See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/12151982

Classification of Sleep-disordered Breathing

Article  in  American Journal of Respiratory and Critical Care Medicine · March 2001

DOI: 10.1164/ajrccm.163.2.9808132 · Source: PubMed

CITATIONS READS

103 333

5 authors, including:

Jean Jacques Hosselet Robert Norman

NYU Langone Medical Center New York University
19 PUBLICATIONS   653 CITATIONS    91 PUBLICATIONS   3,574 CITATIONS   

SEE PROFILE SEE PROFILE

Ana C Krieger
Weill Cornell Medical College
47 PUBLICATIONS   1,222 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

CPAP treatment View project

diagnosis of sleep disordered breathing in children View project

All content following this page was uploaded by Jean Jacques Hosselet on 09 November 2018.

The user has requested enhancement of the downloaded file.

Classification of Sleep-disordered Breathing
JEAN-JACQUES HOSSELET, INDU AYAPPA, ROBERT G. NORMAN, ANA C. KRIEGER, and DAVID M. RAPOPORT
Division of Pulmonary and Critical Care Medicine, New York University School of Medicine, New York, New York

Increasing recognition of sleep-disordered breathing (SDB) and its attempt to quantify sleep-disordered breathing is complicated
morbidity have prompted reevaluation of techniques to identify by the use of different techniques to detect respiratory events.
respiratory events during sleep. The present study was designed Until recently, the standard tool used for monitoring respi-
to evaluate the utility of various metrics of SDB and to identify the ratory airflow in polysomnography has been the thermistor.
optimal respiratory metric that objectively correlates to symptoms This device is effective at identifying complete cessation of air-
of excessive daytime somnolence (EDS). Metrics were derived from flow (apnea). However, detection of intermittent reduction of
combinations of conventional apnea/hypopnea, flow limitation flow (hypopnea) is much more subjective; it requires an arbi-
events (transient elevated upper airway resistance identified by
trarily chosen reduction in flow and has been defined with and
characteristic flattening on the flow/time tracing, using a noninva-
without the need for confirmatory consequences such as de-
sive nasal cannula technique), desaturation, and arousal. A total of
saturation or an arousal (10, 12–14).
137 subjects underwent clinical evaluation and nocturnal polysom-
nogram. In 34 randomly selected subjects, the best metrics for dis-
Furthermore, the thermistor is insensitive to other sleep-dis-
criminating between 13 subjects with no EDS/snoring and 21 pa- ordered breathing events that may complete the spectrum of up-
tients with EDS and snoring were identified by receiver operator per airway dysfunction associated with sleep (15, 16). Work with
curve analysis. Of the metrics and cut points tested, a total respira- more intensive monitoring (esophageal manometry) or using a
tory disturbance index (RDITotal, sum of apneas, hypopnea, and flow more sensitive analysis of traditional signals (respiratory induc-
limitation events) of 18 events/h was found to have the best dis- tance plethysmography plus thermistor [17]) has suggested the
criminant ability (100% sensitivity and 96% specificity). Prospec- existence of a large population whose symptoms appear to be
tive testing of this metric was then performed with the remaining due to frequent upper airway events similar in pathophysiology
103 subjects (14 nonsnoring non-EDS, 21 snoring non-EDS, 68 snor- to apnea, but that are undetected by the thermistor (2, 18). The
ing with EDS). Using this cutoff of 18 events/h, we obtained 71% important conceptual advance provided by esophageal manom-
sensitivity and 60% specificity for identifying subjects with EDS. etry has revolutionized the clinical approach to “unexplained”
We conclude that, in subjects with upper airway dysfunction, an excessive daytime somnolence but remains limited in clinical ap-
index that incorporates all respiratory events provides the best quan- plicability by its invasiveness and the increased patient discom-
titative physiological correlate to EDS. fort necessary to detect the respiratory abnormality.
In part because of the ambiguity in the definition of respi-
In the past decade there has been increasing recognition of the ratory events, treatments to relieve symptoms of snoring and
high prevalence of obstructive sleep apnea syndrome in both excessive daytime somnolence are being applied on clinical
the clinic and general populations (1). It has been proposed grounds with only indirect proof of any respiratory abnormal-
that even mild degrees of sleep-disordered breathing (SDB) ity. Thus, for example, frequent arousals or response to con-
may be associated with significant morbidity, including exces- tinuous positive airway pressure (CPAP) are used as evidence
sive daytime somnolence (2–4), long-term cardiovascular com- that upper airway disease was present and of respiratory origin
plications (5, 6), and significant societal costs (7–9). These pub- (19). For a more rational approach to clinical and epidemio-
lic health considerations, as well as increased public awareness logical issues, it has become imperative to find a simple reli-
of the possible medical significance of SDB, have led to an ex- able diagnostic technique to separate patients from individuals
ponential increase in referrals to sleep disorders centers, espe- who are free of disease. Finding such a metric requires making
cially for evaluation of snoring. the assumption that it is possible, using the metric, to explain
There is still much disagreement on the type and level of the level of symptoms (such as EDS) present in a given indi-
respiratory abnormality that must be used as a “cutoff” for sig- vidual. Any discrepancy between the value of this metric and
nificant disease (e.g., obstructive sleep apnea syndrome [OSAS] the level of symptoms in an individual is then attributed to im-
and upper airway resistance syndrome [UARS]). In a publica- perfection of the metric. An alternative to this conceptualiza-
tion by the American Academy of Sleep Medicine, UARS tion is that even with the most ideal metric for representing the
and OSAS were defined as ⬎ 5 respiratory events/h and the physiology of sleep-disordered breathing, differing sensitivity
presence of excessive daytime somnolence (EDS) (10). Part of of individuals to the same level of respiratory stress results in
the problem lies in not knowing the degree of biological varia- differing levels of symptoms. The problem facing researchers
tion in the number of respiratory events that exists in subjects is that identifying the relative merit of these two paradigms
who are asymptomatic and have no evidence of long-term mor- cannot be accomplished until the “best” metric has been iden-
bidity (“normal” subjects). While this problem is beginning to tified to describe the physiology.
be addressed in large epidemiological populations (11), any We have previously shown that a simple noninvasive tool
exists that can both identify apnea/hypopnea and also detect
additional respiratory events associated with flow limitation
(Received in original form August 26, 1998 and in revised form May 22, 2000) (15, 16) and thus potentially add to the definition of the physi-
Supported by grants from NIH-NHLBI HL53931 and NCRR M01 RR00096, ology. Flow limitation is a temporal pattern of airflow that in-
Mallinckrodt Nellcor Puritan Bennett, the Foundation for Research in Sleep Disor-
ders, and the American Lung Association of New York.
dicates abnormal upper airway resistance and usually results
in arousal (20). We have shown that the nasal cannula system
Correspondence and requests for reprints should be addressed to David M. Rap-
oport, M.D., Department of Medicine, New York University Medical Center, 550 detects the same events classified as respiratory effort-related
First Ave., New York, NY 10016. E-mail: [email protected] arousals (RERAs), using esophageal manometry (21). The
Am J Respir Crit Care Med Vol 163. pp 398–405, 2001 present study was designed to evaluate the utility of various
Internet address: www.atsjournals.org metrics of SDB derived from combinations of events detected
Hosselet, Ayappa, Norman, et al.: Classification of SDB 399

