INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Some patients have symptoms and signs that suggest obstruction of the upper airway during sleep, but standard polysomnography shows too few apneas and hypopneas to make the diagnosis of sleep apnea. Monitoring esophageal pressure (Pes) in such patients during polysomnography may allow identification of a subset who demonstrate repetitive, gradual increases in negative intrathoracic pressures and respiratory effort that terminate in arousals. This condition, called the upper airway resistance syndrome (UARS) (1), responds to the same treatments used for obstructive sleep apnea. Although the prevalence of UARS is unknown, one study demonstrated UARS in 15 of 48 patients previously diagnosed as having idiopathic hypersomnia (2). The condition may be particularly relevant to women and children (3).

Despite increased recognition of and interest in UARS, monitoring of Pes is not widespread, in part because of concern that the procedure may disturb sleep. However, little information has been published on the effect of monitoring Pes on sleep architecture. Several authors have suggested that Pes monitoring with an esophageal balloon disrupts sleep. A popular sleep text reported that 10 "young volunteers" monitored for several nights had less Stage 3 and Stage 4 sleep when an esophageal balloon was used, but the authors gave no further details (7). An abstract (8) that described 21 patients in whom Pes was recorded on two nights, with and without esophageal balloons, was interpreted by others (9) as showing sleep disruption as a result of Pes monitoring, but the original authors' conclusions appear less clear. In another study, use of an esophageal balloon for the first 2 h of a nocturnal polysomnogram may have caused alpha delta sleep to occur in one reported case (9).

A more recently developed and validated alternative to use of an esophageal balloon for monitoring Pes is use of a water-filled catheter connected to a transducer (10). This system resembles that commonly used in intensive care units to monitor intraarterial blood pressure. The water-filled catheter is thinner than the esophageal balloon and may disturb sleep less. To assess the effect of the water-filled catheter on sleep architecture, we performed a retrospective comparison of sleep architecture in patients who had concurrent Pes monitoring in our sleep laboratory and matched patients who did not.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Through the use of a computerized database, we identified all patients who had diagnostic polysomnograms performed with Pes monitoring in our laboratory from December 1994 through November 1996 ("Pes subjects"; n = 160). Split-night studies in which continuous positive airway pressure (CPAP) titration was performed were excluded, as were studies done after surgery for sleep apnea. In all instances, Pes was recorded because UARS was within the pretest differential diagnosis.

To identify polysomnograms suitable for comparison with the Pes studies, we next considered all diagnostic polysomnograms performed from January 1988 through November 1994, when our laboratory did not monitor Pes. From January 1988 through November 1996, scoring rules and laboratory procedures for polysomnography had remained otherwise constant. We again excluded all split-night studies and those done on patients who had had surgery for sleep apnea. We also eliminated all studies on patients less than 3 yr old (no Pes subject was within this age range) and a small number of studies that could not be matched because data were incomplete. The result was a group of 2,281 potential comparison studies.

For each Pes subject's study, we then selected one matched comparison study that met all the following criteria: (1) patient of same sex; (2) patient in same age category (based on ages 3 to 9 yr, 10 to 19 yr, 20 to 40 yr, 41 to 64 yr, and > 64 yr); (3) mean apneas /hypopneas per hour of sleep (apnea /hypopnea index; AHI) within 20% of the Pes subject's AHI; and (4) minimum oxygen saturation (minO₂) within 4 percentage points of the Pes subject's minO₂. Matches were available for 155 of the 160 Pes subjects; one Pes subject could not be matched for AHI, and four could not be matched for minO₂.

