Dear all,
I would highly appreciate your advice on some problem
which came up in a meta-analysis.
In one paper, the following results on the effect of an
ordinal exposure variable (with three categories) on the
rate of a disease have been published:
number of subjects, number of events, and person years
for each category of the exposure variable
and hazard rate ratios plus confidence intervals, adjusted
for several confounders, with "no exposure" as the reference
category.
I now need an adjusted hazard rate ratio for the other
two categories combined, i.e. a hazard rate ratio for
"exposed" versus "not exposed".
A simple estimate would be based on a weighted average
of log(HRR), weighted by the inverse of the squared
standard error, where the standard errors have to be
calculated from the published confidence intervals.
Is this an appropriate method or can you point me
to a better approach?
Kind regards
Juergen
|