The public life of physician Peter Attia began in earnest with a 2013 TEDMED talk. In it, citing two personal anecdotes and some theoretical speculation about insulin and fat cells, Dr. Attia argued that low-carbohydrate diets might be useful in treating obesity. Dr. Attia then made a wild leap: low-carbohydrate diets could in fact solve the obesity epidemic and improve the health of millions of Americans. But there was only one catch: he needed the money to prove this. That is where we, the audience, come in.
Millions of views later (now nearly 10 million in total), the result was the cofounding of the nonprofit 501(c)(3) medical research organization Nutrition Science Initiative (or NuSI for short) with New York Times bestselling writer and low-carb enthusiast Gary Taubes–but also newfound fame for Attia as a TEDMED speaker, spokesman of low-carb diets, and director of NuSI’s fundraising efforts, entailing months of intense networking with some of the world’s richest and most influential philanthropists and donors. NuSI would go largely down in flames–sponsoring work by scientists Christopher Gardner (Stanford) and Kevin Hall (NIH) falsifying (here and here) the theories of Taubes and Attia in a series of now highly cited long- and short-term studies.
Gary Taubes, Christopher Gardner, Kevin Hall (left to right)
After years of promoting low-carbohydrate diets in blog posts and alternative health conferences, Attia would quietly slip away from NuSI while maintaining public silence about what had occurred in the organization that had falsified what he was once publicly largely identified with. The lesson, we might think, is that we should be a bit more circumspect in promoting our hypotheses to the public as if they were true. However, Attia would then continue on as a promoter of a large number of medical interventions similarly unsupported by evidence. He would also leverage the large network that he had cultivated during this period to found a concierge medical practice that promoted itself as the cutting edge of medical knowledge, as well as launch a podcast that gives guests a platform to promote a dizzying array of trendy but similarly unsupported medical theories.
Some of Attia’s podcast guests are noted sources of health misinformation, such as Jason Fung and Mark Hyman. But others are scientists who, in efforts to self-promote, overhype their research findings and elide an appropriate discussion about the role of cell culture or animal studies in the scientific process, frequently giving listeners the false impression that isolated mechanisms and findings in disease model systems are a reliable guide to human recommendations or interventions. (They are in fact notoriously unreliable.) This angle puts Peter Attia in the same orbit as other popular online influencers such as Ben Greenfield, Dave Asprey, Rhonda Patrick, and even David Sinclair, each of whom have made a public brand out of speaking beyond the human clinical trial evidence and therefore beyond evidence-based medicine. Many of these figures are a part of the circuit of wellness guests that once frequented the Joe Rogan Experience and were made famous by their appearances there. (Attia himself was a guest on JRE in 2018.)
However, what distinguishes Attia is his credentials. According to his website, Attia is “a Stanford/Johns Hopkins/NIH-trained physician focusing on the applied science of longevity, the extension of human life and well-being.” These credentials have given Attia unparalleled clout (along with David Sinclair) in this niche of influencers. Correspondingly, Attia has nearly 150,000 followers on Twitter and 177,000 on Instagram. This clout has caused Attia’s health communication to have an enormous influence across social media, so that many less influential influencers merely copy Attia’s messaging and replicate it in a storm of wellness posts and videos.
Continuous glucose monitoring
One of the many interventions that Attia has personally been consistently enthusiastic about (others have included metformin and rapamycin for longevity and fasting for “autophagy”) is continuous glucose monitoring, or CGM. CGM is the real-time sampling and reporting of blood glucose, via an implantable device on the arm or torso. CGM was developed for use for people with diabetes to help provide tighter control over blood glucose levels without needing to take periodic finger sticks, a cumbersome process that reduces quality of life. CGM correspondingly improves the quality of life of people with diabetes and has been rightly celebrated as a breakthrough in diabetes care.
But Attia has been perhaps the most important proponent of CGM for non-diabetics, for which there is scant evidence. In part as a result of Attia’s advocacy, dozens of companies now specialize in promoting CGM for “metabolic health”. And, under his influence, CGM continues to make waves in the wellness world.
Earlier this week, cardiologist Danielle Belardo suggested that, in fact, there is no evidence to support this:
This tweet and posts that followed caused CGM as a topic to go viral in the health, medicine, and wellness space on social media. Day after day, Belardo and supporters tweeted and posted, pointing out that CGM for non-diabetics is supported by negligible evidence and therefore should not be recommended to people without diabetes.
Until Attia responded, yesterday, on his website:
The remainder of this post will consist of a debunking of the claims made in this article.