by the nasal cannula (conventional apnea/hypopnea and flow dency to fall asleep. This was based on patient and/or bedpartner in-
limitation events), desaturation and arousal. The goal was to terview at the time of clinical intake and prior to the NPSG. Global
determine the optimal metric to objectively correlate symp- clinical impression was felt to best characterize the complaint that
toms of EDS with their respiratory causes. causes patients to seek medical attention, and to represent the typical
entry criterion for the diagnosis of OSAS/UARS. The Epworth Sleep-
iness Scale (ESS) (22) was recorded in each patient and Multiple
METHODS Sleep Latency Test (MSLT) (23) or Maintenance of Wakefulness Test
(MWT) was performed as clinically indicated (in 45 patients), but nei-
Subjects ther ESS nor MSLT/MWT was used to define the presence of EDS.
All patients referred to the New York University Sleep Disorders Snoring was defined by a subject’s (or bedpartner’s) report of loud
Center (New York, NY) from January 1998 to October 1999 for eval- snoring on most nights of the week.
uation of significant EDS or severe snoring (with or without EDS)
were considered for inclusion in the present study, supplemented by a Development Set
group of specifically recruited normal subjects from the community. To evaluate potential metrics of SDB and to determine the best cut
All patients whose complete sleep history or physical examination points for each metric, data from 34 subjects were randomly selected
strongly suggested a primary nonrespiratory cause of EDS were ex- without stratification. This group consisted of 13 nonsnorers without
cluded (symptoms strongly suggesting narcolepsy, poor sleep hygiene, EDS and 21 patients with suspected sleep-disordered breathing (based
etc.). All patients who presented with predominant symptoms of snor- on the history of severe snoring and EDS). No snorers without EDS
ing and did not complain of EDS were included (n ⫽ 21). For patients were included in the development set.
presenting with EDS and thus were clinically suspected to have sleep-
disordered breathing, the initial nocturnal polysomnogram (NPSG) Test Set
was reviewed. We discarded all patients with severe obstructive apnea
To test the utility of the best metrics developed above (and the associ-
syndrome; that is, an apnea/hypopnea index (AHI) ⬎ 45/h. For this
ated cut points), we prospectively evaluated the data of the remaining
purpose, apnea/hypopnea was defined as a ⬎ 50% reduction in air-
103 subjects. This group consisted of 14 nonsnorers without EDS, 21
flow lasting ⬎ 10 s. These patients were eliminated because the focus
snorers without EDS, and 68 patients with suspected SDB. The “test”
of the present study was on the clinically relevant question of deter-
set thus contained two groups of subjects who had been clinically clas-
mining the physiologic metric most relevant to EDS in UARS and
sified prior to NPSG: subjects without EDS (nonsnoring or snoring)
mild/mod OSAS, and we did not wish to bias our statistics with easily
and patients with EDS and snoring.
identifiable severe OSAS. In total, we retained 89 patients with sus-
pected sleep-disordered breathing. An additional 27 normal subjects Sleep Studies
were recruited who had no history of either EDS, snoring, or any
other sleep complaints. This resulted in a total of 137 subjects whose A full NPSG was recorded in each subject at the New York University
data were analyzed. The protocol was approved by the New York Uni- Sleep Disorders Center. Recordings of central and occipital electroen-
versity Institutional Board of Research Associates. cephalogram (EEG), electrooculogram (EOG), and submental elec-
tromyogram (EMG) were used to monitor sleep. Sleep was scored us-
ing the criteria of Rechtschaffen and Kales (24). Leg movements were
Evaluation of Symptoms at Baseline monitored with an anterior tibialis EMG. A unipolar electrocardio-
All 110 clinical patients in the present study sought medical attention gram (ECG) was used for cardiac monitoring. Oxygen saturation was
for symptoms that were clinically attributed to SDB (89 for EDS and monitored with a pulse oximeter. Chest wall and abdominal move-
snoring and 21 for snoring alone). EDS was defined as a global clinical ment were monitored with piezoelectric strain gauges. Respiratory air-
impression by the sleep clinician of inappropriate and excessive ten- flow was simultaneously monitored with a nasal/oral thermistor and a