Procedures

Nocturnal polysomnography included four electroencephalographic (EEG) leads (C3-A2, C4-A1, O1-A2, and O2-A1 of the 10-20 international electrode placement system), two electrooculographic (EOG) leads (right and left outer canthi), chin and bilateral anterior tibialis surface electromyograms (EMGs), two electrocardiographic (ECG) leads, nasal and oral airflow, thoracic and abdominal excursion, and finger oximetry. Sleep-stage scoring was performed on 30-s epochs according to standard criteria (11) by technologists who had undergone an extensive training program and had correctly scored (as determined by two physicians board-certified in sleep medicine) at least 90% of epochs in a set of reliability records. An awakening was defined as one or more contiguous epochs scored as wakefulness and flanked by epochs of sleep. Sleep latency was defined as the time from the point at which lights were turned out until the first epoch of any stage of sleep. An apnea was defined as complete cessation of airflow during sleep for at least 10 s. A hypopnea was defined as a reduction in airflow leading to either a 4% or greater oxyhemoglobin desaturation, an arousal, or an awakening.

The Pes subjects also had concurrent Pes monitoring with a water-filled catheter (10), which was inserted after one nostril was anesthetized with topical viscous xylocaine. The pediatric, size 6 French nasogastric tube was passed by a technician to a length calculated on the basis of the patient's height (0.288 × height) to locate the tip of the catheter in a retrocardiac position several centimeters above the gastroesophageal sphincter. Insertion of the catheter was generally well-tolerated by patients. Upon chart review of the 48 planned catheter insertions between May and October 1995 in our laboratory, we found no difficulties except for one instance in which an inadequate signal on the polygraph necessitated catheter reinsertion, and another instance in which the patient refused the procedure.

Among the 155 pairs of subjects, 74 pairs had Multiple Sleep Latency Test (MSLT) results available for both subjects, from studies done on the day following the polysomnograms. The MSLTs were performed in a standardized manner (12).

Analysis

Data on the 155 pairs of subjects, including results of their polysomnograms and MSLTs, were analyzed with the SAS system (SAS Institute Inc., Cary, NC). To compare a proportion among Pes subjects to that among comparison subjects, we used the normal approximation to the binomial. To assess for differences within pairs of patients, we used paired t tests. Although multiple tests were planned, significance was kept at p = 0.05, at the risk of spurious findings, so that any potential relationships between Pes monitoring and sleep architecture would not be missed. We also calculated 95% confidence intervals (CIs) to show the maximum extent of differences between the studies of Pes and comparison subjects.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

The 310 subjects included 66 male and 89 female pairs. The subjects' mean age was 36.4 yr, with a standard deviation of ± 15.9 and a range of 3 to 83 yr. Twenty subjects were 3 to 9 yr old, 26 subjects were 10 to 19 yr old, 140 subjects were 20 to 40 yr old, 114 subjects were 41 to 64 yr old, and 10 subjects were 65 to 83 yr old. The mean AHI was 10.4 ± 11.8, and the mean minO₂ was 86.9 ± 5.6%. As expected, no statistically significant differences between Pes and comparison subjects were detected for variables that had been matched (Table 1). On average, comparison subjects had had polysomnography 4.3 ± 1.9 yr before Pes subjects.

Table 2 lists sleep-related diagnoses that were considered firm or probable by the sleep-medicine specialists who reviewed polysomnographic and clinical data after the patient's sleep studies (13). Only the more common diagnoses are listed, and some patients had more than one diagnosis. No significant differences between the Pes and comparison subjects were noted, except that UARS or increased respiratory effort during sleep was found in 31 Pes subjects but could not be assessed in comparison subjects.

As compared with the 310 subjects in this study, the entire group of patients who had diagnostic studies in our laboratory during the same time frame (n = 3,441) had a mean age of 40.9 ± 19.6 yr, and 37% were female. The mean AHI was 25.0 ± 31.8, and the mean minO₂ was 80.8 ± 13.2. Obstructive sleep apnea or UARS was the primary diagnosis in 56% of the patients.

Questionnaire data on home medication use was available from both members of 72 subject pairs. Eight (11%) of these pairs were discordant for hypnotic use; in seven pairs the Pes subject reported use of an hypnotic, and in one pair the comparison subject did so. Thirteen pairs (18%) were discordant for use of other psychoactive medications; in 10 pairs it was the Pes subject and in three pairs the comparison subject who used an antidepressant, antipsychotic agent, antianxiety agent, or stimulant.

