INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Sleep-related breathing disorder (SRBD) is common in the middle-aged population (1, 2). In spite of a number of reports suggesting an increased prevalence of cardiovascular disease in patients with SRBD (3), the exact influence of SRBD on blood pressure and hypertension development remains disputed (4). Former studies have shown that blood pressure regulation is blunted in SRBD patients compared with control subjects (5). In addition, experimental data from animal studies have demonstrated a strong relationship between repetitive obstructive breathing during sleep and systemic and pulmonary blood pressure changes (9). These changes occurred during sleep and during the daytime, thereby indicating a causative role of SRBD in the development of sustained diurnal hypertension (10).

Epidemiological studies have reported a high prevalence of SRBD in hypertensives (11, 12). Between 50 to 80% of patients with SRBD were found to be hypertensive (2, 13), but the causative association was lost when confounders such as age and obesity were taken into account (14). However, recently published data from a population-based study showed a significant influence of SRBD on daytime blood pressure and the risk for systemic hypertension independent of age and body mass index (BMI) (18). This association is, however, likely to be affected not only by anthropometric data, but also by the methods used for assessment of blood pressure, the definition of the SRBD, and the size of the study sample (Table 1). Moreover, it remains unknown whether blood pressure values obtained immediately before or after sleep study are comparable to daytime blood pressure for the assessment of cardiovascular risk.

The present study was undertaken to investigate the influence of different degrees of SRBD and recognized cardiovascular risk factors on daytime office blood pressure and the prevalence of systemic hypertension in a large sleep study sample. In contrast to previous similar studies, daytime blood gases and cholesterol levels were included in the analysis and blood pressure was recorded in the late morning hours.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

The study sample consisted of 1,642 patients successively seen in the outpatient department of the Marburg Sleep Disorders Center from January 1, 1989 to April 30, 1992. Patients were referred because of clinical symptoms and signs suggestive of SRBD. From the original sample, 452 patients with missing data (blood gas analysis, blood pressure recordings, cardiovascular risk factors, technically inadequate sleep recording, and/or incomplete medical history data) were excluded. The final analysis was performed on a data set of 1,087 men and 103 women (age, 50.2 ± 10.2 yr; BMI 29.2 ± 4.9 kg/m²; respiratory disturbance index [RDI], 25.8 ± 24.6 n/h) (Figure 1).

View larger version (21K): [in this window] [in a new window]   Figure 1.   Patient log showing the study cohort and the different subgroups.

Protocol

Age, sex, race, and BMI were recorded for each patient at study entry. All patients completed a validated questionnaire (19) and underwent a standard medical interview. The presence or absence of commonly encountered SRBD-related symptoms such as snoring, witnessed apneas, excessive daytime sleepiness, insomnia, and falling asleep while driving a car were documented. Smoking habits and current alcohol consumption were recorded. The medical history included known clinical diagnoses, in particular known and/or treated systemic hypertension, and previous drug treatment. Blood samples were obtained and analyzed for determination of total cholesterol. A blood gas analysis (samples obtained at noon) was performed using a Radiometer gas analyzer (Radiometer, Copenhagen, Denmark). Blood pressure readings were obtained with the patient in a sitting position after a minimum of 10 min rest, between 9:00 and 11:00 A.M., using the World Health Organization standard protocol (20). Recordings were made on two consecutive days in the outpatient department. The second reading was used for further analysis in order to avoid the "white coat" effect (21). Heart rate was determined by pulse wave palpation.

All patients underwent unattended home monitoring of nocturnal breathing on two consecutive nights using the MESAM 4 device (MAP; München, Germany). The MESAM system was applied in the afternoon for a recording span between 6:00 P.M. and 8:00 A.M. Time of going to bed, lights out, final awakening, longer periods of sleep interruption, and estimated sleep time were assessed using a patient diary. The first night was an adaptation night and RDI was obtained from the second night. Only when the recording was technically insufficient (< 10% of recordings) or the subjective sleep time was < 5 h (< 1% of recordings) was RDI calculated from the first night.

