Are Carbs the Root of All Evil? Hall et al make a Call.
Is this true? Time to find out
More and more of us are suffering from the weight gain, and other problems caused by the mismatch between our mode of living, and that for which we evolved. And yet there is fierce controversy about how best to tackle this problem when it comes to diet. In one camp we have an enthusiastic movement promoting a 'low-carb' diet, with increasing interest in dropping the carbohydrate intake low enough to cause ketosis. The main premise on which this approach is based is that obesity (and diseases like type-2 diabetes) are caused by excessive intake of carbohydrates, which causes insulin resistance and overconsumption of food. This group are - at least from the viewpoint of this observer - coming close to dominating the conversation in the UK. It is also noticeably an approach that favours the consumption of animal foods and seems to hark back to our huntergatherer past. On the other hand we have an longstanding alliance between those promoting low fat intakes, and those who focus on promoting whole plant foods, sometimes consumed exclusively, and this approach seems to be holding a bit more ground in the US. This approach is popular with people who are happy to reduce or omit animal products, and look back to our long primate ancestry for guidance on long-term health. Amongst the arguments supporting this approach is the idea that the high caloric density of fats mean appetite regulation is easily overridden, particularly in diets lacking bulky fibrous plant-foods.
For those most interested in doing the best for our own health and our patients' what was lacking was a carefully controlled experiment to compare the 'pure forms' of the two main diet types involved in what we could call the 'diet wars'. This would help us to make more sense of the relative merits, and best applications for these different approaches.
Last May, the preprint of Kevin Hall and colleagues' paper was excitedly shared around my lifestyle medicine corner of Twitter, for very good reason: it appeared to give a very good indication as to whether the insulin hypothesis of obesity was correct. But it was a pre-print and fascinating though it was, those who were intrigued by the work waited throughout the whole of 2020 for the final version to appear: many iterations of my courses have passed without being able to share the slides I made (prematurely) in May. Patience has now been rewarded with publication in Nature Medicine
Now that it has been published Hall et al. 2021 is still an exciting paper. Why? It gives us some clear steers, some intriguing details and prompts all sorts of further questions. Also, unlike much nutrition research we know what the people in this study were actually eating, and what was happening in their bodies: this was a 'ward study'. In other words the people taking part were living in completely controlled circumstances where their diet could be defined, and multiple measures taken in real time.
So what happened? What follows is a careful reading of the paper re-ordered into a kind of story ...
Our participants were 20 healthy volunteers with an average BMI of 28 (range 20 - 40) who agreed to spend 4 whole weeks in the Metabolic Clinical Research Unit at the NIH Clinical Centre. Split into two groups of 10 they started on first one diet for two weeks and then swapped to the other, so everyone had 2 weeks on each diet. They were told the trial was looking at how the body handled different nutrients, and that it was not a weight loss trial, so they could eat as much or as little from the food they were supplied. This arrived at 3 meals a day, with available snacks and water. They were supplied or offered twice their daily caloric requirement each day. The diets were both pretty healthy in that both groups were eating close to kilo of non-starchy veg (NSV) a day. That is a massive improvement on almost anybody's diet. Hall and his colleagues note that everyone advocates optimising intake of NSV but that the low-carb approach goes on to recommend consumption of fatty animal products, and advises avoiding starchy veg, and starchy and sugary foods, whilst low-fat or whole-food plant-based diets generally recommend whole grains, legumes ,starchy veg and and avoidance of oils, fats and spreads. (It is worth noting that nobody with an ounce of sense or integrity recommends eating processed white starches or sugar.)
So: the low-carb (LC) group were offered (and ate) 10% of calories from carbohydrates, 75% of calories from fats and 15% of calories from protein, with a total of 61% calories coming from animal products, whilst the low-fat (LF) group were on 75% carbohydrates, 10% fat and 15% protein with 0% calories from animal products. The additional carbohydrates came from naturally low-fat whole food plant sources: whole grains, pulses and starchy vegetables.
The first thing to say about this food - yes, food, albeit forming a 'diet' - was that there were no differences in how it was rated between groups for pleasantness, or familiarity, and both diets scored the same for hunger, satisfaction and fullness. Everything that follows needs to be viewed in that light, which is after all most important, to most of us, most of the time.
