ADVERTISEMENT

Issue Navigator

Volume 08 No. 02
Earn CME
Accepted Papers
Classifieds







Scientific Investigations

Relative Prolongation of Inspiratory Time Predicts High versus Low Resistance Categorization of Hypopneas

http://dx.doi.org/10.5664/jcsm.1774

Anne M. Mooney, M.D.; Khader K. Abounasr, M.D.; David M. Rapoport, M.D.; Indu Ayappa, Ph.D.
NYU School of Medicine, Division of Pulmonary, Critical Care, and Sleep Medicine, New York, NY

ABSTRACT

Study Objectives:

Sleep disordered breathing events conceptually separate into “obstructive” and “central” events. Esophageal manometry is the definitive but invasive means of classifying hypopneas. The purpose of this project was to identify noninvasive markers for discriminating high vs. low resistance hypopneas.

Methods:

Forty subjects with obstructive or central sleep apnea underwent diagnostic polysomnography with nasal cannula airflow and esophageal manometry; 200% resistance relative to reference breaths was used to define “high” resistance. Noninvasive parameters from 292 randomly selected hypopneas in 20 subjects were analyzed and correlated to resistance. The best parameter and cutoff for predicting high relative resistance was determined and tested prospectively in 2 test sets in the 20 remaining subjects. Test Set A: 15 randomly selected hypopneas in each subject; Test Set B: all hypopneas in 7 subjects.

Results:

In the development set, prolongation of inspiratory time during the 2 smallest breaths of a hypopnea (Ti) relative to baseline had the best correlation to high relative resistance. In the Test Set A, relative Ti > 110% classified obstructive events with sensitivity = 72%, specificity = 77%, PPV = 64%, NPV = 83%. Similar numbers were obtained for classification of hypopneas based on presence of flow limitation (FL) alone. When either relative Ti or presence of FL were used to define high resistance, sensitivity = 84%, specificity = 74%, PPV = 65%, NPV = 89%. Similar results were obtained for Test Set B.

Conclusions:

Relative prolongation of Ti is a good noninvasive predictor of high/low resistance in a dataset with both FL and NFL hypopneas. Combination of FL and relative Ti improves this classification. The use of Ti to separate obstructive and central hypopneas needs to be further tested for clinical utility (outcomes and treatment effects).

Citation:

Mooney AM; Abounasr KK; Rapoport DM; Ayappa I. Relative prolongation of inspiratory time predicts high versus low resistance categorization of hypopneas. J Clin Sleep Med 2012;8(2):177-185.


Sleep disordered breathing (SDB) is a term which includes both obstructive sleep apnea/hypopnea (OSAHS) and central sleep apnea (CSAS) syndromes, which are both prevalent conditions.16 Although CSAS is most commonly seen in congestive heart failure (CHF), OSAHS is also often seen in these patients.46 At the root of defining whether a patient has CSAS or OSAHS is the categorization of the individual events as central or obstructive; there are both diagnostic and therapeutic implications of the distinction. In concept, obstructive events are defined by high upper airway resistance, whereas central events are defined by low “drive.” However, high resistance and low drive may not be mutually exclusive. There is increasing recognition that both types of events may exist within the same patient6 and that both etiologies may be invoked in the pathogenesis of some events, leading to difficulty in the overall categorization of a patient's disorder as “obstructive” vs. “central.” Despite this, it is currently assumed that the distinction between OSAHS and CSAS has clinical significance, and may have important implications for both diagnosis and treatment. Thus establishing a noninvasive algorithm to characterize individual events which is useful in general polysomnography remains an important target.

For apnea, distinguishing the central vs. obstructive etiology is usually possible noninvasively by the presence/absence of thoraco-abdominal movement (effort). In contrast, for hypopnea (which accounts for the majority of events in SDB), distinguishing the etiology is much more problematic. Esophageal manometry combined with measurement of airflow is the accepted gold standard,79 as the ratio of effort to flow defines resistance. However, this technique is invasive, disruptive of sleep, and considered by most clinical laboratories as unrealistic for routine use.

BRIEF SUMMARY

Current Knowledge/Study Rationale: On polysomnography it is difficult to separate central from obstructive hypopneas without invasive monitoring. The purpose of the current project was to separate hypopneas into central (low resistance) and obstructive (high resistance) using non-invasive criteria derived from the flow signal alone.

Study Impact: An algorithm based on the inspiratory time and presence of inspiratory flow limitation on the airflow signal was useful in separating obstructive from central hypopneas. This non-invasive technique has the potential to improve clinical decision making and influence research design.

Pulse transit time has been shown in a small study to be a useful noninvasive tool for recognition of central hypopneas,10,11 but this data is not routinely collected. Noninvasive techniques based on airflow exist to identify collapsible upper airway behavior; this collapsible behavior generally identifies elevated upper airway resistance.1214 In particular, we and others have shown that the presence of inspiratory flattening on a nasal cannula/pressure transducer tracing15 strongly suggests flow limitation (FL) which corresponds to inappropriate (obstructive) Starling resistor behavior of the airway. Despite this, it is unclear whether absence of FL on a breath implies a central etiology for a given breath.12Figure 1 illustrates flow and esophageal data in events with and without FL and demonstrates both high and low resistance in at least some events where there is an absence of FL.

