The sleep loss epidemic: hunting ninjas in the dark

Bertrand Russell and Ludwig Wittgenstein once famously had an argument about whether or not there was a rhinoceros in the room. According to Russell, Wittgenstein contentiously refused to admit that there was not a rhinoceros in the small room they were in. One interpretation of this argument is that Wittgenstein refused to rule out the rhinoceros on the basis of what we scientists would call the impossibility of proving a negative. While empirical data do not currently support the hypothesis, it is always possible that some might come forward. In sleep epidemiology we have our own rhinoceros in the room problem: the ‘sleep loss epidemic’. There is a pervasive feeling in our scientific community and widely reported in the media that humans get less sleep than we used to. The narrative is that as a society we have become chronically increasingly sleep-deprived since the halcyon-days of the 1950–60s—one just has to forget the warm comforts of the Cold War and the barbiturate, benzodiazepine and amphetamine epidemics (Rasmussen, 2009). The sleep loss epidemic is a little more subtle than the rhinoceros problem, however. It is more like a Ninja in the room. Certainly, no one can actually reference any high-quality data (Matricciani et al., 2011), but the whole concept simply intuitively ‘feels’ correct. The Ninja is hiding in the room somewhere: we just have to look harder and ignore all evidence to the contrary. Much like Russell's rhino, however, if the Ninja is really not there it is going to be impossible to prove it (additionally, Ninjas have a reputation for being very good at hiding). Our research group, for instance, has failed to locate the hidden epidemic. We looked worldwide using a systematic review to identify all previously published nationally representative repeated cross-sectional studies of sleep duration, and found data from 12 countries dating back to the 1960s (Bin et al., 2012). In six countries sleep durations declined, two had mixed evidence (including the United States) and we found another seven countries where sleep has actually increased (including Britain 1961–84). Reviewers helpfully pointed out to us that perhaps the average sleep duration did not matter, but perhaps there was an increasing prevalence in people sleeping fewer than 6.5 h per night? Some papers in our systematic review had indeed reported that phenomenon. Next, we used Multinational Time Use Survey data because of the resemblance to sleep diary data, and collated information from 10 industrialized countries stretching back from the 2000s to the 1970s in 328 018 people to investigate whether the prevalence of either short (≤6 h per night) or long sleep (>9 h per night) had changed (Bin et al., 2013). Short sleep had increased in Norway and Italy, but had decreased in the United States, the United Kingdom and Sweden, while remaining stable in Australia, Finland, Canada and Germany. Conversely, long sleep had increased in Australia, Finland, Sweden, the United Kingdom and the United States, while decreasing in Italy and Canada and remaining stable in Germany, the Netherlands and Norway. The results are similarly mixed in children, with slight increases in sleep duration and slight decreases in others, but stable with another group of countries in between (Matricciani et al., 2012; see Figure 4 in that paper). All this, despite the rise of the internet, 24-h television and the mobile blue-light-emitting personal computing power now built into our mobile phones (Tavernier and Willoughby, 2014). So why does it feel as if we have lived through a decline in sleep loss? Why does this feel so wrong? It is probably because our personal experience is that we have: but this is an effect of ageing, and not a secular effect (Ohayon et al., 2004). We also have important memorable times in our lives when sleep loss occurs: first job, shiftwork, babies, illness, caring for people who are ill, bereavement and grant-writing season, for example. The media and sleep experts are also constantly informing us that it is true. There are also some cherished hypotheses in sleep that require the existence of the sleep loss epidemic: sleep loss causes the obesity epidemic, for instance. We are personally and professionally primed to see it all around us, but psychologically less eager to see the stable or even improving data. I think there are some possibilities that we have not yet collectively ruled out. It is possible that there has been a sleep restriction epidemic in transitioning siesta cultures (see Italy), and so I would certainly encourage people to publish data from countries such as Spain, Greece or Mexico. The other intriguing possibility is that sleep timing has changed systematically. With the rise of blue-light-emitting devices, both our bedtimes and rise-times may have become later and later. Even if this is true, I would still wonder ‘so what?’, because without linking this purported societal shift to a worsening in health, functioning or quality of life one cannot conclude automatically that this change has been detrimental. As Arthur Conan Doyle might have made Sherlock Holmes phrase it: ‘Data! Data! Data! I cannot make bricks without clay!’ (Doyle, 1892/2011). So why did we not find the ninja that everybody assumes is there? In response to good UK data indicating a secular increase in insomnia and hypnotic use (Calem et al., 2012), my colleague Nick Glozier proposed an explanation for the apparent discord between the lack of a detectable sleep epidemic but the widespread belief that sleep somehow is less than we used to get: it could be that sleep quality has worsened (Glozier, 2012). For instance, the obesity epidemic is probably driving up the prevalence of obstructive sleep apnea (Peppard et al., 2013). By itself, that seems fairly strong evidence that sleep quality has worsened. In this issue of the Journal of Sleep Research, Kronholm et al. (2015) have reported that they have found the sleep quality epidemic and they have linked it to an important functional outcome that society actually cares about. They present meticulous analyses of two cohorts providing very near-universal coverage of the adolescent population of Finland for the years 1984–2011 (n = 1.13 million). They observed a notably rapid doubling in the prevalence of a range of subjective sleep quality complaints from the mid- to the late 1990s in both genders and all ages in the population (14–17 years). The epidemic seems to have levelled-off in the 2000s at this new higher level, but has never recovered to pre-epidemic levels. The important external outcome is that self-reported grade-point averages of the respondents show that while school performance has increased during the decades this effect has only been enjoyed by those who are not chronically tired during the day (Fig. 3a,b). There are always caveats in research studies of this scale: the key ones here, as I see it, are that the exposure (sleep) and the outcomes (grades) are all self-reported. Secular shifts in the way people think about and respond to these surveys may explain the pattern. Our public health efforts at warning the population about the importance of healthy sleep (and mental health awareness campaigns) may have actually worked, and people are potentially more willing to admit that they have a sleep or tiredness problem now than in the past. However, as the authors point out, increased poor sleep may also relate to increased psychiatric morbidity in the adolescent population, and these data indicate that something serious may have occurred. Dr Kronholm and his colleagues are meticulous data analysts, and not prone to overinterpreting their studies. They warn us that despite this link between tiredness and grades being statistically significant, it is quite a weak association, accounting for only approximately 1% of the variation despite, the notable pattern seen in Fig. 3a,b. My additional comments are mainly to recommend that you read it—but also not to over-read it: by necessity of the particular data sets used being of limited geographic, ethnic and age ranges. That limits external validity when talking about a possible worldwide sleep loss epidemic. It is a single but important brick, and not the whole wall. A mathematician might warn us that we have only found one-half of one black sheep in Finland (http://en.wikipedia.org/wiki/Mathematical_joke). Sleep is a complex psychobiological phenomenon, and when it is self-reported, probably even more so. Social and economic conditions or changes in the education system might have driven this simply in Finnish adolescents. Indeed, if one looks at the rise in income inequality in Finland using the widely accepted Gini-coefficient method, the year-by-year graph of this rise throughout the 1990s looks strikingly similar to what Kronholm and colleagues report (Official Statistics of Finland, 2012). My point here is that sleep can be viewed as a social justice issue, where the poor and disadvantaged also suffer economically from the greatest lack of sleep security. The lack of a confirmed global sleep loss epidemic does not mean that insufficient sleep or sleep disorders cease to be a public health problem (Institute of Medicine, 2006). Perhaps instead we should celebrate that the public attitude to sleep on the whole is positive, and that society as a whole, through the rise of the internet and mobile computing, has in fact managed to maintain sleep duration. Conversely, sleep quality and timing remains an open question, and it may be that our young people or other vulnerable groups are bearing the brunt of this effect. I encourage others to publish their data, positive or negative, to try to gain a greater global perspective on this intriguing paper. Secular trends in sleep duration and health research at the University of Sydney has been supported by (NHMRC grants 571421, 1004528 and 1060992). I would like to thank Derk-Jan Dijk for his critical review of the draft of this Editorial. No conflicts of interest declared.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.