Correlating Caffeine and Tiredness  

During a recent dig through our data, we found that we had access to information about whether someone had consumed caffeine during the day. As researchers, discovering a column like this is one of the best things that could happen during work. To uncover what was worth investigating, we ran a linear model across our activity and sleep columns fixing this caffeine flag as a dependent variable.

One of the mechanisms through which we regulate tiredness is adenosine—a neurotransmitter that latches onto to four different receptors in the central nervous system. Caffeine is an adenosine receptor antagonist; it latches onto the receptors that would otherwise be used by adenosine, making us feel less tired [1].

Intuitively, we understand this. Don’t have caffeine late if you’d like to sleep on time, down some coffee in the morning if you’re still sleepy. But what do the numbers say? What’s actually happening on a deeper level? Our investigation led to mixed results—some expected, others that may not be the most obvious.

Let’s start with the most obvious, you sleep less when you drink caffeine. In our sample this showed up as a 10-minute difference with a standard error of 2-minutes. Not surprising. We also looked at how the reduction of time spent in the different sleep stages after caffeine consumption; comparing this to the proportion of time spent in each stage for people who had not consumed any caffeine. Figure 1 shows this as a bar plot across all stages.

The order is clear, the order of impact on stages is REM, deep, and then light. In other words, the parts of your sleep that are most important are most affected by consumption of caffeine. It’s not just duration that’s impacted, it’s quality. This shows up in another one of our columns ‘num_rem_events’. We see a drop in the number of REM events during sleep across with a mean reduction of 0.2 and a standard error of 0.09. Clearly, caffeine consumption is not most conducive to a good night’s rest.

Effects of Caffeine on Sleep

But what about other effects? Let’s look at a somewhat surprising result; drop in blood oxygenation. In our investigation this showed up as a reduction of 1.92, with a standard error of 0.84. This is a big swing for blood oxygenation during sleep, where difference between good and bad typical values are separated by a narrow gap. It’s important to note that this may be an artefact of the fragmented sleep we talked about earlier. We investigated further by plotting the distribution differences among non-caffeinated nights and caffeinated nights. This revealed nothing major, however caffeinated distributions seemed to often have a slightly wider spread (Figure 2). When looking at individual users, this pattern did not persist. This further confirmed to that this blood oxygenation results was perhaps due to the fragmented sleep mentioned earlier.

For a final result, let’s take time spent awake in bed. We surprisingly found no substantial change to the time spent awake in bed. We saw this in two different ways. First, the difference of the caffeine effect between seconds spend in bed and asleep time differed only by five seconds. Second, the p-value for the caffeine effect on awake time in bed was ~0.9. Both results led to the same conclusion—there was no meaningful effect on time spent lying awake in bed.

Still, sleep is only one side of the coin; it’s important to look at activity too. About half the significant effects we discovered had to do with time spent doing an activity! We’ll cover this on another blog soon, so stay tuned!

References

[1] Reichert, Carolin Franziska et al. “Adenosine, caffeine, and sleep-wake regulation: state of the science and perspectives.” Journal of sleep researchvol. 31,4 (2022): e13597. doi:10.1111/jsr.13597