Ideal data for assessing health interventions would come from randomized controlled trials in tens of thousands of people with perfect adherence in free-living conditions, reflecting real-world food choices.

“What we really want is the randomized controlled trial in tens of thousands of people so we can look at subgroups and we'd have a lot of power. We want perfect adherence. We want it free living. We want it in people eating foods under the circumstances about which we're going to make claims.”