- Top of page
- Materials and Methods
- Ethics Committee Approval
- Authors' Contributions
A number of associations between sleep problems and attention deficit hyperactivity disorder (ADHD) have been described (Cortese et al., 2009; Sadeh et al., 2006). These include sleep-onset problems, sleep phase delay syndrome, increased movements in sleep, daytime sleepiness and altered total sleep time (Corkum et al., 1998, 1999; Cortese et al., 2006; Konofal et al., 2001, 2010; Mayes et al., 2009; see Cortese et al., 2009; Sadeh et al., 2006 for reviews).
However, many of the studies come from cross-sectional studies of school-age children with limited controlling for confounding or consideration of the longitudinal sleep trajectory. Interpretations of these associations include poor sleep quality or quantity being part of a causal mechanism of ADHD, ADHD and its treatments causing sleep problems, or that the two share a common aetiology (Gringras et al., 2007).
Prospective cohort studies offer the advantage that developmental trajectories of sleep and their temporal relationship to the onset of ADHD symptoms can be determined, and thus offer more insight about possible direction of causation. Thunstrom (2002) reported a 3-year follow-up study of 27 children with chronic sleep problems, of which seven were subsequently diagnosed with ADHD symptoms at age 5.5 years. Touchette et al. (2007) reported from the Quebec Longitudinal Study of Child Development (Canada) that sleep duration of less than 10 h per night, especially before the age of 41 months, was associated with hyperactivity/impulsivity symptoms and lower cognitive functioning at age 6 years. Cohort studies to date have predominantly considered pre-school sleep patterns, and relied on screening questionnaires for hyperactive symptoms, rather than standardised psychiatric measures for ADHD.
We used prospectively collected data from the Avon Longitudinal Study of Parents and Children (ALSPAC) to investigate sleep patterns and trajectories from 6 months after birth to 11 years old, and their relation to ADHD diagnoses established by standardised psychiatric interviews according to DSM-IV criteria (American Psychiatric Association., 2000).
- Top of page
- Materials and Methods
- Ethics Committee Approval
- Authors' Contributions
In all measures of sleep duration and continuity children with ADHD slept for a shorter time, and woke more than their peers. The main reason for their shorter sleep durations was later bedtimes. The effect sizes were very small, little strong evidence of cross-sectional differences from controls at most time points. The age when children diagnosed with ADHD slept significantly less than their peers seems to be about 5–9 years old. Although children with ADHD slept less than their peers, and stopped requiring a daytime sleep at an earlier age than their peers, we did not find strong evidence of differences in daytime sleep duration. This finding does not contradict the objective evidence reported of increased sleepiness on formal multiple sleep latency testing, which measures sleep pressure during carefully confined conditions, rather than actual daytime sleep duration (Cortese et al., 2006).
A different strategy, adopted in many other clinical domains, of identifying falls across normative centiles in longitudinal data more commonly predated the diagnosis of ADHD, and we suggest this approach warrants exploration in future studies. We found that considering a 1SD fall in z-scores for sleep duration between two adjacent time points was a significant predictor in identifying children later diagnosed with ADHD for six of the seven adjacent time points.
We have previously shown that although children's sleep duration tracks their early sleep duration centiles over short periods of study, this relationship steadily weakens over time (Blair et al., 2012). Thus, as well as a remarkable natural variation in sleep duration throughout childhood, individual children do not remain ‘short’ or ‘long’ sleepers throughout childhood, but ‘move’ between centiles. We suspect that this accounts for the relatively small differences in sleep durations when looking at cross-sectional, absolute sleep durations at every age. However, when measuring changes relative to the individual, the magnitude of falls in sleep duration are more striking clinically. For example, between the ages of 42 and 69 months, a 1SD fall represents a reduction in sleep duration of more than 1 h and 20 min.
The strength of this study is that the DAWBA is a validated diagnostic tool used at a late enough age to allow a confident diagnosis of ADHD. Unlike other studies we did not rely on screening measures for ADHD traits at a young age that arguably may be conflated with sleep problems, or even represent sub-threshold and clinically normal behaviours rather than true ADHD.
There are a number of limitations that need consideration in this study. Although the parent DAWBA interview enquires about behaviours in schools and other environments, ideally there would have been parallel DAWBA teacher assessments on all the children. However, there were practical difficulties in encouraging all the teachers to complete a standardised assessment within a limited time window. Data from a UK population-based study that employed the DAWBA suggest that this potentially results in an underestimate of the prevalence of ADHD (Ford et al., 2003). Thus, we might have underestimated the numbers of children with ADHD, although in fact our overall ADHD prevalence (2.1%) is not out of keeping from that reported from the UK 1999 survey (2.23%) of over 10 000 children (Ford et al., 2003).
