INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

An inadequate measurement of respiratory effort during sleep can be associated with false diagnosis of central apnea or hypopnea syndrome, leading to inappropriate treatment. These misclassifications of respiratory events are relatively rare when only central apneas are considered (1, 2). Conversely, it is quite impossible, using commonly available noninvasive techniques, to distinguish between central and obstructive hypopneas. Thus, the measurement of swings in esophageal pressure (Pes) by means of an esophageal catheter is the reference method (3) used to differentiate between obstructive and central apneas or hypopneas. However, this method is rather uncomfortable and may induce sleep disruption (4, 5), and the presence of an esophageal balloon in the upper airway (UA) may modify dynamic airway collapse (6). Thus, thoracic and abdominal excursions are widely used to detect respiratory movement using the dyssynchrony between rib cage and abdomen in order to estimate a progressive increase in respiratory effort. As UA resistance increases, either the rib cage or abdomen begins to move out of phase, with the "stronger" compartment leading the "weaker." During frank apnea, the movement can be entirely paradoxical. This detection technique depends on good movement sensors, and in the very obese subject it can be impossible to get adequate signals from any of the available transducers. Recently, the detection of flow limitation in the inspiratory airflow signal was shown to reflect an increase in UA resistance (7, 8). This can be recorded with a face mask and a pneumotachograph. This technique is of limited use in sleep studies because the face mask leads to sleep disturbance, particularly in patients with mild hypersomnia (9). More simply, a nasal cannula/pressure transducer could be used to infer flow-limited behavior (10). The potential limitation of nasal pressure recording is the reduction in amplitude and change in shape of the signal in the presence of significant oral breathing during sleep (11). Because of these limitations, the techniques described are inadequate as the only measurement of respiratory effort in many patients. New, validated and noninvasive alternatives to Pes measurement are therefore needed for evaluating respiratory effort during sleep.

Pulse transit time (PTT) is the time taken for pulse pressure to travel from the aortic valve to the periphery, and is measured as the time delay between the R-wave on the electrocardiogram (ECG) and the arrival of the pulse wave at the finger (detected by an infrared oximeter probe). PTT has been shown to be inversely correlated with blood pressure, and is capable of revealing acute changes during pulsus paradoxus, generated by high pleural pressure swings encountered during UA resistance in sleep (12). Pitson and colleagues (13) have demonstrated a close relationship between the increase in Pes due to episodes of UA resistance and a progressive rise in the amplitude of PTT oscillations between inspiration and expiration.

The aims of our study were therefore: (1) to evaluate the reliability of a visual analysis of the PTT signal as an estimation of respiratory effort and to examine interobserver agreement with regard to this analysis; and (2) to assess the ability of PTT measurement to classify sleep respiratory events as central or obstructive, using Pes as the "gold standard."

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Polysomnography

Continuous recordings were made of the electroencephalogram (EEG) with electrode positions C3/A2-Cz/O1 of the International 10-20 Electrode Placement System, as well as of eye movements and of the chin electromyogram (EMG) and electrocardiogram (ECG). Respiration was monitored with strain gauges. Airflow was measured with a pneumotachograph (Kontron Instruments, Saint-Quentin, France) or with oronasal thermistors and/or a nasal cannula when the patient did not tolerate the face mask. Respiratory effort was systematically measured by monitoring Pes with an esophageal catheter (Compliance Catheter, Volgens E.K.G.S.; International Medical, Zutphen, The Netherlands). Oxygen saturation was measured with a Biox-Ohmeda 3700 oximeter (Ohmeda, Louisville, CO).

The polysomnogram was scored manually according to standard criteria (14). Episodes of apnea were defined as complete cessations of airflow for 10 s or more, and episodes of hypopnea as decreases in oronasal airflow of more than 50% and lasting for at least 10 s. Apnea/ hypopnea events were classified as central, obstructive, or mixed according to the Pes signal (Figure 1). A central hypopnea was defined as a reduction in flow proportional to the decrease in respiratory drive. Thus, when flow was measured with a pneumotachograph (or nasal prongs), there was a quasilinear relationship between the reduction in flow and Pes values during central hypopnea (e.g., a 50% reduction in Pes corresponded approximatively to a 50% reduction in flow). When a thermistor was used, a linear relationship was not expected, owing to the qualitative nature of the sensor. In this case, a central hypopnea could be defined only when a clear reduction in respiratory effort was observed. During rapid-eye-movement (REM) sleep, the problem is even more complex, owing to the rapid changes in respiratory drive during such sleep, and especially during phasic events. Thus, the variation in respiratory effort may be more erratic. Only trends occurring during consecutive respiratory cycles could be interpreted. If the trend was a cresendo in Pes, or if Pes remained stable, the hypopnea was scored as obstructive. If the trend was a decrescendo, the hypopnea was scored as central.

