Does cutting carbs lower glucose vs does advising people to cut carbs lower glucose?
How to waste 20 years going round in circles
If the objective of a clinical trial was to understand whether reducing carbohydrate intake lowers glucose in people with type 2 diabetes, you would think that after 20+ years of trials that we’d definitely have a good idea, right? No.
There are issues with many of these trials that prevent us from understanding whether cutting carbs per se lowers glucose. For example:
The calorie-content of the diet was usually reduced as well (so was it weight loss or carb reduction (or both) that had the effect?))
If the trial was weight-neutral (ie aimed to keep people the same weight) then they would have had to replace the carbs with either fat or protein. There you have the same issue - was it the carb reduction or the protein/fat increase which lowered glucose?
But really, the absolute doozy in all of these trials is we weren’t testing whether carb reduction lowers glucose. We were testing whether advising people to cut carbs lowers glucose.
Take a look at this table from Bueno et al’s meta-analysis of low carb trials (they looked at weight loss but the principle is the same). They aimed to include in this meta-analysis trials which limited carb intake to less than 50g per day.
But what did the individual trials actually test though?
WOOOOOPS. Some trials didn’t bother see to whether people even followed the advice. Most of those that did found people were consuming much, much more than 50g per day.
You’ll read reviews of low carb on glucose which say “the evidence to date does not suggest that low-carb diets are superior to other approaches for glucose reduction”. Now, if the evidence is pretty inconclusive (ie there is only a small difference between the low-carb vs high-carb group) then this is an appropriate statement to make. But boy, does it hide A LOT of very important caveats.
Equally, the statement “trials don’t show that reducing carbohydrate is a superior method for lowering glucose” is pretty intellectually dishonest. The trials were not carried out in such a way that they could answer this question.
What I am talking about here is the difference between efficacy and effectiveness.
Dietary studies in general have predominantly studied the effectiveness of giving advice to change a diet, rather than the efficacy of a change in dietary intake.
Efficacy can be defined as the “performance of an intervention under ideal and controlled” circumstances, whereas effectiveness refers to its performance under 'real-world' conditions.
Why does this matter? Well, I have never met a patient who wanted to know what the average outcome was from a pretty uncontrolled dietary intervention trial.
Let’s imagine a 12-month “free-living” low-carb trial with 30 participants randomised to the intervention group. And let’s say for argument’s sake the trial defines low-carb as <120g a day. 4 patients report having <120g, 6 patients report having 130-150g, 10 patients report having 150-200g and another 10 report having >200g. Now of course, this type of study relies on self-reported intake which is notoriously unreliable. In addition, participants in this type of effectiveness trial will make other changes like doing more or less physical activity, eating more or less protein/fat which are either not collected or rely on self-report. It’s a hot mess of poor internal validity and confounding.
What does this type of study tell me or the patient about the likely outcome they can expect if they follow a carbohydrate-reduced diet? Nowt.
In my opinion, these studies are not useful for clinical dietetics where we deal with individuals. My approach with patients is to discuss their needs, talk about what the evidence tells us about “efficacy” and then what those diets might look like in terms of day to day intake. Then together with the patient we work out a diet that gets close to the macros/nutritional composition (if appropriate).
What limits my clinical practice is that we really don’t have much efficacy data to rely on. In fact, this is the entire reason I left the NHS and went into research.
Why haven’t we done much efficacy research in nutrition?
Up until very recently, the only way to clinically support a person in changing their diet was by face-to-face appointments, either one-to-one or in a group. Sometimes telephone follow-ups have been used. Naturally, this is a labour-intensive approach and in any healthcare system, resources are limited. For example when I worked as a dietitian in the NHS (2008-2010) I saw a patient for 30 minutes as a first appointment, and then saw them for a 15 minute follow-up appointment 3 months later. There was not the capacity in the system to provide the support needed to educate a patient and make potentially dramatic changes to their diet. It’s also true that even modest changes in dietary intake can be hard to sustain.
Efficacy trials are also harder and much more expensive to carry out (you have to control EVERYTHING). The gold-standard for efficacy studies is the inpatient ward setting where all food and activity can be closely controlled.
So an understandable reason that we haven’t done much efficacy research in nutrition has been “what’s the point in knowing that a diet comprised of 10% carb & 30% protein lowers glucose when the vast majority of people would not be able to follow (or afford) that diet anyway? Should we really be spending millions on trials which will have very little impact on clinical care for the vast majority of people?” It’s not an unreasonable point.
However, advances in technology in the last decade have meant that we can now deliver care remotely. Artificial intelligence and the use of well-designed algorithms can also deliver personalised dietary interventions at scale. This technology will only get better. The Food4me trial gives a good example of how personalized dietary advice can be delivered remotely based on people’s current dietary intake. Theoretically (it hasn’t been well-tested yet) this type of technology could be much better at helping people follow a more restrictive diet like high-protein, low-carb etc. Wearables and biofeedback may also help people follow what have previously been considered “extreme” diets.
Of course, research funds are finite. Far more intelligent people than me make decisions about where research funds go: diabetes, cancer, neurological disease etc.
Nevertheless, we know that dietary factors are a major contributor to global morbidity and mortality. And in terms of type 2 diabetes management, we have a lot of work to do to help people meet treatment targets. More precise dietary advice and support could play an important role in this. So if we want to know what effect diet change X has on any outcome (including glucose), then only efficacy trials will tell us this. (And by the way, this is also the type of research the public wants us to do.)
If we actually had some solid data on what carbohydrate reduction does to glycaemia in T2D it would also put an end to the endless debates about it. I’m lying, it wouldn’t. But at least dietary guidelines wouldn’t have to write vague statements about the utility of low-carb diets in diabetes management (type 1 and 2) if we actually had some reliable data. We’d actually have data which doesn’t have a 0 rating.
Whatever happens with nutrition efficacy research, I pledge to pay researchers £100 to not do another free-living low-carb vs high-carb trial that stopped telling us anything useful in 1995.
Does cutting carbs lower glucose vs does advising people to cut carbs lower glucose?
Hi Nicola, love this content. I’m a research immunologist by trade but became interested in nutrition science for my own health etc. The lacks of standardisation in trials and weak data collection methods always struck me as bizarre. The firm convictions held by all the quacks out there, even folks with MDs???, is astounding! Thank you for being a responsible voice of reason.
You can instruct each patient to measure blood sugar before and after meals. They will see for themselves what raises blood glucose and what doesn’t. If I eat carbs I am a diabetic if I don’t I am not. You don’t need an unreliable trial to tell you that