The association of sleep disordered breathing with left ventricular remodeling in CAD patients: a cross-sectional study

LVM was divided by height^2.7 and by body surface area (BSA) to obtain left ventricular mass index (LVMI). Because LVM was most strongly correlated with BMI and there is not a specific value adjusted for obesity, for this reason we adopted the most commonly used: LVM adjusted according to height^2.7 (LVM/height^2.7) [32] which is proposed to be similarly sensitive in detecting obesity-related and -unrelated LVH. The height^2.7 index was also important for the subcategorization of patients according to the presence of concentric or eccentric LVH because the prognostic value of such subcategorization was apparent only when the height^2.7-based criterion was applied [39].

LVH was defined as gender-neutral cutoff value for LVM indexed for height > 51 g/m^2.7 [40, 41]. Relative wall thickness (RWT) was computed by the ratio of sum of IVST and LVPWT to LVEDD. LV structure was categorized as normal (RWT ≤0.42 and no LVH), concentric remodeling (RWT >0.42 and no LVH), concentric hypertrophy (RWT >0.42 and LVH) and eccentric hypertrophy (RWT ≤0.42 and LVH) [36]. In addition, for the purposes of this study, we defined the prevalence of LVH on the basis of wall thickness criteria (IVST or LVPWT ≥12 mm) [22, 37].

Statistical analysis

Quantitative variables are expressed as means ± SD or medians (interquartile ranges [IQRs]) or in percentages, when appropriate. The Kolmogorov-Smirnov test was used to assess the normal distribution of continuous variables. To compare basic characteristics between non-SA and SA (AHI ≥ 5) groups the significance of differences was assessed using Mann–Whitney tests (or t tests, if data were normally distributed) or chi-squared tests, respectively. Analysis of variance (ANOVA) was used to determine whether there are any statistically significant differences between the means of three (non-SA group excluded) or four independent SA severity (coded as 1 = non-SA, 2 = mild, 3 = moderate, 4 = severe) groups. For the linear trend the p value for F-ratio was presented. A Bonferroni correction was applied to post hoc pairwise comparisons of means.

To further substantiate the notion that mild-severe SA was an independent determinant of LVH by wall thickness criteria or concentric LV geometry (dependent variable coded 0 = absent/1 = present), bivariate analysis and multiple logistic regression analyses were performed. In the multiple logistic regression analysis, statistically significant variables in the bivariate analysis, as well as the non-significant variables with suggested clinical relevance, were included. The candidate explanatory (independent) variables for entering in the multiple regression model were hypoxemia during sleep SaO₂ and age, AHI (4 categories), BMI (4 categories), age over 60, sex, hypertension, diabetes mellitus, NYHA class, current smoking, history of MI as well as LVEF lower than 45%, excessive daytime sleepiness (ESS ≥ 10). Because daytime sleepiness is common in patients with SA, this factor was taken into account as potential confounding factor. The goodness of fit and Nagelkerke R² for these models presented. Forward stepwise selection was then performed to identify factors independently associated with LVH. Data were analyzed with SPSS 17.0 statistical software (SPSS Inc., Chicago, Illinois, USA). A p-value <0.05 was considered significant.

Clinical characteristics

Patients were predominantly male (76%), mean age was 57 ± 9 years (ranged from 32 to 81 years). The median value of the AHI in the sample was 3.4 and the interquartile range (25th percentile to 75th percentile) was 1 to 9. Sleep apnea defined as five or more episodes of apnea and hypopnea per hour of sleep (AHI ≥ 5 n/h) was present in 300 (39%) of patients and 175 (22.7%) of patients had LVEF ≤ 45%.

General characteristics comparison according to SA status

ESS ranged from 0 to 18 with a median of 6 (interquartile range 4–9). The patients with SA had higher scores on the ESS than those non-SA patients (p = 0.006). No statistically significant between-group differences were found for frequency of excessive daytime sleepiness (ESS ≥ 10) (respectively 16% and 19%; p = 0.14) (Table 1). The mean ESS score was within the normal range in both groups (6.0 ± 3.5 vs. 6.7 ± 3.5). ACE inhibitors/angiotensin II receptor blockers and beta-blockers were prescribed to 81% and 89% of study patients, respectively. Antihypertensive and antiagregant drugs were used by larger number of patients with SA than non-SA patients (p = 0.012 and p = 0.003, respectively; data not shown).

Clinical, polysomnographic and echocardiographic characteristics according to SA severity

We found no statistically significant differences in a frequency of EDS among patients without (15.7%) and with SA despite the severity (17.7% to 24.5%, p = 0.556) (Table 3).

There were not found statistically significant differences in left atrium diameter, in thickness of IVST and LVPWT, in value of LVM and LVMI (g/m⁻² and g/m^2.7) of patients included in any of the groups of SA severity. This suggests that SA severity according to AHI categories was not inducing changes in their LV thickness and mass (ANOVA for SA severity groups all p _{for trend} > 0.05, Table 4).

Left ventricular hypertrophy

In univariate analysis the CAD patients with SA had a 1.9-fold higher prevalence of LVH on the basis of wall thickness criteria (IVST or LVPWT ≥12 mm) compared to non-SA patients (49.3 vs. 33.4%; odds ratio, OR = 1.9; 95% confidence interval (CI), 1.4–2.6; p < 0.001) (Table 4, Fig. 1).

Left atrium diameter, IVST and LVPWT was significantly higher in patients with SA, than those in non-SA patients (all p < 0.001). However, no between-group differences were found for RWT (Table 4).

LVH defined by allometrically corrected (LV mass/height^2.7) gender-independent criteria was present in 86% of the SA patients compared to 74% in non-SA patients group (p < 0.001).

Distribution of LV geometrical pattern

Some form of LV geometric remodeling was present in 92% of patients, with concentric LVH as the most common geometric abnormality. Eccentric LVH occurred in 26.3% (n = 192), concentric LVH in 52.4% (n = 382) of patients, concentric LV remodeling in 13.4% (n = 98). Eccentric LVH is much less prevalent than concentric hypertrophy.

The frequency distribution of the LV geometric groups indicated significant differences between SA and non-SA patients (χ² = 18.9; p < 0.001) (Fig. 2). Percentage of concentric hypertrophy defined as the combination of an increased RWT and an increased LVMI (LVM/height^2.7) was 60.9% in CAD patients with SA compared to 47.1% (p < 0.001) in non-SA patients (Fig. 2).

LVH (high LVMI) is traditionally two right panels designate subjects with increased LVMI (Fig. 2). As it described in Table 4 pattern of concentric remodeling – comprising high RWT without an absolute increase in mass index, pattern of concentric hypertrophy depend on increased LV wall thickness and chamber dilation is present.

Determinants of LV hypertrophy in the CAD patients

Predictors of LVH on the basis of wall thickness criteria

Multiple logistic regression analysis highlighted cross sectional predictors of LVH on the basis of wall thickness criteria (interventricular septal thickness or posterior wall thickness ≥ 12 mm) such as male sex (OR = 2.6) p < 0.001), hypertension (OR = 1.8 p = 0.016), age over 60 years (OR = 1.7 p = 0.002), LVEF ≤45% (OR = 0.7 p < 0.001) and AHI (for trend p = 0.029): mild SA versus non-SA (OR = 1.5; p = 0.030), moderate SA versus non-SA (OR = 2.1; p = 0.012) (Fig. 3).

The prevalence of LVH on the basis of wall thickness criteria was lower in patients with LVEF ≤ 45% and did not differ significantly between the patients with severe SA (AHI ≥ 30) and non-SA (OR = 0.96; p = 0.91). Furthermore, the risk progressively increased with higher BMI values (p for trend <0.001) but this increase slowed down in patients with BMI ≥ 40 versus normal (OR = 3.5 95% CI, 1.3–9.6; p < 0.001) (Fig. 3).

