Im working on a meta analysis and in analysis i have multiple adjusted odds ratio for a drug's efficacy in a disease. the odds ratios are drugs efficacy in 6 months and another is of drug's efficacy in 2 years so can I add both of them up?
They are adjusted odds ratios so I don't have the number of patients getting the disease.
Hello Abdul,
If multiple data sets all had equal sample sizes for the focal and reference groups (and equal n across data sets), then adding odds ratios and dividing by the number of ratios summed would give you a collective/aggregate OR estimate (assuming that all data sets were using the same IV and DV, of course).
The average of a set of OR values (when data sets do not have identical sample sizes between groups and across sets) could be considered an unweighted mean estimate of the OR. However it is not likely to be a good estimator of the population OR.
The ideal solution is if you had the actual data points for all of the cases (on all of the variables, especially if adjusted OR values were the goal). Barring that, if you have many such OR estimates, you can describe the de facto distribution (such as by a simplistic device like Tukey's five-number summary: minimum, lower hinge, median, upper hinge, maximum).
Good luck with your work.
Hi Abd,
I think you can use weight mean of OR
Give a one uint of OR of six months and 4 uints of 2 years, maybe does it make sense.
My regards
Hello Abdul Moeed. Do you mean that the individual studies (or some of them, at least) have a repeated measures design such that the treatment effect is measured at two time points? If so, there are methods that can account for the correlated nature of the data. Here is a recent article by Trikalinos & Ingram Olkin that looks relevant:
Article Meta-analysis of effect sizes reported at multiple time poin...
And FWIW, Ingram Olkin has long been recognized as an expert on meta-analysis.
HTH.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics