Abstract
Sleep hypoventilation (SH) may be important in the development of hypercapnic respiratory failure in chronic obstructive pulmonary disease (COPD). The prevalence of SH, associated factors, and overnight changes in waking arterial blood gases (ABG), were assessed in 54 stable hypercapnic COPD patients without concomitant sleep apnoea or morbid obesity.
Lung function assessment, anthropomorphic measurements, and polysomnography with ABG measurement before and after sleep were conducted in all patients. Transcutaneous carbon dioxide tension (Pt,CO2) was measured in sleep, using simultaneous arterial carbon dioxide tension (Pa,CO2) for in vivo calibration and to correct for drift in the sensor.
Of the patients, 43% spent ≥20% of sleep time with Pt,CO2 >1.33 kPa (10 mmHg) above waking baseline. Severity of SH was best predicted by a combination of baseline Pa,CO2, body mass index and per cent rapid-eye movement (REM) sleep. REM-related hypoventilation correlated significantly with severity of inspiratory flow limitation in REM, and with apnoea/hypopnoea index. Pa,CO2 increased mean±sd 0.70±0.65 kPa (5.29±4.92 mmHg) from night to morning, and this change was highly significant. The change in Pa,CO2 was strongly correlated with severity of SH.
Sleep hypoventilation is common in hypercapnic chronic obstructive pulmonary disease, and related to baseline arterial carbon dioxide tension, body mass index and indices of upper airway obstruction. Sleep hypoventilation is associated with significant increases in arterial carbon dioxide tension night-to-morning, and may contribute to long-term elevations in arterial carbon dioxide tension.
This study was funded by grants from the Australian National Health and Medical Research Council and Air Liquide Australia.
Chronic obstructive pulmonary disease (COPD) is a major cause of morbidity and is now the fourth-leading cause of death in the USA 1. Chronic hypercapnia is known to be a poor prognostic factor in COPD 2 but the causes of hypercapnia are not completely understood 3. Patients with severe COPD are known to be at risk of sleep hypoventilation (SH) 4 and it has been proposed that SH may play a role in the pathogenesis of chronic hypercapnia through blunting of central chemoreceptor responses to carbon dioxide (CO2), secondary to gradual renal HCO3− retention 3, 5. In support of this hypothesis are the results from a trial of nocturnal noninvasive ventilation (NIV) in COPD, which showed improvements in awake arterial carbon dioxide tension (Pa,CO2) 6. Pa,CO2 correlated solely with reductions in transcutaneous carbon dioxide tension (Pt,CO2) during sleep.
Despite its potential importance for prognosis, no study has investigated the prevalence of SH in hypercapnic COPD. However, it is known that transient oxygen desaturation is common in severe COPD 7, and particularly in hypercapnic COPD 8. This phenomenon is at least partly due to hypoventilation, but there may also be a contribution from altered ventilation/perfusion (V′/Q′) relationships 7, 9. The mechanisms of SH in hypercapnic COPD are also uncertain. A fall in central respiratory drive, particularly in rapid-eye movement (REM) sleep, is likely to be important 7 but there may also be a significant contribution from increased upper airways resistance. “Overlap syndrome”, the combination of COPD and obstructive sleep apnoea/hypopnoea syndrome (OSAH), is said to predispose to daytime hypercapnia and hypoxaemia independently of lung function 10. However, the role of the upper airway in SH inCOPD patients without overt OSAH has received little attention. An increase in upper airways resistance at sleep onset may contribute to the fall in ventilation during sleep in normal subjects 11. It is therefore possible that patients whose respiratory muscles are working close to their endurance limits 12 cannot compensate for the extra load imposed by the upper airway during sleep.
The present study was conducted in order to address some of these questions. To do this the baseline data for the first 80patients enrolled in the Australian trial of Noninvasive Ventilation in Chronic Airflow Limitation (AVCAL) was analysed. AVCAL is a multicentre, randomised, controlled trial of chronic domiciliary NIV in stable hypercapnic COPD. Recruitment commenced in 1998. There were three main aims of the present study: 1) to document the prevalence of SH in stable hypercapnic COPD; 2) to determine which factors are associated with SH in this population with the specific hypothesis that factors linked with increased upper airways resistance during sleep (obesity, alcohol consumption, snoring and inspiratory flow limitation in sleep) would worsen hypoventilation; and 3) to determine if awake Pa,CO2 is higher in the morning than in the evening in these subjects, and if the night-to-morning change in Pa,CO2 correlates with severity of SH.
