Incidence and Prognostic Indicators of Dementia in Patients With Covert Brain Infarction, Transient Ischemic Attack, and Stroke: The Population‐Based Rotterdam Study

Abstract

Background

Cognitive impairment is common after transient ischemic attack (TIA) and stroke, but contemporary population‐representative estimates of dementia risk after stroke are scarce, particularly in view of stroke severity and competing risk of mortality.

Methods

We included individuals from the population‐based Rotterdam Study with first‐ever covert brain infarction (n=630), TIA (n=547), minor stroke (National Institutes of Health Stroke Scale score <4; n=392), or major stroke (National Institutes of Health Stroke Scale score ≥4; n=493) between 2002 and 2018, and matched those 1:3 to reference participants on age and sex. We determined 10‐year dementia risks by event severity, comparing cause‐specific and subdistribution hazards models to account for competing risk of death, and explored prognostic indicators of dementia after TIA and stroke.

Results

Of 1431 patients with first‐ever TIA or stroke (mean age 75.2 years, 58.3% women), 161 had pre‐event dementia and 205 developed dementia during a median follow‐up of 6.1 years. After 10 years, 59.4% of patients had died, with highest risk in the first months after major stroke. Compared with reference participants, dementia risk was increased after minor (cause‐specific hazard ratio [HR], 1.60 [95% CI, 1.21–2.12]) and major stroke (HR, 1.72 [95% CI, 1.29–2.30]), but not TIA (HR, 0.97 [95% CI, 0.76–1.23]). Among those with covert brain infarction, dementia risk was between that of TIA and minor stroke (HR, 1.34 [95% CI, 0.98–1.83]). Accounting for mortality, 10‐year dementia risk ranged from 14% (95% CI, 12%–19%) after TIA to 21% (95% CI, 16%–25%) after minor stroke and 16% (95% CI, 12%–20%) after major stroke. These risks were substantially higher in the Kaplan–Meier‐estimations for minor stroke (33%) and major stroke (40%). Prognostic indicators for dementia after TIA and stroke included higher age, less education, premorbid cognition, APOE‐ε4‐carriership, hypercholesterolemia, and on neuroimaging hippocampal and white matter hyperintensities volume, cortical thickness, and ≥2 cerebral microbleeds.

Conclusions

Dementia risk is elevated after stroke, and to a lesser extent covert brain infarction, but not after TIA. Excess risk extends to long‐term follow‐up for minor stroke, whereas competing risk of death attenuates risk after major stroke. Clinical and imaging indicators hold potential for personalized estimation of dementia risk.

Nonstandard Abbreviation and Acronym

SDH

subdistribution hazard

Clinical Perspective.

What Is New?

• In this population‐based cohort, the 10‐year risk of dementia was elevated after stroke and to a lesser extent covert brain infarction but not after transient ischemic attack.

• Several clinical and imaging markers, including premorbid cognition, hypercholesterolemia, hippocampal atrophy, and cerebral microbleeds, were prognostic indicators of dementia risk after transient ischemic attack or stroke.

What Are the Clinical Implications?

• These findings emphasize the importance of long‐term mitigation of dementia risk after stroke and covert brain infarction, potentially guided by clinical and imaging biomarkers for personalized preventive interventions.

Poststroke cognitive impairment is a major source of disability and functional dependence after stroke, ¹ and increases the risk of developing dementia in subsequent years. ² , ³ Precise risk estimates of poststroke dementia vary widely, with population‐based studies reporting a 7% risk in the first year after stroke, and hospital‐based studies showing up to 14% risk after severe stroke, a difference partly due to the inclusion of more severe stroke cases in hospital‐based studies. ² However, risk estimates from unselected populations outside of specialized stroke centers remain scarce. ⁴ , ⁵ Most studies do not differentiate by stroke severity, despite around two thirds of cerebrovascular events in the general population qualifying as minor strokes or transient ischemic attacks (TIA). ⁶ Moreover, few studies evaluate absolute dementia risk following covert brain infarction, which occurs in 8% to 28% of the middle‐aged to older population and has an incidence ∼5 times higher than stroke. ⁷ , ⁸ , ⁹