Each panel shows airflow and esophageal pressure obtained from the PSGs in 3 different subjects

Increasing effort and resistance during a hypopnea with flow limitation indicated by the presence of unequivocal flattening of the flow contour (A). Non-flow limited hypopnea with decreased relative effort and resistance (B). Non-flow limited hypopnea, but with increased relative effort and resistance (C). While the presence of subtle changes in the shape of the inspiratory flow curve may be identified by some observers, it is our experience that this represents cardiogenic oscillations, suggesting low resistance.

jcsm.8.2.177a.jpg

jcsm.8.2.177a.jpg
Figure 1

Each panel shows airflow and esophageal pressure obtained from the PSGs in 3 different subjects

(more ...)

The classification of hypopneas (defined by reduced ventilation), even by invasive means, is not without ambiguity. One conceptual approach is to use changes in “effort” alone during the hypopnea relative to stable breathing during sleep for separation of “central” versus “obstructive” etiologies. Relative increase in effort or maintenance of effort at a time of reduced airflow reflects obstruction of the upper airway. However, events with a relative decrease in effort can be associated with unchanged resistance (proportional decrease in airflow) or increased resistance (disproportionate decrease in airflow); thus decreased effort can be associated with both central or obstructive pathology and effort alone is not sufficient to separate central from obstructive hypopneas. For these reasons, a relative increase in resistance during a hypopnea may be the most appropriate measure for classifying hypopneas as obstructive, but this requires an invasive measurement not usually available in routine polysomnography.

The purpose of the current project was to develop an approach to separating hypopneas into central (low resistance) and obstructive (high resistance) using noninvasive criteria derived from the flow signal alone and validate this in a dataset enriched for central/ambiguous etiology.

METHODS

Subjects

The dataset consisted of 40 subjects selected from patients seen in the NYU Sleep Disorders Center. These subjects were chosen with the target of obtaining a wide spectrum of hypopneas with (FL) and without (NFL) inspiratory flow limitation. Twenty-two of these patients had previously undergone an NPSG with esophageal manometry as part of other research protocols for suspected obstructive sleep apnea (19) or suspected central sleep apnea (3). The remaining 18 patients were recruited because they had a clinical presentation suggestive of central sleep apnea and on NPSG had > 25% of hypopneas without flow limitation. No patient was taking sedative or hypnotic medication, but 4 subjects were on narcotic medications at the time of study (3 in development set, 1 in test set). Aggregate patient characteristics are shown in Table 1.

Patient characteristics

NYears of age Avg ± SDGender (%male)BMI Avg ± SDObstructive apnea index Avg ± SDCentral apnea index Avg ± SDRDI Avg ± SDNarcotic use (% set)
Development Set2050.9 ± 14.395%35.6 ± 9.816.9 ± 24.211.4 ± 27.762.8 ± 33.515%
Test Set2055.9 ± 16.890%33.9 ± 10.619.5 ± 24.63.3 ± 5.656.4 ± 31.15%

[i] Subjects encompassed a wide range of sleep disordered breathing. There were no significant differences be-tween severity of SDB, central or obstructive indices between subjects in the development and test sets.

table icon
Table 1

Patient characteristics

(more ...)

The study was approved by the Institutional Review Board of New York University School of Medicine, and all subjects signed informed consent prior to being included in this study.

NPSG Protocol

Polysomnography was performed using the standard clinical protocol recommended by the AASM8 with the addition of esophageal manometry. A nasal cannula pressure transducer system transducer (Protech PTAF2) was used to measure airflow along with an oral thermistor to detect mouth breathing.15 An esophageal catheter consisting of a thin catheter ending in a 10-cm latex balloon (Ackrad Labs, NJ) was placed transnasally following lidocaine anesthesia, and positioned in the lower third of the esophagus. Esophageal pressure measurements were made with a 100 cm H2O pressure transducer (Validyne, Northridge, CA). Chest wall and abdominal movement were monitored with uncalibrated respiratory inductance plethysmography. A single technician scored all sleep studies for sleep and respiration. Sleep was scored according to AASM criteria.8 Hypopneas were defined as events ≥ 10 sec where the square root of the airflow signal on the nasal cannula was reduced to < 70% regardless of desaturation or arousal, but not reduced to < 10% (in which case the event was rescored as an apnea). To become proportional to actual flow (which was needed to calculate resistance), the pressure change in the nose obtained with a nasal cannula needs to be linearized with a square root transform.16,17

Data Analysis

Each hypopnea used in the development and test set (see below) was characterized as FL or NFL using visual assessment. Specifically we examined the shape of each breath on the airflow/time curve for inspiratory flattening, which we have previously shown to be a reliable surrogate for actual detection of flow limitation on a flow/pressure curve.1820 While subtle changes in the inspiratory flow curve may be detected by some observers, our experience suggests that when these are inconsistent or ambiguous (e.g., Figure 1B, C), the correlation to the flow/pressure curve deteriorates. Each hypopnea was reviewed visually by 2 experienced researchers and classified as FL only if the 2 researchers agreed. For each hypopnea, the 2 consecutive breaths with the lowest flow were identified. Two consecutive reference breaths with the highest flow were identified from the 30 sec before or after the event. For each breath the following variables were extracted: Flow (peak value of square root of nasal cannula signal for NFL breaths, value of square root of nasal cannula signal during the plateau phase for FL breaths), ΔEsophageal Pressure (from end expiratory baseline [ΔEffort]), Resistance (ΔEffort / Flow), Inspiratory Time (Ti), Duty Cycle (Ti/Ttot), and Inspiratory/Expiratory duration (Ti/Te). For each variable, the values for the 2 breaths within the hypopnea were averaged. Similarly, for each variable the values for the 2 reference breaths were averaged. Data were expressed as the ratio of hypopnea-to-reference values in order to represent relative effort, relative resistance, relative Ti, relative Ti/Ttot and relative Ti/Te.

