How to Perform a Systematic Search in Clinical Research

Introduction

A systematic search is the foundation of any high-quality systematic review or meta-analysis. Unlike narrative reviews that may rely on ad hoc literature scanning, systematic reviews demand reproducibility, transparency, and completeness. A well-structured search strategy ensures that all relevant studies, regardless of outcome direction, journal prestige, or accessibility, are identified and included for unbiased synthesis.

This article introduces a step-by-step framework for designing and executing a systematic search, focusing on biomedical databases such as PubMed and EMBASE. The process outlined here not only supports methodological rigor but also fulfills essential reporting standards like PRISMA.

Step 1: Formulate a Clear and Focused Review Question

Every systematic search begins with a sharply defined clinical question, ideally structured using a model like PICO (Population, Intervention, Comparator, Outcome) or DDO (Domain, Determinant, Outcome). The clarity of the question dictates the clarity of the search.

Example: Suppose the research question is:"Among patients with allergic rhinitis receiving house dust mite immunotherapy, do clinical outcomes differ between those with monosensitization and those with polysensitization?"

This question contains the core concepts to guide the search:

Domain: Patients with allergic rhinitis receiving HDM immunotherapy
Determinant: Sensitization status (mono vs. poly)
Outcome: Clinical response measures (e.g., nasal symptom scores, medication usage)

Step 2: Identify the Key Concepts from the Question

Each component of the review question becomes a key concept in the search:

House dust mite (HDM) immunotherapy
Allergic rhinitis
Sensitization type (monosensitization, polysensitization)
Clinical outcomes (e.g., nasal symptoms, medication scores)

These concepts will be translated into search terms using both keywords (free-text terms) and controlled vocabularies (database-specific indexing terms).

Step 3: Generate Keywords and Controlled Vocabularies

3.1 Keywords (Free-text Terms)

Searchers begin by brainstorming synonyms, acronyms, and spelling variations for each concept.

For HDM immunotherapy, possible keywords include:

“House dust mite”
“HDM”
“Allergen immunotherapy”
“AIT” (Allergen Immunotherapy)
“SIT,” “SLIT,” “SCIT” (forms of immunotherapy)

For sensitization, one might include:

“Monosensitiz*”
“Polysensitiz*”

The asterisk (*) allows for variations in word endings (e.g., sensitized, sensitization).

3.2 Controlled Vocabularies

Controlled vocabularies refer to structured indexing terms used in databases to organize literature.

PubMed uses MeSH (Medical Subject Headings)
EMBASE uses EMTREE terms

For example:

MeSH for HDM immunotherapy includes “Desensitization, Immunologic/therapy”
EMTREE includes terms like “house dust allergen” or “dermatophagoides pteronyssinus/exp”

Combining both keywords and controlled vocabularies enhances search sensitivity.

Step 4: Build Search Strings for Each Concept

Each concept should have a search string composed of its related keywords and vocabulary terms, connected with the Boolean operator OR. This ensures that the search captures all possible synonyms.

Example for HDM immunotherapy (PubMed syntax):

"Allergen immunotherapy"[tw] OR "AIT"[tw] OR "House dust mite"[tw] OR "HDM"[tw]
OR "Desensitization, Immunologic/therapy"[Mesh]

Each concept-specific string is then combined with others using AND to narrow the search to studies that address all aspects of the research question.

Step 5: Combine Search Strings and Execute the Search

The final step is to combine all key concept strings using AND to construct the full search query.

PubMed Syntax Example:

((“Allergen immunotherapy”[tw] OR “AIT”[tw] OR “House dust mite”[tw] OR “HDM”[tw]) AND (“Allergic rhinitis”[tw] OR “AR”[tw] OR “Rhinitis, Allergic/therapy”[Mesh]) AND (“Monosensitiz*”[tw] OR “Polysensitiz*”[tw]) AND (“nasal symptoms score*”[tw] OR “medication score*”[tw] OR “Treatment Outcome”[Mesh]))

EMBASE Conversion Notes:

Use single quotes instead of double.
Replace [tw] with :ti,ab,kw,de
Replace [Mesh] with /exp

Example (EMBASE):

('allergy immunotherapy*':ti,ab,kw,de OR 'AIT':ti,ab,kw,de) AND ('allergic rhinitis':ti,ab,kw,de OR 'AR':ti,ab,kw,de) AND ('monosensitiz*':ti,ab,kw,de OR 'polysensitiz*':ti,ab,kw,de) AND ('nasal symptoms score*':ti,ab,kw,de OR 'medication score*':ti,ab,kw,de OR 'treatment outcome'/exp)

Step 6: Search Execution and Reporting

Once the search strings are finalized:

Perform the search in each selected database (e.g., PubMed, EMBASE).
Record the number of results retrieved from each database.
Export or document the full search history.
Include this data as part of the review’s supplementary material to support transparency and reproducibility.

Advanced Search Tools:

Use PubMed’s Advanced Search Builder to test and refine combinations.
Use EMBASE’s Query Builder to ensure accurate translation of MeSH to EMTREE terms.

Conclusion

Performing a systematic search is both a science and an art. It demands precision, strategic thinking, and a deep understanding of both the clinical topic and the databases being searched. When done correctly, it ensures that all relevant evidence is captured—providing a strong foundation for a rigorous, unbiased review.

This structured approach—breaking down a review question, identifying key concepts, translating those into comprehensive search terms, and using logical operators to execute and report the search—ensures that systematic reviews are built on a robust evidence base.

Key Takeaways

A systematic search is essential for evidence completeness and bias reduction.
Each key concept from the review question must be translated into both keywords and controlled vocabularies.
Combine concepts using Boolean logic: OR within concepts, AND between concepts.
Tailor syntax for each database (e.g., [tw] vs :ti,ab,kw,de, MeSH vs EMTREE).
Document the full strategy and report search results transparently.