INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

It is well known that patients with cystic fibrosis (CF) and advanced lung disease develop hypoxemia and hypercapnia during sleep, particularly during rapid eye movement (REM) sleep (1). Hypoventilation has been proposed as a major mechanism for these blood gas changes, although no direct measurements of ventilation have been made in REM sleep. Hypoxemia and hypercapnia are poor prognostic signs in patients with CF (4, 5). Therefore, it is possible that preventing the onset of respiratory failure during sleep in CF may prolong survival in patients.

Studies of nocturnal oxygen therapy in CF have confirmed improved oxyhemoglobin saturation (Sp_O2) during sleep (6, 7), but usually with accompanying increases in arterial Pa_CO2 (7, 8). Moreover, long-term oxygen in patients with CF has failed to show improved survival (9). A number of descriptive noncontrolled retrospective studies have described noninvasive ventilation in patients with CF as being effective in improving or stabilizing arterial blood gas (ABG) tensions and as a bridge to transplantation (10). Gozal (8) showed that noninvasive ventilation was more effective than oxygen therapy in controlling carbon dioxide tension during sleep, but this was without direct measurement of ventilation.

The purpose of this study was therefore to make direct measurements of ventilation in all sleep stages in patients with CF with moderate to severe lung disease and to perform a direct comparison between the effects of low-flow oxygen and bilevel ventilatory support on ventilation and gas exchange during sleep. The goal was to determine if sleep-related hypoventilation in these patients could be ameliorated by the use of oxygen or bilevel therapy.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects and Protocol

We studied 13 adult patients with CF with moderate to severe lung disease while in a stable clinical condition, which was defined according to the criteria of Fuchs and coworkers (17). Subjects, with FEV₁ less than 65% of predicted, underwent a baseline diagnostic polysomnographic study as part of a research protocol designed to determine the prevalence of sleep-disordered breathing in this population, and were then asked to participate in this study if their initial sleep study showed episodes of oxyhemoglobin desaturation to < 90% during sleep. An additional eight patients with CF underwent diagnostic sleep studies during the time frame of this study. All of these additional eight patients had nocturnal oxyhemoglobin desaturation to < 90% but did not consent to participation in this current study.

Patients were required to undergo three sleep studies on room air (RA), low-flow oxygen (LFO₂), or bilevel ventilatory support (BVS) (VPAPII; ResMed, Sydney, Australia) ± O₂, in a random order, with continuous recordings of Sp_O2 (%) and transcutaneous carbon dioxide tension (TcCO₂) (mm Hg). During each of these sleep studies patients were required to wear a nasal mask in order for ventilation to be measured. The three sleep studies were conducted within a 1-wk period.

This study was conducted at Royal Prince Alfred Hospital in Sydney, Australia, and approved by the Central Sydney Area Health Service Ethics Committee (protocol no. X97-0204). Written, informed consent was obtained from all patients.

On the RA and LFO₂ nights all patients received continuous positive airway pressure (CPAP) of 4 to 5 cm H₂O (inspiratory positive airway pressure [IPAP] mode of the VPAP II device) throughout the entire study, during both wakefulness and sleep. This was to ensure that partial closure of the upper airway was not contributing to any decrements in ventilation seen, as well as providing a fresh gas flow into the mask, thereby minimizing any discomfort from the breathing circuit.

All patients were acclimatized to the mask and positive pressure during daytime practice sessions before commencement of the study. In addition, most patients also used the ventilator while sleeping during the week before the study with the machine pressure set at 4 cm H₂O CPAP, without using it on the night immediately preceding the study. No patient refused to participate in this study because of poor tolerance of the nasal mask or nasal ventilation during the acclimatization period. All patients were acclimatized to the point that they were comfortable with the study apparatus.

On the LFO₂ night, oxygen was delivered via the breathing circuit at a rate determined individually over the night to maintain Sp_O2  90% in all sleep stages. Likewise, on the BVS ± O₂ night, the IPAP and expiratory positive airway pressure (EPAP) were adjusted with the aim of maintaining Sp_O2  90% and avoiding increases in TcCO₂ associated with hypoventilation. Supplemental LFO₂ was added if, after optimal IPAP and EPAP were obtained, Sp_O2 remained < 90%.

