Richard Sprague

My personal website

Sleeping better on a different diet

Richard Spraguer / 2018-03-06

After a lifetime of enjoying a banana or two almost every day, a microbiome test from DayTwo suggested that I stop eating them. It also said I stop eating most grains, and if I do, to eat them with fat. Generally the recommendations pointed toward me eating a more fat and fewer carbs. So for three months (Dec - Feb) I switched my diet to more of a low carbohydrate high fat (LCHF), with a goal of getting to under 25% carbs and 50% fat.

This was harder than I thought. I think of myself as a typical ominivore on a reasonably-healthy American diet: no sugary drinks or fast food, plenty of vegetables, mostly home-cooked with minimal processing. Sure, I make exceptions now and then, but my generally good health and weight level mean I don’t worry too much about what I eat. I thought skipping bread and rice would be enough, but I learned that carbs are everywhere. To get under 25%, I’m allowed only 65g/day. Just my morning latte is 15g. A single banana is 23g, and if you add one more apple (22g), I’m over my daily limit. You can easily add fat: a tablespoon of olive oil (13g) or a tablespoon of butter (11g) won’t seem like much food, but it adds up quickly. The problem, I found, is that fatty foods also tend to fill me up, so I eat less. And as a percentage of my calories, that means I’m allowed even fewer carbohydrates.

But I persevered. I’m not going to claim I was perfect – I’m doing this experiment for fun, not for serious scientific research – so there were days when I indulged differently. Still, just making a conscious effort was enough to push my eating habits in the general direction of less carbs, more fat. And soon I began to notice some effects.

Summary of the macro nutrients I ate the year before switching to my Goal diet.Summary of the macro nutrients I ate the year before switching to my Goal diet.

Figure 1: Summary of the macro nutrients I ate the year before switching to my Goal diet.

Three months into the new diet, daily tracking showed my sleep had changed slightly: I seemed to be sleeping longer and feeling better, an average of 25 minutes extra per night – a 5% improvement.

My sleep duration apparently increased by about 5% after switching to the new diet.

Figure 2: My sleep duration apparently increased by about 5% after switching to the new diet.

To find out, I compared four gut microbiome samples taken after starting the new diet with four taken before.

Table 1: Phylum level abundances before and after a change to a low carb, high fat diet.
Regular Regular Regular Regular LCHF LCHF LCHF LCHF
Firmicutes 43.76 67.37 62.27 64.54 52.70 51.03 51.58 52.78
Bacteroidetes 29.11 25.28 30.58 27.20 23.31 34.56 27.30 31.26
Verrucomicrobia 21.55 1.82 2.26 3.24 17.93 6.94 13.05 7.17
Proteobacteria 4.10 2.76 2.72 2.50 4.15 3.00 4.08 4.77
Euryarchaeota 1.76 2.51 2.14 2.39 1.80 3.35 3.76 3.00

Note the substantial difference in the level of Firmicutes. The gut microbiomes of nearly all Americans are dominated by this group of bacteria, which is known to play a role in how the body absorbs energy from food. In fact, it was once thought that obesity might be linked to large amounts of this microbe, though that’s been disproven. All of my LCHF samples seem to have Firmicutes just slightly over 50%, whereas three of the four “Regular” samples are much higher. Is that a coincidence? Is it statistically significant?

To find out, I looked at all my gut samples from the previous six months, when I was on my “regular” diet, a total of 72 samples. The average (mean) abundance of Firmicutes in those samples is shown in Table 2, plus the output of a simple T-Test that shows the likelihood that the difference could be due to chance alone.

Table 2: Mean Abundance of Regular Samples Compared to Low Carb/High Fat
Regular (Ave) LCHS
60.14 52.02
##  Welch Two Sample t-test
## data:  f$abundance and ff$abundance
## t = 6.2828, df = 65.746, p-value = 0.00000003024
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##   55410.56 107037.66
## sample estimates:
## mean of x mean of y 
##  601426.6  520202.5

Low p-values generally mean more significance, and in this case the extremely low number implies that the difference in sleep is unlikely to be due to chance alone. P-values have serious problems as a way to measure whether something is statistically significant, but as a sniff-test, the value in this case is so low that it’s worth investigating further. It would be hard to explain away the new, lower Firmicutes abundance as a coincidence.

Incidentally, among my large number of Regular samples there were a few with under 50% Firmicutes abundance, but were easily explained: for example, three of them happened after my Soylent experiment.

Conclusion: A low-carb high fat diet can affect Firmicutes abundance, and in my case it affects sleep as well. If I can figure out what’s driving this, maybe I can improve my sleep even more – or perhaps offer microbiome-informed suggestions for how you can improve too.