Skip to content

It’s not so much this one as….

Facebook will ban ads that discourage people from getting vaccinated, the social media company announced on Tuesday, as it launches a new public health campaign aimed at spreading flu vaccine information.

The changes are a departure from Facebook’s previous policy, which prohibited ads with vaccine misinformation but allowed ads expressing opposition to vaccines if they did not contain false claims.

….what’s the next thing that will be declared to be misinformation which cannot be advertised? Moderate drinking is beneficial to health? As it often enough is. Higher minimum wages cause unemployment? Health care is a public good?

After all, there really are some people who should not take a vaccine.

To put this another way. Once the shrieking classes have gained this power who thinks they’re not going to expand it?

16 thoughts on “It’s not so much this one as….”

  1. I’d be more sanguine than you, Timmy, because the shrieking classes will end up catering only for their own kind and everyone else will go elsewhere or do something better with their time.

    Christopher Snowdon has a piece today about PHE and the results of their sugar reduction programmes which demonstrates nicely how people have agency.

  2. “Moderate drinking is beneficial to health? As it often enough is.”

    I doubt that the evidence is good enough either way: intake figures will be mere self-reports, and the observations of mortality and morbidity will be distorted by confounders. Except: I’m prepared to believe that happy socialising is good for you, so why not lubricate it with a little booze? If the evidence is feeble then it certainly isn’t good evidence that moderate drinking (whatever that may be) is bad for you. Innocent until proven guilty, m’Lud.

  3. Once the shrieking classes have gained this power who thinks they’re not going to expand it?

    I’ve been pointing out lately to people that are upset about unclear, arbitrary and illogical restrictions that one of the reasons I can be reasonably sanguine is that as a smoker I’ve had 20 years of that already. It’s not really news to me. Of course, when it was just smokers getting shafted no one cared. Same deal here.

  4. Dearieme, the evidence, the J-curve which emerges in all surveys, is pretty convincing. The trouble is that those motivated by the cause of knowing what’s best for you will suppress the truth, because..it’s best for you.

  5. Facebook is not a platform. It is a publisher, and should be treated as such.

    Washington is failing us. Again.

  6. “the J-curve which emerges in all surveys, is pretty convincing.”

    It doesn’t convince me – who reports his own intake with decent accuracy? Which researcher has the magic power to correct the data for God-knows-what confounders?

    What you want is some randomised controlled trials. Which you won’t get lest they show that a wee dram is indeed good for you.

  7. @rhoda

    Dearieme is correct about the problem of confounders though – the group of people who never drink is inherently a bit weird anyway, in ways which often go beyond alcohol consumption, and that’s what causes the controversy when comparing their outcomes to people who drink moderately. There are a whole bunch of reasons why a person might not drink, which are likely to correlate with health outcomes in different ways: being a health-freak versus being a member of a religious minority versus having a medical condition versus being too cash-strapped to drink etc. There’s still some evidence that the “J” curve exists after trying to take confounders into account, but it would be easier to put strong faith in that evidence if the taking-into-account had not been so important (e.g. if never-drinkers seemed to be, in all non-alcohol-related respects, exactly the same as occasional drinkers).

  8. It’s not only the non-drinkers who are inaccurate about their intake, it’s everybody, and the tendency would always be to understate. Which shifts the curve leftwards from what an accurate curve would be. The other confounders? What matter if non-drinkers are weird in other ways? They are still less healthy. The curve says nothing about reasons. And wouldn’t you get the same curve with, say, water? Take none, die. Take too much, die.

  9. @dearieme

    I don’t think the reason we wouldn’t have an RCT is just because of fear of what the results might show, though I suspect you’re jesting. Even aside from ethics consideration and the impossibility of blinding, the whole set-up – particularly to observe long-term effects – just seems physically intractable.

