INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

Techniques currently employed for investigating patients with suspected sleep-disordered breathing are inadequate. They are frequently expensive, labor intensive, and may not accurately define the respiratory events that are responsible for the sleep fragmentation and excessive daytime sleepiness that characterize sleep breathing disorders. In recent years it has become apparent that increases in upper airway (UA) resistance that do not cause complete pharyngeal airway collapse may be just as clinically important as full-blown apneas in terms of producing sleep fragmentation and daytime symptoms. This may result in obstructive hypopneas (1), where there is a significant reduction in airflow or in upper airway resistance episodes (UAREs) (2), where there is high inspiratory resistance without a significant fall in airflow or arterial oxygen saturation. These obstructive nonapneic episodes are much more difficult to detect and classify than apneas unless sensitive measures of respiratory effort and airflow are employed. If such measures are not employed it may be impossible to differentiate central from obstructive hypopneas, and UAREs may be missed. Thermistors are the most widely used tool for measuring airflow in sleep studies. However, as they give a more qualitative than quantitative measure, the scoring of hypopneas is often inadequate (3). Pneumotachography, and more recently nasal pressure measurements, tend to allow more valid and reliable interpretation of disordered sleep (4). This has led the American Academy of Sleep Medicine (AASM) Task Force (5) to propose new definitions of respiratory events and syndromes based on methodologies with an acceptable level of established validity. In the present study, as recommended by the AASM task force, we have used pneumotachography and esophageal pressure as the reference methods for measurement of airflow and respiratory effort, respectively.

Esophageal manometry is generally regarded as the gold standard for measuring inspiratory effort. However, this method is not without drawbacks. It is invasive, often uncomfortable for the patient, and may not be tolerated. In addition there is evidence that an esophageal catheter may modify the pharyngeal airway dynamics (6), and it has been suggested that its presence may itself impair the quality of sleep (7), though this is still disputed (8). For these reasons only a few centers routinely use esophageal pressure monitoring and respiratory effort is often assessed using less sensitive techniques. Pulse transit time (PTT) is a relatively new technique that shows promise in the sleep laboratory as a noninvasive surrogate marker of inspiratory effort. It measures the time interval that the pulse pressure wave takes to travel from the aortic valve to the periphery. For convenience, the electrocardiographic R wave is commonly used as the start point and the arrival of the pulse wave at the periphery is detected with a finger photoplethysmograph. PTT is inversely correlated with blood pressure and has been shown to predict changes in respiratory effort via the blood pressure fluctuations induced by negative pleural pressure swings (9). The PTT signal is not as "clean" as the signal derived from esophageal manometry, as only one PTT reading is available with each cardiac cycle and undersampling inevitably occurs (10), but it does not prevent increases (or decreases) in respiratory effort from being detected. PTT has the advantage over the esophageal pressure (Pes) signal of being able to pick up the surge in blood pressure associated with the microarousal (10, 11) that commonly occurs at the end of an obstructive upper airway event, which can assist in the visual analysis of the signal. Data from our center have shown that PTT has a high sensitivity and specificity at differentiating obstructive from central upper airway events (apneas and hypopneas) (12) when compared with Pes monitoring. However, very few data exist on the ability of PTT to detect nonapneic obstructive episodes, in particular upper airway resistance episodes.

The aim of this study was to compare two measures of respiratory effort, Pes monitoring and PTT when scoring obstructive nonapneic respiratory events (ONAREs) in association with a quantitative measure of flow.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

Nine consecutive male patients with suspected nonapneic sleep-disturbed breathing were prospectively evaluated. All complained of snoring and excessive daytime sleepiness, and all had previously undergone overnight oximetry monitoring (Biox-Ohmeda 3700; Ohmeda, Louisville, KY) thus making the presence of severe obstructive sleep apnea unlikely.

Polysomnography

Each patient underwent full overnight polysomnography in our sleep laboratory. Sleep and microarousals were staged manually with standard criteria (13, 14) using electroencephalography (electrodes C3/A2- Cz/O1), electrooculography, and submental electromyography. Thoracic and abdominal movements were recorded with strain gauges and airflow was measured with a pneumotachograph mounted on a full-face mask (Kontron Instruments, Saint Quentin, France) in seven patients and with a nasal pneumotachograph plus buccal thermistor in two patients who could not tolerate the full-face mask.

