---------------------------------------------------------------------------- Pooling research results can lead to bad medicine, study finds ---------------------------------------------------------------------------- Copyright 1997 Nando.net Copyright 1997 The Sacramento Bee (August 21, 1997 00:34 a.m. EDT) -- A popular kind of medical study has a certain problem. A third of the time, its findings are wrong. At least that is the finding of a new study, published in Thursday's New England Journal of Medicine. In question is a growing trend by researchers to draw medical conclusions by adding up the findings of smaller studies. The mathematical method is known as a meta-analysis. But research is showing that the technique has its limits. And there is potential danger for the patients of doctors who change their practice habits based on these studies. For consumers, "there is need for constant skepticism," said Dr. John Bailar, a University of Chicago medical professor and member of the New England Journal's editorial board. Among the findings from pooling smaller studies that are now in question: Whether potassium or calcium supplements can lower blood pressure, whether aspirin can prevent a dangerous complication during pregnancy and whether nitroglycerin can prevent second heart attacks. "The problem is that if you're not very careful, you might end up with the false results," said Dr. Jacques LeLorier, a medical researcher at the University of Montreal and lead author of the new study. Meta-analyses often are done several years into a new drug's life, LeLorier found. Before the government approves the release of a new drug, a large study must be done to prove it is effective for a specific problem. Once the drug is released, smaller studies follow, perhaps enrolling a few dozen patients, as researchers try to see if the drug has other uses. Those smaller studies provide the fodder for meta-analyses. In theory, a statistically savvy researcher can take many small studies and carefully add up the findings to create the numerical power of a big study, said Dr. Patrick Romano, a University of California-Davis, professor who has used the technique. But sometimes the studies are too different to add up, said Romano. Or if the study was a sloppy one, "you end up with garbage in, garbage out," said LeLorier. The problem, said Bailar, is real. "The literature is littered with the wreckage of studies that weren't worth doing and weren't worth publishing," he said. "Journals ought to do a better job of clearing out the trash." Meta-analyses are also hampered by relying on published studies, which are likelier to focus on situations when drugs appear to do some good rather than none. That leaves the journals filled with a disproportionate percentage of positive-sounding studies, a phenomenon known as "publication bias." "It is a well-recognized problem," said LeLorier. "Always the positive result is more exciting than the negative result." That is why pooling positive-sounding small studies can lead to a wrong conclusion. Fueling the problem are young researchers at academic medical centers who need to publish studies in order to secure lifetime jobs. A meta-analysis is cheap. It requires no patients, only a computer. And it is fast, taking a matter of a few months rather than large studies, which take many years. "For academic physicians who want to get promoted, it is a lot easier to do a meta-analysis than a five-year trial," said LeLorier. LeLorier's study began by reviewing 19 meta-analyses and then comparing their conclusions to the findings of 12 subsequent, larger studies that randomize half the patients onto a placebo drug and half onto the one under investigation. The large studies conflicted with the initial ones 35 percent of the time. Perhaps the most dramatic example was aspirin. A meta-analysis suggested that low doses of aspirin were very effective at preventing a toxic condition in pregnancy known as pre-eclampsia. A subsequent study involving 9,000 women found aspirin to be worthless. To combat high blood pressure, large studies found that calcium and potassium supplements worked no better than taking a sugar pill. The finding conflicted with earlier meta-analyses that found the supplements to be effective. Some skepticism of studies is healthy, said Bailar, since medicine is still very much a work in progress. "Medical science is built brick by brick," he said. "You can't understand the shape of the building from looking at one brick." By TOM PHILP, Sacramento Bee <http://www.nando.net/newsroom/ntn/health/082197/health7_5557_noframes.html> ---------------------------------------------------------------------------- [log in to unmask]