The OSLER test represents a simple alternative to the maintenance of wakefulness test. Standard analysis of OSLER test results yields a mean sleep latency (MSL). The aim of this study was to use both MSL and errors (nonresponses to stimulations) to characterize daytime sleepiness in apneic patients. OSLER test results at 9:00 a.m., 11:00 a.m. and 1:30 p.m. were compared in 27 obstructive sleep apnea syndrome patients (50.4 ± 10.4 years; apnea–hypopnea index: 43.05 ± 25.08) and 20 control subjects (C). Not only did obstructive sleep apnea syndrome patients demonstrate earlier sleep onset than control subjects (MSL: 1,788 seconds ± 511 versus 2,335 seconds ± 139, p < 0.001), but they also spent a greater percentage of time making errors than control subjects (5.4% ± 4.7 of total test time versus 0.4% ± 0.4, p < 0.001) with specific error profiles (high prevalence of three to six consecutive errors). When error profile analysis was added to standard sleep latency assessment, up to 40% of patients with normal sleep latency were exhibiting abnormal fluctuation in vigilance. A single 9:00 a.m. OSLER session appeared as sensitive as three consecutive sessions in its use as a means of identifying patients with significant daytime sleepiness. On the other hand, the 1:30 p.m. OSLER test session was least specific to distinguish apneic subjects from normal subjects, suggesting that the OSLER test can identify the afternoon peak in physiologic somnolence.
Daytime sleepiness is a major symptom of obstructive sleep apnea syndrome (OSAS). Sleep fragmentation and the resulting excessive daytime sleepiness are considered instrumental in causing impairment of daytime functioning. As a result, patients with OSAS become prone to accidents as well as to marital and occupational problems (1, 2).
The quantification of daytime somnolence is not simple. Complaints do not always coincide with an objective impairment of vigilance. Questionnaires used in assessing subjective sleepiness (e.g., the Epworth scale  or the Stanford scale ) do not strongly correlate with objective quantifications (5). Assessing daytime impairment by using objective quantification tools may provide a better evaluation of daytime function alteration in these patients. The most widely used objective test of sleepiness is the multiple sleep latency test (MSLT) (6). This test measures an individual's tendency to fall asleep. However, patients with OSAS complain of an inability to stay awake and thus a maintenance of wakefulness test (7), where the instruction is to stay awake rather than fall asleep, may be more clinically relevant in assessing these patients.
These standard tests for objective quantification of sleepiness are labor intensive because they require electroencephalographic (EEG) recordings to determine the exact time of sleep onset. Another drawback is the need for a technician to constantly monitor the EEG trace during the test, which is time consuming. Hence difficulties arise in using such tests in clinical practice and studies in the community.
The OSLER test has been recently proposed as a behavioral test which simplifies the performance of the maintenance of wakefulness test. Indeed, it has been used as a simple alternative to MWT in a study comparing 10 control subjects to 10 severe OSAS patients (8). In this test, the occurrence of sleep is assessed behaviorally rather than by EEG monitoring. The subject is asked to respond by hitting a button each time a dim light flashes. The light-emitting diode flashes regularly for 1 second every 3 seconds. The subject is instructed to remain awake in this soporific situation for a maximum testing time of 40 minutes. When the subject fails to respond for 21 seconds (i.e., seven consecutive illuminations), the test is ended and it is assumed that the patient has fallen asleep. Thus, the OSLER test reproduces many of the MWT characteristics, with the advantage of being a simpler and less expensive tool, which does not require the presence of a trained technician. The simplicity of the test makes it easy to be used outside the sleep laboratory setting.
