Hello Emily,

The answer likely depends on: (a) what outcome is of chief interest; and (b) for the pool of studies you have that meet your screening criteria, which effect is most often reported? If the outcome is, say, survival (as in time to event), then that would call for a different ES than would, say, a dichotomous outcome (e.g., survived 5 years, did not survive 5 years).

If most of your study pool uses one type of ES (e.g., ORs), then you'll have fewer study results to try to convert if you opt for that metric.

Good luck with your work.

Look at this paper

Article Meta-Analysis of Odds Ratios: Current Good Practices

It appears that you have frequency data. If correct, then odds ratio.