Most biomarkers reported as potential determinants of disease risk, disease progression, or response to treatment don’t live up to their billing as originally reported, according to a statistical ysis (J. Am. Med. Assoc., DOI: 10.1001/jama.2011.713). John P. A. Ioannidis of Stanford University School of Medicine and Orestis A. Panagiotou of the University of Ioannina School of Medicine, in Greece, examined 35 highly cited biomarker associations to find out whether the reported size of each effect was accurate. They selected studies from 24 high-profile journals that indicated a relative risk in their abstract and had been cited at least 400 times. For 86% of the biomarkers, the effect reported in the most highly cited study for each biomarker was greater than that reported in larger follow-up studies. For 83% of the biomarkers, the most highly cited paper found a greater effect than was reported in a corresponding meta-ysis. Because many of these most highly cited studies were among the early reports of a particular association, Ioannidis and Panagiotou suggest that exaggerated effects could be the result of the all scale of the initial studies.