The standard way to analyze the OSLER test is to determine the sleep latency, measured as the delay before the appearance of seven consecutive flashes without response (21 seconds). A 21-second threshold has been chosen because it corresponds to the minimal sleep duration generally used to score one epoch of sleep when using standard scoring rules for night polysomnography (9). However, this threshold could be inappropriate in detecting fluctuation in vigilance occurring before the end of the test (10). Indeed, it is quite possible for a somnolent subject to fall asleep several times during the test for less than 21 seconds without ever missing seven consecutive stimuli, and thus finish the 40 minute test (see example in Figure 1). Alternatively, subjects with low motivation can miss occasional LED flashes without any significant sleepiness. In this case, errors are not expected to be consecutive. Priest and colleagues (10) EEG data recorded during the OSLER test showed that normal subjects missing three consecutive flashes (9 seconds) had a 92% chance of simultaneously manifesting an EEG microsleep. This led the present authors to hypothesize that an increase in the number of consecutive misses would indicate that the subject had fluctuations in vigilance during the intervening period.
Analyzing sleep latencies alone (i.e., seven consecutive misses) does not give a thorough view of fluctuations in vigilance during the period preceding the end of the OSLER test. Therefore, the analysis of all omissions and their distribution during the test could provide a more sensitive index in the assessment of daytime somnolence.
This study was aimed at assessing vigilance in OSAS patients and normal subjects. This was achieved by analyzing both groups' capacity to maintain wakefulness during a soporific task, as well as to maintain a constant level of vigilance for a given time. The capacity to maintain a constant level of vigilance was evaluated by analyzing consecutive errors occurring during the OSLER test. The different error profiles (EP) (short errors or longer errors) were considered as reflecting different levels of vigilance. According to previously published data from normal subjects (10), EP 1–2 (corresponding to one or two consecutive errors) might represent lack of attention, EP 3–6 (three to six consecutive errors) microsleep episodes, and EP 7 (seven consecutive errors) sleep onset. We hypothesized that control subjects would exhibit only short fluctuations in vigilance (i.e., EP 1–2) and no sleep onset during the test whereas OSAS patients would exhibit microsleep and sleep onset episodes that would be related to EP 3–6 and EP 7 identification.
A total of 27 patients who had been referred to a tertiary sleep laboratory for clinical suspicion of OSAS that was confirmed by polysomnography (20 males and 7 females, age 50.3 ± 10.4 years) were studied. 20 nonobese, nonsnorer healthy volunteers (14 males and 6 females, age 44.2 ± 10.3 years) served as control subjects. A polysomnography was performed in a random subset of five control subjects (25% of the control group) to confirm the absence of subclinical sleep breathing disorder. None of the participants were remunerated for their participation in the study.
Continuous recordings of EEG (C3/A2-C4/A1-Cz/O1 from the international 10–20 Electrode Placement System), eye movement measurements, chin electromyogram, and ECG with modified V2 lead were obtained. Airflow was measured with nasal pressure, as well as with the sum of buccal and nasal thermistor signals. Respiratory effort was monitored with uncalibrated inductance respiratory plethysmography. An additional signal of respiratory effort (i.e., pulse transit time or esophageal pressure ) was recorded concurrently. Oxygen saturation was measured using a pulse oximeter (Biox-Ohmeda 3700; Ohmeda; Liberty Corner, NJ).
The polysomnogram was scored manually according to standard criteria (11). Apneas were defined as complete cessation of airflow for more than 10 seconds; hypopneas were defined as a decrease of more than 50% in nasal pressure signal lasting for at least 10 seconds, a decrease of more than 30% associated with a decrease in oxygen saturation of more than 3%, or a microarousal. Apneas and hypopneas were classified as central, obstructive, or mixed depending on the absence or presence of breathing efforts.
Test procedure. The OSLER test consisted of 40 minute sleep-resistance challenges conducted in a dark room isolated from external noise. The subject, dressed and lying in a semirecumbent position, was asked to remain awake without using specific strategies. By hitting a button placed on a box directly connected to a personal computer, the subject was instructed to respond to a visual stimulus (light-emitting diode flash) which appeared for 1 second every 3 seconds. A total of 800 stimulations per test was emitted. All tests were video recorded to verify that the subjects were following instructions. Each subject underwent the OSLER test three times (at 9:00 a.m., 11:00 a.m. and 1:30 p.m.) allowing the assessment of vigilance at different times of day.