Predictors of concentric LVH

Mild SA (AHI = 5–14.9 /h), performing multiple logistic regression analysis, was found as a strong independent cross sectional predictor of concentric LVH in the CAD patients (OR = 1.9). The risk of concentric hypertrophy did not increase with the greater level of SA severity.

For BMI increase every 5 kg/m² (e.g., moving from normal weight to overweight, or from overweight to mild obesity), the risk of concentric LVH also increased by 50% (OR = 1.5). Such factors as gender, history of smoking, diabetes mellitus were not statistically significant (Fig. 4).

Summary of findings

Previous studies, that often lacked adjustment for important cardiovascular risk factors suggested that SDB was associated with changes in heart structure, including LV hypertrophy. The present study was performed in a large sample of consecutive stable CAD patients characterized by a relatively high degree of comorbidity. This study extends these results with the novel finding that even mild level of SDB is strongly associated with LVH, with increase LV wall thickness (both interventricular septum and posterior wall) and concentric LVH, independently of coexisting traditionally recognized risk factors. The risk of concentric hypertrophy did not increase with greater levels of SDB severity.

These results could suggest that the effect of SA on LVH risk is fully realized at certain SA severity level. As hypothesized, there was almost 2-fold increase in risk of having LVH on the basis of wall thickness criteria, and 1.5-fold higher risk of concentric hypertrophy, adjusted for the traditionally recognized risk indicators in CAD patients with the disordered breathing.

Although patients with SA had a higher mean value of the ESS than non-SA patients, this value was lower than the reference value for excessive sleepiness (ESS ≥10). Therefore, this value did not seem to be a determining factor in the diagnosis of SA in these patients. The limitations of ESS are known in cardiovascular patients and the existence of patients with severe SA without excessive sleepiness is controversial [41, 42].

Though we do not claim any causal relation our results could suggest that even mild SA had additive effect in changes to the heart shape, similar to the effects of both hypertension and obesity. This effect includes increased mass and thickening of the heart wall. As patients with stable CAD and with SA often do not show typical SA symptoms, the presence of one or more established predictors of LVH on the basis of wall thickness criteria or concentric LVH, such as male sex, older age, higher BMI, ejection fraction >45%, or SA may constitute independent risk factors for adverse long-term outcomes. The increase recognition of SA as an independent, additive or even synergistic risk factor for CAD brings a growing need for early identification of high-risk individuals and a consensus regarding well defined treatment strategies in such patients.

The prevalence of SA in CAD patients

The data in our cross-sectional study indicates that undiagnosed SA was highly prevalent (39%) among patients with CAD. The prevalence of SA in patients with CAD was higher than in normal population, as stated by numerous studies, and was in range or higher than in our study. The studies reported a wide range of prevalence, ranging from 26% to 66%, partially explained by the different value of AHI used to establish the diagnosis of SA in the different studies [3, 5, 41].

There is some evidence suggesting that the prevalence and the severity of SA are modified along the CAD evolution. Mooe et al. reported that clinically significant sleep apnea (AHI >10) was documented in 37% of investigated 142 patients with stable angina pectoris and angiographically confirmed CAD [43]. Peker et al. reported 30% prevalence of SA in patients admitted with an acute coronary syndrome, while Areias et al. claimed SA to be present in 43% of the study patients [12, 17, 41]. In CAD patients immediately after an acute myocardial infarction and after clinical stabilization of unstable angina pectoris Moruzzi et al. observed a significantly higher AHI compared to stable angina pectoris patients [12, 44]. In the study conducted in Greece, the authors concluded that there was a high prevalence of SA during the acute phase of coronary syndrome ACS (54%); this prevalence decreased 6 months after ACS and persisted in 21% of the patients, which indicates that this abnormality may be transient [41, 45]. Recently, Fox et al. (2016) investigating the prevalence of SDB in cardiac rehabilitation patients, reported a prevalence of SDB (AHI ≥ 15) by different cardiac condition. They found that patients who had undergone coronary artery bypass graft (CABG) had a higher prevalence of SDB and more severe SDB compared to patients with stable CAD or ACS (prevalence of 53%, 46% and 45% respectively) [46]. As noted, patients with heart failure and SA (in particular CSA) do not generally present with symptoms that distinguish them from heart failure patients without SA. Arzt et al. (2016) reported a prevalence of 46% in a large representative population of patients with stable chronic HF receiving optimized medical treatment [47].

