We will implement this protocol according to PRISMA-P [Preferred Reporting Items for Systematic Reviews and Meta-Analyses Protocols] guidelines[28] (see Additional file 1). We have completed the registration of this protocol in PROSPERO (International Prospective Register of Systematic Reviews; registration number: CRD42021244923 ). We will update the modification of the protocol at any time in the future.
Search strategy
We will search in Embase, PubMed, Medline, and Cochrane Database of Systematic Reviews from the time between the establishment of the database and the formal start of retrieval. We will manually search for references to eligible papers. We will only include English literature.
We will search following terms: (1) Probiotic, Prebiotic, and synbiotics; (2) Child, Adolescent,and infants; (3) metanalys*[tw] OR meta-analys*[tw]; (4) randomized controlled trial. Taking PubMed as an example, the detailed search terms are listed in Table 1 below.
Table 1: the search strategy for PubMed
#1
|
Probiotic OR Probiotics OR Prebiotic OR Prebiotics OR synbiotics OR Synbiotic
|
#2
|
Children OR Child OR PRESCHOOL OR Preschool Child OR Children, Preschool OR Preschool Children OR Adolescent OR Adolescents OR Adolescence OR Teens OR Teen OR Teenagers OR Teenager OR Youth OR Youths OR Adolescents, Female OR Adolescent, Female OR Female Adolescent OR Female Adolescents OR Adolescents, Male OR Adolescent, Male OR Male Adolescent OR Male Adolescents OR Infant OR Infants
|
#3
|
metanalys*[tw] OR meta-analys*[tw]
|
#4
|
“randomized controlled trial” OR “clinical trial” OR RCT OR “Clinical Trials” OR “controlled trial” OR “Intervention Study” OR “Intervention Studies” OR randomized OR “randomized controlled trial” or random
|
#5
|
#1 AND #2 AND #3 AND #4
|
Eligibility criteria
We will only include meta-analysis of randomized controlled trials which explored the effect of probiotic, prebiotic, and synbiotics on children, adolescents, or Infants. We will only consider articles that meet the following criteria:
- We will consider including populations who are under the age of 18.
- The difference between the experimental group and the control group is the presence or absence of probiotics, prebiotic, or synbiotic intervention. We will exclude these studies if the effect sizes observed are not related to the effect of probiotics, prebiotic, or synbiotic.
- We will also exclude meta-analyses that do not report the necessary data, such as the relative risk (RR), effective rate, or odds ratio risk (OR) and 95% confidence intervals (CI).
- Meta-analyses of randomized controlled trials that are published in full-text format and in English comply with systematic review criteria. We will include other grey literature sources. We will not include network meta-analysis, letter to the editor, abstract, protocols, or conference poster. We will also exclude literature where complete data are unavailable after connecting with the author.
In order to include the relevant literature as comprehensively as possible, we will not set a limit for the publication date. Systematic reviews conducted by different authors are likely to contain partially overlapping literature if the exposures and results of these studies are the same, leading to biased conclusions.[29] In addition, the median updating time of such literature are 5.5 years for the systematic review.[30] Referencing the criteria of the previous umbrella review, [31] we the literature into two categories, namely, outdated (more than 6 years after the beginning of our search) and current (less than 6 years after the beginning of our search). If there are multiple different studies evaluating the same exposures and outcomes, and some of these studies are outdated and some are current, we will exclude the outdated studies during the full-text screening phase. If multiple articles evaluating the same exposures and outcomes are only outdated or not, we will calculate the degree of overlap by the corrected covered area (CCA). [29] The method used to calculate CCA will refer to previously published literature.[29, 31] We will express CCA by percentage, and the specific formula is (N− r)/(rc − r), where N means the number of primary literature included in the systematic review, r is the number of rows, and c is the number of columns.[29] See the table 2 for specific examples.
