These are the rules that decide which studies make it into your review and which get thrown out. They sound simple, and getting them slightly wrong is one of the easiest ways to compromise a review without realizing it. Criteria that are too vague let bias creep in, because two reviewers reading the same study reach different verdicts. Criteria set after you have seen the results let you tune the review toward the answer you wanted. The whole credibility of your included set rests on getting these rules right and fixing them early.
Inclusion criteria define the characteristics a study must have to be eligible: the right population, intervention, outcomes, study design, and so on. Exclusion criteria define characteristics that disqualify a study even if it otherwise looks relevant, such as the wrong setting or a design you cannot use.
Together they define the exact boundary of your review. They should be specific enough that any two reviewers applying them to the same study reach the same decision, and they must be set before you run your search, ideally in your protocol.
Three reasons, and they compound.
First, reproducibility. A systematic review is meant to be repeatable. If your criteria are precise, someone else can apply them to the same studies and reach the same included set. If they are loose, they cannot, and the review is no longer systematic in any meaningful sense.
Second, bias prevention. Explicit criteria stop you from selecting studies based on their results. Without them, it is easy to unconsciously include the trials that support your expected conclusion and exclude the ones that do not. This rarely feels deliberate, which is exactly why a fixed rule is needed.
Third, scope control. Clear criteria keep the review focused on the question you actually asked. Fuzzy criteria let the scope drift, so you end up screening studies that were never really relevant and defending inclusion decisions you cannot justify.
The cleanest way to derive criteria is straight from your PICO question. Each element becomes an eligibility rule.
Population. Who the studies must be about: age range, condition, setting, and any defining characteristics. Include: adults aged 18 and over with type 2 diabetes. Exclude: studies in children or in type 1 diabetes.
Intervention. The intervention of interest, specified precisely enough to be recognizable. Include: structured aerobic exercise programs. Exclude: mixed programs where exercise cannot be isolated.
Comparison. What the intervention is compared against, if your question specifies one. Include: usual care or waitlist controls. Exclude: studies with no comparator.
Outcome. The outcomes a study must report to be useful to you. Include: studies reporting a validated depression symptom scale. Exclude: studies reporting only qualitative impressions.
Beyond PICO, a few more dimensions usually need rules:
Study design. Which designs qualify, for example randomized controlled trials only, or observational studies too. This is one of the most important criteria and often the most decisive.
Timeframe. Any date limits on publication, and the justification for them. Date limits should have a reason, not just convenience.
Language. Whether you restrict to certain languages. Restricting to English is common but introduces a known bias, so it needs acknowledging rather than assuming.
Publication type. Whether you include only peer-reviewed articles, or also conference abstracts, theses, and grey literature.
A common source of confusion is the overlap between the two. If your inclusion criterion is "adults aged 18 and over," you do not also need an exclusion criterion of "studies in people under 18," because that is already implied. Reserve exclusion criteria for disqualifiers that are not simply the inverse of an inclusion rule.
Good exclusion criteria capture things like: studies where the outcome data cannot be extracted, duplicate publications of the same dataset, or designs that meet the population and intervention rules but are unsuitable for other reasons. If every exclusion criterion is just an inclusion criterion turned around, you are adding clutter without adding clarity.
| Category | Example inclusion | Example exclusion |
|---|---|---|
| Population | Adults with type 2 diabetes | Children; type 1 diabetes |
| Intervention | Structured aerobic exercise | Interventions where exercise is not isolated |
| Comparison | Usual care or waitlist | No comparator group |
| Outcome | Validated symptom scale reported | Outcome not measured |
| Study design | Randomized controlled trials | Case reports; editorials |
| Timeframe | Published from 2000 onward | Earlier studies (with stated reason) |
| Language | English-language reports | Languages the team cannot appraise |
| Publication | Peer-reviewed articles | Conference abstracts only |
Even carefully written criteria have ambiguities you will not spot until you apply them. Before full screening, pilot your criteria on a sample of studies, ideally with both reviewers screening the same set independently. Where they disagree, the criteria are usually the problem, not the reviewers. Refine the wording until two people applying the rules reach the same decisions reliably. Doing this at the pilot stage is cheap. Discovering the ambiguity halfway through screening thousands of records is not.
Criteria that are too broad. Vague rules like "relevant studies on the topic" cannot be applied consistently and let the scope sprawl. Every criterion should be something a reviewer can check objectively.
Criteria that are too narrow. Over-restricting can leave you with almost no studies, or bias the sample toward a particular type of research. Narrow with a reason, not reflexively.
Setting criteria after searching. Deciding what counts after you have seen the results is the cardinal sin. It lets the findings shape the criteria, which is backwards. Fix them in the protocol first.
Redundant exclusion criteria. Listing exclusions that merely restate inclusions in reverse adds noise. Keep exclusions for genuine disqualifiers.
Unjustified restrictions. Date and language limits are fine when they have a rationale, but arbitrary ones introduce bias you will have to answer for. State why each restriction exists.
When should I define inclusion and exclusion criteria? Before you search, in your protocol. Setting them in advance is what stops your results from influencing which studies you count as eligible.
What is the difference between inclusion and exclusion criteria? Inclusion criteria are the characteristics a study must have to be eligible. Exclusion criteria are characteristics that disqualify a study. Exclusions should capture genuine disqualifiers, not just the reverse of inclusion rules.
How specific should the criteria be? Specific enough that two reviewers applying them to the same study reach the same decision. If they cannot, the criteria are too vague.
Can I change my criteria during the review? Only with a documented, justified reason, disclosed in the final report. Undisclosed changes, especially ones made after seeing results, undermine the review.
Should I restrict by language? You can, but it introduces a known bias. If you limit to certain languages, acknowledge the limitation rather than treating it as neutral.
Inclusion and exclusion criteria are the boundary of your review, and their job is to be precise, reproducible, and fixed before you look at the results. Build them from your PICO question, add clear rules for study design, timeframe, language, and publication type, keep exclusions to genuine disqualifiers, and pilot them so two reviewers agree.
Set them early and hold to them, disclosing any justified change. That single discipline does more than almost anything else to keep your included set defensible. Your criteria belong in the protocol, which we cover in how to write a systematic review protocol.
Applying your criteria across thousands of records? Verflux supports independent dual screening with conflict resolution, so your eligibility decisions stay consistent and documented.
All analytical features included in the free trial. No credit card, no installation, no R or Python.
Create free account