Two prior population‐based studies in the United Kingdom and the United States have shown that dementia risk after TIA and stroke increases with event severity. ⁴ , ⁵ Although these studies provide valuable insights, long‐term prognostic risk estimates based on contemporary data from generalizable populations are lacking. Assessing postevent dementia from 1987 to 2021, the ARIC (Atherosclerosis Risk in Communities) Study provides risk estimates over a long calendar period. Improvements in acute stroke treatment, secondary prevention, and changes in poststroke mortality may have meanwhile affected poststroke dementia risk. ¹⁰ , ¹¹ , ¹² Using contemporary data from 2002 onwards, the Oxford Vascular Study underlines the stepwise increase of dementia risk with event severity, up to 5 years of follow‐up. As time progresses after a stroke, secondary prophylaxis may become less stringent, for instance due to reduced medical follow‐up. ¹³ Longer‐term risk estimates could guide prognostic information and secondary preventive strategies, but no published studies have assessed long‐term risk beyond 5 years in a population representative sample. Furthermore, comparisons between clinical cerebrovascular events and covert brain infarctions—both frequently associated with dementia—remain unexplored.

Despite the high risk of dementia after TIA and stroke, there are currently no risk stratification tools that can guide individual patient prognosis in clinical practice. Event severity and lesion characteristics largely affect postevent cognition but cannot explain a significant portion of dementia risk, which may be attributable to premorbid factors, concomitant small vessel disease, and other neurodegenerative pathologies like Alzheimer disease. ² , ⁴ , ⁵ , ¹⁴ Several clinic‐based studies among patients with mostly major stroke have suggested that cardiovascular risk factors and structural brain changes may hold prognostic information for dementia risk after stroke. ² , ³ One prior study identified prognostic indicators for poststroke dementia risk in a population‐based setting, awaiting replication to establish robust predictors for personalized prognosis. ⁵ Importantly, such estimates of cumulative incidence of dementia and its prognostic indicators require incorporation of the competing risk of mortality. ⁵ Risk of mortality is particularly high after major stroke, which can lead to inflation of cumulative incidence figures and overestimate poststroke dementia risk. ¹⁵ , ¹⁶

We aimed to determine the risk of dementia in patients with covert brain infarction TIA and stroke in the community, stratified by event severity, and determined prognostic indicators of postevent dementia risk. We compared prognostic estimates while accounting for competing risk of mortality from subdistribution hazard models (SDH) to commonly used Kaplan–Meier.

METHODS

Data Sharing Statement

Data can be obtained upon request. Requests should be directed to the management team of the Rotterdam Study (secretariat.epi@erasmusmc.nl), which has a protocol for approving data requests. Because of restrictions based on privacy regulations and informed consent of the participants, data cannot be made freely available in a public repository.

Study Population

This study was embedded within the Rotterdam Study, an ongoing population‐based cohort study investigating determinants and occurrence of disease in persons aged 40 years and older. The study started in 1990 and now comprises 17 931 individuals living in the Ommoord suburb of Rotterdam, the Netherlands. The design of the Rotterdam Study has been described in detail previously. ¹⁷ In brief, participants are invited for interview and extensive in‐person examination at a dedicated research center once every 3 to 6 years. The current study includes all patients with first‐ever stroke (n=952) or TIA (n=563) or first diagnosis of covert brain infarction on routine brain imaging in the Rotterdam Study (n=630) between April 1, 2002 and December 31, 2018. We selected 2002 as the starting point for our analysis because it coincided with the initiation of the fourth examination wave of the Rotterdam Study. Before this period, stroke‐specific diagnostic approaches did not align with contemporary clinical standards. By the early 2000s, rapid‐access TIA clinics had been widely implemented in the Netherlands, and statins were increasingly used as standard secondary prophylaxis. ¹⁸ , ¹⁹ Furthermore, alteplase had become an established treatment during this time. ²⁰ , ²¹

Patients with TIA who had a subsequent stroke after a TIA were included for both events. Patients with insufficient cognitive screening at Rotterdam Study entry (n=36) were excluded. Brain magnetic resonance imaging (MRI) was implemented into the core protocol of the Rotterdam Study from 2005 onwards. Hence, for the MRI analyses, we included only patients with TIA and stroke events after 2005. A flow chart is presented in Figure S1.

