et al., Diabetes and Metabolic Syndrome: Clinical Research and Reviews, 14:4, 561–565,
Database analysis of 341,484 patients in the UK with 656 hospitalized confirmed COVID-19 patients and 203 deaths, not showing a statistically significant difference after adjustment. Since adjustment factors may be correlated with vitamin D deficiency, the extent of any causal contribution of both vitamin D and the adjustment factors is unclear.
There was an ~10 year time period between baseline 25(OH)D measurement and COVID-19 infection, with 84% concordance for a subsample with measurements ~4.3 years later. Vitamin D levels may change significantly across seasons and years. People that discovered they had low vitamin D levels may have been encouraged to take steps to correct the deficiency.
Davies et al. raise a number of concerns with this study , reporting that it lacked power, suffered low precision and high bias, used flawed models, and contained many serious statistical errors. 1. Mislabelled data artificially inflated the control set, creating an illusion of high power & precision in an underpowered data set study, 2. Logistic regression violated multiple prerequisite conditions, creating biased results and a further reduction of power (overfit, over-adjusted, poor variable types, and poor handling), and 3. Unreliable data in the logistic regressions caused regression dilution bias, bias amplification, and further loss of power and precision.
See also .
Hastie et al., 8/26/2020, retrospective, population-based cohort, database analysis, United Kingdom, Europe, peer-reviewed, 14 authors.
risk of death, 17.4% lower, RR 0.83, p = 0.31, adjusted per study, >25nmol/L.
risk of hospitalization, 9.1% lower, RR 0.91, p = 0.40, adjusted per study, >25nmol/L.
Effect extraction follows pre-specified rules
prioritizing more serious outcomes. For an individual study the most serious
outcome may have a smaller number of events and lower statistical signficance,
however this provides the strongest evidence for the most serious outcomes
when combining the results of many trials.