FDA takes second look at meta-analysis

The drug industry needs new ways of dealing with conflicts in safety data. The conflicts have to be resolved faster, and more definitively. That's the conclusion of Forbes' "Medicine Show" columnist Matthew Herper.

He's talking about meta-data analyses, the chief tool used by Cleveland Clinic cardiologist Steve Nissen (photo) to support his charge that Avandia should be pulled from the market because of associated cardiac risks. Nissen also used the pooling of data from separate trials in waging his heart-risk battle against Vioxx in 2001.

Herper quotes a New England Journal of Medicine article in which FDA researchers urge meta-caution in reading meta-analysis data. The researchers note that publication of such data "commonly leads to urgent calls to take immediate regulatory action, without acknowledgment of potential pitfalls" of the data analysis technique.

The FDA, to reach sound conclusions, must consider "all available data from many sources." In other words, meta-data analysis represents another data point for the regulator to consider; it's not the superset of all trial data from which it can draw conclusions.

This has been coming. In July, at the peak of the Avandia "trial," the Wall Street Journal noted that meta-analysis can be "unreliable or misleading" because of differences among the trials whose data are being combined and whose objectives differ from that of the meta effort.

- see the column