Dear JULES,
Type 2 diabetes & diet – how low should low carb be?
Executive summary
* A study was published in the BMJ in mid-January 2021. It was a systematic review and meta-analysis of randomised controlled trials, which examined remission of (type 2) diabetes (T2D) with low and very low carb diets.
* The systematic review found 23 relevant trials, with a minimum 12 week duration, involving 1,357 people. Approximately half the trials met the criteria for a very low carb diet (VLCD) (<50g day <10% calories) and half met the criteria for a low carb diet (LCD) (<130g/day <26% of calories).
* The main finding was that, at six months, compared with control diets, LCDs and VLCDs pooled together achieved higher rates of T2D remission. The control diets tended to be low fat diets.
* The Number Needed to Treat for this intervention was 3, which is excellent for a medical intervention, suggesting that a low or very low carb diet should be considered for putting T2D into remission rather than a low fat diet.
* However, this finding ceased to be significant when remission was assessed without medication still being used.
* Weight loss was better at 6 months for the LCD/VLCDs compared to the controlled diet. There was no significant difference at 12 months.
* HbA1c was lower at 6 months for the LCD/VLCDs compared to the controlled diet. There was no significant difference at 12 months.
* When there were sufficient data to compare LCDs and VLCDs with each other, VLCDs were less effective than LCDs for weight loss at six months. This effect was explained by diet adherence i.e. patients who were highly adherent to VLCDs did much better than less adherent VLCD patients.
* Regarding safety, there were no significant differences in adverse events or serious adverse events at 6 or 12 months between the LCDs/VLCDs and the control diets.
* This study prompts reflection that low carb diets may be just as effective, if not more so, than very low carb diets – taking into account adherence.
Introduction
Many thanks to Mark Benson for this week’s topic. Mark is a recently retired ophthalmic surgeon, who was involved in the treatment of diabetic retinopathy throughout his career. That would certainly drive someone towards trying to find a way to avoid having to perform such procedures and it drove Mark towards the dietary guideline/carbohydrate intake debate. Mark spotted an article which was published on January 13th, 2021. The article was called “Efficacy and safety of low and very low carbohydrate diets for type 2 diabetes remission: systematic review and meta-analysis of published and unpublished randomized trial data” (Ref 1). He shared that it was being reported on social media as low carb eating would only benefit diabetics for a few months.
If you recall our review of the hierarchy of evidence (Ref 2), from the title of this paper alone we know this is at the top of the pyramid for quality of evidence. It’s a systematic review – which means that rigorous attempts have been made to find all relevant studies (not cherry-picked studies). It’s a systematic review of randomised controlled trials, which is superior to a systematic review of population studies. It includes published and unpublished data (although the latter has not been peer-reviewed). It also includes a meta-analysis – which is the pooling together of studies to see what the combined data show us. Finally, this is a topic of much interest to us – low carb and very low carb diets for putting type 2 diabetes (T2D) into remission.
The study
The aim of this paper was to review the efficacy and safety of low carbohydrate diets (LCDs) and very low carbohydrate diets (VLCDs) for people with type 2 diabetes. Low carb was defined as <130g carb/day or <26% of a 2,000 calorie a day diet. Very low carb was defined as <50g carb/day or <10% of calories from carbohydrates. Only studies of at least 12 weeks duration were included.
The main outcome of interest was remission of T2D (defined as an HbA1c of below 6.5% or a fasting glucose < 7.0 mmol/L – with or without the use of diabetes medication). Other outcomes of interest were weight loss, HbA1c reduction, fasting glucose and adverse events. Outcomes were reviewed at 6 months and at 12 months, although data at 12 months were sparse. The researchers identified 23 trials, which involved a total of 1,357 participants. The trials mainly included overweight and obese patients with type 2 diabetes (T2D). The majority of studies (14/23) included participants using insulin. The trials ranged in size from 12 to 144 participants with an average (mean) age of 47-67 years. The studies varied in carbohydrate restriction with 12/23 meeting the criteria for VLCDs. Most trials (18/23) used low fat diets as the control comparators. The trial duration ranged from three months to two years. Dropouts were common – 18 out of 23 trials reported missing participant data, with 10 reporting more than 20% of data missing. The paper was so thorough in its explanations of, and tests for, so many things, but it was frustrating in some details. For example, Table 1 summarised the 23 trials but it didn’t group them by LCDs vs VLCDs. Knowing from the narrative of the paper that 12/23 met the criteria for VLCDs, most of these were clear from the diet descriptions, but not all. Shai et al (2008), for example, started off as a VLCD (20g carb/day for 2 months) and then became a LCD (increases in carbohydrate were allowed until 120g/day was being consumed). Vlachos et al (2011) was described as a “low carbohydrate and protein sparing modified fast.” There is a standard definition of such a diet and it is very low calorie, as well as low in fat and carbohydrate. Carbohydrate does increase over time, however, and so it wasn’t clear if this was counted as a VLCD. The headline conclusion The abstract of the paper had a comprehensive report on results. The headline