4). Overall we studied more than 1800 children providing adequate power to support the observed results.
|Figure 4. Cumulative plot: failure cumulative by year|
Most trials were small and of poor to moderate quality. The median quality score according to the Jadad scale was two. Because it is impossible to double blind studies on this topic, the quality of studies was limited to a maximum of three (rather than five). It is important to note that studies that are not double blind can lead to an overestimate of the treatment effect (Colditz 1989) and that this is an inherent limitation in this body of literature. While double blinding is probably not feasible in a trial comparing IVT and ORT, allocation can always be properly concealed. Allocation concealment was unclear in all but two trials; this can lead to an overestimate of treatment effects by as much as 40% (Schulz 1995). These factors could skew the results in favour of either ORT or IVT depending on the biases of the investigators.
Additionally, adverse events were not systematically sought in most of the trials. Though this meta-analysis had adequate power to support the efficacy of ORT, it lacks the power to detect serious but rare adverse events in either treatment group. However, analysis of the currently available data suggests that ORT and IVT are similar in safety profiles. Based on two trials, paralytic ileus occurred significantly more often in the ORT group (using the fixed-effect but not random-effects model); however it would not be deemed common enough to discourage the use of ORT. Further there was a statistically significant difference in the occurrence of phlebitis, but phlebitis cannot occur when an IV is not used, thus the statistical significance does not equate to a clinically important difference.
In applying the evidence to clinical practice, the objective output of the meta-analysis must be weighed with other less easily measured factors that support the use of ORT. On a theoretical basis, IVT should be able to replace the fluid already lost, as well as keep up with the ongoing losses even if those losses are through vomiting, diarrhea, or third spacing within the lumen of the gut. For IVT to be effective, the correct fluid and rate needs to be chosen, for errors in either of these parameters can lead to harm as severe as death. IVT is viewed as the gold standard for children in shock or with severe dehydration, and most of the papers reviewed excluded children from their studies on this basis. Besides the complication of IVT fluid type and rate, starting an intravenous intervention causes pain, the attempt can be unsuccessful, phlebitis (inflammation of the vein) can occur, a cellulitis can result and the intravenous can go interstitial resulting in the intravenous fluid going into the immediate surrounding tissue rather than into the vein and therefore the intravascular space. ORT can be performed by almost anyone with very little training and has the advantage that a child's thirst can moderate the quantity and rate of fluid administration. It does not work if the fluid is not being absorbed by the gut (paralytic ileus), and in some cases, the rate of diarrhea increases as oral fluid rate increases so that the child remains in a net negative fluid balance. Once the ORT is administered via a nasogastric tube, some of the theoretical advantages of ORT disappear, and some theoretical disadvantages need to be considered. The child is no longer able to control the rate or amount of fluid intake and so operator errors may occur just as they do with IVT. Errors in placement of the nasogastric tube can occur; the most severe of these is passing the nasogastric tube into the trachea so that the fluid is going into the lungs rather than the stomach. The passing of a nasogastric tube can result in a bleeding nose and discomfort. The tube does not always pass easily on the first attempt. There is evidence to suggest that ORT is less costly than IVT and can be administered as rapidly (Nager 2002). A randomized controlled trial has demonstrated that the use of ORT in a high-income country pediatric emergency department resulted in statistically significantly lower costs, less time spent in the emergency department, and a more favorable impression of caregivers for this form of therapy (Atherly-John 2002).
Though there was little statistical heterogeneity between trials on failure to rehydrate when the outlying trial was omitted (Gonzalez 1988) (I
Another source of variation was the definition of "treatment failure". We examined the effect of different treatment failure definitions through a post hoc refined definition analysis and found that it reduced heterogeneity and the therapeutic benefits of IVT as compared with ORT (from 4% to 2%). The intention-to-treat version of this model involved reclassifying seven ORT withdrawals as failures; this model also reduced heterogeneity and benefits for IVT (from 4% to 3%). If the seven withdrawals were systematically related to treatment benefit, then this intention-to-treat analysis is less biased.
One trial had a statistically significantly greater failure rate (Gonzalez 1988). The trial authors attributed it to the fact that many of the children who failed were younger than six months of age. (This trial was the only one to include neonates.) The authors argued that the burden of illness can be more severe in younger infants. Our data neither prove nor disprove this statement. When we removed this trial from the analysis, the remaining trial results were homogeneous and the overall risk difference shifted towards the null.