Figure 1. Example of flow limitation event. Note the period of breaths with flattened inspiratory flow contour (marked by the bar) and the rapid
return of the flow contour to a sinusoidal shape on arousal from sleep.
400 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 163 2001

TABLE 1
DEFINITION OF SDB INDICES BY THE EVENTS INCLUDED

Flow Hypopnea Flow Limitation Events

With O2 Without O2 With O2 With No

Apnea Desaturation Desaturation Desaturation With Arousal Desaturation/Arousal

AHIFlow x x x
AHIChicago-Cannula x x x x x
RDIDesat x x x
RDIArousal x x x x
RDITotal x x x x x x

nasal cannula connected to a 2-cm H2O pressure transducer (Pro- AHIChicago-Cannula only in that it does not include flow limitation
tech [Minneapolis, MN] or Validyne [Northridge, CA]) as previously events with desaturation but no arousal (of which there are few). It
described (15, 16, 25). is included because in a previous publication (21) we suggested this
Respiratory events were identified on the basis of nasal cannula index
flow signal and defined as follows: RDITotal: Calculated from the sum of apneas, all flow hypopneas and
all flow limitation events without regard to whether there is associ-
Apneas: Absence of airflow on the nasal cannula and ⬍ 10% baseline ated desaturation or arousal
fluctuations on the thermistor signal, lasting for ⬎ 10 s
Flow hypopneas (based on airflow measurement only): Airflow am- In addition to the above-described analyses based on using a single
plitude on the nasal cannula ⬍ 50% of surrounding baseline, last- index to describe SDB, we generated a bidimensional analysis using
ing for ⬎ 10 s. No O2 desaturation or arousal confirmation was re- the apnea index (AI) and the sum of all other events (RDITotal–AI)
quired to score an event separately. This was motivated by our previously published prelimi-
Flow limitation events: Any series of two or more breaths (lasting nary data (26).
⬎ 10 s) that had a flattened or nonsinusoidal appearance on the in-
spiratory nasal cannula flow signal and ended abruptly with a re- Statistical Analysis
turn to breaths with sinusoidal shape (15, 16). An example of such A receiver operator curve (ROC) was constructed for each index for
an “event” is shown in Figure 1. To be counted as a “flow limita- the development set, and cut points were chosen to optimize the sepa-
tion event,” the individual breaths had to have a peak flow be- ration between EDS and no-EDS groups. Sensitivity and specificity
tween 50 and 70% of the peak flow of the surrounding baseline along with their 95% confidence intervals were then calculated. The
(and thus not be a flow hypopnea) optimal cutoff for separating subjects with and without EDS was then
American Academy of Sleep Medicine (AASM) hypopneas: As pro- chosen for each index to simultaneously maximize the sum of sensitiv-
posed by the AASM Task Force (10), these events include both flow ity and specificity. The “best” two metrics (maximal area under the
hypopneas and any flow limitation event associated with 3% desatu- ROC) were chosen from among the indices evaluated. The test set
ration or associated with an AASM arousal was then evaluated prospectively, using only the best indices.
From these types of events, the following indices of SDB were defined
by dividing the relevant number of events by the total sleep time RESULTS
(TST). Table 1 shows which events are included in each index: Demographic and polysomnographic data are shown for the
AHIFlow: Calculated from the sum of all apneas and all flow hypopneas development set in Table 2A and for the test set in Table 2B.
AHIChicago-Cannula: Calculated from the sum of all apneas and all Both sets of subjects are typical of those seen in clinical prac-
AASM hypopneas. This differs from the AHIFlow in that it includes tice and span the spectrum from normal to mild–moderate SDB.
some additional flow limitation events when these are “upgraded” Figure 2 shows plots of the six respiratory indices we de-
to a hypopnea by being associated with an arousal or a 3% desatu- fined (AI, AHIFlow, RDIDesat, RDIArousal, RDIChicago-Cannula and
ration RDITotal). For each index, the left graph shows raw data for
RDIDesat: Calculated from the sum of all apneas and only those non-
both development set and test sets. Three groups (nonsnorers
apnea events (flow hypopneas and flow limitation events) with O2
desaturation. This differs from the AHIChicago-Cannula because it does without EDS, snorers without EDS, and snorers with EDS)
not include flow hypopneas without desaturation or flow limitation are indicated. The cutoff (based on the development set
events with arousal only alone), which separates those with EDS from those without
RDIArousal: Calculated from the sum of apneas, all flow hypopneas, EDS, is shown by the dashed line. On the right is shown the
and only flow limitation events with arousal. This differs from the ROC curve for that index that was used to choose this cut

