Napping
One of the minor frustrations of the information age is that health advice is frequently inconsistent.
For example, if you Google the phrase "cnn.com napping", here are the first four hits you'll see today:
"Napping regularly linked to high blood pressure and stroke..."
"National Napping Day: 5 reasons you should take a nap..."
"Excessive napping could be a sign of dementia..."
"Napping could keep older adults' brains sharp..."
A little confusing, right?
This newsletter focuses on the first hit, which describes a study, published Monday, claiming to show that napping is bad for you, because it increases your risks of hypertension and stroke.
Since napping is one of my favorite hobbies, I was eager to look this study. What I found is terrifically bad statistics, as well as a nice illustration of why experts often disagree about healthy living.
Napping: The mainstream view
In spite of what you might find in a Google search, nearly two centuries of sleep studies have led experts to at least some consensus about napping. Authorities such as the National Sleep Foundation, the CDC, and the Mayo Clinic offer a fairly consistent message. Here's how I would summarize it:
People differ in their sleep needs. Napping may not benefit everyone.
For most people, napping can reduce fatigue, increase alertness, improve mood, and enhance learning and behavior.
Excessively long naps can cause sleep inertia (disorientation and lingering grogginess after waking up) and disrupt nighttime sleep.
Experts also agree about how to increase the benefits of napping:
Keep your naps short (10 to 20 minutes).
Nap in the early afternoon (before 2 or 3 p.m., when there's a natural dip in alertness, and when naps are less likely to interfere with nighttime sleep).
Try to nap in relatively cool, dark, quiet conditions.
Finally, the experts offer some interesting hacks. Here are two:
Drinking a small amount of caffeine immediately before napping may be beneficial. Because caffeine needs about half an hour to take full effect, taking some before a nap may keep you from oversleeping, and you're more likely to wake up alert.
Although short naps are ideal, longer ones (60-90 minutes) may be beneficial too. Medium-length naps (e.g., 45 minutes) are risky, because you're more likely to wake up from a deep-sleep stage and experience sleep inertia. Same goes for excessively long naps.
Nap study overview
The "nap study", published this July 25 in the American Heart Association journal Hypertension, contradicts the mainstream view on napping, because it links napping to an increased risk of hypertension and stroke. The study has gotten a lot of press. If you Google phrases like "Is napping good for you?" or "Is napping bad for you?", you'll see, among the first few hits, links to USA Today, CNN, ScienceDaily, Prevention, and Health articles that cover this study. In spite of the attention, I don't think the findings are credible.
Methods
Participant data were obtained from the UK Biobank, an extraordinary resource representing over half a million older adults (ages 40 to 69) who agreed to be studied from 2006 until at least 2036. Participants regularly answer questions about their lifestyle and provide blood, urine, and saliva samples, as well as access to their health records. The "nap study" focused on 358,451 individuals who, in 2006, had no prior history of hypertension or stroke. Overall, the approach to sampling was strong, in spite of a somewhat narrow demographic focus (i.e., Europeans over 40).
On four occasions between 2006 and the present, participants were asked whether they nap during the day. (This was one of many lifestyle-related questions.) The three answer options were "never/rarely," "sometimes," and "usually." Three-quarters of participants provided the same answer throughout the duration of the study, and less than one percent changed from one extreme ("never/rarely" or "usually") to the other.
Findings
The main finding was straightforward: "Sometimes" or "usually" taking a nap predicted higher rates of hypertension and stroke.
Specifically, compared to people who never or rarely nap, those who "sometimes" nap were 7% more likely to develop hypertension, and 12% more likely to experience a stroke, while those who "usually" nap were 12% more likely to develop hypertension, and 24% more likely to have a stroke.
These group differences were observed after controlling for age, sex, body mass, exercise habits, alcohol intake, smoking status, prior cholesterol levels, and other key variables. They were also observed when people at high risk for hypertension and/or nighttime sleep disturbances were excluded.
Since none of the participants had experienced cardiovascular problems prior to the study, and since so many pertinent variables were controlled for, the researchers concluded that napping may be a causal risk factor for hypertension and stroke. In other words, they concluded that napping may cause more susceptibility to high blood pressure and stroke. Naturally, the media picked up on this important and scary conclusion. But, if you're a napper, don't be alarmed just yet...
A key limitation
As you may have noticed, the researchers measured napping behavior superficially. Participants were simply asked whether they "rarely/never", "sometimes", or "usually" take naps. That's it. Participants were not asked about the duration of their naps. They were not asked about the quality of their naps. They were given no guidance on what terms like "sometimes" and "usually" mean.
Because napping was measured so poorly, I don't trust the results. But let's dig a little deeper. After all, the napping variable did predict hypertension and stroke. If we're going to pick on the way napping was measured, we ought to ask why this variable still had some predictive power.