View larger version (28K): [in this window] [in a new window]   Figure 1.   Visual analysis of the PTT signal. (A) Undersampling error of PTT signal as reflecting Pes. Cardiac cycles are represented by the points indicating each pulse. Breaths corresponding to a selected obstructive respiratory event are represented, with Pes shown by the thick line and PTT by the thin squared-wave line. Because a PTT value is only available with every QRS complex, and the heart rate is therefore not fast enough to prevent undersampling, a given maximum PTT value may or may not correspond exactly to the Pes nadir. (Modified from Pitson and colleagues [13].) (B) Recognition of an obstructive respiratory event. Pes becomes increasingly negative during obstructive hypopnea, and there is then a clear increase in oscillations in PTT between inspiration and expiration, as indicated by the dotted lines. (C ) Recognition of a central respiratory event. The determination of a central event with the PTT signal, as with the Pes signal, is made from the near disappearance of (central apneas) or reduction (central hypopneas) in respiratory oscillations.

Classification of Central and Obstructive Sleep Respiratory Events with PTT in Comparison with Pes

PTT measurement technique. PTT is most conveniently measured from the ECG R-wave rather than from aortic valve opening, and therefore includes a contribution from the left ventricular isometric contraction time (preejection period [PEP]), in addition to the actual pulse-wave transit time from the aortic valve to the periphery. As described by Pollak and coworkers (15), there was no difference between the ECG Q-wave or the ECG R-wave to begin timing the ECG-radial pulse interval.

In this study, beat-to-beat PTT measurements were made with an RM50 recorder (DeVilbiss, Parcay-Meslay, France). PTT was calculated as the interval between the ECG R-wave and the point on the pulse waveform that is 50% the height of the maximum value (as detected with photoplethysmography). With the RM50 recorder, PTT was calculated for every beat and was recorded at 5 Hz to ensure that no values were missed.

Visual analysis of PTT signal. A close correlation between swings in pleural pressure and swings in PTT has previously been described (12). It should be noted that measurement of PTT is not continuous and is available only with each heartbeat. The measurement of respiratory effort with PTT therefore leads to digital undersampling of the esophageal waveform (Figure 1A). Consequently, a breath-by-breath error may be produced, but this becomes insignificant when the PTT pattern is visually examined over several consecutive breaths.

Figure 1B clearly shows that when Pes becomes increasingly negative during an obstructive apnea or hypopnea, there is a clear increase in the variations of PTT between inspiration and expiration.

Figure 1C shows that the determination of a central event from the PTT signal, as from the Pes signal, is made from the near disappearance (central apneas) or reduction (central hypopneas) of changes produced by respiratory effort.

A typical pattern of successive obstructive events is shown in Figure 2. Obstructive hypopneas interspersed with normal breathing result in clear changes in the amplitude of both the PTT and Pes signals.

View larger version (37K): [in this window] [in a new window]   Figure 2.   Sequence of obstructive and normal breathing events. There is a close relationship between the amplitude of the PTT signal and that of the Pes signal both during obstructive respiratory events (bold lines) and normal breathing (arrows).

Data Analysis

One hundred seventy-seven respiratory events in 13 unselected patients were chosen randomly for analysis. As a first step, respiratory events were scored independently by two observers using Pes and blinded to the PTT signal. The same result was obtained by the two observers for 167 of 177 of the events (94.35% agreement). There was disagreement between the two observers for 10 respiratory events as a result of a poor Pes tracing, and these 10 events were therefore discarded, leaving 167 events to be used for PTT assessment.

As a second step, the PTT signal was scored independently by the two observers, with blinding to the Pes, to classify the 167 episodes of sleep respiratory disturbance on the basis of PTT alone.

With this approach, we were able to evaluate the interobserver variability for both PTT and Pes signal analysis. Furthermore, it was possible to calculate the sensitivity, specificity, and positive and negative predictive value of PTT for the detection of central or obstructive respiratory events in relation to Pes.

Statistical methods. We evaluated the classification of respiratory events with the PTT signal in comparison with Pes by calculating the following parameters: (1) Positive predictive value (PPV); which indicated the probability that an event scored with PTT was a true event in comparison with scoring based on the esophageal balloon signal, and (2) negative predictive value (NPV); which reflects the probability that an event scored with PTT was a false event in comparison with Pes.