Development and test sets

The 40 sleep studies available for analysis were divided equally into 2 sets—one for development and one for testing of the algorithms. Each set contained 10 patients whose diagnostic PSG suggested primarily obstructive sleep apnea, and 10 patients where the esophageal manometry suggested predominantly central sleep apnea (CSA) or Cheyne Stokes respiration (CSR). In the development set, 5 events with inspiratory flattening of the flow/time signal and 10 without inspiratory flattening (where available) were selected from each study. The development set contained 73 FL and 219 NFL events. Equal numbers of events (where available) were selected from each patient in order to avoid overrepresentation from any given patient while generating rules for categorization of events. Data for the test set subjects were not analyzed until after analysis on the development set (and final definitions of noninvasive criteria) were completed. Test Set A contained 15 randomly selected hypopneas from each patient in parallel fashion as extracted from the development set (5 events with inspiratory flattening of the flow/time signal and 10 without inspiratory flattening [where available]). Test Set A contained 72 FL and 185 NFL events. Test Set B contained all the hypopneas observed in a subgroup of 7 patients. Four of these patients had classical obstructive sleep apnea, and 3 of these patients had atypical sleep disordered breathing with periodic breathing and predominantly NFL hypopneas. Test Set B contained 223 FL and 410 NFL events. Events in the development set and all test sets were chosen randomly without respect to body position during sleep.

Reference definition of individual events as “high” and “low” resistance

As stated in the introduction, relative resistance or relative effort are alternative ways to define obstructive vs central events using the “gold standard” invasive methodology of esophageal manometry. However, events with a relative decrease in effort do not all have a low resistance, and can be associated with either a proportional decrease in airflow (unchanged resistance) or disproportional decrease in airflow (representing increased resistance)21; thus decreased effort can be associated with both central or obstructive pathology. For the present analyses, we chose to define events as obstructive vs. non-obstructive using relative resistance and not relative effort. The relationship between these parameters in our dataset is shown in Figure 2. The high correlation (0.87) suggests similar results would be obtained when either relative resistance or relative effort is chosen as the basis for a reference classification of hypopneas, but importantly, 63% of events with reduced relative effort, i.e., < 100%, (a possible definition of “central”) have relative resistances in the 100% to 200% range. Thus, given our choice of relative resistance as the defining variable for separating “low” and “high” resistance events, a decision needed to be made with respect to the cut point separating inappropriately obstructive from central hypopneas. Whereas the breaths chosen within events were clearly during sleep, our reference breaths were chosen from periods surrounding SDB events that were usually but not always associated with subcortical arousal (e.g., heart rate acceleration without clear EEG arousal), cortical arousal, or awakening. A doubling of airway resistance has been described in normal subjects at sleep onset,22 and for this reason, we used > 200% as the cut point to define an inappropriate (above that seen with normal sleep) increase in relative resistance. This definition should identify the level of increased relative resistance having potential clinical significance for inappropriate obstruction of the upper airway. Thus, hypopneas were defined as obstructive if relative resistance of the hypopnea breaths was > 200% compared to the resistance of the reference breaths, and non-obstructive if relative resistance was < 200% compared to the reference breaths.

For each NFL event (n = 219) in the development set, relative resistance is plotted against relative effort of events

Relative resistance is highly correlated to relative effort. Relative resistance and relative effort are calculated as the ratio of the values for the 2 smallest breaths within each respiratory event and the 2 largest breaths within 30 seconds of the event (see methods).

jcsm.8.2.177b.jpg

jcsm.8.2.177b.jpg
Figure 2

For each NFL event (n = 219) in the development set, relative resistance is plotted against relative effort of events

(more ...)

To test the dependence of our conclusions on the cut point chosen to define “high” relative resistance, we repeated our analyses using 150% relative resistance to separate “obstructive” from “non-obstructive” events in defining “truth” in the reference data set against which we compared our noninvasive markers.

Analyses and Statistics

Within the development set, correlations between relative resistance, relative effort and each relative noninvasive parameter (Ti, Ti/Ttot, Ti/Te) were first examined only in NFL events. The noninvasive parameter with the highest correlation to relative resistance was submitted to ROC analysis in order to identify the best cutoff for noninvasive classification of resistance or effort. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for high relative resistance (> 200%) were calculated for the NFL events of the development set and Test Set A. An additional analysis was performed after inclusion of FL events in Test Set A and stratified by sleep stage (NREM/REM). Analyses were repeated using presence/absence of FL as the noninvasive parameters, and for the combination of relative Ti and presence of FL. These analyses were repeated for Test Set B. To test the sensitivity of our definition of “obstruction,” Test Set A was also analyzed according to alternative definitions (effort > 100% = obstructive, resistance > 150% = obstructive).

