Book of note: “The Personalized Diet” by Eran Segal and Eran Elinav

I have blogged earlier about the book by neuroscientist Sandra Aamodt and have discussed there in passing the pioneering work by Weizmann Institute scientists Eran Segal and Eran Elinav on the individual microbiome (our “gut bacteria population”) and how it affects blood sugar levels. Now the duo has teamed up with editor Eve Adamson, and together they have put out a popularized book:

I am familiar with some of the original papers in top scientific journals—the book is of course much more readable, and the authors and editors have done a good job of presenting their work in lay language while preserving the broad strokes of their work.

The bottom line of their research is this: each of us carries a whole ecosystem of bacteria in our intestines, which help us digest and absorb food. The specific mix of bacteria varies between individuals, and hence so do our responses to different foods. While weight gain/loss is best seen as an outcome—one aspect of overall health—glycemic response, the changes in blood sugar levels after a meal (“postprandial glucose response”) are sufficiently rapid that they can be monitored in real time (e.g. with a continuous glucose monitor) and correlated with what the person ate (logged in a smartphone app). Doing this for thousands of people is a big-data project par excellence, and this is how computer scientist Segal teamed up with gastroenterologist Elinav.

But this isn’t where it ends. Gut bacteria populations can of course be obtained from stool samples, and subjected to analysis—another aspect of the massive big-data puzzle. Moreover, some of what they infer from the data can be checked in an animal model—for instance, certain gut bacteria can be administered to sterile mice and their weight gain (or lack thereof) in response to certain food mixtures tested on a much shorter time scale than would be possible in slow, relatively large, and long-lived mammals like us.

The duo brought different, complementary perspectives to the problem, not just scientifically but personally. Elinav always loved to take machines apart and see how they fit together (fittingly, he did his military service aboard a submarine), then became fascinated with living organisms. He ended up studying medicine, then specializing in internal medicine. During his residency, he was exposed to the human suffering caused by “metabolic syndrome” (the term given to the combination of severe obesity, adult-onset diabetes, fatty liver, hyperlipidemia, and the complications thereof). He realized that they spent all their time as doctors dealing with the consequences and complications rather than with the root cause.

Segal, on the other hand, was an avid long-distance runner in his spare time. He started experimenting with different nutritional approaches to improve his endurance as a runner, assisted in this pursuit by his wife, a clinical dietitian. As he dove deeper into this and observed diets of fellow runners, it became increasingly clear to him that there was no one-size-fits-all, and that recommendations that were held to be gospel truth (or Torah from Sinai, in our case) were, in fact, counterproductive for some. Why do some runners who eat dates before a run become energized and others exhausted? Who do some do best with carb-loading, and indeed thrive on high-carb diets, while others quickly pack on the pounds and suffer from low energy?

Segal was already involved in the computational study of the human genome at the time and then started reading about the emergent field of study of the microbiome. One thing led to another, a mutual acquaintance put Segal and Elinav in touch with each other, and together they embarked on the collaboration that eventually morphed into the personalized nutrition project.

One factor that facilitated their research was that rapid, reliable, and minimally-invasive blood glucose monitoring technology has become relatively inexpensive. And here some of their first surprises came. Anybody who has followed a Gary Taubes-type diet, or who is trying to manage diabetes, is aware of the ‘glycemic index’ (GI) of foods—the increase in blood sugar levels caused by eating a given amount of the food, compared to the same amount of pure glucose (for which GI=100 by definition). But how uniform are these values really?

Segal and Elinav found that the GI for some foods (e.g., bananas) differed very little between their test subjects (say, 60-65), while others (e.g., apples) were all over the place (40-90). Moreover, the variation was not random but correlated with the person.

One would expect glycemic response to go up more or less linearly with the amount of the food consumed was a given. They found that this is indeed true for smaller amounts, but at some point saturation sets in as the body manufactures more insulin, and the glucose response levels off. (This, of course, does not mean you can just eat ten times as much: the insulin will cause the excess energy to be stored as fat!)

More surprising, however, was that higher fat content in the meal on average caused a minor decrease in glycemic response. For a nontrivial number of their participants, eating toast with butter or olive oil actually did less glycemic harm than eating the toast on its own.

Now trying to keep blood sugar levels on a more even keel has two major benefits. In the short term, yo-yoing blood sugar levels lead to a reduction in energy, a feeling of exhaustion as the body pumps out insulin in response to a sugar spike and blood sugar dips. As for the long term: Segal and Elinav found across their sample that glycemic response after habitual meals is strongly correlated with BMI. Keeping blood sugar levels on a more even keel turns out to be a win-win on all counts.

And here’s the catch—”thanks” to our microbiome, glycemic response is highly individual. Segal himself ‘spikes’ after eating rice, while Elinav does not. One person spikes after ice cream, while another does not—and the same person who spikes after an evening snack of ice cream can safely have chocolate instead, go figure.

This addresses a seeming paradox. It’s not that diets don’t work—in fact, many do for some people, though long-term compliance can be an issue—it’s that there is no diet that will work for everyone, or even for most people.

So the next step, then, was to have a computer analyze the data for some of the participants in depth, and have it plan out a personalized diet that would keep blood sugar levels as steady as possible for that patient. Guess what? Yup, you guessed it.