I am qualified to do this for the following reasons:
- I have used CGM extensively as a non-diabetic and correspondingly have personal experience;
- I am familiar with the literature cited, have considered it at great length, and have been communicating critically about the use of CGM for non-diabetics for more than a year (and warning people that it is a waste);
- I have followed Peter Attia’s career for nearly a decade, have listened to almost all of his >100 podcast episodes, was once an admirer, and am deeply familiar with his arguments about CGM;
- I have medical training and have nearly completed a PhD;
- Peter Attia once told me he respected me and offered me a job as a fact-checker (which I declined).
I hope that #5 especially qualifies me to dissect and debunk his article. After all, if Attia considered me competent enough to fact-check his content for pay, then certainly I would like to think that I can also competent do it for free.
And now, I will.
Attia’s post consists of 15 paragraphs. We will go through each, one by one, showing why they are wrong, why Peter should not be recommending CGM for non-diabetics, and should not be using it for this purpose in his medical practice.
I will conclude by discussing Attia’s motivations, as well as those of his supporter David Sinclair, and as an MD/PhD student myself, the nature of translational science, medical evidence, and medical recommendations.
A few things should be noted.
First, the reason that this post was written was not “a perspective in JAMA.” It was the embrolgio on Twitter and Instagram that has recently exploded about CGM that caused Attia and his staff to write this post.
For good reason. Attia’s concierge medical service charges $160K/patient/year, roughly the salary of an average physician per patient per year. That is, a doctor could have just one of Peter Attia’s patients in a single year, and make their entire salary from just that patient. And CGM is an important part of how he pitches his practice to these ultra-rich patients. Attia needs a suite of interventions that go beyond the medical standard of care in order to justify such exorbitant prices. Otherwise, why not just get a normal doctor? If Attia is wrong about CGM, and if that becomes the popular wisdom, then he loses revenue, and potentially massive revenue, on the order of millions of dollars.
(As we shall see later, Attia is wrong about most things that he proposes beyond the evidence base, as a matter of empirical fact, and nothing can justify the exorbitant prices he charges his patients.)
Which leads us to the second point. Notice the part that I highlighted. “I want to tell you why I not only disagree with these assertions, but also why this type of thinking may be dangerous to the health of hundreds of millions of Americans.” Not having CGM is dangerous to the health of hundreds of millions of Americans? As in, at least 200 million Americans? Here Attia goes all-in: almost everyone should be using CGM, and it is dangerous to suggest otherwise.
Let us see if the rest of his article can remotely support that claim.
This paragraph boils down to this:
How can we know for sure whether someone has normal glucose responses for absolute certain if they do not use CGM?
The answer is we don’t.
But that prompts another question: why do we want to know whether glucose responses are normal?
This will be a major theme of this article.
These are just assertions without evidence.
In paragraph 4, we start to see that Attia cannot even interpret the medical evidence properly. Excursions that reach 140 mg/dL are not diagnostic of prediabetes. Blood glucose readings that occur two hours after a glucose tolerance test that show a reading of 140 mg/dL are diagnostic of prediabetes.
That notwithstanding, all this study suggests is that glycemic dysregulation is not fully captured by traditional metrics. This does suggest that classification of patients’ glycemic excursions by CGM might provide more information than traditional metrics–or it might not. This would need to confirmed by correlating these glycemic variability readings by outcomes, which the study cited does not do. And even then, this would only justify using CGM diagnostically, for 2-4 weeks, as used in the study. CGM used in this way would identify an underlying biological type–but even then, even this would not show that reducing blood glucose excursions as measured by CGM would necessarily make a person healthier or improve outcomes. It might after all be the case that glycemic variability is only indicative of poor health, not that fluctuations in glucose are themselves the cause of this poor health.
This will become important later. For the time being, it is important to emphasize that the only thing that this paper shows is that CGM is measuring something not captured by traditional tests, and that this might (or might not) be clinically useful. Going from this point to suggesting that CGM should be used even short-term to measure something useful, much less used long-term to modify the diet to reduce glucose excursions to improve outcomes, is an outrageous overreach of the evidence cited.
Paragraphs 5 and 6
These paragraphs add no new information. It does not matter how high the glucose levels swing (especially if these are very short-lived swings), if this swinging has not been correlated with worse outcomes.
And even if they were correlated with worse outcomes, the question still remains whether the glucose swings themselves cause harm or some aspect of the underlying biology causes the harm and glucose is just a marker. But these data do not even get to the point of this question, because there were no outcome data in this paper! We are several massive assumptions from being able to draw the conclusion that Attia seems to want us to draw.
More non-arguments! Attia wants his patients never to go above 140 mg/dL–why?
- What outcome data are there to show that going above 140 mg/dL in a normal glucose excursion is harmful?
