Association of Excessive Daytime Sleepiness with the Zung Self-Rated Depression Subscales in Adults with Coronary Artery Disease and Obstructive Sleep Apnea

Excessive daytime sleepiness (EDS) is a factor associated with both obstructive sleep apnea (OSA) and depressive symptoms. Continuous positive airway pressure (CPAP) treatment may decrease EDS in adults with OSA; however, the modulatory role of depressive symptoms on the improvement of EDS is not known. We aimed to explore the association between subscales of the Zung Self-rated Depression Scale (SDS) and Epworth Sleepiness Scale (ESS) over a 2-year period in coronary artery disease (CAD) patients with OSA. This was a post-hoc analysis of the RICCADSA cohort, in which 399 adults with CAD (155 sleepy OSA [apnea–hypopnea index ≥ 15/h] and ESS score ≥ 10, who were offered CPAP; and 244 nonsleepy OSA [ESS < 10]), randomized to CPAP [n = 122] or no-CPAP [n = 122]) were included. Three factors were extracted from the Zung SDS, based on the principal component analysis: F1, cognitive symptoms and anhedonia; F2, negative mood; and F3, appetite. In a mixed model, the ESS score decreased by 3.4 points (p < 0.001) among the sleepy OSA phenotype, which was predicted by the decline in the F2, but not in the F1 and F3 scores. The fixed effects of time were not significant in the nonsleepy OSA groups, and thus, further analyses were not applicable. Additional within-group analyses showed a significant decrease in all subscales over time both in the sleepy and nonsleepy OSA patients on CPAP whereas there was a significant increase in the nonsleepy OSA group randomized to no-CPAP. We conclude that the improvement in negative mood symptoms of depression, but not changes in cognitive symptoms and anhedonia as well as appetite, was a significant predictor of decline in the ESS scores over a 2-year period in this CAD cohort with sleepy OSA on CPAP treatment.


Introduction
Obstructive sleep apnea (OSA) is a common sleep-related breathing disorder among adults with coronary artery disease (CAD) [1]. Observational studies have suggested that OSA may increase the risk of incident CAD and mortality [1]. OSA may induce excessive daytime sleepiness (EDS), which is the most common treatment indication for continuous positive airway pressure (CPAP). In OSA patients, EDS is shown to be associated with higher health care use for hypertension, diabetes and depression [2]. Moreover, a significant number of patients with OSA experience residual EDS despite an adherent use of CPAP, and the reasons for the residual EDS are not well defined [3,4].
OSA may also cause depression by disturbing sleep patterns and through its cardiovascular effects [5]. In this context, the evaluation of depression in cardiac patients with concomitant OSA is challenging due to the difficulty of distinguishing depressive symptoms from EDS and the multidirectional relationship of EDS, OSA, CAD and depression.
We have recently demonstrated significant reductions in Zung Self-rating Depression Scale [SDS] scores in nonsleepy OSA patients randomized to CPAP compared with those in the no-CPAP group in the RICCADSA trial [6]. Moreover, there was an even higher benefit in the sleepy OSA phenotype following one year of CPAP treatment [7], suggesting that improvement in depression is strongly dependent on the improvement in EDS. Furthermore, EDS scores are also significantly correlated with depression scores [5,6], and improvement of EDS scores is shown to be modulated by the improvement of depressive symptoms [5][6][7][8][9][10]. A better understanding of this bidirectional relationship is therefore needed.
The relationship between EDS and depression may depend mostly on symptoms such as fatigue and changes in psychomotor activity as well as sleep patterns. Depression scales include heterogeneous symptom profiles of somatic, cognitive and affective categories.
In the current study, we aimed to explore the association between different symptom groups of depression and EDS over a 2-year period, which may have prognostic and therapeutic implications in this context.

