INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Inspiratory flow limitation during sleep is defined by a decreasing intrathoracic pressure without a corresponding increase in airway flow rate. This alinearity in the pressure:flow relationship during inspiration is commonly caused by narrowing of a hypotonic upper airway in response to the negative intrathoracic pressure developed during inspiration (1). There are several potential consequences of inspiratory airflow limitation during sleep (see also DISCUSSION). If the increased airway resistance achieved is sufficiently high, the tidal volume will fall (2, 3). In turn, if the high upper airway resistance and reduced tidal volume persist, then sleep-disordered breathing events will occur in the form of hypopneas (over a few breaths) or even prolonged alveolar hypoventilation might ensue with associated CO₂ retention and arterial hypoxemia and disrupted sleep state. Furthermore, if sufficient inspiratory muscle effort and negative intrathoracic pressure are repeatedly generated against a high upper airway resistance, then transient arousals from sleep may occur, sometimes leading to daytime hypersomnolence, i.e., so-called increased upper airway resistance syndrome (4). Continuous positive airway pressure (CPAP) titration studies have shown further, that flow limitation may curtail periods of deep sleep, even when the degree of flow limitation was not sufficient to be associated with transient cortical electroencephalogram (EEG) arousals (7, 8). Finally, the more negative intrathoracic pressure associated with inspiratory flow limitation has been shown to have secondary effects on stroke volume and cardiac output, especially in certain kinds of patients with already compromised left ventricular function (9, 10). So, the presence of flow limitation may herald ensuing sleep-disordered breathing events, and even in the absence of sleep-disordered breathing events it may prevent the occurrence of optimal sleep architecture or even cause transient arousals and cardiovascular sequelae. In turn, these nocturnal perturbations often predispose to chronic daytime pathophysiology.

How this flow limitation might be detected or quantified in sleeping humans has been the subject of several recent investigations (7, 11). Esophageal pressure measurements have been applied in an attempt to diagnose high upper airway resistance (4, 5) and the time profile of inspiratory flow rate has been used to determine "optimal" CPAP titration (11).

Our aim in the present study was to evaluate noninvasive means of assessing flow limitation using measurements that can be tolerated without disruption of sleep. To this end, we derived an index of flow limitation from the flow rate versus time profile using three measurement techniques, namely flow rate using a face mask and pneumotachograph, pressure at the nares (Pn) measured with a nasal cannula, and differentiated sum Respitrace (dRIP), i.e., inductance plethysmography. Each of these flow rate versus time indices was compared with the gold standard measurement of pharyngeal pressure versus flow rate, in asymptomatic sleeping subjects whose breaths showed a marked variation in the magnitude and type of flow limitation.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Seven habitual snorers (2 females, 5 males, 19 to 25 yr old) with no history of respiratory or cardiovascular disease were studied during nonrapid eye movement (NREM) sleep. The study was approved by the Human Subjects Committee at the University of Wisconsin Center for Health Sciences-Madison. Informed consent was obtained from all subjects.

Protocol

Pharyngeal pressure was measured using a pressure-tipped catheter (model TC-500XG; Millar). The nasopharynx was sprayed with xylocaine (10%), the catheter was threaded through a nose mask, passed into one nostril, and placed 2 cm below the base of the tongue, as judged by visualization. Tincture of benzoin was applied to the nose, and tape applied to secure the catheter to the nose. A pressure transducer (Validyne ± 56 cm H₂O) measured Pn changes via a dual pronged nasal cannula (BCI International) with an opening diameter of 1.5 mm. The nasal mask (Respironics CPAP mask) was sealed with theatrical glue and putty, and the mouth taped closed. Inspiratory and expiratory tidal volumes were compared to determine leaks. A heated pneumotachograph (Hans Rudolph Model 5719, 0-100 L/min) was connected to the mask. The pneumotach was connected to a differential pressure transducer (Validyne ± 0.5 cm H₂O) to measure bidirectional air flow rate. Subjects were also instrumented with an ear oximetry probe (Ohmeda Biox 3740). The subjects were instrumented with electrodes to measure central and occipital EEG (C3/A2, C4/A1, O1/ A2, O2/A1), chin electromyogram (EMG) and eye electro-oculographic activity (right and left EOG), and electrocardiogram (ECG). The subjects were also fitted with inductive belts around the rib cage and abdomen. The inductive belts were attached to an oscillator, which measured the sum of the rib cage and abdominal excursions (Respitrace or respiratory inductance plethysmography [RIP]). The sum RIP was calibrated using an isovolume technique, by instructing the subject, and adjusting the relative gains of the abdomen and rib cage, such that the net sum RIP of zero was obtained. The subjects were sleep-deprived the night before the study. Subjects were studied in NREM sleep. A typical polygraph record along with pressure versus flow rate plots for individual breaths are shown in Figure 1.

