I clicked through to your recommendation to floss and saw an associational study with a set of control variables. This is such a horribly bad sign that it makes me doubt the rest of your post.
Floss does have the weakest evidence going for it, hence its position last on the list. It stayed above the “worth it” line due to the low cost and risk. I also believe it has an impact on quality of life even if the mortality effect turns out to be small. I do need to add a discussion of this to my post at some point.
Necroposting, but do you have any more information on mouthwashes as a source of risk? The one I use (Crest pro health) doesn’t appear to contain chlorohexadine, but does contain another chlorine compound (cetylpyridinium chloride).
Wikipedia says cetylpyridinium chloride is an antiseptic. Assuming the blood pressure-raising mechanism is, in fact, killing off beneficial microbes, then we would expect cetylpyridinium chloride to have similar effects.
I don’t know anything about whether this product causes heart attacks, but any mouthwash containing cetylpyridinium chloride may cause horrible stains on your teeth. Check out the customer reviews on Amazon, where this mouthwash has a one star average rating. I can testify from personal experience that this phenomenon is real, and not some negative marketing campaign instigated by their competitors: I used this mouthwash for less than a month, and it took my dental hygienist almost an hour to remove the stains.
I understand your skepticism about associational studies. Clearly, the likelihood ratio from seeing a positive result in such a study should be tiny in most cases. But just out of curiosity, if you automatically discount all cohort studies, where do you expect evidence on the causal effects of lifestyle interventions to come from?
Nobody questions that doing a randomized controlled trial would provide much stronger evidence, but a RCT with a lifestyle intervention as the exposure and mortality as the outcome would take decades to complete, would require a very large sample size, and would have several potential threats to its validity, including low adherence to treatment assignment and loss to followup. Furthermore, you would need a separate arm for every possible variation of the intervention, and you would need to do one of these trials for every possible lifestyle intervention
In the absence of an RCT, the best we can do is a properly designed and properly analyzed cohort study.
As far as I know, instrumental variables are the only other option that is seriously considered, but there are very few perfect instruments, and in most realistic epidemiologic settings, using a weak instrument is probably worse than doing a cohort study. If you want to go into a further discussion on this, as a starting point, see the article “Instruments for Causal Inference: An Epidemiologist’s Dream?” by Miguel Hernan and Jamie Robins, and focus on the section on how minor violations of unverifiable assumptions can blow up the bias.
I am not suggesting that cohort studies are the answer, but rather that we only have four options: Either (1) Conduct a lot of very expensive randomized controlled trials on every possible lifestyle intervention and wait a couple of decades for the results, or (2) do associational studies, or (3) Postulate that we understand physiology and biochemistry well enough that we can learn about the effects of lifestyle intervention simply by reasoning, or (4) accept that we are unable to learn about the effects on lifestyle interventions on longevity
Personally, I am leaning towards option 4, but I am willing to accept properly conducted cohort studies as weak evidence, at least to give us some idea about what randomized trials would be most promising.
What really confuses me about your comment, is that you doubt the rest of his post simply because he cited a cohort study, when it was obvious from just reading the title of the post that the only evidence he could possibly have on the effect of lifestyle interventions, would necessarily come from associational studies.
Either (1) Conduct a lot of very expensive randomized controlled trials on every possible lifestyle intervention and wait a couple of decades for the results,
RCTS are less expensive than you think and correlational approaches more. The alternative is that instead, we run lots and lots of very expensive enormous national surveys and countless analyses of the form ‘blueberry consumption associates with better health (again)’ which still wind up being wrong something like 2/3rds the time, which wrongness itself wastes even more money by sending researchers down dead-ends (looking for the exact flavenoid which improves health) and distorting the general populations’ expenditures & quality of life & trust in science. Correlational trials are only a bargain if you’re trying to maximize citation count and confusion.
On a more positive note: #4 is unacceptable because human life is so valuable. Each year of life is worth scores of thousands of dollars, and good knowledge about lifestyle interventions like resistance exercise or aspirin can be applied to the entire American population of 300 million people indefinitely. So it’s worth paying for lots of trials from any kind of cost-benefit perspective.
And of course there’s all sorts of ways to optimize these trials to reduce the already-trivial cost of running them: factorial trials (why study just one intervention at a time?), trials designed ahead of time to fit into meta-analyses so they can borrow strength, informative priors on parameters like effect size (eg any RR <0.90 is implausible for these kinds of interventions), sequential trials to re-allocate across arms (like Thompson sampling) or just to halt early once enough information has accrued for a decision-theoretic judgement that an intervention has proven useless or useful (and can be rolled out to the population), and use of exotic covariates (the genetics of placebo response looks very interesting for increasing power)...
I agree with almost all of this. I do however think that it would be very hard to convince a sufficient number of people to let their lifestyle choices be assigned by chance, and even harder to convince them to adhere to the assigned randomization arm over several decades.
Note that if you use a factorial design, you are limiting yourself to study only joint interventions. For example, if you conduct an experiment where you first randomly assign alcohol, and then randomly assign smoking, you will be able to figure out the joint effect of these interventions and the interaction between them, but you will not be able to estimate the overall effect of using alcohol, because part of that effect may be mediated by an increased chance of taking up smoking. This can make it difficult to interpret the trials, particularly if we use high dimensional factorial designs.
I am also skeptical of re-allocation across arms, but I’ll have to read up on Thompson sampling.
What about (2′): “do associational studies, but try to implement assumptions needed for g methods to work via study design.” That is, make sure exposures are given only given the observed past, there isn’t interference by construction, etc.