Study Population
This was a secondary analysis of the RICCADSA trial. The RICCADSA study population has been previously described elsewhere [11]. In brief, adults with a history of percutaneous coronary intervention (PCI) or coronary artery by-pass grafting (CABG) within 6 months prior to study start were invited to participate in the Randomized Intervention with CPAP in Coronary Artery Disease and Obstructive Sleep Apnea (RICCADSA) trial. The patients were enrolled into the study between 2005 and 2010, and the follow-up for the primary outcomes was ended in May 2013 [12].
As illustrated in Figure 1, 399 patients met the inclusion criteria and were included in the current subanalysis. The participants were classified as OSA (apnea-hypopnea index [AHI] ≥ 15/h) and no-OSA (apnea-hypopnea index [AHI] < 5/h) based on the baseline sleep recordings with cardiorespiratory polygraphy. The cut-off for AHI (15/h) was chosen for the OSA diagnosis, given that the diagnostic procedure at baseline was based on cardiorespiratory sleep studies, not on gold standard polysomnography, and thus, the total sleep time could not be given exactly. This cut-off value was previously shown to be reliable for the OSA diagnosis whereas the cut-off AHI 5/h was reliable for excluding OSA [13]. Patients with AHI 5.0-14.9/h were considered as a borderline group, and were not included in the main trial ( Figure 1). The individuals with nonsleepy OSA phenotype (Epworth Sleepiness Scale [ESS] < 10) were assigned to CPAP or no-CPAP, and the sleepy OSA phenotype (ESS ≥ 10) received CPAP.

Cardiorespiratory Polygraphy
Baseline sleep studies were conducted by using the Embletta ® Portable Digital System device (Embla, Broomfield, CO, USA). As previously described in detail [11], the Chicago criteria were used for apnea and hypopnea definitions [14].

Epworth Sleepiness Scale
The ESS questionnaire [15] contains eight questions regarding the probability of dozing off under eight different situations. The maximum score is 24, and a cut-off value of 10 was used to define the EDS in the RICCADSA trial.

Zung Self-Rating Depression Scale
The Zung SDS is a well-recognized questionnaire, that gives both a total score, and a categorical rate of depressive mood [16]. In brief, the scale contains 20 questions (Table 1) with the self-rating scale ranging from 1 to 4 points (1 = a little of the time, 2 = some of the time, 3 = good part of the time, 4 = most of the time). All participants were requested to complete the Zung SDS questionnaire at baseline and at 1 and 2-year follow-ups. As the item 4 ("I have trouble sleeping at night") is a sleeprelated question, it was excluded from the factor analysis. Three factors are extracted based on the principal component analysis; F1, cognitive symptoms and anhedonia; F2, negative mood; and F3, appetite (Table 1).

Cardiorespiratory Polygraphy
Baseline sleep studies were conducted by using the Embletta ® Portable Digital System device (Embla, Broomfield, CO, USA). As previously described in detail [11], the Chicago criteria were used for apnea and hypopnea definitions [14].

Epworth Sleepiness Scale
The ESS questionnaire [15] contains eight questions regarding the probability of dozing off under eight different situations. The maximum score is 24, and a cut-off value of 10 was used to define the EDS in the RICCADSA trial.

Zung Self-Rating Depression Scale
The Zung SDS is a well-recognized questionnaire, that gives both a total score, and a categorical rate of depressive mood [16]. In brief, the scale contains 20 questions (Table 1) with the self-rating scale ranging from 1 to 4 points (1 = a little of the time, 2 = some of the time, 3 = good part of the time, 4 = most of the time). All participants were requested to complete the Zung SDS questionnaire at baseline and at 1 and 2-year follow-ups. As the item 4 ("I have trouble sleeping at night") is a sleep-related question, it was excluded from the factor analysis. Three factors are extracted based on the principal component analysis; F1, cognitive symptoms and anhedonia; F2, negative mood; and F3, appetite (Table 1).

