meta - summarize vs. forestplot - CIs based on the t-distribution

[Ryan Walters]

Hi folks,

I am using v. 17.0 BE. When using the meta package, I want to present 95% CIs based on the t-distribution. The tdistribution option only appears to work for summarize, but is ignored by forestplot (which presents a narrower CI based on the z-distribution). The manual for meta says that the tdistribution option is valid for forestplot. My code and results are below. Any insight is appreciated.

Thanks,
Ryan

meta summarize, random(reml) eform("Risk Ratio") tdistribution

RR: 0.39, 95% CI: 0.24-0.63

meta forestplot, random(reml) eform("Risk Ratio") tdistribution

RR: 0.39, 95% CI: 0.26-0.58

meta summarize, random(reml) eform("Risk Ratio")

RR: 0.39, 95% CI: 0.26-0.58

[Tiago Pereira]

I have faced the same problem, and used the option customoverall() to present a second summary estimate (the one using the t-distribution to compute 95% CIs)