Systematic Review

Systematic reviews reduce bias at several choke points: search bias by prespecifying databases and terms, selection bias by applying eligibility criteria consistently, extraction errors by using structured forms and often duplicate review, and interpretation bias by formally judging risk of bias and certainty of evidence. PRISMA is mainly a reporting framework, while tools such as RoB 2, ROBINS-I, or AMSTAR 2 address appraisal of study or review quality more directly.

Why the word “review” hides the real job

In journals, review can mean anything from “an expert talks through a topic” to a tightly scripted evidence project. That is why readers get fooled. Two papers can both be called review articles, while only one actually tells you where the authors searched, which studies were excluded, and why.

The surprise is this: a systematic review is not mainly about summarizing. It is about search discipline before interpretation. Think of it like dragging a magnet across a beach in straight, marked lanes. If you wander around and pick up whatever glints, you may still find coins, but you cannot claim you searched the beach. A systematic review lays out the lanes first, then reports what was found, what was missed, and what got thrown out as junk.

What “systematic” actually means

A systematic review starts with a focused question, then usually sets a protocol in advance: databases to search, keywords, inclusion rules, exclusion rules, outcomes of interest, and how study quality will be judged. Researchers then run the search, remove duplicates, screen titles and abstracts, read full papers, extract the data, assess risk of bias, plainly, the ways a study could tilt the result, and then synthesize the findings.

That synthesis may be narrative or statistical. If the studies are similar enough, the reviewers may do a meta-analysis, which is the number-pooling part. So in the “systematic review vs meta-analysis” debate, the clean answer is: a meta-analysis is sometimes inside a systematic review, not a synonym for it.

This also explains “systematic review vs literature review.” A literature review can be useful, but it may not have a registered protocol, a reproducible search, or a formal bias assessment. A systematic review is supposed to leave footprints another team could follow.

Why the step counts keep changing

You will see people ask for the 5 steps of a systematic review or the 7 steps of a systematic review. Both are simplifications. Different textbooks combine or split steps differently, but the backbone is stable: ask a focused question, plan the protocol, search broadly, screen studies, extract and appraise, synthesize, and report transparently. The number changes; the logic does not.

One decision this helps you make today

If you are reading a supplement claim, do not stop at the phrase “systematic review.” Prefer the review that shows a protocol or registration, reports a PRISMA flow diagram, explains risk-of-bias methods, and makes clear whether it searched only PubMed or multiple databases. That one decision will save you from treating a polished opinion piece as if it were a map of the whole evidence field.

Specific failure mode: do not treat “searched PubMed” as enough. For fast-moving supplement topics, a review that skips Embase, CENTRAL, trial registries, or reference-list searching may miss negative or unpublished studies and leave you with an overly rosy picture.