Systematic Review With Meta-Analysis

Lactobacillus Rhamnosus GG for Treating Acute Gastroenteritis in Children

A 2019 Update

Hania Szajewska; Maciej Kołodziej; Dorota Gieruszczak-Białek; Agata Skórka; Marek Ruszczyński; Raanan Shamir


Aliment Pharmacol Ther. 2019;49(11):1376-1384. 

In This Article

Materials and Methods

The methodology was similar to one followed in our earlier systematic review on the same topic,[3] hence, the protocol was not registered. The guidelines from the Cochrane Collaboration for undertaking and reporting the results of a systematic review and meta-analysis[8] and the PRISMA statement[9] were followed for this systematic review and meta-analysis. Ethical approval was not needed.

Criteria for Considering Studies for This Review

All relevant randomised controlled trials (RCTs) that compared use of L rhamnosus GG (as a single ingredient, in all delivery vehicles and formulations, at any dose) with no L rhamnosusGG (defined as placebo or no treatment) were eligible for inclusion. The primary outcome measures were the duration of diarrhoea and stool volume. The secondary outcome measures were the effects of L rhamnosus GG on the course of diarrhoea, including the percentages of children with diarrhoea at various times intervals (as specified by the investigators), the percentage of children with diarrhoea lasting longer than 7 days, the duration of hospitalisation, and adverse effects. Other outcomes evaluated by the authors of the original trials were also considered, if relevant to this review.

Search Methods for Identification of Studies

The Cochrane Central Register of Controlled Trials (CENTRAL, the Cochrane Library), MEDLINE, and EMBASE databases were searched from May 2013 (end of last search) to January 2019. The principal search text word terms and MESH headings used were as follows: diarrhea/diarrhoea, diarrh*, gastroenteritis, probiotic*, Lactobacillus rhamnosus GG, Lactobacillus GG, LGG (for details, see Table S1). No language restrictions were imposed. The reference lists from identified studies and key systematic review/review articles, including previously published systematic reviews with or without a meta-analysis, were also searched to identify any other relevant studies. The and websites were also searched for RCTs that were registered but not yet published. Letters to the editor, abstracts, and proceedings from scientific meetings were excluded.

Data Collection and Analysis

Using a standardised form, one reviewer (MK) undertook the literature search, data extraction, and quality assessment. The data extracted included baseline characteristics, inclusion criteria, experimental and control treatments, setting, dose, and funding. All newly identified studies were evaluated by other reviewers.

Assessment of Risk of Bias in Included Studies

The Cochrane Collaboration's tool for assessing risk of bias was used. The risk of bias parameters included the type of randomisation method (selection bias), allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), and incomplete outcome data (attrition bias). Additionally, selective reporting (reporting bias) and other types of bias were considered. If an item could not be evaluated due to missing information, it was rated as having an unclear risk of bias.[8]

Measures of Treatment Effect

The dichotomous outcomes, the results for individual studies, and pooled statistics were reported as the risk ratio (RR) between the experimental and control groups with 95% CI. The continuous outcomes were reported as the mean difference (MD) between the treatment and control groups with 95% CI.

Dealing With Missing Data

We assessed pooled data using available case analysis, that is, an analysis in which data are analysed for every participant for whom the outcome was obtained, rather than intention-to-treat analysis with imputation.

Assessment of Heterogeneity

Heterogeneity was quantified by χ 2 and I 2, which can be interpreted as the percentage of the total variation between studies that is attributable to heterogeneity rather than to chance. A value of 0% indicates no observed heterogeneity, and larger values show increasing heterogeneity. All analyses were based on the random-effects model.

Assessment of Reporting Biases

When at least 10 RCTs were available, publication bias was assessed using the funnel plot proposed by Egger et al[10] A P < 0.05 implicates publication bias.

Data Synthesis (Statistical Methods)

The data were analysed using Review Manager (RevMan [Computer program]. Version 5.3. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). As reported earlier, in the case of one RCT (Basu et al 2009[11]), we combined two experimental arms (L rhamnosus GG at 2 × 1010 CFU or 2 × 1012 CFU) into a single experimental group according to the method of Hogg and Craig.[12] In the trial by Misra et al[13] (reporting only mean and median), missing standard deviations (SD) were obtained using the formula recommended by Higgins et al[8] In four trials reporting median and IQR (Schnadower et al,[6] Czerwionka-Szaflarska et al,[14] Aggarwal et al,[15] and Sindhu et al[16]), missing means and SDs were estimated using the formula recommended by Hozo et al.[17]

Subgroup Analyses

For the primary outcomes, a priori subgroup analyses based on factors that could potentially influence the magnitude of the treatment response were planned for the following: (a) Dose of L rhamnosus GG (high dose [≥1010 CFU/day] vs lower dose [<1010 CFU/day]); (b) Setting (studies carried out in geographical Europe vs non-European countries); (c) Type of treatment (outpatient vs inpatient); (d) Aetiology of diarrhoea; and (e) Vaccination against rotavirus status.

We also performed a post hoc analysis based on the duration of diarrhoea prior to randomisation (≤5 days vs >7 days). Additionally, when there was statistically significant heterogeneity in the primary outcome across studies, subgroup analyses were performed to determine the impacts of allocation concealment (adequate vs inadequate and/or unclear) and blinding (open trial vs double-blind trials). We also performed a post hoc analysis based on the duration of diarrhoea to assess whether the findings were affected by including only trials at low risk of bias (defined as those with adequate randomisation, allocation concealment, blinding, and at least 90% follow-up).