Hi Dmitry, thank you for the post. Well written, perfect explanations. My comment can be regarded as critical, but I thought that any comment is better than no comments at all. So…
You've already mentioned that in the introduction, and I agree with you in that: I would also be rather cautious making conclusions from single clinical observations of an association between behavioral patterns (like a healthier diet) or lab test results (e.g., hyperglycemia) and disease outcomes. As wisdom asserts, correlation does not imply causation. A tendency toward higher blood sugar is natural for acute infections, even mild ones. It may sound like truism, but this is the exact reason why, dealing with risks in medicine, people distinguish between risk factors and risk markers. Higher blood sugar in diabetes predicts a worse outcome in many acute conditions, but is it a cause of the bad outcome or just a predictor. Imagine a residential building and the smoke coming out of it. I you see no fuss, it may be a pretty insignificant accident. If you see people running out, there may be a fire. Are those running people the true cause of the smoke accident's terrible outcome or a mere predictor. If plasma glucose is really the villain, to confirm it scientifically is no expense. It would be comparatively cheap to conduct a small randomized trial with low-dose long-acting insulin or sugar-lowering tablets like the newer SGLT2 inhibitors, which make glucose come out with urine (though some believe they make promote ACE2 expression in the kidneys; doi:10.3390/ijms18051083), or any other «gentle» sugar-lowering drugs like tablets with DPP4 inhibitors. The fatality rate is so high that the small sample size won't be statistically an issue, either.