In a word, I think the answer is: fishing.
That's a pretty strong statement – it's more of an accusation than a statement – so I want to explain my rationale as clearly as possible.
Evidence of fishing
The U.K. Biobank database contains information on literally hundreds of variables; there's no particular focus on sleep and cardiovascular health. None of the study's seven authors has a history of researching this topic. Rather, they're all medical experts from universities and centers in China, where they work within fields such as genetics and aging. In short, it's unclear how the researchers arrived at the particular focus of this study. Meanwhile, the statistical approach they used did not allow for any variable other than napping to predict hypertension and stroke. The researchers controlled for other variables, like exercise, but they didn't report whether those other variables had any association with cardiovascular outcomes. My conclusion is that the researchers went fishing for significant effects and found one they liked. It's easy to find significant effects in a sample of 358,451 people, because you're fishing in a very well-stocked lake!
Bad statistics
Even if I'm wrong about the fishing, the statistical approach in this study was terrible.
If you have a background in stats, you'll recognize a couple of classical mistakes in regression modeling. (If you're not a stats person, please be patient with the next sentence.) The researchers' primary model focused on a single predictor (napping) along with numerous covariates, but they failed to describe the impact of those covariates, and they never addressed the possibility that a different model might've revealed no effects of napping.
My concerns here pertain to something known as model specification. A "model" is just a set of variables presumed to be related in some way. In this study, the researchers tested a model in which napping leads to hypertension and stroke. The researchers controlled for other variables, but they didn't describe how well any of those variables predict the cardiovascular outcomes. However, those variables may have been important. They may have been so important that napping doesn't matter at all.
For example, consider exercise. Using the same database, we might find that after controlling for napping behavior, people who exercise more are less likely to experience hypertension or strokes. Here's another possibility: Using the same database, we might find that if exercise and napping are compared, exercise predicts the cardiovascular outcomes, but napping doesn't. The researchers didn't examine these possibilities. Certainly they should have. The same can be said for other variables known to influence cardiac health (diet, smoking, alcohol intake, etc.).
The problem here, crudely speaking, is that the researchers didn't give any variable other than napping a chance to predict hypertension or strokes.
In short, the statistical modeling was badly done. Although other studies have linked excessive napping to a variety of health problems, there's no clear evidence that healthy napping increases your risk of hypertension or stroke. I see no evidence of that in the literature or in the present study.
Why does health advice so often seem inconsistent?
This is a complicated question. Here, in lieu of a complete answer, I'll just share two thoughts:
1. Inconsistent health advice has achieved the status of a meme. (Eggs used to be good for you. Then they were bad for you. Then they were good for you again. But some experts disagree…) Even so, we may overestimate the extent of inconsistency. Research in all fields, not just health, leads to inconsistent advice. Is this phenomenon really more prevalent in the area of health? It's hard to say. We might just be more attuned to inconsistencies here, because health advice is personally relevant and, sometimes, a little scary. Most of us are less troubled by conflicting views on topics like why the dinosaurs became extinct.
2. A theme I've aired repeatedly in newsletters on diet, exercise, longevity, alcohol, and red meat consumption is that measurement in health research is often inherently imprecise. When behaviors like eating, exercise, sun exposure, alcohol intake, sleep, etc. need to be recorded for extended periods of time, researchers have to rely on participant self-report rather than direct observation. Unfortunately, people sometimes lie, or make mistakes, and even when they get it right, the questions presented to them tend to be brief and superficial, as in the napping study. No wonder health researchers end up with conflicting findings.
Why would researchers use brief, superficial questions? Outcomes in health research – hypertension, cancer, obesity, longevity, etc. – are known to be influenced by many variables, and the more variables you measure, the less time you can devote to measuring any one of them. Someone who visits a UK Biobank station for an hour will answer a variety of questions about their lifestyle and provide blood, urine, and saliva samples. Researchers can't afford to spend 20 minutes on detailed questions about napping behaviors.
In short, if research-based health advice is especially inconsistent, the culprit may be imprecise measurement of key variables.
The solution to this problem, in my view, is to either improve the measurement or forego the research. In the case of the nap study, participants should've at least been asked about the duration of their naps, and how many days a week they nap. That's only two more questions. The data would've still been terribly imprecise, but the researchers would've at least been able to identify participants who take excessively long naps most days of the week – a known correlate of emerging cardiovascular problems, and perhaps the source of the significant effects reported in this study.
I hate to think of elderly people giving up their afternoon naps because they've read that napping is dangerous for them. Either they read this, or someone tells them, but either way they don't realize it's bad advice stemming from flawed statistics.
Rest assured, healthy napping is good for you. I personally look forward to the next 20 minute conversation with my pillow!