Effect Methods

Choosing between Fixed and Random effects methods

There are two main types of methods that can be used when performing a meta-analysis. These are called the ‘fixed effect method’ and the ‘random effect method’. The choice over which method is more appropriate depends primarily on the level of heterogeneity between the individual studies. If low levels of heterogeneity are detected, indicated by an I2 value less than 40%, then it is appropriate to use a fixed effect method to pool the data. If there are moderate levels of heterogeneity between the study results, indicated by an I2 value between 40% and 84%, then it would be more appropriate to use a random effect method. A random effect method has the advantage that it allows for the heterogeneity between the studies to be taken into account. This is shown by the width of the confidence intervals for the pooled result getting wider than would be seen if a fixed effect method was considered. However, if an I2 value of 85% or more is seen in a meta-analysis, this would indicate that high levels of heterogeneity are present in the meta-analysis. In this situation even the random effect method can not account for the heterogeneity and so the meta-analysis should be abandoned since the individual results are too different to combine together. In our example, the I2 value was 45%. Therefore it would be more appropriate to use a random effect method. If we change our method from a fixed effect to a random effect we can see that 95% confidence intervals have got slightly wider.