Disability trajectories before and after stroke and myocardial infarction: the Cardiovascular Health Study

Abstract

Importance

Ischemic strokes may accelerate long-term functional decline apart from their acute effects on neurological function.

Objective

To test whether the increase in long-term disability is steeper after than before the event for ischemic stroke but not myocardial infarction (MI).

Design, Settings, and Participants

The population-based prospective cohort Cardiovascular Health Study (1989-2013). Longitudinal follow-up for a mean 13 years. Models based on generalized estimating equations adjusted for baseline covariates and included a test for different slopes of disability before and after event. Participants included Medicare-eligible individuals age >65 years, not institutionalized, expected to reside in the area for 3 years, and able to provide informed consent. Exclusions were needing a wheelchair, receiving hospice care, radiation treatment, or chemotherapy. Mean age of the entire cohort (n=5888) was 72.8 (SD 5.6) years; 42% were male.

Exposure

During follow-up, 382 participants had ischemic stroke, and 395 had MI with >1 disability assessment after the event. Models tested for slope differences before and after these events.

Main Outcome/measure

annual assessments with a disability scale (measuring activities of daily living [ADL] and instrumental ADLs). The number of ADLs and IADLs (range 0-12) that the participant could not perform was analyzed continuously.

Results

Participants had a mean of 4 disability assessments before and after events. The increase in disability around the time of the event was greater for stroke (0.88 points on the disability scale, 95% CI 0.57-1.20, p<0.0001) than MI (0.20, 0.06-0.35, p=0.006). The annual increase in disability before stroke (0.06 points per year, 0.002-0.12, p=0.04) more than tripled after stroke (0.15 additional points per year, 0.004-0.30, p=0.04). The annual increase in disability before MI (0.04 points per year, 0.004-0.08, p=0.03) did not change significantly after MI (0.02 additional points per year, −0.07-0.11, p=0.7).

Conclusions and Relevance

In this large population-based study, a trajectory of increasing disability became significantly steeper after stroke but not after MI. Thus, in addition to the acute brain injury and consequent impairment, ischemic stroke may also be associated with potentially treatable chronic adverse effects on the brain that lead to accelerated functional decline.

According to the traditional view (Figure 1A), stroke is thought to cause disability acutely, followed by a 3-to-6-month recovery period, after which disability stabilizes unless recurrent events occur.^1–4 However, stroke may accelerate the accumulation of disability over time, beyond disability progression due to non-pathological cognitive aging.⁵ According to this new paradigm of the effect of vascular brain injury on functional status (Figure 1A), disability may accelerate over time after recovery from stroke, even without recurrent clinical events.

Figure 1. Conceptual Depiction of Concepts Examined.

A) The trajectory of disability in relation to stroke

B) Estimated parameters

Several lines of evidence support this paradigm. Stroke risk factors of diabetes and hypertension have a cumulative effect on vessel dysfunction, leading to ongoing vascular brain injury. Also, ischemic stroke may cause delayed neuronal death and neurodegeneration,⁶ changes in local inflammation and systemic inflammatory profiles,^7–9 progressive cardiovascular impairment and reduced fitness,¹⁰ and covert brain infarcts in the absence of clinically recognized overt events.

In order to determine the unique effect of stroke on disability trajectories, the comparison group of myocardial infarction (MI) is ideal, because those with MI have similar distributions of vascular risk factors and experience a sudden event, in contrast to a healthy community-based control group with no anchor point from which to measure follow-up. Such an anchor is essential because previous research has demonstrated trajectories that vary based on time from the event. We hypothesized that the slope of increasing disability over the long term is steeper after recovery from stroke than before stroke but similar before and after MI.

Methods

Cardiovascular Health Study (CHS) participants (n=5888) were recruited from 1989-1993 from a sex- and age-stratified random sample of Medicare-eligible individuals in 4 states,¹¹ aged ≥65 years, not institutionalized, expected to reside in the area for ≥3 years, and able to consent. Individuals needing a wheelchair or receiving hospice care, radiation treatment, or chemotherapy were excluded. Baseline sociodemographic, functional, and health data were obtained from interviews, clinical examinations, medical record abstraction, and publicly released Medicare claims data, as previously described.^{11, 12} The institutional review board at each participating center approved the study, and each participant gave written informed consent.