TABLE 2
DEMOGRAPHIC AND POLYSOMNOGRAPHIC DATA FOR THE DEVELOPMENT SET AND FOR THE TEST SET

Sex Age BMI TST AI AHIFlow AHIChicago-Cannula RDIDesat RDIArousal RDITotal

Group No. (f/m) (yr) (kg/m2) ESS (min) (per hour) (per hour) (per hour) (per hour) (per hour) (per hour)

A. Development set (n ⫽ 34)

No snoring, no EDS 13 1/12 23–65 18–35 0–10 226–429 0–2 0–15 1–17 0–14 1–16 2–17
Snoring, EDS 21 5/16 24–67 21–46 4–19 170–458 0–12 1–33 10–48 2–20 8–46 11–71
B. Test set (n ⫽ 103)
No snoring, no EDS 14 9/5 18–58 20–42 1–7 89–454 0–3 0–17 1–20 0–11 1–19 1–21
Snoring, no EDS 21 3/18 14–71 20–51 0–11 122–457 0–45 0–53 0–55 0–49 0–53 2–57
Snoring, EDS 68 17/51 14–69 21–47 2–24 179–463 0–32 1–43 3–53 0–41 3–49 5–59

Definition of abbreviations: AHIChicago-Cannula ⫽ apnea ⫹ flow hypopnea ⫹ flow limitation events with desat/arousal index; AHIFlow ⫽ apnea ⫹ flow hypopnea index; AI ⫽ apnea index;
BMI ⫽ body mass index; EDS ⫽ excessive daytime somnolence; ESS ⫽ Epworth Sleepiness Scale; RDIArousal ⫽ apnea ⫹ flow hypopnea ⫹ flow limitation events with arousal index;
RDIDesat ⫽ apnea ⫹ flow hypopnea with desat ⫹ flow limitation events with desat index; RDITotal ⫽ apnea ⫹ flow hypopnea ⫹ flow limitation event index; TST ⫽ total sleep time.
Hosselet, Ayappa, Norman, et al.: Classification of SDB 401

Figure 2. Plot of AI, AHIFlow, AHIChicago-Can-

nula, RDIDesat, RDIArousal, and RDITotal for all
subjects in the development set (n ⫽ 34)
and test set (n ⫽ 103). Open symbols in-
dicate subjects without excessive daytime
somnolence (EDS): nonsnoring (NoSn) sub-
jects (open circles), snoring (Sn) subjects
(open triangles). Subjects with EDS are
shown with filled symbols (filled circles).
On the right are the receiver operator curves
(ROCs) for each index along with the area
under the curve (AUC). The dashed line
indicates the optimal cut point for each
index obtained using the ROC curve.

point. As can be seen from the ROC curves, the RDITotal ap- used in common practice. Of note, the cut point or 5/h for the
peared to have the best discriminant ability (greatest area un- AHIChicago-Cannula (proposed by the AASM to be of some rele-
der the curve, AUC ⫽ 0.95). For the ROC curve of RDITotal, vance) results in a sensitivity of 100% but a specificity for
the optimal combination of sensitivity and specificity was ob- EDS of only 15%.
tained at a cut point of 18 events/h. In the development set, 18 Using the bidimensional analysis the best cut points (AI ⬎
of 21 subjects with EDS and snoring had an RDITotal ⭓ 18 2 and RDITotal–AI ⬎ 16) resulted in a sensitivity of 86% and
events/h and 13 of 13 subjects with no EDS/snoring had an specificity of 100%, which were identical to those produced by
RDITotal ⬍ 18 events/h. This resulted in a sensitivity of 86% the RDITotal ⬎ 18 events/h alone. We also examined other sta-
and specificity of 100% in the development set for patients tistical classification techniques to separate subjects with EDS
thought by clinical assessment (snoring and EDS) to have up- from those without EDS (principal components, discriminant
per airway dysfunction. It is also apparent from the ROC analysis, and CART [Systat version 9; SPSS Inc., Chicago, IL]
curves that AI alone and AHIFlow have poor discriminant abil- analysis). None of these methods resulted in better separation
ity. RDIDesat, RDIArousal, and RDIChicago-Cannula have a better between groups, and we could not justify their use.
discriminant ability, but are still inferior to RDITotal. Table 4 shows the sensitivity and specificity for EDS ob-
Table 3 shows the sensitivity and specificity in the develop- tained in the test set subjects, using the best indices and cut
ment set for EDS, using the “best” cut point for each index (as points obtained from the development set (see Table 3). For
identified on the basis of the ROC curves for that index). Also RDITotal, 48 of 68 subjects with snoring and EDS (symptom-
shown are sensitivity and specificity data for other cut points atic) fell above the cutoff value of 18 events/h. Twenty-one of
402 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 163 2001

TABLE 3
DEVELOPMENT SET*

Index Sensitivity for Specificity for

(per hour) Cut Point EDS EDS

AI 2† 52% (11/21) 100% (13/13)

