Abstract
A recent meta-analysis identified a prospective association between depression and cardiovascular disease; however, there was no association for studies with long-term follow-up periods. The literature has primarily focused on baseline depression status or symptoms, which may not capture the chronic nature of depression. This study examined the prospective relationship between depressive symptoms and cardiovascular disease up to 15 years later in 274 cardiovascular disease–free older adults. Depressive and anxiety symptoms, mean arterial pressure, and cardiovascular disease status were assessed. Baseline and chronic depressive symptoms predicted increased risk of cardiovascular disease, underscoring the importance of assessing and treating depression in older adults.
Introduction
Cardiovascular disease (CVD) and depression are among the leading causes of global disease burden (Ferrari et al., 2013; Kassebaum et al., 2016; Roth et al., 2015). By 2030, it is projected that 43.9 percent of the US adult population will have some form of CVD, which is associated with US$1.2 billion in direct and indirect costs (Benjamin et al., 2017). Depression is a highly prevalent mental illness associated with significant economic burden. In 2015, an estimated 6.7 percent of adults in the United States experienced a depressive episode (Center for Behavioral Health Statistics and Quality, 2016). Depression was associated with US$210.5 billion in direct and indirect costs in 2010 (Greenberg et al., 2015). Although decades of research, including findings from epidemiological studies, have found associations between depression and CVD (Carney and Freedland, 2016), questions remain about the temporal nature of this relationship.
A recent meta-analysis of 30 prospective cohort studies found that depression was associated with an increased risk of coronary heart disease (CHD) and myocardial infarction (MI; Gan et al., 2014). To examine the role of depression as a pre-morbid risk factor, studies were included only if participants were free of CHD at study entry. Depression was associated with an estimated pooled relative risk (RR) of 1.30 (95% confidence interval (CI) = 1.22–1.40) for CHD and an estimated pooled RR of 1.30 (95% CI = 1.18–1.44) for MI. Associations remained after adjustment for smoking, body mass index (BMI), hypertension, diabetes, physical activity, and socioeconomic status, which suggests that depression is an independent risk factor for some types of CVD including CHD and MI.
Additional subgroup analyses revealed no statistical association between depression and risk of CHD among studies that had a 15-year or greater follow-up period (Gan et al., 2014). One explanation for the attenuated risk between depression and CHD during longer follow-up periods is that participants pursue treatment, recover, or develop depression after baseline, which has classification implications (Almas et al., 2015). Currently, there are only a small number of published prospective studies with a minimum of 15 years of follow-up (Barefoot and Schroll, 1996; Brown et al., 2011; Brunner et al., 2014; Gump et al., 2005; Janszky et al., 2010).
In a 37-year follow-up study, Janszky et al. (2010) examined the prospective relationship between depression and CHD in a sample of Swedish young males. Exposure data were provided by a nation-wide conscription survey, while outcome data were obtained from the Swedish Health registrar. It was found that a diagnosis of anxiety, but not depression, was associated with cardiac events during follow-up. One limitation of this study is that exposure to depression and anxiety was only assessed once at baseline. It is possible that participants developed depression after baseline and were therefore misclassified (Almas et al., 2015; Brunner et al., 2014; Péquignot et al., 2016).
Results from a 27-year follow-up study of older adults revealed that elevated scores on a depression index from the Minnesota Multiphasic Personality Inventory (MMPI; Hathaway and McKinley, 1943) were associated with increased risk of acute MI (Barefoot and Schroll, 1996). It is unclear whether anxiety, which is highly comorbid with depression, contributed to the relationship between depression and CVD. Indeed, research suggests that anxiety may be associated with incident CVD (Batelaan et al., 2016).