Follow-up

Potential events were identified through regular contact with participants and proxies.¹³ Data on vascular events were collected at local sites and adjudicated by centralized endpoint committees for stroke and MI.¹⁴ Stroke was classified as ischemic (lacunar, cardioembolic, atherosclerotic, or indeterminate), hemorrhagic, or unknown.¹⁵ Potential incident MI cases were adjudicated based on a review of clinical history of cardiac symptoms, elevated cardiac enzyme levels, and serial electrocardiographic changes.

Study outcomes

Disability was assessed annually by the activities of daily living (ADL) and instrumental ADL (IADL) scale, modified from the National Center for Health Statistics Supplement on Aging¹⁶ and the New Haven Established Populations for Epidemiologic Studies of the Elderly Study.¹⁷ The scale assesses ability to carry out ADLs (walking around the home, getting out of bed, eating, dressing, bathing, using the toilet) and instrumental ADLs (heavy housework, light housework, shopping, preparing meals, paying bills, using the phone). The scale is scored from 0-12 based on number of ADLs and IADLs with which the participant reported having difficulty or could not perform, and was analyzed as a continuous variable as in previous research.^{12, 18} In secondary analysis, the scale was dichotomized as non-disabled (score of 0) and disabled (score ≥1). Disability assessments were missing in 14% of measurements among those with stroke (8% before and 18% after) and 9% with MI (5% before and 15% after).

Explanatory variables

Covariates were assessed at baseline and included demographic variables (self-reported age, sex, race-ethnicity, level of education), lifestyle variables (self-reported smoking, alcohol consumption, physical activity) and vascular risk factors (body mass index, hypertension, diabetes mellitus, cardiac disease, hyperlipidemia, as previously defined¹⁵) The strength of participants’ social networks was assessed with the Lubben Social Network Scale,¹⁹ a validated 10-item measure that includes assessments of five aspects of social networks. Depression was defined by a score of >9 on the Centers for Epidemiologic Studies Depression (CES-D) scale.²⁰ The Modified Mini-Mental State Examination score²¹ was assessed at 1 year of follow-up. Personal income was defined as total family income before taxes from all sources in the past 12 months and was categorized into <$12,000, $12,000–$34,999, and ≥$35,000, based upon prior analyses.²² Due to associations in prior studies between vascular outcomes and levels of the inflammatory biomarker C-reactive protein (CRP) and the atherothrombotic marker lipoprotein-A, we adjusted for baseline CRP and lipoprotein-A levels, log-transformed because of skewed distributions; measurement and sample processing have been previously described.^{23, 24}

Statistical analysis

Distributions of baseline characteristics and follow-up times, stroke and MI events during follow-up, and disability assessments before and after events were examined.

Stroke cohort

We sought to determine whether the slope of disability differed before and after stroke. Participants without prevalent stroke who experienced ischemic stroke during follow-up and had ≥1 disability assessment post-stroke were included. Due to correlations among repeated measures of outcomes in the same participant, regression models based upon generalized estimating equations (GEE) with an exchangeable correlation structure and robust standard errors were used, with an identity link function for continuous disability and a logit link function for dichotomous disability.

Assessments of disability occurring within 6 months after stroke (n=163) were ignored, since the course of recovery during this period is well-documented, and our interest was the long-term course of disability. Follow-up was censored at the time of recurrent stroke. The primary covariate was time of follow-up, and the parameter term associated with this signified the slope of increasing disability. The model included a product term (between post-stroke status and time of follow-up) that allowed for a different slope before and after stroke, and allowed for a direct test of a difference in slope, as follows:

$$\text{Y }=\text{ intercept }+{\text{ β}}_1\ast \text{FU }+{\text{ β}}_2\ast \text{poststroke }+{\text{ β}}_3\ast \text{FU}\ast \text{poststroke }+\sum \text{ β}\ast \text{covariates},$$

where FU=follow-up time, poststroke=0 if the time of follow-up was before the stroke, and 1 if after the stroke. β₁ estimated annual change in disability before the event (Figure 1B-a), β₂ estimated change in disability around the time of event (Figure 1B-b), and β₃ estimated additional annual change in disability after the event (Figure 1B-c). This modeling strategy has been used in several previous studies.^25–27

In model building, we sequentially added groups of variables, including demographics, vascular risk factors, social variables, and cognitive and mood factors. In order to maximize power to estimate the primary associations of interest, we removed variables not significant at a p-value cutoff of 0.1, while retaining demographic variables. We tested non-linearity of time trends with quadratic terms, and non-linearity was lacking.

