OBJECTIVES: Performing a systematic literature review (SLR) update can be more efficient than starting afresh when new evidence emerges. However, guidance is inconsistent when selecting the best search method for retrieving novel references. This research explored different methods of identifying references in an SLR update.

METHODS: We investigated three search update approaches to identify references in MEDLINE and Embase (via Ovid). Approach 1 involved re-running the original search string and de-duplicating in EndNote against the original SLR’s EndNote library; Approach 2 applied an entry date or revision date limit; and Approach 3 limited searches by publication year. The three approaches were tested on three SLRs across different disease areas (original searches: 2017‒2020; update: 20^th June 2022). Approaches were compared in terms of number of references retrieved and overlap with the original searches.

RESULTS: After de-duplicating against the original SLR libraries, Approach 1 yielded the highest number of novel references, followed by Approach 2, and then Approach 3. Approach 1 identified 4 to 15 references not retrieved by Approach 2, and 51 to 123 references not retrieved by Approach 3. Approach 2 identified 40 to 119 additional references compared to Approach 3. As for the overlap with the original libraries, this was considerably higher for Approach 2 (21%‒44%) than Approach 3 (4%‒13%). Many references identified in the original SLRs were revised after the original search date, hence captured by Approach 2 due to the revision date limit.

CONCLUSIONS: Re-running the original search string and de-duplicating against the original set of results retrieved the greatest number of unique references and resulted in the most complete update search. We would thus recommend this approach if the original SLR library is available. Other approaches exist when the library is not available; however, they increase the risk of missing relevant references.