Thanks for your thoughts, Ted. I wouldn’t be too worried about the timing of the tests coinciding with coming down from a sugar-high. The number of participants and NHANES’ applied interrogation techniques (repeat dietitian-led 24-hr recalls) would make the scenario you describe rather unlikely. But not absolutely impossible , of course.

With respect to the standardization of significance, the applied standard is the 0.05 probability threshold, below which significance is assumed (to be subject to Bonferroni or other corrections in case of multiple variables being tested). What you rightfully point out is the disconnect between statistical significance and clinical relevance. This disconnect gets lost in the reporting of relative risks while omitting the absolute risk. A doubling of risk sounds big, but when it means that 2 in 100 person will have the investigated incidence vs. 1 in 100, then the relevance becomes questionable, as you point out.

In defense of statistics: it’s not so much that the statistics lie, it’s the people who either use the wrong ones or present their results in a misleading way.