RESULTS

Development Set

As previously shown,23 events with inspiratory flow limitation (FL) showed an overall high (289% ± 142%) relative resistance. Events without flow limitation (NFL) showed a significantly lower relative resistance than FL events (151% ± 100%, p < 0.001), but, as shown in Figure 3, there was a substantial overlap of relative resistance between FL and NFL events. Thus flow contour alone was not sufficient to noninvasively predict low relative resistance.

Distribution of the relative resistance for FL and NFL events

Bars represent events having relative resistance within the bin indicated on the x axis. The height of each bar indicates the percentage of all events of that flow type (FL vs. NFL). There were a total of 73 FL events and 219 NFL events.

jcsm.8.2.177c.jpg

jcsm.8.2.177c.jpg
Figure 3

Distribution of the relative resistance for FL and NFL events

(more ...)

Figure 4 shows the relationship of the noninvasive flow parameters (relative Ti, relative Ti/Ttot and relative Ti/Te) to relative resistance in NFL events. All 3 noninvasive parameters were significantly correlated (p < 0.001) with relative resistance; the highest correlation was observed between relative Ti and relative resistance. This was also true when the noninvasive parameters were examined against relative effort (data not shown).

Utility of using noninvasive parameters of flow (x axis) to estimate relative resistance (y axis) of breaths

Data is shown for NFL events in the development set. Relative Ti (A); relative Ti/Te (B); Relative Ti/Ttot (C). Best correlation is with relative Ti (A).

jcsm.8.2.177d.jpg

jcsm.8.2.177d.jpg
Figure 4

Utility of using noninvasive parameters of flow (x axis) to estimate relative resistance (y axis) of breaths

(more ...)

Figure 5 shows an ROC curve for all NFL events in the development set using a cutoff of > 200% relative resistance (solid line) to define obstructive events. Relative Ti > 110% (i.e., prolongation of Ti) was the best cutoff to predict high relative resistance, with an AUC of 0.82. Using an alternative reference definition of obstructive events as relative effort > 100% (dotted line), ROC analysis identifies a similar Ti (> 104%) as the best cutoff to predict high effort, with an AUC of 0.76.

ROC analysis for relative Ti to predict high vs. low relative resistance (solid line) or high vs. low relative effort (dotted line)

Using a cutoff of > 200% relative resistance to define obstructive events, ROC analysis identifies Ti > 110% (relative prolongation of Ti) as the best cutoff to predict high resistance with an AUC of 0.82. With obstructive events defined as effort > 100% of baseline breaths, ROC analysis identifies relative Ti > 104% (prolongation of relative Ti) as the best cutoff to predict increased effort with an AUC of 0.76.

jcsm.8.2.177e.jpg

jcsm.8.2.177e.jpg
Figure 5

ROC analysis for relative Ti to predict high vs. low relative resistance (solid line) or high vs. low relative effort (dotted line)

(more ...)

Table 2 shows a classification analysis for NFL events in the development set. Sensitivity, specificity, PPV, and NPV are listed for using relative Ti > 110% to identify obstructive events defined as relative resistance > 200%. The table also lists a classification analysis for using relative Ti > 104% to identify obstructive events when these are defined as relative effort > 100%. By either criterion, relative Ti shows a high negative predictive value: in NFL events, absence of relative lengthening of Ti strongly suggests that an event is not obstructive. However, in NFL events, relative lengthening has little predictive value (PPV = 32% to 36%).

Development Set (219 NFL events)

NFL events only (n = 219)SensSpecPPVNPV
    Obstructive event defined as Relative Resistance > 200%67%75%32%93%
    Test positive = Relative Ti > 110%
    Obstructive event defined as Relative Effort > 100%75%70%36%93%
    Test positive = Relative Ti > 104%

table icon
Table 2

Development Set (219 NFL events)

(more ...)

Validation of Development Set Results in Test Sets

Test Set A

In NFL events in Test Set A (15 randomly selected hypopneas/patient from 20 patients distinct from patients in the development set), relative prolongation of Ti was again significantly correlated with relative resistance (r = 0.48, p < 0.001), although the correlation was lower than in the development set (where r = 0.65, p < 0.001).

Table 3A shows a classification analysis for NFL events (n = 185) in Test Set A. Sensitivity, specificity, PPV, and NPV for NFL events are shown for the same cutoffs and definitions used for the development set in Table 2. Results are similar to those seen in the development set. Table 3B shows results from a similar classification analysis when both NFL and FL events from Test Set A are included. Of note, the NPV (prediction of central events) remains high, despite the inclusion of the predominantly obstructive events implied by the presence of FL.

Test Set A

A NFL events only (n = 185)SensSpecPPVNPV
    Obstructive event defined as Relative Resistance > 200%63835289
    Test positive = Relative Ti > 110%
    Obstructive event defined as Relative Effort > 100%68693988
    Test positive = Relative Ti > 104%
B NFL + FL events (n = 257)SensSpecPPVNPV
    Obstructive event defined as Relative Resistance > 200%72776483
    Test positive = Relative Ti > 110%
    Obstructive event defined as Relative Effort > 100%82604988
    Test positive = Relative Ti > 104%

table icon
Table 3

Test Set A

(more ...)

