Insights / Topic · Umbrella reviews
Reviewing existing systematic reviews and meta-analyses: eligibility at review level, handling overlapping studies, AMSTAR 2, and grading credibility.
When a field already contains several systematic reviews on related questions, the useful contribution is often a review of those reviews. An umbrella review, also called an overview of reviews, collects the existing syntheses, appraises them, and summarises what the body of evidence shows across outcomes or interventions.
The unit of inclusion is the review rather than the primary study, which changes every step. Eligibility criteria are set at the review level, including the minimum methodological standard a review must meet. Methodological quality is appraised with AMSTAR 2, which identifies critical weaknesses rather than producing a score. Data are extracted from the reviews, with a decision about whether to re-analyse primary data when reviews report it inconsistently.
The central problem is overlap. The same primary studies often appear in several reviews, and counting them each time inflates the apparent weight of the evidence. Overlap is mapped in a citation matrix and quantified with the corrected covered area, and a strategy is chosen in advance, such as selecting the most recent, most comprehensive or highest-quality review for each question. Many umbrella reviews of observational evidence then grade credibility with explicit criteria, including the number of cases, the significance threshold, heterogeneity, prediction intervals, small-study effects and excess significance.
Reviewers of umbrella reviews focus on a few weak points. Whether overlap was measured and handled or ignored, whether AMSTAR 2 ratings were applied consistently and by two people, whether the most recent evidence has been missed because the latest included review is several years old, and whether the credibility grading rests on criteria stated in advance. Addressing each of these explicitly is what separates an umbrella review from an annotated list of reviews.
Reporting follows PRIOR, the guideline for overviews of reviews. The notes in this topic cover framing an umbrella review question, building an overlap matrix, applying AMSTAR 2 consistently, choosing between narrative summary and re-analysis, and presenting credibility classes without overstating them.
Explainer
Reviews of the same question routinely share almost no trials. Calculating overlap first tells you whether a new review is needed or an overview would say more
