That’s a totally fair question. I tend to write about whatever gets me fired up, or at least whatever grabs my attention to the point where I can’t get anything else done until I dig into the subject at hand. Even though I do not enjoy treating diabetes, I am still fascinated by the pathophysiology, especially when longstanding dogma gets challenged and things get controversial.
Recently, I was listening to an episode of Dr. Bret Scher’s Boundless Health Podcast, in which he interviewed Dr. Jeffry (not a typo) Gerber, a proponent of the low-carb lifestyle. Dr. Gerber spoke effusively about Dr. Joseph Kraft, the author of a book I’ll be referencing in this post. Apparently, Dr. Kraft collected glucose and insulin level data on over 14,000 patients over a 20 year period between 1972-1992, constructing what is probably the largest database of such data ever accumulated in one spot. So what did Dr. Kraft figure out with all this data?
He found that a downright scary-high percentage of the people who had normal glucose tolerance (as defined by normal blood sugars on a glucose tolerance test) had hyperinsulinemia (high insulin levels in the blood). An even larger percentage of people with impaired glucose tolerance (bad, but not bad enough to be considered diabetic) also had hyperinsulinemia . He then essentially concluded that almost everyone has diabetes – they just don’t know it yet.
Well that’s disheartening. How come I’ve never heard of this guy?
That’s the question I was asking myself: how is it possible that I went through an entire Endocrinology fellowship without hearing about this man’s work? Did I sleep through that lecture? So I decided to do some digging.
It soon became clear that the reason why I never heard about Dr. Kraft’s work is that, although he has a few publications about this subject in medical journals, a complete synthesis of his theories wasn’t published until 2008, in the form of his book: Diabetes Epidemic & You, subtitled Should Everyone Be Tested? Absolutely Not! Only Those Concerned About Their Future!
I know. The title is a snooze-fest, and the subtitle does not inspire a sense that the book will be well-organized or tight. Spoiler alert: his message has enough validity and importance for me to explore it in this post. However, the book itself is in desperate need of an editor, and Dr. Kraft makes some bold assertions based more on opinion than solid evidence.
What I intend to do in this post is very simple, but it’s going to take me about 4600 words to do it; please consider yourself warned. I will:
- Distill Dr. Kraft’s book down to a few easily digestible concepts by stripping away the complexity and unfocused writing that muddies his message.
- Explain why you should care about his work.
- Critique how some doctors are appropriating his data for use in ways that are either inappropriate or just plain unnecessary.
- Recommend a strategy for incorporating his conclusions into modern medical practice.
Hyperinsulinemia (HI) is bad.
This is Dr. Kraft’s primary message and, because he found that high insulin levels were incredibly prevalent – even in people with normal blood sugars – pretty much everyone reading this should care about it. So what’s the problem with having HI? I’ll cover that to some extent but, as regular Hormones Demystified readers know, this is not a medical-o-pedia site, so I recommend that you follow the links to references if my assumption of your facility with this subject is incorrect.
Catherine Crofts et al from Auckland University of Technology have written a couple of well-balanced papers reviewing the subject of HI*. In Hyperinsulinemia: A unifying theory of chronic disease? the authors do a great job differentiating between what we know about the potential negative effects of HI mechanistically versus epidemiologically.
Let me explain. There is plenty of good evidence linking HI to various pathologies (coronary disease, obesity, diabetes, etc.) from a mechanistic, how-would-this-work-in-theory, perspective. But, when we look for hard data proving that there is a causative effect between HI and an increased risk of things like heart attack, the data pool becomes cloudier. One reason why the data is murky is that it’s quite difficult to tease out the negative effects of hyperglycemia (high blood sugar), versus insulin resistance (tissues are not as responsive to insulin as they should be), versus HI. Moreover, most research over the years has focused on hyperglycemia as the cause of disease. While high blood sugars are toxic to the body, it is increasingly recognized that high insulin levels have an independent (though clearly related) toxicity.
Okay, so high blood sugars are bad and high insulin levels are probably bad. Where does insulin resistance (IR) fit into the equation? For years, the dogma has been that IR leads to HI and, ultimately, hyperglycemia/diabetes. Mechanistically, this makes sense. There are multiple things that contribute to IR, most notoriously obesity. As we gain weight, our tissues become resistant to insulin, which leads the pancreas to crank out way more insulin than it should. The extra insulin will overcome the resistance by binding to insulin receptors that would otherwise be sitting idle in the presence of lower (normal) insulin levels. Upon binding to insulin, the cells will then allow blood glucose inside, hence the blood sugar will be maintained in a normal range…until the pancreas can’t hack it anymore. By the time diabetes (or prediabetes) is diagnosed, insulin resistance and hyperinsulinemia will have been present for many years.
