A "research gap" is one of those phrases everybody uses in introductions and almost nobody defines precisely. "There is a gap in the literature" is not a justification — it is filler. To make the case that your project is worth doing, you need to identify which kind of gap exists and demonstrate it with concrete evidence. There are five gap types worth knowing. Each is found by a slightly different search and each justifies a different kind of project.
Population gap
A population gap exists when an intervention has been studied in one group but not in another clinically distinct group, where the biology or context might plausibly change the effect.
Example: Endovascular thrombectomy for large vessel occlusion stroke has dozens of trials in patients aged 18 to 80 but very few in patients over 85, even though octogenarians make up an increasing share of strokes. The gap is the elderly population. The justification is that age-related changes in collateral circulation, anticoagulation use, and frailty plausibly change the risk-benefit calculation.
How to find it: Read the inclusion criteria in the three largest trials on your topic. The patients they excluded are your candidate population gap.
Methodological gap
A methodological gap exists when the evidence on a question is based on the wrong study design, and a better-designed study has never been done.
Example: The evidence for cervical cancer screening intervals beyond 5 years is mostly modeling studies and decision analyses. There has never been a randomized trial because of the long follow-up required, but observational cohort data from large screening programs could close the gap if analyzed properly. Or: an intervention has dozens of single-center retrospective case series but no prospective cohort.
How to find it: Run your PubMed search and limit by publication type. If "[pt] Randomized Controlled Trial" gives you zero hits and "[pt] Observational Study" gives you 40, you have a candidate methodological gap.
Outcome gap
An outcome gap exists when studies have measured surrogate outcomes but never the patient-centered outcomes that actually matter.
Example: Trials of statins in primary prevention have extensive data on LDL reduction (surrogate) and cardiovascular events (intermediate), but limited data on quality-adjusted life years, deprescribing decisions in elderly patients, or shared decision-making outcomes. A trial whose primary endpoint is quality of life or treatment burden, rather than LDL, fills an outcome gap.
How to find it: Read the primary outcomes of the five most-cited trials on your topic. If they are all biomarkers and your candidate study would measure functional status or patient experience, you have a gap.
Temporal gap
A temporal gap exists when the most recent comprehensive synthesis is old enough that significant new evidence has accumulated since.
Example: A Cochrane review on PD-1 inhibitors in melanoma published in 2017 is structurally obsolete because three major trials have read out since then, and the indication has expanded to adjuvant and neoadjuvant settings. An updated review is a legitimate project even though a previous one exists.
Rule of thumb: If the most recent systematic review is more than 5 years old and at least 10 substantial new studies have been published since, a temporal-gap update is publishable. If 3 years old with 3 new studies, probably not yet.
How to find it: Note the publication year of the most recent SR/MA on your topic. Count primary studies indexed after that date.
Comparator gap
A comparator gap exists when intervention A has been compared to placebo and intervention B has been compared to placebo, but nobody has done a head-to-head A vs B trial — or a network meta-analysis to indirectly compare them.
Example: Both dapagliflozin and empagliflozin have placebo-controlled trials in heart failure. The clinical question of "which one should I prescribe?" requires either a head-to-head RCT or a network meta-analysis. The latter is feasible and publishable from existing data.
How to find it: List the interventions in your topic and the comparators used in each trial. If every cell on the diagonal is filled but no off-diagonal cells exist, you have a comparator gap.
How to verify a candidate gap is real
Spotting a candidate gap is the easy part. The hard part is confirming it has not already been addressed in a paper your initial search missed. Three quick checks before you build a project around a gap:
- Cross-database search. A "gap" that shows up in PubMed might be filled in Embase or a regional database. Run the same search in at least one other database.
- Recent preprints. Search bioRxiv and medRxiv via Europe PMC. A preprint addressing your gap means someone is six months ahead of you.
- PROSPERO and ClinicalTrials.gov. An ongoing review or trial closing the gap soon is functionally a published filling. See our PROSPERO walkthrough.
How to write up a gap in your introduction
Bad: "There is a gap in the literature regarding X."
Better: "The two most recent systematic reviews of X (Smith 2022, Lee 2023) included only patients aged 18 to 70 and used 6-month mortality as the primary outcome. Both reviews flagged the absence of evidence in patients over 75 and the lack of patient-reported outcome data as priorities for future research. Our review addresses both gaps by including all-age cohort studies and extracting the EQ-5D outcomes reported in 14 of the 38 candidate studies."
The second version names the specific reviews, cites the specific gap they identified, and shows exactly how the proposed project closes it. That is what reviewers want to see.
If you use ResearchChecker, the Gap Analysis section automatically flags population, methodological, outcome, temporal, and comparator gaps based on the live evidence landscape. The output is a starting point you verify against the actual papers — the same way you would do manually, just faster.