MESAM 4 Device

The MESAM 4 device (22) records oxygen saturation using finger pulse oximetry, snoring using an electric subminiature microphone placed over the larynx, beat-to-beat heart rate analysis, and body position using a circular sensor taped just below the sternum.

Data Processing

Sleep-disordered breathing events were determined using the previously described method of MESAM 4 evaluation (24). Significant events of sleep-disordered breathing were scored only with a concomitant desaturation of  4% from baseline. Estimated sleep time was determined based on the information from the sleep diary (time going to bed, lights out, final awakening, lights on, longer periods of sleep interruption). Subsequently, the MESAM scorer edited this information by checking the heart rate (significant drop and reduced variability when going to bed, abrupt increase, and plateau after final awakening) and body position signal (e.g., change from upright to supine position). Events were counted, resulting in an RDI based on estimated sleep duration. All evaluations were performed by experienced sleep technicians.

The study cohort was subdivided into five RDI classes in order to permit analysis of the relationship between blood pressure, systemic hypertension, and different degrees of SRBD. The lowest RDI class (< 5) was defined as non-SRBD in accordance with what has been described elsewhere in epidemiological research (1).

A positive history of "hypertension" (Group 1, n = 599) was defined as previously diagnosed hypertension and/or previously initiated antihypertensive drug treatment. Patients free from a history of hypertension or with no previously documented treatment of hypertension were classified as "no hypertension" (Group 2, n = 591); see also Figure 1. In Group 2 we analyzed the current blood pressure status. Because of the different definitions of hypertension used in epidemiological research, two different cutoff points were applied in the subsequent analysis, patients with blood pressure  140/90 mm Hg (Classification 1, n = 363 newly diagnosed hypertensives) and patients with blood pressure  160/90 mm Hg (Classification 2, n = 147 newly diagnosed hypertensives).

Statistics

Results are given as mean ± SD. Statistical analysis was performed using the SPSS software program (SPSS for Windows 7.5; SPSS, Chicago, IL). All reported p values are two-tailed. Pearson's correlation coefficient was used. Evaluation of the independent impact of SRBD on blood pressure and heart rate corrected for possible confounders was performed with stepwise linear regression analysis after explication of data distribution using the Kolmogorov-Smirnov test and curve estimation models. RDI, age, BMI, sex, plasma cholesterol level, current mean nicotine/alcohol consumption, daytime PCO₂ and PO₂ were allowed as possible confounding variables in the stepwise model. Model fit was analyzed by coefficient of determination (R²) test and residual plots. Significance was tested using the Student's t test. Because of the interaction between blood pressure, heart rate, and antihypertensive medication, stepwise linear regression was only performed in Group 2 with no history of antihypertensive drug use (n = 591).

Multiple logistic regression analysis was performed with "hypertension" (Classifications 1 and 2) as the dependent variable. This permitted inclusion of all patients in the analysis, because hypertension as a dichotomous dependent variable remained unaffected by antihypertensive treatment. RDI, age, BMI, and sex (as continuous and as categorized variables) were allowed to enter the stepwise model as possible confounding variables. Linear forward analysis with a probability of 0.05 for the stepwise entry into the model was applied. The odds ratio (OR) indicates the relative risk for a particular predictive variable, compared against a reference category (here RDI < 5, age < 39 yr, BMI < 24 kg/m², female sex), for the presence of systemic hypertension. Model fit was analyzed by comparison of predicted and observed number of patients in each category (Hosmer and Lemeshow test). Significance was tested using the Wald ² test.