From day one those on the LF diet ate fewer calories and continued to do so. Over 600 calories a day less than the LC group in the first week. By week two LC eaters were producing ketones and they then started to slightly reduce their calorie intake, but still the LF eaters were eating over 500 calories a day less. In marked contrast to this finding, the LC group dropped weight rapidly in the first week, and then continued to loose weight at a slower rate, whilst the LF group gradually accelerated into a more modest weight loss. Yet when body composition was analysed it was clear that almost all that first week weight loss in the LC group was 'everything but fat', with clear signs of increased protein turn-over, and by the end of the second week it was clear that only the LF group were in fact shedding statistically significant quantities of fat - at a rate of just over 0.25kg a week. For 500 - 600 calories/day reduction you might expect more weight loss on the LF diet, but going into the fine print, the reason for this is probably explained by a reduction in basic energy expenditure. The reasons for that aren’t given …
Remember the context, they all liked the food, were all able to eat as much or as little as they wanted, and were told this was not a weight loss trial.
The research group also monitored a huge range of other factors. The tables are long and complex so what really stands out - particularly from the perspective of a clinician managing and advising on patients' health overall?
The LC diet had some attractions: HbA1c's (in the normal range to begin with) went down a little, insulin secretion was a lot lower, for those that value ketosis, ketosis was clearly achieved, and there were intriguing reductions in VLDL (very low density lipoprotein) and triglycerides (TG) and HDL stayed higher than the LF group all of which are changes for the better. As a clinician however I was struck that the group went from borderline normal uric acid to very abnormal levels, an average of 7.2. (Raised uric acid will trigger gout in individuals prone to it.)
It was also striking that 9 out of 10 were found to have impaired glucose tolerance at the end of the LC diet (measured using a standard oral glucose tolerance test), as against 3 out of 10 after LF. So, does this mean we can say the trial showed worsened insulin sensitivity on the LC diet? No. It appears the trial appears was not ‘powered’ (with enough participants) to test this. (The derived Matsuda score for insulin sensitivity showed a trend without statistical significance toward worsened insulin sensitivity on the LC diet (p=0.8).)
Meanwhile on the LF diet although insulin production overall was increased, in line with the marked rises in both after-meal insulin and glucose, fasting insulin levels were lower than they had been at the beginning of the study and, as described, more people on the LF diet had normal glucose tolerance tests. This LF group had large reductions in total cholesterol (TC) - roughly 25% and LDL - roughly 30%, whilst triglycerides were raised after and between meals. They showed a slight improvement in uric acid (moving securely into the normal range). In addition to this there was a drop in blood pressure close to that observed with many drug treatments for hypertnesion, and a significantly lower pulse rate was observed.
What stands out to me as a GP is that people on the LF Whole Food Plant Based diet had improvements in all sorts of areas we are constantly striving to improve: body fat, blood pressure, total cholesterol and LDL, and glucose tolerance.
The authors raise concerns about the effects of raised lactate and raised post-prandial glucose (within the normal post-prandial range) on the LF diet. As a clinician it seems these may be more theoretical concerns than real-world, because whole-food plant-based diets are associated with very low levels of the diseases they flag up for concern.
Further Questions …
The findings themselves prompt many questions . Do the differences seen here between the two diets continue over longer time-frames? What would happen if the protein contents of these comparator diets were increased to be nearer those many of us now eat? Would we see bigger or smaller, or different effects in diabetics, or in a group that were all obese? What happens when proportions of fat, carbohydrate and protein change? What would we see if the low carbohydrate diet was plant-based? One can only hope that grant applications are in to answer some or all of these questions
Are carbs the root of all evil? Is the insulin theory of obesity true?
Returning to the main purpose of the study, we can, with the authors, conclude that in this closely controlled environment the insulin theory of obesity did not hold up: pushing down insulin levels overall by eating a ketogenic low-carb diet did not lead to anything like the intake reductions or fat loss shown on the comparator diet. The LF diet raised overall insulin production and so should, according to the theory, have led to weight and fat gain, particularly with the reduced energy expenditure seen in this study. None of the things predicted by the theory shown in the graphic at the head of this post actually happened.
This should give us pause for thought
Whether this paper prompts the kind of debate and discussion it richly deserves remains to be seen, but speaking as a generalist, it has always seemed of concern that so much hope and practice has been constructed on the basis of a single hormonal pathway. There is no doubt that in many patients we can see measurable improvements with a switch from their standard western diet to a diet with much less refined carbohydrate and more non-starchy vegetables, whatever reasoning we use to recommend those changes. The History of Medicine is full of examples of treatments that worked for reasons quite other than those understood at the time. Body chemistry is a complex thing, we aren't all the same, and a much broader conception of the various ways in which diet choices can help people move towards their goals is needed than simple stories reliant on single explanations. We are innately prone to believe our own narratives, particularly those held by a group, and need to always be alive to the possibility that a better story might remain to be told.