I don’t know genetics but this smells like weak instrument to me? The genetic risk score instruments explained only 2% of variation in vitamin D deficiency. Relevance in the first stage of an instrumental variables analysis is crucial to get power. So the analysis is using only 2% of the variation in vitamin D deficiency to try to explain COVID severity, and unsurprisingly getting not significant results.

Can anyone who’s up to speed on their stats figure out how big the causal effect of vitamin D deficiency on COVID severity would have to be to be detected by this data and method?