table 2: the example for calculating CCA
|
Review 1
|
Review 2
|
Review 3
|
Review 4
|
Primary literature 1
|
*
|
*
|
*
|
|
Primary literature 2
|
*
|
*
|
*
|
*
|
Primary literature 3
|
*
|
*
|
*
|
*
|
Primary literature 4
|
*
|
|
*
|
*
|
Primary literature 5
|
*
|
*
|
*
|
*
|
Primary literature 6
|
*
|
*
|
*
|
|
Primary literature 7
|
*
|
*
|
|
|
CCA:(23-7)/(7*4-7)=76.19%
And we will divide CCA into the following levels:very high (CCA >15%), high (CCA 11-15%), moderate (CCA 6-10%), or slight (CCA 0-5%).[29]
We will generate a graphical crossover table (citation matrix) that overlaps the systematic reviews as columns and the primary literature included in these systematic reviews as rows.[29] The selection criteria for overlapping systematic review after review are as follows: [31]
1.We will include all non-overlapping systematic review. For overlapping systematic review, we will first include the non-obsolete Cochrane literature.[32] By examining the effects of different inclusion criteria on the comprehensiveness and complexity of medical intervention reviews, some researchers concluded that the data obtained from the Cochrane review was the most comprehensive, and that the quality of Cochrane reviews was higher than that of systematic reviews in other journals. [31, 33]
2. IF there are a high degree of overlap (CCA ≥11%) between two or more reviews, we will select reviews as following grading criteria: the rating is highest or moderate quality by reference to the AMSTAR 2 quality assessment tool; most recent; pooled effect estimates or a meta-analysis is provided; and the number of studies or participants is the highest.[31, 32]
3. For a slight or moderate degree of overlap (CCA ≤10%), we will retain both reviews, and further compare their results.[31]
Literature screening
Two reviewers will independently select literature by reading the titles and abstracts. After removing articles by reading the titles and abstracts, we will further find the full text for the remained articles. Discrepancy will be solved by discussing with a third reviewer. we will handle the screening of studies in EndNote.
Data extraction and data synthesis
Two reviewers will extract the corresponding content by reference to the JBI Data Extraction Form for Review for Systematic Reviews and Research Syntheses.[34] (see Additional file 2) We will extract the following data items: first author’s name, publication year, name of diseases, the number of studies included in meta-analyses, total sample size, cases and controls, interventions,the dose of intervention, effect sizes and the corresponding 95% confidence interval, publication bias, and heterogeneity. The data will be analyzed by means of Epi Info (V.7.2) after we have entered the information.
We will recalculate the summary effect and 95% CIs using fixed-effect or random-effect models. The I2 statistic or Cochrane’s Q test will be used to determine the magnitude of heterogeneity. For Cochrane’s Q test, the result will be considered as significant heterogeneity when p<0.1; for the I 2 statistic, the result will be classified as significant heterogeneity when 50% ≤ I2. The 95% prediction interval (95% PI) for the effect size will be assessed. Publication bias and small-study effect will be estimated by Egger’s test(a p value<0.1) or Begg’s test(a p value<0.1). All statistical analyses will be conducted using Review Manager (RevMan, version 5.3 for Macintosh; The Cochrane Collaboration) and the statistical package PASW 20.0 for Macintosh (SPSS Inc.). If possible, we will conduct sensitivity, subgroup analyses, or meta-regression.
Assessment of methodological quality and assessment of certainty in the findings
Methodological quality will be assessed by checking A Measurement Tool to Assess Systematic Review version 2 (AMSTAR-2),[35] which including 16 items (7 critical domains and 9 non-critical domains). We will categorize the results as high, moderate, low, and critically low after assessing the methodological quality by AMSTAR-2.[35] The specific classification criteria are as follows: No or 1 non-critical weakness will be appraised as high; more than 1 non-critical weaknes will be appraised as moderate; 1 critical flaw with or without non-critical weakness will be appraised as low; if more than 1 critical flaw with or without non-critical weakness will be appraised as critically low.
In addition, we will take the GRADE (Grading of Recommendations Assessment, Development and Evaluation)[36] approach to assess the strength of evidence. The strength of evidence will be classified as high, moderate, low or very low.[36, 37]