All patients were matched 1:3 to reference participants (N=5251) based on event date, age (±3 years), and sex, using an incidence density sampling approach. ²² As such, reference participants for patients with stroke could not have had a stroke up to the index date, and reference participants for patients with TIA could not have had either a TIA or stroke up till the index date. Reference participants for individuals with covert brain infarction were free of TIA and stroke at time of their brain MRI and did not have signs of brain infarction on their scan. In accordance with the incidence density sampling method, reference participants could later be included as cases if they developed TIA, stroke, or covert brain infarction during the course of the Rotterdam Study. We report our study according to the Strengthening the Reporting of Observational Studies in Epidemiology guidelines.

Ethics Approval

The Rotterdam Study has been approved by the Medical Ethics Committee of the Erasmus MC and by the Ministry of Health, Welfare and Sport of the Netherlands, implementing the Population Screening Act: Rotterdam Study. ¹⁷ All participants provided written informed consent to participate in the study and to obtain information from their treating physicians.

Ascertainment of Clinical Stroke and TIA

Occurrence of TIA and stroke is assessed in all Rotterdam Study participants through routine home interviews and visits to the research center every 3 to 6 years. In addition, participants are continuously monitored for the occurrence of stroke and TIA through linkage of the study database with files from general practitioners and nursing home physicians, which included discharge letters from any hospital admission or outpatient clinic visit. Potential TIA and stroke cases were reviewed by research physicians, and an experienced vascular neurologist adjudicated the final diagnosis, as described in detail previously. ²³ Stroke was defined in accordance with the World Health Organization criteria as a syndrome of rapidly developing clinical signs of focal (or global) disturbance of cerebral function, with symptoms lasting 24 hours or longer or leading to death, with no apparent cause other than of vascular origin. ²⁴ We defined TIA as the presence of focal neurological symptoms, lasting no longer than 24 hours and attributable to dysfunction of one arterial territory of the brain. ²⁵ Stroke severity was assessed using a previously validated, record‐based adaption of the National Institutes of Health Stroke Scale (NIHSS). ²⁶ Minor stroke was defined as NIHSS score <4 and major stroke NIHSS score ≥4.

Information on stroke subtype and vascular territory was obtained from medical records. Stroke events were classified as hemorrhagic, ischemic, or unspecified in case of insufficient information. Vascular territory was based on clinical symptoms and classified as left hemisphere, right hemisphere, posterior circulation, and other (including >1 side and amaurosis fugax). Presence of aphasia was derived from medical records.

Assessment of Covert Brain Infarcts and Other Imaging Parameters

Participants underwent scanning on a single 1.5 Tesla MRI scanner (GE Healthcare) using a multisequence protocol consisting of T1‐weighted, proton density‐weighted, fluid‐attenuated inversion recovery, and proton density‐weighted sequences. All scans were appraised by trained research physicians for the presence of infarcts and cerebral microbleeds (ie, small round to ovoid hypointense areas on T2*‐weighted images). ²⁷ These ratings were done blinded to clinical data. Covert brain infarcts were defined as any cortical infarct or lacune on brain MRI in participants without a history of TIA or stroke.

For brain volumetry, T1‐weighted (voxel size 0.49×0.49×1.6 mm³), proton density–weighted (voxel size 0.6×0.98×1.6 mm³), and fluid‐attenuated inversion recovery (voxel size 0.78×1.12×2.5 mm³) scans were used for automated segmentation of brain tissues, including supratentorial gray matter, white matter, cerebrospinal fluid, and white matter hyperintensities using an in‐house segmentation tool. ²⁸ , ²⁹ All segmentations were visually inspected, and manually corrected if needed. Total brain volume was defined as the sum of white and gray matter. Supratentorial intracranial volume was estimated by summing gray and white matter (consisting of the sum of normal‐appearing white matter and white matter lesion volume) and cerebrospinal fluid volumes. ²⁹ Furthermore, T1‐weighted MR images were processed using FreeSurfer (version 6.0) to obtain measurements of hippocampal volume ³⁰ and cortical thickness. ³¹ Hippocampal volume was defined as the sum of the left and the right hippocampal volumes. We used both the mean cortical thickness as well as a surface area‐weighted average of the Alzheimer disease‐related cortical signature defined by Dickerson and colleagues. ³²

Outcome Assessment of Dementia and Mortality

Participants were screened for dementia at baseline and at each follow‐up interview, ³³ using the Mini‐Mental State Examination (MMSE)and the Geriatric Mental Schedule organic level. Those with an MMSE score of <26 or a Geriatric Mental Schedule organic level score of >0 were further examined using the Cambridge Examination for Mental Disorders in the Elderly diagnostic interview. ³³ Additionally, participants were continuously under surveillance for dementia through electronic linkage between the study database and medical records from general practitioners and the Regional Institute of Outpatient Mental Health Care. The final diagnosis of dementia was made by a consensus panel led by a neurologist based on standard criteria for dementia (Diagnostic and Statistical Manual of Mental Disorders, 3rd edition revised). Follow‐up for dementia was done until January 1, 2020. Vital status was ascertained through linkage with the municipal administration of Rotterdam.

Assessment of Prognostic Indicators

Educational attainment was assessed during interview at study entry, and for the purpose of the current analyses classified into primary education, or intermediate, and higher education. History of diabetes, coronary heart disease, or atrial fibrillation at time of stroke event were ascertained through a combination of in‐person assessments and medical record checks. For all other variables, we used information obtained during the closest available examination round preceding stroke, TIA, or covert brain infarction. Smoking habits were assessed by interview and categorized as never, former, or current smoking. Body mass index was calculated from measured weight and height (kg/m²). Hypercholesterolemia was defined as a non‐high‐density lipoprotein cholesterol ≥3.4 (total cholesterol minus high‐density lipoprotein cholesterol). ³⁴ Blood pressure was measured twice using a random‐zero sphygmanometer, and we used the average of 2 readings. Hypertension was defined as a systolic blood pressure ≥140 mm Hg or diastolic blood pressure ≥90 mm Hg, or use of antihypertensive medication. APOE genotype was determined using polymerase chain reaction on coded DNA samples or biallelic TaqMan assays for rs7412 and rs429358 (Thermo Fisher Scientific, Waltham, Massachusetts). ³⁵ , ³⁶ During home interview, trained research nurses performed an MMSE and enquired about instrumental activities of daily living based on 8 items with a score of 0 to 3 each ³⁷ : telephone use, medication maintenance, shopping, traveling on your own, finance management, laundry, housekeeping, and meal preparation. Impaired instrumental activities of daily living functioning was defined as a score>8. ³⁸ As indicated previously, imaging prognostic indicators included total brain volume, hippocampal volume, cortical thickness, Alzheimer disease cortical signature and white matter hyperintensities, and 0, 1, or ≥2 cerebral microbleeds.

Statistical Analysis

We had complete data on age and sex. Missing data for covariates were imputed using 5‐fold multiple imputations: level of education (1.2%), body mass index (2.5%), hypercholesterolemia (4.3%), smoking status (0.2%), hypertension (1.5%), diabetes (9.3%), atrial fibrillation (15.0%), MMSE score (2.5%), instrumental activities of daily living (35.5%), stroke subtype (0.8%), vascular territory (7.2%), and APOE genotype (5.0%).

First, we calculated the prevalence of dementia, with its corresponding 95% CI, at time of covert brain infarction, TIA, and minor and major stroke. Next, patients who were free of dementia before their event were matched 1:3 to reference participants, using an incidence density sampling approach as outlined previously. We compared baseline characteristics between covert brain infarction, TIA, and minor and major stroke and their respective reference participants.