Now some people might be discouraged by the idea of carrying around a blood sugar monitor for two weeks and carefully logging every meal (and physical activity). But once a large enough dataset has been established, and correlated to analyses of the gut flora composition in all the test persons, it becomes possible to predict glycemic responses to different foods with reasonable accuracy based on a bacterial population analysis of stool samples. A startup company named DayTwo is offering to do exactly that. [Full disclosure: I have no financial interest in DayTwo or in any of Drs. Segal and Elinav’s ventures.]

We are at the dawn of a major revolution in healthcare—a shift away from a paradigm of statistical averages to one of detailed monitoring of individual patients. Call it ‘personalized medicine’ or any other buzzword: it does seem poised to radically change healthcare and individual health outcomes for the better.

 

On dieting, weight, and reductionist fallacies

 

Sandra Aamodt, the former editor of Nature Neuroscience, presents a TED talk where she explains something counterintuitive: not only do most diets fail to achieve permanent weight loss, but in some cases the rebound actually overshoots, and the diet actually causes a weight gain in the long run.

As she describes it: the hypothalamus of the brain acts as a kind of ‘weight thermostat’ (that would be a barostat? :)) that tries to adjust body weight to within about 10-15 lb of a set weight by sending chemical signals that up- or down-regulate appetite, that speed up or slow down metabolism, etc. If weight drops “too” far below the set point, signals to increase food intake are sent out, and if no food intake ensues (because no food is available, or because the person is dieting), then metabolism is slowed down to reduce the base metabolic rate (i.e., the number of calories your body needs to keep basic functions going at rest). Unfortunately, the “set point” can be ratcheted up but not trivially ratcheted down.

People who think it is all about the pounds (or about the BMI) will find this a depressing message. But this is a classic example of the “reductionist fallacy”: weight or BMI are but. one metric of health among many. There are many others that matter, such as percentage muscle mass, blood sugar at rest, blood pressure, cholesterol, blood oxygen levels,… A person who is technically overweight (i.e., BMI between 25 and 30) but eats healthily, exercises at least 3 times a week, does not smoke, and only drinks in moderation actually has a better health prognosis than somebody who has an “ideal” weight (BMI around 20) but smokes and drinks heavily and never does any exercise.

To be sure, she shows that among people who do not have any of these four healthy habits, an obese person (BMI=30 or higher) has seven times the mortality risk of somebody with an ideal BMI=20.oo. However, for those who do observe all four healthy habits, the mortality risks with normal, overweight, and obese patient differ only by statistical uncertainty.

Does that mean that a morbidly obese person who cannot fit in an airplane seat does not need to go on a diet? Of course, it doesn’t — that is a straw man, and “set point” normally don’t go that high unless pushed there by unhealthy habits or regular binge eating.

But somebody who, well, has a naturally zaftig built is probably better off making a fixed habit of exercise, and to eat ‘smart’, than to go on some extreme low-carb diet. (Full disclosure: I do restrict my carbohydrate intake, but not all the way down to “ketogenic”.)

There is an additional factor here: in recent years we are increasingly aware of the role the microbiome (“gut bacteria”) plays in food absorption, and particularly in sugar absorption. For instance, in this very recent paper: http://dx.doi.org/10.1016/j.cmet.2017.05.002

ABSTRACT: Bread is consumed daily by billions of people, yet evidence regarding its clinical effects is contradicting. Here, we performed a randomized crossover trial of two 1-week-long dietary interventions comprising consumption of either traditionally made sourdough- leavened whole-grain bread or industrially made white bread. We found no significant differential effects of bread type on multiple clinical parameters. The gut microbiota composition remained person specific throughout this trial and was generally resilient to the intervention. We demonstrate statistically significant interpersonal variability in the glycemic response to different bread types, suggesting that the lack of phenotypic difference between the bread types stems from a person-specific effect. We further show that the type of bread that induces the lower glycemic response in each person can be predicted based solely on microbiome data prior to the intervention. Together, we present marked personalization in both bread metabolism and the gut microbiome, suggesting that understanding dietary effects requires integration of person-specific factors.

dnuhxwhx4aav-ec

We are only beginning to understand how human digestion, food absorption, metabolism, and the microbiome interact. Eventually, genome analysis combined with microbiomics will bring us into the personalized nutrition era.

 

UPDATE: from the same team, a 2014 paper showing that artificial sweeteners induce glucose intolerance by altering the microbiome.  NATURE’s editorial summary in lay language:

We have been using non-caloric artificial sweeteners for more than a century. Today the food industry is using them in ever-greater quantities in ‘diet’ foodstuffs and they are recommended for weight loss and for individuals with glucose intolerance and type 2 diabetes mellitus. Eran Elinav and colleagues show that consumption of the three most commonly used non-caloric artificial sweeteners saccharin, sucralose and aspartame directly induces a propensity for obesity and glucose intolerance in mice. These effects are mediated by changes in the composition and function of the intestinal microbiota; deleterious metabolic effects can be transferred to germ-free mice by faecal transplantation and can be abrogated by antibiotic treatment. The authors demonstrate that artificial sweeteners can induce dysbiosis and glucose intolerance in healthy human subjects, and suggest that it may be necessary to develop new nutritional strategies tailored to the individual and to variations in the gut microbiota.