PREDIMED trial of Mediterranean diet: retracted, republished, still trusted?BMJ 2019; 364 doi: https://doi.org/10.1136/bmj.l341 (Published 07 February 2019) Cite this as: BMJ 2019;364:l341
- Arnav Agarwal, research associate1,
- John P A Ioannidis, professor2
- 1Department of Medicine, University of Toronto, Toronto, Ontario, Canada
- 2Stanford Prevention Research Center and Meta-Research Innovation Center at Stanford (METRICS), Stanford University, Stanford, CA 94305, USA
- Correspondence to: J P A Ioannidis email@example.com
The Prevención con Dieta Mediterránea (PREDIMED) trial1 is one of the most influential randomised trials ever. It was cited 3364 times in Google Scholar in the five years after its publication. However, in June 2018 the trial was retracted and republished because serious protocol deviations were detected. Moreover, the repercussions of these protocol deviations and of the correction process raise many additional important questions. How do you correct one of the most influential trials and the large universe of its secondary publications?
Initial results and early raised concerns
PREDIMED was originally published in 2013.1 Heralded as a pioneer effort in nutrition,2 it randomised 7447 participants to a Mediterranean diet supplemented with extra virgin olive oil, a Mediterranean diet supplemented with nuts, or a control diet. It showed a 30% relative risk reduction in a composite clinical endpoint of cardiovascular death, stroke, or myocardial infarction in the Mediterranean diet groups.1 PREDIMED was an important effort and one of the few long term trials with clinical outcomes in nutrition.
However, some concerns were raised at the time of the publication. The interventions tested were not a typical Mediterranean diet but single food supplementations. The control group experience was not necessarily non-Mediterranean. The primary outcome was a composite of three endpoints,34 and significant differences were driven by a single endpoint (stroke) without differences in other cardiovascular disease or death. Effect sizes were probably inflated because the trial was stopped early after interim analyses showed benefit. Several systematic reviews and guidelines have either omitted PREDIMED56 or have rated it as having serious risk of bias and being difficult to interpret.7 Moreover, secondary PREDIMED analyses reported results that …