We determined incident dementia risk after cerebrovascular events, stratified by event severity (ie, for covert brain infarction, TIA, minor stroke, and major stroke). We computed cause‐specific hazard ratios (HRs) for incident dementia, using Cox proportional hazard models censoring at time of death, and compared these with HRs from SDH models. All Cox proportional hazards models were assessed for violations of the proportional hazards assumption. This assumption was violated in models comparing TIA, minor stroke, and major stroke for the outcomes of dementia and death (Schoenfeld residuals test, P < 0.01). Despite this, Cox models were retained to allow comparison with SDH models. The latter were considered the primary analyses, as they more appropriately account for competing risks. Cox models are presented to illustrate how estimates differ under conventional survival analysis. In addition, all Cox models were assessed for influential observations and nonlinearity, with no substantial deviations detected. HRs from SDH models are particularly informative in prognostic research, whereas cause‐specific HRs best reflect pathogenesis. ³⁹ We also determined 2‐year, 5‐year, and 10‐year cumulative incidences of dementia by event severity, comparing estimates from the Kaplan–Meier estimator to those from the SDH models. In the presence of competing risk of death, the latter provide more reliable prognostic estimates, as violation of the independence of censoring assumption due to the competing event in the Kaplan–Meier estimator leads to overestimation of absolute risks. ¹⁵ , ¹⁶ In sensitivity analyses, we repeated the analyses for ischemic and hemorrhagic stroke separately, and further distinguished risk between major stroke of NIHSS score 4 to 10 and >10. We did not directly compare absolute risks after covert brain infarction to risk after TIA and stroke, as the requirement to undergo voluntary study brain MRI implies that individuals with covert brain infarction are somewhat healthier than their counterparts with TIA or stroke.

Next, among all individuals with TIA and stroke (i.e., all clinical events), we determined prognostic value of premorbid clinical characteristics and brain MRI measures, using univariable and multivariable SDH models. We excluded participants who were diagnosed with dementia, died, or were lost to follow‐up in the first 3 months after their TIA or stroke, in order to determine prognostic indicators that are applicable after the initial, acute phase of the event, for example during rehabilitation or at outpatient follow‐up. Moreover, we performed sensitivity analysis, including only prognostic indicators that were sampled at a research center visit within 5 years before the TIA or stroke event.

Finally, we explored whether stroke recurrence among patients with TIA and stroke was associated with dementia risk. We used time‐dependent Cox proportional hazard models, adjusted for age and sex, to compute cause‐specific HRs for dementia with recurrent stroke as a time‐varying exposure.

All analyses were done using R (version 4.2.1; using packages “mice”, “survival” and “tidycmprsk”). The α (type 1 error) was set at 0.05.

RESULTS

Participant Characteristics

Of 1755 included individuals with first diagnosis of cerebrovascular disease, 631 had covert brain infarction, 501 TIA, 357 minor stroke, and 394 major stroke. A total of 161 participants had a diagnosis of dementia before their cerebrovascular event (or MRI scan for covert brain infarction), with highest prevalence among those with major stroke (18.3% [95% CI, 15.1–21.9]), followed TIA (7.7% [95% CI, 5.7–10.2]) and minor stroke (7.4% [95% CI, 5.2–10.5]). Thus, for subsequent analyses on dementia incidence, we included all 1252 participants without dementia at time of their cerebrovascular event (Table 1).

Table 1.