Because of the known reduction of effort associated with REM, a separate classification analysis of using Ti to predict resistance by sleep stage is shown in Table 4. The data show that in REM there is an exaggeration of the above findings (in all sleep), with a very high negative predictive value (i.e., a short relative Ti is strongly predictive of a low relative resistance in NFL).

Test Set A, effect of sleep stage (REM/NREM) on test validity

NFL events only (n = 185)SensSpecPPVNPV
    REM (n = 44)
        Obstructive event defined as Relative Resistance > 200%50881797
        Test positive = Relative Ti > 110%
    NREM (n = 141)
        Obstructive event defined as Relative Resistance > 200%64815786
        Test positive = Relative Ti > 110%
NFL + FL events (n = 257)SensSpecPPVNPV
    REM (n = 57)
        Obstructive event defined as Relative Resistance > 200%60793890
        Test positive = Relative Ti > 110%
    NREM (n = 200)
        Obstructive event defined as Relative Resistance > 200%74766980
        Test positive = Relative Ti > 110%

table icon
Table 4

Test Set A, effect of sleep stage (REM/NREM) on test validity

(more ...)

Table 5 shows a classification analysis for events in Test Set A with and without flow limitation (FL and NFL), when the presence or absence of flow limitation alone is used to predict high or low resistance. These results are further analyzed with respect to sleep stage (REM/NREM). The absence of flow limitation in REM has very high negative predictive value.

Test Set A, classification results using presence or absence of flow limitation alone

NFL + FL events (n = 257)SensSpecPPVNPV
    All Events (n = 257)
        Obstructive event defined as Relative Resistance > 200%56887478
        Test positive = Flow limitation present
    REM (n = 57)
        Obstructive event defined as Relative Resistance > 200%80896295
        Test positive = Flow limitation present
    NREM (n = 200)
        Obstructive event defined as Relative Resistance > 200%54887672
        Test positive = Flow limitation present

table icon
Table 5

Test Set A, classification results using presence or absence of flow limitation alone

(more ...)

As a test of the sensitivity of our findings to the definition of “high” relative resistance (> 200%), we repeated the classification analysis of Ti as a predictor of resistance using a cutoff of 150%. Table 6 shows these data and suggests similar trends.

Effect of an alternative definition of resistance as > 150% in Test Set A

NFL events (n = 185)SensSpecPPVNPV
    Disease present = Relative Resistance > 150%48856475
    Test positive = Relative Ti > 110%
NFL + FL events (n = 257)SensSpecPPVNPV
    Disease present = Relative Resistance > 150%63807668
    Test positive = Relative Ti > 110%

table icon
Table 6

Effect of an alternative definition of resistance as > 150% in Test Set A

(more ...)

To incorporate all of the observations made above for noninvasive prediction, we present a combined classification analysis (Table 7) for predicting relative resistance > 200% for events in Test Set A using either the presence of FL or Ti > 110%. By these criteria, sensitivity rises to 84%, while specificity, PPV and NPV remain similar at 74%, 65%, 89%.

Test Set A, two-tiered classification analysis

NFL + FL events (n = 257)SensSpecPPVNPV
    Obstructive event defined as Relative Resistance > 200%84746589
    Test positive = Flow limitation present or if NFL then Relative Ti > 110%

table icon
Table 7

Test Set A, two-tiered classification analysis

(more ...)

Figure 6 compares the validity of each test (1-presence of flow limitation alone, 2-relative prolongation of Ti > 110% alone, 3-either presence of flow limitation or relative prolongation of Ti > 110%) to predict relative resistance > 200%.

Comparison of the sensitivity, specificity, PPV, and NPV for each noninvasive predictor of high resistance, in Test Set A events (FL and NFL)

A two-tiered classification analysis using either the presence of flow limitation or relative prolongation of Ti (> 110%) demonstrates the highest sensitivity while NPV is preserved.

jcsm.8.2.177f.jpg

jcsm.8.2.177f.jpg
Figure 6

Comparison of the sensitivity, specificity, PPV, and NPV for each noninvasive predictor of high resistance, in Test Set A events (FL and NFL)

(more ...)

Test Set B

In contrast to Test Set A, which consisted of an equal number of hypopneas randomly selected from all patients, Test Set B evaluated all hypopneas in 7 patients. Four of the patients demonstrated classic loud snoring and flow limitation throughout the majority of the night. The other three patients had predominantly non-flow limited hypopneas occurring in a periodic pattern. There were 633 hypopneas in Test Set B (410 NFL, 223 FL). Table 8 shows that the 3 noninvasive predictors (presence of FL, relative Ti > 110%, and the combination of the two criteria) show good sensitivity, specificity, PPV, and NPV in the data from Test Set B.

Classification results using the 3 noninvasive predictors in Test Set B

NFL + FL events (n = 633)SensSpecPPVNPV
    All hypopneas in 7 Patients (n = 633)
        Obstructive event defined as Relative Resistance > 200%
        Test positive = Flow limitation present71%89%81%82%
        Test positive = Relative Ti > 110%72%81%72%81%
        Test positive = Either FL present or Relative Ti > 110%84%77%71%88%

table icon
Table 8

Classification results using the 3 noninvasive predictors in Test Set B

(more ...)

