One of my subgroups has a Hedges' g of 1.1. Is that too high?

A Hedges' g of 1.1 is actually fine. It is quite effective, but not abnormal. Values that come from conversion errors are usually around ±3 or ±4, which is extreme. If you have checked carefully, there is no problem.

Hedges' g is usually interpreted like this:

• Small effect = 0.20
• Medium effect = 0.50
• Large effect = 0.80

In practice, sometimes you have checked everything yourself and the overall effect is still around 1.0 or even higher, and the treatment looks a little too miraculous. Our role as meta-analysis authors is to make sure we did not misread the data in the original studies. That is enough.

Why would a treatment look that effective? It is hard to say. Some studies may be of doubtful authenticity; some investigators may have selected their patients too well; some may have copied numbers wrongly when reporting. But unless you have firm evidence, we read the literature in good faith.

This is also why, when I am looking for a meta-analysis topic and find that the field has had data fabrication that led to retractions or even an academic scandal, I stay away from that field. Reviewers are usually experts in the area. If one asks me, "This field has had research fraud. How can you be sure the studies you included will not later be found fraudulent and retracted?" there is really no way to answer that.

If a field raises doubts, don't work on it.

But I think your 1.1 is fine, and you have checked carefully. You have done your part.