Where this all gets really interesting is when we start questioning whether the chicken or the egg came first. Based on available medical literature, it turns out to be pretty difficult to generalize whether IR precedes or follows HI. However, there are some very smart people starting to make the case that HI may be the root cause of IR, obesity, and diabetes – or at the very least, one of the initial steps in the cascade of events that lead to these bad outcomes.
Does it even matter whether HI or IR comes first?
This is the first of two times in this post that I will ask this question. I’m not going to answer it just yet, but I am going to lay some foundational stuff on you. HI is thought to precede hyperglycemia by up to 24 years. If HI is truly the earliest manifestation of a metabolic derangement that will ultimately progress to type 2 diabetes, then it might be helpful to know if it’s present, so we can treat it (more on that later). In our current paradigm, we don’t usually test for HI (more on that shortly), and we certainly don’t test for IR, since there isn’t a great way to do that outside of a research setting. What we do is monitor glucose, which is problematic, since that is usually the last thing to rise. By then, a fair amount of pancreatic beta cell (insulin-producing cells) function has already been lost.
The concept of HI preceding insulin resistance is not new.
This is where Dr. Kraft and his database come in. Remember that he started this work in 1972 and found that people of all ages demonstrated high insulin levels in response to a glucose load during his 100 gram oral glucose tolerance test. This appears to have laid the groundwork for his belief that HI is one of the earliest findings of a metabolic derangement that leads to all kinds of chronic diseases.
Before we guzzle the entire bottle of Kool Aid, we need to look at this critically. First, while Dr. Kraft collected data from over 14,000 glucose tolerance tests, he did not specify how many people referred for the test were healthy volunteers and how many were referred for a suspicion of diabetes. As best as I can tell, most were referred for a suspicion of having diabetes. Is it surprising, then, that more than 80% of the study population demonstrated HI, including nearly 75% of people with normal glucose tolerance?
It should also be noted that the average body mass index (BMI) of people with normal glucose tolerance but high insulin levels was in the overweight range (25-30). So, even though these people were not technically “obese,” their doctors probably recognized that they were overweight, and I suspect they had other markers that flagged them as metabolically unhealthy (high triglycerides, low HDL, high blood pressure, etc.). Unfortunately, none of that data was recorded.
So, what we have here is a study population that is heavily skewed toward the metabolically-challenged end of the spectrum. Where’s the control group? How are we supposed to know what constitutes normal insulin levels? For reasons that are unclear to me, Dr. Kraft used this skewed study population to determine the “normal” reference range for insulin levels. He analyzed people with normal glucose tolerance who had the “best” insulin response, and that became the definition of “normal.”
This would be an apt moment to point out that the normal reference range for any blood test is typically determined by studying people known to not have the disease in question. Would it make any sense for me to determine what a normal TSH should be by running TSHs on the next 100 people who get sent to me for suspicion of having hypothyroidism? Of course not.
Before I get completely bent out of shape about Dr. Kraft’s methods, I will concede that his study design probably makes some of his findings even more impactful. Because many – perhaps most – of the participants were referred for testing because their doctors suspected they had diabetes, we’ve already established that it isn’t surprising to see HI in most of the subjects. And, we know that Dr. Kraft used this study population with such a high prevalence of HI to determine “normal” insulin levels. So, if there is any bias here, it should be that his set point for “normal” is actually too high.
I actually hope that there isn’t an upward bias in the data. If the normal reference range for insulin was set too high, that means that even more than 80% of his study population had HI, which is extremely scary. That would essentially label almost every person who was tested as on their way to diabetes. Or would it?
There was no longitudinal followup in Dr. Kraft’s study.
I don’t mean to be hypercritical here, as following these patients for years was not part of his study design; it would have required a tremendous amount of resources to accomplish. However, with that delimitation, I do think that Dr. Kraft should have applied the brakes before declaring that all patients with HI had “diabetes in situ.” It’s not that I think we should blow off the possibility that many of us are already HI and on our way to developing diabetes. Diabetes is an epidemic and the numbers are getting worse. But we clearly have no idea how many of Dr. Kraft’s subjects went on to develop prediabetes or diabetes.