Materials and methods
Patients
All patients were undergoing baseline assessment in Adelaide, Melbourne or Sydney, Australia, for the AVCAL trial. All respiratory physicians in these three centres were contacted directly to inform them of the trial aims and design. They were provided with inclusion and exclusion criteria and requested to refer all patients they considered appropriate for assessment. Advertisements were also placed in newsletters of the Royal Australasian College of Physicians, the Thoracic Society of Australia and New Zealand and the Australasian Sleep Association.
The inclusion criteria were: 1) a clinical history consistent with stable COPD without an exacerbation of airways disease for at least 4 weeks at the time of evaluation; 2) irreversible airflow obstruction (forced expiratory volume in one second (FEV1) <1.5 L or <50% predicted, FEV1/forced vital capacity <65%, ΔFEV1 with bronchodilator <15%, or if FEV1 <1.5 L, ΔFEV1 <200 mL); 3) daytime awake Pa,CO2 >6.12 kPa (46 mmHg) while in a stable condition; and 4) on oxygen therapy for a minimum of 3 months with a prescription meeting American Thoracic Society guidelines.
The exclusion criteria were: 1) morbid obesity (body mass index (BMI) ≥40); 2) significant OSAH (apnoea/hypopnoea index (AHI) ≥20·h−1 (see scoring criteria below)); 3) previous treatment with domiciliary NIV; and 4) age ≥80 yrs.
The protocol was approved by the Research and Ethics Committees of all participating institutions. All participants gave informed written consent.
Measurements
Height and weight were measured and BMI calculated. Spirometry pre- and postbronchodilator, lung volumes and single-breath diffusion capacity of the lung for carbon monoxide were recorded. Arterial blood was drawn for blood-gas analysis while on the prescribed flow of oxygen, and after breathing room air for 15 min, if tolerated.
Full-attended polysomnography was performed (Compumedics S series, Abbotsford, Victoria, Australia) while on prescribed oxygen flow delivered by nasal cannulae. Measured parameters were electroencephalography (EEG), left and right electro-oculography, submental electromyography, airflow (by thermistor and/or by nasal pressure cannula), thoracoabdominal movement by inductance bands, leg movements, body position, and arterial oxygen saturation.
Pt,CO2 was recorded in all subjects with the TINA TCM 3 device (Radiometer, Copenhagen, Denmark) calibrated according to the manufacturers instructions. Arterial blood gases (ABG) were measured in the evening prior to sleep and again in the morning at least 5 min after final awakening, on an identical flow of oxygen while supine. Wakefulness was confirmed on the EEG. The simultaneous stable Pt,CO2 value at the time of each blood gas sample was noted.
A detailed history of lifetime alcohol consumption was taken using the “timeline followback” technique 13. This involves aiding patients' memories of their past alcohol intake by using key events in their lives, e.g. first job, armed forces service, marriage, divorce, retirement, death of spouse etc. The technique has been validated previously and gives highly reproducible results 13, 14. The information was used to calculate lifetime alcohol consumption and current consumption.
Data processing
Sleep was manually staged in 30-s epochs according to standard criteria 15. Apnoea was defined as a complete cessation of airflow for 10 s. Hypopnoea was defined as a >50% reduction in amplitude in two of the following for >10 s: airflow, thoracic and abdominal respiband signals. Because all patients were on supplemental oxygen, no desaturation criterion was used in scoring hypopnoeas. For this reason, a rather high AHI cut-off (≥20·h−1) was used in eliminating subjects with OSAH. In the Sleep Heart Health Study cohort, it was found that omitting desaturation from the criteria for scoring respiratory events produced up to a 10-fold increase in median AHI 16.
Night-to-morning ΔPa,CO2 were calculated. Pt,CO2 data were sampled at 1 Hz and mean Pt,CO2 values calculated for each epoch of sleep. Evening Pa,CO2 was used as an in vivo calibration (fig. 1⇓) to adjust for baseline offset in Pt,CO2 measurements.