Lung Function

Measurements of spirometry were performed using a Mass Flow Sensor (Sensormedics max 20; Sensormedics Corporation, Yorba Linda, CA) which was calibrated before each study and compared with normal predicted values of Quanjer and coworkers (18). Lung volumes were determined by body plethysmography (Gould 2800; Gould Electronics, Dayton, OH). Results were compared with normal predicted values of Goldman and Becklake (19). Inspiratory muscle pressure at residual volume (PI_max) and expiratory muscle pressure at total lung capacity (PE_max) were recorded using a hand-held pressure gauge and the results were compared with normal predicted values of Wilson (20). A gas chromatograph (1085D Series PF/Dx system; Medical Graphics, St. Paul, MN) was used to measure carbon monoxide transferred per liter of lung volume (KCO) and diffusing capacity of carbon monoxide (DL_CO). The normal predicted value for KCO was 6.9 ml CO/ mm Hg/min/L (STPD), a value based on mean laboratory values for normal nonsmoking healthy adults obtained in our laboratory.

ABG Tensions

Initial awake baseline ABG values were obtained with the patient seated and breathing RA, usually in the late afternoon. Repeat ABG measurements were then taken in the mornings after the oxygen and BVS nights. Subjects were woken for ABGs and remained breathing through the nasal mask during the procedure.

Sleep Study Recordings

During polysomnography, continuous recordings were made on a computerized system (Sleepwatch; Compumedics, Melbourne, Australia). Sleep stage was determined from two channels of electroencephalogram (EEG) (C3/A2, O2/A1), two channels of electro-oculogram (EOG) (ROC/LOC), and from the submental electromyogram (EMG). Respiratory variables were monitored using abdominal and thoracic impedance bands for chest wall movement and diaphragm EMG electrodes to reflect respiratory effort. Sp_O2 was measured with a finger probe (3700e; Ohmeda, Boulder, CO). Both nasal airflow and mask pressure were measured with a pressure transducer (Model DP-45; Validyne Corp., Northridge, CA). TcCO₂ (TCM3; Radiometer, Copenhagen, Denmark) was also measured continuously overnight. Periods of known artefact, including that due to body movement and loss of contact, in the Sp_O2 and TcCO₂ recordings were manually excluded from analysis.

Apnea was defined as cessation of airflow for  10 s, or a cessation of airflow for < 10 s with an oxygen desaturation of  3%. Hypopnea was defined as a reduction in amplitude of airflow, or thoracoabdominal wall movement of > 50% for  10 s, or a reduction in airflow or thoracoabdominal wall movement of > 50% for < 10 s if it was accompanied by an oxygen desaturation of  3%. The number of apneas and hypopneas per hour of non-rapid eye movement (NREM), REM, and total sleep time (TST) were calculated and reported as the respiratory disturbance index (RDI). TST average minimal Sp_O2 and REM average minimal Sp_O2 are calculated by averaging the minimal Sp_O2 recorded in each 30-s epoch of sleep. Sleep stages were scored in 30-s epochs according to the standard criteria of Rechtschaffen and Kales (21). An EEG arousal was defined as an abrupt increase in EEG frequency for  3 s that in REM sleep was accompanied by an increase in submental EMG amplitude.

The change in TcCO₂ (mm Hg) from NREM to REM sleep in our study was calculated for each REM period on each of the three study nights. A line of best fit was drawn between four points on the TcCO₂ slow recording: one at the start and end of each REM period, taking into account the approximately 3-min time delay for the device, and a point 5-min before and after each REM period so long as the TcCO₂ reading was stable. A perpendicular line was then drawn to the peak TcCO₂ reading for that REM period being measured. A weighted average was then obtained for the delta TcCO₂ for each subject on each study night (Figure 1).

