INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Recent data report a decline in mortality from coronary artery disease (CAD) attributed to improvements in treatment of CAD as well as effectiveness of primary and secondary prevention models. However, the mortality rate remains high in this group of patients (1). Case-control studies have demonstrated an independent association between obstructive sleep apnea (OSA) and CAD (2, 3). To our knowledge, studies addressing secondary prevention in patients with CAD (4) have considered obesity and hypertension, which are both common in OSA (5, 6), whereas OSA per se has yet not been adjusted for. Moreover, little is known about the effect of secondary prevention models on CAD when elderly subjects are included in the analysis.

Retrospective studies suggest that crude mortality is increased in untreated patients with OSA (7), which may be due to coexisting cardiovascular morbidity (8). Continuous positive airway pressure (CPAP), which is recognized as a major treatment modality in OSA (9), not only reduces daytime sleepiness and improves cognitive function (10) but also has beneficial effects on hypertension (11), cardiac failure (12), and nocturnal angina (13) in OSA patients. In addition, long-term CPAP treatment has been shown to reduce the need for acute hospital admission due to concomitant cardiovascular and pulmonary disease in OSA patients (14). However, little is known about the long-term effects of CPAP on cardiovascular mortality in OSA.

We have previously demonstrated a high prevalence of OSA, defined as a respiratory disturbance index (RDI) of 10 events/h or more, among consecutively selected patients with CAD requiring intensive care compared with age-, gender-, and body-mass index (BMI)-matched control subjects (3). In the present, 5-yr follow-up study, we aimed to prospectively investigate cardiovascular mortality in the CAD patients with versus without concomitant OSA. The predictive value of RDI on mortality was addressed after statistical adjustment for traditionally recognized risk factors, thereby introducing RDI as an additional factor in the evaluation of secondary prevention models in CAD.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Population

This prospective study was undertaken in consecutive patients with CAD requiring intensive care at the Department of Cardiology, County Hospital of Skaraborg, Skövde, Sweden during a 6-wk period in 1993. The selection criteria for CAD have been described in detail elsewhere (3). In short, among 97 eligible patients with CAD, 81 (83.5%) completed a self-administered questionnaire concerning symptoms associated with OSA, and 62 subjects (63.9%) underwent an overnight sleep study. The sleep questionnaire included questions with yes/no alternatives concerning history of snoring, witnessed apneas, sleep fragmentation, nocturia, and night-sweating as well as excessive daytime tiredness before the current hospitalization period. Reasons for lacking overnight sleep study were early death during the intensive care period (n = 12), development of severe cardiac failure or stroke at the study start or after discharge from the hospital (n = 14), or patient refusal (n = 9). Detailed patient characteristics and sleep questionnaire data are shown in Tables 1 and 2. Three patients with OSA who had received CPAP therapy during the follow-up period were excluded from the final analysis (see RESULTS).

Basal Diagnostic Procedure for OSA

The overnight sleep study was performed in the sleep laboratory of the Department of Otorhinolaryngology at the same hospital. The patients were investigated during a stable clinical condition (functional class I or II, New York Heart Association [NYHA] Classification) (15) and none of them had myocardial infarction, stroke, or hospital admission resulting from cardiac failure within 4 mo of the sleep study. Alcohol consumption or intake of sedative medication was not allowed within 48 h of the investigation. The time interval between hospitalization in the intensive care unit (ICU) and the overnight sleep study ranged between 4 and 21 mo.

The sleep study, lasting at least 6 h after sleep onset, included nocturnal recording of respiratory and body movements and ballistocardiogram (BCG) via a static charge-sensitive bed (SCSB; Bio-matt, Biorec Oy, Turku, Finland) as well as monitoring of arterial oxygen saturation (Sa_O2) via a finger probe (Biox 3740; Ohmeda, Boulder, CO). All signals were continuously sampled, displayed online and stored in a Computer (BR11 software; Biorec Oy, Helsinki, Finland). Increasing amplitude of respiratory movements with simultaneous increase in respiratory variation of BCG and a subsequent body movement (arousal movement) was recorded as an obstructive event according to the definition by Alihanka (16). Central apneas were recorded when the amplitude of respiratory movements decreased with a simultaneous decrease in variation of BCG (17). A decrease in oxygen saturation of 4% or more immediately after baseline was defined as a significant desaturation. The diagnosis of OSA was based on a RDI, defined as the average number of obstructive events with significant desaturations per hour of supervised estimated sleep, of 10 or more. RDI was scored manually by the same physician (S.L.). Of seven patients demonstrating a combination of central and obstructive apneas, only two were qualified for the diagnosis of OSA according to the RDI definition used previously (Table 2).