In our clinical experience, the reproducibility of the fourth session of the OSLER test is less satisfactory than that of the three other sessions. Both patients and control subjects were reluctant to accept an additional 40-minute session. As a consequence, we decided to perform three sessions, instead of four as originally described (8).
All patients with OSAS and five of the control subjects underwent polysomnography the night before the OSLER test. Control subjects who did not undergo polysomnography were instructed to follow their usual sleeping habits. They were asked whether sleep quality had been normal during the night before the OSLER test.
The first OSLER session (9:00 a.m.) was started on average 2 hours after subjects had awakened.
Sleep latency. Each session ended automatically after 40 minutes or before if the subject did not give a response to seven consecutive flashes (i.e., 21 seconds), which was taken as an indication of sleep onset (8). Twenty-one seconds has been proposed as a test termination time because it represents approximately the minimal time needed to score a conventional epoch of sleep (9). The computer stored the time to test termination, which is assumed to represent sleep latency. A sleep latency (test duration) was determined for each session (9:00 a.m., 11:00 a.m., and 1:30 p.m.). The mean sleep latency of the three OSLER tests was calculated.
Errors made by patients and control subjects were compared. The percentage of time corresponding to the appearance of errors during the OSLER test was also taken into account ([3 seconds × number of omissions/sleep latency duration in seconds] × 100).
Lastly, the pattern of consecutive errors during the test was analyzed. The relative proportion of each subtype of consecutive errors (i.e., 2, 3, 4, 5, 6, 7) was used to identify various levels of sleepiness. For each test, consecutive errors were allocated to an EP type. (i.e., two consecutive errors = EP2, three consecutive errors = EP3). The EPs were clustered into type EP1–2 for one and two consecutive errors, indicating lack of attention, type EP3–6 for three to six consecutive errors, indicating a longer period of inattention, and type EP7 for seven consecutive errors, characterizing sleep onset during the OSLER test. Priest and coworkers (10) previously demonstrated in normal subjects that seven consecutive errors (21-second misses) during the OSLER test was almost always associated with the occurrence of an EEG defined episode of microsleep. In addition, the probability of microsleep was 43% and 83% for 3 and 6 second misses, respectively. For 9-second misses, the probability of recording a microsleep on the EEG trace reached 92%.
Percent of patients and control subjects demonstrating fluctuation in vigilance during the OSLER test. Based on the findings of Priest and colleagues (10), one can postulate that one period of 9 to 18 second misses (one EP3–6) or several episodes of 3- to 6-second misses (five EP1–2) are likely to be associated with sleep periods and, therefore, that such errors represent fluctuations in vigilance during the test. For this reason, we extended our analysis to an examination of the percent of patients and control subjects demonstrating such patterns over the three sessions of the OSLER test.
The unpaired t test or Mann-Whitney test was used to compare control and patient groups for quantitative variables. The χ2 square test was used for qualitative variables. Results are expressed as mean ± SD and statistical significance was accepted for p < 0.05.
Age, yr (mean, SD)
BMI, kg/m2 (mean, SD)
ESS (mean, SD)
AHI (mean, SD)
Mean Nocturnal SaO2, % (mean, SD)
Minimal Nocturnal SaO2, % (mean, SD)
Time Spent at SaO2 < 90%, % (mean, SD)
|Control subjects||20, 14/6||44.17, 10.32||22.99, 3.33||5.55, 3.38||11.49, 6.18 (n = 5)||96.4, 1.12 (n = 5)||92.25, 2.37 (n = 5)||0.025, 0.05 (n = 5)|
|Patients||27, 20/7||50.31, 10.37
NS||28.5, 5.12*||13.3, 5.13*||43.05, 25.08*||92.90, 2.68*||74.27, 17.57*||11.15, 18.7*|
None of the subjects was excluded on the basis of monitoring by video recording. All subjects followed instructions correctly.
Sleep latencies for each test throughout the day are shown in Figure 2. For all tests, patients demonstrated significantly shorter sleep latencies than control subjects. The mean sleep latency in the control and the patient groups was 2,335 ± 139 seconds (38.9 minutes) and 1,788 ± 511 seconds (29.8 minutes), respectively (p < 0.001) (Figure 2).