View larger version (38K): [in this window] [in a new window]   Figure 1.   The polysomnograph demonstrates varying degrees of flow limitation in a subject over a 40-s period during NREM sleep. The breaths are numbered on the pneumotach flow trace. Pharyngeal pressure, pressure at the nares (Pn), the differentiated Respitrace, and EEG versus time are shown. Pharyngeal pressure versus flow rate plots are shown for each numbered breath. In these breaths, there are different levels of flow limitation. For example, breath 6 has a fairly linear relationship between pressure and flow (i.e., no flow limitation), whereas breath 11 has a constant flow rate with a decreasing pressure (i.e., severe flow limitation).

Signal Analysis

All signals were sampled at 128 Hz digitally. Every three consecutive samples were averaged and this averaged signal was then resampled at one-third of the original sampling rate (42.67 Hz). This technique is called down sampling. The signals were then digitally filtered with a Hamming filter. A breath was defined by first identifying the peaks and valleys of each flow rate measurement, i.e., pneumotach flow, differentiated RIP, and Pn. Then the point of maximum rate of change for each flow rate signal, within the threshold range for zero flow, was defined as the initiation of inspiration or expiration.

Invasive Method for Determination of Flow LimitationGold Standard

To measure the extent of flow limitation using invasive techniques, pharyngeal pressure and flow rate were analyzed via computer from the beginning of inspiration to the nadir of the pharyngeal pressure signal. The pressure:flow rate relationships were divided into four categories on the basis of the severity of flow limitation. A nonflow limited inspiration (Level 1) was defined as a linear pressure:flow rate relationship (Figure 2a). Mild flow limitation (Level 2) was defined as a mildly alinear pressure:flow rate relationship (Figure 2b). A more severe level (Level 3) was defined as a period of constant flow rate without pressure dependence, the most severe (Level 4) being a period of decreasing flow rate with negative pressure dependence (Figure 2c). As detailed below for Levels 1 and 2 we were able to quantify the pressure:flow rate relationship using curve-fitting techniques and a single model equation; however, with the more severe levels of flow limitation the flow versus pressure relationship frequently showed oscillations throughout inspiration that we could not curve-fit mathematically.

View larger version (25K): [in this window] [in a new window]   View larger version (20K): [in this window] [in a new window]   Figure 2.   (a) The Level 1 pressure:flow rate relationships for nonflow-limited inspiratory cycles. (b) The Level 2 pressure:flow rate relationships for mildly flow-limited inspiratory cycles. (c) Level 3- flow rate with no pressure dependence and oscillating flow rate; Level 4-flow rate with negative pressure dependence and oscillating flow rate.

The breaths were first analyzed for the more severe categories of flow limitation, Levels 3 and 4. For each inspiratory pressure:flow rate relationship, the slope between every sample point was calculated. The breath was first analyzed to determine if it was a Level 4 breath, which required a falling flow rate for at least 1 cm H₂O pharyngeal pressure reduction (Figure 2c). Flow rates that oscillated between falling and constant were also taken into account (Figure 2c). The lower limit required for the pressure:flow rate slope was 0.01 cm H₂O/L/s. If the flow rate oscillated between a falling and constant flow over at least 2 cm H₂O reduction of pressure, then the breath was also categorized as Level 4. The average percentage of the inspiration analyzed that had a falling flow rate was 9.8 ± 4.8%. The average reduction in pharyngeal pressure with falling flow rate was 1.8 ± 1.3 cm H₂O. If a breath met the Level 4 requirement, the pressure:flow rate analysis was completed and the gold standard was defined, otherwise analysis continued.

