Hypertension prevalence in obstructive sleep apnoea and sex: a population-based case–control study

The association between obstructive sleep apnoea (OSA) and hypertension (HT) is an issue of increasing concern 1, 2. A substantial proportion of the adult population has mild-to-moderate OSA characterised by repeated episodes of apnoea and hypopnoea during sleep 3. The prevalence of daytime HT has been estimated to include 50–60% of patients with known OSA 4. Conversely, OSA has been reported in >30% of HT patients 5 and may be a frequent cause of drug-resistant HT 6. The reciprocal association between the two conditions advocates that OSA and HT may share common pathogenetical pathways. However, OSA appears to promote HT development 7 and to sustain a condition with elevated blood pressure 8. Random controlled intervention trials of nasal continuous airway pressure in OSA patients demonstrated a reduction of blood pressure over a period of 4–9 weeks 9, 10.

Studies investigating the potential pathophysiological mechanisms involved in HT development in OSA have proposed the involvement of an increased peripheral vascular resistance 11. Potential mediators include an increased sympathetic activity 12, 13, impaired renin-angiotensin-aldosterone system function 14, 15 and endothelial dysfunction 16, 17. Moreover, the repeated episodes of hypoxia/re-oxygenation in moderate-to-severe OSA may promote vascular inflammation 18, oxidative stress 19 and accelerate the development of atherosclerosis 20. The exact influence of genotypical or phenotypical characteristics on the susceptibility to HT in OSA is still unclear 21–23. OSA may represent a synergistic risk factor for HT development in certain subjects with, for example, comorbid obesity 24, a metabolic disease 25 or certain coexisting traditional risk factors 26. Similarly, this risk may be different in males compared with females, due to a differential risk-factor pattern between the sexes.

Previous data investigating HT prevalence in sleep-laboratory 27, 28 and population– 29, 30 based cohorts, suggest that the risk of HT is essentially equal in male and female subjects with OSA. However, data on cardiovascular mortality 31, as well as recent experimental data on endothelial function in sleep apnoea patients, suggested a differential impact in males and females 32.

The primary aim of the present case-control study was to explore OSA as a risk factor for HT in two community-based samples; patients with HT treated in primary care (cases) and normotensive subjects from a random population sample (controls). An additional aim of the study was to investigate the influence of sex on this association.

SUBJECTS AND METHODS

The Skaraborg Sleep Study subjects

The present study used a case–control design aimed at including 100 male and 100 female patients with HT from the surveyed population, with a corresponding number of population controls previously classified as normotensive. All control subjects with known HT at baseline (n = 60) or with newly developed HT (n = 46) were excluded from the control sample. However, the vast majority of those with known HT at baseline were already included in the patient sample. Participants from the two cohorts, aged 40–65 yrs at baseline, were invited to undergo a full-night ambulatory polysomnography (PSG). Those responding to the invitation were seen in random order for informed consent at the primary care centre (study visit see below). Entry criteria for control subjects included a baseline blood pressure <160/<90 mmHg. When all eligible patients with HT had been considered for participation and a corresponding number of control individuals had been included the recruitment was stopped. Subjects fulfilling criteria subsequently underwent PSG. The recruitment process is summarised in figure 1⇓. From the 290 invited patients with HT, 163 (56%) were included. Of the 293 invited controls 185 (63%) were finally enrolled. Descriptive characteristics of the study populations are shown in table 1⇓. The Skaraborg Sleep Study was reviewed and approved by the local human research ethics review board.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1—

Flow chart displaying the selection undertaken for the study population. HT: hypertension; PSG: polysomnography.

View this table:

• View inline
• View popup

Table 1—

Baseline characteristics of hypertensive(HT) and non-hypertensive (non-HT) males and females

RESULTS

OSA was highly prevalent in these middle-aged to elderly study participants. The AHI levels and OSA prevalence were generally higher in males than in females (table 1⇑ and 2⇓). Overall AHI expressed as mean±se was 34±2.6 and 23±2.4 events·h⁻¹ in hypertensive and normotensive males, respectively (p = 0.004). In contrast, the AHI levels in the two female groups were 23±2.6 and 24±2.5, respectively (p = 0.759). As seen in table 1⇑, hypertensives of either sex tended to be older, more obese and to have a higher WHR when compared with their normotensive counterparts. Blood pressures were almost identical in males and females in both the hypertensive and the normotensive groups. Total sleep time was generally shorter in hypertensives of either sex.

The proportion of males with OSA, using either AHI 10 or 30 events·h⁻¹ as the cut-off for diagnosis, was significantly higher in HT subjects than in the controls (table 2⇑). This difference was not seen in the females. A formal test of a differential effect by sex in the association between OSA and HT by a two-way interaction term (OSA×sex) was nonsignificant (2.2 (0.8–6.0); p = 0.124). The OR for HT was unaffected by AHI severity threshold (10 or 30) in males, while AHI, by either threshold, was not associated with HT in females. In both sexes body mass index (BMI) was associated with HT (p<0.001 in males and p = 0.018 in females), and a corresponding age-independent association was seen between BMI and OSA (p = 0.007 in males and p = 0.004 in females).