Simultaneous Pt,CO2 and Pa,CO2 results were examined before and after sleep. If the Pt,CO2 minus Pa,CO2 had changed across the night, this change was attributed to Pt,CO2 sensor drift. The morning Pt,CO2 value was corrected for this drift, and all Pt,CO2 values in between were corrected using linear interpolation (fig. 1⇑).
SH was defined as >20% of total sleep time (TST) spent with “corrected” Pt,CO2 >1.33 kPa (10 mmHg) above supine waking baseline levels. Severity of SH was assessed in each subject in three ways: 1) the increment in Pt,CO2 (wakefulness–sleep) above which 20% of TST was spent (20%incr; fig. 2⇓); 2) the maximal increment of Pt,CO2 (maxincr; fig. 2⇓); and 3) REM-related hypoventilation, measured as the mean rise in Pt,CO2 from the average in the preceding 5 min of nonrapid-eye movement (NREM) sleep to the maximum in each period of REM (ΔNREM–REM). This value is less likely to be influenced by drift in the Pt,CO2 sensor as the change is calculated over a much shorter time period, usually <15 min.
If a drift of <1.33 kPa (10 mmHg) was seen from night-to-morning in the Pt,CO2 sensor, or if evening and morning Pa,CO2 were not available to validate Pt,CO2, the 20%incr and the maxincr values were excluded from analysis but the ΔNREM–REM values were retained.
Stable sleep was defined as an epoch without a stage change in the previous four epochs. A random sample of 10 nonconsecutive epochs of stable sleep in each stage was chosen by computer in each of the 54 patients in whom nasal pressure was measured. Each breath was visually scored for the presence of inspiratory flow limitation, without reference to the SH or blood-gas data for that subject, by a physician experienced in evaluation of inspiratory flow limitation from clinical data. Breaths were scaled to be of approximately uniform duration and amplitude. Evaluation was made purely on the basis of the shape of the inspiratory nasal pressure trace 17. Breaths were assigned a score of 0, 1 or 2 corresponding to no flow limitation, mildly flow-limited, or markedly flow-limited. A mean score per breath was derived for each stage of NREM sleep. These were then weighted according to the percentage of total NREM sleep occupied by that stage, and a weighted mean for all of NREM sleep was calculated. A separate mean score per breath was calculated for REM (REMflow). A blind rescoring in a random sample of 65 epochs produced a concordance of 87%.
Statistics
The correlations between lung function and blood-gas parameters, as well as reported alcohol consumption, indices of inspiratory flow limitation and polysomnographic variables with measures of SH were assessed using Pearson's product moment with Spearman's rho for non-normally distributed variables. Predictive power of combinations of variables was assessed using forward stepwise-linear regression, with criteria for entry into the model being a probability of F of <0.05 and for removal a probability of >0.1. Comparison of evening and morning ABG measurements was performed using paired t-tests. Results are given as mean±sd unless otherwise stated.
Results
Baseline data from the first 80 patients (54 male, 26 female) enrolled in the AVCAL trial were analysed. Evening and morning blood-gas samples were taken from 67 patients under satisfactory conditions as defined above. There was no systematic overestimation of ΔPa,CO2 by Pt,CO2 as ΔPa,CO2 increased (fig. 3⇓).
Amongst these 67, data on Pt,CO2 trends across the night were excluded from analysis in 13 because of excessive drift of the Pt,CO2 sensor (11 patients), very high Pt,CO2 values that were outside the range of the recording equipment (one patient) and failure to record Pt,CO2 value at the time of ABG (one patient). Data on overnight trends in Pt,CO2 were analysed in the remaining 54 patients. Their ages, baseline lung function and ABG, BMI and polysomnographic variables are shown in table 1⇓. There was no significant difference in any of the parameters in table 1⇓ between the excluded patients and those in whom SH data was retained. In these 54 subjects the mean drift in the sensor across the night was 0.21±0.65 kPa (1.6±4.9 mmHg). Figure 4⇓ is a frequency histogram of the 20%incr values. Of these 54 subjects, 23 or 43% were found to have SH according to the authors' a priori definition.
