Summary measurement lost when adding mods to rma.glmm

Question

At this moment I'm trying to calculate the (adjusted)IRLM with the rma.glmm function of the metafor package.

My data is a dataframe that looks like the following:

head(data)


 patient-years events    age
1  180.0000  4    NA
2  116.2500 13 51.83
3   66.2500  6 48.00
4  423.6333 21 58.00
5  142.1783  7 53.20
6 1117.3167 72 59.90 

The function to calculate IRLM works fine:

y=rma.glmm(xi=events, ti=patent-years, data=data, measure="IRLN",method="ML")

And gives me the following forest plot:

metafor::forest.rma (y)

Forest plot

However, when I want to adjust my model:

    nh=rma.glmm(xi=events,ti=patient-years, data=datanh, 
measure="IRLN", mods = ~ age , method="ML")

(Where age is a numeric vector)

The summary measure is lost

Adjusted forest

I've tried all I can think of, but really don't know how to fix this. Do you have any suggestions?

Answer 1

When you add a moderator to the model, there is no longer the effect (or to be precise, the average effect in a random-effects model). The size of the average effect then depends on the value of the moderator. The gray-shaded polygons in the forest plot then reflect the estimated average effects corresponding to the values of 'age' for the included studies.

You could compute the predicted average effect for a particular value of age with the predict() function, ie,:

predict(nh, newmods = <age value>, transf=exp)

( transf=exp to obtain the estimated average IR for the specified age value).

Some might plug the average of the age values observed in the studies into and interpret this as an adjusted estimate. One can debate whether this terminology ('adjusted effect') is correct.