Narrative reviews versus systematic reviews

The Cochrane Handbook connects the search strategy with previously defined eligibility criteria. To assess a review, look for databases searched, dates covered, search terms and restrictions. These details reveal which evidence could enter and which could be missed.

The phrase ‘the literature was reviewed’ is insufficient. Listing databases without explaining their use is insufficient too. A question about a cell model and one about animal outcomes require searches and criteria addressing those differences.

Cochrane recommends documenting inclusion and exclusion decisions and using independent assessors to determine final eligibility. The record allows readers to check why an apparently relevant study is absent from the synthesis.

As an editorial check, choose a known reference and follow its path. If excluded, there should be a reason linked to the criteria. This does not certify the entire review, but can reveal selection that is difficult to reconstruct.

SANRA was developed to assess narrative reviews through objectives, topic importance, searching, references, evidence level and outcome presentation. Its purpose was evaluating article quality, not converting an article into a systematic review.

A narrative review can organize concepts, relate mechanisms and explain the history of a hypothesis. Its usefulness depends on readers distinguishing interpretative synthesis from an exhaustive search. A coherent narrative does not demonstrate inclusion of every conflicting result.

Cochrane's chapter on other synthesis methods describes options for circumstances in which results are not pooled statistically. Absence of a pooled estimate still requires explanation of grouping, comparisons and data presentation.

In a peptide review, for example, separating cell cultures, animals and formulations may be more informative than combining them into one number. This illustrates a synthesis decision; its justification should appear in the article.

Cochrane warns against classifying studies solely by statistical significance and counting the positive ones. Two studies may estimate similar effects yet fall on opposite sides of a threshold because their precision differs.

Before accepting ‘most studies agree,’ check what was counted as agreement. Direction, magnitude and uncertainty provide different information. A useful review shows these differences and explains how it moved from individual studies to its conclusion.