The science of intermittent fasting
We’re back with the Science Series and today I’ll be discussing what the research says about intermittent fasting for health and performance. This post will be split over a number of parts as there’s quite a lot to cover and the last thing I’d want to do is bore you…
Intermittent fasting (IF) has emerged as a popular strategy in the fitness world, and beyond. In my How to eat for fat-loss nutrition guide, I recommend a 14 hour daily fast. The principle reason for using IF for fat-loss is that it serves as an effective heuristic for reducing overall caloric intake, helping to create the deficit you need in order to lose weight. But there are other benefits to IF that go beyond controlling your intake and it may be a good idea to fast daily even if your primary goal is not fat-loss, but health or performance (I fast and have no intention of losing weight).
Before we go any further. What is intermittent fasting? Intermittent fasting is any protocol that schedules precise periods of fasting and eating. Fasting periods can last anywhere from 12 hours to a day or two. IF includes popular nutritional strategies like 16:8, 5:2 and the more recent TRE (time restricted eating). In this post, I will mostly consider the more common IF protocols that implement daily fasting periods of 14-16 hours, including both those that adhere to circadian rhythms (e.g. TRE) and those that do not. Note that this article will not discuss the many potential benefits of prolonged fasting. I think of prolonged fasts as those that last 3 days or more and the physiological mechanisms and consequent benefits may be quite different.
Insulin, glucose and metabolic disease
In today’s post, I’ll discuss two important recent papers on intermittent fasting: one published by Sutton et al. in Cell Metabolism in 2018 and the other published my Moro et al. in the Journal of Translational Medicine in 2016. I’m looking at these papers in particular for two reasons:
- They are both randomised controlled trials (RCT) in humans => excepting systematic review of RCTs, these provide the best and most unbiased data.
- They both consider IF in the context of stable-weight subjects => in other words, these studies isolate the benefits of IF from the benefits of caloric restriction.
Both papers primarily concern themselves with investigating the potential metabolic benefits of IF, especially as far as glucose and insulin regulation are concerned. But they also look at some markers of cardiovascular health, endocrinology and inflammation.
If you’ve been in the fitness scene for more than 5 minutes, you’ve heard of insulin and you’re probably at least vaguely aware of the concept of insulin resistance. To put it very simply, chronically elevated levels of insulin (hyperinsulinaemia) and blood glucose (hyperglycaemia) are markers of poor metabolic health and are associated with nasty chronic diseases: obesity, type 2 diabetes, heart disease, kidney disease, stroke etc. Keeping insulin and glucose within fairly narrow ranges is therefore critical for optimal health and longevity as well as for weight management. If IF can help us achieve this goal, we should take note.
So what do these two studies say and how did they reach their conclusions? I’ll consider the papers in turn, looking at Sutton et al. first.
Early time-restricted eating: the Sutton et al. paper
The Sutton et al. trial used a classic randomised controlled ‘crossover’ set-up. Participants were first randomised to either eTRE (early TRE) or control groups. The eTRE group ate within a 6-hour window, with dinner finishing before 3pm, while the control group ate within a 12-hour window. All meals were provided and supervised by the researchers and were identical between groups (so-called ‘eucaloric’). The only difference was meal timing. Intake was set so that subjects were eating the same number of calories as before the trial and consequently maintained their weight (a so-called ‘isocaloric’ experiment). They followed this paradigm for 5 weeks with measurements being taken at baseline and at the end of the intervention. Then, following a 7 week ‘wash-out’ period, the participants swapped groups and undertook the opposite protocol. Researchers then compared intra-individual changes observed over 5 weeks of eTRE against the changes observed in 5 weeks in the control group; that is, they looked at the difference between how an individual responded to eTRE as compared with how that same individual responded to the control diet.
Although the study was very small, with just 8 participants, the fact that researchers controlled food intake and timing so rigorously and measured intra-individual differences, rather than inter-individual differences, makes the data compelling.