Are the numbers as bad as Dr. Kraft would suggest? Remember that his data was collected from 1972-1992. With more than 80% of the study population labeled HI, and knowing that HI may lead to diabetes over the next 20+ years, we would expect the current prevalence of diabetes to be close to 80%, no?
In 2015, the prevalence of diabetes (all types) in the United States was 9.4%. Of the 30.3 million people with diabetes, 23.1 million were diagnosed and 7.2 million were undiagnosed. If you know how they figure out the number of undiagnosed cases, please enlighten me in the Comments section. Moving on, what about prediabetes (impaired glucose tolerance and impaired fasting glucose)? Would the number of prediabetics bring us closer to that 80% prevalence of Dr. Kraft’s “diabetes in situ?” Well, 26% of Americans were prediabetic in 2015.
It appears we have a discrepancy that needs to be resolved. Dr. Kraft found that 80% of people had “diabetes in situ,” but 20-40 years later, “only” 35% of Americans have been labeled as diabetic or prediabetic. Yes, it’s bad, but not to the extreme that Dr. Kraft portrayed. Have Americans become significantly healthier over that time frame? Have we all adopted the low carb lifestyle and started high-intensity interval training? Not even close.
So how do we reconcile Dr. Kraft’s estimation of the prevalence of “diabetes in situ” with what we’re actually seeing? For those who have really been paying attention, I’ve spent the last few paragraphs building a straw man argument, just to illustrate a point:
Dr. Kraft’s study population was selected for likely having glucose metabolism problems; this was not a random cross-section of the population.
Therefore, the way to reconcile the discordance between Dr. Kraft’s estimate of how many people would develop diabetes and the actual prevalence of diabetes is to recognize that he found high insulin levels in people who were suspected of having high insulin levels. That makes his findings much less dramatic. But it doesn’t mean that we are out of the woods with this whole HI issue. If HI really is one of the first measurable abnormalities in the metabolically unhealthy, the question is: should we be doing glucose tolerance tests (GTT) with insulin levels on almost everybody?
Before you call your doctor and demand this testing, it’s important to go back to the basics (know your assay) to understand what this testing can and cannot tell us. First, we no longer use glucose tolerance tests (GTT) routinely for the diagnosis of (pre)diabetes (outside of gestational diabetes). Why? They have poor reproducibility and suboptimal reliability. There also isn’t much we can get from a GTT that we can’t get from a hemoglobin A1c and some fasting and post-meal blood sugar measurements.
In the case of trying to diagnose HI early, though, we’re not really expecting to make a diagnosis of diabetes with the GTT. What we are trying to determine is whether someone has inappropriately high insulin levels in response to the sugar shake, even if the blood sugars stay normal. That begs the questions:
- How much faith can we put in the insulin levels that we measure during a GTT?
- Do we have a firm understanding of what is “normal” for each time point during the GTT?
- Are the cutoffs suggested by Dr. Kraft applicable to insulin levels tested using different assays than what he used, or different glucose shakes?
- Does the finding of HI portend a high risk of developing diabetes/other chronic disease and, if so, how are we going to use that information to decrease the risk?
Let’s take these one at a time.
- How much faith can we put in the insulin levels that we measure during a GTT? Not too much. Insulin secretion is pulsatile, meaning that there can be significant changes in short periods of time. Levels can be so variable that it has been recommended to take the mean of three fasting insulin samples at five minute intervals, but that never happens in clinical practice. Single fasting insulin samples can have a coefficient of variation of 25-50%!
Dr. Kraft does note that there is too much overlap in fasting insulin levels between people of normal and impaired glucose tolerance. He found that there was much more separation between groups in the post-glucose load insulin levels; therefore, he does not recommend the use of fasting levels in isolation. However, because the variability in insulin levels likely also applies to the post-shake levels, I am not sure we can put a lot of faith into the numbers we get from a GTT.
- Do we have a firm understanding of what is “normal” for each time point during the GTT? There isn’t extensive evidence answering this, outside of Dr. Kraft’s data and that of Hayashi et al. This provides a natural segue into the next question:
- Are the cutoffs suggested by Dr. Kraft applicable to insulin levels tested using different assays than what he used, or different glucose shakes? This is where I really want you to pause and reflect. When I started digging into Dr. Kraft and the issue of HI, I found people on the internet quoting numbers and cutoffs from his database like they were handed down from Moses himself on stone tablets. If you are going to apply Dr. Kraft’s numbers to yourself or your patients, you need to use the exact same protocol and the same insulin assay. Even then, the whole enterprise is problematic because of the known variability in insulin levels from moment to moment.
