INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Excessive daytime somnolence is a frequent complaint, with severity varying from mild and unobtrusive to severe and disabling. Literally, it refers to the subjective feeling of having difficulties in remaining awake and fully alert and an increased tendency to fall asleep. It is frequently difficult to make a clear difference between the complains of somnolence, fatigue, weakness, and asthenia. As with any symptom, the quantification of daytime somnolence can only rely on questionnaires, of which several different ones have been developed in recent years (1, 2). Symptoms or complaints do not always reflect an objective reality. For instance, a subject complaining of excessive daytime sleepiness does not necessarily actually fall asleep during daytime, nor necessarily present with any discernible decrease in cognitive or decisional abilities. Thus, tests have been developed to measure the propensity to actually fall asleep in favorable conditions (multiple sleep latency test, MSLT [3]), or the capacity to remain awake when placed in conditions theoretically ideal for falling asleep (maintenance of wakefulness test, MWT [4]). A truly somnolent subject should score high in any questionnaire on sleepiness, should have short latencies to sleep in the MSLT, and should be unable to remain awake in the MWT. However, this is not always the case. The MSLT and the MWT are not well correlated (5), and neither is well correlated with widely used scales of subjective somnolence, like the Epworth Sleepiness Scale (ESS, 6). This suggests that our understanding of this common complaint is still far from ideal.

Excessive daytime somnolence (EDS) may be the consequence of short- or long-term sleep deprivation such as in shift workers or busy physicians (especially, though not exclusively, those devoted to sleep medicine), may be secondary to rare diseases such as narcolepsy, or to common diseases such as the obstructive sleep apnea syndrome, but is also extremely frequent in the general population with prevalence ranging from about 15% up to 30% (7, 8). Tests devised to assess objectively EDS should therefore be adapted to the huge number of potential subjects to be tested, and should ideally be simple, inexpensive, reliable, and easily repeatable, and require little training from technical personnel. This is far from being the case for the MSLT or the MWT, which require recording during several hours of relatively complex electrophysiological signals, and experienced trained personnel to read the recordings while they are being performed.

Very recently a simplified version of the MWT has been proposed under the acronym OSLER test (for Oxford SLEep Resistance test) (9). During the MWT, the subject, supine or semirecumbent in a quiet dark room, is specifically asked to remain awake for 40 min four times every 2 h, beginning at about 8:00 A.M., after a normal sleep night. Electroencephalogram (EEG), Electro-oculogram (EOG), and chin electromyogram (EMG) signals are continuously monitored, and subjects are immediately awakened should they fall asleep (to avoid long periods of sleep that could decrease the homeostatic drive to sleep). With the usual sleep scoring technique, at least three 30-s epochs of stage 1 non-rapid eye movement (REM) sleep, or a single 30-s epoch of any other sleep stage, is required before the presence of sleep is clearly identified. Because an epoch is classified according to the predominant vigilance state (i.e., the state present during at least half an epoch), between 16 s and 48 s of sleep is needed to determine the sleep latency during an MWT (10).

The OSLER test uses the same situation and requests the subject, placed semirecumbent in a dark and quiet environment, to remain awake for a maximum testing time of 40 min. This 40-min session is repeated four times at 2 h intervals. However, the presence of sleep is assessed behaviorally: subjects ares asked to touch a nonrecoil button each time a dim light flashes on a socket placed at eye's height in front of them. The LED flashes regularly for 1 s every 3 s. The test is terminated after seven consecutive flashes without response, assuming that the subject has been asleep for 21 s. Thus, each session may last 40 min or less, depending on whether the OSLER test termination criterion is met or not. The only published validation to date has been that the OSLER test gave results similar to the MWT performed another day in two groups of normal subjects and patients with obstructive sleep apnea syndrome (OSAS) (9). It is conceivable that a subject performing the OSLER test could show periods of sleep of less than 21 s, and miss less than seven consecutive stimuli, thus allowing the test to proceed. It is even possible that a somnolent subject could sleep for 20 s every 30 s, thus during most of the performance of the test, without ever missing seven stimuli in a row, and thus finishing the 40-min test. As it has been devised, the OSLER test yields (as the MWT and the MSLT) a single figure: the mean duration of the four tests. Therefore, it is possible to fail to take due notice of short, but frequent, periods of sleep not fulfilling the classic definitions.