Group Assignment, Randomization, Interventions and Follow-Up
As illustrated in Figure 1, all participants with sleepy OSA phenotype as well as the nonsleepy OSA phenotype, who were randomized to CPAP, were provided with a self-titrating CPAP device (S8 ® , or S9 ® ; San Diego, CA, USA) with a nasal or full-face mask and humidifier. All participants receiving CPAP were contacted by telephone after one week, and all patients were followed at 1 year and 2 years [11].

Adherence to CPAP
All patients receiving CPAP brought their devices to the clinic at each follow-up visit, CPAP hours/night and CPAP days/period were downloaded from the device, and accumulated CPAP hours/night/all nights were calculated. All technical adjustments were done according to the clinical routines by the sleep medicine unit staff.

Statistical Analysis
The study sample distribution of demographics and clinical characteristics at baseline was examined using the descriptive statistics. Continuous variables were reported as mean ± standard deviation, or median values with boundaries of the interquartile ranges (IQR), and the categorical variables as numbers and percentages. The chi-square test was used to compare the subgroups on the categorical variables, and for the within-group changes over time, nonparametric tests were used for the continuous variables (median with IQRs).
A principal component analysis (PCA) was conducted on the Zung SDS with orthogonal rotation (VARIMAX). The Kaiser-Meyer-Olkin Measure was used to test the sample adequacy. Bartlett's test was applied to check for the sphericity assumption. The primary item inclusion was set at 0.40 (absolute value). The number of factors were decided A linear mixed model was applied to examine individual changes on the estimated mean ESS scores in response to CPAP treatment among the CAD patients with OSA based on the procedure described by Heck et al. [17]. As illustrated in Figure 2, a two-level growth modeling approach was used to test the effect of "time" as the effects of CPAP treatment on the ESS. ESS scores, as the outcome, were measured three times for each patient to examine the pattern of change over the temporal period of study. The individual change was considered as the expected within-subjects effect for (level 1) whereas the shape of growth trajectory (time-related slope) was considered as the expected between-subjects effect (level 2). The "time" as a linear trend, coded as 0, 1 and 2, was entered into the model as a fixed effect as well as random effects for the intercept and slope. The covariance matrix of the random effects was assumed to be unstructured. The "time" as a quadratic trend, coded as 0, 1 and 4, was also added to the model as the fixed effect. The scores of F1, F2 and F3 were considered between-subject predictors. They were included to the mixed model as the fixed effects to explain possible differences in their individual intercepts and time-related slopes. Moreover, age, body-mass-index (BMI), AHI, CPAP use (hr/night) and comorbidities were tested for possible main effects and interactions. The statistical procedure was performed in the sleepy OSA group on CPAP treatment, and nonsleepy groups randomized to CPAP v.s. no-CPAP, respectively. All statistical tests were two-sided, and p < 0.05 was considered statistically significant. Statistical analysis was performed using the Statistical Package for Social Sciences, version 26.0 for Windows ® system (SPSS ® Inc., Chicago, IL, USA).

Principle Component Analysis
A principal component analysis was conducted on the 19 items with orthogonal rotation (VARIMAX). The Kaiser-Meyer-Olkin measure verified the sampling adequacy for

Principle Component Analysis
A principal component analysis was conducted on the 19 items with orthogonal rotation (VARIMAX). The Kaiser-Meyer-Olkin measure verified the sampling adequacy for the analysis, KMO = 0.884. Bartlett's test of sphericity χ 2 (190) = 2409.6, p < 0.001, indicated that correlations between items were sufficiently large for the PCA. Initially, an analysis was run to obtain eigenvalues for each factor in the data. Five factors had eigenvalues over 1 (Kaiser's criterion) and they explained 52.7% of the variance in the data. The scree plot was slightly ambiguous and showed inflexions. Therefore, another PCA was conducted by setting the number of factor extractions by 3. Table 1 shows the factor loading after rotation. Items 2 and 8 were excluded due to item inclusion criteria (> 40). The items that cluster on the same factors suggest that factor 1 represents cognitive symptoms and anhedonia, factor 2 negative affective symptoms and factor 3 appetite.