Demographic and Clinical Characteristics of the Study Population

	Whole cohort	Covert brain infarction	Transient ischemic attack	Minor stroke	Major stroke
	No.=1755	No.=631	No.=501	No.=357	No.=394
Sex, female	968 (55.2)	298 (47.3)	308 (61.5)	204 (57.1)	218 (55.3)
Age, y, mean±SD	75.58±10.50	70.19±10.16	76.70±9.66	78.21±8.90	80.79±8.82
Education
Primary only	237 (13.5)	57 (9.0)	66 (13.2)	53 (14.8)	75 (19.0)
Intermediate	1231 (70.1)	437 (69.4)	357 (71.3)	258 (72.3)	274 (69.5)
Higher	265 (15.1)	129 (20.5)	69 (13.8)	43 (12.0)	39 (9.9)
Body mass index, mean±SD	27.57±4.05	27.54±4.10	27.73±4.10	27.60±3.98	27.29±3.98
Hypercholesterolemia	1282 (73.0)	443 (70.3)	388 (77.4)	251 (70.3)	288 (73.1)
Smoking status
Never	519 (29.6)	186 (29.5)	170 (33.9)	88 (24.6)	102 (25.9)
Former	886 (50.5)	314 (49.8)	255 (50.9)	182 (51.0)	204 (51.8)
Current	347 (19.8)	128 (20.3)	75 (15.0)	87 (24.4)	88 (22.3)
Hypertension	1406 (80.1)	487 (77.3)	402 (80.2)	293 (82.1)	334 (84.8)
Diabetes	323 (18.4)	114 (18.1)	80 (16.0)	72 (20.2)	90 (22.8)
Atrial fibrillation	193 (11.0)	37 (5.9)	43 (8.6)	50 (14.0)	78 (19.8)
Mini‐Mental State Exam score, mean±SD	27.59±2.04	27.74±2.01	27.73±1.82	27.35±2.13	27.24±2.30
Instrumental activities of daily living, mean±SD	6.92±4.28	5.63±4.07	7.23±4.27	7.26±4.00	7.58±4.43
Aphasia	239 (31.8)	0 (0.0)	0 (0.0)	44 (12.3)	195 (49.5)
Stroke subtype
Ischemic	588 (78.9)	630 (100.0)	501 (100.0)	311 (88.4)	277 (70.5)
Hemorrhagic	89 (11.9)	0 (0.0)	n/a	14 (4.0)	75 (19.1)
Unspecified	68 (9.1)	0 (0.0)	n/a	27 (7.7)	41 (10.4)
Vascular territory
Left hemisphere	448 (40.5)	n/r	160 (32.6)	138 (42.6)	181 (48.5)
Right hemisphere	384 (34.7)	n/r	130 (26.5)	122 (37.7)	152 (40.8)
Posterior circulation	134 (12.1)	n/r	48 (9.8)	61 (18.8)	29 (7.8)
Other*	140 (12.7)	n/r	153 (31.2)	3 (0.9)	11 (2.9)
APOE ε4 carriers	255 (15.3)	172 (27.8)	99 (21.0)	91 (27.2)	91 (24.8)

Numbers are n (%) unless specified otherwise. Data were complete for age and sex. Data were missing for level of education (1.2%), body mass index (2.5%), hypercholesterolemia (4.3%), smoking status (0.2%), hypertension (1.5%), diabetes (9.3%), atrial fibrillation (15.0%), Mini‐Mental State Exam (2.5%), activities of daily living (24.6%), instrumental activities of daily living (35.5%), stroke subtype (0.8%), vascular territory (7.2%), and APOE genotype (5.0%). N/a indicates not available; and n/r, not recorded.

^*Indicates >1 side or amaurosis fugax.

Included individuals with covert brain infarction were younger than patients with TIA or stroke, and those with major stroke more often had atrial fibrillation. TIA patients were less often current smokers compared with patients with stroke (Table 1). Nearly half of patients with major stroke had aphasia, compared with 12% of patients with minor stroke. Of all first‐ever strokes, 588/745 (78.9%) were ischemic and 89/745 (11.9%) had primary intracranial hemorrhage, whereas stroke type was undetermined in 68 (9.1%). Characteristics of each patient group in relation to the matched reference participants are depicted in Table S1.

Cumulative Incidence of Postevent Dementia and Mortality

During a median follow‐up time of 6.1 (interquartile range, 2.9–9.4) years, 974/7133 (13.7%) individuals developed dementia. With increasing event severity, diagnosis of dementia relied more often on medical records, as repeated cognitive screening during in‐person assessment was completed by fewer participants after major stroke (80/394; 20.3%) than after minor stroke (141/357; 39.5%), TIA (253/502; 50.5%), and covert brain infarction (393/630; 62.4%), as well as for reference participants (2999/5251; 57.1%). In total 882/1252 (70.4%) participants with TIA or stroke died during the follow‐up period, of whom 172 (19.5%) within the first 3 months after index event. Mortality was highest after major stroke (85.6%), followed by minor stroke (72.5%), TIA (56.3%), covert brain infarction (35.9%), and 49.7% for reference participants (Figure 1).

Figure 1. Cumulative incidence of dementia after transient ischemic attack and stroke, in relation to mortality risk.

