Repeated measures meta regression

[Al Bothwell]

Hi Everyone,

Is there a way to run a repeated measures meta-regression in Stata?

Thanks!

[George Ford]

metareg?

[Bruce Weaver]

Hi Al Bothwell. By "repeated measures", I suppose you mean that there can be two or more effect size estimates per study, and that they are correlated. Is that right? (E.g., this could happen if each study has 2 or more different dose groups, and the effect sizes are mean differences for each dose group vs the common control group.) Thank you for clarifying.

[Al Bothwell]

Thanks Bruce. There are "two or more effect size estimates per study, and that they are correlated". Each study has 2 or more time points where pain score (the outcome) is calculated.

[Tiago Pereira]

It is generally not clinically appropriate to combine effect sizes from different time points, as pain intensity is inherently influenced by time. A sensible approach is to select one time point per study, the one closest to the time point of interest, or to perform different meta-analyses, each for a separate time point (or group of clinically similar time points).

[Bruce Weaver]

This 2019 article may be helpful:

Cheung MW. A Guide to Conducting a Meta-Analysis with Non-Independent Effect Sizes. Neuropsychol Rev. 2019 Dec;29(4):387-396. doi: 10.1007/s11065-019-09415-6. Epub 2019 Aug 24. PMID: 31446547; PMCID: PMC6892772.

Abstract
Conventional meta-analytic procedures assume that effect sizes are independent. When effect sizes are not independent, conclusions based on these conventional procedures can be misleading or even wrong. Traditional approaches, such as averaging the effect sizes and selecting one effect size per study, are usually used to avoid the dependence of the effect sizes. These ad-hoc approaches, however, may lead to missed opportunities to utilize all available data to address the relevant research questions. Both multivariate meta-analysis and three-level meta-analysis have been proposed to handle non-independent effect sizes. This paper gives a brief introduction to these new techniques for applied researchers. The first objective is to highlight the benefits of using these methods to address non-independent effect sizes. The second objective is to illustrate how to apply these techniques with real data in R and Mplus. Researchers may modify the sample R and Mplus code to fit their data.

I don't know if anyone has developed Stata code for either type of analysis described above.

[Al Bothwell]

Thank you Bruce. I will try doing this in R.

[Bruce Weaver]

Good luck, Al Bothwell. If time permits, let us know how it worked out when you're done.

[Tiago Pereira]

-robumeta- to get a single estimate or -mvmeta- for the multivariate case are options. The latter is more difficult in sparse data cases and lack of suitable approaches to get decent covariances.