Definition of Traditional Risk Factors at Baseline

Clinical characteristics and risk variables at baseline in the CAD patients with OSA (n = 16) versus without OSA (n = 43) are shown in Table 2. Overweight was defined as a BMI  25 but < 30 and obesity as BMI  30 kg/m² (18). Hypertension was defined as ongoing pharmacologic antihypertensive treatment and/or a recorded systolic blood pressure (BP)  160 mm Hg and/or diastolic BP  95 mm Hg, measured on at least three different days (19). Diabetes mellitus was diagnosed when subjects were receiving insulin or oral antidiabetic drugs, or with a fasting blood glucose concentration > 6.7 mmol/L at three separate occasions. Three different cutoff levels of serum total cholesterol concentration were used (see Table 2) adapting a definition of moderate or severe hypercholesterolemia (20). The subjects were classed as current smokers, former smokers (those who had stopped smoking at least 6 mo before the study inclusion), and those who had never smoked.

Observation Period and Follow-up Procedure

A time span of 60 mo after discharge from the ICU was defined as the observation period. The CAD patients with OSA were referred to the Respiratory Division, Department of Medicine, for initiation of CPAP treatment according to the clinical routines, which have been described in detail elsewhere (14). In short, CPAP therapy was initiated depending upon severity of OSA and daytime symptoms of the patients as well as acceptance of the device at the time of the CPAP titration. The final decision on CPAP treatment was made by pulmonologists without knowledge of the ongoing evaluation of outcome measures in this particular population of CAD patients with concomitant OSA. All patients in this study were followed according to the clinical routines, including secondary prevention models in CAD such as pharmacological treatment of hypertension, hypercholesterolemia, and/or diabetes mellitus. In addition, dietary advice was given to the subjects with overweight or obesity as well as diabetes mellitus and/or hypercholesterolemia. The smokers were advised to stop smoking.

Surgical Intervention and Long-term Pharmacological Treatment

Data collection on surgical intervention for CAD such as percutaneous transluminal coronary angioplasty or coronary artery bypass grafting as well as prescription of pharmacologic agents for treatment of cardiovascular disease (International Classification of Diseases, ICD-9 codes 401 to 435) (21) was based on reviews of all clinic charts and confirmed by direct patient interviews (Table 3). At least 1 yr of medication during the 5-yr follow-up was defined as long-term pharmacologic treatment. Registered drugs include those listed within the Anatomical Therapeutic Chemical (ATC) classification system codes C01-C10 (22).

Cardiovascular Mortality

The time and the cause of death were obtained from hospital records and the National Cause of Death Registry in Sweden as well as telephone interviews with the closest relative listed in the charts. Death was attributed to a cardiovascular origin in the case of documentation of significant arrhythmias, cardiac arrest, cardiac failure, myocardial infarction, or stroke.

Statistics

The groups were compared using Student's t test for variables measured on a continuous scale, and where appropriate, Fisher exact test (two-tailed) or chi-square test for categorical variables. To test the correlation between different variables and incidence of death, a special univariate survival test (23) was applied. The following variables were analyzed: RDI, OSA, central sleep apnea, age at baseline, gender, BMI at baseline, acute myocardial infarction (MI) at baseline, history of MI, hypertension, diabetes mellitus, smoking history, current smoking at baseline, hypercholesterolemia, serum total cholesterol at baseline, history of snoring, excessive daytime tiredness (alone and in combination with snoring), witnessed apneas, sleep fragmentation, night sweating and nocturia, time interval (months) between discharge from the ICU and the overnight sleep study, surgical intervention for CAD as well as pharmacologic treatment (for specification of drugs, see Table 3). The significant variables were then included in a Cox multiple conditional regression model by forward stepwise method. In addition, the hazard function of death was estimated by use of a Poisson model depending on RDI, current age, and time since intensive care (24). Truncation to the left was applied when calculating the hazard function at the time of the measurement of RDI (a priori nobody can have a value of RDI measured after the death).