Sleep Architecture in Pes and Comparison Subjects

Sleep architecture in the study subjects is summarized in Table 3. The largest differences between Pes and comparison subjects were in total recording time (TRT) and total sleep time (TST). The TRT was, on average, 467.9 ± 36.8 min in Pes subjects, as compared with 498.7 ± 39.6 min in matched comparison subjects (paired t test, p  0.0001). TST was proportionately reduced among Pes subjects (p  0.0001), and sleep efficiency (TST/TRT) showed only a small decrement (p = 0.04). The 95% CI for sleep efficiency indicated that, based on our data, patients who have Pes monitoring have at worst a 5.7% lower sleep efficiency than subjects who do not have Pes monitoring. Other measures of wakefulness, such as the number of awakenings per hour of sleep, did not show any significant differences between Pes and comparison subjects.

The latency to sleep onset (first epoch of Stage 1 sleep) and latency to Stage 2 sleep did not differ for Pes and comparison subjects. However, the latency to the first five consecutive minutes of sleep was 10 min longer, on average, among Pes subjects (p = 0.02). The latency to the first epoch of rapid-eye-movement (REM) sleep was, on average, 24 min longer (p = 0.01).

Stage 1 sleep as a percentage of TST was no different in Pes and comparison subjects; neither was the number of entries to Stage 1 sleep. Percent Stage 2 sleep and percent REM sleep were reduced, and percent Stage 3/4 sleep was increased to statistically significant degrees among Pes subjects, but the small mean differences and narrow CIs show that the magnitudes of these differences were not large. For example, the most extreme difference, that for percent Stage 3/4 sleep, represents at most a 5 percentage point increase in patients recorded with Pes monitoring. Among the 74 pairs of subjects with MSLT data, the mean sleep latency on the MSLT showed no significant difference between Pes and comparison subjects.

Potential Reasons For Different Total Recording Times

We considered three possible explanations for different total recording times (TRT) among Pes and comparison subjects. The first was that Pes subjects went to sleep later than comparison subjects because of a lengthened setup time. However, this explanation seemed unlikely because insertion of the Pes monitoring catheter takes only about 10 min, and the entire setup is generally done well in advance of subjects' usual (and laboratory) bedtime. A second possible explanation was that studies in Pes subjects could have been terminated early owing to patient discomfort or inability to sleep. This possibility could not be directly tested, but duration of the final morning awakening was no different for Pes and comparison subjects (paired t test, p = 0.98); a longer final morning awakening might have been expected in Pes subjects if the catheter had prevented them from sleeping at that time.

The third possible explanation for the observed difference in TRT between Pes and comparison subjects was that TRT became shorter during the 8 yr of polysomnography in our laboratory, so that the more recent (Pes) studies had shorter TRTs than the older (comparison) studies. A Pearson correlation between the dates of the polysomnograms and TRTs confirmed a decrease in TRT during these years (r = 0.40, n = 310, p < 0.0001). This decrease predated introduction of Pes monitoring in our laboratory, since it was statistically significant before December 1994 (r = 0.26, n = 155, p = 0.0013) but nonsignificant and in the opposite direction after this date (r = 0.067, n = 155, p = 0.41). Furthermore, the changes with time in comparison subjects' TRTs could not fully explain the within-pair differences in TRT; no relation was present between the within-pair difference in TRT and the amount of time that separated the two studies (linear regression, p = 0.61).

We suspect that the shorter TRT in Pes studies partly reflects differences in study goals between the later Pes and earlier comparison studies. In the comparison studies, idiopathic hypersomnia or insufficient sleep syndrome may have been in the differential diagnosis more often, and we try to perform 9-h polysomnograms rather than 8-h studies in these cases. A gradual increase in the proportion of studies done mainly to assess sleep-related breathing could explain the mild decrease with time in the TRT before December 1994.

