Thursday, March 31, 2016

A Meta-Analysis of Low Fat and Carbohydrate Trials for Weight Loss

A meta-analysis was published last December looking at weight loss in low fat diet RCTs lasting at least 1 year [1], where a low fat diet was compared with either.

  • Low carbohydrate diets
  • Higher fat diets
  • The participants’ usual diet 

I have covered most of the low fat vs. low carbohydrate trials they included [2] (I looked at trials with at least 100 participants and lasting at least 1 year).  In the low fat vs. low carbohydrate comparison they also included:

In summary, regardless of whether weight loss was a goal or not:

  • Low carbohydrate > low fat
  • Low fat = higher fat
  • Low fat > usual diet 

Macronutrients and weight loss: These results indicate that the low fat diet only had an advantage when compared to the usual diet and was equal to higher fat diets.  As is pointed out in the discussion, this suggests that dietary fat in these trials isn’t likely to influence weight loss; and that the benefit of the low fat diet compared to the usual diet could be due to an intervention effect (more interaction with the research staff/dieticians, more self-monitoring, etc)

The better performance of the low carbohydrate diet would suggest that replacing carbohydrates with fat aids weight loss efforts, while the results from the low fat vs. higher fat trials would suggest that it wouldn’t make any difference.  The reason here might be because the low carbohydrate diets had a higher protein intake but the higher fat diets did not.  The trials where the low carbohydrate group had a higher protein intake than the low fat group tended to be the ones where the low carbohydrate group outperformed the low fat group [2]

This meta-analysis is timely, considering the recent ‘Paleo’ mouse study, where a low carbohydrate, high fat diet unsurprisingly led to obesity in mice [3].  The senior author has said that "Low-carbohydrate, high-fat diets are becoming more popular, but there is no scientific evidence that these diets work.” [4].  As I mentioned in the post, it’s clear none of the authors have read/acknowledged the results from relevant RCTs that contradict his claim and this can be clearly seen from this meta-analysis: that low carbohydrate, high fat diets work to some degree*  and to a better degree than low fat diets**

* Average weight loss in these trials (low fat, low carb, etc) is generally not very exciting.  So now, as the authors conclude "Additional research is needed to identify optimum intervention strategies for long-term weight loss and weight maintenance, including the need to look beyond variations in macronutrient composition."

** While the difference between the low carbohydrate and low fat groups isn’t large, consider that most of the low carbohydrate diet arms were mostly ad libitum while most of the low fat arms included some level of advice to deliberately reduce calorie intake

A paradigm shift: This paper is primarily a meta-analysis looking at weight loss in RCTs of low fat diets, where the main conclusion is that “Findings from our systematic literature review and meta-analysis of RCTs do not support the efficacy of low-fat diet interventions over higher-fat diet interventions of similar intensity for significant, long term, clinically meaningful weight control”.  Some people may see this as a profound loss for the low fat dogma (which is still very prevalent in the media and in public health), particularly as this paper is co-authored by a few big names (David Ludwig, Walter Willett and Frank Hu).  But before you get too surprised, realise that the Harvard School of Public Health hasn’t been promoting a low fat diet for quite a while [5]

Their ideal slightly higher fat diet is one that’s low in saturated fat and high in polyunsaturated fat, where vegetable oils are included on the Harvard healthy eating plate and are recommended to be a major fat source [6].  These kinds of people can influence policy makers and it’s interesting to note that the new 2015 USDA dietary guidelines have removed the restriction on total fat intake but have made ‘oils’ part of their key recommendations and a kind of pseudo food group [7, 8, 9].  This is mainly based on the idea that replacing saturated fat with polyunsaturated fat will reduce the risk of coronary heart disease, however this isn’t supported by adequately controlled RCTs [10, 11].  The new emphasis of vegetable oils, which are ubiquitous and something westerners have no difficulty getting plenty of, may simply just dilute the nutrient density of the diet


  1. I think the (statistically significant but not massive) advantage of LCHF of LFHC would become even more evident if most of the high-omega6 PUFA vegetable oils consumed on LCHF were replaced with more mono & especially saturated fats. Why? The best reason i've seen is given by Peter of Hyperlipid:

    "My (repetitive) idea is that a certain amount of input to the ETC via ETFdh drives reverse electron flow through complex I to limit adipocyte distention. Stearic acid plus chow starch does this early during a meal. Butter plus potato starch does this early during a meal. Canola oil plus potato starch doesn't.

    We have no idea whether the butter with potato starch would carry on, long term, to a slim phenotype in people. The long term effect of stearate is downwards and of oleate/linoleate is upwards on bodyweight in mice, but the effect is small. How come?"

    1. Thanks Raphi, that's an interesting mechanism. Do you know of other instances where stearate or butter is protective?

  2. I can't say that I do (not for stearate specifically). However, the idea is about SFAs more generally: 2 protons & 2 electrons are supplied to FADH2 at Electron Transfer Flavoprotein DeHydrogenase (ETFdh) inducing more adipocytic IR quicker than with PUFAs [partially via reverse electron flow through Complex 1...if im not mistaken].

    "beta oxidation [of PUFAs] generates a much lower input at ETFdh (1 less FADH2 per double bond) & so insulin sensitivity at peak nutrient uptake is maintained for longer, fat pours in to adipocytes for longer" [again, from Peter]

    There's bound to be additional explanations but this very 'basic' (ETC-centric) explanation satisfies my bias of 'no more complexity (higher level signaling) than needed'.

    What protective instances are you thinking of?

    1. I was just wondering about other mouse studies or human studies, etc. Palm oil is used fairly commonly in HFD mouse studies and is quite obesogenic for them (cocoa butter is much less so, but this appears to be due to not being well absorbed)

  3. Thanks for your very interesting post. Like to hear your view on if low carb (LC) diet can actually reverse diabetes? The idea is being promoted by many, such as at

    Seems to me that LC can be effective to lose weight, but not sure if LC can really reverse diabetes, and seems to me that so far no systematic clinical studies have been conducted/published to support the claim.

    Thanks in advance

    1. Low carb diets make a lot of sense for diabetes management. I don't doubt they could reverse it and I'm sure you could find many testimonials of this happening. But I'm not familiar enough with the research to show you a clinical trial that found that outcome. I am familiar with a clinical trial where a Paleo diet essentially reversed diabetes. Notably the calorie intake in the Paleo group was quite low (1388 kcal)even though it was ad libitum. Short term very low calorie diets and bariatric surgery (forced calorie restriction) are both very successful methods of essentially reversing type 2 diabetes

      One of the problems with the trials included in this meta-analysis is that the all the groups didn't lose much weight (1 to a few kgs), and while this was sufficient to lead to some modest improvements in glucose and insulin levels, it's not near diabetes reversal

  4. This comment has been removed by the author.

  5. This comment has been removed by the author.