A quantitative measure of respiratory effort was available through the use of an esophageal pressure catheter (Compliance catheter Volgens EKGS; International Medical, Kutphen, The Netherlands). PTT measurements were made using the RM50 recorder (DeVilbiss, Parcay-Meslay, France), which calculates the time interval between the electrocardiographic R wave and the point on the finger pulse waveform (detected by photoplethysmography) that is 50% the height of the maximum value. Only one PTT value is available with each cardiac cycle but the RM50 samples at 5 Hz to ensure that no values are missed.

Visual Analysis of Respiratory Signal and ONAREs Scoring

A total of 130 5-min periods of polysomnography trace were randomly chosen to be analyzed visually by two different observers skilled at reading sleep studies. Each 5-min trace was printed out twice: one with Pes and the other with PTT as the measure for respiratory effort. In addition to Pes or PTT each trace also had airflow, thoracic and abdominal movement, and sleep stage analyzed in 20-s epochs. Each observer scored all the traces independently of the other observers. Those with the Pes signal were scored first and, after an interval of 4 wk, those with the PTT signal were scored. The 5-min tracings were presented to the observer in a different order to ensure that the interpretation was not influenced by the previous analysis. This permitted the respiratory events scored with PTT to be compared with those scored with Pes, and in addition interobserver variability could be assessed.

Apneas were scored but were not included in this analysis as it was the ability of Pes and PTT to detect nonapneic events that was the aim of the study. An hypopnea was defined as a decrease in, but not complete cessation of, oronasal airflow of more than 30% lasting at least 10 s and was classified as central or obstructive on the basis of the respiratory effort signal (Pes or PTT). A UARE was defined as an increase in respiratory effort in a crescendo pattern, in the absence of a greater than 30% decrease in airflow.

Although a pneumotachograph was used in all of our patients, making a clear distinction between obstructive hypopneas and UAREs on the basis of subtle airflow differences was not always easy. For this reason the final analysis combined both hypopneas and UAREs to provide a single score of obstructive nonapneic events (i.e., ONAREs) for the purpose of comparing Pes with PTT and evaluating the interobserver variability.

Comparison of Pes with PTT in Scoring 20 Obstructive Apneas, 20 "Classic" Hypopneas, and 20 "Classic" Upper Airway Resistance Episodes (UAREs)

This analysis was done to obtain a reference value of PTT sensitivity and interobserver agreement in frank, clearly recognizable respiratory events randomly selected. "Classic" hypopneas were defined as follows: a clear and discernible reduction in flow (> 50%) that ended with an arousal and with a fall in Sa_O2 of at least 3%. "Classic" UAREs were accepted when a flow limitation without flow reduction occurred concurrently with a crescendo in Pes, ended with an arousal, and was then followed by Pes going back to resting levels.

Data and Statistical Analysis

The interobserver variability of Pes and PTT for detecting respiratory events was calculated. Taking only those respiratory events for which the two observers agreed, the sensitivity of PTT to detect obstructive respiratory events using Pes as the gold standard was calculated (True Positive/[True Positive + False Negative]). The positive predictive value was also calculated (True Positive/[True Positive + False Positive]). Although there were periods with no significant respiratory events or with "stable upper airway resistance," it was not possible to provide true negatives and therefore a calculation of specificity or negative predictive value could not be obtained.

A further post hoc analysis was undertaken of those respiratory events for which there was disagreement between observer 1 and 2 in scoring ONAREs using Pes or PTT. This was to determine if (1) these particular respiratory events were different in terms of flow reduction or increase in respiratory effort. The rate of increase in respiratory effort was assessed, respectively, by the slope of Pes and PTT expressed in arbitrary units (AU) (see Figures 2 and 3). (2) If there was any evidence of either cortical microarousal (ASDA) on the electroencephalographic (EEG) recordings or autonomic arousal on the PTT. An autonomic arousal was defined as a significant dip in the PTT baseline corresponding to the surge in blood pressure provoked by the transient sympathetic activation that usually occurs at the end of obstructive respiratory events (15).