Continuous values are given as means ± SD. A p value (two-sided) of 0.05 or less was regarded as statistically significant.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

As shown in Table 1, patients who were not studied because of early death or severe cardiac failure were approximately 5 yr older than the final study population (p = 0.015) but there was no significant difference between the two groups regarding BMI and sex distribution. Neither age, BMI, and sex distribution nor subjective symptoms including history of snoring, excessive daytime tiredness (alone and in combination with snoring), witnessed apneas, sleep fragmentation, nocturia, and night sweating differed significantly between the final study population and the subgroup of patients who refused participation in the sleep study (Table 1). Overnight sleep recordings revealed OSA in 19 CAD patients (30.7%). However, three subjects (age 63, 51, and 46 yr, respectively) with an RDI of 30, 11, and 30/h, respectively, had complied (elapsed runtime counter) with the CPAP treatment (55, 37, and 36 mo, respectively) and were alive at the end of the follow-up period. These patients were excluded from further analysis as they had no apneas on CPAP treatment.

As shown in Table 2, CAD patients with OSA were older than those without OSA. However, the relative proportions of age  65 or  75 yr, sex, overweight or obesity, hypercholesterolemia, diabetes mellitus, smoking history, hypertension, or history of myocardial infarction at baseline were not different between the groups. Occurrence of concomitant central apnea at the time of the overnight sleep study did not differ significantly between the groups. No patient demonstrated dominating central sleep apnea or Cheyne-Stokes respiration. Mean RDI was approximately three times higher in the CAD patients with OSA (range 10 to 35) compared with those without OSA (range 0-9) (p < 0.0001, Table 2).

During the follow-up period, surgical intervention for CAD was carried through in two OSA subjects compared with nine in the non-OSA group (NS). Long-term treatment with nitrates, calcium antagonists, angiotensin converting enzyme inhibitors, diuretics, digoxin, and/or lipid-lowering agents tended to be more common in the OSA group, whereas beta blocker use was more frequent among non-OSA patients. However, only the usage of diuretics differed significantly between the groups (Table 3).

During the observation period, neither myocardial infarction nor stroke incidence differed significantly between OSA and non-OSA patients, whereas mortality rate was significantly higher in the OSA group (Table 4). As shown in Table 5, the cause of death was considered as cardiovascular in all subjects. One patient died suddenly at home during the night after initial symptoms with fever and dyspnea. No autopsy was performed. Time of death was not related with the normal sleeping period in the majority of the deceased patients (Table 5). Three of six deceased OSA subjects were on the waiting list for initiation of CPAP treatment, the remaining three were not referred owing to low degree of OSA and lack of daytime symptoms. One additional subject in the non-OSA group with a RDI of 7 was on the waiting-list for CPAP prescription. This patient suffered from excessive daytime sleepiness and obstructive events although the number did not qualify for a diagnosis of OSA at the time of the overnight sleep study. Myocardial infarction and cardiac failure were the most frequent causes of death in the CAD patients with concomitant OSA.

As shown in Table 6, univariate predictors of cardiovascular mortality were identified as RDI, OSA, age, hypertension at baseline, history of never-smoking, and digoxin treatment during the observation period. As OSA turned out not to contribute more than RDI to the prediction of death, OSA was removed, and RDI alone was entered into the subsequent Cox multiple conditional regression model. Thus, in this forward stepwise procedure, RDI remained as the only significant independent predictor of cardiovascular mortality (exp  = 1.13, 95% confidence interval [CI] 1.05 to 1.21, p < 0.001). In the Poisson model, after 3 yr assuming an age of 70 yr, increasing RDI was associated with an increased risk of death (Figure 1).