The results may not be generalizable to all children with dehydration secondary to gastroenteritis but may be generalizable only to those with dehydration secondary to diarrhea. The reader must also recognize that most of the trials excluded children in shock, severe dehydration, and paralytic ileus since IVT is the indicated treatment for these clinical scenarios. A post hoc look between trials with different inclusion and exclusion criteria suggests that there may be an important difference in response to ORT among participants that vomited and did not vomit. The risk difference for trials that excluded participants with persistent vomiting was 0% (95% CI -3 to 3) as compared with 4% (95% CI -5 to 13) in trials that did not exclude such participants (Analysis 01.17). The issue of how vomiting affects the efficacy of ORT needs further study. However, in practice treatment failure only means than one switches to IVT.
Implications for practice
There were no important clinical differences between ORT and IVT for rehydration secondary to acute gastroenteritis in children. It seems reasonable that children presenting for medical care with mild to moderate dehydration secondary to acute gastroenteritis should initially be treated with ORT. Should treatment fail, then IVT may be used. In children who have persistent vomiting, ORT may be used, but the child must be closely observed for proof of successful treatment.
For every 25 children treated with ORT, one would fail and require IVT. Clinicians and families need to apply this evidence to individual situations in order to decide whether they are willing to accept this minimal risk.
Implications for research
Further research comparing ORT and IVT for children with dehydration secondary to gastroenteritis is not warranted and may be unethical. If undertaken, further research should focus on evaluating the efficacy of nasogastric rehydration in children who have persistent vomiting.
We thank Ellen Crumley, Marlene Dorgan, and Carol Friesen for their assistance with searching, Ben Vandermeer for assistance re-running the statistical analysis, and Maria Ospina for assistance with interpretation of Spanish articles. The editorial base for the Cochrane Infectious Diseases Group is funded by the UK Department for International Development (DFID) for the benefit of developing countries.
Data and analyses
- Top of page
- Authors' conclusions
- Data and analyses
- What's new
- Contributions of authors
- Declarations of interest
- Sources of support
- Differences between protocol and review
- Index terms
Appendix 1. Search methods: detailed search strategies
Appendix 2. Description of oral rehydration solutions
Appendix 3. Risk of bias (methodological quality)
Appendix 4. Meta-regression results
Last assessed as up-to-date: 28 March 2006.
Protocol first published: Issue 3, 2003
Review first published: Issue 3, 2006
Contributions of authors
L Hartling provided overall project coordination and contributed to protocol development, literature searching, screening of articles for relevance and inclusion, assessment of study quality, data extraction, data analysis, and preparation of completed review. S Bellemare contributed to protocol development, literature searching, relevance screening of articles, assessment of study quality, data extraction, data analysis, and preparation of completed review. N Wiebe conducted the statistical analysis and contributed to the protocol development and preparation of completed review. K Russell contributed to screening of articles for relevance and inclusion, data extraction and entry, and preparation of completed review. T Klassen contributed to protocol development and preparation of completed review, and provided methodological and content expertise. W Craig contributed to protocol development, quality assessment, preparation of completed review, and provided content expertise.
Declarations of interest
Sources of support
- Alberta Research Centre for Child Health Evidence, Canada.
- Alberta Heritage Foundation for Medical Research, Canada.
Differences between protocol and review
2006, Issue 3 (deviations from protocol).
- We used risk difference for our primary outcome instead of risk ratio because we have too many trials with zeros in both treatment groups; we did say how we would calculate baseline risk in the protocol, and we are simply expanding upon this.
- Our primary analysis changed from using the fixed-effect model to the random-effects model before we looked at the data; this change was based on the comments from statisticians in Cochrane's Statistical Methods Group.
- We chose to use the I-squared statistic rather than the chi-square test for heterogeneity based on comments from statisticians in Cochrane's Statistical Methods Group.
- Heterogeneity is always present, but it may not always be quantifiable (too small). We used the fixed-effect model in the sensitivity analyses to give a sense of whether the treatment effect may ever be significant, because the random-effects model may be biased when there is funnel plot asymmetry, and for a conservative approach to an equivalence hypothesis if there could be one.
- Omitting the trim-and-fill method from the protocol was an oversight.
- We added the participant subgroup inpatient/outpatient status before looking at the data, but after the protocol was published.
- We added 'sodium intake and sodium levels' as an outcome measure; since hyponatremia and hypernatremia were part of the protocol, this makes sodium levels and intake relevant, and we therefore chose to include this outcome measure post-hoc as it was frequently reported in the included trials.
Medical Subject Headings (MeSH)
Administration, Oral; Dehydration [etiology; *therapy]; Fluid Therapy [*methods]; Gastroenteritis [*complications]; Infusions, Intravenous; Randomized Controlled Trials as Topic; Rehydration Solutions [*administration & dosage]
MeSH check words