Risk Factors for Poststroke Cognitive Decline
The REGARDS Study (Reasons for Geographic and Racial Differences in Stroke)
Background and Purpose—Poststroke cognitive decline causes disability. Risk factors for poststroke cognitive decline independent of survivors’ prestroke cognitive trajectories are uncertain.
Methods—Among 22 875 participants aged ≥45 years without baseline cognitive impairment from the REGARDS cohort (Reasons for Geographic and Racial Differences in Stroke), enrolled from 2003 to 2007 and followed through September 2015, we measured the effect of incident stroke (n=694) on changes in cognitive functions and cognitive impairment (Six-Item Screener score <5) and tested whether patient factors modified the effect. Median follow-up was 8.2 years.
Results—Incident stroke was associated with acute declines in global cognition, new learning, verbal memory, and executive function. Acute declines in global cognition after stroke were greater in survivors who were black (P=0.04), men (P=0.04), and had cardioembolic (P=0.001) or large artery stroke (P=0.001). Acute declines in executive function after stroke were greater in survivors who had <high school education versus college graduates (P=0.01). Incident stroke was associated with faster declines in global cognition and executive function but not new learning or verbal memory compared with prestroke slopes. Faster declines in global cognition over years after stroke were greater in survivors who were older (P<0.01), resided outside the Stroke Belt (P=0.005), or had cardioembolic stroke (P=0.01). Faster declines in executive function over years after stroke were greater in survivors who were older (P<0.01) or lacked hypertension (P=0.03).
Conclusions—Incident stroke alters a patient’s cognitive trajectory, and this effect is greater with increasing age and cardioembolic stroke. Race, sex, geography, and hypertension status may modify the risk of poststroke cognitive decline.
Whether risk factors contribute to poststroke cognitive decline (PSCD) independent of prestroke cognitive decline is controversial.1,2 For example, older age and fewer years of education seem to increase the risk of PSCD2; however, because most studies lack a description of the trajectory of cognitive decline before the stroke, it is not clear how (or if) the stroke modified a process that was already underway. We previously showed that stroke is associated with an acute decline in cognitive function and also accelerated and persistent cognitive decline over years.3 However, most studies of risk factors for PSCD have not measured both of these cognitive changes. Identifying risk factors for PSCD may suggest potential mechanisms, future research directions, and at-risk groups to target with interventions.
We determined the sociodemographic, geographic, medical, and behavioral predictors of PSCD in a large, geographically dispersed, race- and sex-balanced sample controlling for prestroke cognitive trajectories. Blacks, women, rural adults, and Stroke Belt residents may have greater risk of PSCD owing to higher rates of diabetes mellitus4 and recurrent stroke,5 more severe strokes and cardioembolic strokes,6,7 greater burden of cerebrovascular pathology (eg, white matter hyperintensity and subclinical atherosclerosis),8,9 and lower educational attainment.5 Vascular risk factors, their treatment, or behavioral factors may contribute to PSCD through pathways of oxidative stress, inflammation, and endothelial dysfunction independent of stroke, but this is controversial.1,2,10 We examined 4 research hypotheses: an increased risk of PSCD will be associated with (1) older age and lower educational attainment; (2) black race, female sex, rural location, and Stroke Belt residence; (3) histories of hypertension, diabetes mellitus, and high cholesterol; and (4) physical inactivity and current cigarette smoking. We also examined the influence of stroke type (hemorrhagic versus ischemic) and ischemic stroke subtype on PSCD.2
Study Design, Participants, and Measurements
Study protocol is available at http://www.regardsstudy.org. Statistical code is available through written agreement with authors from Dr Levine (e-mail: ). Data set is available through a data use agreement with University of Alabama at Birmingham (e-mail: ). REGARDS (Reasons for Geographic and Racial Differences in Stroke) is a prospective cohort study of 30 239 non-Hispanic black and white individuals ≥45 years of age examining regional and racial influences on stroke mortality.11 Details are described elsewhere11,12 and in the online-only Data Supplement.
Participants or their proxies were followed every 6 months by telephone with retrieval of medical records for reported hospitalizations. For this study, we followed participants from recruitment (2003–2007) through September 30, 2015. To control for prestroke cognition, we required participants to have ≥1 cognitive outcome measurement and participants with incident stroke to have ≥1 cognitive measurement after stroke. We excluded participants with baseline moderate-to-severe cognitive impairment, including dementia, defined as a Six-Item Screener (SIS) score <5.13 The study was approved by the institutional review boards of all participating institutions, and all participants provided written informed consent.
Cognitive Function Assessments
Formally trained technicians administered cognitive function tests longitudinally by telephone, including (1) the SIS (primary outcome; score range, 0–6) measured global cognition annually starting in 200313 and (2) a battery of 3 cognitive tests (secondary outcomes) measured biannually starting in 2006 that included the Consortium to Establish a Registry for Alzheimer's Disease Word List Learning (new learning; score range, 0–30), Word List Delayed Recall (verbal memory; score range, 0–10), and Animal Fluency Test (AFT; executive function; score range, >0; online-only Data Supplement).14,15 These cognitive measures can be measured reliably and precisely by telephone,16–18 are included in the vascular cognitive impairment harmonization standards,19 and have been validated in blacks and whites.13,20,21 For all cognitive tests, higher scores indicate better performance.
Measurement of Incident Stroke
Physician experts adjudicated incident strokes using published guidelines and reviewed medical records, death certificates, and additional sources (online-only Data Supplement).11,22 Strokes were classified as hemorrhagic or ischemic (subtypes: cardioembolic, large vessel disease, small vessel disease, other, and unknown).
Descriptive characteristics were compared by incident stroke status. Linear mixed-effects models measured changes in cognitive scores over time and included the following: random effects for intercept and slope; covariates, including baseline score, for each cognitive outcome (Table 1); time as years from the date of the first measurement of the cognitive outcome; a time-varying covariate to estimate the effect of incident stroke on the acute decline in cognitive function at the time of the event (change in intercept); and a time after stroke covariate to estimate the effect of incident stroke on the rate of change in cognitive function (change in slope).3 Figure I in the online-only Data Supplement shows the conceptual model. We censored each cognitive outcome at the time of second incident stroke, death, loss to follow-up, or the end of follow-up.
We tested whether the set of risk factors for each hypothesis modified the effect of stroke on the intercept and slope of cognitive trajectories using interaction terms and the likelihood ratio test. If the addition of a set of interaction terms for a hypothesis significantly improved model fit, then we tested whether each individual risk factor in the set modified the effect of stroke on the intercept and slope of cognitive trajectories. We also estimated the odds of incident cognitive impairment (SIS <5/impaired versus ≥5/unimpaired)13 using generalized, linear mixed-effects models for a binary outcome and allowing a participant’s cognitive impairment to vary over time.
We calculated participant-specific (conditional) predicted values for each cognitive score and participant-specific predicted probabilities of incident cognitive impairment during follow-up time for a black woman with covariate values that were representative for the Stroke Belt population because it had a higher risk of cognitive decline relative to the remaining population. Random effects for this prediction were set to zero.
Statistical significance for all analyses was a 2-sided, adjusted P value <0.05. We report nominal P values, unless stated otherwise. All analyses used STATA software, version 14.1 (Stata Corporation, College Station, TX).
The study sample included 22 875 participants, 694 of whom experienced expert-adjudicated incident stroke (637 ischemic, 55 hemorrhagic, and 2 of undeterminable type) during a median follow-up of 8.2 years (interquartile range, 6.0–10.1 years). Figure II in the online-only Data Supplement shows the derivation of the cohort. Table I in the online-only Data Supplement compares characteristics between included and excluded participants. Table II in the online-only Data Supplement presents baseline characteristics of study participants.
Participants had a median of 7 SIS tests (interquartile range, 4–9 tests) and a median of 2 3-test batteries (interquartile range, 1–3 tests). Because the secondary outcome measures were introduced during follow-up and performed less frequently, the word list learning analysis included 12 955 participants, the word list delayed recall analysis included 12 738 participants, and the AFT analysis included 13 792 participants.
Effect of Age and Education
Table 1 presents the base model (model M0) without stroke x risk factor interaction terms. Stroke altered a patient’s trajectory of global cognition and executive function, and this effect was greater with increasing age (ie, baseline age modified the effect of stroke on changes in the slopes of global cognition [P<0.01] and executive function [P<0.01]; model M1 for SIS and AFT; Table 1). We illustrate results using participant-specific predicted values of cognition (Figure). Compared with younger survivors, older survivors had significantly faster declines in global cognition (adjusted difference in slope after incident stroke between survivors aged 75 years and survivors aged 65 years, −0.04 points per year; 95% confidence interval [CI], −0.06 to −0.03 points; P<0.01) and executive function (adjusted difference in slope after stroke between survivors aged 75 years and survivors aged 65 years, −0.6 points per year; 95% CI, −0.9 to −0.2 points; P<0.01) after stroke (Figure). The acute decline (intercept) in global cognition and executive function after stroke did not differ by age.
Unlike age, education level modified the effect of stroke on the acute decline in executive function (intercept; P=0.01). Compared with stroke survivors who were college graduates, stroke survivors who had <high school education had greater acute decline in executive function after stroke (adjusted difference in intercept after stroke between survivors with college degree and survivors with <high school education, −2.08 points per year; 95% CI, −0.45 to −3.72 points; P<0.01; model M1 for AFT; Table 1). However, education did not modify the effect of stroke on the slope of executive function (P=0.62). Changes in new learning and verbal memory after stroke did not vary by age and education (Table III in the online-only Data Supplement).
We also assessed SIS as a binary outcome. The odds of cognitive impairment after stroke were 1.17 (95% CI, 1.06–1.31; P=0.003) greater per each year of follow-up for each 10-year increase in baseline age (age model; Table 2; Figure III in the online-only Data Supplement). For a black woman living in the Stroke Belt with average values for all covariates at baseline and stroke at year 4, baseline age of 75 years compared with baseline age of 65 years was associated with greater risk of cognitive impairment: absolute difference in incident cognitive impairment was 8.4% (95% CI, 4.0%–12.7%) at year 4, 12.8% (95% CI, 8.8%–16.7%) at year 6, and 17.7% (95% CI, 12.1%–23.3%) at year 8 (Table 3).
Effect of Race, Sex, and Region
Race (P=0.04) and sex (P=0.04) modified the effect of incident stroke on the acute decline (intercept) in global cognition (model M2 for SIS; Table 1). The acute decline in global cognition after stroke was greater in blacks compared with whites (adjusted difference in intercept after stroke between blacks and whites, −0.11 points per year; 95% CI, −0.22 to −0.01 points; P=0.04) and greater in men compared with women (adjusted difference in intercept after stroke between men and women, −0.10 points per year; 95% CI, −0.20 to 0.00 points; P=0.04).
Stroke Belt residence (P=0.005) modified the effect of incident stroke on the change in the slope of global cognition (model M2 for SIS; Table 1). Declines in global cognition over the years after stroke were significantly faster in non-Stroke Belt residents compared with Stroke Belt residents (adjusted difference in slope after stroke between survivors residing within the Stroke Belt compared with survivors residing outside the Stroke Belt, −0.04 points per year; 95% CI, −0.07 to −0.01 points). The change in the acute decline in global cognition (intercept) after stroke did not differ by region. Changes in new learning, verbal memory, and executive function after stroke did not vary significantly by race, sex, and region.
Effect of Hypertension, Diabetes Mellitus, and Hyperlipidemia
Hypertension status (P=0.03) modified the effect of incident stroke on the slope of executive function (model M3 for AFT; Table 1). Compared with hypertensive survivors, nonhypertensive survivors had significantly faster declines in executive function after stroke (adjusted difference in slope after stroke, −0.58 points per year; 95% CI, −1.12 to −0.04 points; P=0.03; model M3 for AFT; Table 1). Changes in global cognition, new learning, and verbal memory after stroke did not vary significantly by presence of hypertension, diabetes mellitus, or hyperlipidemia at baseline.
