Psychological distress, major depressive disorder, and risk of stroke

During the past decade, research has focused on increasing understanding of the bidirectional association between depression and stroke. Results from stroke survivor,^1,2 register,³ and population-based⁴ studies have suggested that depression commonly occurs following stroke, necessitates increased need for poststroke healthcare, is associated with an increased risk of suicide, particularly in women and in younger age groups, and following control for vascular and other risk factors, is associated with increased stroke incidence.

However, the evidence that depression may confer increased risk of incident stroke is controversial.^5,6 This follows from study limitations imposed by cohort size and endpoint rarity,^7,8 by restriction to special study groups,⁹ by limited capacity to adjust for stroke risk factors,^7,10 and in particular, by a near-universal reliance on general measures of mental health status defined by symptom count alone rather than by depressive disorder diagnostic criteria that require recognition (for example) of evidence of a sustained period of clinically important distress and impairment.^4,9,11,12

To advance understanding of the origin of the suggested association between depression and stroke, and to consider the clinical implications of such an association, evidence is needed, based upon large population-based prospective healthy (stroke-free) cohort studies, that include assessment of depressive disorder defined by diagnostic criteria. Based on data collected within the United Kingdom European Prospective Investigation into Cancer (EPIC)–Norfolk population-based prospective cohort study,¹³ we evaluated the association between measures of general mental health well-being and of depressive disorder and incident fatal and nonfatal stroke, independent of cerebrovascular risk factors, for both men and women.

METHODS

RESULTS

Of 20,627 study participants aged between 41 and 80 years, 8,939 were men and 11,688 were women. Of these 1,083 (5.3%) reported episodes of MDD within 12 months of HLEQ assessment (336 [3.8%] for men and 747 [6.4%] for women, p < 0.001 for test of difference by sex), and 3,184 (15.4%) reported episodes of MDD at any time in their lives (1,012 [11.3%] for men and 2,172 [18.6%] for women, p < 0.001). Mean MHI-5 score was 76.9 (SD = 16.4), 78.9 (15.8) for men and 75.4 (16.6) for women (p < 0.001 for test of difference by sex), and was 55.2 (20.3) for participants who reported an episode of MDD in the past 12 months, 76.5 (14.4) for those who reported episodes of MDD in their lifetime (that was not in the past 12 months), and 78.5 (15.2) for participants who reported no episodes of MDD at any time in their lives. During a total of 170,593 (median 8.5) person-years of follow-up, 595 incident stroke endpoints were recorded (277 among men and 318 among women). Of these, 167 were fatal and 491 were hospital admissions (with 63 participants admitted to hospital and subsequently dying from stroke). Figure 1 shows the distribution of MHI-5 scores, and table 1 shows the number and percentage of participants reporting 12-month MDD according to stroke endpoints.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Mental Health Inventory (MHI-5) scores

Table 2 shows prevalence rates of 12-month MDD and mean MHI-5 scores according to cerebrovascular risk factors and antidepressant medication use. Twelve-month prevalence rates of MDD were higher and MHI-5 scores were lower (representing greater psychological distress) for younger than for older participants (both p < 0.0001). After adjusting for age and sex, rates of 12-month MDD were higher and MHI-5 scores were lower (representing greater psychological distress) for participants who were current cigarette smokers, were obese, had previously had an MI, were of lower social class, reported hypertension treatment, and reported recent antidepressant medication use. Additionally, participants reported greater psychological distress (lower MHI-5 scores) if they had previously been diagnosed with diabetes, had higher total blood cholesterol, or had a lower level of educational attainment.

Table 3 shows that no association was observed between 12-month MDD and incident (fatal and nonfatal) stroke (hazard ratio 1.08 [95% CI 0.67, 1.75] after adjusting for age, sex, cigarette smoking, SBP, total cholesterol, obesity [BMI ≥ 30], preexisting MI, diabetes, social class, education, hypertension treatment, family history of stroke, and antidepressant medication use). In addition, no association was observed between lifetime MDD and (any fatal or nonfatal) stroke (hazard ratio 1.13 [95% CI 0.85, 1.50]), and no association was observed between either 12-month (0.81 [0.33, 1.98]) or lifetime GAD (0.58 [0.27, 1.24]) and (any fatal or nonfatal) stroke, with all adjustments as above.

Table 4 shows that a 1 SD decrease in MHI-5 scale score (representing greater psychological distress) was associated with an 11% increased risk of incident stroke after adjusting for all risk factors (as above). This association was consistent for men and for women (p = 0.46 for sex interaction), and was more pronounced for fatal than for nonfatal stroke endpoints. Figure 2 shows that this association conformed to a dose-response relationship with the risk of stroke progressively greater for participants reporting increased psychological distress. Table 5 shows that while the association between MHI-5 and stroke was strongest in the first 2 years of follow-up, it remained after excluding the first 6 years, and there was no evidence of nonproportional hazards (p = 0.39).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Hazard ratios (95% CIs) for incident (fatal and nonfatal) stroke, according to quintiles of Mental Health Inventory (MHI-5) score (where a lower score represents greater psychological distress), adjusted for age and sex

DISCUSSION

Based on data from 20,627 initially stroke-free participants in the EPIC-Norfolk study, we found no evidence that episodic mood state as defined by diagnostic criteria was associated with incident stroke. In contrast, we found evidence that a more general measure of current mental health well-being was associated with increased risk of incident stroke in men and in women and independently of known cerebrovascular risk factors. This association conformed to a dose-response relation and did not attenuate with increasing length of follow-up.