View larger version (22K): [in this window] [in a new window]   Figure 1.   Typical example of desaturation and hypercapnia in REM sleep demonstrating the method of summarizing the increase in transcutaneously recorded CO2.

Measurements of Ventilation

Ventilation was measured directly with a pneumotachograph (Res- Med) coupled to a pressure transducer (DP-45; Validyne) placed between the nasal mask and the exhalation port of the mask system. The accuracy of this system was tested and reported in the work of Becker and coworkers (22). To determine whether there was any significant effect of nasally applied CPAP on breathing at rest, they measured minute ventilation, tidal volume (VT), and respiratory rate (RR) in eight healthy subjects with and without CPAP at 4 cm H₂O. They found no significant effect of this level of pressure on any parameter of ventilation. Similarly, it has been found that CPAP delivered at a minimal pressure of at least 5 cm H₂O did not alter VT or RR in subjects with chronic obstructive pulmonary disease (COPD) (23, 24).

The dead space of the mask and the pneumotachograph was approximately 100 ml, and the resistance of the system was 0.2 kPa · s · L¹. Tidal volume (VT) (mean value of inspiratory and expiratory VT), RR, and breath-by-breath minute ventilation (I) were calculated from the flow signal produced by the pneumotachograph. Before each study night the pneumotachograph was calibrated against a syringe of known volume. An average of 45 ± 16 measurements were recorded at frequencies of 10 to 42/min. The measured VT value was 99.9 ± 2.6% of the syringe volume. The following morning the calibration of the pneumotachograph was checked. The measured value was 98.1 ± 5.5% of the syringe volume. During all recordings the flow signal was continuously displayed on a monitor so that significant mask or mouth leaks could be detected as either a deviation from the zero-flow value at end expiration or a difference in the inspiratory and expiratory VT values. Only periods free of leaks were selected for analysis. Mean values for Sp_O2 were calculated for each breath. These data were acquired online with a 12-bit AC/DC converter, with sampling rate at 125 Hz.

Data Analysis and Statistics

Mean values of VT, RR, and I were calculated during wakefulness and during REM and NREM sleep for each subject. Data were assessed for normality and then either an analysis of variance (ANOVA) with repeated measures or a Friedman's two-way ANOVA was used to assess the data, with Bonferroni's and Dunn's multiple comparison tests being used for post hoc analysis, respectively (25). Paired t tests were used to analyze the ABG results. Data are reported as mean ± 1 SD. Statistical significance was assumed at p < 0.05.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

All 13 subjects had moderate to severe lung disease with FEV₁ ranging from 17.6 to 56.3% of predicted values. Subjects had awake Pa_O2 values ranging between 53 mm Hg and 77 mm Hg and six subjects were hypercapnic during wakefulness (Pa_CO2 > 45 mm Hg). Two additional subjects had Pa_CO2 values equal to 45 mm Hg. Anthropomorphic and daytime pulmonary function are presented in Table A in the online data supplement ().

On the BVS study night a mean IPAP of 11.9 ± 1.7 cm H₂O and mean EPAP of 4.5 ± 0.8 cm H₂O were used. Seven of 13 subjects required additional oxygen on the BVS night with a mean requirement of 0.7 ± 0.9 L/min to maintain Sp_O2  90%. On the LFO₂ night all subjects received supplemental oxygen, with a flow requirement of 1.4 ± 0.9 L/min to maintain Sp_O2  90%.

One patient was unable to tolerate the increases in IPAP necessary to maintain his Sp_O2  90% on the BVS night. He was able to tolerate a maximal IPAP of 8 cm H₂O, and as a result he required more supplemental oxygen than other subjects during the BVS night. In the remaining subjects no problems were reported regarding mask fitting or tolerance to the pressure titration procedures.