5 29% (6/21) 100% (13/13)
AHIFlow 5 81% (17/21) 39% (5/13)
10 57% (12/21) 69% (9/13)
15† 48% (10/21) 100% (13/13)
AHIChicago-Cannula 5 100% (21/21) 15% (2/13)
10 90% (19/21) 62% (8/13)
15† 71% (15/21) 92% (12/13)
RDIDesat 5 81% (17/21) 69% (9/13)
6† 71% (15/21) 85% (11/13)
10 38% (8/21) 92% (12/13)
15 29% (6/21) 100% (13/13)
RDIArousal 5 100% (21/21) 15% (2/13)
10 81% (17/21) 62% (8/13)
15 62% (13/21) 92% (12/13)
17† 62% (13/21) 100% (13/13)
RDITotal 5 100% (21/21) 15% (2/13)
10 100% (21/21) 46% (6/13)
15 90% (19/21) 77% (10/13)
18† 86% (18/21) 100% (13/13)
20 67% (14/21) 100% (13/13)
Figure 3. Plot of apnea index (AI) against nonapnea event index
AI 2† (RDITotal–AI) for 103 test subjects. Open circles represent asymptomatic
(RDITotal–AI) 16† 86% (18/21) 100% (13/13) subjects, open triangles represent snorers without EDS, and closed cir-
bidimensional 5 86% (18/21) 77% (10/13) cles represent subjects with EDS and snoring. The region bounded by
15 AI ⬍ 2/h and RDITotal–AI ⬍ 16/h was defined as identifying the “nor-
* n ⫽ 34. mal” region.
†
Best.

for patient complaints. Despite this, a significant number of

35 subjects without EDS (asymptomatic) fell below this value. individuals will have both tests “normal” and still have clinical
Specifically, 13 of the 14 subjects who had no snoring and no complaints of EDS.
EDS had an RDITotal ⬍ 18 events/h. The one subject who de-
nied snoring and had no EDS and had an RDITotal ⭓ 18 DISCUSSION
events/h was observed to have loud snoring during the NPSG.
On the other hand, of the 21 snorers who denied EDS, 8 had In the upper airway resistance syndrome, as in OSAS, sleep
an RDITotal ⬍ 18 events/h. disruption and symptoms of excessive daytime somnolence are
Figure 3 shows the test set data plotted using the two dimen- thought to result from repetitive respiratory events. Many
sional analysis based on apnea index and RDITotal–AI. The two of these events may be missed by conventional (thermistor-
cut points shown are those derived from the development set. based) monitoring (10, 15). A consensus statement from the
Shown in Figure 4 is the ESS for all subjects and mean AASM has suggested that these otherwise missed events can
sleep latency (SL) in a subgroup of 32 subjects who had an be detected by esophageal manometry and should be called
MSLT. In both the normal group and the snoring subjects who RERAs. We have previously shown (21) that one can use flow
denied EDS, both the ESS and SL are in the “normal” range limitation detected on a nasal cannula flow signal to identify
(ESS ⬍ 10 and SL ⬎ 10). However, in the group with clinically these events noninvasively. The present article explores the
defined EDS, these indices cover the entire range of values relationship between various indices of respiratory distur-
(3–24 for Epworth, 3–20 min for SL) and are frequently not in bance and EDS and shows that the simple index based on the
synchrony. If either test is abnormal, this has predictive value total number of apneas, hypopneas, and flow limitation events
provides a better sensitivity and specificity than other com-
monly used indices. This is particularly relevant as the group
TABLE 4 we analyzed was restricted to patients whose equivocal physi-
TEST SET* ology represents a challenge for the clinician.
The current recommendations (AASM) for defining respi-
Index Sensitivity for Specificity for ratory abnormality suggest that one should include RERAs in
(per hour) Cut Point EDS EDS
assessing the amount of sleep-disordered breathing, but give
RDITotal ⭓ 18 71% (48/68) 60% (21/35) little useful guidance as to the cut point between normal and
⭓ 20 66% (45/68) 63% (22/35) disease if these events are included. Our data indicate that typi-
AHIChicago-Cannula ⭓5 97% (66/68) 26% (9/35) cal cut points of 5 and 10 events/h are too low. This is particu-
⭓ 15 69% (47/68) 63% (22/35) larly relevant for the RDITotal, where a cut point of 18 events/h
AI (RDITotal–AI) ⭓2 79% (54/68) 54% (19/35) had the best discriminant utility, purely from respiratory infor-
⭓ 16 mation in the NPSG.
AI ⭓5 81% (55/69) 51% (18/35) Despite optimal separation, 20 of the 68 subjects with EDS
(RDITotal–AI) ⭓ 15
in our test set (clinically suspected sleep-disordered breath-
*n ⫽ 103. ing) had an RDITotal below the best cut point we could identify
Hosselet, Ayappa, Norman, et al.: Classification of SDB 403

Figure 4. Plot of Epworth Sleepiness Scale (ESS) for all

subjects. Sleep latencies for a subset of patients who
had an Multiple Sleep Latency Test (MSLT) are shown
on the right. The dashed line is the conventional cutoff
used for EDS: ESS ⬎ 10 or MSLT ⬍ 10 min.

(⬍ 18 events/h). On further review of all available data in number of flow limitation events, and this cannot be reflected
these patients, including the NPSG, we found that 9 (45%) of in a count of total events. Support for this idea is provided by
these 20 patients had a coexistent nonrespiratory potential the observation of Stradling et al. that arousals following ap-
cause for their EDS that was not apparent at the time of initial neas last longer than those following hypopneas; however,
clinical interview. Examples of these causes were narcolepsy these authors did not relate arousals to symptoms (27). Pre-
and periodic limb movements (PLMs) not initially in the dif- liminary data from our laboratory suggest that the conceptu-
ferential diagnosis. In contrast, the same type of review of the ally attractive but slightly more intricate bidimensional ap-
remaining 48 subjects with EDS who had RDI ⬎ 18 events/h proach may have advantages in evaluating subjects with
identified only 8 subjects (17%) who had a coexistent cause of positional or sleep stage-dependent SDB (28) or in evaluating
EDS. These data suggest that an RDITotal above the cutoff change due to partially effective therapy (e.g., dental appli-
value of 18 events/h is a strong physiologic correlate of a respi- ances) (29). This is particularly true when the comparison be-
ratory basis for the symptom of EDS. Conversely, an RDITotal tween two studies involves conversion between types of
⬍ 18 events/h has utility in predicting that a subject has a non- events (e.g., apnea to hypopnea, flow limitation events or
respiratory contribution to his EDS. Thus, our results suggest RERAs).
that the RDITotal, while not perfect, begins to approach the Several additional issues are raised by our methodology.
sensitivity required for a screening tool, that is, one whose These include (1) the definition of EDS and its choice as the
purpose is to “rule out” significant SDB. outcome measure, (2) the effect on sensitivity and specificity
The level of respiratory abnormality (RDITotal) in our 21 of our criteria for selection of subjects and the classification as
snorers without EDS was highly variable. Thirteen of them had snorers without EDS as “nondisease,” and (3) our definition
RDITotal above our cut point of 18 events/h (range, 18.5–55/h). of “mild” respiratory events.
Not only did these snorers have a total respiratory index simi-
lar to that found in individuals who had clear sleep-disordered 1. Our criterion for the definition of EDS was entirely sub-
breathing (who had EDS), but they also had similar numbers jective and relied on patient report and/or the global interpre-
of apneas and hypopneas. This suggests that different suscep- tation of the history by the expert sleep clinician. The MSLT is
tibility may exist to the same physiologic respiratory stress, at the acknowledged “gold standard” for assessing sleepiness on
least for EDS and that there is no perfect way to separate indi- the basis of the assumption that physiological need (an in-
viduals completely by counting the respiratory events on the creased tendency to fall asleep) leads to symptoms. However,
NPSG. Furthermore, we did not find any significant differ- there is a complex relationship between this test and the sub-
ences in age, sex, weight, or other characteristics in these two jective perception of symptoms that brings a patient to the
groups. Finally, we did not find evidence of the differences be- sleep clinician (30, 31). This is likely to be particularly true in
tween snorers being due to be an error of reporting subjective mild disease. The Epworth Sleepiness Scale (ESS) is a self-
EDS. Of the 21 snorers without EDS, 8 had either an MSLT reported subjective scale used to assess EDS, but this scale has
or MWT performed and there was no correlation between the been shown to have only a modest correlation with either
RDITotal and the sleep latencies. Only two (RDITotal of 15/h MSLT (32) or patient complaints (33), and our own data con-
and 54/h) of these eight snorers who reported no EDS and had firm the lack of a strong relationship. Our goal in the present
an MSLT done had sleep latencies ⬍ 10 min. study was to define an “outcome” measure with which to cor-
In our analysis, the best classification of our subjects was relate the physiologic respiratory data from the NPSG. Given
obtained with a single index, the RDITotal. A similar level of the known limitations of the objective tests of physiologic
separation was achieved with a bidimensional classification sleepiness and of existing subjective scales, we chose to use as
scheme of AI and other nonapnea events (RDITotal–AI). Be- an “outcome” the representation of EDS we believe is most
cause of its simplicity, we have emphasized the use of the sin- general and perhaps most frequently clinically used, that is,
gle RDITotal. Despite this, the bidimensional approach has a the global assessments by the patient and sleep clinician.
conceptual advantage in that different respiratory events may It must be noted that EDS is only one of several outcome
carry different weights in determining symptoms. Thus, a variables that could have been used to test the utility of physi-
small number of apneas may be more important than a larger ologic measurements of SDB. Other studies have raised the
404 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 163 2001