View larger version (23K): [in this window] [in a new window]   Figure 2.   Examples of flow limitation and upper airway resistance episodes. A 3-min polysomnography trace is demonstrated with abdominal (ABD) and thoracic (THO) inductance plethysmography, airflow (FLOW) measured with a pneumotachograph, pulse transit time (PTT), and esophageal pressure monitoring (PES). All the events that have been scored in this 3-min epoch are not shown. Our goal is to focus on typical problems encountered in scoring ONAREs. Thus, in this trace only two ONAREs are shown (both are examples of an upper airway resistance episode). On the FLOW trace there is no clear reduction of airflow during the ONAREs but there is characteristic flattening of the waveform indicating "flow limitation." This "flow limitation" gives a clue to the observer that there is significant upper airway resistance occurring. With the first episode, there is an obvious crescendo increase in inspiratory effort clearly visible with both the PTT and the PES traces. With the second ONARE, the crescendo increase in inspiratory effort is less obvious and could easily be missed using the PES trace alone. However, there is a dip in the PTT signal indicating an autonomic arousal (see arrows), and this, coupled with the presence of "flow limitation," allows an ONARE to be scored. Sleep stages were available for the expert's visual scoring but are not represented in the figure.

View larger version (32K): [in this window] [in a new window]   Figure 3.   (A) Undersampling error of a PTT signal. Each black dot represents a cardiac cycle. The thick and thin lines represent Pes and PTT signals, respectively. Two respiratory cycles are shown. The nadir of the first cycle happens to coincide with the arterial pulse wave and therefore the PTT amplitude is an accurate representation of the Pes swing. The second respiratory cycle does not coincide with a pulse wave and therefore the PTT amplitude underestimates the Pes swing (adapted from Pitson and coworkers [9]). (B) An example of an ONARE scored with Pes but missed with PTT All the events that have been scored in this 3-min epoch are not shown. Our goal is to focus on typical problems encountered in scoring ONAREs. Thus, in this trace only one ONARE is shown (scored with Pes but missed with PTT). In the example shown here a classic pattern of an upper airway resistance episode is detected with Pes but, as the crescendo increase in respiratory effort is subtle, the PTT signal is not sensitive enough to pick it up. In addition there is no significant change in airflow and there is no visible autonomic arousal (dip in PTT baseline), and thus no event is scored with the PTT signal. Sleep stages were available for the expert's visual scoring but are not represented in the figure.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

Patient Population

This was a prospective study including nine unselected male subjects. The mean ± SD, median of patient age, body mass index, and sleep and respiratory parameters are shown in Table 1. Patients had mild to moderate sleep apnea syndrome. Using esophageal pressure classification for the whole night patients had an apnea + hypopnea index of 25.1 ± 10.8/h of sleep, which comprised predominantly obstructive hypopneas. The ONARE index (obstructive hypopnea index + upper airway resistance [UAR] index) was 25.2 ± 10.5. None of the patients had significant cardiac disease other than controlled hypertension.

Interobserver Variability of Pes and PTT Interpretation

As already mentioned, obstructive hypopneas and UAREs were grouped together and the scores for each observer using each method are listed in Table 2. Observer 1 detected a total of 330 ONAREs with Pes (of which 46 were not detected with PTT) and 305 with PTT (of which 20 were not detected with Pes). Observer 2, on the other hand, identified significantly less nonapneic events: only 249 with Pes (of which 49 were not picked up with PTT) and 231 with PTT (of which 20 failed to be identified with Pes). This represents significant interobserver variation irrespective of which measure of inspiratory effort was used (29.7% and 37.1% for Pes and PTT, respectively).

Examples of ONAREs scored using airflow, Pes, or PTT are shown in Figures 1, 2 and 3.

View larger version (40K): [in this window] [in a new window]   Figure 1.   Examples of hypopneas. A 3-min polysomnography trace is demonstrated with abdominal (ABD) and thoracic (THO) inductance plethysmography, airflow (FLOW) measured with a pneumotachograph, pulse transit time (PTT), and esophageal pressure monitoring (PES). In this trace five ONAREs are shown (all are examples of an obstructive hypopnea). There is a reduction in (but not complete cessation of) airflow during each episode associated with an increase in inspiratory effort, which is clearly demonstrated on both the PTT and PES traces. The arrows coincide with the occurrence of a cortical arousal on the EEG traces. Note that slow waves that are not recognizable at this scale may precede the arousal itself. Sleep stages were available for the expert's visual scoring but are not represented in the figure.