An additional logistic regression analysis was performed with hypertension Classification 2 (hypertension  160/95 mm Hg) as the dependent variable, which separated the patients with regard to age (patients  50 yr and > 50 yr).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The RDI was less than 5, indicating the absence of significant SRBD in 231 patients, whereas an elevated RDI ( 5) was found in 959 patients. After allocation of patients into the different RDI classes, there was an obvious stepwise increase of blood pressure and heart rate with severity category in both patient Groups 1 and 2 (Tables 2 and 3). Increasing RDI severity was also associated with an increase in age and BMI and a lower PO₂. Analysis using the Pearson test showed a significant correlation between RDI and systolic blood pressure (r = 0.2, p < 0.01), diastolic blood pressure (r = 0.22, p < 0.01), and heart rate (r = 0.23, p < 0.01) in the patients without a history of hypertension (Group 2).

Multiple linear regression analysis controlling for BMI, age, sex, nicotine/alcohol consumption, cholesterol level, PCO₂, and PO₂ showed a significant, independent association between RDI and resting systolic and diastolic blood pressure (Table 4, A and B). The calculated systolic (diastolic) blood pressure increase with every additional RDI unit in the final linear model was 0.07 (0.07) mm Hg. Increasing daytime PCO₂ was associated with systolic blood pressure (unstandardized coefficient [B] = 0.28, p < 0.05). Total cholesterol plasma level, daytime PO₂, alcohol consumption, and nicotine consumption showed no significant association with resting blood pressure.

SRBD had a highly significant influence on daytime resting heart rate in a multiple linear regression model (Table 4, C). The calculated heart rate increase with every RDI unit was 0.1 beats/min. A decrease in PO₂ was also significantly associated with increasing heart rate, whereas age, BMI, PCO₂, sex, alcohol, and nicotine showed no independent influence. The number of patients with hypertension (Classification 1/2) increased with a higher RDI class (Tables 2 and 3). The relative risk for hypertension was significantly increased by 1.2/5.5% with every further RDI unit (-coefficient, 0.012/0.055; p < 0.01) in a multiple logistic regression. Other significant predictors for hypertension were age, BMI, and sex. The influence of the interaction term BMI×RDI was also significant ( = 0.001, p = 0.043) in more severe hypertension (Classification 2,  160/95 mm Hg), indicating that RDI was a stronger risk factor in less obese patients.

The OR for hypertension ( 140/90 mm Hg) was increased 1.8- to 1.9-fold in the RDI ranges 10 to  40 with only a marginal difference between the RDI groups 10 to < 20, 20 to < 40, and  40 (Table 5). When using the 160/95 mm Hg cutoff point for hypertension (Classification 2) there was an even higher OR: 1.52, 2.07, 2.15, and 4.15 for the groups RDI 5 to < 10, 10 to < 20, 20 to < 40, and  40, respectively.

OR for hypertension were also computed for younger ( 50 yr) and older (> 50 yr) patients. For hypertension Classification 2 OR were in general 2- to 3-fold higher in all the RDI classes in younger patients (Table 6). Moreover, in younger patients the relative influence of BMI was lower in the highest BMI class ( 30) and among those with a BMI of 24 to < 27 kg/ m². When the 140/90 mm Hg cutoff point was used, this age- related difference was no longer apparent. Sex and the interaction term age×RDI were not significant in the final logistic regression model (p > 0.1).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study of a sleep clinic cohort demonstrated a dose-related association between SRBD and daytime office blood pressure independent of age, BMI, sex, cholesterol level, daytime PO₂ as well as PCO₂, nicotine consumption, and alcohol consumption. SRBD was also independently associated with increased resting heart rate. Moreover, the OR for systemic hypertension (systolic/diastolic blood pressure  140/90) was increased approximately twofold in moderate to severe SRBD. The impact of SRBD was even stronger when a higher cutoff point for definition of hypertension (systolic/diastolic blood pressure  160 and/or 95 mm Hg) was used. Finally, the relative risk for SRBD-related hypertension was higher in patients  50 yr compared with older patients. Overall, there was a high degree of interrelation between SRBD and BMI.

