Steps | Description | Example |
---|---|---|
Identify comparisons | Use 2 at a time (pairwise) comparison Separate populations that can be studied within the comparison | Jogging vs running, jogging vs gym work, jogging vs dancing, for weight reduction Compare above in older people |
Identify outcomes and effect measures | Select outcomes for each comparison (as per protocol) Then effect measure to report results | Weight reduction Increase in muscle mass (continuous outcome) MD or SMD |
Collect data | Data collation from each selected study | |
Combine results | Studies with more precise estimate should be give more weight | Variance is used to estimate the weight of the study |
Choose statistical methods | Straightforward method | Inverse variance is used |
Mantel–Haenszel method (dichotomous data) | Suitable for small studies with low event rate | |
Peto method | An additional option for odds close to one | |
Assumptions about heterogeneity | Decision between fixed or random effect analysis | |
Present results | Forest plot | Displaying results in a graph with overall effect estimate at the bottom |