This study was devised to verify the occurrence of sleep periods during the performance of the OSLER test in a group of normal subjects after a non-sleep-deprived night and after a sleep-deprived night. Periods of sleep as short as 3 s were taken into account, and their influence on the responses to the individual visual stimuli was assessed in detail.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Twelve normal young nonobese (less than 27 kg/m² of body mass index) university students (seven female) without any complaint of daytime sleepiness were studied from November to December 1999 (a period without examinations). They were remunerated for the procedures (each subject received the equivalent of $75). They signed an informed consent for the study, which had been approved by the institutional ethics committee. The subjects were studied on two consecutive days, once after a non-sleep-deprived night and the other after a sleep-deprived night. The order of the nights was determined at random; half of the subjects began by a sleep-deprived night and the other half by a non-sleep-deprived night.

Methods

Subjects arrived at the hospital at 6:30 P.M. They were provided with a cassette recorder (Oxford Medilog 9000-II) continuously recording from surface electrodes two channels of EOG, one channel of chin EMG, four channels of EEG (C3-A2, C4-A1, C4-O2, and Fp2-C4), and one channel of ECG. At the end of installation, the subjects left the hospital with instructions to either have a normal sleep night or refrain from sleeping throughout the night (subjects had prepared special activities for that night to help them stay awake). They came back to the hospital between 8:00 and 9:00 A.M. the following morning, and the first measurement was performed immediately, followed by a new session every 2 h for a total of four sessions.

Each measurement session consisted in a recording of the OSLER test. Subjects lay semirecumbent on a comfortable armchair in a quiet and darkened room without windows, with the LED device placed at eye's height at a 2-m distance from the head of the subjects. The LED device and the hand response device with the nonrecoil button were relayed to a personal computer located in an adjacent room. The nonrecoil button is in fact a proximity sensor with a sensing distance of 1 to 2 mm, according to the manufacturer (Stowood Scientific Instruments, Oxford, UK). The software of the device repeatedly monitors the switch, looking for events such as lifting of the finger away from the button and approaching the finger toward the button. If the finger is continuously kept on the button, no event is detected and this is interpreted as a failure to respond to the preceding stimulus (Lyn Davies, personal communication). The computer counted and recorded the responses and the failures to respond. The computer stopped the test immediately after seven consecutive missed stimuli. The subject was then immediately "awakened" at that time by the technician monitoring the computer screen. If, during the session, there were not seven consecutive missed stimuli, the session was stopped after 40 min. Synchronization between the OSLER session and the electrophysiological signals recording was performed by pushing the event button available in the Oxford Medilog device, and starting at the same time a stopwatch. The OSLER session was started from the personal computer keyboard exactly 30 s afterward. It is to be noted that the electrophysiological signals were not available on-line to the technician performing the session, so that the latter did not know whether the subject was actually asleep or not during the session.

The subjects were then submitted to a reaction time test (PVT), measuring the speed of response to a flash. The device (Psychomotor Vigilance Test, PVT 192; CWE Inc., Ardmore, PA) consists of a time counter that starts running on an illuminated window at variable times. The subject is instructed to push a button as soon as the clock starts running. This stops the clock, and the number of (generally) milliseconds needed to stop the clock is stored in memory. After 10 min, the device gives the mean reaction time for the 10-min period. The mean of the four reaction time measurements constitutes the single PVT result.

Subjects were free to move between each session but were reminded that they should not sleep between sessions. At the end of the fourth session (between 2:50 and 3:50 P.M.), a new cassette was introduced into the recorder, the electrodes were verified, and the subject left the hospital to have a sleep-deprived or a non-sleep-deprived night. The following morning, the subject came back to the hospital at the same time as the first morning, and measurements were repeated as on the previous day.