What the trial found is fascinating. The primary end-points under investigation were measurements of “glucose tolerance, postprandial insulin, and insulin sensitivity as measured using a 3-hr oral glucose tolerance test (OGTT)”. An OGTT involves giving a patient a 75g dose of glucose solution and then measuring blood glucose and insulin levels at 30, 60, 90, 120 and 180 minutes afterward. The peak and average values of these biomarkers over the 3 hours gives a good indication of your metabolic health. All else being equal, lower average and peak values for glucose and insulin across the test is preferable, as discussed above.
Interestingly, the experiment revealed no difference between groups in terms of fasting glucose or glucose readings during the OGTT. But insulin dynamics did change significantly. eTRE decreased fasting insulin levels as well as reducing insulin measurements at 60 and 90 minutes following the glucose test. Over the 3-hour window, eTRE lowered both average and peak insulin levels. This indicates an increase in insulin sensitivity. As most of you will know, one of the main functions of insulin is to provide a signal to cells to import glucose from the blood. So if eTRE and control participants showed similar glucose dynamics but the eTRE group achieved these dynamics at lower average insulin levels, then eTRE cells must be responding more sensitively to insulin signalling. Improving insulin sensitivity could have a substantial effect on metabolic health, even if glucose dynamics are left unchanged. There is good evidence to suggest that hyperinsulinaemia is the earliest predictor of future metabolic disease and is a risk factor independently of poor glucose regulation.
What about the size of the effect on insulin? eTRE was observed to bring fasting insulin down by 3.4 mU/L or ~14.5%, down from an average of 23.4 mU/L to 20 mU/L. That takes the average subject from moderate hyperinsulinaemia back into the normal range. So, all in all, you’d have to say that’s a pretty important effect, especially given that the intervention lasted only 5 weeks. Similarly, average and peak insulin during the OGTT both improved by around 20% each.
Figure: glucose and insulin regulation in eTRE and control
Secondary measurements revealed that eTRE may decrease blood pressure. eTRE lead to a drop in both systolic and diastolic blood pressure of 11 and 10 mm HG respectively. This is interesting. Since participants were, on average, just north of prehypertensive (120+/80+), this drop took them back into a fairly optimal range below 120/80. Sutton et al. found that IF improved a marker of oxidative stress, isoprostane-8, but did not improve markers of chronic inflammation (high sensitivity c-reactive protein [CRP], IL-6). Cortisol was also unaffected by IF.
Figure: blood pressure, inflammation and oxidative stress in eTRE and control
Given the highly controlled nature of this experiment, I think the results are worth paying attention to. That being said, the trial was very small and a larger replication would be needed to really corroborate the effects. In addition, the trial participants were pre-diabetic middle-aged men. As with other papers I’ve discussed, the effects observed in this population may not translate to normal, healthy subjects (and perhaps not women either). Luckily for us, the Moro et al. paper looks at IF in a healthier cohort.
IF in weight-trained individuals: the Moro et al. paper
Moro et al. investigated the effects of IF in a cohort of 34 weight-trained young men over the course of an 8-week RCT. These were men with a minimum of 5 years of experience lifting weights, so the bar was fairly high. Subjects were instructed to maintain their previous intake of calories throughout the experiment. Half the participants were randomised to eat all their meals between 1pm and 8pm, while the control group ate between 8am and 8pm. During the experiment, subjects took part in identical training programmes to control for any differences in exercise volume and intensity. Again, the motivation for the trial was to see what effects IF might have beyond effects mediated by reduced food intake and weight-loss.
The headline result was that intermittent fasting led to reductions in body fat, while the control group did not lose fat. IF subjects lost an average of 1.62kg of fat mass over the 8 weeks. That’s not a huge amount but you should bear in mind that these were relatively lean, trained individuals. Subjects began the experiment with an average of ~10kg of fat mass so the fat-loss seen by the IF cohort represented a loss of over 15% of their body fat. And this without a reduction in calories… IF subjects maintained their lean mass too.