Data from the Whitehall II cohort study were used to examine depressive symptoms and future major CHD and stroke events in adults over a 24-year follow-up period (Brunner et al., 2014). Frequency of depressive symptoms measured by the General Health Questionnaire-30 Symptom Scale (GHQ-30; Goldberg, 1972) during the two-decade follow-up period revealed a dose–response relationship with CHD but not with stroke events. However, the GHQ-30 is a general measure of psychological distress, capturing both anxiety and depression, that only had moderate agreement with the Center for Epidemiologic Studies Depression Scale (CES-D; Radloff, 1977) collected at one phase of the study. While the dose–response relationship between depression and future CHD was found, only age, sex, and ethnicity were controlled for in the analyses. It is important that variables commonly associated with risk for CVD (e.g. smoking status and BMI) are taken into account.
An 18-year prospective study followed a sample of men at high risk for CHD after they participated in a multi-year intervention trial to reduce risk of future CHD (Gump et al., 2005). Depression symptoms measured during the sixth year of the trial were associated with higher risk of all cause and CVD mortality, specifically stroke 18 years later (Gump et al., 2005). Findings were obscured by the fact that the sample was restricted to men at high risk for CVD, some of whom (22%) had already developed CVD by the beginning of the 18-year post-trial follow-up period.
Finally, the 15-year prospective relationship between depressive symptoms and risk for MI or CHD was examined in a US sample of older patients, including a large number of women and minorities (Brown et al., 2011). Older adults with significant depressive symptoms at baseline were approximately 1.5 times more likely to suffer a CHD event (i.e. acute MI or CHD-related death) over the 15 year follow-up period, even after controlling for demographics and known cardiovascular risk factors. Findings are limited by the single baseline measurement of depression and lack of information about anxiety symptoms, which tend to co-occur with depression and may contribute to the relationship between depression and CHD.
The aforementioned prospective studies provide important data points toward clarifying the relationship between depression and future CVD. However, given the small number of published prospective studies and the fact that only one study examines the potential confounding role of anxiety, more research is needed to clarify the long-term prospective relationship between depression and CVD.
The aim of this study was to examine the prospective relationship between depressive symptoms and cardiovascular events, while addressing the possible influence of anxiety and other covariates related to CVD risk (i.e. age, sex, education, and mean arterial pressure (MAP)). Analyses were based on data from a population of initially CVD-free subjects. In comparison to baseline exposure, it was hypothesized that cumulative exposure to depressive symptoms would have a more pronounced association with CVD at follow-up.
Methods
Participant selection
Data were drawn from the Maine-Syracuse Longitudinal Study (MSLS), a 36-year longitudinal study of aging, cardiovascular risk factors, and cognitive functioning that has collected seven waves of data. The MSLS sample consists of community-dwelling adults 23–98 years of age living in Central New York. Since its inception in 1975, the MSLS has excluded participants based on diagnosed alcoholism, diagnosed psychiatric disorder which precluded completion of test battery, or an inability to speak English. This and other MSLS studies have applied additional exclusionary criteria depending on the research questions of interest. For this study, participants who endorsed the presence of CVD at wave 4 (W4; 1990–1996) or W5 (1996–2001) of the MSLS were excluded from this study to examine the developmental trajectory of CVD. The University of Maine Institutional Review Board approved this study (reference number: 2005-07-04), and informed consent was obtained from all participants.
Attrition
Our sample consists of eligible participants who completed measures of depression and anxiety at either W4 or W5. After exclusion for the presence of CVD at W4 or W5 (n = 16), 419 older adults were eligible for study. Of those participants whom were eligible at baseline, 65.4 percent participated in data collection during W5–W7 (W7; n = 274).
Of the remaining eligible participants, those retained at W5–W7 follow-up did not differ significantly by age, ethnicity, sex, education, MAP, or chronic depressive and trait anxiety symptoms (p values > .10). Table 1 presents descriptive statistics for participants retained at follow-up, including demographics, mental health, cardiovascular health, and other health variables. Of note, variables were collected at different waves of the study. Superscripts have been used in the table to denote the wave that each variable was collected.
Demographic and health variables by gender.
