I don’t have access to the full paper, but I would expect that they at least tried to control for sun exposure.
In general, if you can think of an obvious confounding factor in about five seconds, then it’s a safe assumption that professional researchers thought of it too.
Or, at least, a safer assumption: it's worth checking to see what they said about it before publicly speculating.
And indeed, it seems they did survey for sun exposure and include it in their analysis, and they caveat a lot of their references to other work in their introduction noting where other studies didn't.
Ok but they tried to control for sun exposure just by asking how much sun exposure have you gotten in your life? A little or a lot (paraphrasing)?
The issue with this is that the amount of vitamin D someone might take is correlated with how much sun exposure is available. And the amount of available sun exposure can impact what is considered a little / a lot to each person.
> The issue with this is that the amount of vitamin D someone might take is correlated with how much sun exposure is available. And the amount of available sun exposure can impact what is considered a little / a lot to each person.
That could be case but not necessarily in this study.
I'm pretty sure XKCD created a name for this, if no-one is replying, give a horribly wrong solution to nerd snipe someone to put the time in to giving a correct solution.
> In general, if you can think of an obvious confounding factor in about five seconds, then it’s a safe assumption that professional researchers thought of it too.
I work in academic medicine. I read a lot of papers. This is not at all a given in my experience, except maybe in the tippy top journals (Nature, NEJM). When in doubt, read the paper, see if they mention the confounder you thought of.
It turns out that there was also a similar non-disclosed drop in non-Covid-related mortality. Either we discovered a magic elixir, or the entire effect is probably just confounding.
The original authors even say in their response that
> However, boosters were generally not administered to hospitalized patients who were at high risk for death from any cause.
They never even attempted to control for it.
Edit: at least NEJM accepts letters to the editor about the crap it publishes.
They probably have, but that doesn’t mean they have the necessary data to actually address the confounds. Often there is a trade off between what is most provable and what is most novel. Publishing incentives being what they are, novel invariably wins.
The study is based on a couple hundred people in a city in Finland, they could at least have tried to collect data on people closer to the tropics to hedge a bit. I doubt this has any validity ignoring such a basic confounding factor like living in a place that does have a lot of sun exposure.
Keep in mind that people from Finland are more likely to have skin that makes them particularly vulnerable to Melanoma.
A loved one is fighting this right now - if something simple like supplements, dietary practices or drug development could reduce the risk, it could potentially prevent alot of suffering. Even if it doesn’t work, perhaps it’s a line of inquiry with some value.
They could at least try to include white people in places where there is more sunlight, both my mom and my wife are white (my mom would likely be at home at a Nordic country, her family is full of redheads) and they both developed melanoma in northeastern Brazil, albeit they were found out quickly and did not cause much trouble.
Not even trying to consider the confounding variables is really sad.
Honestly I’m don’t feel qualified to evaluate the study.
It attracted my attention because the risk factors for melanoma are different than other skin cancers. A smaller number of severe sunburn incidents is correlated with melanoma while
ongoing exposure to more moderate sunlight drives other cancers. We already know that the palest skin types (ie your redhead relatives, etc) are at higher risk than even slighter less pale people.
I hope there’s more research and we learn more about ways to help bend the risk curve.
Accounting for confounders is hard. Otherwise randomised controlled studies wouldn't be needed, and we'd not have taken this long to walk back the consensus that red meat causes cancer.
Here is the description how they measured the impact of sun exposure to the results:
> The exposure of skin to UV radiation was clarified with different questions. The self-estimated lifetime exposure was studied with the following question ‘How often have you exposed yourself to sunlight during your lifetime?’ The answer options were (1) ‘seldom’, (2) ‘occasionally’, (3) ‘often’, or (4) ‘very often’. The sunburn history was studied with the following question: how often has your skin been burned due to sunlight during your lifetime? The answer options were (1) ‘seldom’, (2) ‘occasionally’, or (3) ‘often’. The answer options for the question of ‘Main environment in working history’ were (1) ‘outdoor’, (2) ‘indoor’, or (2) ‘variably both’.
They saw approximately the same distribution of sun exposure across the different test groups, it looks like.
I wish they asked exact rates and timelines. Like living in Florida, getting a bad burn only once per year might be considered seldom. But if you lived in Alaska, I’m sure that would be qualify as often.
I think just as likely is the general problem of people who take supplements being generally more conscientious and less likely to engage in risky behaviours (e.g. wearing sunscreen in summer)
Doctors recommend supplementing vitamin D if you have a measured deficiency. Most people naturally settle to 20 ng/mL or higher, and just incidental exposure, or eating certain foods can help you either absorb or synthesize it.
There are also people, like me, who no matter what, we can't make as much vitamin D for whatever cluster of genetic factors causes that. Some of us are always tired unless we take 50,000 IU of D3 a week.
That’s over double the dose considered “safe”. Obviously, I’m not suggesting you do otherwise; it’s just way outside the bounds of what most adults would do unless directed to by their physician.
The neat thing about vitamin D synthesis in the skin is that it stops once there's enough of it - it is literally impossible to overdose on vitamin D through sun exposure. So, there's a ceiling that makes "must be higher in sunny regions" far from obviously true!
That gets stored in the fat. Overdose through ingestion is a thing, our bodies aren't so smart about that; it's skin synthesis specifically that limits itself.
I’m prescribed 50,000 3x a week. The RDA was possibly miscalculated, but the only real way to know what will work for you is test - dose - test, repeat.
But they don't necessarily say what those confounding factors are in the abstract (I'm assuming they do in the full paper but I don't have access).
It's easy for me to imagine tons of confounding factors that are associated with vitamin supplementation, e.g. wealth (wealth is associated with better health outcomes nearly everywhere you look), diet in general, exercise levels, etc. I highly doubt they accounted for all of these.
The most plausible reason is that the scientific peer-reviewed result is correct, not the tiring "correlation does not imply causatian" commenter on HN who at most skimmed the paper.
2009:
> Epidemiological data show an inverse relationship between vitamin D levels and breast cancer incidence. In addition, there is a well-documented association between vitamin D intake and the risk of breast cancer. Low vitamin D intake has also been indicated in colorectal carcinogenesis. A vitamin D deficiency has also been documented in patients with prostate cancer, ovarian cancer, as well as multiple myeloma. Larger randomized clinical trials should be undertaken in humans to establish the role of vitamin D supplementation in the prevention of these cancers.
> The most plausible reason is that the scientific peer-reviewed result is correct, not the tiring "correlation does not imply causatian" commenter on HN who at most skimmed the paper.
Quite the odd take given the last sentence in the abstract of this paper is essentially "correlation is not causation":
> In conclusion, regular use of vitamin D associates with fewer melanoma cases, when compared to non-use, but the causality between them is obscure.