For each type of event, we evaluated the accuracy of detection by calculating sensitivity and specificity.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patient Population

This was a prospective study including 13 unselected male subjects. Patients had undergone polysomnography for investigation of symptoms suggestive of upper ariway resistance syndrome (UARS) or obstructive sleep apnea syndrome (OSAS). The mean ± SD range and median of patient age, body mass index (BMI), and sleep and respiratory parameters are shown in Table 1. Patients had mild to moderate sleep apnea syndrome. Using a classification by Pes for the entire night, we found that the subjects had an apnea-hypopnea index (AHI) of 25.1 ± 16.6, which was predominantly due to obstructive hypopneas. One patient demonstrated hypertension. This patient and another had a previous history of myocardial infarction, with coronary artery surgery in one case. The first subject was treated with a calcium channel blocking agent and the second with nitrates and diuretics. The other patients were free of any cardiovascular abnormality.

Classification of Respiratory Events

Interobserver agreement. Table 2 shows interscorer agreement for visual analysis of the Pes and PTT signals. Excellent agreement was found between the two observers for the two signals used for classification of respiratory events.

Accuracy of PTT for distinguishing central and obstructive events. The sensitivity, specificity, and positive and negative predictive value of PTT are summarized in Table 3. PTT was highly specific (93%) and sensitive (100%) in recognizing 40 central apneas. It was less sensitive (84.6%) in the detection of 26 central hypopneas as compared with apneas, but remained highly specific (98.6%). Misclassifications of respiratory episodes were due to baseline variations or artifacts in the PTT signal in 57% (i.e., eight of 14) cases. Discrepancies between Pes and PTT often occurred in REM sleep (six of 14 of the false classifications) (Figure 3).

View larger version (36K): [in this window] [in a new window]   Figure 3.   Sequence of respiratory events during REM sleep. In REM sleep the variations in respiratory effort may be erratic, with no steady pattern of increase. This phenomenon is caused by huge variations in respiratory drive coinciding with phasic REM sleep. In such a situation the typical pattern of an increase in oscillations in the PTT signal between inspiration and expiration can be more difficult to recognize.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We found that PTT measurement during sleep is capable of discriminating central and obstructive respiratory events. A correlation between Pes and PTT values has previously been reported (12, 13), but our study is the first to validate PTT as a clinical tool for classifying respiratory events. We found its specificity and sensitivity to be sufficiently convincing to suggest that PTT may become the reference noninvasive method for the classification of respiratory events during sleep. Moreover, the visual pattern observed during obstructive or central events was sufficiently typical to yield good interobserver agreement.

During obstructive events, measurement of Pes remains the reference method for demonstrating repetitive gradual increases in negative intrathoracic pressures that terminate in arousal. Pes is not widely used in sleep laboratories, in part because of its low acceptance by patients and in part because using an esophageal balloon to determine respiratory effort during sleep may itself affect the polysomnographic results. The use of local anesthesia before introduction of the probe, and the presence of the catheter in the UA, may change upper airway characteristics by impairing or modifying pharyngeal reflexes (16, 17). It is not known to what extent the placement of the Pes catheter in the UA interferes with UA dynamics. In using imaging techniques such as computed tomography (CT) or somnofluoroscopy (18) (i.e., without a catheter in the UA), to seek the initial site of pharyngeal collapse during sleep, the results are different from those obtained with manometric determination of the site of UA obstruction (6). Moreover, the use of esophageal catheters for monitoring respiratory effort may also disturb sleep. A recent study by Chervin and coworkers (19) demonstrated decrements in total sleep time, sleep efficiency, and the proportions of Stage 2 and REM sleep with this technique. However, the differences were of such small magnitude that their clinical significance is doubtful. Thus, Pes remains of limited use in clinical practice in the majority of sleep laboratories, owing to the rate of its acceptance by the patients and to unresolved questions about induced sleep disturbances and modifications in UA dynamics. Noninvasive alternatives to measure respiratory effort are therefore undoubtedly needed.