Sensitivity of PTT for the Detection of Obstructive Nonapneic Events

When the findings of the two observers were compared, there was agreement on 239 (70.3%) ONAREs when scored with Pes and 207 (62.9%) when scored with PTT. When taking only those events for which there was agreement between the two observers, 52 events detected with Pes which were not picked up with PTT, and 20 events scored with PTT were not detected with Pes. Among the 52 ONAREs identified by Pes and unrecognized by PTT, 41 were associated with both an autonomic and a cortical arousal. Taking Pes as the gold standard measurement for respiratory effort, PTT had a sensitivity of 79.9% and a positive predictive value of 91.2%.

Sensitivity and Interobserver Agreement of PTT for the Detection of Obstructive Apneas, "Classic" Hypopneas, and "Classic" UAREs

In these so-called "frank respiratory events," when the findings of the two observers were compared, there was an agreement on 90% of obstructive apneas, 95% of hypopneas, and 100% of UAREs when scored with PTT. Thus, in these events, the rate of interobserver agreement and the PTT sensitivity was higher compared with more subtle respiratory events. The 90% rate for apneas was due to two apneas that were wrongly scored as hypopneas by one of the observers in each case. Moreover, the slopes of increase in respiratory effort were significantly higher when comparing frank with subtle respiratory events (Table 4).

Analysis of the Reasons for Disagreement between Observers 1 and 2 in Scoring ONAREs Using Pes or PTT (Post Hoc Analysis)

The events for which there was agreement among observers tended to have a higher reduction in airflow (51 versus 30.3%, p < 0.001), a larger increase in respiratory effort as assessed by PTT (slope of PTT: 23.1 versus 16.4 arbitrary units [AU], p < 0.01), and a greater incidence in the occurrence of autonomic microarousals on PTT (microarousals in 90 versus 45% of the events, p < 0.006).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

Our study clearly illustrates that the scoring of nonapneic obstructive respiratory events leads to significant interobserver variations, particularly when the reduction of airflow is small (mean reduction around 30%) and the increase in respiratory effort is limited. This was particularly the case when PTT was used (interobserver variations: 37.1%) but also occurred to a significant degree when the gold standard measure of respiratory effort, esophageal pressure monitoring, was employed (interobserver variability: 29.7%). PTT provides only a semiquantitative estimate of respiratory effort but has the advantage of being able to detect the autonomic arousal that is commonly associated with the end of an upper airway obstructive event and this added piece of information may allow PTT to score events that have been missed with Pes alone (16). On the other hand subtle increases in respiratory effort are better seen using Pes, and this may explain why 52 ONAREs missed with PTT were identified with Pes. Whatever the tool used to estimate or measure respiratory effort, a quantitative measurement of airflow and/or the recognition of a flow limitation by a pneumotachograph or a nasal cannula are critical for scoring subtle ONAREs.

There is growing evidence that nonapneic obstructive respiratory events are as important as full-blown apneas in causing hypersomnolence (17) and psychological morbidity, and that they may also contribute to the increased cardiovascular risk associated with sleep-disordered breathing (18, 19). Large epidemiological studies may partly answer these questions in the near future (20), but progress is hampered both by the uncertainty of which types of respiratory events are pathological and by the inadequacies of currently used technology employed to detect respiratory events (3, 4, 21, 22). Obstructive nonapneic respiratory events can be characterized by three different polygraphic features: (1) a variable reduction of flow (from inspiratory flow limitation to frank hypopnea), (2) an increase in respiratory effort, and (3) the occurrence of a microarousal ending the respiratory event. We have demonstrated a high level of interobserver variability both with Pes (29.7%) and PTT (37.1%). This was a surprising finding as both observers work in the same sleep laboratory, used the same criteria to score respiratory events, and had a strong agreement when differentiating central and obstructive apneas and hypopneas (12). The two experienced observers were significantly less in agreement for the respiratory events in which the reduction of flow was very subtle and the slope of increase in respiratory effort was difficult to characterize (see Figures 1-3 and Table 4). In frank respiratory events with a more discernible slope of increase in respiratory effort, we found a higher rate of interobserver agreement and a better sensitivity for PTT compared with subtle respiratory events. Thus, a high interobserver variability exists only for subtle respiratory events. In this situation, also, PTT works as well as Pes.

