Caloric Restriction and Longevity
As with any topic, the best place to start is with a literature review, and we'll get to that in a minute. I find the studies on primates most interesting/promising, so I want to go over them briefly first:
- The first study was done by the National Institute on Ageing (NIA). As of 2012, most of the monkeys had died. Researchers found CR had no effect on mortality (p=0.934) Impact of caloric restriction on health and survival in rhesus monkeys from the NIA study. Still, they found improvements in age-related markers and medical conditions.
- A second study was done at the University of Wisconsin (UW). Researchers split the rhesus monkeys into two groups and reduced the food of one group by 30% while insuring the monkeys had enough nutrients to remain healthy. As of 2014, most the monkeys had died and they found an all-cause mortality risk ratio of 0.35 (CI = 0.16 to 0.75). They also found large reductions in age-related markers and medical conditions Caloric restriction reduces age-related and all-cause mortality in rhesus monkeys.
- A third study was done at the University of Maryland, where, rather than reducing calories by a fixed percentage, they targeted a weight that represented a typical "normal lean adult". Compared to the control group (who ate as much as they wanted), the experimental group had an all-cause mortality risk ratio of 0.58 (CI = 0.11 to 1.22) Mortality and morbidity in laboratory-maintained Rhesus monkeys and effects of long-term dietary restriction.
- Finally, one study bucked the trend and studied grey mouse lemurs, reducing the experimental group's calories by 30%. They lived 50% longer. While they also lost more grey brain matter, their cognitive and behavioral performance did not change Pifferi.
The natural question to ask is why the NIA study found no improvements, while the other three studies did. Fortunately, there's a paper examining the differences between the NIA and UW studies Caloric restriction improves health and survival of rhesus monkeys.
One of the differences was that, among males, the weight difference between the control and CR groups were 33% larger in UW than in NIA. This is despite similarly large reductions in diet and despite the fact that UW monkeys actually ate more.
They also cite research suggesting weight changes are responsible for only a small part of the longevity gains and that effects in mice depend a great deal on other factors (wild-caught vs lab, whether they're inbred, initial weight, etc).
They conjecture that its possible lab animals are overfed, and CR merely remedies this problem, suggesting CR might not yield benefits in the wild.
Though the specifics are unclear, the divergence of the second study suggests that the naive argument that "caloric restriction makes you live longer" is too simple.
Alright, with that background in place, lets dig into the recent (2017) literature review Nutrition modulation of human aging: the calorie restriction paradigm. What follows is a summary of it, which I've decided not to color (as I typically do).
Lets start with a quote:
Calorie restriction (CR) is the only nutrition intervention that, based on strong and consistent evidence in a variety of non-human species, is now widely recognized as a highly promising strategy to delay the onset and progression of age-related metabolic diseases and extend lifespan
McCay et al. were the first to show CR effects on age related disease risk reduction and extension of lifespan based on findings in rats. Similar beneficial effects have been subsequently demonstrated in various animal models, including yeast, flies, worms, fish, and in mammals.
Small-animal studies suggest that no single causal pathway or mechanism by which CR attenuates disease risk and improves markers of longevity, instead indicating that the benefits of CR are realized via a complex interplay of biological changes that are both synergistic and mutually non-exclusive. However, reduced metabolic rate, lower energy expenditure, and reduced core-body temperature have been implicated as responses to CR in both rodents and NHP [non-human primates]. The net effect of these physiologic changes is a reduction in oxidative stress-induced cellular damage and improvements in markers of inflammation and immune function; these outcomes are widely captured in CR studies and serve as markers for assessing the benefits of CR in humans. Newer mechanisms have been identified and are discussed in section 6 of this review
With this introduction out of the way, they dig into the the NIA and UW primate studies. While they admit the effect of CR on mortality is not great in NIA, they note that the monkeys did improve on 7 different health metrics, concluding that the two studies
not only validated the findings from smaller animals, but also reinforced the scientific priority for conducting randomized controlled trials (RCT) in humans
Moving on to humans, they mention two observational studies:
- One study examined the elderly in Okinawa, Japan, since its inhabitants historically eat fewer calories than those of other provinces or countries. The population is generally healthier after controlling for age.
- Members of the CR Society are typically healthier than age-and-gendered matched members of the general population.
They also cite some experimental studies:
- The Biosphere experiment accidentally tested the effects of CR on 8 adults for 2 years. This resulted in the participants eating 23% fewer calories, losing ~16% of their weight, and seeing general health improvements.
- A series of studies (CALERIE-1) were conducted on non-obese humans for 6-12 months. There were many findings, some of which were (1) 30% CR was much less sustainable among participants than 25% CR (2) the food's glycemic load didn't affect adherence (3) no safety concerns were observed even in non-obese patients, though the minimal BMI was 22, so safety has not been confirmed for people in the middle and lower portions of healthy BMIs.