The detection of dyssynchrony between rib cage and abdominal components of respiration (20) is the more widely used means for detecting increases in respiratory effort. This detection technique depends on good movement sensors, and in the very obese subject it can be impossible to get adequate signals from any of the available transducers. The forced oscillation technique (FOT) seems to provide a promising signal. This technique consists of superimposing, on the patient's spontaneous breathing, a small pressure oscillation through a nasal mask. In their study, Badia and associates (21) found that in patients with severe sleep apnea-hypopnea syndrome (SAHS), the measurement of oscillatory impedance is a potential alternative or complementary noninvasive method for the diagnosis of respiratory events. Another interesting signal seems to be the signal detected with a nasal cannula/pressure transducer. This measurement is made with a standard nasal oxygen cannula placed in the nares and attached to a sensitive pressure transducer. The technique detects pressure fluctuations caused by inspiration and expiration that are proportional to flow. Condos and colleagues (7) have already pointed out that in patients under nasal continuous positive airway pressure (nCPAP), identification of a plateau on the inspiratory waveform correlates with an increased UA resistance. In a recent study, Hosselet and associates (10) showed that a characteristic flattened flow/time contour, as measured with a nasal cannula during spontaneous breathing, indicates an obstructive respiratory event (high UA resistance). In contrast, a rounded flow/time contour indicates a central-type respiratory episode (low UA resistance). The recognition of these typical waveforms therefore suggests that the use of a nasal cannula/ pressure transducer system may permit the classification of respiratory events as central or obstructive. However, the use of a nasal cannula for this purpose has never been validated in large scale studies. In particular, the extent to which partial mouth breathing, undetected with the nasal cannula, records artifactual flow limitation remains be established (11). Although several noninvasive techniques for assessment of respiratory effort are available, each presents some limitations. Thus, a combination of nasal prongs, for estimation of flow, and of PTT, for noninvasive measurement of respiratory effort, may provide the simplest way for accurately assessing respiration during sleep.

The present study was an attempt to evaluate PTT as a new method for measuring respiratory effort during sleep. PTT is inversely correlated with the systolic blood pressure, and permits a close estimation of respiratory effort. Pitson and colleagues (13) have demonstrated a correlation between swings in PTT and swings in the Pes signal. However, respiratory effort is not the only physiologic component inducing fluctuations in PTT. PTT is also influenced by fluctuations in autonomic tone and by the left ventricular isometric contraction time. Thus, pathologic conditions such as cardiac failure or drugs that modify autonomic tone or cardiac contractility could be confounding factors in analyzing variations in PTT. Moreover, changes in the compliance of the arterial wall due to aging or atheroma are probably associated with a modified relationship between PTT and Pes values. These questions will need close examination in further studies. Apart from these potential physiologic limitations to PTT measurement, we have identified misclassifications of respiratory events induced by artifacts on the PTT signal or occurring in REM sleep. The main causes of these artifacts were finger movements and/or displacement of the ECG electrodes during the night. Technical adjustments and/or algorithm modifications are needed to solve these problems. We also found that misclassification of respiratory events occurred mainly in REM sleep. During an obstructive respiratory episode in non-rapid-eye-movement (NREM) sleep, Pes regularly increases until event termination (Figure 1B). In REM sleep, the variations in respiratory drive may be erratic, making the accurate classification of hypopneas difficult. During phasic REM, this variable breathing pattern is more pronounced, and classifying hypopneas with PTT is particularly difficult (Figure 3). In such a situation, recognizing the typical pattern of increase in oscillations in PTT between inspiration and expiration, as occurs in an obstructive hypopnea, can be more difficult. However, this is not a problem specific to PTT. It occurs with any measure of respiratory effort, including Pes, and is the likely reason why there was a greater frequency of discrepancy between PTT and Pes during REM sleep in our study. Additionally, the visual analysis that we used remains subjective, and variations in observer experience may lead to differences in signal interpretation. The development of an algorithm for interpretation of the PTT signal that permits the automated classification of sleep-disordered breathing events would further enhance the clinical usefulness of this measure. This algorithm could also contain some simple mathematical tools that minimize the influence of artifacts (e.g., finger movements) and baseline variations.

Despite the drawbacks described here, we have demonstrated that visual analysis of the PTT signal is an excellent tool for distinguishing between central and obstructive events. For example a negative predictive value (NPV) of 95% for excluding central hypopnea is adequate for avoiding the false diagnosis of central apnea or hypopnea syndrome. Furthermore, although this question was not specifically addressed in our study, PTT also appears to be a potentially useful method for recognizing high-resistance respiratory events in the UARS. As recently reported by Stradling and coworkers (22), PTT may also permit evaluation of the mean level of respiratory effort throughout the night. Because PTT is capable of detecting both changes in respiratory effort and microarousals (23), an ambulatory device including PTT measurement and nasal prongs may provide a simple and robust way for monitoring sleep-disordered breathing.

In conclusion, our study has shown that PTT is a noninvasive method that accurately distinguishes between central and obstructive sleep respiratory events. Development of algorithms that improve interpretation of the PTT signal, particularly for the elimination of artifacts, are needed. Such a system would open the possibility of automated classification of respiratory events and, if clinically validated, could give PTT a central role in the investigation of sleep-disordered breathing.