In der Businessweek von gestern steht ein Lesemuß für Medizinbibliothekare: When Medical Studies Collide: Contradictory reports? Meta-analysis may make things more confusing [via Auburn Steward, medlib-l].
On this roller coaster of health advice, caffeine, carbohydrates, and replacement hormones for women after menopause are bad one day, good the next. The problem is, the world of medical and health research is messier than most people realize. Black-and-white answers are rare, even when it comes to a single drug trial. […] That’s when researchers often lump data from a number of trials together in a meta-analysis, hoping the sum will be greater than the parts. But the approach often has pitfalls. In addition, Coleman used only published studies, while Barrett included two unpublished ones. That decision can have a big impact. „We know there is publication bias,“ says Frank E. Harrell Jr., chair of biostatistics at Vanderbilt University. It’s much easier to get a study published that says, „something works!“ than one saying, „Oops, the treatment had no effect.“ Using published data alone thus typically makes the final result more positive.
Und hier noch gleich ein Argument für Open Access zu den rohen Original-Studiendaten (vor dem Publication Bias):
Meta-analyses may also mislead by relying on data reported in papers rather than on original raw data, which are usually kept secret. „Good raw data from one study can be worth 50 studies in a meta-analysis,“ says Vanderbilt’s Harrell.
Alles in allem ein Plädoyer nicht gegen Metaanalysen aber für solche mit identischen Einschlußkriterien (und keiner Abhängigkeit von der Yellow Press), wie z.B. die der Cochrane Library.