⇓Tables 2 and 3⇓ give univariate correlation coefficients between measures of SH and putative associated factors. Multiple linear-regression analysis produced the following equations: where SWS is slow-wave sleep.
Morning ABG samples were taken 33.5±30 min after final wakening. Table 4⇓ gives mean values for evening and morning Pa,CO2, pH and H+ concentration. Pa,CO2 rose by 0.70±0.65 kPa (5.29±4.92 mmHg) across the night (p<0.001). Night-to-morning ΔPa,CO2 correlated with all measures of SH (table 5⇓).
Multiple regression analysis produced the following equation:
Discussion
This study has shown that SH is common in hypercapnic COPD, occurring in 43% of the population. Severity of SH correlated with daytime Pa,CO2 and with BMI, and during REM with severity of inspiratory flow limitation and with AHI. In contrast to earlier reports in normal subjects 18, 19, the authors found a significant increase in waking carbon dioxide tension (PCO2) between night and morning in these hypercapnic COPD patients. The night-to-morning rise in Pa,CO2 was highly correlated with severity of SH. This is the first study to document the prevalence of SH in hypercapnic COPD. A number of studies have examined the prevalence of REM-related oxygen desaturation in sleep in COPD 8, 20, 21. However these desaturations may be partially due to V′/Q′ disturbance 7, 9. One previous study measured Pt,CO2 during sleep in 23 COPD patients and 33 normal controls 22. Mean maximal increase in Pt,CO2 was 0.80 kPa (6 mmHg) in the COPD patients and was not different to the controls. These results differ from those in the current study, probably because few of the patients in the study by Midgren and Hansson 22 were hypercapnic (Pa,CO2 5.59±0.80 kPa (42±6 mmHg)). In the present study, there was a strong correlation in the subjects between daytime Pa,CO2 and severity of SH. Previously, hypercapnia has been shown to be a risk factor for REM-related desaturation in COPD 8, 21, 23.
OSAH appears be an important cause of hypercapnia in some patient groups 10, 14, 24. Continuous positive airways pressure has been shown to restore eucapnia and increase ventilatory response to CO2 in morbidly obese hypercapnic patients with OSAH 24. Hypercapnia that appears disproportionate to level of lung function impairment has been associated with the presence of obstructive sleep apnoea (OSA) in COPD patients 10. Chan et al. 14 showed that hypercapnic COPD patients had many more sleep-disordered breathing events, had higher BMIs and smaller upper airway cross-sectional areas than eucapnic controls matched for lung function.
In the current study, the hypothesis that lesser degrees of upper airways obstruction, other than OSAH, could contribute to sleep hypoventilation in severe COPD was addressed. Even after exclusion of patients with OSAH, severity of inspiratory flow limitation in REM, and AHI in both REM and NREM sleep were found to be predictive of REM-related hypoventilation. BMI was significantly correlated with SH across the whole night.
A surprising finding was an inverse correlation with alcohol consumption. This negative correlation seems counterintuitive given what is known about the acute effects of alcohol in sleep, especially in males 25, 26. These findings could not be explained by a reduction in REM sleep due to chronic alcohol abuse. It is possible that heavy drinkers who were susceptible to the effects of alcohol on sleep-disordered breathing were excluded from the study due to a diagnosis ofOSAH, leaving a “survivor population” who either drank minimal alcohol or were not susceptible to sleep-disordered breathing. Ultimately, the authors do not have a definite explanation for this finding, which may be due to chance.
Another interesting result was the finding that %SWS wasan independent predictor of NREM-to-REM change in Pt,CO2. This may also be a chance finding but may relate to the fact that the upper airway is more stable in SWS than in lighter NREM stages; in REM it is at its most vulnerable to collapse 27. Therefore the transition from SWS to REM would be more likely to produce the greatest change in ventilation.
Daytime Pa,CO2 was the factor most strongly associated with both 20%incr and maxincr. This is consistent with a number of studies that have shown daytime hypercapnia and awake ventilatory response to CO2 correlated with propensity to REM-related desaturation in COPD 8, 20, 23. Correlations cannot prove causation, and even if the factors are causally related, they do not establish direction of causation. It is possible that worsening daytime hypercapnia is the result, not the cause of SH. The largest study of predictors of hypercapnia in COPD included BMI in the final regression equation for CO2 28. While obesity may influence daytime Pa,CO2 via mass loading of the chest wall, both in wakefulness and sleep, it may also increase upper airway resistance in sleep, worsening SH and blunting chemosensitivity.
