[Skip to Content]
[Skip to Content Landing]
February 5, 1997

Large Trials vs Meta-analysis of Smaller Trials-Reply

Author Affiliations

Pfizer Central Research Groton, Conn
National Institutes of Health Bethesda, Md
New England Medical Center Boston, Mass

JAMA. 1997;277(5):377-378. doi:10.1001/jama.1997.03540290029026

In Reply.  —Dr Klebanoff and colleagues bring into focus the controversy of fixed vs random effects models of combining data.1 We think it was fair and reasonable for us to report both effects and to point out that, when diversity is present, a random effects model is probably more appropriate. It is puzzling why these authors attribute to us an a priori belief that large trials usually support the results of meta-analyses of small trials.Their second concern is imprecise. We reported that a higher event rate was associated with a greater benefit of treatment in 4 cases (not 5 cases), but with a smaller benefit in 1 case. Klebanoff et al seem to believe that de facto all treatments should work the same for all patients regardless of the underlying risk of the disease. We do not share this notion and believe that most clinicians would not either.2,3 More