The mean percent of subjects making errors during the OSLER test was significantly higher in the OSAS group than in the control group (91.3% of patients with OSAS versus 51.7% of control subjects).
For all three sessions, patients with OSAS had more errors in each category of EP (1–2, p < 0.0001; 3–6, p < 0.00001; 7, p < 0.0001) than control subjects. Moreover, the difference in the number of errors between the two groups was probably underestimated, the recording time being shorter in the OSAS group (i.e., shorter sleep latency).
An analysis of the distribution of errors shows that both groups did not demonstrate the same error profile (Table 2)
Patients %, SD
Control Subjects %, SD
|EP1-2, %||72.8, 29.2||99.0, 3.0||0.0009|
|EP3-6, %||12.8, 15.1||1.0, 3.0||0.0100|
|EP7, %||14.5, 28.6||0||0.0040|
|EP1-2, %||78.3, 19.0||98.0, 6.3||0.0010|
|EP3-6, %||16.7, 18.1||2.0, 6.3||0.0060|
|EP7, %||5.0, 10.3||0||0.0100|
|EP1-2, %||75.7, 23.3||70.8, 43.3||NS|
|EP3-6, %||12.2, 11.6||3.3, 6.0||0.0100|
|EP7, %||12.1, 21.2||25.9, 44.8||NS|
A range of 9 to 18 second misses, 21 second misses, and more than four episodes of 3 to 6 second omissions were used for both control subjects and patients as a surrogate marker of vigilance fluctuation. Using the standard 21 second misses threshold, 14 of 27 patients (51.8%) were considered to be exhibiting excessive daytime somnolence as defined by a reduced sleep latency during the 9:00 a.m. OSLER test session (Figure 4). When taking into account the 9 to 18 second misses, the percentage increased by 14.8%. Moreover, by adding patients with more than four episodes of 3 to 6 seconds omissions, the percentage increased by an additional 22.2%, compared with the initial scoring. As expected, owing to the decrease in vigilance at the beginning of the afternoon, a significant proportion of control subjects also experienced errors (20% [21 second misses], 30% [one EP3–6] and 45% [five EP1–2]) during the 1:30 p.m. OSLER test.
When we excluded the 1:30 p.m. OSLER test because of the loss of specificity, the 9:00 a.m. and 11:00 a.m. sessions provided the same information in terms of percent of patients classified as exhibiting daytime hypersomnolence. Only two patients with OSAS out of 27 demonstrated errors during the 11:00 a.m. session only, whereas the 9:00 a.m. session was normal. In other words, analysis of the 11:00 a.m. OSLER session resulted in classifying only two additional patients with OSAS as hypersomnolent.
This is the first study using both sleep latencies and error profiles during the OSLER test as a means of characterizing daytime sleepiness in patients with OSAS. Patients with OSAS not only fell asleep earlier as compared with control subjects (mean sleep latency: 1,788 seconds ± 511 versus 2,335 seconds ± 139, respectively; p < 0.001) but they also demonstrated specific patterns of error profiles (a high prevalence of three to six consecutive errors). When an error profile analysis was added to the classical sleep latency assessment, a loss of vigilance could be suspected in up to 40% of patients who presented normal sleep latencies. A single 9:00 a.m. OSLER session appeared as sensitive as three consecutive sessions were in identifying patients with significant daytime sleepiness. Indeed, only two patients with OSAS out of 27 demonstrated errors on the 11:00 a.m. session, whereas the 9:00 a.m. session was normal. On the other hand, the 1:30 p.m. OSLER test session was less specific in distinguishing patients with OSAS from normal subjects, indicating that the OSLER test may identify the afternoon peak in physiologic somnolence. Our findings suggest that the OSLER test is a sensitive tool in the detection of mild fluctuations in vigilance.