When nonapneic respiratory events are scored with only one channel used in isolation, it is usual to find a relatively low concordance between different scorers (23). When a corroborative criterion is used (i.e., a microarousal or a desaturation) the interscorer reliability improves. This is particularly true for desaturation, which is easy to recognize and is associated with more pronounced respiratory changes. In our study, oximetry was not made available to the observers on the traces and although none of the patients experienced marked desaturations, it might have made interpretation easier in some cases. In the Whitney study (23), the added information due to the occurrence of a microarousal for scoring a hypopnea appeared negligible probably due to the difficulties and discrepancies in visual scoring of EEG microarousals (24). Pes gives information on respiratory effort, whereas the PTT signal is novel in that it allows both measurement of respiratory effort and detection of microarousals. Moreover, the PTT pattern of autonomic arousal is undoubtedly much more simple to detect visually than that of cortical arousal on the EEG. This explains why in our study PTT was able to score events that were missed using Pes (Figure 2). Nevertheless, even with this attribute, the interscorer agreement using PTT was relatively low. Post hoc analysis demonstrated that this occurred mainly with respiratory events associated with little reduction in flow and/or increase in respiratory effort (Table 3), whereas the events for which the observers agreed were associated with higher flow reduction, a larger increase in respiratory effort as assessed by PTT, and a higher incidence of autonomic microarousals on PTT. The recognition of an inspiratory flow limitation with or without reduction in the amplitude of airflow signal was also a key factor in recognizing respiratory events, although its exact contribution was difficult to quantify (Figure 2). The 37.1% of interscorer variability that occurred when using PTT as the reference signal was probably due to individual differences in interpreting the three physiological signals (airflow, respiratory effort, and autonomic arousals) together.

During PTT measurement, a pulse wave may not coincide for each respiratory cycle with the nadir of Pes and this produces an undersampling phenomenon (Figure 3A). In addition PTT, unlike Pes, provides only a semiquantitative measure of respiratory effort at a given point of time. This may explain why the more subtle increases in respiratory effort, which may be associated with nonapneic respiratory events, can be more accurately described with Pes (Figure 3B). In our study Pes identified 52 ONAREs unrecognized by PTT of which 41 were associated with both an autonomic and a cortical arousal. Although PTT picked up these microarousals, they were not scored as respiratory microarousals as PTT underestimated the variations in respiratory effort (see above) and the small decrease in airflow was not deemed significant (Figure 3B).

Our study suggests that accurate measurement of airflow, respiratory effort, and microarousal is needed to accurately score subtle nonapneic respiratory events. We have demonstrated that both PTT and Pes, when used as measures of respiratory effort, are effective at detecting ONAREs. Although Pes is highly sensitive at detecting changes in respiratory effort, and as such remains the gold standard, it does not provide any information on the presence of microarousal and it has the added disadvantage of being an invasive and uncomfortable technique. Pes is therefore not necessarily the best measure for respiratory effort in an investigative package for scoring obstructive nonapneic events in the clinical setting. PTT, on the other hand, is noninvasive and provides not only an estimation of respiratory effort but also has the advantage of being able to detect the autonomic arousal that is usually associated with the ending of obstructive events. For the detection of ONAREs a quantitative measurement of flow is crucial. The wide use of thermistors during polysomnographic recordings that cannot detect subtle changes of respiratory flow can lead to a considerable underestimation, at least for hypopneas and even more for upper airway resistance episodes (25). This technical limitation associated with the high intraobserver variability in scoring that we have described, even using reference methods, undoubtedly leads to major underestimation of nonapneic respiratory events in routine clinical practice. This has implications for both patients' diagnosis and therapy and epidemiological studies.

	CONCLUSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

Pulse transit time can be considered as an alternative to Pes as a means of measuring changes in inspiratory effort when scoring ONAREs. However, both PTT and Pes are not without their drawbacks. This study has demonstrated that significant interobserver variability can occur even when well-defined scoring criteria are employed. The ability to accurately measure flow limitation, respiratory effort, and microarousals is crucial for the correct detection of nonapneic obstructive events. Refinements in the definition and in the recording of these respiratory events are needed.