03 · What You Need to Know
What a Systematic Review Can Tell You and What It Can Hide
A Systematic Review Is Usually a Starting Point, Not an Obstacle to Avoid
Systematic reviews exist partly because reading primary studies individually is an inefficient way to understand a large evidence base. A well-conducted review identifies eligible research using explicit methods, organizes it around a defined question, assesses important limitations, and synthesizes findings across studies.
Cochrane's guidance on overviews of reviews illustrates the same efficiency principle one level higher: existing syntheses can allow researchers and decision makers to address broad questions without reconstructing every primary-study analysis from the beginning. At the same time, Cochrane notes that reviewing primary studies can provide more information because data can be extracted directly and standardized for the question being asked.
So the default should not be "never trust a review." Nor should it be "never read the originals." The useful question is whether the review contains enough trustworthy information for your purpose.
Use the review level
When you need a defensible synthesis of the overall evidence and the review is current, relevant, methodologically trustworthy, and sufficiently detailed.
Return to the study level
When your interpretation depends on details, evidence, or methodological judgments that cannot be adequately evaluated from the review alone.
Go Back When a Crucial Methodological Detail Is Missing
A systematic review necessarily compresses information. A primary study containing thousands of words of methods, results, tables, appendices, and supplementary material may receive only a few rows in a characteristics table and a brief risk-of-bias assessment.
That compression is usually necessary. It becomes a problem when the omitted detail matters to your question.
For example, you may need to know exactly how an intervention was implemented, how an outcome was operationalized, how missing data were handled, which covariates were included in an analysis, whether groups differed at baseline, or how a particular subgroup was defined. If the review does not provide enough information, inspect the primary report rather than guessing what the original investigators did.
Go Back When a Risk-of-Bias Judgment Seems Consequential or Unclear
Review authors frequently make structured judgments about limitations in individual studies. Those judgments can influence sensitivity analyses, certainty assessments, and ultimately the review's conclusions.
If a pivotal study is classified as having low, high, or serious risk of bias and that classification materially affects the synthesis, you may need to inspect the original study to understand the basis for the judgment.
This does not mean casually overruling review authors whenever you dislike an appraisal. It means verifying consequential judgments when the review's explanation is insufficient for your purpose.
Go Back When the Review and the Primary Study Seem to Say Different Things
Sometimes a review's summary of a study appears inconsistent with the study itself. Before assuming an error, check whether the review extracted a particular outcome, follow-up period, subgroup, or analysis rather than the study's headline result.
A single study can report multiple outcomes and analyses, while only some may satisfy the review's eligibility criteria. The review and study may therefore both be accurate while answering slightly different questions.
If the discrepancy remains after checking those details, the primary report becomes necessary for determining what was actually measured and reported.
Go Back When You Need to Understand Why Studies Disagree
A forest plot can show that effect estimates vary, but it may not explain why. Study-level details can reveal differences in participants, implementation, exposure definitions, outcome measurement, follow-up, adherence, attrition, analytical methods, or context that are compressed in the synthesis.
This becomes particularly useful when substantial heterogeneity makes a pooled estimate difficult to interpret. The more the practical meaning of the synthesis depends on why individual effects differ, the stronger the reason to inspect the studies driving that variation.
Go Back When a Few Studies Drive the Pooled Result
Studies rarely contribute equally to a meta-analysis. A small number of large or statistically precise studies may account for much of the weight. Alternatively, an unusual study with an extreme effect may materially influence the pooled estimate or heterogeneity.
If your conclusion depends heavily on those studies, they deserve closer inspection. Examine whether their populations, interventions, outcomes, and methods genuinely support the interpretation being placed on the pooled result.
This is particularly important because a statistically precise synthesis can still depend on methodologically weak evidence.
Go Back When the Review Is Outdated
A review represents the evidence available through its search date, not indefinitely. Important primary studies may appear afterward.
Check the final search date rather than the publication date. If substantial new evidence has accumulated, the review can still provide a valuable map of the earlier literature, but its conclusion may no longer represent the complete evidence base.
You do not necessarily need to abandon the review. A practical approach is to use it for the evidence covered by its search and examine important eligible studies published afterward. If enough new evidence has accumulated that the old synthesis is no longer adequate, a newer review or formal update may be preferable.
Go Back When Systematic Reviews Disagree
Conflicting reviews may differ because of their questions, eligibility criteria, searches, study groupings, analytical methods, or interpretation. Often you can identify the source of disagreement by comparing the reviews directly.
Sometimes that comparison leads back to a small set of primary studies. Perhaps one review included studies another excluded, two reviews extracted different results from the same trial, or a heavily weighted study was handled differently.
When systematic reviews reach different conclusions, examining the studies responsible for the disagreement can be more informative than reading yet another review abstract.
Go Back When Reviews Overlap Heavily
Several systematic reviews do not necessarily represent several independent bodies of evidence. The same primary studies may appear repeatedly across reviews.
Cochrane recommends mapping primary-study overlap when working with multiple reviews because double-counting the same underlying data can give particular studies excessive influence and, when re-analyzed, produce overly precise estimates.
If you need to determine whether several reviews genuinely provide independent corroboration, inspect their study lists and, where necessary, the primary-study reports themselves.
Go Back When the Review's Scope Does Not Quite Match Yours
A review may include adults and children when your question concerns only children, several intervention variants when you care about one, or multiple outcomes when only one is relevant to your decision.
Cochrane's overview guidance explicitly recognizes situations in which only a subset of primary studies contained within an included review matches a narrower question. In such cases, primary studies may need to be assessed against the narrower eligibility criteria.
This principle extends beyond formal overviews. If the review is broader than your question, inspect whether the subset relevant to you actually supports the general conclusion.
Go Back When You Need Data the Review Did Not Extract
Your research question may require information the review authors did not need. Perhaps you need intervention dosage, implementation characteristics, a secondary outcome, subgroup data, measurement instruments, contextual variables, or a particular follow-up period.
Cochrane notes that referring back to underlying primary studies can increase comprehensiveness and rigor when needed information is missing or inadequately reported in systematic reviews, although doing so requires additional time and resources.
At that point, the primary studies are not being used to challenge the review. They are simply the correct source for information the review was never designed to preserve.
Do Not Go Back Merely to Cherry-Pick an Appealing Study
There is an important danger on the other side. Once you open individual studies, a dramatic finding, prestigious journal, enormous sample, or result matching your expectations can become psychologically persuasive.
If a rigorous review shows a mixed evidence base, finding one compelling primary study does not automatically invalidate the synthesis. The study should be interpreted within the broader evidence, including its own risk of bias and the reasons its result may differ.
Watch Out
Returning to primary studies should increase resolution, not provide an excuse to replace a systematic body of evidence with whichever individual paper produces the conclusion you prefer.
The Authors' Conclusion Is Not a Substitute for the Underlying Results
Even at the systematic-review level, conclusions involve interpretation. Cochrane advises review authors to communicate uncertainty, applicability, benefits and harms carefully and cautions against turning review findings directly into context-free recommendations.
If the practical claim you want to make is stronger or more specific than the review's evidence summary, inspect the underlying evidence before extending the conclusion. Sometimes the primary studies reveal that an apparently simple average depends on populations, conditions, or outcomes quite different from the ones you intend to discuss.
