The multiple sleep latency test and Epworth sleepiness scale in the assessment of daytime sleepiness

Dr Murray Johns' recent article argues that the Epworth Sleepiness Scale (ESS) not only costs about 1000 times less than the Multiple Sleep Latency Test (MSLT) but also serves as a superior gold standard measure of excessive daytime sleepiness (Johns, 2000). The article provides no new data but relies on published ESS and MSLT results from normal and narcoleptic subjects to calculate sensitivities and speci®cities of the ESS and MSLT for these states, which are assumed to represent fully alert and pathologically sleepy individuals. Receiver-operator curves then show that the ESS assigns subjects to their correct groups with nearly perfect accuracy, whereas the MSLT does not perform as well. Unfortunately, these analyses are ̄awed. The narcoleptics were de®ned, in large part, by their subjective complaint of excessive daytime sleepiness (American Sleep Disorders Association 1997; Mitler et al. 1998; US Moda®nil in Narcolepsy Multicentre Study Group 1998). The normal subjects' MSLT data were derived from a study in which no subject with subjective daytime sleepiness could be included (Levine et al. 1988). Normal subjects' ESS data (Johns and Hocking 1997) were taken from the 72 (22%) of 331 Australians who were, as described by Johns, `selected by strict criteria derived from a detailed sleep questionnaire' (Johns 2000); this sample, de®ned by self-report, could be labeled `super subjective normals'. The result is that in John's re-analysis, the normal subjects had no subjective sleepiness and the narcoleptics did have subjective sleepiness, by de®nition. Subjects were grouped more by virtue of their subjective sleepiness than by objective criteria. Johns then showed that the subjectively derived ESS assigned persons to their subjectively de®ned groups more accurately than did the objective MSLT. The ESS previously has been shown to correlate well with patients' own perceptions of their sleepiness, and to have weak or no correlation with objective measures of sleepiness (Chervin and Aldrich 1999): Johns' recent analysis only seems to con®rm these observations. Several additional problems with the ESS are not addressed in Johns' article. Increasing evidence suggests that in the assessment of sleepiness, the ESS is subject to undesirable confounding variables, including gender (Chervin and Aldrich 1999), psychological in ̄uences (Olson et al. 1998), and subjective perception of fatigue, tiredness, and lack of energy (Chervin, 2000a). Although Johns repeatedly argues, based on face validity, that the ESS measures sleep propensity in eight speci®c situations rather than just one (like the MSLT) (Johns 1991, Johns 1993; Johns 1994; Johns 1998; Johns 2000) he has provided no criterion validity to substantiate this argument. In one study that did test his hypothesis, subjective responses to the ESS item that asks about `lying down to rest in the afternoon when circumstances permit' failed to show any robust association with objective measures in this speci®c situation, namely the afternoon naps of MSLTs (Chervin et al. 1997). Finally, Johns' recent article was somewhat selective in its review of existing literature. The largest existing studies of sleep apneics assessed with both MSLTs and Epworth scales showed no statistically signi®cant associations between the two measures (Chervin and Aldrich 1999; Benbadis et al. 1999). Some studies suggest that apnea severity as determined by polysomnography is associated with MSLT-measured sleepiness but not with ESS scores (Chervin et al. 1997; Chervin and Aldrich 1999; Kingshott et al. 1995). A recent study of 1824 individuals did show a highly signi®cant ± but impressively weak ± association between apnea severity and ESS scores (Gottlieb et al. 1999). For example, the subjects with little or no sleep apnea had a mean ESS score of 7.2 ‹ 4.3 while those with the most severe apnea had a mean score of 9.3 ‹ 4.9. Johns is correct in writing that the MSLT is unlikely to be a perfect gold standard, but in that respect the test is similar to many other medical gold standards. His point is also well taken that the 5 and 10 minute `rule of thumb' for MSLT interpretation should not be misused. Strict cut-points on a continuous unimodal measure will almost always serve a patient poorly if results are not carefully integrated and weighed with data derived from the patient's medical history (Chervin, 2000b). Neither the MSLT nor the ESS have been well-validated against objective sleepiness-related outcomes of importance to patients, such as motor vehicle crashes or work performance. Until those important data become available, Correspondence: Ronald D. Chervin, MD, MS, Michael S. Aldrich Sleep Disorders Laboratory, University Hospital 8D8702, Box 0117, 1500 E. Medical Center Dr, Ann Arbor, MI 48109±0117, USA. Tel.: (1) 734 647 9064; e-mail: chervin@umich.edu J. Sleep Res. (2000) 9, 399±401