conclusion was that, at six months, compared with control diets, LCDs achieved higher rates of T2D remission. At 12 months, data on remission were too sparse to be of significance. Another main conclusion was that VLCDs were less effective than less restrictive LCDs for weight loss at six months. This effect was explained by diet adherence i.e. patients who were highly adherent to VLCDs achieved a “clinically important reduction in weight”, compared with less adherent patients on VLCDs. The takeaway message from that could be – don’t bother with a VLCD – unless you’re going to be super adherent, a LCD will do better for weight loss at six months. Readers of the paper might also be discouraged that there’s not much evidence of benefit at 12 months, so why bother. (The social media interpretation). Anyone thinking this should be reminded of our review of the evidence of Virta (VLCD) vs DiRECT (very low-calorie diet) for remission of T2D (and other outcomes) at two years (Ref 3). Virta wasn’t included in this systematic review as it wasn’t strictly randomised, and thus didn’t meet the randomised controlled trial criteria. Those headline messages need some unpacking, so let’s proceed… The results The paper summarised the main result as “at six months, compared with control diets, LCDs achieved higher rates of diabetes remission.” This conclusion was based on just 8 of the 23 studies, which reported on remission of T2D (HbA1c <6.5%) at 6 months – when medications were still being taken. These eight studies included three VLCDs (Dyson et al 2010, Saslow et al 2014, and Saslow et al 2017) and five LCDs (Jönsson et al 2009, Tay et al 2014, Yamada et al 2014, Sato et al 2017, and Morris et al 2019). When these 8 studies (allowing for diabetes medications still being taken) were pooled together, 76 out of 133 people (57%) in the LCD intervention and 41 out of 131 people (31%) in the control group achieved remission of T2D at 6 months. Meta-analysis established that this was a risk difference of 0.32 (95% confidence interval 0.17 to 0.47). In medical interventions, the Number Needed to Treat (NNT) is a really useful way of evaluating the benefit of a treatment. The NNT is the number of people that need to be treated for one person to benefit. If an intervention has an NNT of 1 – every person is helped by the treatment. If an intervention has an NNT of 100, it means that 100 people need to be treated for 1 person to benefit. The NNT for the LCD intervention was 3, which is extraordinary in healthcare. To put this in perspective, statins, for people without heart disease, have an NNT of 104 to prevent a non-fatal heart attack. Meanwhile they have a number needed to harm of 1 in 10 for muscle damage and 1 in 50 for development of type 2 diabetes (Ref 4). The main finding ceased to be significant when remission was assessed without medication still being used. There were five studies that could be examined for this research question. Pooling these five studies together in meta-analysis failed to achieve a (statistically) significant result. These findings were followed by two non-findings: - When 3 studies were examined for remission of T2D at 12 months, there was no difference between LCDs (and they were LCDs, not VLCDs – Tay et al 2014, Sato et al 2017, and Morris et al 2019) and the control diet. - When 2 studies were examined for remission of T2D at 12 months – without medication – there was no difference between LCDs and the control diet. The results from the absolute outcomes of interest (weight loss and HbA1c) were: - When 18 studies were examined for weight loss at 6 months, weight loss among the LCDs/VLCDs was an average of 3.46kg lower than the control diet. - When 7 studies were examined for weight loss at 12 months, there was no significant difference between the LCD/VLCDs and the control diet. - When 17 studies were examined for HbA1c at 6 months, this was an average of 0.47% lower among the LCDs/VLCDs than the control diet. - When 8 studies were examined for HbA1c at 12 months, there was no significant difference between the LCDs/VLCDs and the control diet. - When 9 studies were examined for adverse events, there was no significant difference between the LCDs/VLCDs and the control diet. There were also no significant differences in adverse events at 12 months, although just 3 studies were available to examine this. The paper doesn’t tell us if VLCDs did better than LCDs for T2D remission – just that VLCDs and LCDs combined did better at six months than the comparator diets (which were generally low-fat diets). The fact that LCDs couldn’t be compared with VLCDs was disappointing, but it was an unavoidable outcome from the limited number of studies with the data needed for each research question. The Cochrane Handbook mandates that researchers must “Ensure that there are adequate studies to justify subgroup analyses… at least ten observations (i.e. ten studies in a meta-analysis) should be available for each characteristic modelled” (Ref 5). Subgroups themselves should have a minimum of four studies (Ref 6). Subgroup analysis The researchers examined a number of subgroups and found significant results in three. These were reported in the narrative of the paper and further information was given in the supplementary tables. One of the three subgroup findings was that studies that included patients using insulin achieved fewer remissions at six months compared with studies that didn’t include patients on insulin. That makes sense. It suggests that it is more difficult to achieve remission when the T2D has reached the insulin medication stage than when it hasn’t. The other two subgroup results were relevant to the issue of LCD vs VLCD performance. First, the VLCDs achieved smaller weight loss at six months than the LCDs (Ref 7). Second, this was explained by diet adherence. For the first of these findings, when VCLDs