DISCUSSION

The present analysis shows that useful predictions about whether a hypopnea is “obstructive” or “central” (defined by relative resistance) can be made from airflow alone. Our data show that the relative prolongation of Ti can be used to correctly classify the majority of hypopneas without use of invasive monitoring in a dataset enriched for ambiguous (non-flow limited) hypopneas. In this dataset, the combined assessment of relative Ti and inspiratory flow limitation had greater accuracy than either alone. An algorithm categorizing each hypopnea event first according to FL and then according to relative Ti, not only helps classify individual events, but may also help in the clinical classification of patients and thus potentially inform treatment decisions.

The present study was predicated on the assumption that there is a clear dichotomy between “central” and “obstructive” etiologies that can be applied to each individual hypopnea, and that this was represented by the relative resistance obtained using the invasive measurement of esophageal manometry. We acknowledge that there may be considerable overlap in the pathophysiology underlying a given event, and as a result there may not exist an unambiguously correct distinction between an obstructive and a central hypopnea. However, it is currently assumed that this distinction exists and has clinical significance as well as implications for treatment and prognosis2428; a dichotomous classification scheme is currently standard clinical practice.8

In the absence of consistent definitions of hypopnea,2931 we chose to use a very inclusive definition of hypopnea based on reduced airflow alone for our analysis,9 rather than one that disregards events without 4% desaturation or EEG arousal.8 In addition, our methodology for classifying hypopneas requires comment. The signals we used for flow are not calibrated. Thus all measurements (such as resistance) were not absolute, and needed to be related to a “baseline.” By calculating relative resistance, and relative Ti we build on the logic used to define hypopnea using the accepted change in relative amplitude. We used high vs. low relative resistance as the variable to provide the reference categorization of each SDB event, with > 200% being the cut point between obstructive and non-obstructive (see justification in methods).

Having chosen a definition for “correctly” classifying hypopneas as “obstructive” or “central,” the second component of our analysis was to use the now classified events as a test of various noninvasive algorithms to accomplish the same classification. Although relative prolongation of Ti alone provided a highly specific marker for high resistance (similar to flow limitation), it lacks sensitivity. Using the presence of either flow limitation or the prolongation of Ti (relative Ti > 110%) improves sensitivity while maintaining NPV for classifying events as “obstructive.” An automated technique utilizing discriminant analysis of flow limitation and other features extracted from the airflow signal was recently reported to demonstrate similar accuracy in predicting relative changes in effort32,33 and supports our conclusions.

The observed low sensitivity for detecting events in Test Set A with high relative resistance is predictable from the choice made in constructing this dataset. We specifically enriched the data set with hypopneas without flow limitation as we felt that these posed the highest ambiguity based on prior work.12 The sensitivity of our algorithm to high resistance events necessarily increases as one includes more patients (and events) that are flow limited and this is confirmed by the higher sensitivity seen in Test Set B that has greater number of FL events.

There are obvious advantages to an algorithm for separating obstructive from central events based on noninvasive measurements. Patient willingness, use of lidocaine, sleep effects, potential extension to home study data and, finally, cost, all favor a noninvasive test over esophageal manometry if comparable classification of events can be achieved. A possible limitation of our data is that quantitative flow was not measured using a pneumotachograph, and thus resistance cannot be quantitatively determined. Our choice of the nasal cannula signal was intentional in this regard, as we were trying to develop a technique was that was generalizable to routine clinical use, and pneumotachographs are generally reserved to specialized research applications. We and others14,18,34,35 have previously shown that the nasal cannula qualitative flow signal can accurately determine presence or absence of flow limitation. Furthermore, breath duration (used to define the respiratory variables we examined relative Ti, relative Ti/Ttot, relative Ti/Te) does not require quantitative airflow measurement. While quantitative resistance cannot be calculated from the nasal cannula signal, relative resistance should not suffer from this limitation.

Other aspects related to our methodology require comment. First, since airflow was collected with a nasal cannula/pressure transducer system where the calibration is only valid across short time periods (3 min) we needed to define “baseline breathing” for each event rather than using one single baseline period per subject We chose periods of breathing within 30 seconds of the hypopnea. Second, in order to maximize reproducibility of breaths chosen for analysis we use the 2 biggest breaths for use as reference and 2 smallest breaths within the hypopnea.

An interesting and striking finding in our study was the behavior of our algorithm specifically in REM. In this state, absence of flow limitation was almost always associated with low resistance. Thus, in REM there was little need to explore additional noninvasive parameters beyond flow limitation to classify hypopneas. One possible explanation is that the upper airway during REM is so susceptible to collapse that any increase in effort brings out flow limitation. In NREM the lesser collapsibility of the airway makes for more overlap between the level of effort and the expression of flow limitation.

A limitation of our dataset is the predominance of male patients. Gender related differences in load response and arousal threshold may affect the appearance of FL and relative prolongation of Ti.36 However, data from Pillar, et al.37 show that despite the greater prevalence of flow limitation and the reduced ventilation seen in men with inspiratory resistive loading, men and women had similar prolongation of inspiratory time. Further testing in a dataset with more women may be desirable before generalizing our conclusions.