Measurements

Continuous sleep recordings were assessed off-line in two different ways. Except for the OSLER sessions periods, sleep was analyzed according to Rechtschaffen and Kales in 1-min epochs (10, 11). For the recording periods corresponding to the OSLER sessions, we assessed the occurrence of microsleep defined as follows: a period of at least 3 s with a  (4-7 Hz) rhythm replacing an  rhythm or appearing on a background of desynchronized EEG on all four EEG channels, and without eye-blinking artifacts. Slow eye movements were accepted (12). The presence of  rhythm was considered as wakefulness, and thus excluded microsleep (see Figure 1). No attempt was made to differentiate between microsleep and established sleep. In other words, no maximal length was defined for microsleep, so that it could evolve into established sleep if it lasted long enough. Recordings where microsleep periods had been identified were reviewed a second time by the same expert reader, at least 2 wk apart. The second time, the reader had the results of the first scoring, and introduced any desired correction. In fact, no microsleep episode identified on the first reading was deleted, and less than 10% of the number of microsleep periods originally identified were added during the second reading for any given recording. The data present the results of the second reading.

View larger version (17K): [in this window] [in a new window]   Figure 1.   Example of microsleep periods. From top to bottom, channels 1 and 2: EOG; channel 3: chin EMG; channels 4, 5, 6, and 7: EEG (respectively, C3-A2, C4-A1, Fp2-C4, and C4-O2), and channel 8: ECG. The small vertical bars below the ECG are 1-s marks. The time is given in hh:mm:ss at the bottom left corner of each screen page. Upward arrows signal the start of two microsleep periods of 9 and 4 s, respectively. The downward arrows show the end of each microsleep period.

The OSLER test paper output gives the individual missed stimuli as a function of time (see Figure 2 for an example). The screen time scale (as well as the paper output time scale) can be modified from 5 s to 40 min. From the paper output of the OSLER recordings, we analyzed the correspondence between the misses (consecutive or not) and the presence or absence of periods of microsleep. We also studied the relationship between the OSLER test results and the PVT results, the total sleep time (TST) during the previous night, as well as between the OSLER session results and the cumulative time in microsleep during the performance of the OSLER sessions.

View larger version (11K): [in this window] [in a new window]   Figure 2.   Two examples of the number of missed stimuli as a function of time during two portions of the OSLER tests. The hatched bars at the bottom of the traces indicate the presence of microsleep periods. Note that microsleep periods do not perfectly overlap with the missed stimuli. Stimuli can be missed in the absence of microsleep (see the five and seven consecutive missed stimuli in A); microsleep can be seen without stimuli being missed (B). Note that the seventh missed stimulus leads to test termination without delay.

Statistical Analysis

Correlations were performed by the least-squares method, whereas comparisons were performed with the Student's t test for paired or unpaired samples, as required. A p value < 0.05 was considered significant. Specificity and sensitivity of the OSLER test in detecting sleep (of  3 s duration) were calculated according to the following formulas: Sensitivity = OSLER sessions terminated before 40 min/All sessions showing microsleep periods during the OSLER tests; and Specificity = OSLER sessions of 40 min duration/All sessions showing no microsleep periods during the OSLER tests.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In two of the female subjects, the recordings of the second night (in both cases the non-sleep-deprived night) were of very poor technical quality. Scoring was impossible for long periods amounting to a total of 4 h. Therefore, their results were excluded from any further analysis. Data on the total sleep time during the two nights, and mean and SD for the OSLER and PVT tests after the non-sleep-deprived and sleep-deprived nights, are given in Table E1 in the online data supplement for the remaining 10 subjects. These subjects strictly complied with the expected behavior: only two subjects had any sleep during the sleep-deprived nights (respectively 3 and 4 min of stage 1 sleep). No subject entered stage 2 sleep during the sleep-deprived night. No subject slept between the OSLER and PVT measurements. Subjects having a non-sleep-deprived night first had TST values in the range of expected normal sleep time. By contrast, subjects having the sleep- deprived night first had significantly longer total sleep periods on the second (non-sleep-deprived) night (7 h 04 min, range 5 h 12 min to 8 h 37 min versus 11 h 28 min, range 10 h 24 min to 12 h 51 min, p < 0.0005).