Changes in Ventilatory Pattern

Ventilation was measured directly in sleep in patients with CF using technology recently developed in our laboratory and described in detail by Becker and coworkers (22). Our direct measurements confirmed hypoventilation in REM sleep in patients with moderate to severe lung disease due to CF. I was reduced on the RA night from awake to REM sleep (p < 0.001) and from NREM to REM sleep (p < 0.01). Similarly, I was reduced on the LFO₂ night from awake to REM sleep (p < 0.001) and from NREM to REM sleep (p < 0.001). No significant fall in I was seen from awake to NREM sleep on any night. There was no significant difference in I between NREM and REM sleep with BVS, although there was still a reduction in I from awake to REM sleep (p < 0.001) (Table 1, Figure 2A). In this regard BVS was more effective than oxygen therapy in attenuating hypoventilation during sleep.

View larger version (38K): [in this window] [in a new window]   Figure 2.   Changes in I (A), VT (B), and RR (C ) associated with wakefulness, NREM and REM sleep on each of three study nights (RA, LFO2, and BVS). *p < 0.01 REM versus NREM and awake, p < 0.001 REM versus awake only, **p < 0.01 awake versus NREM and REM, p < 0.01 awake versus NREM only, ¶ p < 0.05 NREM BVS versus NREM RA and NREM LFO2, p < 0.05 REM BVS versus REM RA and REM LFO2.

The reduction in I with REM sleep on the RA and LFO₂ nights appeared to be the result of a reduction in VT from NREM to REM sleep that did not occur with BVS, although this did not reach statistical significance. VT was significantly lower during sleep than wakefulness on all three nights (p < 0.001) (Table 1, Figure 2B). There was no difference in awake VT between the three study nights, but the VT values during both NREM and REM sleep were significantly greater with BVS than RA or LFO₂ (Figure 2B).

On each of the three study nights RR was higher during NREM sleep than during wakefulness (p < 0.01). On the RA and LFO₂ nights the RR during REM was also greater than that measured during wakefulness (p < 0.01), and there was no significant difference between the NREM and REM sleep RR. Interestingly, the REM sleep RR on the BVS ± O₂ night was not greater than the awake RR, as had been seen on the LFO₂ and RA nights (Table 1, Figure 2C). The RR in REM sleep with BVS was less than with RA or LFO₂ (p < 0.01, p < 0.05, respectively) (Figure 2C).

Pulse Oximetry, TcCO₂, and ABG Changes

Both LFO₂ and BVS ± O₂ improved the nocturnal Sp_O2 % in this group in REM sleep (p < 0.01) and overall as a percentage of TST (p < 0.05) (Figure 3). The percentage of REM with Sp_O2 greater than 90% on RA was 56 ± 41%, compared with 90 ± 25% with LFO₂ and 93 ± 11% with BVS. The percentage of TST with Sp_O2 greater than 90% on RA was 64 ± 42%, compared with 94 ± 17% with LFO₂ and 93 ± 11% with BVS. There was no significant difference in the percentage of NREM time with Sp_O2 > 90%.

View larger version (25K): [in this window] [in a new window]   Figure 3.   Percent (%) of time spent with SpO2 > 90% in NREM, REM, and TST on each of three study nights (RA, LFO2, and BVS). *p < 0.05, **p < 0.01 RA versus LFO2 and BVS.

The mean maximal change in TcCO₂ from NREM to REM on the RA night was 4.1 ± 1.5 mm Hg, 4.7 ± 2.9 mm Hg during LFO₂, and 1.6 ± 0.9 mm Hg with BVS ± O₂ (Figure 4). The post hoc analyses on the repeated measures ANOVA performed on these results showed no difference between the RA and LFO₂ nights. However, BVS ± O₂ significantly attenuated the rise in TcCO₂ seen with REM sleep compared with both the RA and LFO₂ nights (p < 0.05, p < 0.01, respectively).

View larger version (12K): [in this window] [in a new window]   Figure 4.   Change in TcCO2 (mm Hg) from NREM to REM sleep on each of three study nights (RA, LFO2, and BVS).