The OSLER test has been proposed as a simplified MWT which does not require EEG recordings or technician supervision. Its ability to distinguish normal subjects from severe OSAS patients (8) in terms of hypersomnia and to detect vigilance fluctuations after sleep deprivation in normal subjects (10) has been demonstrated. Moreover, the OSLER test appears sensitive enough to identify improvement in sleepiness after continuous positive airway pressure treatment (12). However, normative data are available only from two small groups of 12 and 10 subjects, respectively (8, 10), and a complete analysis of error profiles has never been reported in unselected groups of OSAS and control subjects.
In our study, all the normal subjects were able to stay awake for 40 minutes during the OSLER tests performed at 9 and 11:00 a.m. In the reports by Bennett and coworkers (8) and Priest and coworkers (10), the majority of control subjects were able to finish both 40 minute sessions without 21 second misses. On the other hand, the significant reduction in sleep latencies observed in OSAS patients in our study confirms the OSLER test's capacity to distinguish between control subjects and OSAS patients. The mean sleep latency in our subjects tended to be longer than in the initial validation study (29 minutes versus 10 minutes, respectively). This initial study was performed in a group of severe, highly symptomatic patients with OSAS (Epworth sleepiness scale = 17, O2 saturation dip rate = 32.7/hour [9.7–65.6]) (8). Mean sleep latencies later reported by the Oxford group in a more representative OSAS population (12) were in fact close to the 30 minutes reported in the present study.
By analyzing the errors made before the end of the test (i.e., seven consecutive omissions), we found that errors systematically occurred during the test in most OSAS patients (91.3% of patients with OSAS versus 51.7% of control subjects). Overall, control subjects made fewer errors than patients with OSAS. Moreover, the number of errors would have presumably been higher in patients with OSAS had the test been prolonged beyond the sleep latency cut off (EP 7). In the OSAS group, error time (time spent making errors) represented 5.4% of their sleep latency, whereas it represented only 0.4% in the control group. This observation suggests that patients with OSAS have difficulties not only maintaining wakefulness during a soporific task, but also in maintaining a constant level of vigilance during the test. This could explain the results obtained from driving simulation in patients with OSAS (13). Not only do patients with OSAS get into accidents such as crashes (during sleep episodes) but also show a greater tendency to deviate from the ideal lane position (short fluctuations in vigilance) than control subjects. In our study, type 1–2 errors were the most prevalent, both in control subjects and patients. However, type 3–6 errors accounted for more than 12% of patients total errors, in contrast with the control group (only 4% of their total errors). Seven type errors never occurred in the control group during morning OSLER sessions. The different error profiles are likely to be related to different aspects of vigilance fluctuations.
In a recent study, Priest and colleagues (10) hypothesized that the probability that a subject was asleep would increase as the number of consecutive misses increased. Indeed, these authors were able to show a 92% probability of microsleep onset in subjects who presented three consecutive errors. Microsleep was defined as the occurrence of a period of at least 3 seconds where Θ rhythm replaces α rhythm or appears on a background of desynchronized EEG without eye-blinding artifacts. Another intermediate state of vigilance termed ‘attention lapses’ (13, 14) has been defined as the appearance of α rather than Θ rhythm for less than 5 seconds on EEG tracings. These different definitions underscore the difficulty in characterizing attention and vigilance fluctuations. Different error subtypes, defined by the number of consecutive misses, and their frequency during the OSLER test are likely to represent a range of situations, from physiologic variations in attention to sleepiness that could potentially lead to increased accident risk. Based on findings in normal subjects (10), it is difficult to attribute the 3–6 EP in the OSAS group to a simple lack of attention. The number of 3–6 errors gives an indication of overall microsleep time before the onset of sleep (seven consecutive errors). Further studies are needed to determine whether an increase in the number of errors with normal sleep latencies is correlated to modifications of outcome measures such as subjective sleepiness, abnormal cognitive functioning, and automobile crashes.