If the breath did not meet the requirement for Level 4, the breath was then analyzed to determine if it should be placed in the Level 3 category. The breath was first analyzed to determine if the flow rates were constant for a period of pressure reduction. If flow was constant for at least 1 cm H₂O pressure reduction, the breath was placed in Level 3 (Figure 2c). The limits for slopes were ± 0.01 cm H₂O/L/s. Breaths that oscillated between constant and negative flow rates for at least 0.75 cm H₂O of pressure reduction, but less than 2 cm H₂O were also placed in Level 3. The 0.75 cm H₂O of pressure reduction threshold was chosen because the level of flow limitation was concluded to be more severe when both negative and constant flow rates occurred within a breath than a constant flow rate alone. If the breath met this requirement, then the pressure:flow rate analysis was completed and the gold standard was designated Level 3, otherwise analysis continued. Level 3 breaths had an average pharyngeal pressure with falling or constant flow rates of 0.93 ± 1.34 cm H₂O, which on average was 10.3 ± 7.1% of the inspiratory time.

If the breath was not found to be a Level 3 or 4 breath, then the least-squares error method was used as a curve-fitting technique. This produces an exponential equation of the form V = a · Pph^b where V is flow rate, a is a constant which in combination with Pph defines the magnitude of the flow rate, and b defines the shape of the curve by indicating the amount of curvature in its shape. This equation was chosen for its ability to reflect a "bend" in the pressure:flow rate relationship between slopes of 0 and 1. If b  0.8, the pressure:flow rate relationship was categorized as linear (Level 1) (Figure 3a), and if 0.8 > b > 0, then the pressure:flow rate relationship was categorized as mildly alinear (Level 2). (Figure 3b). The selection of b = 0.8 as the qualifying cutoff point for mild flow limitation was made subjectively based on the visual analysis of many breaths. For this analysis, breaths were compared with the fitted curve by successively using values for b from 0.5 to 1 in 0.05 increments.

View larger version (16K): [in this window] [in a new window]   Figure 3.   Curve-fitting is used to define (a) linear and (b) nonlinear pressure:flow rate relationships (open circles indicate actual data points, solid line indicates the line curve has been fitted). The pharyngeal pressure was first normalized to 2 cm H2O, owing to the nature of the curve-fitting equation which applies the constraint of V = a when Pph = 1 cm H2O.

Determination of Flow Limitation by Noninvasive Methods

To evaluate flow limitation from noninvasive measurements, the time profile of pneumotach flow rate, differentiated sum Respitrace, and nasal pressure during inspiration were analyzed. An individual nonflow limited breath template was signal averaged from a series of awake breaths using peak flow rate as a fiduciary point. The template breaths were close to sinusoidal for all subjects. Analysis was done on both the template and the breath to be evaluated for flow limitation using the middle portion of the breath, taken between 25% to 75% of the total inspiratory time (see DISCUSSION for rationale of choice).

The middle 50% of the template was matched in time as well as amplitude to the middle 50% of the breath being evaluated for flow limitation. The area difference between the template and breath being analyzed was calculated and represented as the percentage of the total area under the breath being analyzed, i.e., "area index." As the breath deviated further from the template with increasing levels of flow limitation, the greater was the area index (see Figure 4).

View larger version (17K): [in this window] [in a new window]   Figure 4.   A mildly flow-limited breath with a nonlinear relationship between pressure and flow rate will have a small area index (a), but the area index will be large for a breath with negative flow dependence (b).