The cumulative incidence of dementia during 10 years of follow‐up based on cause‐specific hazards (A) and subdistribution hazards (B). C shows the mortality risk, stratified by event severity. Patients with pre‐event dementia were excluded. NIHSS indicates National Institutes of Health Stroke Scale.

Risk of dementia increased with increasing event severity in cause‐specific hazard models (Figure 1A). Compared with age‐ and sex‐matched reference participants, risk of dementia was higher for patients with minor stroke (HR, 1.60 [95% CI, 1.21–2.12]) and major stroke (HR, 1.72 [95% CI, 1.29–2.30]), but not TIA (HR, 0.92 [95% CI, 0.71–1.20]) Excess dementia risk after major stroke was observed across the entire 10‐year follow‐up period, whereas excess risk following minor stroke increased over time (Figures 1A and 2). For individuals with covert brain infarction, cause‐specific hazards of dementia were between those for TIA and minor stroke (HR, 1.34 [95% CI, 0.98–1.83]; Figure 2).

Figure 2. Excess risk of dementia with covert brain infarction, transient ischemic attack, and stroke.

The cumulative incidence of dementia by diagnostic severity in comparison to matched reference participants. Results are shown for (A) covert brain infarction, (B) transient ischemic attack, (C) minor stroke (NIHSS score <4), and (D) major stroke (NIHSS score ≥4) across 10‐year follow‐up. The top row shows the cumulative incidences from the cause‐specific hazard models, whereas the bottom row illustrates cumulative incidences from the subdistribution hazard models. Patients with pre‐event dementia were excluded. CBI indicates covert brain infarction; NIHSS, National Institutes of Health Stroke Scale; SDH, subdistribution hazard; and TIA, transient ischemic attack.

Accounting for competing risk of mortality from a prognostic perspective, excess dementia risk attenuated, in particular for major stroke and to a lesser extent minor stroke (Figure 1). Compared with age‐ and sex‐matched participants, risk remained significantly increased after minor stroke (SDH ratio, 1.36 [95% CI, 1.03–1.79]) but not after major stroke (SDH ratio, 0.87 [95% CI, 0.65–1.17]; Figure 2). These results were in line with the high mortality in the early phase after major stroke, particularly for severe stroke with NIHSS score >10 (Figure S2). Accordingly, when assessing dementia risk from 3 months post stroke onwards, risk estimates after major stroke somewhat increased (SDH ratio, [95% CI, 1.16 0.84–1.61]) (Figure S3).

Absolute risk of dementia at 2 years after the index event ranged from 4.8% (95% CI, 3.2%–6.9%) for TIA, to 5.4% (95% CI, 3.4%–8.1%) for minor stroke and 8.5% (95% CI, 6.0%–11%) for major stroke (Figure 1B). On the basis of Kaplan–Meier estimates, 10‐year risk of dementia ranged from 20% (95% CI, 16%–25%) after TIA, to 32% (95% CI, 24%–39%) after minor stroke, and 40% (95% CI, 29%–48%) after major stroke (Figure 1). In contrast, accounting for mortality risk (Figure 1C), 10‐year dementia risk after major stroke was 16% (95% CI, 12%–20%), compared with 21% (95% CI, 16%–25%) after minor stroke, and 16% (95% CI, 12%–19%) after TIA (16%; 12–19).

Absolute dementia risk was similar after ischemic and hemorrhagic stroke (Figure S4), due in part to high mortality after hemorrhagic stroke (3‐month survival: 61/121, 50.8%). Risk of dementia after covert brain infarction was 1.3% (95% CI, 0.6%–2.4%) at 2 year after follow‐up in SDH models, and increased up to 3.8% (95% CI, 2.5%–5.6%) at 5 years and 11% (95% CI, 8.6%–15%) at 10 years after the index MRI‐scan (Figure S5). Compared with reference participants, dementia risk curves for covert brain infarction started to diverge from 4 years after study MRI (Figure 2).

