In this issue of the Journal of Clinical Sleep Medicine, Cook and colleagues present a detailed evaluation of Actiwatch-2 actigraphy settings and their impact on the accuracy of sleep time estimation in people with clinically-diagnosed idiopathic hypersomnia (IH).1 In their patients, a low sleep-wake activity threshold and an immobility time of 25 minutes resulted in the best concordance between actigraphic estimates of total sleep time and total sleep time as measured by polysomnography (PSG).
This work is an important step in validating current IH diagnostic criteria. By necessity, the history of IH diagnostic criteria has been one of increasingly specific “best guess” clinical phenotyping, with subsequent refinements as studies support or fail to support criteria. The International Classification of Sleep Disorders, Second Edition, introduced a Multiple Sleep Latency Test (MSLT) cutoff of 8 minutes with the acknowledgement that published values for patients with IH ranged outside this cutoff.2,3 Subsequent investigations confirmed the inadequacy of this MSLT threshold as the only objective criterion for IH,4,5 and the International Classification of Sleep Disorders, Third Edition (ICSD-3) incorporated measured sleep time of ≥ 11 hours6 based on PSG data.4 Allowing actigraphic estimation of sleep times ≥ 11 hours was a nod to practicality where 24-hour PSG is infeasible, in the stated absence of validation.6 Thus, the work by Cook et al., as a first step in this validation process, is extremely welcome and well-timed. Notably, this is not the only aspect of ICSD-3 criteria currently lacking validation for IH. Important questions also remain about measurement of sleep drunkenness,7 evaluation of comorbidities as causal versus coincidental,8,9 sleep efficiency relative to controls,10 and whether long sleep should be a separate clinical entity.2,6
There are multiple strengths to the work of Cook et al. In particular, allowing ad lib sleep during PSG is relatively uncommon in clinical practice but is clearly aligned with actigraphy's intended use. Despite the recommendation to wean patients off psychotropic medications “whenever possible” prior to MSLT,11 the reality of clinical practice is that this is often infeasible and this recommendation is frequently discounted.12 As such, by including participants on psychotropic medications, Cook et al. provide a more real-world assessment of actigraphic performance generalizable to practice.
Several issues of actigraphic validation for IH diagnosis remain. In the Cook study, PSG lights out and lights on times were used to set the Actiwatch rest period.1 In the controlled laboratory setting, bed time and wake time are closely tied to lights out and lights on, sleep is only allowed to occur between these time points, and these times are precisely known. In contrast, in the ambulatory setting where actigraphy is used, lights out and lights on time can be estimated from the actigraph's light sensor; however, these times may or may not align with bed, wake, or sleep times. Furthermore, assuming that sleep occurs between lights out and light on does not, for example, capture sleep time in a patient with hypersomnolence who falls asleep while watching TV. Daytime naps may also contribute a substantial amount to 24-hour sleep time in patients with IH. Thus, validation of different methods of defining major and minor rest periods in IH actigraphy is still needed.
Additionally, the use of actigraphy in the home setting and over multiple nights has yet to be validated for IH. A multitude of factors may affect actigraphy accuracy differently between the laboratory and home setting, including mattresses,13 bed partners, and pets. Nightly variability of periodic limb movements14 and other nocturnal movements may plausibly result in night-to-night variability in actigraphy accuracy. Finally, additional studies with large sample sizes are needed to determine the full range of individual variability in actigraphy accuracy with each device, despite an encouragingly close average difference from PSG of 3.4 minutes in the Cook study. In the much-larger insomnia literature, it is clear that total sleep times are accurately estimated by actigraphy on average (mean difference of only 9.5 minutes in meta-analysis of over 650 individuals), but studies have shown both over- and underestimation of more than an hour in individual participants.15 Application of actigraphy to individual patient diagnosis thus requires understanding of both average and ranges of accuracy. Many clinical factors are known to affect actigraphy accuracy in other conditions, including antidepressant use, body mass index, sleepiness severity, sleep efficiency, and periodic limb movements,16 so larger studies are needed to capture these effects in IH.
In addition to issues of validation, one of the most notable characteristics of the Cook study was the very high percentage of patients diagnosed with IH who did not meet ICSD-3 criteria. Out of 33 patients clinically-diagnosed with IH, only 13 (39%) had objective demonstration of hypersomnolence via MSLT or total sleep time. As correctly pointed out by the authors,1 the ICSD-3 does allow for diagnosis of IH based on clinical grounds alone when objective criteria are not met. However, according to the authors of the ICSD-3, this circumstance should occur “occasionally.”6 Other than the data from Cook et al., what is the evidence for how commonly this “occasional” occurrence takes place? In a consecutive series of 100 patients being evaluated for suspected central nervous system hypersomnia, 33% had normal MSLT results, compared to 22% meeting MSLT criteria for IH.5 The mean 24-hour sleep time in the normal MSLT group was 494 minutes, ie, many could not be diagnosed with IH by ICSD-3. In a series of 105 consecutive patients with clinically-diagnosed of IH, 47 met MSLT criteria, 28 met total sleep time criteria, while the remaining patients (29%) did not meet either.4 Taken together, it appears that a clinical phenotype of IH accompanied by normal MSLT results and sleep times is common. Whether this reflects heterogeneity in the diagnosis or a pure failure of current tools remains to be determined. Ultimately, it points to the need for ongoing validation of existing tools, development of disease-specific tools that capture the full phenotype, and identification of biomarkers to guide diagnosis and management of people with IH.
This work was supported by the National Institute of Neurological Disorders and Stroke (K23 NS083748) from the National Institutes of Health. Dr. Trotti reports no conflicts of interest.