Effect of Physical Activity and Smoking
Changes in global cognition, new learning, verbal memory, and executive function after stroke did not vary by physical activity and smoking status at baseline (model M4 in Table IV in the online-only Data Supplement for SIS and AFT; Table III in the online-only Data Supplement for word list learning and word list delayed recall).
Effect of Stroke Type and Ischemic Stroke Subtype
Ischemic and hemorrhagic strokes were associated with acute declines in global cognition and faster declines over years, although some results for hemorrhagic stroke were no longer statistically significant (model M5 for SIS; Table 1). Among ischemic stroke, cardioembolic and large vessel subtypes were associated with significant acute declines in global cognition (both P<0.001). In the years after cardioembolic stroke, global cognition declined significantly faster than it did before the stroke (P=0.01).
In this national cohort of black and white US residents of ≥45 of age, incident stroke accelerates declines in global cognition and executive function during 8 years, and this effect is greater with increasing age. The age-dependent risk for PSCD is over and above the age-associated risk for cognitive decline in those without stroke. The odds of cognitive impairment after stroke were 17% greater per 1 year of follow-up for each 10-year increase in baseline age.
Greater acute declines in global cognition after stroke were observed among stroke survivors who were black, men, and had cardioembolic or large vessel ischemic strokes. Greater acute declines in executive function after stroke were seen in survivors with <high school education. In the years after stroke, global cognition declined significantly faster in survivors who had cardioembolic ischemic stroke or who resided outside the Stroke Belt, and executive function declined significantly faster in survivors without hypertension, compared with prestroke slopes.
Our study has several strengths. We had repeated cognitive measurements over time ≤14 years of follow-up in a cohort and stroke subset of sufficient size to estimate stroke-related changes in cognition and to examine potential effect modification by participant factors. Our work extends previous research because we accounted for prestroke cognitive trajectories. REGARDS systematically measured cognitive domains commonly affected by stroke: global cognition, learning, memory, and executive function.19
Incident stroke may have a more detrimental effect on the cognitive trajectory of older adults through several mechanisms. Older adults are more likely to have neurodegenerative disease, atrophy, cerebrovascular disease, and comorbidity (eg, atrial fibrillation and other vascular risk factors) than younger adults, and these may amplify brain injury and cognitive deficits.23,24 Stroke may exacerbate25 or induce neurodegenerative disease.26 Older adults are more likely to have strokes that are severe or cardioembolic, and the latter were associated with faster cognitive decline perhaps because of injury to the middle cerebral artery brain territory. It is also plausible that the older brain is less efficient at generating and operating compensatory mechanisms to support cognitive decline associated with stroke and aging.27,28 Although we censored participant’s cognitive observations at the time of recurrent stroke, it is possible that older stroke survivors more frequently had subclinical vascular brain injury (eg, microinfarcts or microbleeds) after their index stroke that contributed to subsequent cognitive decline.
Modification of the effect of stroke on changes in cognition by race, sex, region, education, and hypertension history should be interpreted cautiously because we performed multiple comparisons. It is possible that we were unable to detect the effect of stroke on cognition and effect modification because of inadequate power particularly for secondary outcomes (although linear mixed-effects models maximize statistical power); the cognitive tests fail to capture PSCD; and the timing of cognitive measurements 12 to 24 months apart may underestimate early PSCD. The primary outcome was measured using the SIS—a brief cognitive test. Although the Figure seems to show that the exemplar patient age 65 years at baseline may experience deceleration of executive function decline during follow-up time after stroke, these results of a secondary outcome require confirmation.