- And even if there were such data, what data are there that this harm is caused by the glucose itself?
Until these questions are answered, there can be no question: there is no evidence to support the use of CGM to modify diet to achieve glucose readings below 140 mg/dL.
Furthermore, it is possible that chasing such readings might in fact compromise the healthfulness of diet or lifestyle in other ways. Let us take the classic example. What if one finds that blueberries increase blood glucose beyond 140 mg/dL, so one decides to replace them with bacon. Better CGM readings right? But a worse diet.
There is a lot going on here. Let us go through each study, each line, one by one.
Attia’s cardiovascular disease reference– I will allow readers to read the concluding paragraph themselves.
Yes, as the paragraph says: “the current net balance of attained evidence is not favorable that we should care [about postprandial glucose for cardiovascular disease].”
This is because there are no trials demonstrating that reducing blood glucose improves superior cardiovascular outcomes. And this was in patients with frank impaired glucose tolerance and type 2 diabetes.
Blood glucose variability might associate with worse outcomes according to epidemiology. But reducing this variability so far has shown no benefit, i.e. this variability might not be causal or clinically significantly causal. This variability was also not measured by CGM: it was measured using traditional metrics, such as oral glucose tolerance test (OGTT).
And certainly no study has tried to reduce blood glucose below a 140 mg/dL cutoff on CGM, a completely arbitrary value.
In other words, Attia’s use of this review was at odds with the conclusions that the authors themselves gave… And that’s even if this study had anything to do with CGM, which it did not.
Attia’s Alzheimer’s disease reference– Here is where things get cringe. While Attia wrote “Prospective studies show that higher glucose variability in nondiabetics is associated with an increased risk of… Alzheimer’s disease…”, in fact, this study did not look at glucose variability but hyperinsulinemia.
This is a gross mis-citation. There is nothing else to say here. Onto the next reference.
Attia’s frailty reference– Attia and his research team is really screwing up. Like the last reference, this one has nothing at all to do with glucose variability but with HOMA-IR, a traditional metric of metabolic syndrome.
Attia’s cardiovascular death reference– You guessed it, didn’t you? Yep, this study looked at fasting insulin and HOMA-IR, traditional metrics of glucose metabolism that have nothing to do with CGM, glycemic variability, lifestyle modification, glucose excursions, or 140 mg/dL.
Attia’s cancer death reference– Nothing to do with glycemic variability or CGM. But, rather hyperinsulinemia. Cringe.
Attia’s death from any cause reference– Another mis-citation that has nothing to do with glucose variability.
Quite disappointing! Attia promises us bread but gives us rocks. At best. And expects us not to notice.
Let us continue.
Let us focus now on the six studies highlighted.
- OGTT and cancer;
- OGTT and cancer;
- HOMA-IR and frailty;
- HOMA-IR and frailty (same study as #3 linked twice);
- 2hr post-lunch and cardiovascular death;
- OGTT and mortality.
Yet in all of these cases:
- None use CGM;
- The studies that use OGTT do not establish that glucose variability above 140 mg/dL is important;
- None show that the high glucose itself is causal or that reducing it via diet will improve outcomes.
Indeed, that is why Attia moves to the next section, mechanistic plausibility:
He cites four studies. These say what he says they say. For a change. Postprandial hyperglycemia does indeed cause endothelial dysfunction. But so what?
Because postprandial lipemia, caused by the ingestion of fat, also causes endothelial dysfunction. This has also been hypothesized to cause atherosclerosis. That is to say, swapping carbohydrates for fat, to reduce glycemia to avoid atherosclerosis, might induce postprandial lipemia, which might cause atherosclerosis.
Indeed, postprandial triglyceridemia is associated with cardiovascular disease risk, just like postprandial hyperglycemia.
And Kevin Hall’s recent metabolic ward study showed the following: a clear trade-off between postprandial triglycerides and postprandial glucose, with triglycerides higher during a high-fat diet and glucose higher during a low-fat diet, naturally:
And, emphatically, Dr. Hall’s study did indeed show a decrease in glucose with carbohydrate restriction.
So which is it? Do we want to use CGM to reduce our carbohydrate intake to reduce our blood glucose to reduce our heart disease risk, thereby increasing our fats to increase our triglycerides to increase our heart disease risk? Or do we want to not use CGM and do the opposite?
Do you see the problem? If we focus on a single factor that we propose to be causal, and ignore other factors thought to be causal, we end up doing things that may actually reduce one risk factor while increasing others.
What’s more, we do not know whether either factor is really causal, or to what degree. Maybe neither factor is causal and both indicate an underlying state of poor health, and minimizing either our CGM or some hypothetical continuous triglyceride monitor (CTM) gets us precisely no benefit whatsoever, because the real dysfunction is in some aspect of biology that causes both problems.