possibility of a link between SDB and hypertension, cardio- questioned whether it might be possible to use the current
vascular and cerebrovascular morbidity, and mortality. There data set to evaluate an index calculated without an EEG-
is no a priori reason to assume that the same link would exist based denominator of sleep period. Using the time from lights
between an index of SDB and each of these outcomes. In the out to lights on as the denominator for calculating a non-EEG
present study we looked only at EDS, because this variable RDITotal, we repeated the ROC analysis in the development
seems most logically linked to the disruption of sleep experi- set and found that a lower optimal cutoff (15 events/h) re-
enced by our patient population. If a “best” respiratory index sulted in a sensitivity of 86% and specificity of 100%, and a
can be identified that correlates with EDS, it would be logical sensitivity of 69% and specificity of 60% in the test set. These
to begin testing this index as an exposure variable for other results were similar to those obtained with the EEG-based
outcomes, rather than using indices that do not appear to have RDITotal with the cutoff of 18 events/h. Thus, in our data set, a
a tight association with EDS. non-EEG-based index did not reduce the ability of the best
2. The selection of our sample is discussed in METHODS. We respiratory index to discriminate between disease and normal.
specifically excluded subjects with EDS and severe obstruc- However, it is important to note that all of our NPSGs were
tive apnea (AHIFlow ⬎ 45/h) in order to focus our results on collected in a monitored setting and that the extrapolation of
the clinically most vexing group (UARS and mild/moderate our conclusions to an unmonitored setting needs further eval-
OSAS). By definition, all subjects eliminated by this restric- uation to determine the effect on the denominator.
tion would have been true positives in our analysis. The effect In conclusion, our data reemphasize the complex relation-
of this is that we decreased measured sensitivity in our test set, ship between clinical symptoms and the magnitude of the respi-
thus making our results more conservative. ratory disturbance during sleep despite everything done to
A separate issue affects our analysis of specificity. In our maximize detection of respiratory events. However, a respira-
study, specificity was defined for EDS. Of more clinical inter- tory index (RDITotal) that includes apneas, hypopneas, and
est is the specificity of the obtained respiratory indexes and events defined by flow limitation is superior to other indices in
cut points for disease. In this regard, the decision we made to its relationship to the clinical symptom of EDS. We propose
categorize snorers without EDS as nondiseased, as opposed to that this index improves the physiological description of the full
using snoring as a marker of disease, will affect the calculation spectrum of SDB from normal to UARS and OSAS and may
of specificity of any respiratory index. At this time, the clinical be more useful than the currently defined AHI in evaluating
significance of otherwise asymtomatic snoring is an unsettled the relationship between SDB and various clinical outcomes.
issue. It can be argued that a snorer with an AHI ⬎ 30 events/h
Acknowledgment : The authors acknowledge the assistance of Harold Diet-
(by any definition of AHI) should be considered as having dis- zius, Rakhil Kanevskaya, Rebecca O’Malley, Alison Rosen, Helen Valentine,
ease even without the symptom of EDS, rather than being and Maureen Wallace in scoring of the records.
classified as normal because he is asymptomatic. This would
convert one of our errors of classification into a correct result References
and improve specificity of the respiratory index; however, it
1. Young T, Palta M, Dempsey J, Skatrud J, Weber S, Badr S. The occur-
represents circular reasoning if one wants to test the respira- rence of sleep-disordered breathing among middle-aged adults. N
tory metric. Engl J Med 1993;328:1230–1235.
3. The use of the nasal cannula signal and our classification 2. Guilleminault C, Stoohs R, Clerk A, Cetel M, Maistros P. A cause of ex-
of event types require some discussion. The AASM criteria for cessive daytime sleepiness: the upper airway resistance syndrome.
sleep-disordered breathing suggest it is appropriate to detect Chest 1993;104:781–787.
3. Fleetham JA. A wake up call for sleep disordered breathing. Br Med J
apnea and hypopnea by nasal cannula, and respiratory event-
1997;314:839–840.
related arousal by esophageal manometry and EEG arousals. 4. Phillipson EA. Sleep apnea—a major public health problem. N Engl J
Because we have previously shown that an easily identified Med 1993;328:1271–1273.
pattern of changes in shape of the inspiratory airflow derived 5. Guilleminault C, Stoohs R, Shiomi T, Kushida C, Schnittger I. Upper
from the nasal cannula signal (flow limitation event) occurs airway resistance syndrome, nocturnal blood pressure monitoring, and
simultaneously with a classic RERA defined by esophageal borderline hypertension. Chest 1996;109:901–908.
6. Silverberg DS, Oksenberg A. Essential hypertension and abnormal up-
manometry, we did not use this technique. Although flow lim-
per airway resistance during sleep. Sleep 1997;20:794–806.
itation events can occur without arousal (depending on the 7. American Thoracic Society. Sleep apnea, sleepiness, and driving risk.
definition of arousal [20]) flow limitation events terminated by Am J Respir Crit Care Med 1994;150:1463–1473.
arousal are nearly identical to the standard RERA (21). By 8. Dement WC. The perils of drowsy driving. N Engl J Med 1997;337:783–784.
one interpretation, the strict application of the AASM rules 9. Mitler MM, Miller JC, Lipsitz JJ, Walsh JK, Wylie CD. The sleep of
forces a flow limitation event (defined by a visible change in long-haul truck drivers. N Engl J Med 1997;337:755–761.
10. American Academy of Sleep Medicine. Sleep-related breathing disor-
flow contour most often associated with an amplitude reduc-
ders in adults: recommendations for syndrome definition and mea-
tion [our unpublished data]) terminated by an arousal to be surement techniques in clinical research. Sleep 1999;22:667–689.
classified as a hypopnea. However, to accommodate various 11. Quan SF, Howard BV, Iber C, Kiley JP, Nieto FJ, O’Connor GT, Rapo-
indices of abnormal respiration and be more consistent with port DM, Redline S, Robbins J, Samet JM, Wahl PW. The Sleep Heart
definitions of hypopnea based on other tools, we tabulated Health Study: design, rationale, and methods. Sleep 1997;20:1077–1085.
flow limitation events separately from hypopnea and defined 12. Kryger MH. Controversies in sleep medicine: terminology and defini-
tions in sleep-disordered breathing. Sleep 1997;20:1208.
our summary indices by summing various categories. Only in
13. Moser NJ, Phillips BA, Berry DT, Harbison L. What is hypopnea, any-
calculating AHIChicago-Cannula were the flow limitation events way? Chest 1994;105:426–428.
with arousal reclassified as hypopnea. 14. Redline S, Sanders M. Hypopnea, a floating metric: implications for
prevalence, morbidity estimates, and case finding (review, 26 refer-
Our focus in this data analysis was on defining a metric to ences). Sleep 1997;20:1209–1217.
quantify sleep-disordered breathing. However, the calculation 15. Norman RG, Ahmed MM, Walsleben JA, Rapoport DM. Detection of
respiratory events during NPSG: nasal cannula/pressure sensor versus
of an index requires a denominator of time, generally taken
thermistor. Sleep 1997;20:1175–1184.
as the time of EEG documented sleep. There is a growing in- 16. Hosselet JJ, Norman RG, Ayappa I, Rapoport DM. Detection of flow
terest in systems that propose to detect sleep-disordered limitation with a nasal cannula/pressure transducer system. Am J
breathing without incorporating sleep monitoring (34). We Respir Crit Care Med 1998;157:1461–1467.
Hosselet, Ayappa, Norman, et al.: Classification of SDB 405