In general, there is a stronger relationship between SDB and CAD in clinical cohorts than in the general population because clinical cohort studies are particularly influenced by comorbidity and confounding factors, including obesity, hypertension, smoking, and hyperlipidemia. This circumstance may also suggest that SDB constitutes an additive or synergistic risk factor for development of CAD [14, 16].

Cardiac remodeling in hypertension and obesity

Despite the high prevalence of SA reported in patients with CAD, it is frequently underdiagnosed [12]. In addition, comorbidities such as obesity and hypertension that coexist with the majority of SA patients make it more difficult to assess the independent risk of SA on vascular disease. Both hypertension and obesity strongly affect cardiovascular structure and function. Since obesity and hypertension are closely linked and often coexist, while obesity is an independent risk factor for arterial hypertension these two disorders often exert a dual burden on the left ventricle, leading to both dilatation and hypertrophy [30] as well as markedly increasing the premature risk of CV disease [19]. The presence of hypertension in patients with SA likely contributes to the association with LVH or increased wall thickness [48]. Data supporting a possible cause and effect relationship between SA and LVH were published by Cloward et al., who found that 6 months of nocturnal positive airway pressure administration to patients with severe OSA was associated with a significant reduction in LV wall thickness [49].

Many studies have shown that obesity is associated with LVH, a potential contributor to heart failure [29]. It is unclear to what extent LVH results directly from obesity or from associated conditions, such as hypertension, impaired glucose homeostasis, or obstructive sleep apnea [23]. Early studies from Ochsner demonstrated that the cardiac adaptation to these two disorders differed considerably [19]. Hypertension causes fundamental response to an isolated increase in pressure load or afterload such as an increase in LV wall thickening without chamber dilatation. On the other hand, obesity leads to LV chamber dilatation with only minimal increases in LV wall thickness. Cardiac remodeling in obesity is characterized by concentric geometry at least as often as eccentric geometry [33]. However, some investigators have found evidence of eccentric LVH (an increase in cavity volume that is greater than the increase in wall thickness) in obese subjects [30, 31, 48].

In fact, LVH in SDB is as closely related to the severity of the sleep disturbances as it is to the levels of blood pressure and BMI [23]. Our study findings also confirm the strong independent association between concentric LVH and obesity at any value of BMI. It helps to dispel the notion that LVH in this condition is exclusively eccentric. When obesity and hypertension coexist, there is a substantial increase in the prevalence of LVH at any value of BMI confirmed in many studies [23, 33, 49]. However, the increase in LV mass/height^2.7 is more pronounced in patients with hypertension than in those without [33]. MacMahon and colleagues demonstrated that weight loss in hypertensive subjects with overweight reduced LVM and posterior wall thickness [50]. In contrast, Niroumand et al. concluded that SA is not associated with increased LVM independently from obesity, hypertension or advancing age [21].

LV geometry

We found that male sex, age 60+, hypertension, higher BMI, higher than 45% LVEF and SA were independent predictors of LVH on the basis of wall thickness criteria and predictors of concentric LVH (with the exception of sex). These indicators showed additive effects of SA and are consistent with those of other studies. Drager et al. (2007) looked at arterial stiffness by measuring pulse wave velocity and heart structure in OSA patients with and without hypertension. They demonstrated that patients with untreated OSA had increased left atrial diameter, septal wall thickness, LV posterior wall thickness, LV mass index, and LVH, even controlling for cardiovascular risk factors [51]. These indicators showed additive effects of hypertension and OSA.