MI cohort

We conducted an analysis similar to that outlined above, except that the event of interest was MI instead of stroke, and participants were free of prevalent MI at baseline. Hence, the models assessed the slope of increasing disability before and after MI in those who had MI during follow-up allowing for a drop in function after MI. Disability assessments within six months of MI were included, since the 3 to 6-month course of recovery documented with stroke does not exist with the same biological implications as with stroke.^{28, 29} Follow-up was censored at the time of recurrent MI.

Entire cohort

In order to compare directly the change in disability around the time of event (Figure 1B-b) between stroke and MI, we performed another analysis in which we included the entire CHS cohort. We used GEE models as above. For the determination of events, we considered the first stroke or MI only.

Sensitivity analyses

First, different cutoffs of the disability scale were tested systematically, dichotomizing at each level of the scale in unadjusted and adjusted models to determine whether there was a threshold effect at a particular cutoff. In order to determine whether the trajectories of disability before stroke and MI were different between those who had these events and those who did not experience stroke or MI, we compared these slopes to those in the whole cohort excluding those who experienced stroke, MI, or both, in unadjusted and fully adjusted models.

The impact of ischemic stroke subtype on disability trajectories was tested by stratifying models by stroke subtype and comparing trajectories of disability before and after stroke. In order to assess for bias due to differential mortality between MI and stroke, we performed an analysis in which the worst possible disability score was assigned at the time of death.

Results

Among those free of stroke at baseline (n=5639), during follow-up, 415 incident strokes occurred with ≥1 disability assessment after stroke, of which 382 were ischemic; 305 had >1 post-stroke disability assessment. Mean follow-up time was 11.1 (SD 5.0) years in this group, with a mean of 3.7 (SD 2.4) assessments before stroke and 3.7 (SD 2.3) after stroke. Among those free of MI at baseline (n=4734), 395 incident MIs occurred with ≥1 disability assessment after MI; mean follow-up time was 12.4 (SD 5.4) years, with a mean of 3.8 (SD 2.5) assessments before MI and 3.8 (SD 2.4) after MI. The mean baseline age was similar across the entire cohort, stroke cohort, and MI cohort (Table 1). Men were more common than women in the MI cohort. The prevalence of vascular risk factors was higher in the stroke and MI cohorts than in the overall cohort. In the overall cohort, the mean baseline disability score was 0.59 (SD 1.13).

Table 1.

Baseline characteristics of study population

Variable	Entire cohort	First ischemic stroke with ≥1 follow-up assessment	First MI with ≥1 follow-up assessment
Number of participants, No. (%)	5888 (100)	382 (100)	395 (100)
Age, mean (SD), y	72.8 (5.6)	74.1 (5.7)	73.2 (5.3)
Age at event, mean (SD), y	–	78.3 (5.8)	77.5 (5.7)
Body mass index, mean (SD), kg/m2	26.7 (4.7)	26.6 (4.4)	26.9 (4.5)
Male, No. (%)	2495 (42.4)	162 (42.4)	221 (56.0)
Non-Hispanic white, No. (%)	4925 (83.6)	332 (86.9)	351 (88.9)
Non-White, No. (%)	963 (16.4)	50 (13.1)	44 (11.1)
Received at least high school education, No. (%)	3352 (57.1)	234 (61.6)	275 (69.6)
Marital status, No. (%) married	3893 (66.2)	255 (66.9)	249 (63.2)
Yearly income, No. (%)
<$12,000	1470 (26.7)	118 (32.5)	95 (25.4)
$12,000–$34,999	2779 (50.5)	175 (48.2)	194 (51.9)
≥$35,000	1259 (22.9)	70 (19.3)	85 (22.7)
Hypertension	3457 (58.8)	281 (73.6)	265 (67.1)
Diabetes mellitus, No. (%)	1739 (29.9)	143 (37.9)	140 (35.6)
Current smoking, No. (%)	601 (11.6)	38 (10.9)	48 (13.2)
Hypercholesterolemia, No. (%)	1241 (21.1)	86 (22.5)	89 (22.5)
Atrial fibrillation, No. (%)	236 (5.3)	31 (11.4)	15 (5.1)
History of coronary heart disease, No. (%)	1154 (19.6)	106 (27.8)	76 (19.2)
History of myocardial infarction, No. (%)	562 (9.5)	63 (16.5)	0
Modified mini mental state score, mean (SD)	90.6 (5.7)	89.6 (6.1)	90.2 (5.7)
Depressed (CES-D score >9)	809 (13.8)	55 (14.5)	58 (14.8)
Arthritis	3025 (52.0)	219 (57.9)	231 (58.8)
Ischemic stroke subtype, No. (%)
Lacunar	N/A	75 (19.6)	N/A
Cardioembolic		107 (28.0)
Atherosclerotic		28 (7.3)
Hemorrhagic transformation		4 (1.1)
Indeterminate		179 (48.0)

