Does keto only work for six months?

Girls enjoying healthy breakfast

A meta-analysis, published in The BMJ, suggests low-carb diets are better than control or low-fat diets for weight loss and diabetes remission at six months. But the benefits appear to diminish at 12 months.

How should we interpret this data? Does the data prove that low-carb diets lose their beneficial effect over time? And why would something that works for six months stop working at 12 months?

Turns out, the study doesn’t answer any of these questions. It is worth looking deeper into the included individual studies to understand how this meta-analysis reached its conclusions.

First off, this meta-analysis is confusing and potentially in error. For the 12-month outcome data, the authors cite studies that are 12 weeks, 24 weeks, and six months long.

I wrote to the authors asking for clarification. One of them responded, stating they chose to use one citation for each study but assured me the longer-term data exists for each study.

I find this troubling. If authors are citing 12-month data, they should link to the study showing the 12-month data. Anything else seems misleading and potentially irresponsible.

It is impossible to evaluate claims without being able to see the evidence, especially when the intervention part of the trial is finished and there is no record of what the participants ate at the 12-month mark.

A meta-analysis is only as good as the data included.

The authors of this meta-analysis stated they only included studies that used diets with less than 26% carbohydrate or less than 130 grams per day. In many of the included studies, this was untrue or unknown.

For example, the authors cite one study where the low-carb group ended up averaging 33% of their calories from carbs (Davis et al.).

Also, some studies specified the advice for carbohydrate goals but didn’t report the actual amount of carbs eaten. It is left up to us to guess if the participants complied with the diet.

Other studies probably should not have been part of the meta-analysis at all, including one with a 50% dropout rate and another that was not a diet-to-diet comparison, but one that compared a low-carb diet to a low-fat diet plus a weight loss drug.

Because the authors included studies with different low-carb definitions, protein and calories protocols, and medication usage, it makes it difficult, if not impossible, to lump them together and draw meaningful conclusions.

So, what can we conclude from the meta-analysis?

First, meta-analyses are challenging to do. The authors need to be meticulous with the studies included. In this case, it does not appear that they were.

Second, media headlines do not capture the nuance needed to understand the implication of studies like this one.

Some promoted the abstract headline to show low-carb diets are great, and others used the same headlines to show they are no better than other diets. Confusing, right?

And third, nutrition science is messy and complicated.

Many questions remain. Do low-carb diets only work for blood sugar control because of weight loss? Does it matter as long as they lead to improvements? Does the amount of protein matter? Is it all about compliance?

At this point, the data is clear that low-carb diets work for weight loss and improving diabetes control. But how do you know if a low-carb diet is right for you and if will work long term?

As many doctors are beginning to realize, a CGM can help determine which eating pattern is best for a particular individual’s blood sugar.

You can even do your own experiment. Eat high carb, low fat for two weeks while wearing a CGM. Then, switch to low carb, high fat, and compare the results. Does low carb work better for you? Alright! Stick with it. Does it work the same? OK! Then you can pick the diet you prefer.

Preference is a factor we may not always consider strongly enough.

From trials of very low-carb diets, we can infer that compliance matters. For example, the Guldbrand trial cited in the meta-analysis demonstrated a dramatically better reduction in medication use and glucose control for those who were most compliant with the low-carb recommendations.

Also, non-randomized trials, which include subjects who choose to eat keto diets, have excellent compliance and results.

It should be no surprise that a ”diet” works better when you can stick to it. However, that doesn’t mean all diets are equivalent. Monitoring results still matters.

If you enjoy eating low carb and incorporate it as a lifestyle — rather than a diet — it should be an easy decision to give it a try. Eating low carb improves blood sugar control and weight loss for most people, and you can see if it works for you.

If you prefer to eat a low-fat vegetarian diet and will stick with it, then you should try that. Just don’t assume it will improve your health markers. In this case, you should also use your health markers as a guide — just like you would with low carb.

Here is the formula you should shoot for:

Scientific evidence + dietary preference + monitoring results = success.

Let’s not make it more complicated than it has to be.

Thanks for reading,
Bret Scher MD FACC


US magazine once again misinforms readers about effective diets

Dr. David Ludwig on the history of carbohydrate reduction

Keto video