BTW, this article suggests that legumes have some positive effect on longevity, fish might have a smaller such effect, meat may have a negative effect, alcohol has a rather broad confidence interval (which I suspect is because they’re trying to model a non-monotonic effect with a linear model), and none of the other food groups they considered have much of an effect.
I would guess that legume consumption is positive for the elderly because their folate absorption declines with age (along with other B vitamins like b12). I wonder if anyone has tried to test that yet.
And what little evidence we have points towards ovo-lacto and pescatarians having better health
Um, the article you linked seems to say that vegans are healthier:
Vegan All-cause mortality: HR 0.72, 95% CI 0.56-0.92
Pesco All-cause mortality: HR 0.81, 95% CI 0.69-0.94
Lacto-ovo All-cause mortality: HR 0.91, 95% CI 0.82-1
[Meat eaters presumably have an HR of 1]
The difference might not be significant, so I don’t know that we would call this conclusive proof. But it seems like if you’re going to lean one way, it would be towards vegans being healthier.
Especially since “animal products are bad” is a much simpler model than “animal products are bad, except for these few exceptions.”
I guess I completely failed to discuss that the studies I linked to do not constitute the entire set of studies I drew from for the recommendations. I will expand on some of the points when I have time.
Just reading the wikipedia page#Health_studies) on eggs seems to indicate that evidence for their health benefits is questionable at best, (and even though you were trying to make the argument that eggs were healthy you couldn’t find the evidence to do so at first) so given that you’re only mentioning “the largest high level features of a diet that have positive or negative impact”, I’m not convinced eggs are worth including at all.
Yeah, I believe choline is more important than the conventional wisdom suggests given its interaction with various nootropics. It’s really hard to get enough without eggs. Eggs are also absurdly bioavailable compared to everything else.
Er, that’s not what I asked; averting a deficiency presumably takes less consumption than that. Do you have evidence about choline levels, and what does that evidence say about how many eggs you’d need to eat per week to avoid it?
Averting an acute deficiency is completely different from optimal for health. I don’t have a simple cite saying this amount of choline is optimal. I have an impression based on peoples response to extra choline.
Edit: to clarify, choline is not the sole reason I strongly recommend eggs. It is possible to get enough choline without eggs, but the fact that the overwhelming majority of the populace does not meet the adequate intake makes me suspect most diets don’t fulfill this.
Yeah, I believe choline is more important than the conventional wisdom suggests given its interaction with various nootropics. It’s really hard to get enough without eggs.
It seems pretty easy to supplement with soy lecithin. Is there any reason not to do that?
Not particularly for choline other than my normal anti-processed-food-until-proven-otherwise heuristic, but eggs do also contain lots of b12, selenium, and a smaller amount of a ton of other nutrients.
Eggs are very high in methionine, though, and there’s evidence that methionine restriction can increase both mean and maximum lifespan. Some very knowledgeable folk, like Michael Rae, have dropped eggs from their diet for this reason.
“CR works in humans” evidence is bad. “CR works in primates” is bad. “CR works in mice” is shakier than it has been presented.
I don’t think this is an accurate characterization of the state of the evidence. See here for a rigorous examination of the relevant issues.
Also, it seems inconsistent to dismiss the evidence for CR in humans as “bad” and yet praise intermittent fasting (IF), given that (1) IF has been studied much less extensively than CR, (2) IF hasn’t generally shown health benefits comparable to those of CR, and (3) it is generally believed that the benefits that IF does confer are explained by its ability to mimic CR.
My praise for IF can not be based on longevity considerations because the evidence simply isn’t there. It is based on shorter term considerations regarding blood markers that should lead to better health outcomes, as well as quality of life considerations.
I read the linked examination of various CR studies, if anything I am even more dubious than before. There are simply way too many free parameters in the studies that have been done so far for me to feel confident in their results. CR requires an extraordinary lifestyle change that induces some changes that seem quite bad (hormonal and immune system), it would take excellent evidence of benefit to be worth these downsides. The methionine evidence seems even more preliminary, and I’m not going to get rid of nutrient dense food items in my diet based on it until there is stronger evidence.
I’d specifically point to the “Diminishing returns hypothesis” section of the paper as representative of my impression across all studies so far.
It is based on shorter term considerations regarding blood markers that should lead to better health outcomes, as well as quality of life considerations.
We observe even better blood markers (and other biomarkers) on people doing CR. My point is that there is no relevant line of evidence that provides stronger support for IF than for CR. So it’s hard to square your claim that the evidence for CR is bad with your praise for IF.
Upon reading your second paragraph, it now seems to me that you are actually using different evidential standards to assess these two dietary interventions, on the grounds that one (but presumably not the other) “requires an extraordinary lifestyle change”. I think however that it’s much clear to keep the evidential question (“does CR work?”) separate from the practical question (“should I try CR?”), and use consistent evidential standards across the board. Once you reach a position about the degree to which the existing evidence supports the claim that CR has various health benefits, you can then proceed to tackle the issue of whether the benefits are worth the costs. And note that methionine restriction doesn’t require anything remotely like the sacrifices required by CR, so we should clearly keep the two questions separate in this case.
We observe even better blood markers (and other biomarkers) on people doing CR.
No we don’t, we see immunosupression, elevation of some stress markers, and sex hormones crashing. The stronger support for IF is that you can lead a normal life and meet all the usual criteria we use for assessing health. With CR we improve some markers for health at the cost of others and hope we have chosen the correct ones to downgrade in our estimation of effectiveness for longevity.
If I based lifestyle changes on studies with similar levels of evidence to methionine restriction, I would be pulling a Ray Kurzweil and taking dozens of supplements. Rat/mouse models should NOT inform lifestyle changes. Their purpose is to explore hypothesis space cheaply.
Rat/mouse models should NOT inform lifestyle changes. Their purpose is to explore hypothesis space cheaply.
Your epistemology seems flawed. We have multiple lines of evidence and a proper Bayesian agent will take these all into account. This is especially the case when certain types of evidence will only be available after a period of several decades or even several generations. In such cases, we need to rely on the limited evidence we have at present.
In any case, your original point was that the evidence for methionine restriction was worse that the evidence for CR, which you claimed was “pretty bad”. Yet your grounds for concluding that the evidence for CR was “pretty bad” were partly based on practical considerations which do not apply to methionine restriction, so that argument doesn’t really work.
Let’s drop CR for now then and focus on Methionine. The hypothesis here is that I should be dropping foods with known health benefits like salmon[1] and whey[2] from my diet based on a rodent model. I can’t find any studies on human methionine restriction, do you have any pointers?
Also, is there any reason I couldn’t achieve the same effect with a methionine inhibitor like SAMe? Another plausible idea is that methionine is not a problem itself but is merely a symptom of a different issue, such as glycine deficiency (we don’t eat connective tissue much anymore).
The dairy meta-analysis you provide was supported by the “Dairy Innovation Australia Limited”, which makes me skeptical of their findings, given the multiple ways in which data can be aggregated to yield specific conclusions, and the evidence documenting the degree to which financial incentives can affect scientific findings. I haven’t looked at the other study, but I’m happy to accept that the evidence for the health benefits of fish is strong.
I haven’t researched the literature on methionine restriction extensively, but here’s a paper I found after a quick Google Scholar search.
More generally, I think having answers to questions of the following sort would be very helpful in these discussions: “To what degree do hypotheses confirmed by non-human animal models are later vindicated by experimental studies?” and “To what degree do hypotheses confirmed by correlational human studies are later vindicated by experimental studies?”
“To what degree do hypotheses confirmed by correlational human studies are later vindicated by experimental studies?”
That depends on what you mean by a ‘correlational study.’ People who analyze observational data with causality in mind spend a lot of time thinking about potential confounding and what to do about it. For example, here’s an analysis of a very large longitudinal dataset with the aim of determining a causal effect:
which does very sensible things. If you look at sensible analyses of observational data, then ‘the vindication rate’ will be related to how often the needed assumptions actually hold. If you look at non-sensible analyses (e.g. that aren’t adjusting for confounder bias, and so on), then it’s just garbage, no reason to expect better than chance then.
Note the conclusion: even though several RR’s look better for vegans, the data can’t yet make a strong case that veganism is actually better than pesc or ovo-lacto vegetarian diets. In particular, 1.0 RR is often within the 95% CI.
Right. So given that we don’t actually have any evidence to support claims like “Ovo-lacto vegetarians live significantly longer than vegans” don’t you think it makes sense to remove those claims?
No, I’m in agreement with the article that this meta analysis is the best data we have. It finds significant improvements for fish, dairy, and eggs vs vegans.
There’s a 2012 meta-analysis of longevity which did not separate out vegans, and a 2014 meta-analysis on blood pressure which also did not separate out vegans. If you have any pointers I’m glad to look at more.
again look at confidence bounds. Most of the studies you’ll find to simply lack the statistical power to make concrete recommendations. Fish seems unambiguously good and shows the largest effect sizes vs vegans (e.g. http://archinte.jamanetwork.com/article.aspx?articleID=1710093), I agree that ovo-lacto evidence is weaker, but I’ll maintain that there is slight evidence in favor of it. Given that a diet including fish, eggs, and milk, is much much easier to adhere to it remains something I recommend. Remember that my approach to nutrition in the OP is that effect sizes are small and you should focus your efforts elsewhere.
I do appreciate you taking the time to argue this point, smacking various claims with a hammer is essential.
I agree that ovo-lacto evidence is weaker, but I’ll maintain that there is slight evidence in favor of it. Given that a diet including fish, eggs, and milk, is much much easier to adhere to it remains something I recommend. Remember that my approach to nutrition in the OP is that effect sizes are small and you should focus your efforts elsewhere.
At last, we have reached convergence! I disagree slightly (the most recent article you linked again does not find significant differences between vegans and vegetarians as far as I can tell) but I’m fine calling that “slight evidence”. The problem was that the OP said:
Ovo-lacto vegetarians live significantly longer than vegans
Which doesn’t sound like it’s true in either the statistical nor the colloquial sense of the word. Right? So can we just remove that sentence pretty please?
A lot of these studies point to the same small amounts of data. This article for example discusses a new study that again reanalyzes the Adventist study data http://www.nleducation.co.uk/resources/reviews/vegetarians-live-longer-and-healthier/
We don’t really have anything better though. And what little evidence we have points towards ovo-lacto and pescatarians having better health.
BTW, this article suggests that legumes have some positive effect on longevity, fish might have a smaller such effect, meat may have a negative effect, alcohol has a rather broad confidence interval (which I suspect is because they’re trying to model a non-monotonic effect with a linear model), and none of the other food groups they considered have much of an effect.
I would guess that legume consumption is positive for the elderly because their folate absorption declines with age (along with other B vitamins like b12). I wonder if anyone has tried to test that yet.
Um, the article you linked seems to say that vegans are healthier:
The difference might not be significant, so I don’t know that we would call this conclusive proof. But it seems like if you’re going to lean one way, it would be towards vegans being healthier.
Especially since “animal products are bad” is a much simpler model than “animal products are bad, except for these few exceptions.”
I guess I completely failed to discuss that the studies I linked to do not constitute the entire set of studies I drew from for the recommendations. I will expand on some of the points when I have time.
Sounds good.
Just reading the wikipedia page#Health_studies) on eggs seems to indicate that evidence for their health benefits is questionable at best, (and even though you were trying to make the argument that eggs were healthy you couldn’t find the evidence to do so at first) so given that you’re only mentioning “the largest high level features of a diet that have positive or negative impact”, I’m not convinced eggs are worth including at all.
Yeah, I believe choline is more important than the conventional wisdom suggests given its interaction with various nootropics. It’s really hard to get enough without eggs. Eggs are also absurdly bioavailable compared to everything else.
How many eggs per week would you need to eat in order to avoid choline deficiency?
I eat 2-3 eggs a day. You do get a little choline from other sources.
Er, that’s not what I asked; averting a deficiency presumably takes less consumption than that. Do you have evidence about choline levels, and what does that evidence say about how many eggs you’d need to eat per week to avoid it?
Averting an acute deficiency is completely different from optimal for health. I don’t have a simple cite saying this amount of choline is optimal. I have an impression based on peoples response to extra choline.
Edit: to clarify, choline is not the sole reason I strongly recommend eggs. It is possible to get enough choline without eggs, but the fact that the overwhelming majority of the populace does not meet the adequate intake makes me suspect most diets don’t fulfill this.
The last time I tried doing this I ended up with some constipation. It’s possible I wasn’t drinking enough water at the time, though.
It seems pretty easy to supplement with soy lecithin. Is there any reason not to do that?
Not particularly for choline other than my normal anti-processed-food-until-proven-otherwise heuristic, but eggs do also contain lots of b12, selenium, and a smaller amount of a ton of other nutrients.
Eggs are very high in methionine, though, and there’s evidence that methionine restriction can increase both mean and maximum lifespan. Some very knowledgeable folk, like Michael Rae, have dropped eggs from their diet for this reason.
Thanks for the pointer, I am reading the rat and mice studies. So far the evidence seems weaker than the CR evidence, which is pretty bad.
Do you mean that the CR evidence is bad, or that it’s bad that the evidence for methionine restriction is weaker?
“CR works in humans” evidence is bad. “CR works in primates” is bad. “CR works in mice” is shakier than it has been presented.
I don’t think this is an accurate characterization of the state of the evidence. See here for a rigorous examination of the relevant issues.
Also, it seems inconsistent to dismiss the evidence for CR in humans as “bad” and yet praise intermittent fasting (IF), given that (1) IF has been studied much less extensively than CR, (2) IF hasn’t generally shown health benefits comparable to those of CR, and (3) it is generally believed that the benefits that IF does confer are explained by its ability to mimic CR.
My praise for IF can not be based on longevity considerations because the evidence simply isn’t there. It is based on shorter term considerations regarding blood markers that should lead to better health outcomes, as well as quality of life considerations.
I read the linked examination of various CR studies, if anything I am even more dubious than before. There are simply way too many free parameters in the studies that have been done so far for me to feel confident in their results. CR requires an extraordinary lifestyle change that induces some changes that seem quite bad (hormonal and immune system), it would take excellent evidence of benefit to be worth these downsides. The methionine evidence seems even more preliminary, and I’m not going to get rid of nutrient dense food items in my diet based on it until there is stronger evidence.
I’d specifically point to the “Diminishing returns hypothesis” section of the paper as representative of my impression across all studies so far.
We observe even better blood markers (and other biomarkers) on people doing CR. My point is that there is no relevant line of evidence that provides stronger support for IF than for CR. So it’s hard to square your claim that the evidence for CR is bad with your praise for IF.
Upon reading your second paragraph, it now seems to me that you are actually using different evidential standards to assess these two dietary interventions, on the grounds that one (but presumably not the other) “requires an extraordinary lifestyle change”. I think however that it’s much clear to keep the evidential question (“does CR work?”) separate from the practical question (“should I try CR?”), and use consistent evidential standards across the board. Once you reach a position about the degree to which the existing evidence supports the claim that CR has various health benefits, you can then proceed to tackle the issue of whether the benefits are worth the costs. And note that methionine restriction doesn’t require anything remotely like the sacrifices required by CR, so we should clearly keep the two questions separate in this case.
No we don’t, we see immunosupression, elevation of some stress markers, and sex hormones crashing. The stronger support for IF is that you can lead a normal life and meet all the usual criteria we use for assessing health. With CR we improve some markers for health at the cost of others and hope we have chosen the correct ones to downgrade in our estimation of effectiveness for longevity.
If I based lifestyle changes on studies with similar levels of evidence to methionine restriction, I would be pulling a Ray Kurzweil and taking dozens of supplements. Rat/mouse models should NOT inform lifestyle changes. Their purpose is to explore hypothesis space cheaply.
Your epistemology seems flawed. We have multiple lines of evidence and a proper Bayesian agent will take these all into account. This is especially the case when certain types of evidence will only be available after a period of several decades or even several generations. In such cases, we need to rely on the limited evidence we have at present.
In any case, your original point was that the evidence for methionine restriction was worse that the evidence for CR, which you claimed was “pretty bad”. Yet your grounds for concluding that the evidence for CR was “pretty bad” were partly based on practical considerations which do not apply to methionine restriction, so that argument doesn’t really work.
Let’s drop CR for now then and focus on Methionine. The hypothesis here is that I should be dropping foods with known health benefits like salmon[1] and whey[2] from my diet based on a rodent model. I can’t find any studies on human methionine restriction, do you have any pointers? Also, is there any reason I couldn’t achieve the same effect with a methionine inhibitor like SAMe? Another plausible idea is that methionine is not a problem itself but is merely a symptom of a different issue, such as glycine deficiency (we don’t eat connective tissue much anymore).