In the fully adjusted model of the stroke cohort (Table 2), the mean change around the time of stroke was 0.45 points (95% CI −0.05, 0.95, Figure 1B-b). The annual change in disability score before stroke was 0.06 points per year (95% CI 0.002, 0.12, Figure 1B-a), with an additional 0.15 points per year after stroke (95% CI 0.004, 0.30, Figure 1B-c). In these models, assessments of disability were censored after recurrent stroke. The pattern of associations was similar when a dichotomous definition of disability was used (0 vs. ≥1, results not shown). Different cutoffs of the disability scale were tested systematically in adjusted and adjusted models, and no definite threshold effect was found for a particular cutoff of the disability score (results not shown).

Table 2.

Trajectories in the stroke cohort of disability before and after stroke

Variable	Change in disability score	95% confidence limits	p-value
Unadjusted model:
Annual change before stroke*	0.16	0.11, 0.21	<.0001
Change in disability score around time of stroke†	1.21	1.62, 5.75	<.0001
Additional annual change after stroke‡	0.09	0.22, 1.24	0.2
Adjusted for demographics:§
Annual change before stroke	0.15	0.10, 0.20	<.0001
Change in disability score around time of stroke	1.21	0.78, 1.63	<.0001
Additional annual change after stroke	0.12	−0.02, 0.26	0.09
Fully adjusted:‖
Annual change before stroke	0.06	0.002, 0.12	0.04
Change in disability score around time of stroke	0.45	−0.05, 0.95	0.08
Additional annual change after stroke	0.15	0.004, 0.30	0.04

^*see Figure 1B-a

^†see Figure 1B-b

^‡see Figure 1B-c

^§adjusted for age at time of stroke, sex, race, marital status, and income

^‖additionally adjusted for: arthritis, depression, log of lipoprotein A levels, Modified Mini-Mental State Examination score, and social network score

In the MI cohort (Table 3), the mean change around the time of MI was 0.34 points (95% CI 0.07, 0.61) in a fully adjusted model. Because the term for additional annual change after MI was nonsignificant in all models (Figure 1B-c), the slope of change was similar before and after MI in unadjusted and adjusted models with recurrent MI censored.

Table 3.

Trajectories in the myocardial infarction cohort of disability before and after myocardial infarction

Variable	Change in disability score	95% confidence limits	p-value
Unadjusted model:
Annual change before MI*	0.13	0.09, 0.18	<.0001
Change in disability score around time of MI †	0.36	0.11, 0.60	0.004
Additional annual change after MI‡	−0.04	−0.13, 0.04	0.3
Adjusted for demographics:§
Annual change before MI	0.15	0.10, 0.20	<.0001
Change in disability score around time of MI	0.35	0.11, 0.60	0.004
Additional annual change after MI	−0.03	−0.12, 0.05	0.4
Fully adjusted:‖
Annual change before MI	0.04	0.00, 0.08	0.03
Change in disability score around time of MI	0.34	0.07, 0.61	0.014
Additional annual change after MI	0.02	−0.07, 0.11	0.7

^*see Figure 1B-a

^†see Figure 1B-b

^‡see Figure 1B-c

MI=myocardial infarction

^§adjusted for age at time of MI, sex, race, marital status, and body mass index

^‖additionally adjusted for diabetes, arthritis, depression, log of C-reactive protein levels, Modified Mini-Mental State Examination score, and social network score

In the entire CHS cohort (Table 4), the change around the time of stroke (0.88 points, 95% CI 0.57, 1.20) was greater than around the time of MI (0.20 points, 95% CI 0.09, 0.20, p-value for difference=0.037). Also, the slope of increasing disability was steeper after stroke compared to before stroke (0.14 additional points per year, 95% CI 0.09, 0.20), but not after MI (0.01 points per year, 95% CI −0.02, 0.04).