Potential Confounding Variables

To check whether within-pair time between studies might have affected other variables and acted as a confounding factor in the relationships we identified between sleep-architecture variables and Pes monitoring, we regressed the within-pair difference in each remaining sleep-architecture variable on the within-pair difference in study dates. None of these regressions showed significance, except that the number of entries to Stage 1 sleep increased slightly with time between studies (p = 0.049); the number of entries to Stage 1 sleep had in any case not proven higher in Pes studies than in non-Pes studies (Table 3). Our results suggest that none of the relationships in Table 3 were significantly confounded by changes with time in our laboratory procedures or patient population.

The greater frequency of drug use among Pes subjects also created a potential confounding factor in our data. We therefore repeated all the calculations listed in Table 3 for three subsets of subjects: (1) the 72 pairs for which each subject's medication information was available; (2) the 64 pairs for which hypnotic use was concordant; and (3) the 59 pairs for which psychoactive medication use was concordant. None of the within-pair differences in sleep-architecture variables showed substantial changes when the first subset was compared with the second or third.

Children, Older Subjects, and Women

Within-pair differences in all polysomnographic variables listed in Table 3 were calculated separately for three other subgroups of subjects: (1) among children aged 3 to 9 yr (n = 10 pairs), no statistically significant differences between Pes and comparison subjects were detected (p > 0.10 in each case); (2) among subjects who were at least 55 yr old (n = 13 pairs), only TRT was significantly different (p < 0.05) between Pes and comparison subjects; and (3) among women (n = 89 pairs), the only variables that showed statistically significant differences (p < 0.05) were TRT, TST, percent Stage 2 sleep, and percent Stage 3/4 sleep. MSLT data for nine pairs of older subjects and for 40 pairs of women showed similar mean sleep latencies in Pes and comparison subjects (p > 0.10).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study shows that sleep architecture in patients who have polysomnography with Pes monitoring closely resembles sleep architecture recorded in matched patients who do not have Pes monitoring. The data we report, which to our knowledge represent the largest published series of Pes recordings, revealed small, statistically significant differences in sleep architecture between the two groups of patients, but also quantified these differences with sufficient precision to suggest that they are not likely to be clinically important. When we limited the analyses to include only children aged 3 to 9 yr, adults aged 55 yr or more, or women, we still obtained results similar to those in the entire group in each instance; fewer of the differences between Pes and comparison studies reached statistical significance, in part because of the reduced sample sizes. Moreover, results of MSLTs, performed in about half our patients, showed virtually identical degrees of daytime sleepiness whether or not Pes monitoring had been performed on the previous night, further suggesting that any disruption of sleep associated with Pes monitoring was not clinically significant.

Patients who underwent Pes monitoring were matched with those who did not on the basis of sex, age, AHI, and minO₂. However, uncontrolled differences between the Pes and comparison groups could have arisen by chance, confounded our results, and obscured a real difference between groups in sleep architecture. Specific diagnoses showed similar frequencies in Pes and comparison subjects, and therefore could not have acted as confounding variables. The amount of time elapsed between Pes and comparison studies generally did not affect differences in sleep architecture. Furthermore, hypnotic and psychoactive drug use may have been more frequent among Pes subjects, but evidence suggested these were not significant confounding factors: among the subset of 72 subject pairs for whom we had medication information, only a small minority were discordant for drug use, and exclusion of discordant pairs from the analysis had no substantial effect on results.

Little prior work has been published on the potential effect of Pes monitoring on sleep architecture. One group studied 32 subjects on two nights, with and without Pes monitoring; rather than using a water-filled catheter, they used a catheter of similar thickness that contained several transducers (14). The presence of this apparatus produced small but not statistically significant differences in measures of sleep architecture, including TST, number of sleep-stage shifts, sleep efficiency, wakefulness, percent REM sleep, and number of microarousals. These results parallel our own, and might also have shown small, statistically significant effects of Pes monitoring if the sample size had been larger.