HDL: high-density lipoprotein; LDL: low-density lipoprotein; BP: blood pressure; METs: metabolic equivalents.
Self-rated health score = how would you describe your health now? very poor = 1; poor = 2; average = 3; above average = 4; excellent = 5. Hypertensive (%) = ever on medication for hypertension, systolic blood pressure above 140, or diastolic blood pressure above 90.
Data present at either waves 4 or 5.
Data averaged across waves 4 and 5 (baseline).
Data available from wave 6 (follow-up).
Data available from wave 7 (follow-up).
p < .05, **p < .01, ***p < .001 refers to differences in category prevalence by gender.
Study design
This study used mean depressive symptoms and covariate information from W4 (1990–1996) and W5 (1996–2001) of the MSLS as baseline data. CVD status at W6 (2001–2006) and W7 (2006–2011) of the MSLS was used as follow-up data.
Measures of depressive symptoms and CVD
Depressive symptoms were assessed using the Zung Depression Inventory (ZDI; Zung, 1965). Chronic depressive symptoms were calculated by taking the mean ZDI ratings at W4–W5. The ZDI is a 20-item self-reported questionnaire designed to assess depressive symptom severity within the past week. Responses are scored on a 4-point Likert-type scale ranging from 1 “a little of the time” to 4 “most of the time.” Total raw scores are calculated by summing all the items resulting in an index score range from 20 to 80. Missing items were mean-imputed from available ZDI data. Participants missing greater than two items were excluded from analysis. Higher total scores indicate higher depressive symptom severity. At W4 and W5, the ZDI items showed good internal consistency (W4 α = .80 and W5 α = .81).
Information regarding CVD status was obtained through a researcher-administered medical interview. The medical interview contained questions regarding the presence and absence of MI, coronary artery disease (CAD) with and without MI, heart failure, angina pectoris, transient ischemic attack (TIA), and acute stroke that were posed by a trained researcher. Acute stroke was defined as a focal neurological deficit of acute onset persisting more than 24 hours. Cardiovascular events were only considered present if the participant’s self-report was confirmed by his or her medical records.
Covariates
Covariates were selected based on their association to depression and CVD. Adjustments were made for age, sex, education in years, MAP, antidepressant use, disability status, and chronic trait anxiety symptoms. Age, education in years, and mean MAP were calculated by taking an average of W4–W5 ratings. Specifically, MAP was calculated by taking the mean of 15 blood pressure measurements, five obtained while the patient was in each of the sitting, standing, and recumbent positions, which was used as an estimate of baseline cardiovascular functioning. Information regarding antidepressant use was collected from self-reported prescription medication use. Antidepressant use was rated as present at W4 or W5. Information regarding disability status was based on self-reported occupational status. Disability status was rated as present if participants endorsed the presence of disability at W4 or W5.
Mean chronic trait anxiety symptoms were calculated using W4 and W5 trait anxiety scale scores from the State-Trait Personality Inventory (STPI; Spielberger, 1996). The trait anxiety scale of the STPI consists of 10 self-reported items designed to assess anxiety symptom severity that an individual generally experiences. Responses are scored on a 4-point Likert-type scale ranging from 1 “almost never” to 4 “almost always.” Total raw scores are calculated by summing all the items. Missing items were coded as “2.” Participants missing all items were excluded from analysis. Higher total scores indicate higher anxiety symptom severity. At W4 and W5, the STPI trait anxiety items showed good internal consistency (α = .80).
Pulse wave velocity (PWV), BMI, smoking status, and alcohol consumption were considered as covariates because previous research has indicated that they are associated with an elevated risk for CVD. PWV was measured at W7 only and used as an estimate of current cardiovascular functioning at follow-up. BMI was calculated (kg/m2) from height and weight measurements at W4 and W5. The mean BMI was used to approximate physical health at baseline. However, PWV, BMI, smoking status, and alcohol consumption were dropped from the final models due to a lack of association with either depressive symptoms or CVD. Their inclusion did not change the results reported below.
