Mass Research (Volume 4 - Issue 3)
There are numerous “biological clocks” operating in our bodies that influence the cyclical nature of sleep, hunger, food seeking behavior, digestion, energy homeostasis, reproductive hormonal patterns, and more. Probably the most well-recognized is the sleep-wake cycle, in which we synchronize our sleeping patterns with the light and dark cycles of our environment. Additionally, our sleep-wake cycles are modified by the type, timing, and amount of food we consume. From an evolutionary perspective, our ability to modify when we are awake and asleep based on food availability likely increased our ancestors’ survival rates. In the modern world, these adaptations interact with artificial light, work and training schedules, and the food environment in ways we often consider unfavorable. In this systematic review (1), the authors examined 15 cross-sectional studies and four randomized controlled trials to determine if there was a relationship between macronutrient distribution and sleep quality. Overall, a pattern emerged in which better sleep was observed among those consuming higher protein, but lower fat and lower carbohydrate diets. How sleep quality was assessed, and whether relationships were based on cross sectional or controlled trial data adds granularity to these findings. Additionally, the authors used regression to determine if there was a dose-response relationship between percentage of calories from protein and sleep quality, but found none. Read on to learn what we can and can’t conclude from this study, and how it might impact your diet.
The goal of this systematic review was to assess the association between sleep quality and macronutrient distribution in healthy adults.
No hypotheses were stated.
Systematic Search and Data Inclusion Details
As a systematic review with a meta-regression, the authors established a set of search terms and inclusion criteria to locate relevant research and extract its data. The search strings and database search results and inclusion criteria are shown in Tables 1 and 2.
In addition to these inclusion criteria, studies were excluded if they reported evening meals (some data suggest that eating near bedtime can independently impact sleep – more on this later), or if they only reported subjective measures of sleep quality. As shown in Figure 1, a total of 19 studies were included in this review, consisting of 15 cross-sectional studies (analyses at a single time point that examine relationships between variables of interest; can only establish correlation) and four randomized controlled trials (analyses of groups over time in which interventions were compared; can suggest causation).
Analysis of Variables
All meta-analyses are systematic reviews, but not all systematic reviews are meta-analyses. The systematic – and ideally, repeatable – process of finding and selecting articles is what makes something a systematic review, while a meta-analysis goes a step further by attempting to quantitatively summarize the body of literature, often by calculating a pooled effect size of some type. This particular study did not calculate a pooled effect size, but it did include a meta-regression. Like a regression analysis in an isolated study, a meta-regression attempts to establish a mathematical relationship between two variables, such that you can forecast an outcome variable, based on a predictor variable (e.g. if someone does “x” number of sets to failure, I can forecast “y” increase in muscle cross sectional area, with “z” amount of estimation error). However, a meta-regression establishes this mathematical relationship using groups from multiple studies as the data points, while a regression in a single study uses data from individual subjects. A regression analysis can be used to determine if there appears to be a dose-response relationship between two variables; for example, Morton and colleagues (2) recently reported a positive dose-response relationship between increases in strength and lean mass with protein intake up to a plateau at 1.6g/kg/day (with an upper 95% confidence limit of 2.2g/kg/day) by adding the results of several independent studies into a big regression model. In the present study, the authors used meta-regression techniques to assess the effect of macronutrient intakes on sleep duration in six of the included studies from which they could extract the data required for their analysis. They performed this regression to determine if there was a dose-response relationship between percentage of energy from each macronutrient and sleep duration. The stats section of this paper is very concise to the point of not being sufficiently informative, in my opinion (it’s literally a single sentence describing their regression methods). Thus, we’re left to accept their regression analysis at face value, without having sufficient details to critique their choices related to the structure of the regression models.
The macronutrient distributions for those categorized as good sleepers and bad sleepers are shown in Table 3. As you can see, protein intake was higher across the board among good sleepers for every category. Most notably, the randomized controlled trial data showed much higher protein intakes among good versus poor sleepers: 30% versus 18.3% for sleep duration, 22.5% versus 16.7% for general sleep score, 31.1% versus 20% for sleep latency, and 33.8% versus 20% for sleep efficiency. Also, when just looking at the controlled trials, it seems fat intake was lower too, although the differences are less striking. With that said, when examining the broader data set of both cross sectional and controlled trials, there is not as clear of a pattern in the fat and carbohydrate distributions. However, as a function of consuming more calories from protein, the good sleepers generally had either a lower energy intake from carbohydrate, fat, or both.