There was a dose-response relationship between AHI categorised in tertiles and HT in males (fig. 2⇓). The OR for HT increased across AHI tertiles from 1.0 to 2.1 (95% CI: 0.9–4.5) and from 1.0 to 3.7 (95% CI: 1.7–8.2) in males (p-value for trend was 0.005), but not in females where the OR increased from 1.0 to 1.8 (95% CI: 0.8–3.9) and from 1.0 to 1.6 (95% CI: 0.7–3.5; p-value for trend was 0.310). In a logistic regression analysis of the association between OSA (AHI ≥10 events·h⁻¹) and HT, adjustment for differences in age and BMI resulted in a moderate weakening of the OR in males to 2.3 (95% CI: 1.0–5.1; p = 0.043). However, the association remained nonsignificant in females (table 3⇓). Age and smoking had less of an influence. Factors related to the severity of overnight hypoxaemia had an insignificant influence (data not shown). Confounding from body composition was also accounted for by repeating analyses of associations between OSA and HT in strata of overall obesity (BMI World Health Organization-categories) and abdominal obesity (WHR ≥1.0, or not). However, there was a consistent and remaining pattern suggesting an independent association between OSA and HT in males. All analyses were repeated after the exclusion of subjects with a total sleep time <4 h (n = 22) and results were essentially unchanged.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2—

Odds ratio (OR) for hypertension (HT) in a) males and b) females divided in tertiles of apnoea-plus-hypopnoea index (AHI) severity. The AHI severity within tertiles is indicated. – – –: an OR of 1.0 for the reference group (tertile one) to facilitate comparisons with tertile two and three, respectively.

DISCUSSION

The present study demonstrated that OSA is a highly prevalent phenomenon in these middle-aged to elderly cases of HT and population controls. An independent association between OSA and HT was confirmed and the association was confounded by, though independent of, body composition. However, there was a difference in risk between males and females suggesting that the susceptibility to HT development in OSA may be sex dependent and further studies should focus on this issue.

The current study confirms two aspects of previous cross-sectional population studies. First, unrecognised OSA was highly prevalent in the general population 3 and sleep apnoea was over-represented in subjects with a known HT diagnosis 41. The individual awareness of OSA in the current population was low. In fact, only one subject, who enrolled in the study, had previously received medical attention for symptoms of sleep-disordered breathing. The average BMI of the current subjects were similar to that of several previous populations investigated in the same context, e.g. the Sleep Heart Health Study 30. Yet, the prevalence of OSA appeared to be considerably higher. Methodological differences are likely to explain at least part of this discrepancy. The current study used both a nasal pressure cannula and a thermistor for assessment of oro/nasal airflow. This procedure has previously been demonstrated to result in a considerably higher calculated AHI in sleep studies 42. The method is more sensitive at detecting sleep-disordered breathing, particularly in mild OSA. Therefore, a larger number of subjects with mild OSA are likely to have been identified in the current study.

Secondly, OSA was independently associated with HT. This association was valid irrespective of the cut-off level applied for diagnosis (AHI 10 or 30 events·h⁻¹), suggesting that even a proportionally low intensity OSA was associated with an increased risk for HT. The overall OR for HT was in the same order of magnitude as that previously described in a sleep laboratory cohort 28, a population of middle-aged US state employees 7 or in the Sleep Heart Health Study cohort 30. In contrast to the findings in any of these cohorts, the current study found that the association between OSA and HT was almost entirely restricted to the male sex. The current study had a case-control design and did not constitute a combined cohort study. Thus, the effect of OSA exposure in the cases was compared with that of the controls. By contrast, in the previous cohort trials, the analysis was based on the exposure of OSA and the effect on association with HT. Moreover, HT cases in the current study had been accumulated over an extensive time period and the HT diagnosis was evidently well established. It is, therefore, apparent that a different subset of HT sufferers was addressed in the present study. It cannot be excluded that the composition of the HT subset in the current study accounted for the lack of association between OSA and HT in females. However, previous studies investigating drug resistant HT 6, 8 have not pointed towards evident sex differences in the association between HT and OSA.

The current study differs from previous studies in several other important aspects. It was sex balanced and addressed an ethnically homogenous population. The female subjects were almost exclusively post-menopausal. This may have accounted for a proportionally higher number of females with OSA in the present study 43, and therefore, may have resulted in a dilution of the association between OSA and HT in females. However, the lack of an association in females should be interpreted with great caution. Not only were the age adjusted OR for OSA (AHI ≥10), and the upper AHI tertile in females increased to 1.3 and 1.6, allowing for up to 2.6 and 3.5, respectively, but a formal interaction term was also nonsignificant, although again with a wide 95% CI. Although the effect seems to be smaller in females, the current findings indicate too low a power to conclusively rule out an association between OSA and HT in females. The high criteria for HT used in this study may also have contributed to a dilution of an existing association.

The current study has several strengths and weaknesses. The response rates for participation in the PSG surveys were ∼60%, which should be considered as high, given the commitment required from the participants. However, it is not high enough to confidently exclude selection bias. This should be kept in mind when generalising the findings of this study. PSG data was acquired in a primary healthcare ambulatory setting. Results that were close to the complete set of recordings were technically acceptable for adequate sleep stage scoring. Elimination of the subset of subjects with a short total sleep time did not influence the principal results of the study. This procedure may be argued to have changed the validity of the sleep recordings, however, many studies have shown the similarity of in-lab PSG and home PSG 44, 45. A considerable proportion of the cases in this study were taking antihypertensive medications. However, HT was defined as a categorical variable in the current study's analysis. There is no evidence to suggest that antihypertensive medications would have affected the occurrence or severity of OSA differently 46. The incomplete risk factor classification was a weakness of the current study. Traditional risk factors were thoroughly documented, but blood sampling was systematically performed only at the time of the original survey. Although, this does not provide a limitation in terms of investigating an association between OSA and HT it restricts the possibility to perform an evaluation of more recently identified confounder influences.

In conclusion, the current data confirms previous findings of an independent association between obstructive sleep apnoea and hypertension. This association was evident in males but not identifiable in these essentially post-menopausal females. The findings suggest a possible sex-dependent difference in susceptibility to hypertension development in sleep-disordered breathing. Due to the high prevalence of obstructive sleep apnoea in patients with hypertension, sleep-disordered breathing should be considered when patients with high blood pressure are evaluated in primary care.