The possible influence of antidepressant use and disability status was explored in separate analyses. Antidepressant use at baseline was not associated with either depressive symptoms or CVD. However, antidepressants at W4 and W5 were associated with W6–W7 cardiovascular events. Disability status at baseline was not associated with depressive symptoms but was associated with CVD in all models.
Statistical analysis
Hierarchical multivariate logistic regression models were used to investigate the longitudinal association between depressive symptoms (baseline and cumulative) and CVD with adjustment for covariates.
Three separate sets of analyses were conducted to examine depressive symptoms in the prospective prediction of CVD. In each set of analyses, four hierarchical regression models were employed with W7 CVD as the outcome of interest in the first analysis and W6–W7 CVD as the outcome of interest in the second and third analyses. In the first analysis, ZDI scores from W4 were employed as the index of symptoms of depression and constituted Model 1. In Model 2, the covariate W4 trait anxiety was added to Model 1. Model 3 included all covariates from Model 2 in addition to sex, mean age, and education in years from W4 to W5. Model 4 included all covariates from Model 3 in addition to mean MAP from W4 and W5. At each step, inspection of the models indicated appropriate fit to the data (Hosmer–Lemeshow tests: p values > .05).
In the second analysis, the same set of hierarchical multivariate regression models was used with W6 to W7 CVD status as the outcome of interest. At each step, inspection of the models indicated appropriate fit to the data (Hosmer–Lemeshow tests: p values > .05).
The final corresponding analysis was conducted to examine cumulative depressive symptoms predicting CVD. W6–W7 CVD status was the outcome of interest and mean ZDI scores from W4 to W5 composed Model 1 and mean trait anxiety symptoms from W4 to W5 was added in Model 2 as a covariate. Model 3 included all covariates from Model 2 in addition to sex, mean age, and education in years from W4 to W5. Model 4 included all covariates from Model 3 in addition to mean MAP from W4 and W5. At each step, inspection of the models indicated appropriate fit to the data (Hosmer–Lemeshow tests: p values > .05).
Analyses with the inclusion of a quadratic term for mean ZDI scores were conducted to explore curvilinearity. Within each set of analyses, the mean ZDI quadratic term was added in Model 1.
Statistical analyses were conducted with SPSS software (version 23.0). Graphical interpretation was conducted in R (version 3.4.3) using the ggplot2 package.
Results
Demographics
Demographic and health-related variables are presented in Table 1. By W7, the total number of individuals who were diagnosed with at least one CVD was 51 (n = 10 MI, n = 8 CAD with MI, n = 19 CAD without MI, n = 11 heart failure, n = 16 angina pectoris, n = 12 TIA, and n = 13 acute stroke).
Model analyses
Results of analyses for multivariate logistic regression models are presented in Table 2. Of note, sample sizes varied by analyses depending on which waves were utilized. Sample sizes are reported in each section of the table. For all models, Naglekerke R2 was used to assess model fit.
Relationship between depressive symptoms and CVD.
STPI: State-Trait Personality Inventory; MAP: mean arterial pressure; OR: odds ratio; CI: confidence interval; ZDI: Zung Depression Inventory.
p < .05, **p < .01, ***p < .001.
Baseline depressive symptoms and CVD outcome at 15 year follow-up
Regression models examining the longitudinal association between depressive symptoms at baseline (W4) and CVD status 15 years later (W7) included adjustments for covariates. The final model accounted for 21.5 percent of the total variance. Across all models, odds ratios (OR) ranged from 1.09 to 1.12. W4 ZDI predicted elevated risk for CVD at W7 in all four models (p values < .05). At each step, the ZDI quadratic term was non-significant (p values > .14). Results are presented in Table 2 and visually represented in Figure 1.
Relationship between depressive symptoms at W4 and predicted likelihood of CVD at W7.
Baseline depressive symptoms and CVD outcome at 10- to 15-year follow-up
Regression models examining the longitudinal association between depressive symptoms at baseline (W4) and CVD status 10–15 years later (W6–W7) included adjustments for covariates. The final model accounted for 23.2 percent of the total variance. Across all models, OR ranged from 1.09 to 1.13. W4 ZDI predicted elevated risk for CVD at W6–W7 in all four models (p values < .01). At each step, the ZDI quadratic term was non-significant (p values > .43). Results are presented in Table 2.
Chronic depressive symptoms and CVD outcome at 5- to 10-year follow-up
Regression models examining the longitudinal association between mean chronic depressive symptoms across 10 years (W4–W5) and CVD status 5–10 years later (W6–W7) included adjustments for covariates. The final model accounted for 27.3 percent of the total variance. Across all models, OR ranged from 1.11 to 1.18. W4–W5 mean ZDI predicted elevated risk for CVD at W6–W7 across all four models (p values ⩽ .001). At each step, the ZDI quadratic term was non-significant (p values > .44). Results are presented in Table 2.
In addition to mean ZDI, the covariate of W4–W5 mean chronic trait anxiety symptoms predicted increased risk for CVD at follow-up. In the mean STPI model, higher levels of trait anxiety symptoms were associated with lower rates of CVD at follow-up in Model 2 (OR = 0.89; CI = 0.80–0.99; p < .05). This association was no longer significant when adjusting for demographic variables in Model 3 (OR = 0.94; CI = 0.84–1.06; p > .31) and W4–W5 mean MAP Model 4 (OR = 0.94; CI = 0.84–1.06; p > .30).
Chronic depressive symptoms and CVD outcome at 5- to 10-year follow-up by gender
Finally, because the sample was composed of both males (n = 115) and females (n = 159), regression models were used to examine if the association between depressive symptoms and CVD is comparable among genders. The final model accounted for 28.9 percent of the total variance. Across all models, OR ranged from 1.15 to 1.21. W4–W5 mean ZDI predicted elevated risk for CVD at W6 to W7 across all four models (p values < .001). In the first two models, the ZDI by gender term was significant (p values < .01), with the OR equaling .97. However, in the last two models, the ZDI by gender term became non-significant (p = 0.058 at step 3 and p = .079 at step 4), with OR ranging from .90 at step 3 and .91 at step 4. Results are visually represented in Figure 2.
Relationship between depressive symptoms at W4–W5 and predicted likelihood of CVD at W6–W7 by gender.
Discussion
The goal of this population-based prospective cohort study was to examine the length of exposure and outcome period on the relationship between depressive symptoms and prospective cardiovascular events. Depressive symptoms measured on one occasion at baseline (W4) were associated with cardiovascular events at 15 years of follow-up (W7). These findings are inconsistent with a recent meta-analysis, which did not find a statistically significant association between depression and prospective CHD among studies with a 15-year or greater follow-up period (Gan et al., 2014). Analyses also revealed that baseline depressive symptoms (W4) were more strongly associated with cardiovascular events measured over a 10-year exposure period, at 10- to 15-year follow-up (W6–W7). These results might be attributable to a shorter follow-up (10–15 years) and longer (10 year) exposure period.
With respect to chronic depressive symptoms, results suggest that the chronic impact of depressive symptoms over time is associated with future cardiovascular events. Chronic depressive symptoms (W4–W5) predicted cardiovascular events during a 10-year outcome period, at 5- to 10-year follow-up (W6–W7). This association was the most robust of all analyses with depressive symptoms accounting for 12.7 percent of the variance associated with cardiovascular events. Importantly, all significant associations between depression and CVD remained so after adjusting for baseline depressive symptoms, chronic trait anxiety symptoms, age, sex, education, and MAP, with the final model accounting for 27.3 percent of the variance.
The interaction between chronic depressive symptoms and gender on CVD was explored. While the occurrence of CVD at W6–W7 did not significantly differ by gender, it is speculated that the pattern of results may suggest that males experience a higher likelihood of CVD at lower levels of self-reported depression. More research with a larger sample size is needed before definitive conclusions can be drawn.
Implications
Depression is an important cardiovascular risk factor with a large impact on public health. The present results, which found that both baseline and chronic depressive symptoms prospectively predict CVD, provide additional evidence in support of depression as a causal risk factor for CVD. Indeed, evidence exists for the mediating role of several biological and behavioral mechanisms related to depression and CVD (e.g. unhealthy lifestyle, metabolic, immune-inflammatory, autonomic, and hypothalamic-pituitary-adrenal axis dysregulation; refer Penninx, 2017 for a review). Research is needed to clarify the extent to which these mediating mechanisms contribute to increased risk for CVD.
The majority of past longitudinal studies on the relation between depression and CVD have measured depression on one occasion. This study filled an important gap in the literature by examining the prospective impact of both baseline and chronic depressive symptoms on cardiovascular events across a 15-year follow-up period. More research is needed to identify the most effective health-care intervention for identifying and treating depression with the objective of reducing the impact of CVD.
Strengths
The study has several strengths, including a community-based cohort design consisting of participants who were initially free of CVD as determined by a medical interview and record review. A broad range of CVD and stroke outcomes were considered. This study has addressed several potential confounding factors in the association between depression and CVD; MAP was included as a covariate rather than self-report hypertension, which is often prone to recall errors (Gorber et al., 2008). Similarly, the inclusion of trait anxiety symptoms as a covariate allowed for examination of the specificity of depressive symptoms in the prediction of CVD. In addition, the association between antidepressant use and cardiovascular events was assessed and found that antidepressant use at W4–W5 was associated with cardiovascular events at W6–W7. Some observational studies have shown that antidepressant use is associated with increased risk of CVD (Hippisley-Cox et al., 2001; Nabi et al., 2010). An explanation for this finding is that antidepressant use may be confounded by depression severity. As such, it is not clear whether antidepressants contribute to the development of CVD. Finally, repeated measurements permitted examination of exposure risk, length of follow-up, and subsequent CVD.
Limitations
With respect to limitations, the sample was predominately Caucasian and well-educated. The extent to which results generalize to individuals from other racial, ethnic, and educational backgrounds is unclear. In a similar vein, results are based on self-report measurement of depressive symptoms in a relatively healthy population. It is unknown if the findings extend to a clinical sample diagnosed with major depressive disorder according to the Diagnostic and Statistical Manual of Mental Disorders, 5th Edition: DSM-5 (American Psychiatric Association (APA), 2013). Nonetheless, research suggests that subclinical levels of depressive symptoms, particularly in older adults, is an important phenomenon associated with lower quality of life, physical health decline, and increased health-care burden (Meeks et al., 2011). Finally, results revealed that disability status was related to CVD but not depressive symptoms. An important caveat is that these findings were based on a crude measure of disability (occupation related). Future research should include measures of activities of daily living to more accurately examine the relationship between disability, depressive symptoms, and CVD.
Conclusion
To date, very few prospective long-term (15 years or more) studies investigating the association between depression and CVD have been published. These results contribute to the limited evidence base and suggest that baseline and chronic depressive symptoms are associated with future cardiovascular events at 10- to 15-year follow-up. Notably, results remained after adjustment for symptoms of trait anxiety symptoms and other variables associated with increased CVD risk. These findings contribute to the growing body of evidence that suggests depression is an independent risk factor for cardiovascular events.
Footnotes
Acknowledgements
Olivia E Bogucki and Peter J Dearborn contributed equally to this study.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The Maine-Syracuse Longitudinal Study was supported by grants from the National Heart, Lung, and Blood Institute (grants no: R01HL67358 and R01HL81290) and a research grant from the National Institute on Aging (grant no: R01AG03055). The content in this paper does not necessarily reflect the official views of the National Institutes of Health.
