Effect Measures

What information do I need to perform a meta-analysis?

A meta-analysis uses the quantitative results from the individual studies, pools them together using a statistical method, and produces a new single quantitative result. This new quantitative result is a known as the ‘pooled result’ and is presented as a weighted average treatment result and a measure of variation.

The quantitative results from the individual studies and the pooled weighted average treatment result calculated from a meta-analysis are commonly called ‘effect measures’. There is a range of effect measures used within research, and the type of effect measure one uses depends of the type of outcome measure of interest. For example, if our outcome of interest was death, there are only two options available, where the participant is either dead or alive. This outcome measure is known as a binary outcome. For an outcome measure like this, the commonly used effect measure would be the Odds Ratio. The Odds Ratio let us describe the likely harm that a treatment or intervention might cause.

The most commonly used effect measure in meta-analysis is the Odds Ratio. Other common effect measures used within research, and in particular in meta-analyses, are the Risk Ratio, Risk Difference, Mean Difference and Standardised Mean Difference. The Risk Ratio and Risk Difference are used for binary outcomes where there are only two options available. Where as the Mean Difference and Standardised Mean Difference are used for outcomes which are continuous. Continuous outcomes would include measures such as blood pressure or weight, where there are a defined range of possible values.

In addition to the meta-analysis producing a pooled treatment result, a measure of variation is also produced. This measure of variation lets us describe how precise the pooled treatment result is. Common measures of variation include the standard deviation, variance and standard error. These measures of variation are commonly used to calculate 95% confidence intervals for the pooled treatment result.