It's because of old-school research methods. They haven't yet developed a method which correlates any diet with any effect, so they have to look for correlations from a specific subset of data, which is assumed to return results.