Book review: Age Later, by Nir Barzilai.

Books by serious researchers on how to defeat aging are now coming out almost as fast as I have time to read them.

This one mostly aims to enable us live in good health to 115, preferably via a few simple pills.

Age Later is fairly similar to Sinclair's Lifespan. At least, the differences are small compared to how they differ from Aubrey de Grey's Ending Aging. I'm a bit concerned by this, since anti-aging research has not yet demonstrated enough results to justify converging on a single strategy. Maybe that's just an artifact of who's writing books?

The book has only a little bit of theorizing about the causes of aging and about why we should expect it to be curable. Barzilai endorses Sinclair's unimpressive attempts at a theory.

Barzilai waits until the last page to mention an argument that aging can, in principle, be fully reversed (as opposed to delayed). To wit, that fetal cells start with an age of zero, even though the egg and sperm show clear signs of being as old as the parents.

Studies of Centenarians

A fair amount of the book is devoted to evidence from centenarians.

Barzilai provides plenty of examples which indicate that they have far from perfect genes and lifestyles. It's unlikely that we can get much in the way of useful advice from this. E.g. exercise clearly helps some people live longer, while others get can away without it. Since I can't tell which category I'm in, I treat exercise as necessary.

One questionable example:

Most striking, two of the centenarians had variants that are a major risk factor for Alzheimer's (APOE4) - the textbooks say they should have been suffering from dementia at age seventy and dead at eighty

I suggest that Barzilai look for a better textbook. Experts who study pre-agricultural tribes say that people there don't get dementia. I'm pretty sure that's due to lifestyles, not a shortage of APOE4. Also, APOE ε4 Is Not Associated with Alzheimer's Disease in Elderly Nigerians. APOE4 seems well adapted for the lifestyles of many of our distant ancestors, but not so well adapted to the lifestyles of the most developed countries. Presumably, the APOE4 centenarians had enough features of ancestral lifestyles to meet their particular needs.

Centenarian biomarkers seem to be a more interesting topic. High humanin and HDL levels in centenarians might hint at some promising interventions.

Calorie Restriction

Age Later contains a disturbing claim about the semi-famous University of Wisconsin calorie restriction study (Caloric Restriction Delays Disease Onset and Mortality in Rhesus Monkeys). Barzilai, one of the reviewers of the study, noticed that the monkeys weighed more than was plausible for calorie restricted animals, and got the researchers to admit that a caretaker had given the calorie restricted monkeys almost as much food as the controls.

This apparently means that the study's report of 30% calorie restriction was bogus. Yet my attempts to find discussion of this accusation on the web have turned up nothing. I'm very confused. How can Science just ignore this? Did everyone involved just decide that the deviation didn't last long enough to matter?

Does it mean we should dismiss the whole study? Or conclude that a smaller than reported degree of calorie restriction produced the reported benefits (i.e. that the study understates the benefits)? Barzilai seems uninterested in pursuing this issue. He thinks calorie restriction "works", at least for some humans, but that the costs outweigh the longevity benefits (he prefers a form of intermittent fasting).


Barzilai's favorite near term hope for treating aging is metformin. It has been shown to extend life in a number of species, and there's some weak evidence of large benefits in healthy humans. Barzilai makes the human evidence sound stronger than it actually is - the most interesting study showed that diabetics getting metformin outlive non-diabetic controls. However, metformin is only given to relatively healthy diabetics. It's not hard to imagine that patients who were selected to get metformin were unusually healthy except for their diabetes.

See here and here for balanced summaries of metformin. My guess is that metformin is valuable for the large number of people who are developing insulin resistance, although it's healthier, and usually possible, to avoid insulin resistance in the first place.

Barzilai does note that he considers metformin just a first step in handling aging, and he hopes it will pave the way for more ambitious research.


There's a good deal of money for research into patented treatments for aging that would likely be expensive, but not much funding for cheap treatments such as Metformin.

