Right, but I think that that’s predominately because the sample size is small. We certainly can’t conclude, from this evidence, that it does not affect mortality.
That it’s always the case doesn’t mean it’s not relevant. The evidence showed us a small effect that may or may not be a real effect. Gwern demanded to know why there was no effect. These are not the same thing.
Right, but that strikes me as a failure of n. (They started this study with 30 monkeys; what the hell were they thinking? They expanded it to 76, but it’s still too small.) It is noteworthy that the hazard rate of 3 for age-related deaths becomes only 1.5 when you look at all deaths. But what hazard rate do they expect controls to have relative to CR? If it’s around 1.5, then they should have expected from the beginning that even if this study gave the modal result, it would not be statistically significant. For comparison, smoking has a hazard rate in the neighborhood of 2- which this study would have been too small to detect most of the time (as p=.03 for the age-related death hazard rate of 3, just below the .05 level).
When you look at the individual age-related causes, the differences are dramatic. 5⁄38 of the controls were diagnosed with diabetes, and 11⁄38 were diagnosed as pre-diabetic. 0 of the experimental animals developed diabetes (and going from 40% to 0% is a big jump!). Heart disease and cancer were both reduced by 50%. Age at first age-related diagnosis was significantly later in experimental animals (p=.008). 20% of the control group had been diagnosed with an age-related condition by the time the first experimental subject was diagnosed, and at age 30, when half of the control group was dead, 66% of the experimental group had not been diagnosed with an age-related condition, compared to 23% of the control group.* Even age-related lean muscle mass deterioration was less among the CR group.
It would be nice to know about energy levels, fragility, and so on, but from reading the study it seems pretty clear something is better about the CR group than the control group, and reasonable to suspect CR is better overall than neutral or worse. It’s actually not even clear the CR group was more susceptible to injury, because the higher rates of non-age death among CR monkeys (9 instead of 7) could just be due to there being more CR monkeys (since more of the controls are buried).
* I suspect that if an animal died from a non-age cause, it is recorded as never developing an age-related disease. If you assume the best for the control group- that everyone who died was old first- and the worst for the experimental group- everyone who died wasn’t old first- then you get 23% unaged in the control and 30% unaged in the experimental group, which is still striking. If you assume things are more even, then CR wins by a large margin.
It would be nice to know about energy levels, fragility, and so on, but from reading the study it seems pretty clear something is better about the CR group than the control group, and reasonable to suspect CR is better overall than neutral or worse. It’s actually not even clear the CR group was more susceptible to injury, because the higher rates of non-age death among CR monkeys (9 instead of 7) could just be due to there being more CR monkeys (since more of the controls are buried).
The age-related diseases was encouraging, yes, but we could’ve said as much just by pointing to the human study of CRs which has thus found various heart and other improvements.
Energy levels and fragility are huge questions, since they are the obvious answer to ‘what is the tradeoff for fewer age-related diseases through CR’ - they make one fragile or reduce one’s quality of life, and additional deaths manifest some other way.
We, or at least I, are not interested in CR (as compared to gorging ourselves ad libitum) as a method of reaching our life expectancy of 77 or whatever in good shape; I’m barely even interested in it if it gains me a decade or so. (It’s a serious hit to self-control and quality of life, I think.) I’m interested in its potential for serious life-extension and breaking the usual 120-year-old limit to one’s lifespan. If it can’t deliver that, I will be disappointed.
Are there people who hit 120 who were not in good shape at 77? If not, then it seems like improving your status at 77 is a necessary step.
Beyond that, if you wouldn’t do it for an EV of 1 decade, but would do it because it gives you a chance at hitting 120, I think our priorities are very different. (Indeed, referring to 120 as the ‘usual’ limit seems odd to me. It’s five years longer than any man has survived, and there are only 6 men above 110. I wouldn’t call that usual- indeed, the ‘usual’ limit seems like it would be the median age at death. But I count my remaining lifespan up from 0, which is not a universal approach.)
Are there people who hit 120 who were not in good shape at 77? If not, then it seems like improving your status at 77 is a necessary step.
Necessary, but not sufficient, even if one grants the claim that CR results in an overall improvement (rather than a dramatic improvement on some age-related diseases and unknown compensating penalties).
Beyond that, if you wouldn’t do it for an EV of 1 decade, but would do it because it gives you a chance at hitting 120, I think our priorities are very different.
Yes. A treatment that gives another 20 years and breaks the 120 barrier is many times more interesting and full of potential than a treatment that only gives 10 years and respects the old longevity barrier.
