[ineptech]

> This is coming from someone who spends a lot of time in the sun.

Is it coming from someone with a lot of experience evaluating the protocols of epidemiological studies? Sunlight (whether too much or too little) affects us in all sorts of poorly-understood ways; that's why it's useful to do a big correlation study instead of just examining melanomas.

[oaktrout]

"Is it coming from someone with a lot of experience evaluating the protocols of epidemiological studies?"

It is, but I prefer to not involve my own credentials as a boring appeal to authority and focus on the argument at hand. I mentioned spending time in the sun to show I have no horse in the race. I only bring up melanoma because the cohort is “Melanoma In Southern Sweden”, but I think your point is fair regarding likely multifactorial causes.

“Sunlight (whether too much or too little) affects us in all sorts of poorly-understood ways; that's why it's useful to do a big correlation study instead of just examining melanomas”

Still, if I were designing a study to measure the impact of exercise on all cause mortality, I would not have a question like “Do you frequently jog to the farmers market to pick up vegetables?”. Half of the questions ask about going on vacation, maybe we are seeing that people who have enough time to “go abroad” twice a year live longer, I would believe that. I hope this is obvious and that even someone who doesn’t have “a lot of experience evaluating” these studies can see the flaw.

The study didn’t include any men!!! A study of women only might not generalize to the whole population, especially when it comes to all cause mortality. One obvious example: men have 1.5x higher risk of dying from melanoma than women. The study excluded individuals with a history of malignant melanoma.

The no-exposure group exercised less and had lower education, and there is a large risk for residual confounding that could have likely been accounted for by using categorical variables with more than 3 options.

Further criticisms include the “no exposure group” having much more women aged 55-64 (with likely much more comorbidities). They used length to follow up time to account for this which has no place in an observational study. The study excluded women who had a history of cancer.

I appreciate the work the authors did, but remain unconvinced that it is sunlight that caused lower mortality in their study.

[ineptech]

> Half of the questions ask about going on vacation, maybe we are seeing that people who have enough time to “go abroad” twice a year live longer, I would believe that.

First of all, they explicitly mention controlling for income disparity. Second, the questions seemed reasonable to me. One for sunbathing in summer, one for sunbathing in winter which mentions travel, one for tanning beds, and one for specifically about vacationing in sunny places.

Keep in mind the context. This study was not designed to answer "is sunlight good for you" or "do people who sunbathe a lot live longer". AFAICT it was designed to answer "is avoiding sunlight bad for you." Those questions seem reasonable to suss out how much someone seeks/avoids sun exposure. There might be better ones, but it seems hubristic to just assume so, so casually.

> The study didn’t include any men!!!

I don't know why they limited the study to women. I assumed it was to make it easier to control for confounds, but they could've had some other reason. Maybe sunscreen ads are mostly targeted towards women in Sweden, and they thought a larger number of women would turn out to be sun-avoiders. In any event, this isn't a problem with the study, this is just complaining that they studied something different than what you're interested in knowing.

> [differences between the most-sun and least-sun in education, exercise, age, etc]

Of course there were differences, those things correlate with pretty much everything worth studying. Are you saying that the authors didn't control for them (then why collect data on them and include it?), or that they did so badly (why do you think that), or something else?

> I appreciate the work the authors did, but remain unconvinced that it is sunlight that caused lower mortality in their study.

I don't think epidemiological studies ever prove that anything causes anything, we're talking about correlations here. The conclusion was that avoiding sun is associated with higher all-cause mortality among women. I'll believe that until there's a reason not to.

[oaktrout]

Re studying only women: it's a problem when you are making a claim about all cause mortality but failing to mention that it's all cause mortality "in women", you can't generalize.

"Are you saying that the authors didn't control for them (then why collect data on them and include it?), or that they did so badly (why do you think that), or something else"

Yes, I addressed this above, see my comment on residual confounding, amongst my many other criticisms.