An incredible study reveals that studies with incredible results usually are false
After reviewing 228 220 different clinical trials in a wide range of disciplines, statistical analysis found that studies claiming to have found ‘large effects’ in its results, are rarely able to reproduce the same results when other researchers try replicate.
The study is called ‘Empirical Evaluation of Major Impact on Medical Interventions ‘(Very Large Empirical Evaluation of Treatment Effects of Medical Interventions in English), and was written jointly by Tiago Pereira, Ralph Horwitz and the Director of the Institute of Preventive Medicine, Stanford, John Ioannidis (the latter recognized as one of the most critical voices about the ambiguities of scientific research in certain fields).
“By far the effects fade over time, become much smaller,” said Ioannidis. “Probably the most effective interventions has very modest,” said the academic, team leader and recognized by the 2005 report manipulations and exaggerations that violate scientific ethics of medical research in a paper titled ‘ For most published discoveries are false ‘.
To analyze 228,220 trials, the researchers grouped around 85,000 ‘categories’ which consisted of a single study with medical intervention (such as a specific anti-inflammatory drugs for pain after an operation) and a single result (as experience a decrease 50% after six hours of pain).
In 16% of the ‘categories’, at least one study found that the intervention group patients could feel five times better (or worse depending on what you are investigating) compared to patients in the control group received no treatment, and in 90% of these studies, the research found that reduced subsequent clinical trials these alleged effects.
What I noticed was that these studies with ‘amazing’ usually were small, with fewer than 100 participants, so these great effects ended up being the result of chance. “Trials should be of sufficient magnitude to provide useful information,” said Ioannidis.
The investigation also revealed that the studies that delivered the most relevant results were usually those measuring intermediate outcomes (such as lower cholesterol), rather than as sick or die effects. So the team made a call to always be cautious and wait for more evidence before believing that a drug has the effect that any claim his first studies.
Link: Medical studies with striking results Often Prove false (Los Angeles Times)Tags: analysis, Medicine, research, Science, Trial