We simply do not know. Unless we have a randomized controlled trial.
Attia goes on, citing a study showing that “higher postprandial glucose levels are also associated with higher carotid intima-media thickness”.
Unfortunately for Attia’s hypothesis, precisely the same association holds for postprandial triglycerides:
All the previous discussion holds. Replacing carbohydrate with fat to minimize CGM would be trading one risk factor for another. And. We. Still. Do. Not. Know. What. Is. Causal.
Attia then goes on to discuss mouse studies:
But the same criticisms hold. And since the references are not provided, I cannot easily look at the methodology. Do the animals also show a normal body weight? Or do these drugs mimic calorie restriction? Would a drug that caused decreased triglycerides cause the same phenotype?
Without answering these questions, we are back to selective citation city. Which is where you would want to reside if you are pushing an undemonstrated intervention to patients paying $160K/year. Which brings us to the concluding sentence of the paragraph. I quote:
I wouldn’t be surprised if there’s a quote out there attributed to Einstein that says confirmation bias is one of the most powerful forces in the universe. When it comes to observational epidemiology, there’s always a selective interpretation of the evidence.
You don’t say!
This paragraph causes ghostly echoes of Attia’s 2013 TEDMED talk. “If only Americans used keto/CGM, they would not have obesity/diabetes.” The same kind of messianic, unsupported claim, characterized by the same kind of naive oversimplification. There is nothing here but grandiose promises unsupported by evidence. Except in this case, Attia is not trying to have a trial conducted–he is only trying to sell something. Which is even worse.
How precisely is CGM supposed to prevent diabetes? Diabetes is caused by excess weight on a genetically susceptible background. Yet weight creeping upward, which everyone knows causes poor health, has never been a strong enough “biomarker” for most people to act on to reverse a downward slide into poor health. As a result, nearly 4/5 Americans are overweight or obese. Why should CGM be able to do what the weight scale cannot?
We can speculate all day long. Where are the data? We will return to this. But, where are the data???
“To recap my position and interpretation of the data available (more of which you can find in the AMA 24 show notes)…”
What data? What position? The data simply do not remotely show what is being claimed: that CGM (like low-carb eight years before) will turn the boat around. And there is every reason to believe that this is pure hype.
Let us look at the final sentence: “If he can maintain this way of living in the long-run, it’s likely to translate into an improvement in healthspan and reduce his risk of glucose impairment.”
First, there is absolutely no evidence provided that this is true, that this will result in an improved long-term outcome for this man. Second, there is absolutely no evidence provided that he could not have made all of these changes without CGM, nor even that he needed CGM to know to make these changes. Nor that these changes are sustainaable. There is no evidence for any of it. But it makes for a nice story.
This paragraph amounts to saying that Attia believes he does not need to test his hypotheses. We know clearly however from the randomized controlled trial evidence that only a small minority of highly plausible hypotheses prove successful in randomized controlled trials. This is a plausible hypothesis, requiring a test from a randomized controlled trial. Yet according to Attia, we are made to believe that it could entirely stave off the epidemic of poor health in America, without a randomized controlled trial.
I wrote above that “As we shall see later, Attia is wrong about most things that he proposes beyond the evidence base, as a matter of empirical fact, and nothing can justify the exorbitant prices he charges his patients.” And here is the payoff. We empirically know that new plausible hypotheses are overwhelmingly likely to prove disappointing in randomized controlled trial. (See immediately above.) This is why it is a very bad idea to place your bets with someone like Peter Attia: without an RCT showing that you are right, you are likely to be wrong. This is why Attia is wrong about most things that he hypes without a randomized controlled trial. And this is why he was wrong about low-carb and the carbohydrate-insulin model and is likely to be wrong about CGM, rapamycin, metformin, fasting and autophagy, and it goes on and on. He is wrong about most of these things because that is how most plausible hypotheses work: they are usually wrong. And a physician building a “cutting-edge” medical practice for $160K/patient/year based on plausible hypotheses is for this reason going to be giving out a lot of ineffective and/or harmful information or treatments.
This is also why I would not want to be one of Attia’s patients: I would be possibly receive substandard care. There is a word for the kind of medicine Attia practices: VIP medicine. And it is often lamented by physicians writing about the subject, because it risks leading to worse outcomes. In this case, that is because much of Attia’s medical practice is often based on untested, plausible hypotheses like this one about CGM, which he promotes to hundreds of thousands if not millions of people. Which is terrifying and incredibly unethical.