If you’re drinking a 75-gram glucose shake for a 2-hour GTT with 2-3 data points, as opposed to ingesting Dr. Kraft’s 100-gram glucose shake for a 3-5 hour GTT, you need to be very cautious about plugging your insulin values into his reference ranges. Also, when comparing your values to a reference range that is not the one your laboratory provided, you must be aware that insulin levels using different assays may not be comparable. For example, assays testing serum insulin will get higher values than assays testing plasma insulin. I know I’m a broken record, but this means that you need to know your assay.
In my opinion, if you are going to do a GTT with insulin levels, it would be best to talk to your laboratory and see what kind of reference range they are using and how they (or the manufacturer of the assay) validated it. In my experience, the reference range for insulin is usually quite broad, so expect that your values will only be helpful if they are very low or very high. Anything in the middle will be difficult to interpret.
- Does the finding of HI portend a high risk of developing diabetes/other chronic disease and, if so, how are we going to use that information to decrease the risk? If you truly have high insulin levels, I think it is fair to say that it’s not a good thing. As stated earlier, there is a growing body of evidence that HI plays a role in chronic diseases as diverse as diabetes, cardiovascular disease, Alzheimer’s, acne, gout, erectile dysfunction, polycystic ovarian syndrome, vertigo, and tinnitus. As far as assessing your absolute risk of developing any of these in the near or distant future, that’s a much harder question to answer. But I think a fair assumption would be that the risk is elevated. If the risk is elevated, what are we going to do? Before answering that, I want to back up a few steps and point out:
The application of common sense often obviates the need for testing.
I think I coined that phrase, but the concept is not new. Think about the typical patient being evaluated for a suspicion of glucose intolerance. This person usually eats a suboptimal diet and has a BMI that places him in the overweight or obese category. Let’s say that we test his blood sugar and hemoglobin A1c, and they are both normal. Do we offer congratulations and kick him out of the office? Recalling Dr. Kraft’s data in similar patients, nearly 75% of people with normal glucose tolerance had evidence of HI when tested. That’s a pretty high number – so high, in fact, that it makes you wonder whether going through the hassle and expense of a GTT with insulin levels is worth it. Wouldn’t a reasonable alternative be to assume that this person either has HI/IR now, or is at risk of developing HI/IR and then diabetes if he continues his lifestyle unchecked?
Are we really going to recommend a different treatment for this guy if we check insulin levels, or are we going to counsel him to lose weight, avoid simple sugars and simple carbohydrates, and exercise – regardless? Let’s explore it: if we check insulin levels and they come back high, he’s getting the advice I mentioned. If we check insulin levels and they come back normal, we tell him that things look okay right now, but they probably won’t continue looking okay if he doesn’t make some changes – the ones I’ve already mentioned. If we don’t check insulin levels at all, you guessed it – he gets the same counseling.
I do understand that numbers on a page can be a powerful motivator for certain people, so I admit that it might be easier to get them to buy into my recommendations if I prove to them that they have HI. However, in my experience, the people who are most likely to make changes are also the people who are seeking out their physician’s advice – not being told what to do by their physician. That may seem like a subtle distinction – it isn’t. The folks who suspect they are on the wrong track are usually right, and they are much more motivated to make changes than the people I try to convince that their lifestyle needs a lot of work. Do my already-receptive patients need me to show them high insulin levels on a piece of paper to realize that they need to lose weight? Probably not. Would it be extra-motivating to see high insulin levels that turn normal after intervention? Perhaps. Unfortunately, we also know – from Dr. Kraft’s data – that low carbohydrate diets lead to low-normal insulin levels on GTTs. So if I was going to use a GTT with insulin levels, I would use it for initial diagnostic purposes, but I would not use it to gauge the success of treatment. Instead, I would follow glucose and hemoglobin A1c, recognizing that even a “normal” A1c is probably going to go down as someone takes strides to reduce HI through better diet and weight loss.
Even patients seeking out health care optimization don’t need GTT/insulin response testing.
I have a good friend – let’s call him Jon – who is an optimizer. Jon is passionate about decreasing his risk of chronic disease as low as it can go. He is also fascinated by the concept of longevity or, as it is increasingly referred to – health span. He has meticulously gone about tweaking his exercise routines to the point where his body looks more ripped than a Netter drawing. Despite a body fat percentage near or in the single digits, Jon’s A1c was still running in the upper 5’s, putting him in the “at risk for diabetes” category.
