Thanks Robert – we haven’t included physical activity in this project, although others might have in the department. Physical activity didn’t come up in our search of the Global Burden of Disease study for risk factors of poor health that cause a lot of disability/death, but that’s either because we just missed it or it’s difficult to measure. Either way, physical activity is important.

However, I did a quick search for GWAS that looked at physical activity, and one from 2018 () makes it look like genetics don’t explain a large percentage of the variation in physical activity (0.06%), making Mendelian randomization not a good choice of analysis method. There just wouldn’t be any power to detect problems. It would also be hard to disentangle whether the genetics predicted reduced physical activity, or whether they increased risk of disease that then reduced physical activity (even pre-diagnosis). It’s certainly not impossible, just tricky.

I think it unlikely that psychological factors passed down generations would affect (many) MR analyses. The key point is that the genetic variants we use to proxy the exposures (health conditions, risk factors etc.) *should* be randomly distributed, and so nothing should influence any individual’s genetic variants except chance.

However, non-random mating can introduce bias here, but that’s true of non-random mating for any reason, psychological, biological, sociological, anything.

The point is that anything passed down the generations (psychological, genetic, sociological) should be relatively independent – if you receive genetic variants predisposing you to an increase in BMI, that shouldn’t be correlated with the chance of receiving *genetic variants* predisposing you to an increase in heart disease risk, except when the assumptions within MR are broken.

LikeLike

LikeLike

LikeLike

LikeLike

LikeLike

LikeLike

LikeLike

If that were the case I would have no problem with them, but that’s not what is happening here, it’s not even remotely true. They are wedded to the idea that if consumption of alcohol amongst all demographics is reduced, this will automatically mean that harmful drinkers will curtail their intake. This is quite clearly nonsense, yet ‘public health’ clings to this as if it is the silver bullet. They even say, without any nod to the real world, that the most addicted will be the most likely to reduce their drinking when prompted. It’s absurdly stupid to think that let alone advance it as policy, but that’s exactly what Sheffield University’s mendacious minimum alcohol pricing ‘research’ claims. It’s not research, nor should it be even considered to be genuine public health work, it is paid-for lobbying material, made plain for the fact that they objected to PHE telling them to change their results because it didn’t fit in with pre-determined policy goals, but caving in anyway as they derived their income from the same source.

The public is being lied to on a daily basis. I admire your faith in the ‘public health’ profession adhering to scientific standards but they left those behind decades ago, the ‘public health’ movement is now nothing but a gravy train for junk scientists and prohibition activists who have seen an untouchable income stream on offer just for producing garbage. If you challenge their garbage, you are accused of wanting to kill children or in the pay of industry.

As I said, I admire you for sticking up for the principles of scientific rigour and objective epidemiology, but ‘public health’ abandoned that kind of approach at least a decade ago. The study that Snowdon criticises deliberately cherry-picks its sources and is a perfect mirror for the objectionable and deliberately corrupt policy-based evidence that anti-tobacco extremists have been peddling for a long time. If you are seriously interested in protecting scientific rigour within ‘public health’ circles, you should be demanding they are defunded so that they produce proper studies instead of corrupt junk science which perpetuates their income stream.

Standard denouement after every nonsense ‘study’ that ‘public health extremists produce is “more research is required”.

LikeLike

I admit to complete ignorance of who those people are, and what they do – it’s not something I’ve heard of or experienced. Note that I work in a university doing epidemiological research, although not specifically alcohol or smoking research.

While I don’t doubt your account, links to any public information would be handy. I’ve worked slightly with one of the groups in the UK centre for alcohol and tobacco studies, and from my experience, there is no agenda to stop people drinking. To reduce the amount people drink, sure (it’s on their webpage), but not stop it. And the reduction is, to my knowledge, targeted at people who are drinking enough to be experiencing problems, and those that want to reduce their drinking. But the research I’ve seen has been more of the observational type of stuff that Snowdon linked to in his article, and genetic studies. The purpose isn’t to tell people what to do, so much as find out what is going on. And in fairness, smoking is pretty bad for people – lung cancer is bad, but it’s more the COPD and emphysema that I’ve seen that looks awful. At medical school they had devices you breath into that simulate COPD and other breathing problems, and COPD in particular was pretty scary. Granted, air pollution is bad too. Walking to work in Bristol can get me coughing on certain days.

Let me reassure you and say that if there are people who make causal claims without appropriate evidence, I will wholeheartedly reject what they are saying, whether they are pro- or anti-alcohol.

But who pays them, and why?

For what it’s worth, epidemiologists at universities tend to be paid by research councils and charities (usually big ones, CRUK/Wellcome trust etc.). They aren’t in the pocket of pharma or any other vested party. I suppose they might *be* the vested party, but I believe it would be unusual. The researchers I know and work with are only interested in finding out how things are related, in order to improve health. I mean, they are probably more interested in making sure they and/or their team have jobs next year too, but in terms of research, they generally don’t want to push any particular agenda. Not least because there tends to be loads of people involved with epi papers now – making them agree on the order of points in an introduction is hard enough, let alone making them all agree to push any one agenda.

LikeLike

Thanks for the response, very happy to discuss.