The OSLER test duration was significantly shorter after the sleep-deprived-night (25 min 33 s ± 7 min 29 s) than after the non-sleep-deprived night (38 min 20 s + 4 min 10 s, p < 0.001). All five subjects with the non-sleep deprived night first succeeded in finishing the 40-min OSLER test on all four sessions during the first day, whereas only three of the subjects with the sleep-deprived night first arrived at the end of the four 40-min sessions after the non-sleep-deprived night. There was a significant positive correlation between the individual TST and the mean OSLER test duration on the following day: the longer the TST, the longer the mean OSLER test duration (r = 0.65, p < 0.005). There was by contrast a significant negative correlation between the individual mean OSLER test duration and the PVT results: the shorter the OSLER test, the longer the reaction time (r = 0.48, p < 0.025). The PVT results were significantly longer after the sleep-deprived night (241 ± 21 ms) than after the non-sleep-deprived night (224 ± 20 ms, p < 0.005)

When we analyzed the individual OSLER sessions as the number of missed responses, whether consecutive or not (in other words, including those missed responses that did not lead to session termination), we found a great variability of the number of missed responses per session. Table 1 shows the total number of isolated and consecutive missed responses, and the presence or absence of simultaneous periods of sleep as well as their length. In general, as the number of consecutive misses increases, the probability that the subject was asleep during the intervening period also increases: 43% of isolated misses occur during a period including a microsleep; when two successive stimuli are missed, the probability of finding a microsleep is 83%; this figure increases to 92% for three consecutive misses, 96% for four consecutive misses, 95% for five consecutive misses, 100% for six consecutive misses, and 94% for an interrupted session (seven consecutive missed stimuli, see below). In the absence of sleep, a single response was missed 299 times. On the other hand, there was a response 38 times despite the identification of a microsleep period at that moment.

The number of missed stimuli does not exactly correspond to the duration of the sleep period recorded at that particular moment. For instance, seven consecutive misses (i.e., a 21-s period without responses) could correspond to concomitant continuous sleep lengths of 14 to 28 s (see Figure 2). In other cases, a number of consecutive misses (less than seven) could be followed by one response and then another series of consecutive misses attaining or not the seven miss limit of the session (see Figure 3). This could correspond either to two consecutive microsleep periods separated by a short period awake or to uninterrupted sleep. On the contrary, on three occasions, the subjects slept for 21 s but only missed six consecutive stimuli, therefore not ending the session.

View larger version (12K): [in this window] [in a new window]   Figure 3.   Example of the number of missed stimuli as a function of time during a portion of an OSLER test. The hatched bars at the bottom of the time scale indicate the presence of microsleep. Note that within a period of continuous sleep (36 s), not all stimuli are missed: five consecutive stimuli are missed, then there is an isolated response, and only thereafter the seven consecutive missed stimuli criteria of test termination is attained. Note that the seventh missed stimulus leads to test termination without delay.

Two different subjects failed an OSLER session once each without the presence of microsleep as we defined it (i.e., false-positive session). One of these two "exceptions" corresponded to a predominant  rhythm throughout the 21-s period after the non-sleep-deprived night. The other one corresponded to a fully and clear awake period at the start of the first OSLER session of that subject after a sleep-deprived night.

If we consider sleep as present or absent during all the OSLER sessions, irrespective of the duration of both the OSLER session and of the sleep periods, the sensitivity of an OSLER session to detect sleep is 85%, whereas its specificity is 94%. Indeed, for sessions lasting 40 min, sleep periods were present in six sessions, and sleep was not detected in 39 sessions. For sessions lasting less than 40 min, sleep was detected in 33 sessions, and no sleep was detected in two sessions. If all sleep periods during each OSLER session are summed up, so as to obtain an OSLER session TST, these values are highly significantly correlated to the number of misses per session (r = 0.94, p < 0.005). In other words, the number of misses accurately reflects the total time asleep during each session, even for those sessions lasting 40 min but including periods of sleep. There were six of these "false-negative" sessions, with TST varying from 11 to 250 s (see Table 2).