Baseline Characteristics of the Study Population
As shown in Table 2, the sleepy OSA patients were slightly younger and more obese and had higher AHI and ODI than the patients with non-sleepy OSA. Baseline comorbidities were similar.

Within-Group Differences Over Time
As expected, there was a significant decrease in ESS scores over time among the sleepy OSA phenotype on CPAP ( Figure 3). Interestingly, the decline in ESS scores was observed even in the nonsleepy phenotype on CPAP. ESS scores did not change significantly in the nonsleepy no-CPAP group. Average CPAP usage was 2.8 ± 2.9 h/night at 1 year, and 2.6 ± 2.9 h/night at 2 year follow-up in the nonsleepy group, and the corresponding values were 3.4 ± 2.6 h/night v.s. 3.3 ± 2.7 h/night in OSA patients with the sleepy phenotype.
To examine the linear trends on the Zung SDS subscales over time, the mean scores were computed among each group at baseline and follow-ups, respectively. As illustrated in Figure 4, there was a decreasing trend over time on the mean scores of F1 and F2 among both the sleepy and nonsleepy OSA patients on CPAP treatment (F1 14. AHI, apnea-hypopnea index; BMI, body mass index; ESS, Epworth Sleepiness Scale; ODI, oxygen desaturation index; OSA, obstructive sleep apnea; PCI, percutaneous coronary intervention; SDS, Self-rating Depression Scale. * p < 0.05.

Within-Group Differences Over Time
As expected, there was a significant decrease in ESS scores over time among the sleepy OSA phenotype on CPAP (Figure 3). Interestingly, the decline in ESS scores was observed even in the nonsleepy phenotype on CPAP. ESS scores did not change significantly in the nonsleepy no-CPAP group. Average CPAP usage was 2.8 ± 2.9 h/night at 1 year, and 2.6 ± 2.9 h/night at 2 year follow-up in the nonsleepy group, and the corresponding values were 3.4 ± 2.6 h/night v.s. 3.3 ± 2.7 h/night in OSA patients with the sleepy phenotype. To examine the linear trends on the Zung SDS subscales over time, the mean scores were computed among each group at baseline and follow-ups, respectively. As illustrated in Figure 4   Regarding the appetite subscale scores, there were significant within-group changes at 1-year and 2-year follow-ups compared to baseline in all subgroups (p = 0.04 vs. p = Regarding the appetite subscale scores, there were significant within-group changes at 1-year and 2-year follow-ups compared to baseline in all subgroups (p = 0.04 vs. p = 0.001 among the sleepy OSA patients on CPAP, p < 0.001 v.s. p = 0.002 among the nonsleepy OSA patients randomized to CPAP, and p = 0.03 vs. p = 0.01 among the patients randomized to no-CPAP).

Null Model
In the null models, the mean for the ESS scores, including baseline, after 1 year and 2 years, were 10.2 in the sleepy OSA group who received CPAP, and 5.5 in the non-sleepy OSA patients who were allocated to CPAP, and 5.2 in the non-sleepy OSA patients who were randomized to no-CPAP (Table 3). Within-individual variances (Level 1), which summarizes the population variability in average ESS scores around personal growth trajectory, was estimated around 9.9 for the sleepy OSA patients, 3.3 for the non-sleepy patients randomized to CPAP, and 3.3 for the non-sleepy patients with no CPAP, respectively. Between-individual variance (Level 2) was estimated 2.0 in the sleepy group on CPAP, 3.8 in the non-sleepy group on CPAP, and 4.1 in the non-sleepy group who were randomized to no CPAP (Table 3).