View larger version (14K): [in this window] [in a new window]   Figure 1.   By use of a Poisson model the death hazard was calculated as a function of RDI, current age, and time elapsed after the intensive care episode for CAD. The bolded curve gives the function at the current age 70 yr and 3 yr after intensive care. The dotted curves represent 95% CI.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

To our knowledge, this study is the first prospective investigation of the impact of OSA on cardiovascular mortality, which includes elderly subjects, and which introduces RDI in a secondary prevention model in a previously hospitalized group of patients with CAD. The patients were consecutively recruited based on a diagnosis of angina pectoris and/or myocardial infarction and treatment in the cardiac ICU. Of 97 initially eligible patients with a cardiovascular event leading to hospitalization, approximately two-thirds remained to be studied in the sleep laboratory. Although this may have led to a negative bias resulting from unintentional exclusion of the subjects with the most severe CAD, the approach was selected to avoid inclusion of patients with congestive heart failure (CHF) and thereby potentially frequent central apneas (2). We lack complete echocardiographic data regarding left ventricular dimension and ejection fraction. Available data included measurements obtained during the intensive care period and consequently therefore considered not to be representative of cardiac structure and function at the time for the sleep study (4 to 21 mo after discharge from the ICU).

As previously shown by Polo and coworkers (25), the SCSB method to record respiratory and body movements in combination with the monitoring of a BCG and Sa_O2 has a high sensitivity (0.92 to 0.98) for detection of OSA although the specificity is lower (0.61 to 0.68). However, the false-positive findings were not considered as a severe methodological limitation as they were explained by frequent episodes of hypopnea accompanied by arterial oxygen desaturation and arousal (25). This method was also shown to have a high validity in discriminating between central apneas or Cheyne-Stokes respiration (CSR) and obstructive apneas (17, 25). Despite the fact that seven patients demonstrated a combination of central apneas, only two were classed with OSA according to the RDI definition used previously. At the time of the sleep study, the CAD patients with OSA tended to use more cardiac medications (significant for diuretics) than the patients without OSA, suggesting a possible relationship between OSA and CHF in CAD patients. Regarding the functional class of the patients according to NYHA criteria (I-II), as well as an absence of dominating central apneas, these drugs may also reflect an optimal treatment of CHF in this population at the time of the sleep study.

In previous studies addressing mortality in patients with OSA, increased crude mortality was found by some researchers (7, 8) but not by others (26). Moreover, in the elderly, there was no association between apnea-hypopnea index (AHI) and mortality in three prospective studies (27), whereas in another study there was a significant association, but only in women (30). Explanations for this discrepancy may include the different methods used in overnight sleep studies and different cutoff levels of AHI or RDI that were applied. However, impact of OSA on mortality in epidemiologic studies or in sleep clinic cohorts may not necessarily reflect the situation in particular populations, such as CAD patients who already are at high risk of mortality (1). In our original case-control study addressing the association between OSA and CAD, we considered a RDI cutoff level of 10 for the diagnosis of OSA (3). When applying the same cutoff value in the present study of the CAD patients, there was an increased cardiovascular mortality in the OSA group compared with non-OSA patients at the time of the sleep study. Even if the mean age of the OSA group was higher, there was a similar proportion of elderly subjects in both groups. Moreover, age was statistically adjusted for when investigating the predictors of mortality in this particular population. Finally, there was no significant difference between these two groups regarding gender, obesity, smoking history, hypertension, hypercholesterolemia, diabetes mellitus, or history of myocardial infarction at the time of the study start. Although the time interval between the discharge from the hospital and overnight sleep study varied in the patients, it was not found to be significantly predictive of either RDI or mortality in the statistical analysis. The possible impact of surgical intervention and concomitant pharmacologic treatment on mortality was also considered. As discussed previously, the use of pharmacologic agents for CAD and cardiac failure tended to be more common in OSA patients, potentially suggesting more pronounced sickness and disability. However, usage of any of these agents was not found to be significantly predictive of mortality in multivariate analysis.

The cause of death was considered as cardiovascular in all subjects based on the clinical diagnosis. It may be argued that autopsy might have led to a more distinct classification of the cause of death in at least some cases but was not advocated for ethical reasons in accordance with the viewpoint of the closest relatives. Myocardial infarction and cardiac failure were the most frequent causes of death. Although death may have been expected to occur during sleep, at least in the some patients (31), any such association may have been masked by the low absolute number of the deceased subjects.

In a statistical analysis of variables associated with CAD in this population, we found a significant association between cardiovascular mortality and RDI, OSA, age, hypertension, never-smoking history at baseline as well as digoxin treatment during the follow-up period. No association was found between cardiovascular mortality on the one hand and gender, BMI, diabetes mellitus, and hypercholesterolemia at baseline on the other. By including both RDI and OSA simultaneously in the regression model, we were able to show that RDI provided a stronger contribution than OSA to the prediction of mortality. OSA was therefore removed in the final multivariate analysis.

As the present study focused mainly on the impact of OSA on cardiovascular mortality, we lack complete data about the changes in profile of traditionally recognized risk factors during the follow-up period. However, subjects with obesity and diabetes mellitus as well as those with hypercholesterolemia received dietary advice. It is noteworthy that, in the latter group, only four subjects were on treatment with lipid-lowering drugs during the observation period, which reflects the treatment routine that prevailed at the time of the study start (1993). Somewhat surprisingly, we found an increased mortality among the never-smokers during the follow-up period. This observation may potentially be explained by the fact that eight of the 18 current smokers stopped smoking and six of them underwent coronary artery bypass grafting during the follow-up period, which might have favored the prognosis of current smokers. Hypertension at baseline was also found to be a predictor of mortality in the univariate analysis but not significantly so in the multivariate analysis. This lack of significance may have been influenced by the fact that pharmacologic agents, such as nitrates, beta-blockers, calcium antagonists, and angiotensin converting enzyme inhibitors, were widely used to treat CAD and/or CAD-related complications (e.g., cardiac failure). The resulting blood pressure reduction may have caused a lower than expected prevalence of hypertension in the study population. Increased risk of mortality in patients on digoxin treatment in the univariate analysis could be a parallel phenomenon reflecting the severity of the cardiac failure rather than a causal relationship. Moreover, there was no significant association between digoxin treatment and mortality in the multivariate analysis.

Seppälä and coworkers reported an increased cardiovascular mortality among snorers (31). However, a more recent prospective epidemiologic study by Lindberg and coworkers identified a predictive value of snoring on cardiovascular mortality only when a history of excessive daytime sleepiness (EDS) was combined with snoring (32). In the present smaller study, we demonstrated no association between snoring alone or in combination with daytime tiredness and cardiovascular mortality. However, symptom scores traditionally used to identify symptomatic OSA are likely to be less useful in a particular population of CAD patients compared with the general population for several reasons. Most importantly, symptoms suggestive of excessive daytime tiredness or sleepiness may be related to CAD itself. Additionally, pharmacologic treatment such as beta-blockers, used in patients with CAD may increase the frequency of such symptoms.

In the present study, only three of the 19 patients with OSA were on CPAP treatment during the observation period. The low number of treated patients was partly due to the time interval between the initial hospitalization at the ICU and the overnight sleep study as well as the time interval between the diagnostic sleep study and initiation of CPAP treatment. However, CPAP therapy was initiated depending upon severity of OSA and daytime symptoms of the patients as well as acceptance of the CPAP device at the time of the CPAP titration. In fact, approximately one-third of the OSA subjects were asymptomatic. This issue raises questions regarding the indications for treatment of OSA in this particular population of CAD subjects. CPAP therapy is recognized as a major treatment modality in OSA to improve daytime function (10). Moreover, beneficial effects have been demonstrated on hypertension (11), cardiac failure (12), and nocturnal angina (13) in patients with OSA. These beneficial effects of CPAP may not be directly extrapolated to a population of CAD patients in view of the high proportion of asymptomatic subjects. A number of studies (33) have demonstrated a reduced acceptance of CPAP therapy in OSA patients without significant daytime hypersomnolence.

What level of RDI should be considered as pathologic in CAD patients, especially when including elderly subjects into analysis? Previous reports suggest that nocturnal respiratory events are common in an elderly healthy population (34) and lack significant impact on mortality (27). The situation appears to be different in elderly patients with CAD. Defined by the conventional RDI cutoff level of 10, the diagnosis of OSA was found to significantly predict mortality. However, RDI was found to be an even stronger predictor than OSA. RDI was therefore entered into the multivariate analysis. The stronger predictive value of RDI could potentially suggest that an RDI below 10 had a prognostic value for the identification of individuals at risk in this particular population.

We conclude that untreated OSA is associated with an increased risk of cardiovascular mortality in patients with CAD. Furthermore, it appears appropriate that RDI is taken into consideration when evaluating secondary prevention models in CAD with regard to cardiovascular mortality.