This large community-based population study has a number of strengths including prospective determination of endpoints through death certification and hospital record linkage for all study participants (with 167 fatal and 491 nonfatal strokes in 170,000 person-years of follow-up), availability of data for both men and for women, a measure of MDD defined according to core DSM-IV diagnostic criteria and a more general measure of mental health well-being, together with baseline assessments that allowed adjustment for a wide range of cerebrovascular risk factors.

However, the study has a number of important limitations. First, the restricted age range and characteristics of the cohort may reduce the generalizability of findings. Participation in the study involved commitment to future collection of detailed biologic and dietary data. While this resulted in a cohort that included fewer current cigarette smokers than the general resident population of England, the sample was representative in terms of anthropometric variables, blood pressure and serum lipids,¹³ and in terms of physical and mental functional health.²⁵ Secondly, given that health care systems vary in their use of inpatient services for the assessment and treatment of patients with stroke, and perhaps because these might vary most for less severe stroke,²⁶ less severe stroke endpoints may have been underascertained through use only of data from death certificates and from hospital admissions.

Limitations of measurement and of power should be taken into account when interpreting the finding of no association between episodic mood state and incident stroke in these data. The HLEQ structured self-assessment of MDD represents a pragmatic solution to enabling such measures to be included in a large-scale chronic disease epidemiology setting. Previous work (based upon a subsample of these data) has shown DSM-IV MDD prevalence estimates and associated demographic risk profiles to be broadly similar to those derived from dedicated large-scale psychiatric epidemiology research studies.¹⁶ This study had 80% power (at the 5% level of significance) to detect a relative risk of magnitude 1.33 for lifetime and 1.53 for 12-month MDD. While this sample was large, with nearly 600 stroke endpoints observed in over 8 years of (median) follow-up, the relatively low prevalence of the exposure variables limits the capacity of the current study to have detected associations of more modest magnitude. In addition, the potential for differential recall bias arising from the requirement of a 4-week retrospective recall period for the MHI-5 vs 1-year and lifetime periods of recall for the MDD and GAD diagnostic modules may provide an alternative explanation to the negative findings observed in relation to these measures here.

This community-based cohort study has been able to report associations between mental health status measures defined by a measure of psychological distress (previously identified as an effective screening instrument for mood disorders²¹), and by diagnostic criteria and incident (fatal and nonfatal) stroke. Our results are consistent with previous evidence, based upon a range of self-report depression symptom scales, that greater emotional symptomatic distress is associated with increased stroke risk.^4,8,9,11,12 A previous attempt to evaluate the association between a lifetime history of depressive disorder, defined through use of restricted diagnostic criteria, and stroke (involving a 13-year follow-up of 1,703 study participants) reported evidence of a two- to threefold increased risk of stroke but was limited by endpoint rarity (66 self-reported strokes and 29 stroke-related deaths during the follow-up), and by incomplete adjustment for cerebrovascular risk factors.⁷ A further registry study also reported evidence of association between patients admitted to hospital for depression and subsequent stroke but was unable to adjust for relevant risk factors.¹⁰ While previous evidence has shown anxiety, including GAD, to be common in patients for up to 3 years following stroke,^27,28 the authors are unaware of any previous examination of the association between GAD and stroke incidence. Given that depression is commonly comorbid with anxiety,²⁹ and that the MHI-5 includes a question concerned with current anxiety, our results suggesting no association between GAD and incident stroke, though limited by power considerations, are of interest in aiding collective interpretation of these findings.

Stroke is among the leading causes of long-term disability and death worldwide.³⁰ While evidence, based upon the Framingham Study original and offspring cohorts, has reported decreases in the incidence of stroke and atherothrombotic brain infarction over the past 50 years, achieved in part through improved risk factor control, continued efforts are needed to identify new risk factors that may aid primary prevention and reduce post-stroke disability and mortality.³¹ Our results respond to calls for work to aid insight into understanding the origin of the relationship between mood status and stroke risk.⁶ They focus attention on the conceptually more general nature of psychological distress and well-being, represented by the MHI-5, rather than the specific represented by MDD and GAD. However, limitations of statistical power in relation to both the measures of MDD and GAD diagnostic status warrant caution in discounting their respective association with incident stroke. In addition, difficulty in being able to demonstrate discontinuities or points of rarity between related psychiatric syndromes³² has remained a challenge in the specification of diagnostic criteria.^33,34 It is possible that psychological distress, within a broader spectrum of mental health variation, is part of a prolonged prodrome to stroke or that the association may be due to residual confounding. Further work is needed to replicate these findings, to consider emotional health associations according to stroke subtypes, and to evaluate potential mechanisms such as for example the debated association between homocysteine concentration and depression and stroke.^35–37

ACKNOWLEDGMENT

The authors thank the participants and general practitioners who took part in this study and the staff associated with the research program.