I clicked through to your recommendation to floss and saw an associational study with a set of control variables. This is such a horribly bad sign that it makes me doubt the rest of your post.
Floss does have the weakest evidence going for it, hence its position last on the list. It stayed above the “worth it” line due to the low cost and risk. I also believe it has an impact on quality of life even if the mortality effect turns out to be small. I do need to add a discussion of this to my post at some point.
Some mouthwashes may be risky
Necroposting, but do you have any more information on mouthwashes as a source of risk? The one I use (Crest pro health) doesn’t appear to contain chlorohexadine, but does contain another chlorine compound (cetylpyridinium chloride).
Wikipedia says cetylpyridinium chloride is an antiseptic. Assuming the blood pressure-raising mechanism is, in fact, killing off beneficial microbes, then we would expect cetylpyridinium chloride to have similar effects.
I don’t know anything about whether this product causes heart attacks, but any mouthwash containing cetylpyridinium chloride may cause horrible stains on your teeth. Check out the customer reviews on Amazon, where this mouthwash has a one star average rating. I can testify from personal experience that this phenomenon is real, and not some negative marketing campaign instigated by their competitors: I used this mouthwash for less than a month, and it took my dental hygienist almost an hour to remove the stains.
I understand your skepticism about associational studies. Clearly, the likelihood ratio from seeing a positive result in such a study should be tiny in most cases. But just out of curiosity, if you automatically discount all cohort studies, where do you expect evidence on the causal effects of lifestyle interventions to come from?
Nobody questions that doing a randomized controlled trial would provide much stronger evidence, but a RCT with a lifestyle intervention as the exposure and mortality as the outcome would take decades to complete, would require a very large sample size, and would have several potential threats to its validity, including low adherence to treatment assignment and loss to followup. Furthermore, you would need a separate arm for every possible variation of the intervention, and you would need to do one of these trials for every possible lifestyle intervention
In the absence of an RCT, the best we can do is a properly designed and properly analyzed cohort study.
As far as I know, instrumental variables are the only other option that is seriously considered, but there are very few perfect instruments, and in most realistic epidemiologic settings, using a weak instrument is probably worse than doing a cohort study. If you want to go into a further discussion on this, as a starting point, see the article “Instruments for Causal Inference: An Epidemiologist’s Dream?” by Miguel Hernan and Jamie Robins, and focus on the section on how minor violations of unverifiable assumptions can blow up the bias.
I am not suggesting that cohort studies are the answer, but rather that we only have four options:
Either (1) Conduct a lot of very expensive randomized controlled trials on every possible lifestyle intervention and wait a couple of decades for the results, or (2) do associational studies, or (3) Postulate that we understand physiology and biochemistry well enough that we can learn about the effects of lifestyle intervention simply by reasoning, or (4) accept that we are unable to learn about the effects on lifestyle interventions on longevity
Personally, I am leaning towards option 4, but I am willing to accept properly conducted cohort studies as weak evidence, at least to give us some idea about what randomized trials would be most promising.
What really confuses me about your comment, is that you doubt the rest of his post simply because he cited a cohort study, when it was obvious from just reading the title of the post that the only evidence he could possibly have on the effect of lifestyle interventions, would necessarily come from associational studies.
RCTS are less expensive than you think and correlational approaches more. The alternative is that instead, we run lots and lots of very expensive enormous national surveys and countless analyses of the form ‘blueberry consumption associates with better health (again)’ which still wind up being wrong something like 2/3rds the time, which wrongness itself wastes even more money by sending researchers down dead-ends (looking for the exact flavenoid which improves health) and distorting the general populations’ expenditures & quality of life & trust in science. Correlational trials are only a bargain if you’re trying to maximize citation count and confusion.
On a more positive note: #4 is unacceptable because human life is so valuable. Each year of life is worth scores of thousands of dollars, and good knowledge about lifestyle interventions like resistance exercise or aspirin can be applied to the entire American population of 300 million people indefinitely. So it’s worth paying for lots of trials from any kind of cost-benefit perspective.
And of course there’s all sorts of ways to optimize these trials to reduce the already-trivial cost of running them: factorial trials (why study just one intervention at a time?), trials designed ahead of time to fit into meta-analyses so they can borrow strength, informative priors on parameters like effect size (eg any RR <0.90 is implausible for these kinds of interventions), sequential trials to re-allocate across arms (like Thompson sampling) or just to halt early once enough information has accrued for a decision-theoretic judgement that an intervention has proven useless or useful (and can be rolled out to the population), and use of exotic covariates (the genetics of placebo response looks very interesting for increasing power)...
I agree with almost all of this. I do however think that it would be very hard to convince a sufficient number of people to let their lifestyle choices be assigned by chance, and even harder to convince them to adhere to the assigned randomization arm over several decades.
Note that if you use a factorial design, you are limiting yourself to study only joint interventions. For example, if you conduct an experiment where you first randomly assign alcohol, and then randomly assign smoking, you will be able to figure out the joint effect of these interventions and the interaction between them, but you will not be able to estimate the overall effect of using alcohol, because part of that effect may be mediated by an increased chance of taking up smoking. This can make it difficult to interpret the trials, particularly if we use high dimensional factorial designs.
I am also skeptical of re-allocation across arms, but I’ll have to read up on Thompson sampling.
What about (2′): “do associational studies, but try to implement assumptions needed for g methods to work via study design.” That is, make sure exposures are given only given the observed past, there isn’t interference by construction, etc.
Out of curiosity, do we have hard data on the reliability of this vis-a-vis RCTs?