Systematic review and meta analysis SOPs

From Wikipedia

Jump to: navigation, search

Contents

[edit]

Formulate research question

[Tip:start with very strict inclusion and exclusion criteria and then expand progressively depending on the number of available articles. this will avoid getting into a study with hundreds of articles which is possibly difficult to handle with quality]

[edit]

Define inclusion and exclusion criteria

Must define the following, apart from any other additional criteria

  • the diagnosis/indication should be same
  • the intervention, outcome and study sample
    • the intervention should be same
    • the intervention period and time interval should be similar
    • outcome measures should be the same or at least interchangeable (you can convert all the studies to a common measure) including the time when the outcome was measured
  • all papers should have the same study design (clustered within RCTs, cohorts, diagnostic or qualitative studies)
[edit]

Locate sources to find studies

The search for articles should be systematic, following the steps outlined below, to ensure not to miss any major study

  • identify bibliographic databases (examples given below)
  • if there is a reason that the intervention being studied for the review might have early studies, then print versions of the databases like Index Medicus (includes citation from 1879) and Excerpta Medica (includes citations from 1948) can be explored

Some useful short videos for searching the literature and keeping up-to-date for any additional relevant information

[edit]

Search strategy to identify relevant information

Template for systematic literature search

  • Maintain a detailed record of searches and retrieval of studies
    • database names
    • period searched
    • specific search strategies
    • material retrieved for each search strategy from each of the databases.
    • date when search was done (day, month and year)
    • languages
    • boolean combination of terms
    • use of filters
    • communication done with the authors of the studies, if any
  • enlist/derive synonyms for the relevant terms to be used as key words
  • define literature search strategy
  • select high-yield journals for manual search
  • contact first/corresponding authors in search of additional references
  • hand search contents of the key journals relevant to the field, including those that are not enlisted in the databases searched
  • review bibliographies/reference list of the review papers
  • contact experts
  • published and unpublished reports of the studies
    • locate conference proceedings, abstracts of dissertations and other "gray" literature (the one not procurable through normal published sources). This can be achieved by contacting colleagues, first authors of the selected studies, experts in the said area of research and pharmaceutical companies as well as other informal channels can be used.
  • book searches
  • ongoing research (exploring the websites/databases of funding agencies/Trials Central/Current Controlled Trials
  • citation searching using the index terms of retrieved articles
  • check citations of the articles retrieved. Use Google Scholar for retrieving list of articles that cited the articles that we have retrieved
  • check the reference list of articles retrieved
[edit]

Screen the retrieved literature information to identify the studies meeting inclusion exclusion criteria

  • preferably by two reviewers, independently
  • note agreements and disagreements for possible calculation of Kappa value for inter-reviewer agreement/variability
  • archive the excluded studies separately
[edit]

Appraise the quality of selected studies (in terms of methodological quality-Methods as well as reporting quality-Results)

Tamplate for measuring the quality of studies (Downs and Black, 1998)

  • preferable to use a formal checklist or scale to appraise individual studies in order to avoid biased quality appraisal.The checklist includes questions like,
    • whether the individual studies are affected by significant bias (this will decide the weightage to be given to the findings of the study)
    • whether the study represent the wider population?
    • whether the study was confounded by other factors affecting the interpretation of its results
    • whether the study reports incomplete information- the authors of the study should be contacted to get additional information
  • If all the selected studies are RCTs, the minimum quality criteria should be specified and applied for appraising the quality of studies.For example what constitutes a proper RCT, true randomization(classifying the selected RCTs for their quality is an ideal scenario: Peter Juni et al, 1999)
[edit]

Prepare flow chart of how the selection of studies done- from beginning to end

For example, Flow of information (Alessandro Liberati et al, 2009)

  • No. of records identified through database search + other sources (if any)
  • No. of duplicates records excluded
  • No. of records screened
  • No. of records excluded during screening (language)
  • No. of full text articles assessed for eligibility (Inclusion-Exclusion criteria)
  • No. of full text articles excluded- providing reasons for exclusion
  • No. of studies included in qualitative synthesis (Systematic review)
  • No. of studies included in quantitative synthesis (Meta analysis)
[edit]

Prepare sheet-data extraction form/tabular format summarizing the included articles

Template for Data abstraction

  • reference
  • quality of the study
  • study design
  • study sample/population
  • setting
  • inclusion/exclusion criteria
  • number of participants
  • intervention
  • interval of intervention
  • length of follow up
  • outcome
  • method of measuring the outcome
  • point estimate
  • main findings
    • while capturing the outcomes, it is important to capture rates as well apart from the numbers (frequency)of outcomes
  • confidence interval or measure of dispersion
[edit]

Qualitatively identify homogeneous clusters of papers in terms of outcomes, intervention, and inclusion/exclusion criteria

[edit]

Combine results either through qualitative methods (SR) OR quantitative/statistical methods (MA)

  • homogeneity tests
  • select fixed vs. random effect models
[edit]

References

Retrieved from "http://www.rar.duhs.duke.edu/wiki/index.php/Systematic_review_and_meta_analysis_SOPs"