Data Analysis

Data from the seven subjects were analyzed, with the pressure:flow rate relationship evaluation used as the gold standard. The breaths defined by the differentiated sum Respitrace and the pressure transducer signal at the nares were matched in time with the breaths from the flow rate signal. The gold standards were then used to evaluate the noninvasive area index, using sensitivity and specificity. Three types of sensitivity/specificity analysis for evaluating the noninvasive measurements were completed (12). First, we determined the area index at which sensitivity and specificity were equal in terms of the technique's ability to distinguish between several different severities of flow limitation. Choosing this point of equal sensitivity and specificity minimized the error in classification of flow limitation by requiring that the same percentage of breaths were correctly classified as flow limited (true-positive) as were correctly classified as nonflow limited (true-negative). Second, we determined the effects of changing values for area index on sensitivity and specificity for each measurement, in terms of their ability to distinguish no or mild from moderate and severe flow limitation. Third, we compared the three noninvasive tests by plotting receiver operating characteristic (ROC) curves which contrast the true-positive rate versus the false-positive rate over the entire continuum of area indices for each measurement.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Pressure:Flow Rate Relationship

For each gold standard group, the median and interquartile range of the tidal volume, resistance at peak pressure and peak flow rate, time of inspiration (TI), peak flow rate, and peak pressure is listed in Table 1 along with the number of breaths in each category. With increasing severity of inspiratory flow limitation, mean peak respiratory flow rate was reduced and the median of pharyngeal resistance at peak flow rate and especially at the nadir of pharyngeal pressure also increased. Also, note the large intersubject variability in airway resistance within each of the flow limitation categories. For example, across the breaths categorized as Level 1 with "linear" pressure:flow relationship, average resistance at peak inspiratory flow rate varied from 5 cm H₂O/L/s in one subject to 24 cm H₂O/L/s in another subject. In other words, the latter subject had a relatively high but unchanging resistance to air flow throughout inspiration (see DISCUSSION). Mean tidal volume changed very little across the categories of flow resistance, although TI was prolonged between Levels 1 and 2 versus Levels 3 and 4 of flow limitation.

Noninvasive Measures of Flow Limitation

Table 2 lists the number of breaths categorized by the severity of flow resistance, as judged by the pharyngeal pressure:flow rate relationships, i.e., "gold standard." The 26 fewer RIP than pneumotach flow breaths analyzed reflects irregularities in some of the RIP signals. The lesser number of breaths analyzed for the nasal pressure signal reflects the reduced number of subjects studied with this method.

Pneumotach Flow Rate

The sensitivity and specificity at the point of equality was identified for the separation between levels of air flow limitation as defined by the pressure:flow rate gold standard in Table 3 for pneumotach flow rate. As an example, for the separation between gold standard Level 1 (linear) and Levels 2, 3, and 4, the sensitivity and specificity at the point of equality at an area index of 30.7% is 72%. With each increasing level of flow limitation, the area index cut off became larger as we expected. The area index was effective in separating gold standard Levels 3 and 4 from Levels 1 and 2, as well as Level 4 from all others, with a sensitivity and specificity of 82% and 84%, respectively. However, it did not perform well in separating Level 1 from other groups, with a sensitivity and specificity of 72%. This was due mostly to the inability to separate Level 1 from Level 2. For example, 210 breaths out of 382 that were identified by the gold standard as Level 2 were categorized by the area index as Level 1. 

Pressure at the Nares

The Pn was analyzed using the noninvasive area index and compared against the gold standard grouping. Table 4 displays the sensitivity and specificity at the point of equality for the separation of gold standard groups. The sensitivity and specificity for all levels was lower than that found using the pneumotach flow rate (Table 3). Again we observed a low sensitivity and specificity for the separation between Level 1 and other levels. The area index was effective at separating more severe flow limitation (Levels 3 and 4) from mild and no flow limitation (Levels 1 and 2).

Differentiated Sum Respitrace

The dRIP was analyzed using the noninvasive area index and compared against the gold standard grouping. Table 5 displays the sensitivity and specificity at the point of equality for the separation of gold standard groups. The sensitivity and specificity for all groups was lower than that found using the pneumotach flow rate (Table 3) and Pn (Table 4). We observed a lower sensitivity and specificity (71%) than with either pneumotach flow rate (82%) or Pn (76%) for the separation between mild (Levels 1 and 2) and severe (Levels 3 and 4) flow limitation.