    To be honest I’m not sure the issue about the “J” curve is of any great medical importance at an individual level, since even those who propose that moderate drinking produces a reduction in risk don’t seem to think it’s a really big one – if you’re a non-drinker of otherwise average habits and health there are likely other things you could do to give yourself a far bigger health boost then having the odd glass of wine. Besides, there are few areas of life whereby people deliberately set out to attain the minimal level of risk, rather they like to do what they feel like doing, within levels of risk they find tolerable. The main medical take-away seems to be that reasonably low levels of drinking don’t generally put you at particularly high risk, with the contention being over whether the minimal risk is at zero or slightly higher consumption. As Christopher Snowdon points out, the debate seems to be of more concern as a rhetorical tool for people arguing for further alcohol restrictions or reduced guidelines.

    I note that there have been quite a few RCTs in diet and nutrition, another area where we have gaping holes in our knowledge and some fiercely contested and utterly contradictory views, and yet even they don’t seem to have resolved very much – nutrition perhaps being even worse for multifactorial causation and confounders popping up everywhere.

  10. Dearieme

    ‘I doubt that the evidence is good enough either way…’

    It is. Alcohol breaks down fats in the stomach so helps digestion. In moderate amounts it gives a sense of well-being which is good for the morale and mood and thus mental health.

  11. @rhoda

    “The other confounders? What matter if non-drinkers are weird in other ways? They are still less healthy. The curve says nothing about reasons.”

    “What matters” is that it would be nice to untangle what effect the alcohol is having from what effect those other factors are having. (In fact causal effect is a tricky business, but at least, we’d like to know what association remains once other factors are taken into account). One way to think of controlling for confounders is that by splitting your sample into strata (subgroups who are more homogeneous in the factor you’re trying to control for) you end up not with “the curve” but with several curves.

    As a hypothetical example, we might stratify by presence or absence of some pre-existing condition. Let’s say we plot health problems on the vertical axis (higher is worse health outcomes) against alcohol consumption on the horizontal axis (rightwards is more alcohol consumed), separately for the those with (high-risk) and without (low-risk) the condition. Perhaps we’d find that both strata showed upwards-sloping curves, with the curve for the high-risk group parallel but above the low-risk curve, since at any level of alcohol consumption, on average the high-risk group suffers more harm. In this case it seems that, whichever of the two strata you’re in, increased alcohol consumption is associated with increased harm, and the harm-minimising consumption level is zero regardless of whether you’re low or high-risk.

    If so, the “J” curve is just a statistical artefact related to Simpson’s paradox, aka the “amalgamation paradox” – when we combine the two groups and plot their overall curve, non-drinking appears to produce greater harm because this left-hand side of the curve contains disproportionately many members of the high-risk group, who are understandably keen to keep their alcohol consumption low. Go a bit to the right and moderate drinkning contains disproportionately many low-risk people, which pulls the average harm down sufficiently so to make up for the fact that even among the low-risk group, moderate drinking produces more harm than never-drinking. Go even further to the right, which is dominated by the low-risk group (since few high-risk people dare to consume this level of alcohol), and the increased harms of higher consumption pull the curve up again. This set-up certainly doesn’t prove alcohol causes health harms or that zero is the optimal level of consumption, but it does suggest that alcohol is associated with harms once the pre-existing condition is taken into account.

    However, there is another possibility. Perhaps when we split out the high-risk and low-risk strata, we find that the high-risk curve still lies, as expected, above the low-risk curve, but both still follow the characteristic J-shape. In this case, even after controlling for the pre-existing condition, we still observe that the lowest harm (whichever stratum you’re in) occurred with above-zero alcohol consumption. If the J-shape really is just an artefact due to confounding, at least we know it isn’t due to this factor – though perhaps the picture would change if we controlled for another factor too.

    Yes the picture is muddied by measurement error both horizontally (poorly self-reported consumption) and vertically (e.g. from using a proxy measure for the health harm) but that doesn’t mean we don’t care about confounding and have no need to take it into account, if what we’re really interested in is the potential harm done by alcohol itself. Unfortunately there may be many candidates for potential confounders, likely including several we didn’t even measure.