Barzilai sees it as critical that the FDA now recognizes aging as a disease. I see that as only having moderate effects on the cost of getting a treatment for aging approved (e.g. any aging treatment ought to qualify as a cancer treatment, but trials might need twice as many patients if only the cancer benefits count). Barzilai recounts evidence that progress is being held back much more by the reluctance of funding agencies, such as the NIH, to consider aging a disease (or is their reluctance due to a presumption that it's untreatable?), yet somehow he doesn't say that it's important for the NIH to classify aging as a disease.

Barzilai has done a great job of convincing the FDA to more or less agree to a definition of aging that recognizes it as a disease which is worth curing.

The definition seems mostly appropriate. Barzilai complains that the FDA rejected the use of diabetes as evidence of age-related loss of health. Barzilai is mostly right here. The FDA seems to be demanding that aging be defined in terms of outcomes that are clearly quite harmful (stroke, dementia, death). Diabetes doesn't guarantee a big decline in quality of life. Cancer is a more complicated story, and is likely the most critical aspect of aging to fix - how did the FDA decide to include it but not diabetes?

The definition is pretty likely to exclude just about any treatment that doesn't help with aging. But I have some concerns that it might end up locking in criteria that are overly expensive to meet. E.g. if epigenetic clocks prove to be reliable measures of age-related health, trials which use those clocks as the primary outcome might be an order of magnitude cheaper.

Barzilai expresses doubts about epigenetic clocks, because "methylation may not be such a reversible process." That's odd, since there was a study showing reversal in humans which was published in time that it could have been squeezed into the book, and an RCT showing reversal published shortly after Age Later. It seems clear that our main doubts about epigenetic clocks ought to be the risk that the clocks are just symptoms, not causes, of aging.


Age Later is a moderately good book, probably appropriate for influencing funding agencies to shift a bit more money to aging research. It's a bit more focused than I'd like on small improvements.

It didn't quite convince me to buy stock in Barzilai's company, CohBar.

New to LessWrong?

New Comment
3 comments, sorted by Click to highlight new comments since: Today at 1:42 PM

On calorie restriction: The last time I looked into this the data was very thin, such that finding out a major study had faked data might reduce the experimental evidence base to zero. If researchers were giving the calorie-restriction group extra food that they weren't supposed to, it's also likely that it was different food, which might itself explain a lifespan difference (if that food was healthier).

I think the prior is mostly against calorie restriction; looking at what the marginal energy expenditures are which would be reduced during calorie restrictions, two big ones are immunity and damage-repair, both of which are things where you'd worry about under-expenditure to produce aging-like problems. (Worse, if calorie restriction weakens immunity and this causes aging, this isn't likely to show up in animal studies, since animals have a different set of pathogens to worry about.)

On metformin: I followed the Campbell et al to the first of the four metformin-mortality studies it cites, Bannister et. I'm not entirely sure what exactly the trick is with this study, but I'm pretty sure it's tricky in some way. This observational study found lower mortality in type 2 diabetics with a mean A1c of 8.6% on metformin than in matched controls. Given that metformin is already the standard of care for T2DM, this would seem to imply that being diabetic reduces risk of death, rather than increasing it, which is nonsense.

I give more weight than you do to the variety of species in which calorie restriction has shown benefits.

It does seem quite possible that calorie restriction does worse in high disease environments. But humans in the developed world are exposed to a good deal less infectious disease than what we're evolved for. That suggests that it might be healthy to weaken our immune system slightly (but there's almost certainly an age past which that ceases to be true).

Still, I'm not at all tempted to do continuous calorie restriction. I currently do one or two days per week of limiting myself to <1000 calories, and eat until I feel full on most other days. I completely stopped calorie restriction for about three weeks near the start of the pandemic, due to increased concerns about immune system risks.

The metformin studies are downplaying some possibly important selection effects. I don't see a need to hypothesize tricks beyond that.

Both the diabetic group and the non-diabetic group had some undiagnosed health problems. For some unknown fraction of diabetics, those health problems caused them to be excluded because doctors saw that their poor health called for a switch from metformin to a more drastic treatment. It's unclear how a study could produce comparable selection effects in non-diabetics.

Note: Due to a bug in the LW API that GreaterWrong ran into, this post wasn't visible until now, but was published 3 days ago. I reset its date to give it proper time on the frontpage.