Four subjects had incomplete ABG data, with one or more of the ABG punctures being unsuccessful, leaving only 10 to 11 paired data points in the ABG comparisons. Three ABG comparisons were made, two being the change in ABGs from evening to morning on the two treatment nights and the third being the direct comparison of the morning ABGs after treatment. No significant difference was seen in morning Pa_O2 between the LFO₂ and BVS nights. Pa_O2 was also noted to remain unchanged from the evening to the morning measurement irrespective of the treatment received overnight. A rise in Pa_CO2 was seen from evening to morning on both the LFO₂ and BVS nights (p < 0.01), but there was a trend toward a lower morning Pa_CO2 after BVS ± O₂ compared with the LFO₂ night (p = 0.06). This was supported by the pH results showing that these subjects were significantly more acidotic after a night on LFO₂, from an awake evening pH of 7.4 ± 0.02 to a morning pH of 7.36 ± 0.03 (p < 0.001). Likewise, these subjects showed a decrease in pH after the BVS night to 7.39 ± 0.02 (p < 0.05), but when comparing the morning pH values between the two treatments these subjects were significantly more acidotic after LFO₂ than BVS ± O₂ (p < 0.05).

Sleep Architecture and RDI

Neither LFO₂ nor BVS significantly modified sleep architecture with regard to time available for sleep, TST, sleep latency, REM sleep latency, REM and NREM minutes, percent REM/NREM sleep, arousal index, percent stage 1 and 2 sleep, and percent slow-wave sleep. Sleep efficiency was a little higher with BVS when compared with RA (p < 0.05), but there were no differences between RA and LFO₂ or between LFO₂ and BVS (Table B, online data supplement).

There was a significant difference in overall RDI (events per hour) between the three nights (p < 0.001), with the TST RDI on the BVS night being lower than on the RA (p < 0.001) or the LFO₂ (p < 0.01) nights (Table B, online data supplement). The highest RDI was seen in REM sleep on the RA night and was associated with hypopneas, of which there were significantly more on the RA (3.4 ± 2.4 hypopneas/hour) and LFO₂ (2.3 ± 1.7 hypopneas/hour) nights than the BVS night (0.1 ± 0.2 hypopneas/hour) (p < 0.01). The REM RDI was significantly greater than the NREM RDI on all three study nights using the Wilcoxon signed rank test (25). The differences were large on the RA (p = 0.0015) and LFO₂ (p = 0.0005) nights and were smaller on the BVS night (p = 0.05). There was no significant improvement in RDI, either TST RDI, NREM RDI, or REM RDI with the addition of LFO₂ to the circuit. BVS led to a reduction in TST RDI and in REM RDI (p < 0.05).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We measured ventilation directly in a group of patients with CF, moderate to severe lung disease, and oxyhemoglobin desaturation during sleep. Our results clearly showed that I in REM was significantly lower than in NREM or wakefulness during both RA and LFO₂ therapy. This REM-related hypoventilation was markedly reduced with BVS so that ventilation was maintained from NREM to REM sleep.

There is limited information on quantitative changes in ventilation during sleep in patients with lung disease. I during sleep has been measured in patients with COPD (26) and with CF (3). These early studies were done using an inductance vest, or respiratory inductance plethysmography. The accuracy of inductance plethysmography in patients with lung disease and over longer recording periods with changes in posture has been questioned (29) because of poor correlation in all sleep stages between plethysmographic VT and expired volume measured by a pneumotachograph. By contrast the measurement of T with a pneumotachograph is accurate and reproducible, and does not depend on body position. The technique to measure ventilation used in our institution with use of low-level CPAP has been shown to be reliable over long periods of recording (22). Although it is possible that similar low levels of CPAP used in the current study had some impact upon ventilation and oxygenation, we do not believe that this effect would be large, as low levels of CPAP have been shown not to affect breathing at rest in normals (22) or in stable COPD (23, 24). In a study by Regnis and coworkers higher levels of CPAP were required to improve sleep oxygenation and decrease respiratory disturbance (30); therefore we believe that low levels of CPAP are not likely to have had major impact. Moreover, the level of CPAP used in our study remained the same on all study nights for each subject.

