"The largest UK study (Hastie et al. 2020) found an association between vitamin D status and COVID‑19 only in a univariable analysis (with this single potential

"The largest UK study (Hastie et al. 2020) found an association between vitamin D status and COVID‑19 only in a univariable analysis (with this single potential causative factor). Importantly, no causal relationship between vitamin D status and COVID‑19 was found after adjustment for confounders such as comorbidity, socio-demographics, ethnicity, BMI and other baseline factors." A couple of problems with interpreting that study: First, it only looks at infections, not course of infection. The million-dollar question is not whether Vitamin D prevents an infection altogether, but how high the risk for a severe course of COVID-19 is, given a Vitamin-D deficiency. Secondly, it adjusts for ethnicity and the "Townshend deprivation index" as a proxy for socio-demographics. You have to look at the huge impact of being classified as "black" (i.e. very dark skin color), and the large impact of being classified as "south asian" (likely moderately dark skin color). No other factor besides "poor health" even comes close. These people are largely Vitamin-D deficient, so by controlling for that factor, you are implicitly lowering the impact of Vitamin-D. It's like saying "Oh yeah your Vitamin-D levels are low but really the problem is that you're black". If you look at the controlled numbers, being insufficient in Vitamin-D becomes protective (OR ~= 0.9), which doesn't make any sense. "Townshend deprivation" measures household overcrowding and lack of access to a car, which are obviously factors that increase the risk of contracting any viral infection. Again, this shouldn't be controlled for in order to answer the question. Of course a single-variate analysis is not enough, because many diseases are causal for low Vitamin-D levels, but a multi-variate analysis with too many variables isn't meaningful.