Can you do a meta-analysis with only 2 studies?

Technically yes, a meta-analysis can be performed with as few as 2 studies. However, a 2-study meta-analysis provides very limited information: you cannot assess heterogeneity reliably, cannot test for publication bias, and the pooled estimate is essentially a weighted average of just two data points. Most methodologists recommend reporting the individual study results alongside the pooled estimate and interpreting the result with substantial caution.

What is the minimum number of studies for a funnel plot?

Funnel plots require a minimum of 10 studies to be visually interpretable, and statistical tests for funnel plot asymmetry like Egger's test have very low power with fewer than 10 studies. With fewer than 10 studies, asymmetry in a funnel plot is essentially uninterpretable, and formal publication bias testing should not be conducted.

How many studies do you need for subgroup analysis?

Each subgroup should contain at least 2 studies for a pooled estimate, but 5 to 10 studies per subgroup are recommended for meaningful between-group comparisons. With fewer than 5 studies per subgroup, tests for subgroup differences have very low statistical power and findings should be considered exploratory rather than confirmatory.

Is a meta-analysis with 3 studies worth doing?

A meta-analysis with 3 studies can be worthwhile if the studies are methodologically sound, reasonably homogeneous, and address a clinically important question. The pooled estimate provides a more precise answer than any individual study alone. However, heterogeneity assessment will be limited, and you should not attempt publication bias testing. Present individual study results alongside the pooled estimate for transparency.

What should I do if I have too few studies for meta-analysis?

When you have too few studies for meaningful meta-analysis, conduct a narrative synthesis instead. Present individual study results in a structured table, describe the direction and magnitude of effects across studies, discuss consistency or inconsistency of findings, and provide a qualitative assessment of the overall evidence. Narrative synthesis is a valid approach recognized by PRISMA and Cochrane for situations where meta-analysis is not appropriate.

How Many Studies for a Meta-Analysis? Minimum Explained

You need a minimum of 2 studies to conduct a meta-analysis, as this is the smallest number that allows statistical pooling of effect estimates. However, 2 studies is a theoretical minimum, not a practical recommendation. Most methodologists and the Cochrane Handbook recommend having at least 5 studies for basic meta-analysis and 10 or more studies for reliable assessment of heterogeneity, publication bias, and subgroup differences.

The question of how many studies you need depends on what you want to do with the results. A simple pooled effect estimate can be calculated from 2 studies. But if you want to assess whether the effect varies across populations (subgroup analysis), investigate sources of variation (meta-regression), test for funnel plot bias detection methods, or have confidence in the stability of your estimate (sensitivity analysis), you need substantially more studies.

The Theoretical Minimum: 2 Studies

A meta-analysis pools effect sizes from individual studies to produce a combined estimate with a narrower confidence interval than any single study. Mathematically, this pooling requires at least 2 data points. With 2 studies, you can calculate a weighted average effect size, a 95% confidence interval, and a basic Q-statistic for heterogeneity.

However, a 2-study meta-analysis has severe limitations:

Heterogeneity is essentially unassessable. The Q-test has extremely low statistical power with 2 studies, and I-squared is unreliable with fewer than 5 studies
The pooled estimate is fragile. If one study has a methodological flaw, it directly drives 50% of the result. There is no stability from additional studies to buffer against individual study weaknesses
No publication bias assessment. learn about funnel plots and statistical tests for asymmetry cannot be used with 2 studies
No subgroup analysis possible. You cannot investigate whether the effect varies by population, setting, or intervention characteristics

Despite these limitations, a 2-study meta-analysis is preferable to no synthesis when only 2 relevant studies exist. The Cochrane Handbook states that meta-analysis of 2 studies "may be valuable if both are large, rigorous, and clinically similar." Present individual study results alongside the pooled estimate and clearly communicate the limitations.

Practical Minimums by Analysis Type

Practical minimum studies needed for different meta-analysis types — Studies needed by analysis type

Analysis	Minimum Studies	Recommended	Rationale
Basic pooled estimate	2	5+	Stability of the combined effect
I-squared heterogeneity	3	10+	I-squared has low precision with few studies
Q-test for heterogeneity	3	10+	Very low statistical power below 10 studies
Subgroup analysis	2 per subgroup	5-10 per subgroup	Test for subgroup differences needs power
Meta-regression	10	20+	Rule of thumb: 10 studies per covariate
Funnel plot (visual)	5	10+

Why 10 Studies Is the Common Benchmark

Three statistical reasons why ten studies is the common meta-analysis benchmark — Why 10 studies is the common benchmark

The number 10 appears frequently in meta-analysis methodology guidelines as a minimum threshold for several reasons:

Heterogeneity assessment. The I-squared statistic measures the percentage of variability across studies that is due to true differences rather than chance. With fewer than 10 studies, the confidence interval around I-squared is so wide that the estimate is essentially uninformative. A meta-analysis of 4 studies might report I-squared of 50%, but the 95% confidence interval could range from 0% to 90%, making the value meaningless for decision-making. See our heterogeneity guide for interpretation details.

Publication bias detection. explore funnel plots require at least 10 studies to show visually interpretable patterns of asymmetry. Egger's regression test has very low statistical power below 10 studies, meaning it frequently fails to detect real publication bias when it exists. The Cochrane Handbook recommends against using these methods with fewer than 10 studies.

Model selection. Choosing between random effects and fixed effects models becomes more consequential with fewer studies. Random effects models, which are usually preferred because they account for between-study variation, produce wider confidence intervals that may encompass clinical irrelevance when the number of studies is small. With very few studies, the between-study variance estimate (tau-squared) is imprecise, affecting the validity of the random effects model.

How Many Studies Do You Need for a Meta-Analysis? Minimum Requirements Explained

Key Takeaways

The Theoretical Minimum: 2 Studies

Practical Minimums by Analysis Type

Why 10 Studies Is the Common Benchmark

Frequently Asked Questions

Related Articles

Reading About Meta-Analysis? Our PhD Team Runs Them Every Day.

Dr. Sarah Mitchell

Reading About Meta-Analysis? Our PhD Team Runs Them Every Day.

Quality Matters as Much as Quantity

When to Use Narrative Synthesis Instead

What the Cochrane Handbook Recommends

Related Articles