The authors found that Pa,CO2 remained elevated over evening levels 33.5±30 min after final wakening. They believe this is the first report of such a change in the literature. This contrasts with normal subjects who show no difference in PCO2 before and after sleep 18, 19. ΔPa,CO2 was highly correlated with severity of SH (table 3⇑) and the combination of 20%incr and age explained 66% of the variance. The authors believe this provides supporting evidence for the theory that nocturnal hypoventilation can influence daytime Pa,CO2 in severe COPD.
It has been proposed that SH contributes to development of chronic hypercapnia in COPD through the mechanism of gradual renal HCO3− retention. Night-to-morning changes in HCO3− observed in the present population were compared with the change predicted for acute on chronic hypercapnia in published equations 29, but the authors were unable to demonstrate any renal HCO3− retention across the night due to SH. Therefore, if little or no renal compensation occurred, why did these patients not return to their presleep ventilatory status on wakening? There are a number of possible explanations for this. First, renal HCO3− retention may have occurred in the subjects but was too small to be detected because of thewide confidence intervals of the predicted response 29. Second, compensation may have occurred in the cerebrospinal fluid overnight. With the fall in Pa,CO2 on awakening, the cerebrospinal fluid would become relatively alkalotic, increasing pH in the environment of the medullary chemoreceptors and so reducing respiratory drive. A third possible explanation for the results is that acute hypercapnia has significant negative effects on the ability of the respiratory muscles to generate force and predisposes them to fatigue 30. Therefore, it may be that despite similar respiratory centre output, the mechanical response of the muscles is temporarily impaired in the first few hours after waking in patients with significant SH. Finally, there may be a contribution from the normal circadian rhythms of hypercapnic ventilatory response and end-tidal CO2 31.
Methodological considerations
Study population
As with all clinical trials the study population consisted only of those subjects who were both referred for assessment and consented to be enrolled in the AVCAL trial. This process always includes some referral bias. However, the authors attempted to avoid studying a population only referred for sleep assessment as much as possible.
Patients who were morbidly obese (BMI ≥40) were excluded and the mean BMI of the study population was 25.7±5.7 kg·m−2. In addition patients with concomitant OSA were excluded. Mean AHI of the study population was 4.7±5.0. The lowest quartile of median AHI scores in the 5,046 participants in the Sleep Heart Health Study 16 had an AHI up to 18.3 when no desaturation criterion was included in the definition of hypopnoea, compared with the lowest quartile of up to AHI 3.7, when using a definition also requiring 3% desaturation. The authors believe therefore that using an AHI cut-off of 20 to exclude patients with OSA was justified.
From 80 patients studied, 26 were excluded for technical reasons. However, there was no difference in any of the baseline characteristics recorded between the excluded subjects and those whose data was retained. The authors believe that the group studied was truly representative of the general hypercapnic COPD population.
Use of supplemental oxygen
Supplemental oxygen may worsen hypercapnia during sleep in COPD 32. Therefore, it is possible that SH would have been less severe if these patients had been studied without oxygen. However all subjects were studied on their usual prescribed oxygen flow. Therefore, the authors believe the data more truly represents SH in the clinical situation. In addition, they felt that if severe desaturation was seen, or if the patient became very dyspnoeic during the night it would have been ethically unacceptable to withhold oxygen from patients for whom it was part of their usual therapy. If oxygen were recommenced part way through the study this would have invalidated Pt,CO2 and Pa,CO2 measurements, as any changes could have been due to reintroduction of oxygen.
Transcutaneous carbon dioxide tension measurement
SH has been defined as an abnormal increase in Pa,CO2 during sleep 4. Diagnostic criteria proposed by this task force required an increase in Pa,CO2 >1.33 kPa (10 mmHg) from awake supine values during sleep or oxygen desaturation during sleep not explained by apnoea or hypopnoea events.