We have analyzed the relative value of each of the four sessions performed per day in each subject. If only the first two sessions are retained for analysis, the mean results are not significantly different from the mean of the four sessions, but the sensitivity of the test decreases. To retain the same average results, but also the same specificity and sensitivity, as the full four 40-min session test, the first three, or the last three, sessions have to be included in the analysis (see Table E2 in the online data supplement).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We have found that the OSLER test, without perfectly reflecting sleep occurrence, appears to be a convenient tool to accurately detect sleep latency even for quite short periods of sleep. With respect to the other tools available up to now to objectively assess daytime somnolence, the OSLER test offers the advantage of simplicity, low cost, automatic reading, and low requirements for technical personnel.

We have studied a small group of normal subjects under normal conditions and after a single night of sleep deprivation. Our aim was to place subjects under conditions of normal or increased daytime sleepiness. We have objectively verified that there was no sleep at all (or almost so) during the sleep-deprived night, and we are thus sure that the subjects were indeed in a condition of increased daytime somnolence during the following day. This was also reflected in the long TST during the non-sleep-deprived night in the five subjects having the sleep-deprived night first. In these particular subjects, daytime somnolence could still be increased after the non-sleep-deprived night, as suggested by the fact that two of the five subjects had OSLER test results below 40 min. However, this is not a drawback for our study as our main aim was to obtain a detailed understanding of the evolution of vigilance states leading to an absence of response to visual stimuli under varying levels of daytime somnolence. It is not at all established that the level of somnolence experienced after a single night of sleep deprivation in normals is similar in intensity or quality to the one experienced by patients with chronic sleep abnormalities, such as narcolepsy, obstructive sleep apnea, or periodic leg movements. Therefore, further studies are warrented in those specific groups of patients in order to verify the predictive value of the test in each case. Our subjects received a financial remuneration for their participation in the study. This could have influenced their motivation, either by pushing them to resist falling asleep, or by pushing them to try to fall asleep, during the OSLER test performances. However, the subjects were not told what we expected from the results and were not paid as a function of their particular individual results. Even the two subjects excluded from the study were paid for their participation. Therefore, it is unlikely that this could have influenced our results.

We have assessed sleep in two different ways. To make sure that there was no sleep during the sleep-deprived night, and to assess sleep during the non-sleep-deprived night and between the OSLER tests, we used the standard Rechtschaffen and Kales scoring rules in 1-min epochs (10, 11). To identify sleep during the performance of the OSLER tests, we used a definition allowing the detection of much shorter periods of sleep, so-called microsleeps. These are short-lasting (a few seconds) episodes of slow eye movements, or interruption of the blinking artifacts characteristic of full wakefulness, accompanied by the appearance of a  rhythm on the EEG. We used a minimum 3 s time for scoring a microsleep for two reasons: to follow Harrison's definition (12) and because shorter durations are extremely difficult to detect visually. The distinction between microsleep and sleep is at least partly a matter of length, as a microsleep lasting for more than 15 s will be scored as sleep according to the standard rules. In fact, we made no distinction between microsleep and sleep during the OSLER test: only a minimum duration of 3 s was defined, but no maximum length was fixed to separate microsleep from "standard" sleep.

It appears from our data that stimuli can be frequently missed in the absence of sleep. This may be due to a decrease in the level of vigilance, the so-called attention lapses (13), that we have disregarded in this study. Attention lapses can be detected by the appearance of an  rhythm on EEG tracings and have already been described by Daniel (14) and used recently by Risser and coworkers (13). Missed stimuli in the absence of sleep can also be due to a number of other factors unrelated to vigilance in strict terms, such as distraction, a decrease in the amplitude of the movements of the finger with respect to the button, or a simple loss of one stimulus during eye blinking. Nevertheless, as the number of consecutive missed responses increases, the probability of finding a microsleep period also increases and is equal to or greater than 95% for four or more consecutive missed responses.