Sensitivity versus Specificity

The effects of a changing area index for each of the noninvasive techniques over the entire continuum of sensitivity and specificity are shown in Figure 5. These data were based on the capability for distinguishing flow limitation Levels 1 and 2 from 3 and 4 as defined by the gold standard. For each of the noninvasive techniques, we observed that increasing the difference in area between the template breath and the unknown breath (i.e., area index) resulted in a progressive decline in sensitivity and increase in specificity of detection. These trends were quite similar among the three methods, although the cross-over point (i.e., of equal sensitivity and specificity) occurred at the highest values for pneumotach flow followed by nasal pressure and dRIP.

View larger version (20K): [in this window] [in a new window]   Figure 5.   Changes in sensitivity (closed circles) and specificity (open circles) in distinguishing flow-limitation Levels 1 and 2 from 3 and 4 of the noninvasive measurements with increasing values for the area index of flow limitation.

The ROC curves shown in Figure 6 contrast the three noninvasive methods with respect to true-positive versus false-positive rates of distinguishing gold standard flow limitation Levels 1 and 2 versus 3 and 4. In terms of accuracy, the pneumotach flow method was superior followed by nasal pressure and dRIP. For example, at a high sensitivity of 85%, the pneumotach flow method yielded 15 to 20% false-positives, the Pn technique 35% false-positive, and the dRIP over 50% false-positive. Similarly, for a low false-positive rate of about 15%, the pneumotach flow method showed the highest sensitivity of 75 to 80%, followed by 65% for Pn and 50% for dRIP.

View larger version (22K): [in this window] [in a new window]   Figure 6.   ROC curves contrasting true-positive versus false-positive rates for distinguishing between flow limitation Levels 1 and 2 from 3 and 4 for each of the noninvasive methods for detecting flow limitation.

Additional Approaches

The sensitivity and specificity for each noninvasive measurement of flow limitation compared with the pressure:flow rate gold standard were lower than desired, especially in separating the milder levels of flow limitation. In an attempt to raise sensitivity, we tried many other noninvasive indices. Varying percentages of each analyzed breath were evaluated with the same techniques used above. Other indices included time to peak flow rate over total inspiratory time, the average slope of varying percentages of the breath, and the standard deviation from a line fit for a percentage of the breath. We also combined indices together. However, the area index from the middle 50% of the breath had the highest level of sensitivity and specificity when compared with the gold standard method.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We have addressed the problem of detection and quantification of air flow limitation during sleep. First, we computerized the shape analysis of the continuous pharyngeal pressure:flow rate relationship during inspiration. We defined the shape of the pressure:flow relationship and categorized the degree of flow limitation. We offer this analysis of the pressure:flow rate relationship as a more valid approach than the traditional method of single point resistance calculations. Second, we compared the area between a nonflow limited template breath to the unknown breath in order to assess the degree of flow limitation, for three noninvasive and/or indirect measurements of air flow versus time within the breath. We found that two of the three methods were capable of distinguishing the most severe flow limitation from no flow limitation at a sensitivity/specificity of greater than 80%, and a mild to moderate flow limitation from severe flow limitation at a sensitivity/ specificity of greater than 75%. All of the noninvasive methods distinguished no flow limitation from mild flow limitation with a sensitivity/specificity of only 62 to 72%. We propose the use of the pneumotach flow or nasal pressure noninvasive measurement techniques used in combination with an area breath analysis approach as reasonably reliable measures of most types of moderate/severe flow limitation encountered in sleeping subjects. On the other hand, high levels of "fixed" flow resistance causing little or no alteration in the shape of the flow:time relationship will not be detected by this analysis.

