Meta Analysis in Practice

The goal of a meta-analysis is to synthesize data from a series of studies.  The meta-analysis allows us to see the treatment effects from the various studies in context, and to determine if the effects are consistent across studies or if they vary.  The meta-analysis yields a more precise estimate of the treatment effect than any of the included studies, and may yield a statistically significant effect even when the separate studies did not.

A case in point is the analysis published by Cannon et al (2006).  Canon synthesized data from four studies, each of which compared the impact of high-dose statins vs. standard-dose statins for preventing myocardial infarctions in a specific population of patients (for details, see the paper).  One study was statistically significant, while three were not, leading to a perception that there was no clear evidence in favor of the higher dose.

By contrast, the meta-analysis showed clearly that the increased dose was more effective, and that this effect was consistent in all four studies.  The reason that three of the four studies failed to yield a statistically significant effect was simply that they were under-powered.  When Cannon et al combined the data in a meta-analysis the p-value for combined effect was 0.00003.  Equally important, they were able to compute a precise estimate of the treatment effect, which was a risk ratio of 0.85, with a 95% confidence interval of 0.79 to 0.92.

Cannon, C.P., Steinberg, B.A., Murphy, S.A., Mega, J.L., & Braunwald, E. (2006). Meta-analysis of cardiovascular outcomes trials comparing intensive versus moderate statin therapy. Journal  of the American College Cardiology, 48, 438–445.