How to calculate absolute risk in a meta analysis

[Rasmus Tolstrup]

Hello everybody.
I would like to know how to calculate an absolute risk in a meta-analysis, for developing osteoarthritis after an injury (AR exposed) and for developing OA if you do not have an injury (AR not exposed). I have already calculated the OR, RR and RD using the metan commands.
I have aprox 50 studies, and I have all the data (inj+OA, notinj+OA, injwithoutOA and notinjwithoutOA) for all of the studies.

I have been using this command to find the OR, RR and RD: metan OA_inj OA_not_inj not_OA_inj not_OA_not_inj, rr lcols(Author OA_inj OA_not_inj not_OA_inj not_OA_not_inj) random

But somehow I cannot find the command for absolute risk. Can anybody help me?

Best regads, Rasmus

[Marcos Almeida]

Hello Rasmus,

Welcome to the Stata Forum!

The absolute risk is just the difference in the proportions. i.e., the risk difference.

Shall you need to pool the risk differences, you may wish to check the -rd- option in - metan - commands.

Best,

Marcos

[Rasmus Tolstrup]

Thank you Marcos.
So I can use this: metan OA_inj OA_not_inj not_OA_inj not_OA_not_inj, rd lcols(Author OA_inj OA_not_inj not_OA_inj not_OA_not_inj) random as the Risk Difference?
This gives me RD: 0.436 95%CI(0.396 to 0.477)
Hence the absolute risk for having OA with an injury is 43.6% ?

Best, Rasmus

[Marcos Almeida]

I think so, Rasmus. Actually, you may check it out, for the formula is easily applied 'by hand' as well.

[Rasmus Tolstrup]

I thought about doing it in a regular meta analysis like this:
Metan logrisk log_serisk, random eform

1. I would define the 'risk' var like this: generate risk= OA_inj/(OA_inj+ not_OA_inj)
2. I would generate the _serisk (standard error of the risk) but I do not know how to do this correct?
3. I would generate log var for both 'risk' and '_serisk' to combine it in a meta-analysis.

So by hand, I guess you mean this? Or am I wrong?

Best regards