Limitations of Noninvasive Techniques for Detection of Flow Limitation

All of the noninvasive methods we tested were not sufficiently sensitive in detecting relatively mild from moderate levels of flow limitation. This insensitivity may be accounted for in part simply by inaccuracies in flow measurement by nasal pressure and especially the RIP techniques. More importantly the reduced sensitivity and specificity for all noninvasive techniques is accounted for primarily by hysteresis in the pressure:flow rate relationship (see Figure 7). Because flow limitation was defined as a decreasing intrathoracic pressure without a corresponding change in flow rate, it is inappropriate to analyze the inspiratory pressure:flow rate relationship beyond the point at which pressure is decreasing, i.e., beyond the nadir of pressure. Any air flow that occurs beyond the nadir of pressure is not flow limitation because pressure is increasing; however, this period of increasing pressure will be included in the flow rate versus time relationship. This is particularly a problem with breaths that are either not flow limited or only mildly flow limited because the pressure nadir usually occurs early in the breath. Thus, because of the hysteresis, a Level 1 breath with a very open airway may exhibit a constant flow rate beyond the nadir of pressure. In the flow rate versus time profile, this constant flow rate profile will be flat, appearing incorrectly as a Level 3 high flow resistance breath. In essence, with a measurement of flow rate alone, we do not know where the nadir of pressure is located and the accuracy with which actual flow limitation is detected will suffer.

View larger version (27K): [in this window] [in a new window]   Figure 7.   Effects of hysteresis in Pph versus flow rate on detecting flow limitation. This breath is categorized as linear according to the Pph:flow rate analysis. After the nadir of pressure, alinearities are present in the pressure:flow rate relationship and are reflected in the flow-versus-time relationship. Thus, this is a Level 1 breath with a very open airway exhibiting a constant flow rate beyond the nadir of pressure; however, in the flow rate versus time profile, the constant flow rate profile will be flat, appearing to be a Level 3, i.e., high flow-limitation breath.

Another limitation of noninvasive techniques was the failure to define high resistance breaths that have a "fixed resistance" and therefore a near linear pressure:flow relationship throughout the breath. Our algorithm evaluates the shape of the pressure:flow rate relationship. The great majority of high resistance breaths in sleeping subjects result from a highly compliant, floppy airway which starts off with a steep pressure:flow slope but then shows a sharp decline in flow rate as pressure continues to fall as inspiration proceeds. Our shape algorithm would correctly identify these alinearities in pressure versus flow or flow versus time. Six of the seven subjects we tested showed these types of high resistance breaths and there are several published reports showing the high prevalence of these types of pressure:flow alinearities in subjects with snoring and/or sleep-disordered breathing events (2, 7, 13).

However, some subjects with a narrowed but stiff upper airway will also show a high fixed resistance to air flow but a linear or near linear pressure:flow rate or flow rate versus time relationship. A high fixed resistance throughout inspiration might occur in the presence of neck flexion or an enlarged mass (e.g., tonsils) in the upper airway. This situation was most likely present in one of our subjects during sleep who showed an average resistance of 24 cm H₂O/L/s at the nadir of pressure. According to our shape analysis of Pph:flow rate, these breaths were classified as Level 1 breaths (i.e., a linear Pph:flow rate relationship) (see Table 1). Thus, our noninvasive algorithm will only respond to a changing time-dependent shape of the pressure:flow rate relationship, but if that relationship is linear with high resistance, we will not detect this high resistance with flow measurements alone. At first glance it may appear that incorporating a measure of flow amplitude would help detect this type of high resistance. Although it is likely that flow rate would be relatively depressed with high resistive loads (especially during sleep), one could not distinguish between a low "central" neural respiratory motor output versus high airway resistance as a cause of the decreased flow rate.

The gold standard of flow rate is pneumotach flow rate. This was definitely shown to be the best measurement in identifying the pressure:flow rate relationship from the flow rate versus time profile. However, tight-fitting masks frequently disrupt a subject's sleep. The next best alternative was the Pn as an indicator of flow shape. This method was able to distinguish mild from severe flow limitation and causes negligible disturbance to the patient. It is important to note that this study was conducted under ideal conditions, because the subject's mouth was taped shut, forcing all air to travel via the nares. Although the oral route for air flow is not commonly used during sleep, if indeed some flow was directed through the mouth, then the nasal pressure measurement would miss this flow and likely not be as effective as a sensor of actual flow shape or certainly of flow amplitude.