For the reasons elaborated above, the authors felt they could not withhold oxygen from these patients during their polysomnographs. Furthermore, comparison of severity of hypoventilation using oxygen saturation is difficult when patients begin at different points on the oxyhaemoglobin desaturation curve. Finally, desaturation is a nonspecific finding, which may also be caused by change in lung or closing volume with deterioration in V′/Q′ inequality 4, 7. Therefore, oxygen saturation was not used as an index of SH.
Insertion of an indwelling arterial line to measure Pa,CO2 could not be justified in this clinical trial and was likely to deter patients from enrolling. It is accepted that Pt,CO2 is an imperfect measure of Pa,CO2. There are many reasons for the differences in the two values when measured simultaneously 33 and commercial Pt,CO2 monitors incorporate software tocorrect for some of these factors. A number of studies in adults have found high correlations with simultaneous Pa,CO2 33–35. Control of sleep-related increases in Pt,CO2 has also been found to be a predictor of improvement in daytime Pa,CO2 with noninvasive positive-pressure ventilation in COPD 6. Nevertheless, two studies in the setting of SH have found large discrepancies between Pt,CO2 and simultaneously measured Pa,CO2 36, 37. Neither of these studies used the approach of in vivo calibration of the device, though the study of Rosner et al. 37 suggested accuracy and precision would be improved if this were performed. After in vivo calibration the question becomes whether Pt,CO2 accurately reflects change in Pa,CO2. Some studies have found a systematic increase in Pt,CO2–Pa,CO2 as Pa,CO2 increases 34, 37 although others have found no such bias 35, 36. There was no evidence of a systematic overestimation of ΔPa,CO2 by Pt,CO2 as ΔPa,CO2 increased in these data (fig. 3⇑). The authors therefore concluded that changes in Pt,CO2–Pa,CO2 across the night were due to sensor drift. Afterexcluding from analysis subjects in whom drift was >1.33 kPa (10 mmHg), the authors corrected for drift using the morning Pa,CO2 sample. The mean correction introduced was 0.21±0.65 kPa (1.6±4.9 mmHg).
Using these methods, SH, as measured by Pt,CO2 was highly correlated with night-to-morning change in Pa,CO2. Thus, while the authors cannot be certain of the precision of the Pt,CO2 measurements, they believe they are likely to accurately represent trends in Pa,CO2 across the night.⇑
Definition of sleep hypoventilation
There is no universally accepted definition of SH. The recent consensus conference 4 chose a threshold value of an increase in Pa,CO2 >1.33 kPa (10 mmHg) during sleep, without specifying a duration for this increase. The authors therefore chose a threshold increase of 1.33 kPa (10 mmHg) in Pt,CO2. It was felt that 20% of TST constituted a significant portion of the night spent with abnormally elevated Pt,CO2 levels. However, it is clear that the prevalence of SH will vary depending on the definition used. Figure 4⇑ illustrates how prevalence changes depending on the increment in Pt,CO2, which is defined as significant. Likewise, if SH is defined as >10% of TST spent at >1.33 kPa (10 mmHg) above baseline waking levels, the prevalence in this population would be 48%, if >40% of TST the prevalence would be 31%.
To conclude, the authors have found a high prevalence of sleep hypoventilation in a large hypercapnic chronic obstructive pulmonary disease population. Sleep hypoventilation was primarily related to baseline arterial carbon dioxide tension, body mass index and, in rapid-eye movement sleep to increased upper airway resistance. These patients often have increases in arterial carbon dioxide tension between night and morning, which are strongly related to the severity of sleep hypoventilation. Further studies are needed to investigate the long-term influence of repetitive increases in carbon dioxide due to sleep hypoventilation on respiratory drive and daytime blood gases in these patients.
Acknowledgments
The complete list of AVCAL investigators. Adelaide: R.D. McEvoy, F.J. O'Donoghue, P.G. Catcheside, L.S. Rowland, R. Antic, H.W. Greville, R.E. Ruffin, A.M. Southcott. Sydney: R.R. Grunstein, D.J. Barnes, I.H.Young, E.E. Ellis, E.R. Collins. Melbourne: R.J. Pierce, K. Detering, D.A. Campbell, L.B. Irving, C.F.McDonald, C.J. Worsnop, L.M. Schachter, S.E. Rochford.
- Received July 24, 2002.
- Accepted January 15, 2003.
- © ERS Journals Ltd