After defining a clear research question and eligibility criteria, the next critical step in a meta-analysis is developing a comprehensive, transparent, and reproducible search strategy. The quality of the search directly determines the validity of the final results—an incomplete search leads to biased conclusions.

Below is a step-by-step guide to building and executing a robust search strategy.

1. Use Multiple Databases (Not Just One)

A comprehensive search typically involves at least three databases, with search strategies tailored to each database. Commonly used databases include:

• MEDLINE

• Embase

• CENTRAL (Cochrane Central Register of Controlled Trials)

Depending on the topic, additional specialized databases may be appropriate (e.g., PsycINFO, CINAHL, Scopus).

Platform vs Database (Important Distinction)

It is essential to understand the difference between a platform and a database:

• Platform: The interface used to access databases (e.g., PubMed, Ovid, Web of Science)

• Database: Where the indexed literature is stored (e.g., MEDLINE, Embase)

Each platform has unique search syntax, filters, and indexing behavior, meaning search strategies must be adapted for each platform, even when searching the same database.

Although optional, collaboration with a professional medical or academic librarian is strongly encouraged, as they can significantly improve search sensitivity and reproducibility.

2. Identify Core Concepts Using PICO

Search strategies are built around the key concepts of the research question. In most cases, the focus is on:

• P (Population)

• I (Intervention or Exposure)

Occasionally, O (Outcome) or study design may be added if specified in the eligibility criteria.

For each key concept:

1. Identify the main term

2. Compile a list of synonyms, related terms, acronyms, and alternate spellings

3. Consider how authors would describe the concept in titles and abstracts

3. Use Boolean Operators Strategically

Boolean operators form the backbone of the search strategy:

• OR → combines similar terms within the same concept

• AND → combines different concepts to narrow the search

Example logic:

• Population terms combined with OR

• Intervention terms combined with OR

• Population set combined with Intervention set using AND

This approach maximizes sensitivity while maintaining relevance.

4. Combine Subject Headings and Keywords

Effective searches use both subject headings and keywords.

Subject Headings

Subject headings are controlled vocabulary terms assigned by indexers:

• MeSH terms in MEDLINE

• Emtree terms in Embase

To use subject headings:

1. Enter the key concept into the database

2. Review suggested subject headings

3. Select the most appropriate term

4. Decide whether to explode (include narrower related terms) or focus the heading

Multiple subject headings may be needed for a single concept.

Keywords

Keywords are uncontrolled terms that appear in titles, abstracts, and other fields.

When selecting keywords, consider:

• Synonyms

• Acronyms

• Alternate spellings

• Truncation (e.g., dislocat* → dislocate, dislocation, dislocations)

In many databases, adding .mp. allows the keyword to be searched across multiple fields.

Unlike subject headings, keywords are not standardized, so all relevant variants must be included.

5. Build and Test the Search Iteratively

Search development is an iterative process:

1. Start broad

2. Run the search

3. Review the first ~30 results for relevance

4. Adjust terms, headings, or operators as needed

If results are too broad:

• Focus subject headings instead of exploding them

• Add additional concepts or limits

If results are too narrow:

• Add synonyms

• Remove unnecessary limits

Expected hit counts depend on topic scope:

• Broad topics: often >2000 results

• Narrow topics: substantially fewer

6. Finalize and Run Searches

Once finalized:

• Run all database searches on the same day

• Export results from each database

• Import them into reference management or screening software (e.g., Covidence, Rayyan)

This ensures consistency and accurate reporting.

7. Screening Studies

After importing results, the screening process begins.

Duplicate Removal

Duplicates are removed prior to screening.

Title and Abstract Screening

• Conducted in duplicate by two independent reviewers

• A pilot screening of a small subset is recommended to align interpretations of eligibility criteria

• Any uncertainties or conflicts should move forward to full-text screening

Full-Text Screening

• Specific reasons for exclusion must be documented for each article

• Disagreements are resolved by discussion or a third reviewer

• Inter-rater reliability should be calculated at both stages, typically using Cohen’s kappa (κ)

Additional Searches

• Manually screen reference lists of included studies

• Review references of similar systematic reviews to identify missed articles

8. Reporting the Search and Selection Process

The PRISMA flow diagram should be used to document and report:

• Databases searched and hits per database

• Total records screened

• Records excluded at each stage

• Reasons for full-text exclusions

• Articles identified through manual searches

• Final number of included studies

Transparent reporting is essential for reproducibility and credibility.

Final Takeaway

A rigorous search strategy is systematic, transparent, database-specific, and reproducible. Investing time in careful planning—and refining the search iteratively—pays off by minimizing bias and strengthening the validity of the meta-analysis.

https://axeusce.org/courses/systematic-review-meta-analysis-training/