I am trying to conduct a meta-analysis of incidence of a disease in an exposed population. I have the observed cases and person-years for the exposed group, but the included studies sometimes use control (non-exposed) incidence data of the national population (stratified by age and sex) as cases per 100,000 PY, while others use incidence from a control group in the study.
When entering my data, I need to enter the following:
[Cases within exposed] [PY for exposed] [Cases within unexposed] [PY for unexposed].
For the unexposed data, should I be entering the cases per 100,000 PY, or standardize the ratio to match the PY of the exposed population (e.g. 5 per 100,000 or 1 per 20,000, if the exposed population had 20,000 PY)? This decision significantly effects the weighting of each study depending on how much more or less PY the exposed patients were followed up for.
Thank you in advance for any help.
Excellent question! For studies with "control" data coming from the general population (rather than a study control arm), I would "standardize the ratio to match the PY of the exposed", keeping in mind that such "historical control" reduces the quality of the study, which you may also have to weigh/explore.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis