“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.
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.