The OSLER test has been devised as a simplified MWT, giving a single mean duration of the test over four measurement periods or sessions. In other words, the test retains as the only meaningful value the average session termination criterion. Our data suggest that the total number of missed responses, reflecting the cumulative microsleep time, could add valuable information as a quantification of the sleep pressure or propensity, perhaps allowing for better discrimination between subjects. This figure can be easily retrieved from the computerized output of each session. Adding this value could also help to take due account of the microsleep episodes leading to five or six missed responses but not to the seven missed response definition of session interruption (see Table 1).

Figure 4 shows the total number of missed stimuli per session versus the total duration of each individual OSLER session (i.e., four data points per subject and per day are shown). To take into account the variable duration of each OSLER session, the number of missed stimuli per session is expressed normalized for the respective session duration. For instance, a subject showing no stimuli missed during 7 min, and then missing seven consecutive stimuli, would have one missed stimulus per minute duration of that OSLER session. The same one missed stimulus per minute duration of the OSLER session would be calculated for a subject showing 40 missed stimuli but successfully completing the 40-min session. Figure 4 shows several interesting features. The reproducibility of the four OSLER sessions performed after the sleep-deprived night is rather poor, and significantly higher than after the non-sleep-deprived night. When the reproducibility of the four OSLER sessions of each day is assessed by the standard deviation of the four results, there is a significant difference between the two days, variability being significantly higher by paired t test after the sleep-deprived night, p < 0.05 (see also Table E1). Figure 4 also shows that there is a clear difference between the frequency of missed stimuli and the duration of the OSLER sessions. For instance, subjects 3a and 7b have the same frequency of missed stimuli per minute, but subject 3a fails the session after 2 min, whereas subject 7b goes on for 28 min before the session ends. Similarly, failure rates of 0.5 missed stimuli per minute can be seen for sessions of 17 min (subject 7b), 24 min (subject 4b), or 40 min (subject 1b). Finally, Figure 4 also shows that, in general, there is an inverse significant correlation between both variables (i.e., the higher the number of missed stimuli per minute, the shorter the session duration; r = 0.70, p < 0.005). However, the strength of this correlation is not excessively high, with an r² of 0.49. We do not propose that the number of missed stimuli per minute duration of the OSLER sessions should replace the originally proposed duration of the OSLER test, but rather that both should be taken into account. A note is worthwhile concerning the two false-positive sessions, where there were seven consecutive missed responses (resulting in session termination) without sleep being detected. In one subject, this corresponded to a continuous period of  rhythm after the recovery night and could therefore be interpreted as a long attention lapse, which could have a similar meaning as a microsleep in the context of this study. The other one corresponded to the first session, coincided with a clear full wakefulness polygraphic tracing, and could be due to a misunderstanding by the subject of the request to touch the button with each flash, and to lift the finger between flashes. Even if these two sessions are taken at face value, in an "intention to treat" analysis, as we have done, the sensitivity and specificity of the OSLER test appear satisfactory.

View larger version (11K): [in this window] [in a new window]   Figure 4.   The number of missed stimuli during each individual OSLER sessions is displayed on the y axis, against the duration of each individual OSLER session. The number of missed stimuli is normalized per minute duration of the OSLER session, to take into account the varying length of different sessions. Each subject is represented by a different symbol. The a and b refer to OSLER tests performed on Day 1 and Day 2, respectively. Note that there are four sessions per day, but that one of the data points of subject 3a represents two different sessions (duration of OSLER sessions: 16 min 43 s and 16 min 46s; number of missed stimuli per minute: 1.25). For comments, see text.