Though Jon did not have a GTT with insulin levels at this point, I’m sure he would have jumped at the chance to do it if offered by his doctor. But did he need it to push him toward further health optimization? No. He researched the heck out of Intermittent Fasting (IF) and decided to embark on an experiment doing Time Restricted Feeding (TRF) – I will dive into this more in a future post. For now, my point is that he has dropped his A1c to the low 5’s, his cholesterol is down, and he shaved the few remaining adipose cells off his midsection. Oh, and he feels amazing.
Who might be a good candidate for GTT/insulin response testing?
If you are not overweight, you think you’re pretty good at this whole health thing, your A1c and glucose are clearly within normal limits, and you’re looking to see if your metabolic status is as good as you think it is, then you might be a good candidate for testing. Let’s explore it: if I test you and your glucose levels are normal but the insulin levels are high, we’ll talk about decreasing carbohydrates in your diet and perhaps we’ll look at things like the Ketogenic Diet or Intermittent Fasting. If I test you and you’re normal, then I’ll pat you on the back and kick you out of my office with a smile.
In the above situation, you can see that the recommendations will be different based on the results of testing, which makes testing a more reasonable proposition.
While there are probably other conditions under which testing would be reasonable, there is only one I can think of where I regularly do this testing. In lean women with suspected polycystic ovarian syndrome (PCOS), I like to do a GTT with insulin response to get a sense of whether their PCOS has an underlying pathophysiology involving IR/HI. If they have high insulin levels, I am more likely to consider metformin therapy to see if we can normalize periods using an insulin sensitizer. Disclaimer: I am operating outside the guidelines on this one, as oral contraceptive pills are considered first-line therapy for androgen suppression/normalization of periods. It is my opinion, though, that the added bonus of decreasing the risk of progression to type 2 diabetes with metformin is a valid reason for using it in HI/IR patients.
Even though I use the GTT/insulin response test in some women with PCOS, I do not use it in all of them. If I see an overweight/obese woman with PCOS, my suspicion of HI/IR is automatically higher, based on what we know about the pathophysiology of the disease. Additionally, these women often have a darkening/roughening of skin on the back of the neck or elsewhere (acanthosis nigricans) which is caused by high insulin levels. This physical exam finding obviates the need for the GTT/insulin response – the recommendations would not change based on testing.
Wrapping up, what comes first, HI or IR? Is this the million dollar question?
Emerging data suggests that HI may precede IR. But as I said earlier, these two concepts are so intertwined that it is very difficult to separate them. Even if HI leads to obesity and IR, we know that obesity and IR also lead to HI. The most likely answer here is that there is a bidirectional relationship. Should we care which comes first?
In a word, yes. While we are accustomed to seeing overweight/obese people with IR and counseling them about their risk of progressing to type 2 diabetes, we are not used to questioning whether normal-weight people are metabolically healthy. If it is increasingly recognized that normal weight people can have HI, and HI leads to IR and obesity, then it creates a greater sense of urgency to counsel everyone to eat and exercise better.
Yes, we should already be counseling everyone in this fashion, but the reality is that doctors tend to be firefighters, focusing our attention on putting out the blazes that are immediately threatening the structure (body). If HI is proven to be a much bigger problem than we’ve realized, striking much earlier than previously thought, that could change the way doctors approach their “healthy” patients. I’m excited to see where this will all sort out, especially because we don’t know nearly as much as we need to about what causes HI, so we also don’t know nearly enough about how to treat it. The best advice we can give right now is: control your weight, avoid excessive carbs, and exercise. Hmmm…maybe medicine hasn’t changed much over the last 40 years!
*Remember there was an asterisk at the end of a sentence way back when you started reading this monstrous piece, like an hour ago? If you’re still interested in reading more about Dr. Kraft’s work, but you don’t want to slog through his book like I did, then check out: Identifying hyperinsulinemia in the absence of impaired glucose tolerance: An examination of the Kraft database. Crofts, C et al. Diabetes Research and Clinical Practice, 118 (2016) 50-57. The authors make Dr. Kraft’s material way more accessible.
Are you a doctor who is using GTTs and insulin levels? Why or why not? Do you agree that most lifestyle recommendations would not change based on the results of the testing? Are you a health enthusiast/optimizer who has been found to be HI/IR despite having normal blood sugars? How do you manage your condition? Comment below!
By reading this site and interacting with me in the Comments, you agree to abide by my Disclaimer.