Maybe It's Quality Not Length That Matters

A widespread claim in popular media and health circles is that we live in an increasingly sleep deprived society, with technological intrusion into our lives, and greater working, educational, and travel demands. The academic world,^1–3 including this author,⁴ have helped drive this view, claiming to having identified a “hidden epidemic” and demanding attention of the public health agenda.⁵ However, several recent reviews have questioned this zeitgeist.^6–8 The paper by Calem and colleagues⁹ in this issue of SLEEP provides an interesting insight into both how to do this research well and where we should really be looking.

To adequately demonstrate secular changes in populations, epidemiological studies must fulfill a number of basic criteria. They must repeatedly sample from the same population representative sampling frame, using the same methods and assessment procedures, at the same time each year, obtaining high, unbiased response rates. In other words we are attempting to identify exactly the “cohort effects” many cohort studies try to diminish. It is preferable still if respondents are blinded to the study aims and “objective” measures of sleep duration are obtained.

The paper by Calem et al. fulfills most of these criteria. It uses data from the UK National Psychiatric Morbidity Surveys of community dwelling adults from 1993 to 2007. These data were restricted to those comparable between surveys—English, aged 18-64, and sampled in January to April. They used a symptom based approach to define outcomes of poor sleep quality including “insomnia.” Although not identical to DSM or ICSD insomnia criteria, the definition is similar and, importantly, identical in each survey. The assessment method varied a little over time but both interviewers and respondents were blinded to any sleep disturbance hypothesis. Response rates, as with community surveys worldwide, declined over time from 79.4% to 56.6%, leading to a likely reduced response from those with ill health and substance abuse.¹⁰ They observed respondents to be more educated and employed in later surveys, making their findings a conservative estimate.

Calem et al.⁹ demonstrate an increase in all definitions of sleep disturbance, the most marked being the near doubling of the prevalence of the severe “insomnia diagnosis” group, which rose to 5.8% in 2007. This confirmed a less methodologically robust Finnish study.² Furthermore, these increased rates were not attributable to changing associations of insomnia with any particular demographic factor, were not accounted for by any change in depression prevalence, and the reporting of other psychological symptoms did not increase in a similar fashion over this period. This latter counteracts the argument that we are merely seeing a rise in complaining.

The information upon sleep medication use is less informative because of self-report limitations, data quality and terminology, and changing patterns of hypnotic use. Notably, fewer than 1 in 50 with insomnia symptoms in this study were taking a hypnotic.

Well-constructed studies and acknowledgement of conflicting evidence are important in determining the legitimacy of claims of sleep disturbance epidemics. For instance, an elegant piece of investigative literature genealogy⁶ demonstrated significant citation bias in studies of shorter sleep duration trends in children. Among a total of 51 papers, only 17 contained any data, and only 6 of these showed consistent decreases over time. Those papers citing other evidence referred exclusively to original papers showing a decrease, and ignored conflicting evidence. The same authors⁷ undertook a systematic review of original data in the area. A complex Monte Carlo modeling with multiple data imputations suggested a worldwide average of 1-min per year decrease in children's sleep—hardly an epidemic. Some regions, such as Australasia and Europe, demonstrated an increase in average sleep duration. A systematic review of secular changes in adult sleep duration⁸ showed as many studies demonstrating an increase as a decrease, with conflicting evidence available for the United States. Many of the adult studies in this review⁸ did not meet the basic standards for study quality identified above, and Matricciani et al.⁷ do not provide the information with which to judge in the child studies. Some highly cited surveys that have different sampling frames year on year¹¹ do not meet these criteria.

So where does this take us? Firstly, Calem et al.⁹ demonstrate a robust epidemiological approach that should be adopted by the field. Second, it shows how we have missed an opportunity in the methods of many repeated epidemiological surveys that could have collected good sleep data, such as the National Comorbidity Surveys, but have been limited by the screen and module approach. Third, it suggests that we should be evaluating secular trends in other aspects of sleep than mean population sleep duration. Sleep quality, timing and variability, and actual diagnoses would seem the obvious ones. Although any shifts in the bell curve of a normally distributed phenomenon may have dramatic effects at the ends of the distribution, the at-risk groups can also change if, for example, the population distribution becomes less kurtotic (same mean but greater variance). It may be that we should be looking simultaneously at both the impact of disturbed sleep and short sleep duration, as these appear to interact in their effects.¹² As studies have shown marked inter-individual variation in sleep deprivation tolerance,¹³ an insomnia definition may partly capture this. A useful addition to the field would be repeated population representative studies using objective measures. This could, for instance, be obtained by the National Health and Nutritional Examination Survey (NHANES), which has been able to collect over 4000 participants' accelerometer data.¹⁴ There is no mention of any research to underpin the claims of population changes in sleeping patterns in the 2011 NIH Sleep Disorders Research Plan.¹⁵ As this states that “influencing public attitudes depends in part on the availability of evidence-based recommendations on the amount of sleep needed,” determining whether there are secular changes in duration, disturbance and/or quality and the causes and impact of these would seem a fundamental task.

DISCLOSURE STATEMENT

Dr. Glozier has indicated no financial conflicts of interest.

CITATION

Glozier N. Maybe it's quality not length that matters. SLEEP 2012;35(3):313-314.