Vitamin D Does Not Explain the Race/Ethnicity-COVID-19 Relationship

While I was out, a peer-reviewed paper was published in the journal, Diabetes and Metabolic Syndrome: Clinical Research & Reviews that strongly challenges the idea that vitamin D status explains the relationship between race and ethnicity and COVID-19 risk, and adds to the growing data suggesting the relationship between vitamin D and infection risk is weak.

Pubmed dates the paper to May 7, but the journal has it listed for their July-August issue and Wayback Machine doesn't show it existing prior to May 25.

This paper used data from the UK Biobank.

The UK Biobank recruited just over 500,000 participants between the ages of 37 and 73 years old from England, Scotland, and Wales, between the years 2006 and 2010. Information about vitamin D status, relevant health conditions and diseases, and socioeconomic factors was collected at baseline, and COVID-19 testing was done between March 16 and April 14, 2020.

Just under 350,000 people had a complete set of baseline data, just under 1500 people had COVID-19 testing, and just under 450 had a positive test.

Vitamin D Below 20 ng/mL Is Associated With Infection Risk

Without adjusting for any potential confounders, 25(OH)D under 20 ng/mL (50 nmol/L) was associated with a 19% higher risk of testing postive and 25(OH)D under 10 ng/mL (20 nmol/L) was associated with a 37% higher risk of testing positive.

The risk associated with levels under 20 ng/mL just missed statistical significance (p=0.068), while the risk associated with levels under 10 ng/mL achieved statistical significance (p=0.011). This means we can be very sure we would not have observed the second association if random chance were the only thing operating (we might observe an outcome of that strength or greater 1.1% of the time when randomness prevails), while we can't be so sure about the first association (we might observe an outcome of that strength or greater 6.8% of the time when randomness prevails).

They didn't report anything specific about levels between 20-30 ng/mL or about levels greater than 30 ng/mL. Since only the levels under 10 ng/mL achieved statistical significance, and since the relative risk was only 19% greater at levels 10-20 ng/mL, it seems doubtful there was an even lower risk associated with levels above 30 ng/mL. That's consistent with the one other study that correlated pre-infection vitamin D status to infection risk, which found that risk bottomed out above 20 ng/mL.

The Association Disappears When Adjusted for Confounders

When adjusted for ethnicity, sex, age, month of testing, socioeconomic status, self-reported health rating, smoking, BMI, diabetes, blood pressure, and long-standing illness or disability, the association with vitamin D status disappeared. In fact, although everything was very far from statistical significance (p=0.232-0.564), in the adjusted model, 25(OH)D under 20 ng/mL was associated with an 8% lower risk of testing positive and a 25(OH)D under 10 ng/mL was associated with a 12% lower risk of testing positive.

It isn't clear whether all of these adjustments are fair. Some of them might cause poor vitamin D status, and poor vitamin D status might also cause some of them. When vitamin D and the variable being adjusted for lie in the very same chain of cause-and-effect, the adjustment runs the risk of distorting the true relationship. Still, the association must be rather weak if it cannot withstand these adjustments and the fact that a number of those variables, discussed below, unlike vitamin D, remained significant in the adjusted model, speaks poorly of vitamin D as a cause of low infection risk.

Vitamin D Cannot Explain the Association With Race and Ethnicity

In the adjusted model, a number of predictors maintained statistical significance: males were 41% more likely to be positive; those with the highest level of socioeconomic deprivation were 89% more likely to test positive than those with the lowest; those who were overweight were 34% more likely to test positive and those who were obese were 62% more likely to test positive; blacks were 4.3 times as likely to test positive as whites, South Asians were 2.4 times as likely, and those with an “other” non-white ancestry were 1.87 times as likely.

Adjusting race and ethnicity for vitamin D status did not alter these outcomes at all.

The elevated risk for blacks was 5.5-fold unadjusted, 5.3-fold adjusted for vitamin D, and 4.3-fold adjusted for all the other variables. The elevated risk for South Asians was 2.8-fold unadjusted, 2.7-fold adjusted for vitamin D, and 2.4-fold adjusted for all the other variables. The elevated risk for other non-whites was 1.95-fold unadjusted, 1.9-fold adjusted for vitamin D, and 1.87-fold adjusted for all the other variables.

I don't think this study is that valuable for the correlation with vitamin D status, since the vitamin D status was taken 10-14 years before the COVID-19 testing.

However, if skin pigmentation is a major driver of vitamin D status, then there should be at least some degree to which that effect persists over the course of 14 or more years.

Furthermore, the unadjusted associations between vitamin D and infection risk in this study (19-37% greater risk of testing positive) were similar to the two other studies on infection risk, which showed a 73% greater risk under 20 ng/mL, and an uncalculated risk limited to those over 70, which probably equated to a roughly 50% greater chance of testing positive under 20 ng/mL. None of these vitamin D associations are even remotely close to the 4.3-5.5-fold greater risk in blacks or the 2.4-2.8-fold greater risk in South Asians.

Therefore, it seems wildly improbable that vitamin D, were it measured close to the time of infection rather than 10-14 years prior, would better explain the association with race and ethnicity in this population.

The Association With Race Is More Likely Driven by Genetics

It also seems unlikely that socioeconomic factors are driving these strong associations. All of the participants spoke English, though perhaps with different levels of fluency. Adjusting race for all of the variables did slightly reduce the associations, but this went from 5.5-fold to 4.3-fold among blacks and it went from 2.8-fold to 2.4-fold among South Asians. It seems that ancestry itself is tremendously more associated with disease risk than socioeconomic variables, age, sex, vitamin D status, and even preexisting medical conditions. This makes me suspect that differences in genetics that underly the ability of the virus to infect cells or key components of the antiviral defense that are highly relevant to this specific virus are what actually underly these associations.

