August 31, 2020
On August 14, I covered a Mendelian randomization study from the UK Biobank that found no association between genetics that impact vitamin D status and COVID-19. Mendelian randomization studies are thought to be one of the best “natural” substitutes for randomized controlled trials when trying to understand whether an association represents cause-and-effect because we are born with our genes and they are distributed to us by randomly shuffling the genes our parents had. As such, we know that they aren't caused by our health conditions or our behaviors.
Although this paper found no evidence of a causal association, I pointed out that genetics only account for a small portion of vitamin D status. Further, only a small number of the genes among the 143 that had been tied to vitamin D status actually have specific relationships to vitamin D status. Many of them have broad effects on the transport of lipids in the blood or the transport of a multitude of things within cells. For these two reasons, I dismissed the study as being unable to tell us much about whether the vitamin D-COVID-19 association is causal.
The most recent study is a preprint* released on August 22, where researchers from China, Scotland, and the Netherlands performed a more complex analysis of the UK Biobank data. The data are a bit puzzling, but they offer some modest evidence in favor of causality.
Whereas the previous paper had just looked at the interaction between genes that correlate with 25(OH)D (the main marker of vitamin D status) and COVID-19 outcomes, this new paper adds interactions with the UVB light exposure in the person's locality (but not any of their lifestyle habits that would influence how much of the local UVB they were exposed to), and polymorphisms in the vitamin D receptor (which could influence the biological activity of a given 25(OH)D concentration).
The local UVB explained 12.4% of the person's 25(OH)D and genetics explained 4.2%. Together, the two factors synergized to explain 23.1% of the person's 25(OH)D.
Here are the key findings:
- Compared to those with less than 12.5 ng/mL 25(OH)D, those with 12.5-20 ng/mL had a 16% lower infection risk and those with greater than 20 ng/mL had a 31% lower infection risk.
- Neither local UVB nor vitamin D genetics predicted infection risk.
- Local UVB was associated with less hospitalization and death.
- Neither local UVB nor genetics that impact 25(OH)D had any interaction with vitamin D receptor genetics.
- When stratified into low, medium, and high local UVB, vitamin D genetics correlated with a lower infection risk, but only in the high-UVB group. They only presented this data for whites, who made up 86.5% of the sample. They didn't report whether the association held across races.
- These authors performed five different types of Mendelian randomization analyses on the vitamin D genetics. Although they all trended toward suggesting an inverse causal relationship between 25(OH)D and infection risk, only the weighted mode analysis achieved statistical significance. As described here and here, this is the type of Mendelian randomization analysis that allows the majority of the genes to be violating basic assumptions, such as, in this case, direct and specific effects on the infection risk mediated by 25(OH)D status. This is consistent with some of the genes having such a direct and specific effect, while the majority of them do not.
The most puzzling part of these findings is the fourth bullet point: genes that correlate with higher 25(OH)D levels only correlate with lower infection risk among whites in the highest third of local UVB exposure.
As the authors note in their discussion, heritability of vitamin D status in winter can reach 70-90% but can be close to zero in the summer. This is because genetics often influence the amount of dietary vitamin D or UV-B exposure that we need to maintain a certain level in our blood. UV-B exposure, in particular, tends to be self-limiting. One person might maximize their vitamin D status in ten minutes while another might take 2 hours, but if both people spend 3 hours in the sun, they're getting about the same amount of vitamin D from it because they've both gone beyond their point of maximal benefit. In the winter, those who need more UV-B may have much less vitamin D than those who need very little; in the summer, everyone is getting enough.
From this, one would think that genes make the biggest impact in low-UVB conditions, not high-UVB conditions.
On the other hand, while heritability overall is greatest in the low-UVB conditions, perhaps some genes are more relevant in high-UVB conditions. For example, any genes that influence the skin concentrations of 7-dehydrocholesterol, the precursor to vitamin D, might fall into this class. In low-UVB conditions, skin 7-dehydrocholesterol might have little value. In high-UVB conditions, it would serve as the source of a rich supply of vitamin D.