In conclusion, we describe a noninvasive algorithm applied to data derived solely from the nasal cannula flow signal which is easily automatable and has good sensitivity and specificity for prediction of relative resistance of hypopneas. The utility of this algorithm lies in its ability to noninvasively classify obstructive and central hypopneas in clinical and research data sets of NPSGs, and thus test the utility and implications of this distinction for diagnostic, therapeutic and long term-outcomes.

DISCLOSURE STATEMENT

This was not an industry supported study. Dr. Rapoport has received support for research from Fisher & Paykel Healthcare, Ventus Medical, speaking and consulting engagements for Fisher & Paykel Healthcare. Dr. Rapoport holds multiple US and foreign patents covering techniques and analysis algorithms for the diagnosis o f OSAHS and techniques for administering CPAP. Several of these have been licensed to Biologics, Fisher & Paykel Healthcare, Advanced Brain Monitoring and Tyco (Health C'Aire). Dr. Indu Ayappa has received support for research from Fisher & Paykel Healthcare, Ventus Medical. Dr. Ayappa holds multiple US and foreign patents covering techniques and analysis algorithms for r the diagnosis of OSAHS and techniques for administering CPAP. Several of these have been licensed to Fisher & Paykel Healthcare and Advanced Brain Monitoring. The other authors have indicated no financial conflicts of interest.

ACKNOWLEDGMENTS

The authors thank Rakhil Kanevskaya for manual scoring of the sleep studies. Supported by grants from: American Sleep Medicine Foundation, NHLBI K25 HL04420, NCRRM01RR0096, Foundation for Research in Sleep Disorders.

REFERENCES

1 

Young T, Palta M, Dempsey J, Skatrud J, Weber S, Badr S, authors. The occurrence of sleep-disordered breathing among middle-aged adults. N Engl J Med. 1993;328:1230–5. [PubMed]

2 

Javaheri S, Parker TJ, Wexler L, et al., authors. Occult sleep-disordered breathing in stable congestive heart failure. Ann Intern Med. 122:487–92. [PubMed]

3 

Javaheri S, author. Central sleep apnea-hypopnea syndrome in heart failure: prevalence, impact, and treatment. Sleep. 1996;19 10 Suppl:S229–S231. [PubMed]

4 

Javaheri S, Parker TJ, Liming JD, et al., authors. Sleep apnea in 81 ambulatory male patients with stable heart failure. Types and their prevalences, consequences, and presentations. Circulation. 1998;97:2154–9. [PubMed]

5 

Redeker NS, Rapoport DM, authors. Central vs. obstructive respiratory events and daytime function in stable heart failure. Am J Respir Crit Care Med. 2007;175:A577Paper presented at.

6 

Redeker NS, Muench U, Zucker MJ, et al., authors. Sleep disordered breathing, daytime symptoms, and functional performance in stable heart failure. Sleep. 2010;33:551–60. [PubMed Central][PubMed]

7 

Kushida CA, author. The use of esophageal manometry in the diagnosis of sleep-related breathing disorders. Conf Proc IEEE Eng Med Biol Soc. 2004;5:3860–3. [PubMed]

8 

Iber C, Ancoli-Israel S, Chesson AL, Quan SF, authors; Westchester IL, editor. The AASM Manual for the Scoring of Sleep and Associated Events: Rules Terminology and Technical Specifications. 2007. 1st ed. American Academy of Sleep Medicine;

9 

American Academy of Sleep Medicine. Sleep-related breathing disorders in adults: Recommendations for syndrome definition and measurement techniques in clinical research. Sleep. 1999;22:667–89. [PubMed]

10 

Argod J, Pepin JL, Levy P, authors. Differentiating obstructive and central sleep respiratory events through pulse transit time. Am J Respir Crit Care Med. 2000;162:87–93. [PubMed]

11 

Argod J, Smith RP, Pepin JL, Levy PA, authors. Pulse transit time and esophageal manometry in the detection of non-apneic obstructive respiratory events. Am J Respir Crit Care Med. 1999;159:A791

12 

Clark SA, Wilson CR, Satoh M, Pegelow D, Dempsey JA, authors. Assessment of inspiratory flow limitation invasively and noninvasively during sleep. Am J Respir Crit Care Med. 1998;158:713–22. [PubMed]

13 

Mansour KF, Rowley JA, Meshenish AA, Shkoukani MA, Badr MS, authors. A mathematical model to detect inspiratory flow limitation during sleep. J Appl Physiol. 2002;93:1084–92. [PubMed]

14 

Mansour KF, Rowley JA, Badr MS, authors. Noninvasive determination of upper airway resistance and flow limitation. J Appl Physiol. 2004;97:1840–8. [PubMed]

15 

Hosselet JJ, Norman RG, Ayappa I, Rapoport DM, authors. Detection of flow limitation with a nasal cannula/pressure transducer system. Am J Respir Crit Care Med. 1998;157(5 Pt 1):1461–7. [PubMed]

16 

Thurnheer R, Xie X, Bloch KE, authors. Accuracy of nasal cannula pressure recordings for assessment of ventilation during sleep. Am J Respir Crit Care Med. 2001;164:1914–9. [PubMed]

17 

Montserrat JM, Farre R, Ballester E, Felez MA, Pasto M, Navajas D, authors. Evaluation of nasal prongs for estimating nasal flow. Am J Respir Crit Care Med. 1997;155:211–5. [PubMed]