We were also reliant on subjective (parental reported) measures of sleep patterns and sleep behaviours. There are often differences between sleep difficulties reported by parents and those shown on objective sleep measures (Cortese et al., 2009). This study did collect contemporary sleep reports, reducing likelihood of any recall bias, and had a large enough sample size to allow comparisons with typically developing peers. For a study of this size, objective assessments over time are extremely difficult. In the paper (Blair et al., 2012) we have discussed the limitations of relying on subjective sleep data, but also why the data are still helpful and reliable. In this particular case the subjective nature of the data means that although children with ADHD are reported as sleeping less, this does not take into account their actual, objective sleep efficiency. Total sleep duration was calculated from parental reports of estimated bedtime, wake-time and daytime naps. This did not account for sleep latency, as an estimated bedtime did not differentiate between when the child goes to bed and goes to sleep. Also, night-time sleep duration did not take into account duration of any waking events. If one extrapolates from polysomnographic and actigraphy data, we feel it is likely that the reduced sleep duration and increased night waking reported is likely to be subjectively and objectively real.
A limitation of most longitudinal studies conducted over several years is that missing data and loss to follow-up are more likely in the most socioeconomically deprived groups. The ALSPAC study is no different, but of sufficient size that although some of the vulnerable groups were lost through this attrition, enough families remained in the study to differentiate between even relatively small social groupings.
We have not addressed whether psychosocial problems in the family might be associated with poor sleep and ADHD in children (Thunstrom, 2002). Dorris et al. (2008) suggest that it is important to consider parental limit setting and contingency management when assessing and managing sleep problems in children with ADHD. It is important to note, however, that in a study considering sleep hygiene and bedtime routines, these were equally well implemented for children with ADHD as for other children (Van der Heijden et al., 2006). In a review of 22 longitudinal studies, Hemmi et al. (2011) concluded that regulatory problems in infancy, including sleep, were associated with externalising behaviours and ADHD.
We do not have accurate information on which children were prescribed stimulant or other medication for their ADHD. At the time of the study it is very unlikely in the UK that medication treatment with stimulants would have commenced before 6 years old. The UK has historically treated far fewer children with ADHD than other countries. In 1995, for example, only 0.03% of children in the UK were on medication for ADHD in contrast to 3% in the USA and 1.7% in Australia (Parliamentary Office of Science and Technology., 1997). Mayes et al. (2009) and others have noted that medicated children had greater difficulty falling asleep than unmedicated children, although they suggest the medication prescription might be a proxy marker for ADHD severity. This uncontrolled factor might therefore have influenced the older children with ADHD in the cohort.
Comorbidities commonly associated with ADHD include anxiety, depression oppositional defiant disorder and autistic spectrum disorders (ASD; Faraone et al., 1998; Greene et al., 2002; Jensen et al., 1993; Wilens et al., 2002), but this study is underpowered to consider the impact of comorbidities in ADHD on sleep disturbance. The exclusion of children with ASD from DSM-IV (American Psychiatric Association., 2000) ADHD is arguably a weakness of this version of the diagnostic classification system, as in clinical practice the two are often comorbid. Our data on sleep trajectories in autism will be presented separately.
Recent research has shown links between ADHD and restless legs syndrome (Cortese et al., 2005; Konofal and Cortese, 2005). As this is a relatively recent finding, we did not collect data on this in our study. This may, however, be a potential area for further investigation by others.
In a recent review paper, Corkum et al. (2011) highlighted the importance of assessment and treatment of sleep issues in children with ADHD. We have demonstrated on subjective measures used here that the sleep disturbances begin in infancy and that significant reductions in sleep duration at most ages predict an increased risk of ADHD. Sleep duration appears to be another potential endophenotype for ADHD that might help in both early identification for potential interventions, and also to better understand possible mechanisms. Clinically we feel there are enough normative data to allow developmental paediatric clinics to usefully collect and track sleep trajectories in the same way as other growth and physical parameters. Significant and consistent perturbations for an individual should be carefully considered in the overall developmental context for that child. We do not feel the magnitude of difference in total sleep time we determined is enough to be considered causal in the evolution of the ADHD, although we acknowledge the growing body of evidence suggesting that smaller increase in total sleep time may be worthwhile over cumulative nights (Jan et al., 2010). Our preference is to speculate that these early sleep peturbations are subtle early markers in some children that reflect a shared underlying pathophysiology between sleep and ADHD.