Today the study “Vitamin D Status in Hospitalized Patients With SARS-CoV-2 Infection” was published. The key point from the abstract is
Vitamin D deficiency was found in 82.2% of COVID-19 cases and 47.2% of population-based controls (p<0.0001)… No causal relationship was found between vitamin D deficiency and COVID-19 severity as a combined endpoint or as its separate components.
This strongly suggests a causal role for vitamin-D deficiency in covid hospitalization, and probably in risk of infection. This might explain some of the curious patterns, where the previously healthiest countries are the ones where COVID-19 has spread fastest.
There are several other such studies and a website that summarizes them.
Let’s take a look at the basis for this statistic:
including 216 patients aged ≥ 18 years, with confirmed COVID-19 admitted to the University Hospital Marqués de Valdecilla in Santander, Northern Spain from March 10 to March 31, 2020, and 197 sex-matched population-based controls recruited from the Camargo Cohort (14,15) during their last follow-up visit on January–March of the past year.
I’m a bit fuzzy on all this statistics stuff, so I’m going to walk through this in baby steps to make sure I get it right.
During that period Spain went from 3'258 confirmed covid cases to 111'541, out of a population of 47'400'000. This range of 34× during the study period makes it a bit difficult to do the calculations I want to do, but let’s use, say, 50'000, since presumably the vast majority of the patients admitted to hospitals during that period were admitted toward the end. (And as we’ll see at the end, this doesn’t matter anyway.)
Presumably only a fraction of the people with COVID-19 were hospitalized; at present 157'881 people have been hospitalized out of 1'046'132 PCR-confirmed cases, or 15%. So maybe 7500 people were hospitalized with covid during that time, and maybe we can assume that the 82.2% number is typical of them: 6200 hospitalized covid patients with vitamin-D deficiency, 1300 hospitalized covid patients without it. Out of the total population, if we assume the 47.2% number from the previous year is typical, we have 22 million people who weren't hospitalized with covid and were vitamin-D deficient, and 25 million people who weren't hospitalized with covid and weren’t vitamin-D deficient either.
So, 6200 out of 25 million vitamin-D-deficient people were hospitalized with covid at the time (250 out of every million people), and 1300 out of 22 million non-vitamin-D-deficient people were (59 per million). So the relative risk is 250 ÷ 59.
That’s a relative risk of 4.2. If you were vitamin-D deficient in Spain at that point, you were 4.2 times as likely to get hospitalized with covid than if you weren’t deficient, probably because the deficiency raises your risk of catching covid. A lot. This is a huge relative risk. (Or is it? Apparently risk ratios are the thing to use instead of odds ratios.)
Note that 6200 and 1300 are products of my estimate of the number of people hospitalized with 82.2% and (100% - 82.2%) respectively. So you get the same relative risk regardless of whether the actual number of hospitalized people was 750, 7500 or 75'000. (At 750'000 or more it might start to matter if people who later got covid were excluded from the “population-based control group” or not, but even now, in October, Spain hasn’t hospitalized nearly that many covid patients.)
Are there other explanations, other than vitamin D deficiency causing an increased risk of serious covid, probably through causing an increased risk of covid?
Well, the most obvious is that covid could cause vitamin D deficiency, for example by interfering with digestion or by directly depleting vitamin D stores. I don’t know enough about vitamin D metabolism to be very confident in this, but I don’t think it’s very likely; as a fat-soluble vitamin, it can be stored for long periods of time, so I think the body usually contains a fairly huge amount compared to what it can use in the first week or two of a covid infection.
A second possible connection is for a common cause to produce both vitamin D deficiency and covid susceptibility. As Aaron Ferrucci points out, this could be something as simple as spending time indoors and not getting exercise outside.
The above is not exhaustive, but it hopefully clarifies that the posited protective effect of vitamin D against covid might not really exist, despite the astounding risk ratio computed above.
There are other recent papers like “Vitamin D and COVID-19”, Bilezikian et al., that strongly suggest a causal mechanism, though that one cautions that it’s “a putative clinical link that at this time must still be considered hypothetical.”
Normally 40 IU is 1 μg. I’m not sure if the weird IU-density variability thing that comes into play with some other vitamins is at play here, but for now I’ll assume it’s not.
The US RDA is 600 IU or 15 μg, with the tolerable upper intake level being 4000 IU or 100 μg; Australia and NZ instead recommend 10–80 μg/day, and the EU 15–100 μg/day, same as the US.
Given this, I’d think supplementing with a dose of some 2000 IU/day would be strongly advisable, as well as getting lots of UV-B exposure.