One potential source of measurement error in our study may have been mouth and mask leaks. These were detected as described, and recording periods in which leaks occurred were not used in our analysis. It could be argued that a potential source of bias in our study was introduced by the eliminating of periods of breathing when leaks occurred, as this could result in selection of periods of greater ventilation for measurement. This strategy of analysis was used as our aim was to compare the effects of LFO₂ and BVS when delivered optimally. Air leaks during the use of BVS may reduce the effectiveness of this therapy. In our study mouth leak was minimized by the use of a chin strap, and leaks were rectified as soon as possible after detection.

Using a horizontal body plethysmograph, with airflow being measured with a pneumotachograph, Ballard and colleagues (31) observed that I decreased by 18.5% during stages 3 and 4 NREM sleep, and by 35.5% during REM sleep as compared with wakefulness in five patients with COPD. Using similar methodology Ballard and colleagues studied spontaneous ventilation during sleep in six patients with CF (32), reporting a reduction in I from wakefulness to NREM sleep resulting from a decrement in VT. No data were obtained during REM sleep in this study. In the current study there was no significant difference between awake and NREM I on any of the three study nights, but the decrease in VT as described by Ballard and colleagues was seen. Figure 2 illustrates that although the decrease in VT from awake to NREM sleep remained significant with BVS, the VT on the BVS night was significantly greater in both NREM and REM sleep than on the RA or LFO₂ study nights.

In contrast to the decrement in I between NREM and REM seen on our RA and LFO₂ study nights, there was no significant reduction on the BVS night. This indicates that BVS is able to prevent the reduction in ventilation that occurs in REM sleep in patients with CF and thus prevents oxyhemoglobin desaturation during sleep. Our results do not exclude the possibility that ventilation/perfusion inequality contributed to sleep-related hypoxemia. However, they do indicate that reduction in ventilation was a very important cause of oxygen desaturation during sleep.

Use of both LFO₂ and BVS ± O₂ improved nocturnal Sp_O2 (%) in all sleep stages. In addition, BVS was able to successfully attenuate the rise in TcCO₂ that occurred on both the RA and LFO₂ nights. Further evidence that BVS was effective in minimizing the degree of hypoventilation occurring during sleep was provided by the ABG samples, which showed an improvement in pH and a trend toward a lower Pa_CO2 after a night of BVS compared with LFO₂.

An increase in the sleep-related rise in carbon dioxide levels with the use of supplemental oxygen versus room air during sleep has previously been reported in a number of studies (6- 8). Gozal (8) compared BVS ± O₂ to oxygen therapy alone in six patients. No direct measurements of ventilation or ABG tensions were made. With this night-to-night comparison of BVS and LFO₂, it was found that BVS was as effective as supplemental oxygen in improving Sp_O2 during REM and NREM sleep in this patient group. Gozal reported significant increases in TcCO₂ with oxygen therapy that were not found in the BVS group. The rise in TcCO₂ from NREM to REM sleep in our study was not significantly higher with LFO₂ than with RA. However, we have shown a reduction in the rise in TcCO₂ with BVS. In view of the attenuation of the rise in TcCO₂ together with the maintenance of minute ventilation with BVS in REM sleep, we conclude that BVS ± O₂ led to improvements in ventilation during sleep in these patients with CF.

In conclusion, noninvasive positive pressure ventilation, with oxygen if required, was as effective as LFO₂ alone in preventing sleep-induced hypoxemia in patients with CF and moderate to severe lung disease. Improved oxygenation was achieved without modification of sleep quality and efficiency. An additional and potentially clinically important benefit of noninvasive positive pressure ventilation was improvement of alveolar ventilation. Overcoming REM-related hypoventilation over a longer time period with the use of BVS may be important in delaying the onset of awake hypoxemia and hypercapnia, both of which are poor prognostic markers for survival in CF (4, 5). However, further studies are required to make conclusions about long-term outcome with nocturnal BVS compared with nocturnal LFO₂. Despite concerns about reduced patient tolerance and compliance expressed by other researchers (8), noninvasive positive pressure ventilation should be considered as a feasible therapeutic modality in CF patients with advanced lung disease with better efficacy at preventing sleep-related hypoxemia and hypercapnia than nocturnal LFO₂ therapy.