Table 4.

Trajectories of disability before and after stroke and myocardial infarction in the entire cohort (n=5888)

Variable	Change in disability score	95% confidence limits	p-value
Annual change	0.11	0.10, 0.12	<.0001
Change in disability score around time of stroke	0.88	0.57, 1.20	<.0001
Change in disability score around time of MI	0.20	0.06, 0.35	0.006
Additional annual change after stroke	0.14	0.09, 0.20	<.0001
Additional annual change after MI	0.01	−0.02, 0.04	0.4

In unadjusted and fully adjusted models, the disability trajectory before stroke was similar to the trajectory in the whole cohort excluding stroke, MI, or both. When the worst possible disability score was assigned at the time of death, results were similar to the primary analysis: in a fully adjusted model, the mean increase in the disability score around the time of event was significant for stroke (0.68 points, 95% CI 0.41, 0.96) but not for MI. Increasing disability after event was present for stroke (0.05 points per year additional increasing disability score after event, 95% CI −0.001, 0.10) but not for MI.

Trajectories of disability before and after stroke were examined in ischemic stroke subtypes. Among those with lacunar stroke (n=75), change in disability score around the time of stroke was not significant and the magnitude was small, but a trend was evident for increasing disability score after stroke (0.33 additional points per year, 95% CI −0.06, 0.72). For cardioembolic stroke (n=107), the change in disability score around the time of stroke was significant (1.52 points, 95% CI 0.67, 2.37), and a trend was evident for increasing disability score after stroke (0.25 additional points per year, 95% CI −0.02, 0.53). For “other” ischemic strokes (n=211), the change in disability score around the time of stroke was significant (1.37 points around the time of stroke, 95% CI 0.85, 1.90), but a trend was lacking for increasing disability score after stroke.

Discussion

In this large population-based study, the slope of increasing disability after recovery from stroke was higher compared to before stroke, a pattern not evident before and after MI. A significant increase in disability around the time of event was evident for stroke and less so MI. Among the cohort of those who had stroke during follow-up, the slope of increasing disability after stroke was more than 2 times the slope before stroke. In all of these models, disability measurements after recurrent stroke were censored, so the estimated disability trajectories were independent of recurrent clinically overt stroke. This study is novel for at least two reasons. First, we examined disability not at a single follow-up time but estimated trajectories over time, including the extent to which stroke and MI altered these trajectories around the time of the event (Figure 1B-b) and over time after the event (Figure 1B-c). Second, instead of comparing trajectories after stroke to a stroke-free population with fewer vascular risk factors and without an acute event to anchor the estimation of trajectories, we compared to disability trajectories in those with MI, who have similar vascular risk profiles as stroke and experience an acute vascular event requiring hospitalization. This study provides unique and valuable information about the effect of stroke on disability trajectories in the elderly.

Stroke is the leading cause of serious disability in the U.S.³⁰ Stroke is traditionally seen as a discrete, monophasic event, and functional status has been assumed to stabilize following the 3 to 6-month recovery period after stroke, unless recurrent events occur. However, we present evidence that a single ischemic stroke continues to be associated with a gradual increase in disability over the long-term after stroke. The discrete stroke event may have long-term and ongoing effects on disability. We showed that participants who eventually have a stroke do not have a higher slope of increasing disability before stroke than those who do not eventually have a stroke.

Although the power to test subtypes of ischemic stroke was limited, slopes of disability among different ischemic stroke subtypes were similar. Further study would clarify whether a trajectory of increasing disability is steepest for lacunar subtype, which could reflect underlying covert but progressive small vessel disease. Cardioembolic and “other” subtypes were associated with an increase in disability around the time of stroke, whereas this was not found for lacunar strokes, reflecting a relatively milder phenotype with lacunar strokes.