Gwern suggests it’s important to take it in the morning, not at night, and reports that he’s taking 5000 IU per day. He says overdose starts around 70'000 IU, so it might be a good idea to start the vitamin-D regimen with a single dose on the order of 20'000 IU. He also suggests that “an hour on the beach” is likely to give you 10'000 IU, and so this should be a safe daily dose. The Endocrine Society Clinical Practice Guideline on the subject counts 4'000 IU daily as “maintenance tolerable upper limits”, and suggests that adequate blood levels “may require at least 1500–2000 IU/d[ay]”. It confirms Gwern’s thing about sunlight: a minimal erythemal dose (mild first-degree sunburn) is 20'000 IU!
Gwern also recommends vitamin D supplementation for life extension, quite aside from covid and nootropic reasons: it extends your life by an expected four months or so.
ChristianKI wrote a vitamin D primer on Lesswrong, recommending among other things to take vitamin K2 as well; this is a common practice for OTC supplements.
The CPG also mentions that the circulating half-life of the 25(OH)D form is 2–3 weeks, which reinforces my earlier-mentioned skepticism that a covid infection could drop vitamin D levels rapidly enough to provoke a deficiency in the study linked. And it mentions that body fat sequesters the vitamin, increasing the risk of deficiency, which might explain several puzzling things about covid, including how smokers are at lower risk for covid in countries with high obesity — except that smoking lowers vitamin D in Copenhagen and in Guangzhou, so the smoking link is nonexistent.
Damn, this CPG is a fucking goldmine.
Getting such large quantities of vitamin D from food is very difficult.
Vitamin D₃, cholecalciferol, is used as rat poison and possum poison with a LD₅₀ of about 10 mg/kg, so if I were a possum the acutely lethal dose would be about 1.2 grams, about 48 million IU. In humans there are concerns with continued doses over 4000 IU per day, as mentioned above.
This can of jurel (“jack mackerel”, the marketing name for horse mackerel, which is not a mackerel) says it contains 300 grams of fish and 12.5 grams of fat. This supposedly contains 4.6 IU of vitamin D per gram, so the can I just ate should have given me 1380 IU, two days’ worth of the minimal allowance but only about 8 hours of the upper limit. I don’t remember how much it cost, but maybe AR$150, 88¢, 640 microdollars per IU, or US$1.28/day for 2000 IU/day.
An egg only has about 44 IU, 1% of the upper intake level, 2% of my goal, and 14% of the US RDA. Eating four dozen eggs a day to get to 2000 IU is probably not a good idea. I don’t think eggs contribute enough to be worth consideration here.
Cod liver oil (aceite de higado de bacalao) as a supplement is 100 IU/g or 450 IU per spoonful, and eating several spoonfuls of it per day seems plausible (and is the recommended dose). 150 mℓ of cod liver oil goes for AR$825 in a bottle, which is US$4.85,or 3.2¢/mℓ, which is basically a gram I guess. This works out to 72 microdollars per IU, or 14.4¢ per day for 2000 IU/day.
This box of La Serenísima instant dry whole milk says it contains 400 g to make 3 ℓ, and a 200-mℓ serving contains 2.1 μg (84 IU, 42% of the daily value, which I guess we can deduce is 5 μg, ⅓ of the US/EU value and ½ of the .au/.nz value.) This serving supposedly has 26⅔ g of dry milk in it, so it’s about π IU of vitamin D per gram of dry milk, a bit less than the fish. I think the price per gram is also similar or maybe a little higher. I’d need to eat 600 g, a box and a half of dried milk, per day, to reach 2000 IU per day. Eating nearly a kilo of dried milk per day, consisting mostly of lactose, seems even less appealing than eating hundreds of grams of fish.
On the plus side, it’s a lot more feasible to eat 600 g of dried milk than it would be to drink 4.8 liters of milk.
Perhaps not entirely coincidentally, this supplementation level is precisely the maximum that would be allowed in the US.
The Armonía brand cut-rate instant dry whole milk is basically the same.
There are several brands of vitamin D supplements available; Puritan’s Pride sells 100 softgels of supposedly 250 μg each (10k IU) for AR$3120 (US$18.35). This is 18.35 microdollars per IU or 3.7¢ per day. Now Foods sells 120 softgels of the same dose for AR$3850 (US$22.60), or 22.6 microdollars per IU or 4.5¢ per day. Their recommended dosage is one pill every three days, which seems pretty reasonable.
Lower dosages tend to cost about the same per pill rather than per IU.