In Figure 2, you can see the authors’ grouping of the good and bad sleepers’ macronutrient intakes in the randomized controlled trials in reference to the USDA’s acceptable macronutrient distribution ranges (AMDR). In general, most of the intakes fell within the AMDRs, except for a few study groups being a bit over on fat or under on carbohydrate. Among good sleepers, protein was higher in all cases and fat tended to be a bit lower, while carbohydrate had the most variance between and within groups.
Finally, the authors found no relationship in their meta-regression between sleep duration and macronutrient distribution in the six studies analyzed. However, as I mentioned, their explanation of the model-building process leaves us with a lot of uncertainty regarding key details, so take this specific finding with a grain of salt.
The basic interpretation of this review is pretty straightforward. Groups with better sleep quality and duration consumed higher protein in both cross sectional and longitudinal studies. The findings were actually stronger when assessing the controlled trial data, which makes me think this is probably a legitimate association. With that said, the authors thought it was worth pointing out that reverse causation could potentially be at play. Any time you establish a correlation between two variables, you don’t know which is the causative variable (or if, in fact, it’s some other variable impacting both variables you’re assessing). In this case, for example, are we sure a higher protein diet is causing better sleep, or does more sleep influence people to eat more protein? Indeed, there is data suggesting that poor sleep lowers satiation and increases hunger (3). Further, some data report poor sleep can result in an increased intake of fat (4) or carbohydrate (5), necessarily reducing the proportion of protein in the diet. With that said, I don’t think this is a case of reverse causation. The authors put forth reverse causation as a potential limitation, but I think they were just being cautious. As I mentioned before, the strongest relationships between protein intake and sleep quality were in the controlled trials, in which the researchers specifically implemented diet interventions to see the effect on sleep. Meaning, these studies specifically compared the impact of high, moderate, and low protein diets on sleep duration and quality. I think simply because this systematic review included 15 cross-sectional studies, but only four controlled trials, the authors were more conservative in their interpretation (in my opinion overly so).
With that said, being cautious is better than being too speculative or exaggerating data. So, I commend the authors for leaving no stone unturned when it comes to stating potential limitations. But while the authors over-delivered in stating limitations, I think they fell a little short in their meta-regression. As I mentioned earlier, a properly done meta-regression should include a number of details regarding how the models were constructed and a thorough description of the results. The way this paper was reported leaves us with a lot of uncertainty about some pretty important details. That said, based on the number of data points available for their meta-regression and a visual assessment of the plots, they were probably destined for non-significant findings regardless of the choices made in the model-building process. I think it’s safe to say that there probably isn’t enough data to establish a linear dose-response relationship between sleep duration and protein intake (if one exists) and make a subsequent protein intake recommendation.
