I recently posted this graph on Twitter which suggests that Covid-19 mortality is related to latitude:
There’s a particular type of person on the internet who sees a graph like this and reaches deep into their data science boot-camp memories to exclaim “Correlation doesn’t imply causation!” or “There’s no randomized clinical trial!” or even “There are differences in testing strategy!”
These responses are stupid not because they’re wrong but because they ignore how decision-making works. It would be great if we had several high quality clinical trials for every decision. But clinical trials are difficult, expensive, and take a long time to complete, so we usually have to make medical decisions without clear clinical direction. Even in cases where there is a lot of clinical data, that data may not map perfectly onto a doctor’s actual decision. A drug may have been approved based on a large, multi-center randomized trial, but maybe they have to treat a pregnant diabetic patient, and the trial didn’t include any of them. As a result, doctors usually have to make decisions with vague, provisional information.
This is especially true in the case of Covid-19. The disease has only existed for 9 months, so obviously we’re not going to have a randomized clinical trial for most medical interventions. It would be a kind of malpractice to wait to treat patients, or engage in public health interventions, until we have clear guidance from a large, multi-center clinical trial. Medical personnel have to make medical decisions today with today’s knowledge, even if that knowledge is provisional.
The way to make these decisions is by combining prior knowledge with risk assessment.
For example, we have a lot of prior knowledge about treating acute respiratory distress with ventilators.
Since Covid-19 looks like a respiratory disease, doctors have a prior expectation that mechanical ventilation will help Covid patients when they enter respiratory distress.
Note that we do this without a clinical trial. It’s entirely possible that Covid-19 patients are not helped by mechanical ventilation or that it’s even harmful. But we shouldn’t wait to put people on ventilators until we have a Covid-19 ventilation study because patients would die in the time it takes to conduct such a study. Similarly, public health interventions were put in place without perfect certainty about how Covid is transmitted. Public health officials didn’t wait for a large clinical trial on social distancing; instead they relied on their prior knowledge about how viruses like Covid are transmitted and decided that the risks of letting the virus run wild are probably higher than putting in overly stringent public health interventions.
The second consideration is how the risk of action compares with that of inaction. Given that we don’t know much about this virus, what’s the downside risk if you’re totally wrong about the intervention? If the risks of intervention are very low and the upside is high, it’s often worth pursuing even if there’s a low probability of success. For example, we don’t really have good evidence that widespread mask usage prevents Covid-19, but since masks are very unlikely to be harmful and there’s a plausible way they could reduce transmission, we might decide to promote them. Similarly, we don’t know that closing playgrounds reduces transmission, but closing them isn’t going to put anyone at major risk so it’s probably a good bet.
Prior knowledge about Vitamin D and Covid
Our prior knowledge about vitamin D and Covid should be based on the same logic as mechanical ventilation: Covid looks like a respiratory virus, so until we learn otherwise we should treat and prevent it like a respiratory virus. A meta-analysis found two important facts about vitamin D supplementation and other respiratory conditions:
- Overall. vitamin D supplementation reduced the incidence of respiratory distress by 20%.
- For the very deficient, supplementation reduced the incidence by around 70%.
These results aren’t that impressive for most respiratory viruses, which have a low rate of complication. I might not get that excited about reducing my risk of dying from a cold from 0.05% to 0.04%, but those are big numbers for Covid because the baseline risk is higher. If we could reduce the rate of Covid respiratory distress by 20% - 70% that would not only save a lot of lives but also allow us to relax certain types of social distancing.
Again, this is just a prior. As we learn more about the virus we might discover that Covid is different from other respiratory viruses and vitamin D has no real effect. However, so far almost everything we’ve learned has supported a connection between vitamin D and Covid outcomes. For example:
- A re-analysis of 107 Swiss blood samples found that PCA positive patients had 25-hydroxyvitamin D concentrations half that of PCA negatives. This finding held after stratifying for age and gender.
- A restrospective study in the Philippines found a significant association between vitamin D status and severe Covid-19 infections.
- An Indonesian study of 712 people found that after controlling for age, gender, and co-morbidity, vitamin D deficiency was associated with a twelve times higher risk of death.
- Hospitalized male Covid-19 patients were found to have lower vitamin D levels than controls.
- A study out of Northwestern University found that the vitamin D status of a country’s elderly population was associated with the number of severe cases of Covid-19 in that country.
None of these data points are particularly strong, but they all point in the same direction and reinforce the prior expectation that vitamin D reduces incidence of Covid infection in the same way it does other respiratory conditions.
Risk of action and inaction
What happens if we supplement everyone with vitamin D and it doesn’t help Covid patients one bit?
First, people are unlikely to be harmed by supplementation. Vitamin D supplementation is extremely safe. The above meta-analysis found that supplementation didn’t increase death or adverse events from any cause. Second, even if widespread vitamin D supplementation doesn’t mitigate Covid, it will still help prevent other respiratory conditions, which is positive. In the scenario where supplementation is not helpful to Covid, it would help decrease infection and hospital visits from things like non-Covid pneumonia, and as a result there would be more ventilators and hospital resources for Covid patients. Similarly, it probably improves Covid outcomes indirectly because people who get Covid-19 after an unrelated respiratory ailment probably do worse.
So the upside of vitamin D supplementation is a significant reduction in Covid-19 transmission and mortality.
The downside is that we don’t change Covid-19 rates at all but reduce the rate of other respiratory conditions, which in turns helps our Covid-19 response.
The upside is large, the downside is also positive: therefore we should supplement with Vitamin D.
This argument should sound familiar because it’s exactly the same one that convinced so many people to wear homemade masks. There are no clinical trials testing mask use among the general population, and indeed it’s pretty difficult to imagine such a trial ever being done. There aren’t even any observational studies of populations in real world conditions. All we have are a few laboratory studies that show that masks can trap viral particles, and an association between a small number of mask-wearing countries and low Covid rates. Despite that, we have a prior expectation that since Covid mostly comes out of people’s mouths, covering those mouths probably has some protective effect. Since mask use is basically risk-free, any protective effect is worth the effort. The clinical data on vitamin D is quite a lot stronger than universal masking because we have high quality randomized clinical trials demonstrating that vitamin D protects against other respiratory viruses, and we have at least a few observational studies showing that vitamin D deficient patients are at a high risk for severe Covid. Even if vitamin D has no effect on Covid, we can be fairly sure that supplementation will free up health care resources by reducing other illnesses, and we can be very sure that it won’t cost much or cause any harm.
There are a number of clinical trials for vitamin D in Covid-19, but it would be a mistake to wait for the results of these trials because there’s substantial upside to supplementation, and no real downside.
- If you are vulnerable to low vitamin D you can get safe sun exposure or start supplementing with vitamin D.
- Facilities such as prisons, long term care facilities, and hospitals where sun exposure is limited should provide vitamin D supplementation.
- Governments should develop vitamin D plans for winter months. Argentina and New Zealand should start doing this now; Canada and Europe should plan for the fall.
- Health care workers, especially people of colour, should be tested and treated for vitamin D deficiency.
Again, these things may do nothing but reduce the number of non-Covid respiratory infections. But as interventions go, they are cheap, safe, and easy to implement. If there’s any chance they can reduce Covid transmission, they are worth doing.