It might help if we better understood which set of genes are acting through 25(OH)D in a direct and specific way to reduce COVID-19 infection risk. The differential success of the weighted mode analysis suggests that this could be a small subset of the 143 genes, or at least, in any case, a minority of them. Maybe it is those that enhance the effect of UVB light on vitamin D status.
The differential success of the weighted mode method also seems consistent with my criticism of the first Mendelian randomization study, that only a few of the genes involved would actually have direct and specific effects on 25(OH)D.
Overall, this study provides weak and inconsistent support for a causal effect of vitamin D. Compared to the first Mendelian randomization study that found no evidence for causality, that is an upgrade.
At the end of the day, however, Mendelian randomization studies on this topic with these types of sample sizes just don't have the juice to address the question. The small number of COVID-19 patients and the small effect of genetics on 25(OH)D (only 4.2%) mean that this study only had 46% statistical power to detect a 20% difference in risk conferred by the genes. Generally a study needs 80-90% statistical power to be considered adequate.
My Other Vitamin D Posts
My other vitamin D posts help put this set of findings into the larger context:
- My Response to Rhonda Patrick on Vitamin D and COVID-19
- Update on Vitamin D and COVID-19 Using the First Observational Study Released
- The Second Study on Vitamin D and COVID-19 Is Now Out
- Two New Vitamin D Studies
- Vitamin D: The First Study on COVID-19 Infection Risk
- The First Vitamin D Study With Pre-Infection Levels Weakens the Association
- Vitamin D Does Not Explain the Race/Ethnicity-COVID-19 Relationship
- Five New Vitamin D and COVID-19 Studies
The Bottom Line
The ultimate arbiter of whether the vitamin D-COVID-19 connection is causal will be the randomized controlled trials that are currently underway.
In the meantime, it makes tremendous sense to keep 25(OH)D at 30-40 ng/mL. This is associated with the lowest all-cause mortality, is good for bones, hormonal health, and many other health endpoints, and is consistently associated with the lowest risk of COVID-19 incidence, severity, and mortality.
Stay safe and healthy,
Please Support This Service
Support the service by purchasing a copy of The Food and Supplement Guide for the Coronavirus.
Get the guide for free when pre-ordering my Vitamins and Minerals 101 book.
Or, you can offer support by buying Testing Nutritional Status: The Ultimate Cheat Sheet
Get the guide for free, get 30-50% off the pre-orders of my book, and get the Cheat Sheet for 50% off when you sign up for the CMJ Masterpass. Masterpass membership also includes a storefront where 305 brands of supplements are all 35% off, permanently. These include over 500 different vitamin D supplements. Use the code COVID19 to save 10% off the membership price.
Here's a list of other ways to support my work: How You Can Support My Work
Here are three ways to discuss this topic, including asking me questions and getting a response:
- The Masterpass FREE Forum. This forum is free and open to anyone to participate. Anything related to health and nutrition, including all aspects of the coronavirus, is welcome. I will do my best to participate several times a week, though I expect this to eventually be very large and may at some point have to participate on a weekly basis if it starts to take on a life of its own.
- The Coronavirus Forum. This is for anyone who purchases The Food and Supplement Guide for the Coronavirus, pre-orders my upcoming Vitamins and Minerals 101 book, or joins the CMJ Masterpass (if you join, use the coupon code COVID19 for 10% off the membership price). This forum is dedicated specifically to the coronavirus, has subsections based on topics (nutrition, medicine, lifestyle, mechanisms of disease), and has a section where the archive version of this newsletter is directly linked and each newsletter can be discussed as an individual thread. I consistently participate in this forum 3-5 times a week.
- The Masterpass Discussion Group. Preserved for those who join the CMJ Masterpass, it's the best place to ask me questions in a fairly intimate environment and get a rapid response. All topics I cover are fair game, and I consistently participate approximately five times per week. The Masterpass also has monthly live Zoom Q&As that are even more intimate.
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.
* 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.