What about the confounding variable that your test population all have something non-nutritional in common: the institution.

Large-scale high-quality research on humans isn't possible.