Motor responses can be obtained in (or in spite of) the presence of periods of microsleep as we defined them. This emphasizes that the transition from wakefulness to sleep is less sharp than our polygraphic definitions might suggest, which is not at all a new finding. Automatic-type behavior has been already described during stage 1 and even stage 2 non-REM sleep (15). Nevertheless, the probability of finding a response in the presence of microsleep decreases as the length of the microsleep increases, and becomes very low for microsleep periods lasting more than 8 s (see Table 1). Indeed, 11.5% of sleep periods lasting from 3 to 5 s, and 5.4% of sleep periods lasting from 6 to 8 s, are not detected by the OSLER test, whereas all periods of sleep lasting for more than 8 s lead to at least one missed stimulus. Thus, although strictly speaking, the OSLER test may sometimes fail to detect 21-s sleep periods, and certainly fails to detect many shorter microsleep episodes, it is nevertheless quite accurate in distinguishing the absolute presence or absence of sleep during the four 40-min opportunities required to perform it.

The notion of somnolence is difficult to comprehend in quantitative terms. The emphasis on patient subjectivity (16) implicit in the modern shift of medical thought from physician to patient (or client) may obscure the fact that quantification of complaints or symptoms just does that: quantitates subjective feelings, and does not necessarily represent truth. For instance, it is known that a placebo treatment may have powerful effects on the results of the ESS, or of any other subjective scale of sleepiness (or any other subjective perception) (17). Similarly, treatments without any real effect on the disease may result in a decrease in the subjective perception of somnolence without any changes in objective tests of excessive daytime somnolence (18). Therefore, it is always difficult to interpret subjective scales in the absence of confirmatory objective data. The extreme somnolence of patients with severe OSAS, and its disappearance with continuous positive airway pressure (CPAP) treatment, makes sense intuitively, as they correspond respectively to the presence and suppression of clear-cut sleep-disturbing apneas. The situation is less clear in many other cases of excessive daytime somnolence. Thus, having an objective, simple, repeatable test to assess somnolence is almost a necessity. The usual instruments (MSLT or MWT) are certainly not appropriate, both because of their complexity and because of their artificial context that cannot reflect somnolence in real-life conditions. Is the OSLER test a better instrument? The answer to this question depends in fact on the way somnolence is reflected by the presence of repetitive episodes of microsleep. Our study does not allow us to analyze this issue in depth. It can only be suggested that the somnolence secondary to a sleep deprivation night in a normal subject seems to be characterized by the presence of microsleep episodes and that this seems to be well-reflected by the OSLER test. We have purposely refrained from submitting our subjects to a somnolence questionnaire, because our aim was to assess the performance of the OSLER test in detecting objective elements typical of excessive somnolence (microsleep) and because our protocol was devised to put subjects in a situation of clear excessive somnolence.

There was a significant but poor (r² = 0.23) negative correlation between the OSLER test and the PVT. However, the differences in the PVT results between the test performed after a normal night and after a sleep-deprived night were small, even if significant. Moreover, most if not all the PVT results remained within the limits of normality even after the sleep-deprived night. For instance, the normal subjects studied by Barbé and coworkers had a mean PVT result of 262 ± 5 ms (19). The Osler test represents a boring task, whereas the PVT test has a "competitive" nature ("as soon as possible, as fast as you can"), so that each type of test might explore a different aspect of a given individual's daytime somnolence.

If the OSLER test is to be applied in a wide scale, to assess all cases of somnolence, and to verify the decrease in somnolence required, for instance, in some European countries to allow somnolent patients to resume driving motor vehicles, then the duration of the testing procedure becomes an important issue. We have found that reducing the number of sessions from four to three (and thus shortening the test by 2 h) does not decrease its value. This reduction in the time needed to perform the complete testing, added to the low requirements for technical personnel (when compared with the MWT), makes the OSLER test better adapted for wide-scale objective assessment of daytime somnolence.

In conclusion, we have found that the OSLER test, without perfectly reflecting sleep occurrence, appears to be a useful, accurate, convenient, and simple way to detect the occurrence of microsleep episodes in sleepy normal subjects.