Although the linear mixed-effect model is an approximation and may not work well at the tails of the distribution, we verified linearity and normality of residual errors assumptions of the model by inspecting residual plots. We did not have information on stroke features (laterality and severity), acute stroke treatments, brain imaging (infarct location and atrophy), incident dementia, or functional impairment. Results are generalizable only to community-dwelling stroke survivors not requiring a proxy respondent (eg, without aphasia). The slight increases in global cognition, new learning, and verbal memory over time before stroke may be attributable to practice effects, selective outmigration of cognitively impaired adults, and healthy survivor effect. Previous analyses that accounted for loss to follow-up or death did not change results.3
Incident stroke alters a patient’s cognitive trajectory, and this effect is greater with increasing age and cardioembolic stroke. Race, sex, geography, and hypertension status may modify the risk of PSCD.
We thank the other investigators, the staff, and the participants of the REGARDS study (Reasons for Geographic and Racial Differences in Stroke) for their valuable contributions. A full list of participating REGARDS investigators and institutions can be found at http://www.regardsstudy.org.
Sources of Funding
This research project is supported by a cooperative agreement U01 NS041588 from the National Institute of Neurological Disorders and Stroke (NINDS), National Institutes of Health (NIH), Department of Health and Human Service. Representatives of the funding agency have been involved in the review of the article but not directly involved in the collection, management, analysis, or interpretation of the data. This work also was supported by the National Institute on Aging (NIA)/NIH grant K23AG040278. Dr Levine received grant support from NIA/NIH R01 AG051827 and NIH/NINDS R01 NS102715.
Dr Levine reports consultant/advisory board work for Astra Zeneca/University of California at San Francisco (SOCRATES trial [Study to Prevent Major Vascular Events With Ticagrelor Compared to Aspirin in Patients With Acute Ischemic Stroke or TIA] event adjudicator), University of California at San Francisco (POINT trial [Platelet-Oriented Inhibition in New TIA and Minor Ischemic Stroke] event adjudicator), Program Advisory Committee of the Kaiser Permanente Northern California—University of California at San Francisco Stroke Prevention/Intervention Research Program. Dr Galecki received grant support from the Institute on Aging/National Institutes of Health P30AG024824. The other authors report no conflicts.
This content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of Neurological Disorders and Stroke or the National Institutes of Health.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.117.018529/-/DC1.
- Received July 5, 2017.
- Revision received December 8, 2017.
- Accepted January 3, 2018.
- © 2018 American Heart Association, Inc.
- Ivan CS,
- Seshadri S,
- Beiser A,
- Au R,
- Kase CS,
- Kelly-Hayes M,
- et al
- 4.↵Centers for Disease Control and Prevention. National Diabetes Fact Sheet: National Estimates and General Information on Diabetes and Prediabetes in the United States. Atlanta, GA: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention; 2011.
- Desmond DW,
- Moroney JT,
- Sano M,
- Stern Y
- Roquer J,
- Campello AR,
- Gomis M
- de Leeuw FE,
- de Groot JC,
- Achten E,
- Oudkerk M,
- Ramos LM,
- Heijboer R,
- et al
- Unverzagt FW,
- Monahan PO,
- Moser LR,
- Zhao Q,
- Carpenter JS,
- Sledge GW Jr,
- et al
- Wilson RS,
- Leurgans SE,
- Foroud TM,
- Sweet RA,
- Graff-Radford N,
- Mayeux R,
- et al
- Hachinski V,
- Iadecola C,
- Petersen RC,
- Breteler MM,
- Nyenhuis DL,
- Black SE,
- et al
- Ferraro FR
- Lucas JA,
- Ivnik RJ,
- Smith GE,
- Ferman TJ,
- Willis FB,
- Petersen RC,
- et al
- Halanych JH,
- Shuaib F,
- Parmar G,
- Tanikella R,
- Howard VJ,
- Roth DL,
- et al
- Whitehead SN,
- Cheng G,
- Hachinski VC,
- Cechetto DF
- Tatemichi TK,
- Foulkes MA,
- Mohr JP,
- Hewitt JR,
- Hier DB,
- Price TR,
- et al