Good sleep is important for health and optimal function. That much is pretty clear from the totality of sleep research. Lack of sleep also exacerbates many chronic conditions, can be a trigger for things like migraines and seizures, and correlates with poor cognitive function and memory. Having good sleep habits, and good overall sleep is therefore an important lifestyle factor for health and function.
A new study adds to this data by showing a correlation between decreased sleep duration and chronic illness in older adults (>50), but the data is complex. Most mainstream outlets oversimplified the outcome of the study in a way similar to this BBC headline: “Five hours’ sleep is tipping point for bad health“. This is ultimately a misunderstanding of how statistical significance works, but let’s review the study itself: “Association of sleep duration at age 50, 60, and 70 years with risk of multimorbidity in the UK: 25-year follow-up of the Whitehall II cohort study“.
The Whitehall II cohort study follows a group of UK civil servants starting in 1985. The strength of this data set is the long duration, several decades, and size (it started with 10,308 individuals). These are large numbers for a study, but not really for an observational study such as this. Experimental studies where some variables are determined typically involve 50-100 individuals at the low end, with larger studies in the hundreds and rare very large studies in the thousands. Observational studies come in various types but they look at people in the world without determining specific variables. Observational studies can have tens of thousands to millions of individual subjects, depending on their methods, and can also follow subjects for long periods of time.
It is generally difficult to make causal conclusion from observational studies, which can only determine correlations by themselves. We can say, for example, that some potential risk correlates with a negative outcome based on observational data, but we cannot say it causes it because subjects were not randomly selected to be exposed or not to the risk. There are therefore multiple confounding factors and possible causal relationships.
So while this is a decent-sized cohort study, it is not a particularly large one, and since it is observational we have to be on the look out for possible confounding factors. Let’s look at what they found:
Compared to 7-hour sleep, sleeping ≤5 hours at age 50 was associated with an increased risk of a first chronic disease (1.20, 1.06 to 1.35; p = 0.003) and, among those who developed a first disease, with subsequent multimorbidity (1.21, 1.03 to 1.42; p = 0.02). Sleep duration ≥9 hours was not associated with these transitions. No association was found between sleep duration and mortality among those with existing chronic diseases.
Multimorbidity simply means two or more diseases out of the 13 chronic illnesses they tracked. Therefore, sleeping less than 5 hours a night had an increased risk of developing chronic illness but not of dying when compared to those who sleep 7 hours a night. This does not mean that 5 hours is a “tipping point,” it just means that the results did not become statistically significant until this point. It is possible, even probable, that the risk of association is continuous. That is how the researcher plotted their data, although they did not have very much resolution as they broke down sleep duration into one hour increments.
You may also wonder why <5 hours is compared to 7 hours. That’s because 7-hour sleep duration had the lowest morbidity association. The relative risk curves are actually U-shaped, with increasing risk for 8 hour and >9 hour sleep durations (although did not reach statistical significance primarily because of the small numbers of subjects in the >9 hour category). The U-shaped risk curve is interesting, and may suggest a strong possibility that at least some of this risk association has a reverse causality – chronic illness causes poor sleep, rather than the other way around. This cannot be excluded by this study.
In fact clinicians often think of poor sleep as “comorbid” with specific chronic illnesses – they likely worsen each other, and it is difficult to tease out the specific cause and effect. For example, poor sleep can exacerbate weight gain, and being overweight can worse sleep quality.
Another issue is that the sleep duration was mostly self-reported, with some data from subjects wearing sleep trackers. Sleep duration is not the only measure of overall sleep benefit. It’s possible that some of the long sleepers are staying in bed for 9 or more hours because their sleep quality is very low.
What about the fact that in this study (because of how it was conducted) short sleep duration preceded the development of chronic illnesses? This does not preclude confounding factors. Poor sleep can be an early symptom of a chronic illness. This is true for some neurological diseases, like dementia and Parkinson’s disease, for example. Also, other factors may cause poor sleep and chronic disease, such as being overweight, lack of physical activity during the day, or chronic conditions not on the list followed by this study.
There are a few take-aways from this study and other sleep research. First, given the totality of data it is reasonable to conclude that poor sleep is a risk factor for many chronic illnesses, exacerbates many symptoms and can impair function. None of my criticisms of this study or its reporting change that. For any individual, optimizing their sleep quality is likely to have health benefits.
For many of these lifestyle risk factors that are difficult-to-impossible to control for in an experimental setting, and therefore we have to do our best with observational data, we tend to take a practical approach. This gets back to the “comorbid” approach – I cannot necessarily tease apart the exact lines of cause and effect, so essentially we treat everything in hopes of reducing the complex of comorbidities – optimize sleep and lose weight and treat the depression, for example.
Also, if your sleep duration or quality is poor, or you do not feel rested in the morning, see your physician. This can be a primary sleep disorder, a risk factor, or it can be a symptom of an existing chronic illness. Either way, poor sleep is just as important to discuss with your health care provider as other chronic symptoms.
Further – observational data is complex, with many possible arrows of causation and confounding factors. Be skeptical of simplistic reporting about correlations found in such studies. This study does not show that 5 hours of sleep is a “tipping point” for chronic illness. It shows a complex correlation between mostly self-reported sleep duration and the risk of developing certain chronic illnesses, and provokes as many questions as it answers.
Follow up research can include more objective measures of sleep duration and quality, higher resolution data, larger cohorts with a more representative sampling, and gather data that can help address some of the possible causal relationships. But no single study is going to answer such complex questions. The relationship between sleep and chronic illness is one of those questions that will take dozens of high-quality studies to flesh out. This study is a good addition to the overall data, but we have to recognize its limitations.