Your central point wasn’t referenced (excepting the journal articles I read, which didn’t seem to be what you meant), so I couldn’t say whether anti-alcohol campaigners are either prevalent or whether their bars are set too high. If anti-alcohol campaigners are making causal claims without the evidence to back them up I’d quite happily write something about their articles too. My ire is with causal claims from observational evidence without good mechanisms and reasonable assumptions, I have no desire to make people stop drinking.

They are the usual criticisms of epi because they are the reality of working with observational data. I work with a lot of it, so have to be aware of as many flaws as possible, either to account for them, or to acknowledge the limitations of my work.

I’m only not directing the criticism at the alcohol -> cancer studies because it wasn’t part of your argument that moderate drinking is beneficial, which was the one I was rebutting. You’re right, all studies should be held to exactly the same level of rigor. But studies showing alcohol causes harm often have the advantage of good mechanisms of action, increasing the plausibility of any causal claims. “Excess alcohol consumption over decades causes cirrhosis” is plausible, given we have a good mechanism (alcohol -> acetaldehyde in liver -> damage -> cirrhosis), as well as the observational evidence, including dose-response (the more you drink, the worse the problem). “Moderate alcohol consumption is beneficial to the heart” doesn’t have the same level of observational evidence, nor the same evidence for mechanisms, as I discussed. Held to the same standard, the evidence for lots of alcohol causing harm is way more persuasive than moderate alcohol being beneficial.

How has the teetotalers dying younger than moderate drinkers been tested? All the evidence from your article was observational, excepting the trials looking at biomarkers, which, of course, doesn’t count for questions of all-cause mortality. Adjusting for confounders makes results less likely to be confounded. But unmeasured confounding is onmipresent. In the studies showing positive or negative effects of alcohol use.

“I know of no reason to think that people who are religious are more likely to get heart disease” – depends on the religion, and how it affects culture including diet and exercise, as well as exposure to any number of environmental conditions. However, I listed potential confounders off the top of my head. The point is that one’s environment, upbringing, social circles, genetics, country, religion, past alcohol use, everything about how a person interacts with the world could affect their alcohol intake, and any health outcome you cared to measure. I know that studies usually account for former drinkers (indeed, I have done so), but how much is left missing? Defining someone as a “sick quitter” basically makes a complex multi-dimensional continuous trait into a binary variable based on a few measured variables. It loses information, probably quite a lot, and is thus non-perfect. It doesn’t “debunk” either that health is relevant to drinking status or confounding as a source of bias.

Writing this, I see your point about setting the bar too high – because alcohol intake is a complex behavioural exposure, there is basically no threshold that observational evidence could meet to convince me moderate alcohol is beneficial to mortality, because I don’t believe you could ever truly say that never-drinkers and moderate drinkers are the same apart from alcohol intake. Without a clear, proven mechanism of action, it would come down to faith – if you believe alcohol is good, you trust the observational evidence, if you believe it is bad, or that you can’t be certain one way or the other, you can’t assume that confounding and reverse causation are not issues.

I would say the burden of proof is tilted more toward people who think alcohol is beneficial to find out why it has a J-shaped association with mortality/CVD, rather than on those that say there is no safe level. I say this mostly because there is plenty of robust evidence heavy alcohol use causes harm (even if you want to debate cancer links, the number of drink-driving accidents and accidents caused by alcohol are pretty clear-cut), and the effects of alcohol are pretty clearly negative on the body – dehydration, nausea, vomiting, etc.

If you want to point out any factual errors, I would gladly consider them and revise if warranted, especially if they staggered you. Translation from word to twitter to blog no doubt incorporated errors, and my knowledge of the alcohol literature is incomplete.

But then I’d also like to hear about why you don’t consider the genetic studies a source of evidence, and if you do, what you make of their lack of a J-shape curve? People without the ability to process alcohol are clearly one of the best groups to study what happens when you don’t drink alcohol. There’s potential issues with pleiotropy sure, but given the clear mechanism of action, this is dwarfed by unmeasured confounding as a source of bias.

Do you believe alcohol causes an increase in HDL, and that this means it is protective to the heart? If so, why? The RCTs are clearly biased. I haven’t seen many funnel plots that show small study effects so clearly. If you don’t think alcohol protects the heart through HDL, then how does it do it? There isn’t, so far as I know from both reading you article and the journal articles we’ve both cited any *trials* looking at alcohol and health outcomes over longer periods, and you can’t have causal claims without potential mechanisms.

I do want to say again I can see how frustrating it must be to be asked to provide evidence at ridiculously high standards to show causality in one direction, and then see things be accepted in the other direction without those same standards being applied. However, I am not saying alcohol drinking is harmful at moderate levels. Overall, it probably isn’t. My argument is solely in negating the claim that moderate alcohol intake is protective for heart health. I realise that by focusing on negating the causal claim I made that unclear.

People who drink moderately are most likely fine. I really couldn’t care less whether they reduce their drinking. What I don’t want to see happen is people who don’t drink now being told they *should* drink because it would be beneficial to their health and hearts. Equally, I wouldn’t want people who drink a lot to be put off reducing because they think alcohol is good for them.

LikeLike