A treatment that gives another 20 years and breaks the 120 barrier is many times more interesting and full of potential than a treatment that only gives 10 years and respects the old longevity barrier.
I agree that 20 more years is more interesting than 10 more years. What I find confusing is the probability on living past 100 without CR that’s implied by the longevity barrier being relevant.
So it doesn’t make even the too-weak significance of p=.05.
Right, but I think that that’s predominately because the sample size is small. We certainly can’t conclude, from this evidence, that it does not affect mortality.
Right, but that’s always the case- you can never accept the null hypothesis.
That it’s always the case doesn’t mean it’s not relevant. The evidence showed us a small effect that may or may not be a real effect. Gwern demanded to know why there was no effect. These are not the same thing.
Right, but that strikes me as a failure of n. (They started this study with 30 monkeys; what the hell were they thinking? They expanded it to 76, but it’s still too small.) It is noteworthy that the hazard rate of 3 for age-related deaths becomes only 1.5 when you look at all deaths. But what hazard rate do they expect controls to have relative to CR? If it’s around 1.5, then they should have expected from the beginning that even if this study gave the modal result, it would not be statistically significant. For comparison, smoking has a hazard rate in the neighborhood of 2- which this study would have been too small to detect most of the time (as p=.03 for the age-related death hazard rate of 3, just below the .05 level).
When you look at the individual age-related causes, the differences are dramatic. 5⁄38 of the controls were diagnosed with diabetes, and 11⁄38 were diagnosed as pre-diabetic. 0 of the experimental animals developed diabetes (and going from 40% to 0% is a big jump!). Heart disease and cancer were both reduced by 50%. Age at first age-related diagnosis was significantly later in experimental animals (p=.008). 20% of the control group had been diagnosed with an age-related condition by the time the first experimental subject was diagnosed, and at age 30, when half of the control group was dead, 66% of the experimental group had not been diagnosed with an age-related condition, compared to 23% of the control group.* Even age-related lean muscle mass deterioration was less among the CR group.
It would be nice to know about energy levels, fragility, and so on, but from reading the study it seems pretty clear something is better about the CR group than the control group, and reasonable to suspect CR is better overall than neutral or worse. It’s actually not even clear the CR group was more susceptible to injury, because the higher rates of non-age death among CR monkeys (9 instead of 7) could just be due to there being more CR monkeys (since more of the controls are buried).
* I suspect that if an animal died from a non-age cause, it is recorded as never developing an age-related disease. If you assume the best for the control group- that everyone who died was old first- and the worst for the experimental group- everyone who died wasn’t old first- then you get 23% unaged in the control and 30% unaged in the experimental group, which is still striking. If you assume things are more even, then CR wins by a large margin.
The age-related diseases was encouraging, yes, but we could’ve said as much just by pointing to the human study of CRs which has thus found various heart and other improvements.
Energy levels and fragility are huge questions, since they are the obvious answer to ‘what is the tradeoff for fewer age-related diseases through CR’ - they make one fragile or reduce one’s quality of life, and additional deaths manifest some other way.
We, or at least I, are not interested in CR (as compared to gorging ourselves ad libitum) as a method of reaching our life expectancy of 77 or whatever in good shape; I’m barely even interested in it if it gains me a decade or so. (It’s a serious hit to self-control and quality of life, I think.) I’m interested in its potential for serious life-extension and breaking the usual 120-year-old limit to one’s lifespan. If it can’t deliver that, I will be disappointed.
Are there people who hit 120 who were not in good shape at 77? If not, then it seems like improving your status at 77 is a necessary step.
Beyond that, if you wouldn’t do it for an EV of 1 decade, but would do it because it gives you a chance at hitting 120, I think our priorities are very different. (Indeed, referring to 120 as the ‘usual’ limit seems odd to me. It’s five years longer than any man has survived, and there are only 6 men above 110. I wouldn’t call that usual- indeed, the ‘usual’ limit seems like it would be the median age at death. But I count my remaining lifespan up from 0, which is not a universal approach.)
Necessary, but not sufficient, even if one grants the claim that CR results in an overall improvement (rather than a dramatic improvement on some age-related diseases and unknown compensating penalties).
Yes. A treatment that gives another 20 years and breaks the 120 barrier is many times more interesting and full of potential than a treatment that only gives 10 years and respects the old longevity barrier.
I agree that 20 more years is more interesting than 10 more years. What I find confusing is the probability on living past 100 without CR that’s implied by the longevity barrier being relevant.