SDB produces a series of mechanical, hemodynamic, chemical, neural, and inflammatory responses (i.e. intermittent apnea-induced hypoxia, hypercapnia, increases in sympathetic drive and LV myocardial afterload, daytime hypertension, loss of vagal heart rate regulation and systemic vasculature remodeling) [23, 26, 49]. These strongly impact the CV system, being all potent stimuli to myocyte necrosis and apoptosis, myocardial ischemia, and finally, predict development of concentric LV geometry [26].

The associations between SDB and LV geometry have been previously investigated in a lot of studies conducted on patients with obesity and hypertension, yet the results are irreconcilable [21, 23, 26, 29, 31]. Both overweight and increased blood pressure determine a state of chronic myocardial LV overload and represent two of the main predisposing factors for LVH and systolic dysfunction. In theory, the two conditions are hemodynamically different and should lead to different LV geometry adaptations [30]. In clinical practice, such dichotomy is less evident because obesity and hypertension lead to heterogenic and unpredictable changes in LV geometry [18, 19, 23].

Although less well established than for hypertension, there is growing evidence that SA is a risk factor for incident diabetes [52]. In addition, some reports have demonstrated metabolic syndrome independent association of SA with LV structure [53, 54]. The relative absence of effects of diabetes mellitus on LVM in our study could be explained by relatively low prevalence of this CAD risk factor in our patient population, and the fact that this prevalence was similar in patients with SA and non-SA (21%). The impact of diabetes mellitus on LVM is predominantly through an interaction with obesity and age rather than a direct and independent effect. After adjusting for confounding factors there was no difference between diabetic and non-diabetic participants with regard to apnea-hypopnea events [14, 52].

Despite significant scientific evidence pointing to SA as an emerging CV risk factor, SA is still underdiagnosed in several cardiology subspecialties (hypertension, coronary, arrhythmia, heart failure, valvular heart disease) and consequently undertreated [55]. On account of the high prevalence of SA in the general population, the sleep medicine field is confronted by a demanding challenge to provide sufficient resources for the management of this disorder. In spite of the differences in the diagnostic procedures as well as different AHI cut-off values for definitions of SA, there seems to be enough evidence to warrant overnight sleep screening in individuals with CAD, given that concomitant SA may worsen long-term outcomes. Furthermore, several subsequent studies have demonstrated that patients with untreated severe SA exhibit an increased incidence of fatal and nonfatal cardiovascular events (including MI and ACS) compared with those without SA or even with those with untreated mild to moderate SA [6, 14, 16]. In a longitudinal study of 408 patients with CAD, Mooe and co-workers (2001) noted that SA independently increased the risk for incident cerebrovascular events [13]. Yet such an association suggests but does not demonstrate a causal relationship, and the role of treating SA to prevent CV diseases remains unclear [6]. SA is prevalent in our study population, especially in its mild form. Although severe SA is associated with adverse health consequences, the association with mild disease is unclear. Reports differ regarding the clinical relevance of mild SA. Improved diagnostic techniques and evidence based approaches to management in mild SA require further research. Exploring potential causes and improving recognition as well as SA treatment in cardiology might have benefits in the reduction of CV morbidity and mortality [56].

Limitation and strength of the study

We acknowledge a few limitations of this study. First, the cross-sectional design does not permit us to evaluate the longitudinal impact of SA on cardiac structure, as the real duration of the SA is not known. A potential limitation is that our study was performed in a clinic-based sample. Second, in our study we do not separately analyze the influence of central and obstructive events. The causative mechanisms of obstructive and central type sleep apnea differ, but the physiological and hence cardiovascular sequelae can be considered in unison [4, 5, 7].

The strengths of this study include relatively large sample size. First, by including a large sample of patients and numerous confounding covariates, our study has overcome many of methodological limitations and provides strong support for an independent cross-sectional association between SA and LV. Second, adjustments of echocardiographic parameters to BMI and hypertension have been successfully implemented. Third, all of our subjects had comprehensive overnight standardized polysomnography data.