Hello, We are conducting a meta-analysis to assess the potential relationship between a dichotomous outcome (access to appropriate care) and a variety of exposures based on the PROGRESS-PLUS list. For instance, this list include the variable ?race?, which in most studies is a categorical variable with choices such as African-American, Caucasian, Asian, etc. However, in some studies the reference is set as 'caucasian' while in other it is 'african-american'. Say we have 2 studies with caucasian as the reference and 2 as african-american. By default we could perform 2 meta-analyses of 2 studies each. But in order to have more power and be more comprehensive, we'd like to include all 4 by obtaining a common reference. We do not expect to obtain original data to perform an individual-patient data MA, so let's assume that all we'll have access to are the Odds ratios with 95% confidence intervals, and hopefully the proportion of outcome for every category of the exposure variable. How could we achieve this using any of the R package? Thanks, Pier-Alex
[R-meta] common reference for categorical variable in MA
2 messages · Pier-Alexandre Tardif, Michael Dewey
Dear Pier-Alex Do you mean all four just compared Caucasian with African-American but two did it one way and the other two the other? If that is true you can just invert the odds ratios by taking reciprocals or for log odds ratios just change the sign. If I have misunderstood then perhaps you can clarify? Michael
On 24/03/2023 15:23, Pier-Alexandre Tardif via R-sig-meta-analysis wrote:
Hello, We are conducting a meta-analysis to assess the potential relationship between a dichotomous outcome (access to appropriate care) and a variety of exposures based on the PROGRESS-PLUS list. For instance, this list include the variable ?race?, which in most studies is a categorical variable with choices such as African-American, Caucasian, Asian, etc. However, in some studies the reference is set as 'caucasian' while in other it is 'african-american'. Say we have 2 studies with caucasian as the reference and 2 as african-american. By default we could perform 2 meta-analyses of 2 studies each. But in order to have more power and be more comprehensive, we'd like to include all 4 by obtaining a common reference. We do not expect to obtain original data to perform an individual-patient data MA, so let's assume that all we'll have access to are the Odds ratios with 95% confidence intervals, and hopefully the proportion of outcome for every category of the exposure variable. How could we achieve this using any of the R package? Thanks, Pier-Alex [[alternative HTML version deleted]]
_______________________________________________ R-sig-meta-analysis mailing list @ R-sig-meta-analysis at r-project.org To manage your subscription to this mailing list, go to: https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis