Depressive Symptoms and Risk of Stroke in a National Cohort of Black and White Participants From REGARDS

Abstract

Objective

The purpose of this study was to examine depressive symptoms as a risk factor for incident stroke and determine whether depressive symptomatology was differentially predictive of stroke among Black and White participants.

Methods

The study comprised 9,529 Black and 14,516 White stroke-free participants, aged 45 and older, enrolled in the REasons for Geographic and Racial Differences in Stroke (2003–2007). Incident stroke was the first occurrence of stroke. Association between baseline depressive symptoms (assessed via the 4-item Center for Epidemiologic Studies Depression Scale [CES-D-4]: 0, 1–3, or ≥4) and incident stroke was analyzed with Cox proportional hazards models adjusted for demographics, stroke risk factors, and social factors.

Results

There were 1,262 strokes over an average follow-up of 9.21 (SD 4.0) years. Compared to participants with no depressive symptoms, after demographic adjustment, participants with CES-D-4 scores of 1–3 had 39% increased stroke risk (hazard ratio [HR] = 1.39, 95% confidence interval [CI] = 1.23–1.57), with slight attenuation after full adjustment (HR = 1.27, 95% CI = 1.11–1.43). Participants with CES-D-4 scores of ≥4 experienced 54% higher risk of stroke after demographic adjustment (HR = 1.54, 95% CI = 1.27–1.85), with risk attenuated in the full model similar to risk with 1–3 symptoms (HR = 1.25, 95% CI = 1.03–1.51). There was no evidence of a differential effect by race (p = 0.53).

Conclusions

The association of depressive symptoms with increased stroke risk was similar among a national sample of Black and White participants. These findings suggest that assessment of depressive symptoms should be considered in primary stroke prevention for both Black and White participants.

Stroke is a leading cause of death and disability in the United States. A variety of stroke risk factors have been identified and consistently confirmed.¹ A growing body of research suggests that depressive symptoms may also increase stroke risk.^2,3 A 2011 meta-analysis demonstrated a positive association between depressive symptomatology and future risk of stroke. The association was present in crude analyses and after adjustment for age, medical comorbidities, and other common risk factors.⁴ Meta-analyses in 2012 and 2015 also indicated that risk for stroke was increased by depression,^5,6 noting that this relationship was likely independent of other risk factors such as diabetes and hypertension.⁶

A notable limitation of previous research is the lack of racial and ethnic diversities among the study populations.^7,8 The results have been mixed among the few investigations that enrolled Black participants and examined race and depressive symptoms in relation to incident stroke.^2,9–12 Previously, we found that approximately 50% of the Black-White difference in incidence is attributable to “traditional” stroke risk factors, with 4 factors likely contributing the remaining 50%: differential susceptibility, residual confounding, nontraditional risk factors, and measurement error.¹³ One of these nontraditional risk factors may be depression or depressive symptoms. Depression often goes undetected and undiagnosed in Black participants, who are frequently less likely to receive effective care and management.^14,15

The purpose of this study was to examine depressive symptoms as a risk factor for incident stroke in a general population sample of Black and White participants and determine whether depressive symptomatology was differentially predictive of stroke between these groups.

Methods

Design and Participants

The REasons for Geographic and Racial Differences in Stroke (REGARDS) study is a national, population-based longitudinal study designed to examine risk factors associated with Black-White and regional disparities in stroke incidence and mortality. From January 2003 to October 2007, a total of 30,239 Black and White men and women, aged 45 years and older were enrolled. Participants were randomly sampled through mail and telephone with a goal of recruiting 30% of participants from the “Stroke Belt” to ensure representation of individuals from states—including Alabama, Arkansas, Georgia, Louisiana, Mississippi, North Carolina, South Carolina, and Tennessee—with the highest stroke mortality rates. Data on demographic and psychosocial characteristics, lifestyle patterns, and other stroke risk factors were collected via telephone interview. Physiologic measures and written informed consent were obtained during a subsequent home visit. Ongoing follow-up is conducted through biannual telephone calls to participants or proxies identified by the participants at the time of enrollment. The institutional review boards of institutions involved as REGARDS collaborators reviewed and approved the study methods. The telephone response rate, defined according to American Association for Public Opinion Research standards, was 33%, and the cooperation rate was 49%, rates similar to those in other cardiovascular cohort studies.^2,3,16–18 Additional details regarding the study design and data collection procedures have been published elsewhere.^19–21 The sample for the present study consisted of 24,045 participants who were stroke free at baseline, had complete data on depressive symptoms and all covariates, and were followed for potential stroke events.

Standard Protocol Approvals, Registrations, and Participant Consents

Consent was obtained initially by telephone and later in writing during the in-person evaluation. The institutional review boards of participating institutions approved the study methods.

Depressive Symptoms

Depressive symptoms were assessed using the 4-item version of the Center for Epidemiologic Studies Depression Scale (CES-D-4)^22,23 administered during the baseline evaluation of each participant. The four-item scale evaluates a subset of symptoms from the larger measure and assesses how often respondents felt depressed, sad, lonely, or had crying spells. Scores on the CES-D-4 range from 0 to 12, with higher scores reflecting a higher level of depressive symptoms. Scoring guidelines for the CES-D-4 recommend a cut point of 4, with scores of 4 or greater indicating depressive symptomatology. The CES-D-4 correlates well with the full CES-D (α = 0.87) and has demonstrated adequate sensitivity (79%) and specificity (87%) in comparison to the full scale.²³

Stroke Incidence

Incident stroke was defined as the first occurrence of a physician-adjudicated stroke. Potential strokes were identified through follow-up calls made to participants at 6-month intervals after the baseline interview. At least 2 members of a physician committee independently reviewed medical and death records obtained after a reported or suspected stroke event. When deaths occurred with no relevant records available, committee members evaluated suspected events based on completed or informant questionnaires.^19,24 This analysis included World Health Organization-defined stroke²⁵ or clinical stroke meeting the American Heart Association criteria,²⁶ ischemic and hemorrhagic, reported as of October 2017.

Baseline Variables

Additional variables assessed at baseline included demographic characteristics, traditional stroke risk factors, and social factors. Geographic location was identified based on the participant's address at the time of enrollment. Participant's age, sex, race, marital status, education, household income, cigarette smoking status, use of medications for diabetes and hypertension, and social network were obtained by self-report. Participants were also asked about their history of physician-diagnosed atrial fibrillation, stroke, or heart disease. During the in-person evaluation, physical measurements, an ECG, a fasting blood sample, and a medication inventory were obtained.

Marital status categories included married, divorced, widowed, never married, and other. Education level was categorized as less than high school, high school graduate, some college, or college graduate and above. Income was categorized as less than $20K, $20K–$34K, $35K–$74K, $75K and above, or not reported. Social network was defined according to the number of family and friends seen monthly. Smoking was categorized as ever vs never. Hypertension was defined as systolic blood pressure greater than 140 mm Hg, diastolic blood pressure greater than 90 mm Hg, or self-reported use of antihypertensive medications. History of heart disease was defined as any self-reported myocardial infarction (MI) or “heart attack,” coronary artery bypass surgery, coronary angioplasty or stenting, or evidence of MI from ECG. Diabetes was defined as a fasting glucose level greater than 126 mL/dL (or greater than 200 mL/dL if the participant was nonfasting) or self-reported medication use for glucose control. Atrial fibrillation was defined by self-report or the centrally read ECG. Left ventricular hypertrophy (LVH) was defined by ECG.

Statistical Analyses

We hypothesized that depressive symptoms would have a larger impact on stroke risk in Black than in White participants. Cox proportional-hazards modeling was used to examine the association between baseline depressive symptoms and incident stroke. Scores on the CES-D-4 were treated categorically, with participants classified into one of the following categories: (1) a score of 0, indicating that the participant reported no symptoms, (2) a score of 1–3, indicating that the participant reported experiencing some symptoms, but not at a level reflective of depressive symptomatology, and (3) a score of 4 or greater, which was indicative of depressive symptomatology. Although the CES-D-4 has conventionally been treated as a dichotomous variable, the scoring pattern of our sample (with over 60% of participants obtaining a score of 0) suggested that using a cut point of 0—in addition to the traditional cut point of 4—would be an appropriate and potentially more informative means of categorizing the sample.

Analyses were adjusted for demographic factors (age, race, and sex), traditional stroke risk factors (hypertension, diabetes, smoking, atrial fibrillation, heart disease, and LVH), and social factors (education, income, marital status, and social network). Factors were selected because of their known association with stroke risk or depressive symptoms. Adjusted models included an age by race interaction because of the documented higher stroke incidence for Black than White participants at younger ages.²⁷ In addition, because previous studies of stroke and depressive symptoms suggest differential associations by sex,^2,28 we conducted a three-way interaction test between race, sex, and depressive symptoms. Tests for interaction were conducted between (1) race and depressive symptoms, (2) sex and depressive symptoms, (3) race, sex, and depressive symptoms. Evidence of interaction was determined by an a priori α level of 0.10. All statistical analyses were conducted using SAS 9.4 data management software.

Data Availability

The data used in these analyses include potentially identifying participant information and therefore are not publicly available because of legal and ethical restrictions. Qualified investigators may request access from the University of Alabama at Birmingham (regardsadmin@uab.edu) to obtain deidentified data.

Results

Of the sample of 24,045 participants, 40% were Black and 55% were women. The mean age of participants was 64.5 years at enrollment. Over an average follow-up period of 9.21 years, 1,262 incident stroke cases were identified, of which 535 (42%) occurred among Black and 727 (58%) among White participants. The mean CES-D-4 score at baseline was 1.32 among Black and 0.91 among White participants, suggesting low prevalence of depressive symptoms in general. Additional characteristics of the sample are presented in table 1.

Table 1.

Baseline Characteristics Classified by Race and CES-D-4 Score (N = 24,045)

Within the combined sample of Black and White participants (table 2) in the unadjusted model, participants who reported any symptoms on the CES-D-4 were more likely to experience a stroke compared with participants who reported no symptoms. Participants with scores of 1–3 had a 37% increased risk of incident stroke (hazard ratio [HR] = 1.37, 95% confidence interval [CI] = 1.21–1.54). The results were similar after adjustment for demographic factors (HR = 1.39, 95% CI = 1.23–1.57) and only slightly attenuated after adjustment for stroke risk factors (HR = 1.32, 95% CI = 1.17–1.49) and social factors (HR = 1.27, 95% CI = 1.11–1.43). Participants with CES-D-4 scores of 4 or greater experienced a 36% higher risk of stroke (HR = 1.36, 95% CI = 1.14–1.64) in the unadjusted model. Risk of incident stroke remained elevated after adjustment for demographic factors (HR = 1.54, 95% CI = 1.27–1.85), stroke risk factors (HR = 1.36, 95% CI = 1.13–1.64), and social factors (HR = 1.25, 95% CI = 1.03–1.51).

Table 2.

Cox Regression Model Predicting First Stroke—Hazard Ratios (95% Confidence Interval) by Scores on 4-Item Center for Epidemiologic Studies Depression Scale (N = 24,045)

A priori, we hypothesized a differential effect by race, but the interaction was not significant (p = 0.53); however, for comparison purposes, we present both combined and race-stratified results in table 2. Comparing those with no depressive symptoms with those with scores of 1–3 on CES-D-4, the results for Black and White participants were similar to the combined results, with a significant association between scores of 1–3 on CES-D-4 and incident stroke. Both Black and White participants with scores of 1–3 had a 25%–40% increased risk of incident stroke. In comparing participants with CES-D-4 scores of 4 or greater to those with no depressive symptoms, in White participants, the risk of incident stroke was significantly elevated in the unadjusted model and remained so in all subsequent adjusted models, but for Black participants, the association was significant only in the demographic adjusted model (HR = 1.37, 95% CI = 1.05–1.79). The sex-by-depressive symptoms interaction (p = 0.45) and the 3-way interaction of race, sex, and depressive symptoms (p = 0.36) were not significant.

Discussion

In this national sample of Black and White adults, baseline depressive symptoms were associated with a significantly increased risk of incident stroke, even after adjustment for demographic factors, stroke risk factors, and social factors. Our results are consistent with previous research demonstrating a significant association between depression and increased stroke risk^3,5,6,29,30 and do not suggest that this association differs between Black and White participants.

The lack of a significant interaction between race and depressive symptoms suggests that the association between depressive symptoms and stroke risk for Black and White participants is more similar than different, and the race-specific results should be interpreted with caution. However, the results suggest that for White participants, experiencing 4 or more symptoms of depression increased the stroke risk by 73% within a 9-year follow-up period, after adjustment for demographic characteristics, and was lessened to a 41% increased risk after adjustment for traditional stroke risk factors and social factors. However, in Black participants experiencing 4 or more depressive symptoms,the risk of stroke was not increased except after adjustment for demographic factors. We note that the sample size and number of stroke events is smaller for Black participants than that for White participants, and as such, there is less statistical power to detect associations. It is possible that differences in symptom reporting may account for the demonstrated effects because previous research suggests that older Black participants may under-report depressive symptoms on self-rated scales.³¹ However, evidence regarding potential racial differences in symptom reporting is mixed, with some studies finding that Black participants report more depressive symptomatology than White participants, whereas others indicate that Black participants disclose depressive symptoms at comparable or lower levels than White participants, particularly when the studies control for socioeconomic status.³² In a cohort study of community-based Black participants, an association was found between major depressive symptoms and risk of incident stroke, with the risk increased 2-fold.¹¹ Furthermore, recent item-level analyses of the CES-D demonstrated no differences between Black and White participants in responses to the 4 items we used to assess depressive symptoms.^33,34 It is thus unlikely that differential response patterns influenced the results in this study. The possibility that other forms of psychological distress may contribute to stroke risk among Black participants deserves additional study. The Northern Manhattan Study, a prospective cohort study that included mostly Hispanics but also African Americans, found a 12% greater risk of ischemic stroke for every 5-point increase in CES-D score (HR = 1.12, 95% CI = 1.01–1.25).³⁰ Further research is needed to examine potential racial differences in the risk of incident stroke conferred by depressive symptoms.

In the National Health and Nutrition Examination Survey (NHANES) self-reported depressive symptomatology was predictive of stroke in risk-adjusted models for all participants and for White men, whereas significance was borderline in White women.¹² Depressive symptomatology showed a higher risk of stroke in Black men and women.¹² Our results may diverge from the previous findings because of measurement differences. For example, stroke incidence was defined in NHANES based on the review of death certificates and discharge codes after self-reported hospital stays, whereas stroke incidence in our study was rigorously confirmed through physician adjudication using such documents along with other review and classification methods. Moreover, we assessed depressive symptoms with one of the most widely used measures of depressive symptomatology. Depressive symptoms in NHANES were assessed using the depression subscale of the general well-being schedule (GWB-D). Caution is advised when using this scale with ethnic minority populations because research on the psychometric properties of the GWB did not support the existence of a depression subscale factor when tested among a sample of Black women.³⁵ For this reason, the authors recommended that the full GWB be used as a single measure of well-being among such populations. Given this proposed limitation of the GWB, it is possible that these NHANES findings better reflect the effects of general psychological distress or poor well-being, rather than depressive symptomatology.¹²

Additional research findings that Black respondents are more likely than White participants to endorse interpersonal items on the CES-D (e.g., “People were unfriendly” and “I felt people disliked me”)^33,34 suggest that perceived prejudice or discrimination may be more salient contributors to psychological distress among Black participants. Such alternative forms of distress, rather than depression, may be more strongly associated with stroke risk among Black participants. Ongoing depressive symptoms among Black participants may exacerbate other more prevalent risk factors or contribute to physiologic dysfunction, resulting in a greater risk of stroke.^2,10 Findings are mixed regarding some physiologic factors that can contribute to the mechanisms associated with depressive symptoms with stroke risk such as inflammation, sedentary behaviors, improper nutrition, or allostatic load.^36,37 This issue merits further study and underscores the importance of further delineation regarding stroke risk attributable to depressive symptoms.

An important limitation of this study is the use of an abbreviated version of the CES-D to assess depressive symptoms. Because the shortened version of this inventory does not tap the full range of depressive symptoms and does not include items found to be more salient among Black participants, potential effects associated with those items could not be considered. The CES-D-4 is also slightly less reliable and less sensitive than the 20-item measure.²³ In addition, the CES-D-4 only assesses a portion of depressive symptoms—depressed mood, which limits the other dimensions measured through the full scale. Clinically diagnosed depression is often characterized by a group of symptoms; therefore, given the capacity of the CES-D-4, other dimensions of depressive symptoms are not captured. Consequently, our results may underestimate depressive symptoms and the effect of such symptoms on stroke risk. Moreover, because no version of the CES-D is intended as a clinical diagnostic instrument, conclusions regarding the effect of clinical depression on stroke risk cannot be drawn from this study. We also acknowledge that in larger field studies, a balance must be found between finding meaningful and validated measures and addressing concerns related to time and expense. Finally, scores on the CES-D and CES-D-4 being categorized trichotomously is uncommon. Although our treatment of this variable is unconventional, it allowed for greater examination of the effects of subthreshold symptoms among nondepressed individuals and did not seem to bias our results. Another limitation is the measurement of depressive symptoms at only one point in time. There was potential for selection bias because those who participated in REGARDS may not be representative of the general population.

Despite its acknowledged limitations, this study has several strengths. Unlike many previous investigations, this study includes a large cohort of Black and White participants. In addition, we have a large, national sample that includes a substantial proportion of participants from geographic areas heavily impacted by stroke (i.e., the southeastern US “Stroke Belt”). It has been speculated that the lack of a significant race-by-depressive symptoms interaction for stroke in a preivous study may have been because of insufficient statistical power.³⁸ Although our results warrant replication, our large sample size facilitated adequate power to support our confidence in the observed results.

Our study provides further evidence that depressive symptoms are associated with an increased risk of stroke. Our findings indicate the need for studies to examine diverse populations. They also provide a basis to consider issues beyond the traditional risk factors for stroke. We provide support for depressive symptoms to be given the same consideration as other traditional risk factors such as hypertension and diabetes. It also suggests the importance of including screenings for depressive symptoms among individuals at risk for stroke.

Appendix. Authors

Study Funding

This research project is supported by cooperative agreement U01 NS041588 cofunded by the National Institute of Neurologic Disorders and Stroke (NINDS) and the National Institute on Aging (NIA), NIH, Department of Health and Human Service. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NINDS or the NIA. Representatives of the NINDS were involved in the review of the study but were not directly involved in the collection, management, analysis, or interpretation of the data. The authors thank the other investigators, the staff, and the participants of the REGARDS study for their valuable contributions. A full list of participating REGARDS investigators and institutions can be found at: https://www.uab.edu/soph/regardsstudy/. Additional support was provided by the Deep South Resource Center for Minority Aging Research (RCMAR) P30 AG031054.

Disclosure

The authors report no disclosures relevant to the manuscript. Full disclosure form information provided by the authors is available with the full text of this article at Neurology.org/cp.

TAKE-HOME POINTS

• → Stroke is a leading cause of death and disability in the United States.

• → Depressive symptoms should be considered in primary stroke prevention for both Black and White participants.

• → Further study is needed with diverse populations to explore nontraditional risk factors for stroke.