OBJECTIVES:

When designing a search strategy for a systematic literature review (SLR) in a niche population, striking the right balance between sensitivity and specificity whilst keeping the review manageable can prove challenging. In this case study, a targeted search was developed for a clinical efficacy SLR, restricting the population to patients with both a specific type of cancer and a specific genetic mutation and protein overexpression. Concerns were raised that relevant records may have been missed when both terms were not mentioned. To ensure identification of all publications relating to relevant trials, a supplementary search of named randomized controlled trials (RCTs) was undertaken.

METHODS:

Searches were initially conducted in Ovid Medline, Ovid Embase and the Cochrane Library. Given the tight population, no interventions were included in the search strategy and the population terms were combined with a validated sensitive Cochrane RCT filter. A supplementary search was then undertaken in Embase of named trials identified during title and abstract screening, and results were cross-referenced against the initial searches.

RESULTS:

The supplementary search identified 126 unique references that were not captured in the initial database searches. Screening against predetermined eligibility criteria identified four relevant records, of which two were unique. Publications were not picked up with the initial strategy for several reasons. These included poor reporting (e.g. no RCT terms), broader cancer population terms, spelling errors, and the use of symbols which cannot be searched for in Ovid.

CONCLUSIONS:

Including a pragmatic and unbiased supplementary search step in SLRs with narrow populations could usefully supplement the main database searches, whilst also limiting screening burden and ensuring comprehensiveness when resources are limited. To avoid relevant research being missed, it is recommended that authors identify publications as trials. Syllables and symbols should be spelled out in the first instance and abstracts should be checked for spelling errors.