After this, a second study (CALERIE-2) was conducted in which they attempted a 25% CR in non-obese non-elderly healthy adults. Both the control and experimental group received vitamin and mineral supplements to ensure those needs were being met to isolate the effects of the CR. The study lasted two years and was conducted on non-obese people to tease apart effects from weight loss with effects from CR. Over the two years, most participants failed to adhere to the 25% goal, instead achieving an average CR of 12%. The researchers found CR
was generally safe and well tolerated in participating non-obese individuals without adverse effects on quality of life, mood, cognition, hunger or sexual function.
Finally, this didn't fit in above, but they also mention that studies in rats suggest CR should begin shortly after puberty.
I'm not researcher, but as far as I can tell, there is essentially no evidence suggesting CR is net-harmful. While we lack the gold standard (long-term RCT on humans), there is good evidence in primates and other mammals and moderate evidence in humans that CR is preferable to an ad libitum diet.
So What?[This section is just a record of a mistake I made. Most people probably won't find it illuminating.]
I feel like I'm missing something important. As far as I can tell, how much you eat determines your weight (controlling for exercise), so I'm not really sure how CR differs from just generic long-term weight loss.
If there is no real difference, then the studies in normal weight people showing improvements is equivalent to saying that these normal weight people should lose weight. Note, that when the researchers say "normal weight", they don't really mean "healthy weight". Most of these people had BMIs above 25 - the upper limit for the "healthy" range.
One of the trials had a subgroup with BMIs in the 22 - 25 range and gave them a 25% CR. Unfortunately, their paper Martin lumps that subgroup in with the overweight group in all but one comparison. That one comparison was on a depressive questionnaire, where the CR made the overweight subjects do better but the normal-weight ones do worse.
I asked one of the authors for the data and he kindly linked me to his organization's home page, which provides the data in exchange for your name (see also the documentation). The data is stored in a series of CSV files but with a very SQL-esque philosophy. I still can't decide whether this storage method is genius or an abomination.
In any case, I examined various metrics, assuming I'd find a U-curve where some BMI was optimal.
I did not.
Instead, I discovered what seems obvious to me with hindsight: most body measurements are monotonic.
For example, I looked at changes in diastolic blood pressure and found that CR reduced it for both sexes for all BMIs in the range (22 - 28). In hindsight (and for most people in foresight), this was obvious. I had no reason to believe blood pressure would go up for people who were underweight. Likewise, as you lose weight your pulse tends to drop - as far a I know, there's no reason to expect this to reverse.
Instead, the main problems caused by being underweight are fundamentally caused by malnutrition (anemia, osteoporosis, etc) while the problems caused by being overweight are caused by (afaikt) inflammation and chemical disregulation (heart attacks, strokes, diabetes, etc). Because of this disconnect, there's no reason to think any single metric will make a U-shaped curve.
To be honest, I find this lack of critical thinking on my part kind of disarming. As a sort of post-mortem, I think two different things primed my thinking to be wrong here. First, it is almost certainly true that there is a BMI that (for the average person) would minimize mortality risk - it just can't be derived by these other proxies for health. Second, I had just spent a couple weeks examining the effect of vitamins and other nutrients on mortality. The effect of nutrient consumption on mortality frequently does exhibit U-shaped curves, since overconsumption can cause health problems too. I believe both these factors primed me to irrationally expect to find an "optimal" BMI by looking for extrema among the health metrics.
Like I just mentioned, most of the problems of being underweight are caused by malnutrition. I've discovered a diet of mostly broccoli and mushrooms that can give you all your daily recommended nutrient intakes for ~1000 calories. I'm no health scientist, so I'm deeply unsure, but this naively seems like it'd solve the malnutrition problems caused by being underweight.
Likewise, in most human and animal trials, the subjects are provided with recommended amounts of micronutrients even as calories are reduced, allowing these studies to presumably sidestep most of the health downsides of a light diet. In short, it seems likely that fewer calories are generally better holding nutrition constant.
The natural direction to take this research is to reduce the BMI. Based on what I'm seeing, there were no significantly negative effects in the 22-25 BMI group and the average BMI loss was ~2.4, indicating that, with proper support, modern people can support a BMI of 20 with no significant negative health effects.
On the other hand, extending this research much farther may fall afoul ethical considerations. For instance, the authors intended a BMI loss of ~4, which would've brought a 22 person down to 18, just outside the "healthy" range. Studying a calorically restrictive diet among people with BMIs much below 22 may be deemed unsafe.