To our knowledge, this is the largest clinically based cohort study with ambulatory recording of SRBD and careful assessment of blood pressure and heart rate. The strengths of the study include determination of daytime office blood pressure, which is the most frequently used cardiovascular risk marker (25). Moreover, the inclusion of daytime blood gases, cholesterol levels, and heart rate into the analysis provided new information.

Limitations include the cross-sectional nature of the study, which invalidates definite statements of a causal relationship between SRBD and hypertension. Also, the study setting with a sleep study patient sample and the selection of the control group from patients with RDI < 5 are likely to introduce a bias as snoring itself has been claimed to constitute a risk factor for hypertension (26, 27). However, if anything, this would result in an underestimation of the OR for hypertension in patients with SRBD (RDI  5). The data may therefore not be directly applied in assessments for the association between SRBD and hypertension in the general population.

Some further methodological considerations need to be made. It may be argued that home monitoring of nocturnal breathing better reflects true "SRBD prevalence and severity" because regular habits of sleep, food, and alcohol intake and physical activity are maintained. The assessment of office blood pressure, a recognized cardiovascular risk factor (25), may have provided an advantage compared to early morning or evening blood pressure readings. Experimental data (28) suggest that nocturnal asphyxia-induced sympathetic activation and thereby potentially blood pressure elevation may persist after awakening. Blood pressure obtained during the late evening hours may be influenced by circadian rhythm. In fact, such methodological differences might in part explain discrepancies in previous epidemiological data (Table 1). Additionally, the MESAM 4 system has previously been applied in epidemiological studies (23). Because this system does not provide direct measures of oronasal airflow or respiratory effort, we used RDI rather than separately analyzed apneas and hypopneas as a measure of SRBD activity. Body position was monitored during sleep, but RDI was not adjusted for the amount of time spent in the back position. However, a recent study has demonstrated that no additional information regarding the relation of SRBD and hypertension is provided by this parameter (29). RDI was calculated based on estimated sleep time, which is less exact than total sleep time determined by polysomnography. Using this procedure previous studies have demonstrated an approximately 95% correlation between polysomnography and MESAM 4 evaluation for determination of RDI (24). Finally, BMI rather than waist-hip ratio, waist circumference, or skinfold thickness was used to assess the influence of being overweight. Although BMI may be less accurate (30), this is not supported by other data investigating the association between SRBD and hypertension (18, 29).

The present data support the previous investigations from Young and coworkers (18) in a population-based cohort. In their study, systolic (diastolic) blood pressure increased by 0.24 (0.12) mm Hg with every RDI unit and there was a 2- to 3-fold increase of hypertension risk in moderate to severe SRBD. However, an age-related effect was not stated and patients in this study had less severe SRBD and were more frequently overweight. In fact, more severe SRBD and prominent comorbidity, which are common in the clinical setting, were underrepresented in this population-based study. Grunstein and coworkers (30) described an association between RDI and supine morning blood pressure independent of age and BMI in a clinically based study. Interestingly, in their statistical model, morning systolic and diastolic blood pressures showed 18% and 12% explained variance, which were comparable to 12% explained variance in our study. Diastolic blood pressure was more sensitive to SRBD than systolic blood pressure and waist circumference was superior to BMI as a predictor for SRBD in the final model. Another clinically based cross-sectional study by Carlson and coworkers (31) found a 2.1- to 2.2-fold increased risk for hypertension in patients with mild to moderate SRBD but the study population was leaner and had less severe SRBD.