    Moreover the “weirdness” of non-drinkers can substantially undermine our faith in the results, because more work has to be done unpicking the effects of the confounders. As an extreme example, go back to the scenario where high and low-risk strata both have parallel, increasing curves (with high-risk above low-risk), the sort of set-up that when you see it, would strongly suggest that regardless of risk-group the least harm approach is to reduce consumption to zero. Recall that a “J-curve” can arise as a statistical artefact in this case if the never-drinking and rarely-drinking left-side of the graph are disproportionately drawn from the high-risk group, while the moderate-drinking and above side of the graph graph is disproportionately drawn from the low-risk group. The “weirder” it is to never or very rarely drink, the more extreme this separation becomes, until “disproportionately” becomes “predominantly” becomes “almost exclusively”. At this stage, how do I even know that the two stratified curves actually run in parallel? At high levels of alcohol consumption, I’ve got almost no datapoints for the high-risk group (most of them are too scared to drink this much), so how do I know their harm curve keeps going a fixed height above the low-risk group? Even worse, at low levels of alcohol consumption, I’ve got almost no datapoints for the low-risk group (they’re all far too normal to drink this little), so how can I tell for sure they don’t actually follow a J-curve with increased harm at zero or near-zero consumption?

    When doing statistical modelling to try to control for confounders, it’s common not to incorporate an interaction term, and failing to do so effectively imposes an assumption of parallelism. (So in my purely hypothetical example, if the high-risk group, who could be observed at low levels of alcohol consumption, follow an upwards-sloping harm curve in this region – let’s say their condition means even small amounts of alcohol are bad for them – this would be taken as evidence that the low-risk group does so too.) I hope this sketch suggests why we care about confounders but also why, the weirder it is to never or rarely drink, the more problematic it is to decide whether the apparent harms they suffer from not drinking would apply to somebody else. The intuition is pretty simple – if the only difference between them and you seems to be that they don’t drink, you might reasonably anticipate the same thing to happen to you as to them if you stopped drinking; if they’re completely different to you in many ways other than their avoiding alcohol, that’s a much tougher prediction to make – but explaining it in terms of separate, stratified curves gives a rough idea of how multiple regression techniques actually work at untangling confounders.

  12. As far as I can see “nutrition” as a science, is on about the same level as alchemy. Humans have evolved as omnivores. The digestive system reduces pretty well anything consumed to either sugars or amino acids. Once it’s finished with digestion, the differences between any particular foods is trivial. As long as you eat a varied enough diet to get sufficient vitamins & trace minerals, what you actually choose is irrelevant. Since we’re all repeatedly told we’re all the same species, that there’s populations seem to get by on very little else but rice whilst there’s others manage on mostly animal flesh, seems to confirm that.
    Barring some rare conditions, all dietary problems are psychological. If you can’t cease eating when you’ve had sufficient, if your suffering from veganism, vegeterianism or think you can live on chips, cola & cake or lettuce & chardonay you’re mentally ill & need treatment for it.

  13. Richard Doll, the ne plus ultra of epidemiologists, demonstrated (back in the 90s) that moderate alcohol consumption (1-2 units a day for women; 2-4 for men) reduces all-cause mortality. Christopher Snowdon has written a Spectator article (possibly behind paywall) on the subject. Doll’s first paper is here.

    The problem that the prohibitionists have is that there’s well-documented and understood negative consequences of heavy drinking – nobody thinks a bottle of vodka a day is good for you (except Russians 🙂 ), so the simplistic approach is to assume a linear relationship – a little bit of alcohol is slightly bad for you. But (as with radiation and BMI) there’s little evidence to support this view.

  14. @gamecock
    To be accurate, fatty acids.
    But the point is, the digestive system breaks down what you eat into a kit of chemicals it uses to build or repair the organism or to use as energy. Little you eat is used directly, unchanged. So, since everything is going to be broken down, it doesn’t matter a damn what form what you eat takes as long as it contains what you need. It doesn’t matter whether you eat starch or sugar, since all carbohydrates end up as sugars. Your fleshy bits are built of proteins & fats but they are not the proteins & fats you eat. Those are broken down then recombined into the ones that are you. So which particular ones are in your diet is immaterial.
    So the idea that this or that is the wonder food & something else is death on a plate is pure bollocks

Leave a Reply

Your email address will not be published. Required fields are marked *