The Shape of Growth Trajectory
The fixed effects are presented in Table 4. The estimate was 12.2 points for the intercept (p < 0.001), −4.00 points for the linear growth rate (p < 0.001) and 1.1 points for the quadratic growth rate (p < 0.001). These results indicate that the time-dependent polynomials were significant predictors for the change in ESS, and thus, both the linear and quadratic polynomials should be retained in subsequent analyses in sleepy OSA patients on CPAP. Therefore, it was decided that the time-related slopes were included to the next mixed model. As presented in Table 4, the estimate of the intercept for ESS at baseline was 5.5 in the non-sleepy OSA group randomized to CPAP as well as in the no-CPAP group (p < 0.001 for both). However, the linear and quadratic polynomials were not significant predictors of mean ESS scores (Table 4), and convergence has not been achieved. Thus, there was no need to run further analyses in the nonsleepy OSA patients.

Varying Time-Related Slope in the Sleepy OSA Patients on CPAP
As demonstrated in Table 5, the estimate for Level 1 was 4.9 (p < 0.001). At Level 2, random intercept for the between-individual variance was 2.1 (p = 0.02). Furthermore, the estimate for the random slope (linear polynomials) was 1.2, which also significantly varied between individuals (p = 0.04). However, the estimate of the covariance between the random slope and intercept was not statistically significant. Table 5. The random effect of the "time" as a linear trend in the model.

Adding the Between-Subject Predictors
In order to explain variability between individuals, F1, F2 and F3 were included to the model, respectively, followed by the predictors age, gender, BMI, CPAP hours, AHI, ODI and comorbidities.
The fixed effects are presented in Table 6. The estimate of the intercept for ESS was 7.1 (p = 0.006) in the sleepy OSA group. The estimate for the fixed effect was 0.28 points for the F2 (p < 0.001), −2.88 points for the linear growth rate (p < 0.001), and 0.6 points for the quadratic growth rate (p = 0.03). For the rest of the predictors, the mixed model showed no significant associations, and these predictors were therefore removed from further analyses.

Final Mixed Models in Sleepy OSA Patients on CPAP
The estimate of the intercept for ESS was 8.9 points, 0.26 points for the F2 (p = 0.001). The time related polynomials were −3.4 and 0.87 points, respectively (p < 0.001 for both). The within-individual variance was 4.67 points (p < 0.001) while the random intercept for the between-individual variance was 2.16 points (p = 0.01) ( Table 7).