Now that we’ve established that there is a potentially beneficial impact of high protein diets on sleep, we should try to understand why. Often understanding the mechanism at play influences practice in meaningful ways. The authors propose two possible mechanisms by which high protein diets might positively impact sleep. For our American readers, you will probably be aware of the first potential mechanism, as it’s something brought up annually by your uncle at Thanksgiving after everyone gets sleepy: “You know it’s the tryptophan in the turkey making us sleepy.” Now first, let me just say your uncle is wrong in the specific case of Thanksgiving dinner. It was probably the 2000-calorie meal and the alcohol making him sleepy, not tryptophan. Yes, the amino acid tryptophan is a precursor to melatonin, and yes turkey has tryptophan in it; however, turkey is comparable to other meats as a source of tryptophan. That said, where your uncle is sort of right is that an overall higher protein diet, in the ranges observed in the “good sleepers,” does lead to higher circulating concentrations of tryptophan compared to protein intakes representative in the “poor sleepers,” which might result in better sleep (6). However, to throw a potential wrench in the mix, not all data show a positive relationship between protein and sleep. Indeed, Santana and colleagues actually observed an inverse relationship between sleep quality and protein intake in elderly men (7), and the authors stated the possible explanation is that a high protein intake not only increases levels of tryptophan, but also large neutral amino acids which compete with tryptophan for uptake in the brain (6). However, Santana and colleagues also found a negative relationship with sleep for total energy intake and cholesterol. This makes me think the poor sleepers they observed consumed more fatty meats, and it was the high fat component of their diets that negatively affected sleep (more on this later), rather than the high protein component. With that said, the large neutral amino acids are valine, tyrosine, isoleucine, leucine, and phenylalanine (8); you might recognize valine, isoleucine, and leucine as our friendly neighborhood BCAAs that bodybuilders always find an excuse to take. Back in the day, a few of us nerdy bodybuilders tried taking leucine or BCAA between meals (often including after dinner, before bed) in an attempt to overcome the “refractory period” for muscle protein synthesis, AKA the “muscle full effect” (9). This was done in an attempt to make more of our day “anabolic.” I sometimes used to take BCAA pills before bed because I didn’t want to deal with preparing a meal or a shake. When I did this, I started having trouble sleeping, and I knew a few others who did the same and reported sleep issues as well. Again, this is just anecdotal, but it is interesting. Now, overall, I absolutely don’t think a high protein diet will negatively impact sleep (in fact, the data suggest the opposite), but I don’t recommend taking leucine or BCAA (and maybe not whey either) right before bed. Doing so could result in a fast appearance of these tryptophan-competing amino acids, which could theoretically impede sleep. But, to be clear, I don’t think this is likely to occur from a high protein diet or meal, which is digested much more slowly.
The second potential mechanism by which a high protein diet might cause better sleep is simply by making you less hungry. Most people who have dieted have experienced disrupted sleep due to hunger, at least at some point. Indeed, if you inject rats with ghrelin (a hunger stimulating hormone), they will show increased wakefulness due to suppressed sleep (10). As you have probably read or heard me discuss previously, high protein diets reduce hunger, and do so by suppressing ghrelin secretion (11), which could possibly prevent hunger-induced sleep disruption.
I should be clear that these two mechanisms haven’t been thoroughly examined in humans, nor established as the causative factor as to why higher protein diets are associated with better sleep. In the end, both are speculative mechanisms, but they do give us some plausible ways that a high protein diet could be helpful for sleep. Even in the worst case scenario – that this is just a non-causative correlation – we can at least have some confidence that high protein diets don’t negatively impact sleep.
Finally, while this systematic review specifically did not include studies of pre-bed protein consumption, it’s worth mentioning that this practice is probably not an issue. If you read the end of my interpretation in my review of pre-sleep protein consumption (here), you’ll recall there are data indicating that night time feeding can degrade sleep quality (12, 13). However, most data show this is related to high calorie meals and, specifically, high-fat meals. Thus, 30-50g of protein from low fat, low carbohydrate Greek or Icelandic yogurt, low fat meat, or a casein shake will provide 120-300 calories at most, coming predominantly from protein, and is unlikely to be an issue. Consuming a high protein diet, so long as you don’t take a large dose of leucine or BCAA (or maybe whey) right before bed, is unlikely to cause sleep issues, and it could even result in improved sleep.
As there were only four longitudinal studies in this review and the authors were concerned that we could be observing reverse causation, longer term trials that manipulate protein intake while observing sleep are needed. Once these are completed, a well-done meta-analysis and/or meta-regression would be great to see. Further, because the mechanisms are speculative, it would be helpful if tryptophan concentrations and levels of ghrelin were measured in these studies to see if they were related to sleep improvement. Finally, I’d love to see protein-matched comparisons of different dietary protein sources, including BCAA, whey, casein (or Greek yogurt or cottage cheese), and meats to see if certain protein sources were better at aiding sleep, or were potentially harming it.
If you sleep poorly and you consume a lower protein diet, it’s worth bumping protein to 25-30% of calories to see if it has any positive impact. Also, you probably don’t need to worry about pre-sleep protein negatively impacting sleep if consumed in isolation. However, I would avoid leucine, BCAA, and potentially whey consumption immediately before heading to bed, unless data comes out showing this practice isn’t harmful.