There is a strong interrelation described between BMI and RDI as risk factors for hypertension in all of these studies. In fact, obesity and SRBD may provide additive risk (31). However, different statistical procedures have been employed. Our data and those of Young and coworkers (18) suggested a significant negative interaction variable of BMI*RDI in the logistic regression,  = 0.0012 and  = 0.00198, respectively. Thus, the influence of SRBD on hypertension may be reduced in patients who are extremely overweight. Second, the choice of different cutoff points in the analysis affects the results. Carlson and coworkers (31) described an increased risk for hypertension in the patient group with oxygen desaturation index (ODI)  5 compared with non-SRBD patients, but no further increase in risk with rising ODI. In contrast, Young and coworkers (18) who used an RDI  5 cutoff point demonstrated a stepwise increase in the OR for hypertension with increasing RDI severity. Our data indicate that differences in the definition of systemic hypertension may explain this discrepancy. When hypertension was defined as a blood pressure  140/90 mm Hg, RDI  10 was associated with a significantly increased risk (p = 0.01). Using the blood pressure definition  160/95, an increased risk was identified already at an RDI  5 (p = 0.054).

Our findings are in conflict with other clinically based studies (Table 1), which failed to show an independent relationship after correction for age and body weight (15, 16, 32). Differences in study populations, diagnostic procedures, definitions of SRBD, desaturation and/or hypertension cutoff points as well as statistic procedures should be taken into account. Different methods for blood pressure determination were used and it is also possible that the sample size in some of these studies was too small to permit the detection of a possible association between SRBD and hypertension. In fact, all three studies with sample sizes larger than 1,000 subjects showed a significant and independent influence of SRBD on hypertension and/or blood pressure after control for confounders. It may also be speculated that obesity, hypertension, and SRBD have common genetic traits (33). Statistical control for obesity would in such a case underestimate the association between SRBD and hypertension.

Three new aspects of the present study need to be highlighted. First, there was an age-dependent effect on the interaction between SRBD, BMI, and hypertension. The SRBD-related risk to develop hypertension was increased in younger SRBD patients compared with the older age group. This is in accordance with a recently published prospective study from Lindberg and colleagues (27), who documented an increased risk in young (< 50 yr) but not in older male snorers to develop systemic hypertension. It has been previously claimed that systolic blood pressure increases with age because of progressive wall stiffening in the large arteries (25), whereas diastolic blood pressure tends to fall in the older age groups. The same age-related pattern of blood pressure was found in our study. Although the interaction term age*RDI was not statistically significant, it may be speculated that the previously described pathophysiological mechanisms for hypertension in SRBD (5) provide a stronger impact in a younger age group. At an older age, other risk factors such as being overweight may dominate over SRBD, so that the independent contribution of SRBD for the development of systemic hypertension is reduced or even lost.

Second, RDI showed an independent influence on resting heart rate, which was stronger than that found for systolic or diastolic blood pressure. Resting heart rate is mainly determined by the balance between sympathetic and parasympathetic tone. It may be speculated that blunted baroreflex control (7) and the elevated sympathetic activity induced by the nocturnal breathing disorder (34) play a critical pathophysiological role. Interestingly, increased resting heart rate has been associated with increased cardiovascular risk (35, 36). It has been shown that the increased risk was in part independent of other known cardiovascular risk factors such as age, blood pressure, or cholesterol levels (37). Thus, resting heart rate may be considered as an important predictor for cardiovascular morbidity and mortality in SRBD patients in future studies.

Third, in SRBD patients daytime blood gases were found to independently influence blood pressure and heart rate at rest. Repeated nocturnal asphyxia related to SRBD is claimed to increase blood pressure acutely and it may be a major cause for elevated nocturnal blood pressure. The influence of carbon dioxide is less well known but hypoxia and hypercapnia have been attributed a hypotensive effect in patients with chronic obstructive pulmonary disease (38). Although the exact mechanism remains unclear, the influences of blood gases on office blood pressure and heart rate were independent of the severity of SRBD.

In conclusion, our data suggest an independent dose-response relationship between SRBD and daytime office blood pressure as well as resting heart rate. SRBD has been shown to be a risk factor for systemic hypertension independently of BMI and age with an OR that increased with SRBD severity. Moreover, this risk was elevated in young patients (< 50 yr). Future prospective studies addressing the possible causal relationship between SRBD and systemic hypertension are warranted but our findings demonstrate that high awareness of SRBD should constitute an important part of modern management of hypertension.