In our study, the largest differences in sleep architecture associated with Pes monitoring were mild reductions in TRT and proportionate reductions in TST. The reduced recording times in Pes studies may have been partly related to procedural differences during the 8 yr period which studies were performed. If an unidentified effect is also present, a total recording time that is half an hour shorter on average is still not likely to have substantial clinical impact in practice. In our Pes patients, sleep efficiency was only slightly reduced, and the rate of awakenings was unaffected. Our patients who slept with the water-filled catheter actually had a slightly higher percent Stage 3/4 sleep than did their matched controls, a result contrary to that reported for 10 subjects who slept with esophageal balloons (7). One potential explanation for our finding is that Pes subjects, who had a shorter recording time and perhaps later bedtimes, may have exhibited a compensatory increase in deep non-REM sleep (15).

Although we collected data on the number of awakenings30-s epochs during which the patient was awake at least 50% of the timeas defined by standard scoring criteria (11), we did not tabulate briefer arousals in our patients, and this is a potential limitation of our study. Suggested guidelines for scoring arousals of as short as 3 s were not published until 1992 (16). In experimental paradigms, brief arousals can affect daytime sleepiness, and in clinical practice brief arousals may be an important cause of sleepiness in UARS (17). Our results cannot exclude the possibility that Pes monitoring causes an increase in brief arousals. However, we do not believe such an effect is likely, for several reasons. First, the relatively immobile esophageal catheter is unlikely to generate the multiple, short, and discrete stimuli usually necessary to produce arousals. Second, in our patients, the number of entries to Stage 1 sleep, which is probably a reasonable approximation of the number of arousals, was no different on average whether or not Pes monitoring was used. Third, if arousals were increased to a clinically significant degree during esophageal monitoring, we would have expected to see lower MSLT scores the next day; instead, the average MSLT scores were no different whether or not Pes monitoring had been performed.

The objective of the current study was to examine the effect of Pes monitoring on sleep architecture, rather than to assess the utility of Pes monitoring. However, a growing body of literature suggests that monitoring respiratory effort by recording intrathoracic pressure is important for some patients with sleep-disordered breathing. Monitoring of Pes facilitates the diagnosis of UARS, which causes excessive daytime sleepiness and may contribute to hypertension (2, 18). Children with narrow upper airways may be particularly prone to UARS (3, 4), and a recent report based on a large series of pediatric patients suggested that Pes monitoring can be superior to both end-tidal and transcutaneous capnography for detection of subtle obstructive events during polysomnography in children (19). In our laboratory, insertion of the catheters used for Pes monitoring generally has not been overly upsetting to children or their parents.

Monitoring of Pes may also be particularly useful in female, young, or thin adults who have symptoms of obstructed breathing during sleep (5, 6). In patients with apneas and hypopneas, esophageal manometry allows quantification of increased respiratory effort that may be responsible for arousals (20). In some patients, Pes readings can help distinguish central from obstructive respiratory events. Measurements of Pes can be used to help titrate CPAP to the optimum level (21), and may be useful postoperatively in some cases to identify a reason for continued sleepiness after surgery for sleep apnea (22).

Despite potential applications of Pes monitoring during polysomnography, few sleep laboratories currently employ the technique. Some of the hesitation to adopt this relatively inexpensive technology may relate to the concern that the esophageal catheter could cause sleep disruption. The American Thoracic Society (ATS) recently recommended that esophageal monitoring not be routinely required in pediatric polysomnography, in part because of the suspicion that "the presence of the catheter may contribute to sleep disruption, leading to a less-than-optimal study" (23). Our data suggest that Pes monitoring does not disrupt sleep to an extent that is clinically significant. The decision of whether or not to use Pes monitoring during polysomnography should therefore depend to a larger extent on the need for a quantitative measure of respiratory effort.