This study is one of the few that provides data on disability trajectories related to stroke. Among initially stroke-free participants in the Northern Manhattan Study,²⁵ 210 participants experienced an ischemic stroke during follow-up and lived more than 6 months after stroke. When stratified by insurance status, among those with Medicaid or no insurance, in a fully adjusted model, slope of change in functional status before and after stroke differed significantly (p=0.04), with a decline in the 100-point Barthel Index of 0.58 BI points per year before stroke (p=0.02) and 1.94 after stroke (p=0.001). In the Health and Retirement Survey, the course of functional and cognitive impairment was compared before and after 232 hospitalizations for stroke and 450 hospitalizations for MI.³¹ Using a combined measure of ADLs and IADLs, disability increased more around the time of stroke than MI, similar to our findings in CHS.

Several lines of evidence support the paradigm, proposed here, of progressive brain dysfunction caused by cerebrovascular injury (Figure 1A). First, stroke is caused by conditions, including vascular risk factors and inflammation, that may have an ongoing and cumulative effect on vessel and neuronal function, including small vessels in the case of lacunar stroke, and carotid arteries in the case of large artery strokes.^{32, 33} In addition to causing recurrent strokes, vascular risk factors cause subclinical or covert brain injury manifest as infarcts and leukoaraiosis that may reduce functional status over the long term.^{34, 35}

Alternatively, an individual stroke may cause brain injury that leads to a chronic and degenerative process with progressive damage, dysfunction, and functional decline. Neuronal death through apoptosis and necrosis in the ischemic penumbra may be delayed.⁶ Furthermore, a single ischemic stroke may cause local changes in inflammation and an increase in systemic inflammatory profiles⁷ resulting in ongoing deleterious effects on brain structure and function⁸ that may persist years after stroke.⁹ Recent animal work and human pathological studies, for example, demonstrate that B-lymphocyte-mediated autoimmunity after stroke, due to exposure to neuronal antigens, is associated with delayed cognitive decline.³⁶ These data suggest that some stroke patients may develop a B-lymphocyte response to stroke that contributes to cognitive and functional decline.³⁷

Another possible mechanism of delayed post-stroke decline involves progressive cardiovascular impairment and reduced fitness due to static functional impairment that disproportionately affects stroke patients compared to those with MI. This cardiovascular, non-neurological impairment adversely effects performance in ADLs.¹⁰ Although we adjusted for baseline depression, post-stroke depression may also be responsible for a proportion of the decline we observed. Finally, the concept of “cognitive reserve” explains differing susceptibility to cognitive impairment based on variables such as education, literacy, intelligence quotient, and engagement in leisure activities.³⁸ An analogous concept, “functional reserve,” may explain how a deficit caused by stroke may result in a depleted functional reserve and a consequent failure to compensate for brain aging. This would appear as accelerated increase in disability after recovery from stroke, as we found in this study.

This study has several strengths. CHS is a large, nationally representative cohort of elderly community-dwelling participants with long-term follow-up. A sensitive measure of disability, including both ADL and IADL items, was measured regularly. Surveillance and adjudication of vascular events resulted in a substantial amount of data surrounding vascular events to estimate trajectories, with a mean of 4 annual measurements of disability before and after both stroke and MI. The study also has limitations. Since we required disability measurements before and after the event to estimate trajectories reliably, sample sizes were not large. Also, detailed information about the stroke, such as location, size, and severity, were lacking, as was neuroimaging data through follow-up to determine the influence of recurrent covert infarcts and progressing leukoaraiosis on accumulating disability.

Several implications follow from our findings. First, stroke intervention trials that measure recurrent events or disability at a single follow-up time as outcomes may miss the progressive disability that we observed. Second, some therapies may be effective at altering the adverse functional trajectories that we found, even if they do not prevent strokes. For example, psychoactive or immunotherapies may influence the trajectories of decline. Third, recognition of the role of stroke in exacerbating neurodegeneration could have implications for development of therapies for neurodegeneration.

Question

Is the slope of disability different before stroke and after recovery from stroke?

Findings

In this large population-based study, the slope of increase in disability was threefold greater after recovery from stroke compared to before stroke. The slope before and after the comparison event, myocardial infarction, was not different.

Meaning

Stroke may be associated with potentially treatable chronic adverse effects on the brain that lead to accelerated accumulation of disability.