The performance of the differentiated sum Respitrace as an indicator of flow rate shape was the least sensitive noninvasive method. The sum Respitrace was carefully calibrated for each subject during wakefulness and the bands taped in place. However, the Respitrace did not capture all flow rate in the respiratory system throughout inspiration during sleep. This discrepancy is to be expected because the decreased tonicity of the postural muscles during sleep would lead to some degree of chest wall distortion especially with the more negative levels of intrathoracic pressure generated with dynamic airway narrowing during inspiration. While the sensitivity values for RIP may be marginally acceptable for identifying the extremes of flow limitation, we need to point out that our subjects were not extremely overweight and RIP body surface measurements might be much less reliable measures of air flow shape in the obese patient.

With the caveat of recognizing the ideal conditions for nasal pressure measurements used in our study, we would therefore recommend the use of the nasal pressure measurement combined with area analysis as the noninvasive measure of flow limitation that best combines specificity, sensitivity, and practicality. For purposes of CPAP titration through a nasal mask, applying the pneumotach flow rate signal to area analysis would be the preferred noninvasive method. In either case, we reemphasize that any of these measurements of flow rate "shape" will leave undetected the near linear but high (fixed) resistance breaths.

Finally, we also point out that our study was not carried out in symptomatic patients with syndromes associated with sleep-disordered breathing. Rather, our aim was to study methods for determining within-breath flow limitation and to this end we studied habitual, asymptomatic snorers in NREM sleep who displayed many types and a wide range of severity of flow-limited breaths. While these represent the great majority of the types of flow limitation encountered in epidemiologic studies of the working nonclinical population (16), and in many subjects with very high levels of airway resistance during sleep (2, 7, 11, 13, 15, 17) we cannot be sure that we have also included all of the abnormalities in upper airway mechanics encountered in clinical populations with severe sleep-disordered breathing. Accordingly, it is important that our algorithm for flow limitation detection and discrimination be evaluated in a larger clinical population.

What Severity of Flow Limitation Is Biologically Significant?

Significant flow limitation is important to detect as a means of assessing both the causes and acute and chronic consequences of sleep-disordered breathing and their sequelae. As outlined at the beginning of this report, if flow-limited breaths are sustained (i.e., beyond single breath occurrences) then some of these consequences might include transient arousals and disrupted sleep architecture, prevention of sustained periods of deeper sleep, reduction in ventricular stroke volume and transient hypopneas, or even prolonged hypoventilation with accompanying hypoxemia.

We expect that our noninvasive methods should be sufficiently sensitive and specific to consistently detect most types of flow-limited breaths of a magnitude that commonly leads to such undesirable acute responses as transient arousal, hypoventilation, prevention of slow wave sleep and/or perhaps limitation of cardiac function. In fact, given the minimal changes in peak negative pharyngeal pressure and single point estimates of upper airway resistance between linear and alinear levels of flow limitation (i.e., Level 1 versus 2), it may not be critical that our noninvasive methods were unable to distinguish with high sensitivity between these milder levels of flow resistance.

On the other hand, we caution that the biologic significance acutely and/or longer term, of a given degree of flow limitation is likely to vary widely and often unpredictably. For example, healthy subjects and those in heart failure will vary greatly in the effects of negative intrapleural pressure on left ventricular afterload and stroke-volume (10). Furthermore, arousability in response to a given level of pleural pressure development will vary even within a subject as sleep state changes (18, 19). Sleep state will also affect chest wall compliance (13) thereby influencing: (1) intrathoracic pressure development and the level of sensory input during inspiration for a given degree of airflow limitation; and (2) the effect on tidal volume of a given level of increased airway resistance.

Clearly, there is a need to examine further the biologic significance of flow-limited breaths during sleep, both in terms of their acute biologic effects (see above) and their longer term consequences on daytime blood pressure and hypersomnolence. We also need to understand more about what factors contribute to variability among subjects in terms of their acute and chronic biologic responses to a given magnitude (and duration) of air flow limitation. The study of the "upper airway resistance syndrome" is an important advance on these difficult questions (4). Indeed, the need continues for further work to define even a biologically significant apnea and hypopnea and ultimately a more useful sleep-disordered breathing index. Recognizing and quantifying air flow limitation is an important part of this index.