Background: Sample size at least required in a randomized trial to detect a realistic intervention effect is [formula can not be rendered]. Meta-analyses should also evaluate the evidence according to the number of participants they analyse. Meta-analyses should not analyse included participants as coming from one trial. A realistic intervention effect could be applied in sample size calculation if the degree of evidence is wanted. Meta-analyses ought to consider heterogeneity when sample size is calculated to evaluate evidence when a random effects model is used.

Objectives: To illustrate the calculation of a heterogeneity sample size correction factor to preserve power and level of significance from a fixed into random effects models.

Methods: Please contact authors for information about methods used.

Results: Sample size calculation can be corrected to preserve power and level of significance in a random effects model. The general expression of the correction factor is provided by the product of ratios of squared intervention effects and sums of precision's in the models. The factor simplifies to the inverse homogeneity in case of equal weights and intervention effects.

Conclusion: Sample size should be inflated in the same degree as homogeneity is decreased to detect the same intervention effect irrespective of the selected level of significance and power.