Month 50: caffeine #50to50

my mug of coffee

I drink too much caffeine. In college, I was notorious for drinking multiple cans of coke a day. In grad school, I made the switch to Diet and lost 10 pounds in a month. That was my peak Chipotle phase, too, clocking in around 175.

In the past few years, since moving to California, I’ve drastically reduced my caffeine intake from Coke (having made the switch to Coke Zero, which my crude palette likes as much if not better than original Coke, and miles better than Diet). I basically order it when we eat out, which is a few times a month, and on weekends once in a while. I estimate that I currently drink 5-10 cans a month.

However, I inhale coffee. I have at least two (large) cups in the morning (one on waking up either at home or at Starbucks, another on reaching work) and another in the afternoon (usually to stave off hunger from intermittent fasting). I also will make a cup in the evening if I need to stay up, and on weekends I typically have a few cups too. I estimate that I drink about 75 cups of coffee a month.

I can’t give up on coffee. I’m not really an aficionado of the taste – I can tell great coffee from terrible, but good vs bad coffee is not always clear. Low standards indeed, because for me it is purely functional. I did go nuts over Pumpkin Spice for a while but I’m well and truly over that now. My go-to drink at Starbucks is a flat white with sugar free hazelnut and vanilla, just enough flavor to keep me interested but not attentive. The bottom line is, I don’t get enough sleep, so I drink coffee.

However, caffeine after 2:00 pm has long been shown to interfere with sleep:

  • Caffeine consumed 0, 3, and 6 hours before bedtime significantly reduced total sleep time. Even caffeine consumed 6 hours before bed reduced total nightly sleep amounts by more than 1 hour.
  • Caffeine consumed at all three points diminished sleep quality. Caffeine taken 3 and 6 hours before bedtime, as well as caffeine consumed at bedtime, significantly increased the amount of time spent awake during the night.
  • Disruptions to sleep as a result of caffeine were perceived by volunteers (as recorded in sleep diaries) for caffeine consumed at bedtime and 3 hours before bed, but were not reported for caffeine taken 6 hours before bed. However, sleep monitors measuring total sleep time and sleep efficiency (time spent sleeping relative to total time spent in bed) showed that caffeine consumed 6 hours before bedtime had significant detrimental effects to both.

That third point is the most surprising. It means that even if you don’t perceive an impact on your sleep, there still may be one. I haven’t tried to replicate the study on myself because I don’t have/want a sleep tracker, but the point is compelling.

(Here’s a full-text link to the actual study)

This makes it easy to resolve to stop drinking caffeine after 2pm and limit myself to two cups of coffee a day. I’ll just switch to water (which I drink a lot more of since buying myself a cool hydroflask with a Joshua Tree design).

Table 3 from the study referred to above. Click for full-text link.

6 thoughts on “Month 50: caffeine #50to50”

  1. It’s possible, and for many people likely, but the linked paper isn’t persuasive. Their sample size was only 12 adults (4 additional data points were thrown out) aged 19-48 with self-reported data on baseline caffeine consumption habits and sleep quality, and they only tested each condition (0, 3, 6 hours before fixed bedtime) once. Also, I see no attempt to control for foods consumed before or during the study; it doesn’t even look like they had the participants keep an eating/drinking diary, and they were allowed unrecorded amounts of caffeine before 4pm.

    As a scientific study, this barely rises to the level of “anecdote”.


  2. I just spotted a real howler: in the table you reprinted, the “latency to persistent sleep” for the 6-hour test is reported as 45 minutes plus-or-minus 55 minutes. Never mind reporting the averages of 12 numbers with ridiculous 0.01-minute precision, some of the test subjects must have fallen asleep on the way to bed.

    The “wake time during sleep” is bogus in all columns; even the placebo column has 10 minutes plus-or-minus 15.


  3. i personally wouldn’t call the results bogus – data with a large standard deviation can often have a statistically meaningful correlation. The +/- isnt the whole story. Also, this is not the only study out there though on this (quite a few hits on PubMed) – though I included the direct link to the study, and the table of results, precisely so the reader can judge for themselves 🙂

    Note that they do not claim that the “Latency” result is significant. I dont see any fault in their fairly straightforward calculation of p-value. There is no p-hacking going on here, at the least.

  4. Ah, but how many of the other papers on PubMed also have incredibly tiny samples and basic methodological flaws? I’d hope this is an outlier (and a student project that got no better than a C, with the three PhDs just adding their names to pad it out), but what if they’re all this bad?

    Seriously, there are so few data points and so much variance that the entire thing could be the result of one college kid pigging out at Taco Bell the night of his 0-hour pill. Since the researchers don’t know what the subjects ate and drank (chocolate? sugar? alcohol? hash brownies?), they can’t rule out many other possible causes for the claimed effect.

    For instance, here’s a simple, obvious question: were the pills consistently taken with or without food? Anyone who’s ever taken NoDoz can tell you the answer matters.

    Is the result plausible? Yeah, sure. Does this paper provide any evidence that cutting off caffeine 6 hours before bedtime improves sleep? No, none at all; the methodology seems to be entirely based on what they could afford (only 16 sets of sleep monitors, used only one night for each test condition, with no control group).


  5. fair enough the reason that peer-reviewed studies have a Methods section is precisely so readers can make their own assessment. To be fair there are definitely studies out there that contradict these results from tis paper. In fact i saw a link to one yesterday that I meant to link, but I can’t find it now. That said, the context of my first resolution is to improve sleep and reduce my personal dependence on caffeine. To that end, the study gives me some needed encouragement to modify my habits. The ideal would be to buy a sleep-tracking device and perform my own study on myself, but i cant really justify a splurge on yet more electronics for purely egotistical reasons 🙂 The 2pm limit on caffeine is an easy rule that hopefully will cumulatively combine with the rest of these upcoming 49 resolutions to give me a better overall state of mind, body, and soul.

  6. I had a sleep tracker, went through a formal sleep study, got a CPAP, and it worked pretty well. Going to an 18/6 intermittent fasting schedule and buying a big box of Breathe Right strips worked better. Turned out my actual sleep problem was non-acid reflux, and not having solid food after 6pm reduced it to the point that the strips are enough to correct my chainsaw-like snoring and occasional apnea. Limiting and even giving up caffeine is one of the many things we tried that didn’t help at all, based on the sleep-tracking data. (the commonly-prescribed sleeping pills simply didn’t work on me)

    I really hope that the general quality of research in the field is higher than this paper suggests, but I fear that it’s probably about average. The three PhDs who added their names to it are all employed at a sleep clinic, which is worrisome.


What do you think?