17. Redline S, Kapur VK, Sanders MH, Quan SF, Gottlieb DJ, Rapoport 26. Hosselet JJ, Norman RG, Ayappa I, Oen K, Walsleben JA, Rapoport
DM, Bonekat WH, Smith PL, Kiley JP, Iber C. Effects of varying ap- DM. Use of flow limitation events to characterize sleep disordered
proaches for identifying respiratory disturbances on sleep apnea as- breathing [abstract]. Am J Respir Crit Care Med 1998;157:A650.
sessment. Am J Respir Crit Care Med 2000;161:369–374. 27. Stradling JR, Pitson DJ, Bennett L, Barbour C, Davies RJO. Variation
18. Loube DI, Andrada T, Howard RS. Accuracy of respiratory inductive in the arousal pattern after obstructive events in obstructive sleep ap-
plethysmography for the diagnosis of upper airway resistance syn- nea. Am J Respir Crit Care Med 1999;159:130–136.
drome. Chest 1999;115:1333–1337. 28. Hosselet JJ, Rapoport DM, Rosen A, Norman RG, Ayappa I. Effect of
19. Strollo PJJ, Sanders MH. Significance and treatment of nonapneic snor- position and sleep stage on flow limitation events in mild–moderate
ing. Sleep 1993;16:403–408. obstructive sleep apnea syndrome [abstract]. Sleep 1998;21:414.K1.
20. O’Malley EB, Walsleben JA, Norman RG, Rapoport DM. Detection of 29. Hosselet JJ, Martinez JM, Gelb M, Binder D, Norman RG, Rapoport
unappreciated respiratory-related EEG arousals [abstract]. Am J Respir DM. Utility of identifying flow limitation in assessing the efficacy of
Crit Care Med 1996;153:A568. oral appliances in mild sleep disordered breathing [abstract]. Am J
21. Ayappa I, Norman RG, Krieger AC, Rosen A, O’Malley R, Rapoport Respir Crit Care Med 1999;159:A248.
DM. 2000. Non-invasive detection of respiratory effort related arous- 30. Chervin RD, Aldrich MS, Pickett R, Guilleminault C. Comparison of
als (RERAs) by a nasal cannula pressure/transducer system. Sleep 2000; the results of the Epworth Sleepiness Scale and the Multiple Sleep La-
23:763–771. tency Test. J Psychosom Res 1997;42:145–155.
22. Johns MW. A new method for measuring daytime sleepiness: the Ep- 31. Chervin RD, Kraemer HC, Guilleminault C. Correlates of sleep latency
worth Sleepiness Scale. Sleep 1991;14:540–545. on the multiple sleep latency test in a clinical population. Electroen-
23. Carskadon MA. Measuring daytime sleepiness. In: Principles and prac- cephalogr Clin Neurophysiol 1995;95:147–153.
tice of sleep medicine. Kryger MH, Roth T, Dement WC, editors. 32. Benbadis SR, Mascha E, Perry MC, Wolgamuth BR, Smolley LA, Din-
Philadelphia: W.B. Saunders; 1994. p. 961–966. ner DS. Association between the Epworth Sleepiness Scale and the
24. Rechtschafen A, Kales A. A manual of standardized terminology, tech- Multiple Sleep Latency Test in a clinical population. Ann Intern Med
niques, and scoring system for sleep states of human subjects. Washington 1999;130:289–292.
DC: U.S. Government Printing Office; 1968. NIH Publication No. 204. 33. Chervin RD, Aldrich MS. The Epworth Sleepiness Scale may not reflect ob-
25. Guyatt AR, Parker SP, McBride MJ. Measurement of human nasal ven- jective measures of sleepiness or sleep apnea. Neurology 1999;52:125–131.
tilation using an oxygen cannula as a pitot tube. Am Rev Respir Dis 34. Stradling JR, Davies RJ. Is it necessary to record sleep? (review, 25 ref-
1982;126:434–438. erencess). Sleep 1996;19:S251–S254

View publication stats