Discussion
The main finding of the current study was that improvement in negative mood was significantly associated with improvement in EDS over a 2-year period on CPAP among the CAD patients with sleepy OSA at baseline. The changes in cognitive distortions and anhedonia as well as in appetite were not related. Previous reports were focused on the association between total depression scores and EDS [18]. More recently, we have demonstrated the beneficial effects of CPAP on depression in the RICCADSA cohort, especially among the sleepy OSA phenotype [7]. The current paper revealed that the improvement in EDS was mainly based on the improvement in negative mood independent of age, BMI, AHI at baseline and CPAP use (h/night).
This finding is unique as it shows that the overlap of depression and EDS is mainly through negative mood, instead of anhedonia as the loss of ability to get joy from activities, cognitive distortions and thought content as feeling useless. It is reasonable that a patient with negative mood may be less likely to be adherent to CPAP treatment and get less improvement in EDS as opposed to a non-depressed patient. It is also plausible that adding an intervention (i.e., CPAP in this case) may improve negative mood via a placebo effect. Notwithstanding, adding interventions to improve mood in CAD patients with sleepy OSA may promote additional benefits in improvement in EDS, probably by concomitant improvement in adherence to CPAP treatment.
As extensively discussed in our previous publications, EDS is one of the most important symptoms of OSA, and there seems to exist an interaction between EDS and depression in OSA patients (6,7). The beneficial effects of CPAP treatment for EDS are already wellknown [19]. We demonstrated improvements in both SDS and ESS scores following 1 year of CPAP treatment in both phenotypes [7]. Moreover, there was a significant relationship between CPAP use in h/night and improvement in mood in multivariate analysis adjusted for confounding factors, and the cut-off value of CPAP usage for improvement in mood was 4 h/night when the ESS score change from baseline was entered into the model [7].
In the current study, we did not perform subanalyses within the CPAP usage categories since our aim was to explore the association between the SDS subscales and EDS, in which CPAP usage was included as one of the predictors in the final model. However, it was not a significant predictor in this subgroup of patients with the sleepy OSA phenotype. Though not related to the improvement in EDS, there was a significant decrease in cognitive symptoms and anhedonia scores as well as appetite scores among OSA patients who were treated with CPAP. These results suggest that our previous reports, showing a significant decrease in the total Zung SDS scores (6,7) were covering all subscales, though only the changes in negative mood were significantly associated with the changes in EDS among the sleepy OSA phenotype.
The reduction in appetite scores (increased appetite and weight gain) in the whole cohort was indeed interesting. This might explain a natural course of reduced depressive mood during the recovery period following a revascularization. This may also partly explain weight gain observed among patients with OSA following CPAP treatment [20].
To the best of our knowledge, association of sub-dimensions of depression with OSA as well as with EDS has yet not been studied in neither general population nor in sleep clinic cohorts, and nor in cardiac populations. The extent of the decline in negative mood and cognitive symptoms scores was significantly greater among both sleepy and nonsleepy OSA patients who were on CPAP treatment compared to CAD patients with untreated OSA. Taken together, improving a patient's EDS appears to be crucial for improvement in negative mood in CAD patients with OSA. Similarly, adding interventions to improve mood may improve EDS in CAD patients with OSA.
The current study has several strengths and some limitations of note. To the best of our knowledge, this is one of the largest existing studies in CAD patients with OSA on CPAP treatment addressing the association between sub-dimensions of depression and EDS changes over a 2-year period. As previously shown in a meta-analysis [21], the occurrence rates of depression range from 5% to 47% in cardiac patients. It has been suggested that these differences may reflect the various methodological instruments used in the evaluation of depression [22]. The National Heart, Lung, and Blood Institute (USA) had therefore recommended an investigation to address whether or not depression questionnaires used in the general population were applicable to cardiac patients [23]. In this context, a study by Perez Adel et al. [24] supported the use of the Zung SDS scores with justifications such as (A) it was a widely used self-questionnaire [25]; (B) it was easy to understand [26]; and (C) it predicted mortality and new hospitalizations in cardiac failure patients [27] as well as the risk for cardiovascular death in adults free of CAD at baseline [28]. It has also been argued that the use of total scores of a questionnaire may be problematic since different clusters of symptoms might be underestimated [23], and it might be difficult to diagnose and treat various types of depression. Here, by conducting a two-level linear mixed model, we could explore the association between the subscales of Zung SDS and EDS changes over a 2-year period. The long-term follow-up is an additional strength of the study. We should also acknowledge our study limitations. Firstly, the depression diagnosis was not validated with a structured clinical interview. Since the analyses are mostly related to self-reported questionnaires, we cannot exclude inherent biases. Secondly, sleepy and nonsleepy OSA patients were categorized based on the ESS threshold, which may not be precise. However, this questionnaire is a largely accepted tool used in clinical cohorts. Finally, the patients with sleepy OSA phenotype were more obese and had more hypoxemia than the patients with nonsleepy OSA at baseline. However, these variables were included in the multivariate analysis as covariates but they were not significant predictors of EDS in this cohort.

Conclusions
We conclude that improvement in negative mood symptoms of depression, but not changes in cognitive symptoms and anhedonia as well as appetite, was associated with the improvement in EDS in the sleepy RICCADSA cohort, who were treated with CPAP. Moreover, the magnitude of decline in negative mood and cognitive symptoms scores was significantly greater among both sleepy and nonsleepy OSA patients who were on CPAP treatment. Thus, improving a patient's EDS seems to be important to achieve improvement in negative mood in CAD patients with sleepy OSA. Likewise, adding interventions to CPAP treatment of OSA to improve mood may promote CPAP adherence, and, similarly, better adherence to CPAP treatment and alleviation of the underlying OSA may improve mood in patients with CAD.