were compared with LCDs for weight loss at six months, the VLCDs averaged about 1kg weight loss, but the results were not statistically significant. The LCDs averaged nearly 6kg weight loss and the results were significant. That’s quite a difference. The second part examined just the VLCD group and compared those who were highly adherent with those who were less so. This found that the highly adherent VLCD people averaged 4.5kg weight loss and this was statistically significant. The less adherent VLCD people averaged barely half a kg and this was not statistically significant (Ref 8). Taking the two observations together, this means that the LCD people averaged together still did better than even the most adherent VLCD people. Only by 1.5kg in 6 months, but that is still an interesting finding. Reflections I have had the privilege of hearing Dr David Cavan speak at conferences about the work he has done in Bermuda. He has managed to help people to lose a lot of weight and many to achieve type 2 diabetes remission with carbohydrate intake of approximately 100g/day. If someone is diabetic (either type) there is a strong argument for “the lower carbohydrate intake the better”, but much can still be achieved with low rather than very low. Dr Cavan (and others) have shown that low can get good results. Very low might get better results, but this study has given us evidence to consider. If people can’t stick to very low, then low seems to be better than giving up on very low. My first book (Why do you overeat? When all you want is to be slim) was written back in 2004 and it was about eating real (non-processed food) and then managing carbohydrate intake. I introduced the concept of ‘carb/protein’ meals (grains, legumes, fruits) and ‘fat/protein’ meals (meat, fish, eggs, dairy), followed by the concept of not mixing carb/proteins and fat/proteins at meals as far as possible. The rationale for this was threefold: i) nature provides carb/proteins or fat/proteins but rarely carb/fat/protein combos and thus eat as nature intended; ii) the unnatural carb/fat combo is the domain of fake food (cake, biscuits, confectionery, ice cream, muffins, doughnuts etc) and it can be overconsumed; and iii) eating carbs (insulin stimulating) and fat at the same time provides for an ideal fat storing environment. Even Phase 1 of the diet I suggested allowed unlimited non-starchy vegetables and salads and 50g a day of a non-wheat carbohydrate (e.g. brown rice, oats, quinoa). Initial weight loss turned out to be high (several pounds in a few days). People then settled into a regime of (real food) ‘fat’ or ‘carb’ meals (protein being in everything other than sucrose and oils, so the word protein could be dropped). People soon realised that the more ‘fat’ meals they had, the more satiated they felt, the less hungry they got and the better they lost weight. Many would thus gravitate towards low carb on their own weight and health journey. Some ended up very low carb/keto. I wonder how well people would have done had they all started out on a VLCD. Would they have immediately embraced the vast difference between a VLCD and government dietary guidelines? Would they have found it a step too far and given up? I was vegetarian at the time I wrote my first book. I adapted the carb intake to be up to 150g in the first few days to give vegetarians enough to eat. People still did well and lost weight well. Vegetarians made the same discovery with carb vs fat meals and they too would gravitate more to dairy and eggs rather than pasta and rice. The diet is still a good route in for people to move away from dietary guidelines to real food and lower carb intake. Maybe we should be more open to people reducing carb intake and moving from there? While 50g may be better than 100g, 100g is definitely better than the government recommended 300-400g. If we can just get people away from the absurd government guideline carb intake, they will discover for themselves the power of carb restriction for weight, hunger and much more. They can then manage their own journey at their own pace. If we try to get people to the finishing line right from the start, they may not continue the race and they also have nowhere left to go if very low carb stops working for them (as it does with many people). A bit of flexibility in the low carb world may go a long way. This study has been an interesting reminder of that. Until the next time All the best - Zoë References Ref 1: Goldenberg et al. Efficacy and safety of low and very low carbohydrate diets for type 2 diabetes remission: systematic review and meta-analysis of published and unpublished randomized trial data. BMJ 2021. https://www.bmj.com/content/372/bmj.m4743 Ref 2: https://www.zoeharcombe.com/2020/08/the-hierarchy-of-evidence/ Ref 3: https://www.zoeharcombe.com/2020/09/low-calorie-vs-low-carbohydrate-for-t2d-remission-direct-vs-virta/ Ref 4: https://www.thennt.com/nnt/statins-for-heart-disease-prevention-without-prior-heart-disease-2/ Ref 5: Higgins et al. Cochrane Handbook for Systematic Reviews of Interventions version 6.0 (updated July 2019). Cochrane 2019. Ref 6: Fu R, Gartlehner G, Grant M ea. Conducting Quantitative Synthesis When Comparing Medical Interventions: AHRQ and the Effective Health Care Program. . In: Methods Guide for Effectiveness and Comparative Effectiveness Reviews [Internet] Rockville (MD): Agency for Healthcare Research and Quality (US), 2010 Ref 7: VLCDs led to smaller weight loss at six months (mean difference –1.05, –2.27 to 0.17) than did LCDs (mean difference –5.22, –8.33 to –2.11) (test for subgroup difference P=0.01). Ref 8: Among VLCDs to which the patients were highly adherent, a larger clinically important weight loss occurred (mean difference –4.47, –8.21 to –0.73) compared with patients less adherent to VLCDs (mean difference –0.55, –1.76 to 0.66) (test for subgroup difference P=0.05).