By considering sleep latency as the only measurement of vigilance in the OSLER test, one risks neglecting interesting indicators, such as error analyses, as a useful means of identifying patients with hypersomnolence. For example, some patients with OSAS were able to complete all 40 minutes of testing time although their level of vigilance was not constant throughout the test. Analyzing error profiles may provide useful clinical insight and eventually contribute to a more sensitive diagnosis of daytime sleepiness. In other words, using the sleep onset criterion alone, a number of patients were categorized as “not hypersomnolent” when they should have been diagnosed as “hypersomnolent.” Using a more sensitive assessment of sleepiness, one can expect a better correlation between the number of microarousals during the night and daytime functioning. Moreover, OSAS treatments could be aimed not only at normalizing sleep latencies but also at suppressing fluctuations in vigilance.
The results of this study confirm that the OSLER test is a strong indicator of daytime sleepiness in patients with OSAS. However, the standard way of analyzing the OSLER test (determination of sleep latency) does not identify abnormal fluctuations of vigilance levels during the test. Profile analysis, in conjunction with sleep latency, shows that OSAS patients have difficulty both in maintaining wakefulness and maintaining a constant level of vigilance over short periods. In light of these results, one might argue for a multifactorial approach to vigilance deficits in OSAS. Whether OSAS treatment is accompanied by improvement in sleep latencies and the capacity to maintain a constant level of vigilance remains to be determined.
The authors are indebted to C. Grunwald, S. Launois, and J. Plante for their helpful editorial suggestions, and C. Loiodice and M. Selek for their technical assistance.
Supported by a grant from the French Ministry of Health, PHRC 1997.
|1.||Engleman H, Joffe D. Neuropsychological function in obstructive sleep apnoea. Sleep Med Rev 1999;3:59–78.|
|2.||Georges CF. Reduction in motor vehicle collisions following treatment of sleep apnoea with nasal CPAP. Thorax 2001;56:508–512.|
|3.||Johns MW. A new method for measuring daytime sleepiness: the Epworth sleepiness scale. Sleep 1991;14:540–545.|
|4.||Hoddes E, Zarcone V, Smythe H, Phillips R, Dement WC. Quantification of sleepiness: a new approach. Psychophysiol. 1973;10:431–436.|
|5.||Sangal RB, Sangal JM, Belisle C. Subjective and objective indices of sleepiness (ESS and MWT) are not equally useful in patients with sleep apnea. Clin Electroencephalogr 1999;30:73–75.|
|6.||Carskadon MA, Dement WC, Mitler MM, Roth T, Westbrook PR, Keenan S. Guidelines for the multiple sleep latency test (MSLT): a standard measure of sleepiness. Sleep 1986;9:519–524.|
|7.||Mitler MM, Gujavarty KS, Browman CP. Maintenance of wakefulness test: a polysomnographic technique for evaluation treatment efficacy in patients with excessive somnolence. Electroencephalogr Clin Neurophysiol 1982;53:658–661.|
|8.||Bennett LS, Stradling JR, Davies RJ. A behavioral test to assess daytime sleepiness in obstructive sleep apnoea. J Sleep Res 1997;6:142–145.|
|9.||Rechtschaffen A, Kales A. A manual of standardized terminology, techniques and scoring system for sleep stages in human subjects. Washington, DC: US Government Printing Office; NIH Publication 204, 1968.|
|10.||Priest B, Brichard C, Aubert G, Liistro G, Rodenstein DO. Microsleep during a simplified maintenance of wakefulness test. Am J Respir Crit Care Med 2001;163:1619–1625.|
|11.||The report of an American Academy of Sleep Medicine Task Force. Sleep-related breathing disorders in adults: recommendations for syndrome definition and measurement techniques in clinical research. Sleep 1999;22:667–689.|
|12.||Bennett LS, Langford BA, Stradling JR, Davies JO. Sleep fragmentation indices as predictors of daytime sleepiness and nCPAP response in obstructive sleep apnea. Am J Respir Crit Care Med 1998;158:778–786.|
|13.||Risser MR, Ware JC, Freeman FG. Driving simulation with EEG monitoring in normal and obstructive sleep apnea patients. Sleep 2000;23:393–398.|
|14.||Daniel RS. Alpha and theta EEG in vigilance. Percept Mot Skills 1967; 25:697–703.|