Where Does This Fit With Other Vitamin D Studies?

This study is one of three that looked at the risk of testing positive, instead of looking at disease severity or mortality.

The studies on severity and mortality (here, here, and here) all pointed toward 30-35 ng/mL as the sweet spot for lowest risk. This study and the previous one on infection risk both point toward 20 ng/mL as the point after which risk is lowest. The one other study on infection risk looked at the distribution of vitamin D status in those who tested positive or negative, but didn't calculate the risk of testing positive in different categories of vitamin D status. There was no association in people under the age of 70, while in those over the age of 70 it looked like 30-35 ng/mL was associated with little or no risk of infection.  Thus, the infection risk studies point more in the direction of 20 ng/mL as maximally beneficial, but they don't all do so. Still, it's notable that the only two studies that used pre-infection levels to assess infection risk were those that pointed toward 20 ng/mL as the point after which risk is lowest.

Taking them altogether, 30-35 ng/mL looks harmless in all studies across the board, it represents the sweet spot for infection risk in those over 70 in one study, it is more than enough to minimize infection risk in the other two studies, and it is the sweet spot for lowest risk of severe or fatal disease in all the studies so far.

The infection risk studies are clearly weaker than the severity and mortality studies. The first infection risk study found no association in those under 70; the second required a very complex statistical adjustment and imputation of “likely deficient” and “likely sufficient” levels at the time of COVID-19 testing to achieve statistical significance; the third (this one) lost statistical significance and nominally reversed direction when adjusted for potential confounding factors.

The European and American studies are clearly weaker than the South and Southeast Asian Studies. The first severity study outside of South or Southeast Asia was done in Belgium. It found no association with severity in females, while the second overall vitamin D and COVID-19 study, done in Indonesia, found that the association with mortality was not impacted by sex. In the first study, done in South Asia, only 5% of patients with COVID-19 pneumonia had 25(OH)D above 30 ng/mL; in the Belgian study, 25% of those with COVID-19 pneumonia had 25(OH)D this high.  All the infection risk studies were done in Europe (Switzerland and the United Kingdom) or America (Chicago), and they look weaker, as described above.

Despite these two differences (infection risk vs severity or mortality, and the country or continent), there still may be a weakening trend in the strength of the correlation over time. For example, the first vitamin D study found 86% of mild cases had 25(OH)D above 30 ng/mL, while the third study, also done in South Asia but limited to those over the age of 60, found that only 54% of mild cases had 25(OH)D this high. The first paper found only 3.6% of severe cases had 25(OH)D that high, while the third paper showed a much greater 14% had vitamin D that high. Thus, even when looking at severity and limiting to South Asia, the apparent effect of vitamin D seemed smaller as more papers were published. This is consistent with the initial papers overestimating the effect and additional papers regressing toward the true effect.

This new paper has nothing to add on the question of a U-shaped curve. The Swiss study suggested a possible elevation of risk over 40 ng/mL in those under the age of 70. The Chicago study found no difference in risk in the 40-100 ng/mL range, but gave no information about the distribution of levels within that range. For example, we have no idea if most people in that range had 40-45 ng/mL and one had 100 ng/mL, or if they simply ended the category at 100 ng/mL arbitrarily. Data thus remains inadequate to assess the risk of infection at levels greater than 35 ng/mL.

None of the papers released yet address whether the association represents a cause-and-effect relationship.

The Bottom Line

I stand by my current position to maintain 25(OH)D in the 30-35 ng/mL range, representing the apparent sweet spot for severity, mortality, and, in those over the age of 70, infection risk, and where there is no evidence of an increase of severity, mortality, or infection risk in any group studied. I believe supplementation should be used when necessary to stay in this range.

The association between COVID-19 risk and race deserves further study, and it is unlikely it can be explained by vitamin D status, socioeconomic factors, or preexisting medical conditions, although these factors explain a small component of the association. It most likely has a basis in the genetics underlying susceptibility to infection with this specific virus, or the defense against this specific virus.

Stay safe and healthy,

Please Support This Service

These research updates are made possible by purchases of The Food and Supplement Guide to the Coronavirus. The guide contains my most up-to-date conclusions about what we should be doing for nutritional and herbal support on top of hygiene and social distancing for added protection. Due to the absence of randomized controlled trials testing nutritional or herbal prevention, these are my best guesses for what is likely to work without significant risk of harm, based on the existing science. By purchasing the guide, you are enabling me to continue devoting my skills to the most important issue we now face. I am genuinely grateful for your contribution. You can purchase the guide using this link.

For other ways to support my work, please see here.


I am not a medical doctor and this is not medical advice. I have a PhD in Nutritional Sciences and my expertise is in conducting and interpreting research related to my field. Please consult your physician before doing anything for prevention or treatment of COVID-19, and please seek the help of a physician immediately if you believe you may have COVID-19.


If you aren't subscribed to the research updates, you can sign up here.


You can access an archive of these updates here.

Comments and Questions

To leave a comment or question, please use the Facebook post for this newsletter.


*  The term “preprint” is often used in these updates. Preprints are studies destined for peer-reviewed journals that have yet to be peer-reviewed. Because COVID-19 is such a rapidly evolving disease and peer-review takes so long, most of the information circulating about the disease comes from preprints.