You might wonder what mechanism could mediate this kind of fat-loss. The authors point out that adiponectin was higher in IF subjects. Adiponectin is a hormone produced by fat-cells (adipocytes) that controls metabolism. Chronically low levels are often observed in obese people. Higher levels of the hormone are associated with an increase in energy expenditure and this may account for the fat-loss seen in the trial. Adiponectin causes increased glucose uptake by muscle cells, increased fat oxidation in the liver and in fat cells and decreased production of glucose in the liver. These effects are mediated through the PPAR Alpha, AMPK and p38 pathways as illustrated below. Whatever the mechanism, I think we would all agree that losing fat without decreasing food intake is pretty much the dream. So if this effect is real, we should take note.
Figure: the mechanisms of adiponectin
In agreement with Sutton et al., this trial found that IF lead to decreased fasting insulin. Moreover, they found that IF had a positive effect on fasting glucose (in contrast to Sutton et al.). Given that these trained individuals started with healthy measurements of fasting glucose and insulin, these results are surprising and suggest that we might be able to use IF to optimise already healthy lifestyles. It might also be the case that blood glucose is more responsive in currently healthy individuals, which would explain why we see a change in glucose dynamics in this trial but not in the trial conducted on pre-diabetic men.
Moro et al. observed even larger effects on fasting insulin than Sutton et al. did. The average IF subject decreased their fasting insulin by 10.1 mU/L, or 36%, from 27.8 to 17.7 mU/L! I do wonder why these young weight-trained men had worse baseline fasting insulin than the pre-diabetic middle-aged men, however… Too much pasta in Padova perhaps? Fasting glucose dropped by 10.7 mg/dL, or 11%, from 96.64 to 85.92 mg/dL.
Figure: fasting glucose and insulin in weight-trained IF and control subjects
Moro et al. measured a number of other biomarkers. They found, in contrast with Sutton et al. that IF improved markers of inflammation, reducing IL-6 and TNF-alpha (tumour necrosis factor alpha).
Figure: inflammatory markers in weight-trained IF and control subjects
IF appeared to reduce both testosterone and insulin-like growth factor 1 (IGF-1). This is not necessarily surprising since insulin, being an anabolic hormone, is known to be correlated with testosterone but you might wonder if lower testosterone and IGF-1 is desirable. You should note, first of all, that the IF cohort were able to maintain muscle mass and strength so the decreased levels of testosterone does not seem to have had a negative effect in that regard. However, the study was only 8 weeks long and so the long-term effects of this drop in testosterone is unknown. The average level of testosterone after IF was still well within normal ranges so I don’t think it’s a significant cause for concern. But if your primary goals are strength and athletic performance, this might be a consideration. Just be aware that, as with anything, higher levels of testosterone are not necessarily better: just consult your local steroid-user… Anecdotally, I’ve seen no change in my testosterone levels since starting TRE; in fact, they’ve increased a little bit.
Figure: testosterone and IGF-1 in weight-trained IF and control subjects
The story with IGF-1 is a little more nuanced. IGF-1, which is a downstream product of growth hormone, is an important anabolic agent and so critical for muscle protein synthesis and neurogenesis. But, we know from the data that when it comes to longevity, both high and low IGF-1 levels are associated with increased mortality risk. So a balance is critical. Thinking of IGF-1 and testosterone as GOOD, because of their role in building muscle and fueling performance, is a dangerous game. If health is something you prioritise, balance is always key. I would recommend that if you try IF, you just keep an eye on testosterone (and IGF-1 if possible). If it drops too low, you may need to stop fasting, reduce the fasting periods, or simply increase your food (and particularly carbohydrate) intake.
Figure. Predicted HR for the association between IGF-I and all-cause mortality. [Burgers et al., 2011]
You might be wondering why I didn’t discuss the Moro et al. paper first, given the more positive findings and the more relevant population under investigation. The reason is that while I found the Sutton et al. experimental set-up to be highly rigorous, there were some limitations to the Moro et al. methodology.