It’s great. Attia is confident about a lot of things. And wrong.
“It’s simultaneously a behavioral and analytical tool that can track and uncover strategies and tactics which can actually save an enormous amount of time and money by preventing bad outcomes in the future.”
Evidence? No? Zero? Absolutely zero?
You mean, there is absolutely zero evidence to justify this staple of patient care used to treat patients for $160K/year?
So, lots of words but no substance, no evidence, pure conjecture and (probably) false promises. Not even a promise to do a randomized controlled trial with the millions of dollars made by Attia Medical LLC each year, while promoting these unsupported hypotheses to hundreds of thousands of people on the Internet.
Well, then. I will therefore move to the summary.
My summary of arguments against use of CGM in healthy nondiabetics are as follows:
- There is no evidence that the unique glucotype characterized by CGM differentially predicts outcomes compared to traditional metrics of glucose dysregulation;
- There is only weak evidence that glucose dysregulation as measured by OGTT, HbA1c, etc. is directly causal in disease in nondiabetics;
- There have been no RCTs demonstrating that better glycemic control in nondiabetics leads to better outcomes, much less using CGM;
- There have not been any long-term studies using CGM in nondiabetics whatsoever;
- Some of the interventions to reduce CGM readings (say, carbohydrate restriction) also would increase postprandial triglycerides, also associated with worse outcomes;
- There is therefore an unclear tradeoff between hyperglycemia and hypertriglyceridemia, and that is assuming that they are each causal for bad outcomes in healthy nondiabetics;
- The cutoffs for unacceptable glucose excursions by Peter Attia are not based on any evidence whatsoever and are arbitrary;
Now for a few things I didn’t discuss in this article but will add:
8. It is probably the case that the metabolic phenotype associated with both postprandial hypertriglyceridemia and hyperglycemia is the very same, is the same metabolic phenotype that underlies fatty liver and type 2 diabetes, and can be improved with weight loss, exercise, and muscle building;
9. We should therefore improve this metabolic phenotype by staying lean, exercising, and maintaining muscle mass;
10. It is unclear whether the multifactorial characteristics of the healthfulness of a food can be captured by CGM, or whether people simply need good nutrition and lifestyle education, which would capture those things more effectively than CGM could. In fact, CGM may undermine good nutrition literacy by arbitrarily favoring (by historical accident of having a CGM rather than a continuous triglyceride monitor) unhealthy high-fat foods over healthy high-carbohydrate foods;
11. It is possible that CGM could provide a real-time biomarker that could increase adherence to lifestyle interventions. But this needs to be demonstrated with a randomized controlled trial and not simply assumed and promoted to recruit patients for a multi-million-dollar luxury medical business, much less spread potential medical misinformation to credulous laypeople.
The article by Peter Attia was promoted with particular zeal by David Sinclair, a Harvard professor and fellow biohacker.
Like Attia, Sinclair has his own checkered history of promoting wildly unsupported hypotheses. In 2008, he sold his company Sirtris, which was developing resveratrol, for $720 million to Glaxo-Smith-Kline. This made Sinclair rich, and rightly so, if his claims were borne out: “[Resveratrol] as close to a miraculous molecule as you can find…. One hundred years from now, people will maybe be taking these molecules on a daily basis to prevent heart disease, stroke, and cancer.” Unfortunately, they were not. Resveratrol didn’t pan out and its effects in humans were nowhere remotely as positive as once promised by Sinclair. Within a few years, to great disappointment, Glaxo-Smith-Kline shelved Sirtris and resveratrol.
This has not stopped Sinclair from continuing to make wild claims about aging. He claims to be growing younger and claims that just three supplements he takes have made him “fitter and stronger than he was 20 years ago”. He has moved onto new compounds besides resveratrol, promoting them with the same zeal, despite the failure of the former, as if self-awareness eludes him completely. And perhaps it does.
Unsurprisingly Peter Attia and David Sinclair have become good friends. So when doctors across social media began piling on Attia over this past week, Sinclair responded by waxing poetic about the “best article @PeterAttiaMD has ever written”. Despite this article being, in fact, horrible. But perhaps the most ridiculous part was the last bit: the appalling, anti-doctor “let patients get sick before we act” rhetoric. Because, yes, we want patients to get sick before we act. I mean of course.
I should remind Sinclair therefore that while he and his friend Peter Attia have been busy making grand promises whose only beneficiaries have been their bank accounts, physicians have been running the following trials:
These are all trials conducted overwhelmingly by physicians–who want to see their patients do better. Doctors around the world look forward to the day when Sinclair and Attia will run their own such trial–and do something more than provide empty and grandiose promises and criticize doctors who practice according to the evidence.