[1] http://circ.ahajournals.org/content/109/22/2705.short
[2] http://www.ncbi.nlm.nih.gov/pubmed/21338538
The dairy meta-analysis you provide was supported by the “Dairy Innovation Australia Limited”, which makes me skeptical of their findings, given the multiple ways in which data can be aggregated to yield specific conclusions, and the evidence documenting the degree to which financial incentives can affect scientific findings. I haven’t looked at the other study, but I’m happy to accept that the evidence for the health benefits of fish is strong.
I haven’t researched the literature on methionine restriction extensively, but here’s a paper I found after a quick Google Scholar search.
More generally, I think having answers to questions of the following sort would be very helpful in these discussions: “To what degree do hypotheses confirmed by non-human animal models are later vindicated by experimental studies?” and “To what degree do hypotheses confirmed by correlational human studies are later vindicated by experimental studies?”
That depends on what you mean by a ‘correlational study.’ People who analyze observational data with causality in mind spend a lot of time thinking about potential confounding and what to do about it. For example, here’s an analysis of a very large longitudinal dataset with the aim of determining a causal effect:
http://www.hsph.harvard.edu/wp-content/uploads/sites/1138/2012/09/ije_2009.pdf
which does very sensible things. If you look at sensible analyses of observational data, then ‘the vindication rate’ will be related to how often the needed assumptions actually hold. If you look at non-sensible analyses (e.g. that aren’t adjusting for confounder bias, and so on), then it’s just garbage, no reason to expect better than chance then.
Not sure of actual base-rate but I know it is very poor, which is why most researchers don’t take animal models very seriously.
Not sure, I think also very poor, but having a measure of this would be very valuable.
Reminder to expand on this. (Someone sent me a link saying that eggs are terrible for life expectancy, and I found it dubious so I came back here to look for links to studies.)
Link added to parent thread. Still have more to investigate. This area is extremely frustrating because of the decade-plus lead times on studies.
This overview of studies is a reasonable place to look: http://www.veganhealth.org/articles/dxrates
Note the conclusion: even though several RR’s look better for vegans, the data can’t yet make a strong case that veganism is actually better than pesc or ovo-lacto vegetarian diets. In particular, 1.0 RR is often within the 95% CI.
This is also worth looking at if I forgot to link it anywhere else: http://ajcn.nutrition.org/content/93/1/158.short
Right. So given that we don’t actually have any evidence to support claims like “Ovo-lacto vegetarians live significantly longer than vegans” don’t you think it makes sense to remove those claims?
No, I’m in agreement with the article that this meta analysis is the best data we have. It finds significant improvements for fish, dairy, and eggs vs vegans.
It’s 15 years old. I’m pretty sure there is more data available today.
There’s a 2012 meta-analysis of longevity which did not separate out vegans, and a 2014 meta-analysis on blood pressure which also did not separate out vegans. If you have any pointers I’m glad to look at more.
No pointers, sorry. But for fun I searched PubMed for “vegan” and it came up with 3200 hits...
Random example thought not meta.
again look at confidence bounds. Most of the studies you’ll find to simply lack the statistical power to make concrete recommendations. Fish seems unambiguously good and shows the largest effect sizes vs vegans (e.g. http://archinte.jamanetwork.com/article.aspx?articleID=1710093), I agree that ovo-lacto evidence is weaker, but I’ll maintain that there is slight evidence in favor of it. Given that a diet including fish, eggs, and milk, is much much easier to adhere to it remains something I recommend. Remember that my approach to nutrition in the OP is that effect sizes are small and you should focus your efforts elsewhere.
I do appreciate you taking the time to argue this point, smacking various claims with a hammer is essential.
At last, we have reached convergence! I disagree slightly (the most recent article you linked again does not find significant differences between vegans and vegetarians as far as I can tell) but I’m fine calling that “slight evidence”. The problem was that the OP said:
Which doesn’t sound like it’s true in either the statistical nor the colloquial sense of the word. Right? So can we just remove that sentence pretty please?
Sure, seems reasonable.