The most important of these is the inherent risk of what is known as performance bias. Performance bias occurs where subjects in an experiment know what is expected of them and understand the outcomes researchers are hoping for. As a result of this, they change their behaviour, or the way in which they report their behaviour, to match with these expectations. In the current study, participants were told to keep their total caloric intake constant while adopting the fasting protocol. Now, food intake was measured using a weekly questionnaire and, unsurprisingly, the results of these questionnaires showed that subjects achieved caloric constancy. But there is a definite risk here that participants reported what they knew they should have eaten rather than what they did in fact eat. What’s more, they may have done so without knowing it. Food Frequency Questionnaires, or FFQs, are one of the major problems with the papers linking red meat with mortality and cancer (see my post here). The questionnaires used in this study were better in that they were weekly rather than yearly (how crazy is that?!) but there is still room for error and bias.
And if calories were misreported, in this case being over-reported, there could be a very different explanation for the results seen in the experiment. If the IF group were eating fewer calories than previously, then we would be forced to concede that it was most likely this caloric deficit which mediated the fat-loss and improvements in fasting glucose and fasting insulin. Now, I’m not saying that this is definitely what happened; just that there is some non-negligible risk that it did. And so I think a replication of the study would be needed to confirm the results. In the meantime, we should consider the evidence as preliminary, if very interesting.
Primary outcome bias
The other slight issue I have with this paper is related to the concept of ‘bait and switch’ and primary outcomes (which you can read more about in Ben Goldacre’s excellent book Bad Pharma). In a proper medical clinical trial it is incredibly important to define the primary outcome of the experiment before it is conducted and you should not switch that outcome during the experiment. Why? This is a little subtle but it’s related to the concept of statistical significance… stick with me, please! Most studies choose a 5% level of significance, which means that there is a 5% chance that the experimental results are purely a chance effect; a 5% chance that the effects measured in the trial don’t really exist. Now, let’s suppose you choose blood pressure as your primary outcome. But then half-way through the trial, when the blood pressure results don’t look promising, you switch the primary outcome to LDL-cholesterol. No big deal right? Wrong! Because what is the probability that the results of your trial represent pure chance now? It’s not 5% any longer because you already had a 5% shot with your blood pressure test and now you’ve got an additional 5% shot with LDL-c too (assuming these are independent variables). So, in actual fact, the chances of your finding a positive result, either with blood pressure or cholesterol, purely by chance, is 10%. And you can see what happens when you add further primary outcomes and more switching… In this way, you can simply add or remove outcomes until you get a positive result, purely by chance.
Now, with the Sutton et al. paper, the authors very clearly set out the primary outcomes of their trial: glucose and insulin dynamics as defined over a 3-hour OGTT. They measured other data-points too, but they were very precise about their primary focus. With the Moro et al. paper, the primary outcome is never defined and the number of biomarkers measured was enormous. If you measure enough data points, you’ll eventually find a positive one! That being said, this is only a marginal issue and certainly doesn’t invalidate the study because Moro et al. did not just find one positive result; they found many. So the results are unlikely to represent pure chance. I mention it only as one more thing to have on your radar when it comes to reading studies!
Putting it all together
Putting this all together, I think these two trials provide strong evidence that intermittent fasting improves insulin regulation since this effect was observed in both. Decreased fasting insulin, as well as lower mean and peak insulin during an OGTT, suggests improved insulin sensitivity and metabolic health. There is good reason to believe that hyperinsulinaemia lies at the heart of many chronic diseases and so tools to help prevent and reduce this condition may be crucially important for long-term health. As for the regulation of glucose, the evidence is mixed but it appears that healthier subjects may respond more favourably to IF in this regard.
The Moro et al. paper indicates that IF may lead to fat-loss without a reduction in overall calories, perhaps mediated by increased adiponectin. But further studies in a more highly controlled experimental set up would be needed to confirm this finding, I think.
Both studies also indicate that IF may have other important roles to play in a healthy lifestyle. IF may decrease blood pressure, oxidative stress and inflammation. Other studies, considering these data points as their primary outcomes would be needed to really tease out these benefits. On the potentially negative side, IF may lead to reduced IGF-1 and testosterone. This is probably not as bad as it sounds since both these hormones should be kept balanced as both high and low levels can be harmful. But if you are using IF, it may be a good idea to get regular bloods taken to make sure nothing untoward is happening. See below for a graphical representation of the findings of these two papers: