Depression and Risk of Stroke Morbidity and Mortality:  A Meta-analysis and Systematic Review FREE

Stroke is a leading cause of death and permanent disability, with significant economic losses due to functional impairments.¹ Depression is highly prevalent in the general population, and it is estimated that 5.8% of men and 9.5% of women will experience a depressive episode in a 12-month period.² The lifetime incidence of depression has been estimated at more than 16% in the general population.³ Depression has been associated with increased risks of diabetes,⁴ hypertension,⁵ and cardiovascular disease.⁶ However, whether depression increases the future risk of stroke remains unclear.

A number of studies have assessed the association between depression and subsequent risks of stroke morbidity and mortality, suggesting that depression could be a modifiable risk factor for stroke.^7- 8 A previous meta-analysis that focused on cardiovascular outcomes pooled results from 10 studies published before 2005 as a secondary analysis and reported a positive association between depression and risk of stroke.⁹ Since then many more studies have been published, which allow more detailed analysis of the association between depression and stroke morbidity and mortality. Therefore, we conducted a systematic review and a meta-analysis of prospective cohort studies to describe the association between depression and future risk of total and subtypes of stroke.

We conducted a systematic literature search (up to May 2011) of MEDLINE, EMBASE, and PsycINFO for studies describing the association between depression (defined by self-reported scales or clinician diagnosis) and stroke morbidity and mortality. In addition, we searched the reference lists of all identified relevant publications and relevant reviews.^7- 9 Only articles published in the English language were considered. Two search themes were combined using the Boolean operator and. The first theme, depression, combined exploded versions of the Medical Subject Headings (MeSH) depression, depressive disorder, or depressive disorder, major. The second theme, stroke, combined exploded versions of MeSH terms stroke, cerebrovascular disorders, or intracranial embolism, and thrombosis.

Two investigators (A.P. and Q.S.) independently assessed literature eligibility; discrepancies were resolved by consensus. Articles were considered for inclusion in the systematic review if the authors reported data from an original, peer-reviewed study (ie, not review articles or meeting abstracts); the study was a cohort study (prospective cohort or historical cohort) consisting of noninstitutionalized adults (>18 years old); and the authors reported the risk estimates of stroke morbidity or mortality in depressed participants compared with nondepressed individuals. We used broad inclusion criteria for studies, including all types of stroke (total, fatal, nonfatal, ischemic, and hemorrhagic) and depression status (assessed by different scales or clinical diagnosis). We identified articles eligible for further review by performing an initial screen of identified titles or abstracts, followed by a full-text review.

We extracted the following information about the studies: study characteristics (study name, authors, publication year, journal, study site, follow-up years, and number of participants), participants' characteristics (mean age or age range, sex), main exposure depression (self-reported scales or clinician diagnosis, assessed at baseline or updated), main outcome stroke (morbidity or mortality, types, assessed by self-report, death certificates, or medical records), and analysis strategy (statistical models, covariates included in the models). Quality assessment was performed with consideration of the following aspects: study design, response rate, follow-up rate, follow-up years, exposure and outcome measurements, statistical analysis, and generalizability to other populations (eTable 1).

The hazard ratios (HRs) were used as the common measure of association across studies, and the relative risks (RRs) were considered equivalent to HRs. If the result on total stroke were not available, we used data from ischemic stroke, nonfatal stroke, or fatal stroke (in the sequential order) as a surrogate for total stroke. Forest plots were produced to visually assess the HRs and corresponding 95% confidence intervals across studies. Heterogeneity of HRs across studies was evaluated by the Cochrane Q statistic (P <.10 was considered indicative of statistically significant heterogeneity) and the I² statistic (values of 25%, 50%, and 75% were considered to represent low, medium, and high heterogeneity, respectively).^10- 11 The HRs were pooled using the fixed-effect model if no or low heterogeneity was detected, or the DerSimonian and Laird¹² random-effects model otherwise, and the weights were equal to the inverse variance of each study's effect estimation. The possibility of publication bias was evaluated using the Begg test¹³ and visual inspection of a funnel plot.¹⁴ The Duval and Tweedie¹⁵ nonparametric trim-and-fill procedure was used to further assess the possible effect of publication bias in our meta-analysis. Moreover, stratified analyses and sensitivity analyses were performed to evaluate the influences of selected study and participant characteristics on study results. The analyses were performed with Stata statistical software version 11.0 (StataCorp, College Station, Texas). P values were 2 sided with a significance level of .05.

We calculated absolute risk differences associated with depression by multiplying the background incidence rate of stroke in the general US population with (estimated HR−1). Population attributable risk (PAR) was calculated based on the following equation: PAR%=100 × P_e(HR−1)/(P_e[HR–1]+1), for which P_e is the prevalence of the exposure (depression) in the population and the HR was derived from this meta-analysis.

The search strategy identified 10 075 unique citations. After the first round of screening based on titles and abstracts with the aforementioned criteria, 302 articles remained for further evaluation. After examining those articles in more detail, 274 articles were excluded for reasons shown in Figure 1. Finally, 28 articles met the inclusion criteria and were included in the meta-analysis.^16- 43

Among the included articles, 2 studies were retrieved from the reference lists,^16- 17 and one was from our recent publication.¹⁸

Among these 28 articles, 8 studies specifically reported results on fatal stroke,^{16- 17,21,26- 27,29,33,38} 3 studies on nonfatal stroke,^26,31,38 6 studies on ischemic stroke,^{16,18,24,26,32,35} and 2 studies on hemorrhagic stroke.^18,24 Six studies^{18,28,32,37,40,42} reported the crude association between antidepressant medication use and total stroke risk (Wassertheil-Smoller et al²⁸ reported the results in a separate article⁴⁴).

Characteristics of the 28 selected studies are shown in Table 1. The total number of participants included in this meta-analysis was 317 540, with 8478 reported stroke outcomes (1 study did not report the number of stroke cases¹⁹). The studies varied with regard to how results were presented. Two studies reported results separately by age group: younger than 65 and 65 years or older (Salaycik et al³⁴), 65 to74, and 75 years or older (Avendano et al³⁰); 2 studies reported their results separately by baseline history of cardiovascular disease,^28,40 with 1 study providing unpublished data for the total sample⁴⁰ and 3 studies providing results stratified by sex along with the results from total samples.^35,37,41 With regard to study location, most of the studies were from US or European countries. Three studies were conducted in Japan,^24,27,33 1 in Australia,¹⁶ and 1 in Taiwan,³⁶ and 1 was an international collaboration.⁴³ The study samples ranged from 401 to 93 676, and the follow-up durations ranged from 2 to 29 years. Most of the studies comprised both men and women, while 2 studies included only men,^26,29 and 3 studies only women.^17- 18,28

In most of the studies, depression was measured by self-reported scales, such as the Center for Epidemiologic Studies Depression Scale (CES-D),^{16,20,25,28- 30,32,34- 35,40- 41} Zung Self-Rating Depression Scale,^24,33 30-item General Health Questionnaire,^26- 27 Geriatric Depression Scale,^17,43 Beck Depression Inventory,⁴² Human Population Laboratory Depression Scale,²¹ 9-item Patient Health Questionnaire,³⁹ and 5-item Mental Health Index.¹⁸ Four studies used the Diagnostic Interview Schedule to define depression as the exposure,^23,36- 38 2 studies included antidepressant medication use as a component of depression definition,^18,34 and 4 studies used combined methods.^{18,33- 34,40} The depression status was only measured at baseline in the majority of studies, whereas 3 studies used updated depression assessments.^18,21,33 In most of the studies, stroke was assessed by death certificates or medical records, and some studies combined self-reported measures with medical records; only 1 study relied solely on self-reported outcomes.⁴¹ Three studies included outcomes comprising stroke and transient ischemic attack.^16,34,39 Baseline stroke cases were not excluded in 7 studies^{16- 17,20,27- 28,39,43}; we included those studies in the main analysis, but conducted a stratified analysis by presence or absence of baseline stroke cases.

Adjusted HRs could be determined for most studies, except that 2 studies reported the crude results without adjustment (eTable 2).^20,33 Most of the results were adjusted for age (25 studies), smoking status (20 studies), body mass index (BMI) (14 studies), alcohol intake (9 studies), physical activity (7 studies), and comorbidities (23 studies; such as diabetes, hypertension, and coronary heart disease).

Among the 31 reports from the 28 studies of results on total stroke, the majority of studies reported a positive association (ie, HR>1.00), with 14 of them being statistically significant. Only 4 studies reported an HR of less than 1.00 but these values were not statistically significant. A moderate to high heterogeneity was detected with an I²=66.0% (Cochrane Q statistic=88.1, P <.001), the HR from random-effects model was 1.45 (95% CI, 1.29-1.63; Figure 2). A sensitivity analysis of omitting 1 study in each turn showed that the study by Lee et al³⁶ had the largest influence on the results: the pooled HR without this study was 1.36 (95% CI, 1.24-1.49). Another sensitivity analysis, in which we excluded studies that imputed the risk estimates from other stroke outcomes (ischemic, nonfatal, fatal stroke) if data on total stroke were not available, revealed similar results (HR, 1.46; 95% CI, 1.26-1.69; 23 reports from 20 studies; I²=73.7%; random-effects model). We conducted a secondary analysis to combine the unadjusted results on the association between antidepressant medication use and stroke risk, and the HR was 1.41 (95% CI, 1.25-1.59; I²=0%; eFigure 1).

Fatal stroke results were available from 8 studies with a pooled HR of 1.55 (95% CI, 1.25-1.93) from fixed-effect model (Figure 3). A modest heterogeneity was found with an I²=15.8% (Cochrane Q statistic=8.32, P =.31). Most of the studies found an HR above 1.00 except 1 study with an observed HR of 0.45.³⁸ Ischemic stroke results were available from 6 studies with a pooled HR of 1.25 (95% CI, 1.11-1.40; fixed-effect model). A low heterogeneity was found with an I²=12.3% (Cochrane Q statistic=5.70, P =.34). Similar sensitivity analyses for fatal stroke and ischemic stroke did not appreciably change the results (data not shown). Results for nonfatal stroke and hemorrhagic stroke were not significant (1.21; 95% CI, 0.91-1.62 and 1.16; 95% CI, 0.80-1.70, respectively; both from fixed-effect model with I²=0%; Table 2), however, the number of studies (n=3 and n=2, respectively) that separately addressed these stroke types was small.

The corresponding absolute risk difference associated with depression based on the most recent stroke statistics for the United States⁴⁵ was estimated to be 106 cases for total stroke, 53 cases for ischemic stroke, and 22 cases for fatal stroke per 100 000 individuals per year. According to the most recent statistics, 9.0% (21 million) of US adults meet the criteria for current depression.⁴⁶ Using the risk estimates from our meta-analysis, we estimated that 3.9% (n=273 000) of stroke cases in the United States could be attributable to depression.

Depression was associated with an increased risk of stroke in most subgroups (eFigure 2). The increased risk was more evident in several strata of study characteristics (Table 2): using clinical diagnosis to define depression, with a low study quality (< than the median score), featuring younger participants (mean age <65 years), having a relatively small study sample (n<5000), with Asian participants, and lack of statistical control for smoking status or BMI. Twenty-two reports (7334 cases) adjusted for smoking in the multivariate models, whereas 9 reports (1144 cases) did not. The pooled HR controlling for smoking (1.28; 95% CI, 1.21-1.36) was lower than the pooled HR without smoking in the models (1.92; 95% CI, 1.28-2.86). Likewise, the pooled HR controlling for BMI (1.28; 95% CI, 1.20-1.36; 15 reports, 6718 cases) was lower than pooled HR without BMI in the models (1.76; 95% CI, 1.33-2.32; 16 reports, 1760 cases). No between-group differences were observed for other variables (eTable 3). Nevertheless, moderate to high heterogeneities were observed in most of the subgroups.

Visual inspection of the funnel plot revealed asymmetry (eFigure 1A), and the Begg test was significant (z =2.33; P =.02). A sensitivity analysis using the trim-and-fill method was performed with 6 imputed studies, which produced a symmetrical funnel plot (eFigure 1B). The pooled HR incorporating the 6 hypothetical studies was smaller than the original results, but it remained statistically significant (HR, 1.28; 95% CI, 1.12-1.47; P <.001). No significant publication bias was observed for fatal stroke (P =.22), and a moderate bias for ischemic stroke (P =.04; HR, 1.19; 95% CI, 1.03-1.38 after trim-and-fill method).

Quiz Ref IDThe results of this meta-analysis demonstrate that depression is prospectively associated with a significantly increased risk of developing stroke. Furthermore, the association persisted and remained statistically significant across several subgroups stratified by various study and participant characteristics. We also found a positive association of depression with fatal stroke and ischemic stroke.

Our results are consistent with a previous meta-analysis of 10 studies published before 2005 (HR, 1.43; 95% CI, 1.17-1.75).⁹ Our current meta-analysis, with 5 times more cases, provides strong evidence that depression is associated with increased risks of total stoke, fatal stroke, and ischemic stroke. The result is also consistent with a large case-control study, the INTERSTROKE study,⁴⁷ for which the investigators found that self-reported depression (for ≥2 or more weeks in the last year) was associated with a significantly increased risk of stroke (OR, 1.35; 99% CI, 1.10-1.66) in 3000 cases and 3000 matched controls from 22 countries. Several studies that did not meet the inclusion criteria for the meta-analysis also found a positive association between depression and stroke. For example, Simonsick et al⁴⁸ found that the stroke incidence rates were 2.3 to 2.7 times higher in most subgroups with high depressive symptoms compared with their nondepressed counterparts in a population of older adults with hypertension (n=3461); Nilsson and Kessing⁴⁹ found that patients with depression severe enough (n=11 741) to be hospitalized had an increased future risk of stroke (HR, 1.22; 95% CI, 1.06-1.41) compared with patients with osteoarthritis (n = 81 380) in Denmark. Using a continuous variable of 20-item CES-D score, Ostir et al⁵⁰ found that depressive symptoms were associated with an increased stroke risk (HR, 1.01 per score increase; 95% CI, 1.00-1.02) in 2682 Mexican Americans aged 65 years and older.

Quiz Ref IDDepression may contribute to stroke through a variety of mechanisms. First, depression has known neuroendocrine (eg, sympathetic nervous system activation, dysregulation of the hypothalamic-pituitary-adrenocortical axis, platelet aggregation dysfunction)⁶ and immunological/inflammation effects,⁵¹ which could influence stroke risk. A recent meta-analysis suggests that depression is positively associated with C-reactive protein (CRP), IL-1, and IL-6 in clinical and community samples,⁵² and these inflammatory factors have been associated with an increased risk of stroke.⁵³Quiz Ref IDSecond, depression is associated with poor health behaviors (ie, smoking, physical inactivity, poor diet, lack of medication compliance)⁵⁴ and obesity,⁵⁵ which might increase the risk of stroke. Adjusting for smoking or BMI somewhat attenuated the association between depression and stroke, suggesting that smoking and obesity may confound or mediate the association between depression and stroke. The magnitude of the depression-stroke association observed in this study is similar to the associations between smoking and stroke (HR, 1.51; 95% CI, 1.45-1.58; from a meta-analysis),⁵⁶ and between obesity and stroke (HR, 1.26; 95% CI, 1.07-1.48; from a meta-analysis).⁵⁷ Third, depression is correlated with other major comorbidities, such as diabetes⁴ and hypertension,⁵ both of which are major risk factors for stroke. Finally, antidepressant medication use may contribute to the observed association. We found a positive association between the medication use and stroke risk; however, the results should be interpreted cautiously because medication use can be a marker of depression severity, and many studies lacked information on dose and duration of medication use.

Several limitations of this meta-analysis should be considered. First, we found significant heterogeneity across studies, which may result from differences in study designs, sample sizes, depression and stroke measures, analysis strategies, and participants' characteristics. Although moderate to high heterogeneities still remained in many subgroups, the pooled HRs showed consistent positive associations in most subgroups. Quiz Ref IDSecond, the funnel plot indicated a possible publication bias; however, the trim-and-fill sensitivity analysis did not materially change the results (although the pooled HR was modestly attenuated). Nevertheless, the possibility of publication bias could not be fully excluded by this method. Moreover, the meta-analysis was limited to English-language publications, and there is the possibility of unidentified unpublished reports. Data extraction and analyses were not blinded to the authors, journals, or institutions of the publications; however, the literature screening and data extraction were conducted independently by 2 investigators, and thus, selection bias was unlikely. Quiz Ref IDFurthermore, most studies did not have information on depression treatment and antidepressant medication use. The role of depression treatment in modulating subsequent risk of stroke needs to be studied further. Finally, further studies are needed to determine whether depression is associated with hemorrhagic stroke.

In conclusion, this meta-analysis provides strong evidence that depression is a significant risk factor for stroke. Given the high prevalence and incidence of depression and stroke in the general population, the observed association between depression and stroke has clinical and public health importance. More studies are needed to explore the underlying mechanisms and elucidate the causal pathways that link depression and stroke.