Depression and Risk of Stroke in Midaged Women
A Prospective Longitudinal Study
Background and Purpose—Depression is known to increase stroke risk. Although limited, there is some evidence for age differences, with a suggestion for a stronger association in younger groups. We investigated the effect of depression on stroke incidence in a large cohort of midaged women.
Methods—We included 10 547 women without a history of stroke aged 47 to 52 years from the Australian Longitudinal Study on Women’s Health, surveyed every 3 years from 1998 to 2010. Depression was defined at each survey using the Center for Epidemiological Studies Depression Scale (shortened version) and antidepressant use in the past month. Stroke was ascertained through self-report and mortality data. We determined the association between depression and stroke at the subsequent survey, using generalized estimating equation analysis, adjusting for time-varying covariates.
Results—At each survey, ≈24% were defined as having depression. During follow-up, 177 strokes occurred. Depression was associated with a >2-fold increased odds of stroke (odds ratio, 2.41; 95% confidence interval, 1.78–3.27), which attenuated after adjusting for age, socioeconomic status, lifestyle, and physiological factors (odds ratio, 1.94; 95% confidence interval, 1.37–2.74). Findings were robust to sensitivity analyses addressing methodological issues, including definition of depression, antidepressant use, and missing covariate data.
Conclusions—Depression is a strong risk factor for stroke in midaged women, with the association partially explained by lifestyle and physiological factors. Further studies of midaged and older women from the same population are needed to confirm whether depression is particularly important in younger women and to inform targeted intervention approaches.
Stroke is a major cause of death and disability worldwide.1 The burden is particularly large among women, given their longer life expectancy, higher strokes rates in older age compared with men, and lower physical functioning poststroke.2 Similarly, depression is a global health problem, with a high prevalence3 and ranked as the fourth leading cause of morbidity globally.4
Although the development of depression after stroke is well-recognized,5 the role of depression as a risk factor for stroke is less well understood. Conflicting findings from existing studies led to recent systematic review and meta-analyses of studies that prospectively examined depression and stroke risk.6,7 Although the authors concluded that depression increases stroke risk, there was substantial heterogeneity between studies, which was not fully explained in subgroup analyses.7 Specifically, subgroup analyses revealed substantial heterogeneity between studies of women.7 This could be a result of differences in the age of study populations because meta-analysis revealed a stronger association in studies where the mean age was <65 years. Few studies have reported age-stratified results. However, the Framingham and Established Populations for Epidemiologic Studies of the Elderly studies both found evidence of age interactions, with depression associated with increased stroke risk in younger but not older participants.8,9 An association between depression and stroke was observed in women included in the Nurses’ Health Study, but age-specific effects were not reported.10 Important age group differences may therefore have been masked in previous studies where associations were not stratified by age.
In addition, the extent to which depression operates through conventional stroke risk factors is unclear. Socioeconomic status has rarely been accounted for,6 and few studies have been able to adjust for time-varying covariates. The latter is particularly important in long-term studies, which should ideally control for the development of conditions, such as hypertension. Also, by measuring depression at only baseline, many studies may have underestimated the association with stroke.
We aimed to determine the association between depression and incident stroke in a population-based cohort of midaged women with repeated exposure measures during 12-year follow-up.
We included participants from the Australian Longitudinal Study on Women’s Health, a population-based study of women born in 1921–1926, 1946–1951, and 1973–1978. Women were randomly selected from the Medicare database, which covers all citizens and permanent residents of Australia, including refugees and immigrants. Women born in 1946–1951 were surveyed, using self-completed questionnaires, in 1996 (survey 1, S1), 1998 (survey 2, S2), 2001 (survey 3, S3), 2004 (survey 4, S4), 2007 (survey 5, S5), and 2010 (survey 6, S6). Full details of recruitment and response rates are reported elsewhere.11
We included women from the 1946–1951 cohort, which recruited 13 715 women at S1, 12 338 (90%) of whom returned S2. For these analyses, we followed women from S2 onward because information on depression was not collected at S1. We excluded women as indicated in Figure 1.
Incident stroke was determined at S3 to S6, when women were asked, “In the past 3 years have you been diagnosed or treated for stroke?” We identified deaths and cause of death through linkage of Australian Longitudinal Study on Women’s Health to the National Death Index. Stroke deaths were determined using the following International Statistical Classification of Diseases and Related Health Problems, 10th revision (ICD-10) codes from the principal or secondary diagnosis fields of death certificate data: I60–I60.9, I61.0–I61.9, I63.0–I63.9, and I64.
Measures of Depression
Depressive symptoms experienced in the past week were identified using the Center for Epidemiological Studies Depression Scale shortened version (CESD-10).12 It is well validated and has good test–retest reliability and predictive validity compared with the original 20-item format.12–14 In S2 to S4, women were also asked “During the past 4 weeks have you taken any medications for depression,” and in S5, women were asked to list the medications they had taken in the past month. For the latter survey, we ascertained antidepressant use from the anatomic therapeutical chemical system, by identifying medications coded as N06A or N06B.
In primary analyses, we defined depression at each time point as being present if women reported taking antidepressant medication or if they scored ≥10 on the CESD-10 scale.
Women were also asked at each survey “In the past 3 years have you been diagnosed or treated for depression?” which was used in sensitivity analyses to investigate whether the definition of depression impacted the results.
Other Exposure Variables
Lifestyle and physiological stroke risk factors were determined at S2 to S5. Smoking was dichotomized into current smoker or not. Body mass index was (kg/m2) calculated from self-reported weight and height. Physical activity was assessed using questions from the Australian physical activity survey15 and defined according to minutes of moderate activity per week: nil/sedentary (<0–10 min/wk); low (11–150); moderate (151–300); and high (>300). Alcohol intake was defined in light of the Australian National Health and Medical Research Council guidelines,16 with risky drinkers (15–28 drinks per week) and high-risk drinkers (>28 drinks per week) categorized accordingly. For women identified as low risk by the National Health and Medical Research Council guidelines, we separately categorized those classified as low-risk drinkers from those reporting that they drink only rarely. Non-drinkers were classified separately.
At S1 and S2, women were asked whether they had ever been diagnosed with or treated for hypertension, diabetes mellitus, or heart disease. Subsequently, they were asked whether they had been diagnosed with these conditions in the period since the previous survey. Women were also asked whether they had undergone a hysterectomy or oophorectomy.
Relationship status at S2 was categorized into married/defactor partner, divorced/separated/widowed, and single. Socioeconomic status (SES) was measured by education level at S1 (no formal qualifications; school leaving certificate; high school leaving certificate; trade or apprenticeship; certificate or diploma; and university degree) and home ownership at S2. These SES indicators were included because they were independently associated with stroke in this cohort, whereas occupation and managing on income were not. A small number of women were inconsistent in returning surveys. Where women returned S4, S5, or S6, but did not return the prior survey, we carried forward the values for exposure variables from the immediately preceding survey, if returned. Therefore exposure variables for 5% of women at S3, 3.5% at S4, and 2.7% at S5 were derived from the previous survey.
All analyses were performed using Stata version 12.0 (StataCorp; College Station, TX.)
We compared characteristics of women according to depression status at S2 using the Pearson χ2 test and Student t test. To account for multiple observations for each participant, we used generalized estimating equation regression models for binary outcome data (using an unstructured correlation structure and a logit link function). We calculated odds ratios (ORs) with 95% confidence intervals (CIs) for the relationship between depression and first-ever stroke, including depression as a time-varying covariate. Women who reported stroke did not contribute to the analyses of time periods thereafter. We calculated crude ORs before adjusting for present age and SES and then additional confounding factors. All physiological and lifestyle variables were included as time-varying covariates. Once women reported a chronic condition, they were considered to have it at each subsequent survey. Time lags were used so that exposures, including depression, were associated with stroke occurring at the subsequent survey. We determined model fit at each time point using the Hosmer–Lemeshow goodness-of-fit test, which indicated that the model was well calibrated.
We performed sensitivity analyses, where we repeated analyses using alternate definitions of depression: CESD-10 ≥10; doctor diagnosed depression or CESD-10 ≥10; doctor diagnosed depression, CESD-10 ≥10 or taking anti-depressants. We also repeated the primary analysis but excluded women taking antidepressants. Finally, a proportion of returned questionnaires (5%–17% across surveys) were incomplete with respect to some covariates. We therefore performed a sensitivity analysis, where we repeated the primary analyses, but performed multiple imputation for all covariates with missing values.
Among 12 338 women who returned S2, 12 191 did not have a history of previous stroke. Follow-up rates at S3, S4, S5, and S6 were 88%, 85%, 82%, and 78%. We excluded 1644 women (Figure 1). Of 12 191 women without stroke at S2, we included 10 547 (87%) in the analyses, 79% of whom returned all subsequent surveys. Almost all women (99.4%) were nonindigenous. There were some baseline differences between included and excluded women, with the latter being less well educated, generally less healthy, and more likely to have depression (Table I in the online-only Data Supplement).
At S2, the mean age was 52.5 (±1.5 SD) and prevalence of depression was 25.1%, with a similar prevalence at subsequent surveys (Table II in the online-only Data Supplement). Depression at S2 was associated with SES, marital status, lifestyle behaviors, and physiological stroke risk factors (Table 1).
During follow-up, 177 first-ever strokes occurred, 5 of which were fatal, giving a stroke prevalence of 1.5%. Of these, 143 were included in the primary analyses, and 170 were included in the sensitivity analysis after multiple imputation of missing covariate data. In the primary analyses, depression was associated with >2-fold greater odds of stroke (OR, 2.41; 95% CI, 1.78–3.27). This association attenuated but remained statistically significant on controlling for age, SES, lifestyle, and physiological stroke risk factors (OR, 1.94; 95% CI, 1.37–2.74; Table 2).
We found similar results in sensitivity analyses where we used alternate definitions of depression (Table 3). The association between depression and stroke also was almost identical when we repeated the analyses after multiple imputation of missing covariate data (OR, 1.98; 95% CI, 1.44–2.72).
In our study, depression was associated with a marked increased risk of incident stroke among midaged women, even after adjusting for age, SES, lifestyle, and physiological risk factors. Our findings suggest that the association between depression and stroke risk among midaged women, in particular, may be stronger than previously thought. Interestingly, the role of depression as a preventable risk factor for stroke is often overlooked, with depression notably absent from reviews and guidelines for the primary prevention of stroke.20,21 The association seems to be only partially accounted for by traditional stroke risk factors, highlighting the possible role of other novel risk factors, or biological mechanisms, such as the proposed neuroendocrine17 and immunologic/inflammatory18,19 pathways.
Comparisons With Other Studies
Our finding that depression increases stroke risk is in keeping with results of most other prospective studies,6,7 although the magnitude of the effect in our study population is larger than in many previous studies. However, existing studies have generally included only a broad age range or elderly people, without stratification by age. No previous study has included a female-only population as young as the Australian Longitudinal Study on Women’s Health 1946–1951 cohort. The closest comparison is with the Nurses’ Health Study, which found that depression was associated with a 30% increased stroke risk. However, the mean age of this cohort was 14 years higher than in our study, and results were not reported by different age groups.10 In subgroup analyses of a recent review, the authors compared studies with a population mean age of <65 versus ≥65 years and obtained higher pooled hazard ratios for the former than the latter (1.77; 95% CI, 1.30–2.41 and 1.30; 95% CI, 1.18–1.44, respectively).7 The magnitude of effect observed in our study is consistent with these findings. Furthermore, despite including a much smaller number of stroke outcomes in the younger age group than in our cohort, the Framingham Study found that depression was associated with a 3-fold increased stroke risk in those aged <65 years but found no association in those aged ≥65 years.9
It has been suggested that antidepressants themselves may increase stroke risk, perhaps via the inhibition of platelet aggregation increasing risk of bleeding. A recent review concluded that use of selective serotonin reuptake inhibitors increased risk of brain hemorrhage.22 In our study, the association between depressive symptoms and stroke remained, when we excluded women taking antidepressant medication from our analyses.
Our study benefits from a number of strengths. This was a community-based study, allowing us to extrapolate our findings to the general female population of this age group. Women were followed for a long period and repeated exposure measures were obtained. This enabled us to model depression more accurately over time, rather than relying on 1 measure of depression at baseline and to control for the development of physiological factors, such as hypertension, heart disease, and diabetes mellitus, during follow-up. This is especially important for this particular age group because women were in their late 40s or early 50s when follow-up started, and so may not yet have been diagnosed with these conditions. In general, previous studies of the relationship between depression and stroke have controlled only for baseline measurements of confounders, which may have led to residual confounding.
Our study has some limitations. Incident stroke is based largely on self-report, which may have introduced some errors. However, in a validation study of self-reported stroke in a subset of this cohort for whom hospital discharge data were available, we found moderate agreement between self-reported and hospital-recorded stroke. In general, studies that compared self-reported stroke against a comprehensive comparison group identifying strokes from multiple sources of information, and not only hospital records, report very good validity of self-reported stroke.23–26 A recent study found that 89% of self-reported strokes were verified by hospital or primary care records.23 The prevalence of self-reported stroke in our study is also in keeping with the stroke prevalence in this age group from population-based studies in similar high-income countries.27 Some women may have misreported transient ischemic attacks as strokes. However, given the etiological similarities of these clinical conditions, which are distinguished only by duration of symptoms, we might expect associations with depression to be similar. It is difficult to predict how including nonstroke events will have affected our results. If misclassification of stroke is independent of depression status, then the association will be underestimated. However, if misclassification of stroke is related to depression status at the previous survey, then the effect is more difficult to predict.
The relatively small proportion of women not included in the analyses, largely due to noncompletion of surveys, were generally less healthy than included women, which means we probably underestimated both depression and stroke occurrence. However, although this is a limitation, we cannot be sure that this will have actually affected the strength of the observed association between depression and stroke.
In predominantly white populations, ≈10% of strokes are hemorrhagic,21 and there are known differences in the risk factor profiles of pathological stroke subtypes. Unfortunately, we did not collect information on stroke type. Few studies have investigated the association between depression and pathological stroke type, but limited evidence suggests the association may be weaker with hemorrhagic stroke.7
There may be some residual confounding because we did not collect information on hypercholesterolemia or atrial fibrillation for instance, and exposures were self-reported, which may have led to some misclassification.
Finally, depressive symptoms were identified using the CESD-10 tool in a self-administered questionnaire, which may have introduced errors in depression ascertainment. Our results were, however, robust to sensitivity analyses, where we used alternate definitions of depression. In addition, we are likely to have underestimated rather than overestimated the association between depression and stroke because studies where depression was measured via physician diagnosis show an even stronger relationship.7 Also, the prevalence of depression in our study population at baseline is similar to that expected for women of this age.28
Our findings contribute to the currently limited evidence on potential age differences in the association between depression and stroke, and suggest that the effect of depression may be even stronger in younger women. Further research investigating age differences within the same cohort is needed because the identification of such differences will have important implications for policy and practice. In particular, this will inform the development of effective targeted prevention and intervention approaches. Depression seems to operate only partially through known conventional stroke risk factors, suggesting that other mechanisms may be important.
Drs Jackson and Mishra designed the study. Dr Jackson performed the analyses and drafted the article. Drs Jackson and Mishra interpreted the results, and Dr Mishra commented on the draft article.
Sources of Funding
Australian Longitudinal Study on Women’s Health is funded by the Australian Commonwealth Department of Health and Ageing. The funding organizations had no role in the design and conduct of the study or in data collection, analysis, interpretation of results, and preparation of the article.
Drs Jackson and Mishra were supported by the Australian National Health and Medical Research Council (grant number: APP1000986).
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.113.001147/-/DC1.
- Received February 11, 2013.
- Revision received March 17, 2013.
- Accepted March 26, 2013.
- © 2013 American Heart Association, Inc.
- 4.↵World Health Organisation. The global burden of disease: 2004 update. 2008
- Hackett ML,
- Yapa C,
- Parag V,
- Anderson CS
- Dong JY,
- Zhang YH,
- Tong J,
- Qin LQ
- Avendano M,
- Kawachi I,
- Van Lenthe F,
- Boshuizen HC,
- Mackenbach JP,
- Van den Bos GA,
- et al
- Salaycik KJ,
- Kelly-Hayes M,
- Beiser A,
- Nguyen AH,
- Brady SM,
- Kase CS,
- et al
- Pan A,
- Okereke OI,
- Sun Q,
- Logroscino G,
- Manson JE,
- Willett WC,
- et al
- Lee C,
- Dobson AJ,
- Brown WJ,
- Bryson L,
- Byles J,
- Warner-Smith P,
- et al
- Furukawa T,
- Anraku K,
- Hiroe T,
- Takahashi K,
- Kitamura T,
- Hirai T,
- et al
- Armstrong T,
- Bauman A,
- Davies J
- 16.↵National Health and Medical Research Council. Australian alcohol guidelines: health risks and benefits. Canberra 2001.
- Goldstein LB,
- Bushnell CD,
- Adams RJ,
- Appel LJ,
- Braun LT,
- Chaturvedi S,
- et al
- Britton A,
- Milne B,
- Butler T,
- Sanchez-Galvez A,
- Shipley M,
- Rudd A,
- et al
- Engstad T,
- Bonaa M,
- Viitanen M
- O’Mahony PG,
- Dobson R,
- Rodgers H,
- James OF,
- Thomson RG