18 

Ayappa I, Norman RG, Krieger AC, Rosen A, O'Malley RL, Rapoport DM, authors. Non-Invasive detection of respiratory effort-related arousals (REras) by a nasal cannula/pressure transducer system. Sleep. 2000;23:763–71. [PubMed]

19 

Hosselet J, Ayappa I, Norman RG, Krieger AC, Rapoport DM, authors. Classification of sleep-disordered breathing. Am J Respir Crit Care Med. 2001;163:398–405. [PubMed]

20 

Norman RG, Rapoport DM, Ayappa I, authors. Detection of flow limitation in obstructive sleep apnea with an artificial neural network. Physiol Meas. 2007;28:1089–100. [PubMed]

21 

Badr MS, Toiber F, Skatrud JB, Dempsey J, authors. Pharyngeal narrowing/occlusion during central sleep apnea. J Appl Physiol. 1995;78:1806–15. [PubMed]

22 

Hudgel DW, Martin RJ, Johnson B, Hill P, authors. Mechanics of the respiratory system and breathing pattern during sleep in normal humans. J Appl Physiol. 1984;56:133–37. [PubMed]

23 

Condos R, Norman RG, Krishnasamy I, Peduzzi N, Goldring RM, Rapoport DM, authors. Flow limitation as a noninvasive assessment of residual upper-airway resistance during continuous positive airway pressure therapy of obstructive sleep apnea. Am J Respir Crit Care Med. 1994;150:475–80. [PubMed]

24 

Arzt M, Floras JS, Logan AG, et al., authors. Suppression of central sleep apnea by continuous positive airway pressure and transplant-free survival in heart failure: a post hoc analysis of the Canadian Continuous Positive Airway Pressure for Patients with Central Sleep Apnea and Heart Failure Trial (CANPAP). Circulation. 2007;115:3173–80. [PubMed]

25 

Javaheri S, author. CPAP should not be used for central sleep apnea in congestive heart failure patients. J Clin Sleep Med. 2006;2:399–402. [PubMed]

26 

Teschler H, Dohring J, Wang YM, Berthon-Jones M, authors. Adaptive pressure support servo-ventilation: a novel treatment for Cheyne-Stokes respiration in heart failure. Am J Respir Crit Care Med. 2001;164:614–9. [PubMed]

27 

Philippe C, Stoica-Herman M, Drouot X, et al., authors. Compliance with and effectiveness of adaptive servoventilation versus continuous positive airway pressure in the treatment of Cheyne-Stokes respiration in heart failure over a six-month period. Heart. 2006;92:337–42. [PubMed Central][PubMed]

28 

Kasai T, Usui Y, Yoshioka T, et al., authors. Effect of flow-triggered adaptive servo-ventilation compared with continuous positive airway pressure in patients with chronic heart failure with coexisting obstructive sleep apnea and Cheyne-Stokes respiration. Circ Heart Fail. 2010;3:140–8. [PubMed]

29 

Redline S, Kapur VK, Sanders MH, et al., authors. Effects of varying approaches for identifying respiratory disturbances on sleep apnea assessment. Am J Respir Crit Care Med. 2000;161(2 Pt 1):369–74. [PubMed]

30 

Redline S, Budhiraja R, Kapur V, et al., authors. The scoring of respiratory events in sleep: reliability and validity. J Clin Sleep Med. 2007;3:169–200. [PubMed]

31 

Ruehland WR, Rochford PD, O'Donoghue FJ, Pierce RJ, Singh P, Thornton AT, authors. The new AASM criteria for scoring hypopneas: impact on the apnea hypopnea index. Sleep. 2009;32:150–7. [PubMed Central][PubMed]

32 

Morgenstern C, Schwaibold M, Randerath W, Bolz A, Jane R, authors. Automatic non-invasive differentiation of obstructive and central hypopneas with nasal airflow compared to esophageal pressure. Conf Proc IEEE Eng Med Biol Soc. 2010:6142–5. [PubMed]

33 

Morgenstern C, Schwaibold M, Randerath WJ, Bolz A, Jane R, authors. An invasive and a noninvasive approach for the automatic differentiation of obstructive and central hypopneas. IEEE Trans Biomed Eng. 2010;57:1927–36. [PubMed]

34 

Heitman SJ, Atkar RS, Hajduk EA, Wanner RA, Flemons WW, authors. Validation of nasal pressure for the identification of apneas/hypopneas during sleep. Am J Respir Crit Care Med. 2002;166:386–91. [PubMed]

35 

Hernandez L, Ballester E, Farre R, et al., authors. Performance of nasal prongs in sleep studies: spectrum of flow-related events. Chest. 2001;119:442–50. [PubMed]

36 

Jordan AS, McEvoy RD, Edwards JK, et al., authors. The influence of gender and upper airway resistance on the ventilatory response to arousal in obstructive sleep apnoea in humans. J Physiol. 2004;558(Pt 3):993–1004. [PubMed Central][PubMed]

37 

Pillar G, Malhotra A, Fogel R, Beauregard J, Schnall R, White DP, authors. Airway mechanics and ventilation in response to resistive loading during sleep: influence of gender. Am J Respir Crit Care Med. 2000;162:1627–32. [PubMed]