Clinical Prognostic Indicators

Dementia risk was higher for older patients (per 1‐year increase, SDH‐ratio, 1.05 [95% CI, 1.03–1.06]), those with lower educational attainment (SDH‐ratio, 1.69 [95% CI, 1.18–2.40]), and genetic predisposition to dementia (APOE‐ε4‐carriership, HR, 1.97 [95% CI, 1.42–2.74]). Most cardiovascular prognostic indicators were not significantly associated with incident dementia except hypercholesterolemia (HR, 1.78 [95% CI, 1.15–2.75]; Table 2). Participants with premorbid cognitive impairment (MMSE score <24) showed higher dementia risk (HR, 2.01 [95% CI, 1.23–3.30]), whereas no such association was observed for pre‐event impairment in instrumental activities of daily living (HR, 1.11 [95% CI, 0.83–1.48]). However, in multivariable models that adjusted for other clinical prognostic indicators, the association between premorbid cognitive impairment and dementia risk was attenuated and no longer statistically significant (HR, 1.74 [95% CI, 0.95–3.18]). Dementia risk was lower after a left hemispheric event, unrelated to the presence of aphasia (Table 2). Prognostic indicators were broadly similar when including the first 3 months of follow‐up and when restricting analyses to patients with TIA and minor stroke only (Table S2). Moreover, prognostic indicators remained similar when stratifying and including only those indicators within 5 years before the cerebrovascular event.

Table 2.

Clinical Prognostic Indicators of Postevent Dementia

	Crude model	Adjusted for age and sex	Multivariable model
	Subdistribution hazard ratio (95% CI)	Subdistribution hazard ratio (95% CI)	Subdistribution hazard ratio (95% CI)
Female sex	1.28 (0.96–1.69)	1.10 (0.83–1.47)	1.23 (0.89–1.72)
Age, per y	1.04 (1.02–1.05)	1.04 (1.02–1.05)	1.04 (1.03–1.06)
Primary education only	1.54 (1.11–2.15)	1.31 (0.93–1.85)	1.37 (0.91–2.06)
Smoking status (ref: never smokers)
Former	1.15 (0.84–1.58)	1.34 (0.96–1.86)	1.40 (0.97–2.03)
Current	0.75 (0.49–1.17)	1.01 (0.64–1.61)	0.92 (0.56–1.52)
Hypercholesterolemia	1.67 (1.14–2.43)	1.62 (1.11–2.37)	1.78 (1.15–2.75)
Body mass index (kg/m2), per unit	0.99 (0.96–1.02)	0.99 (0.96–1.03)	0.98 (0.94–1.02)
Hypertension	1.20 (0.81–1.77)	1.01 (0.68–1.50)	1.06 (0.70–1.62)
Diabetes	1.19 (0.86–1.64)	1.21 (0.88–1.67)	1.33 (0.91–1.96)
Atrial fibrillation	0.89 (0.59–1.33)	0.76 (0.50–1.16)	0.69 (0.43–1.10)
Premorbid cognitive impairment (Mini‐Mental State Examination score <24)	2.01 (1.23–3.30)	1.94 (1.18–3.19)	1.74 (0.95–3.18)
Premorbid impaired instrumental activities of daily living score (<9)	1.11 (0.83–1.48)	0.85 (0.62–1.17)	0.91 (0.65–1.29)
Aphasia	0.77 (0.49–1.21)	0.72 (0.46–1.13)	1.09 (0.67–1.78)
Vascular territory (ref: right hemisphere)
Left hemisphere	0.69 (0.50–0.95)	0.67 (0.49–0.93)	0.69 (0.48–0.98)
Posterior circulation	1.03 (0.68–1.56)	0.98 (0.64–1.49)	1.06 (0.66–1.69)
Retina/other	0.79 (0.48–1.27)	0.80 (0.49–1.29)	0.85 (0.49–1.47)
APOE ε4‐carriers	1.82 (1.36–2.43)	1.85 (1.38–2.48)	1.97 (1.42–2.74)

Analyses of all 1046 participants with transient ischemic attack and stroke. Subdistribution hazard ratios are presented with 95% CIs. All models further included event severity (ie transient ischemic attack, minor stroke, and major stroke).

Imaging Prognostic Indicators

Of 913 patients who had a TIA or stroke after implementation of brain MRI in the Rotterdam Study protocol, 436 (47.8%) had had a brain scan, a median 1.9 years (interquartile range, 0.8–3.05) before their event. Participants that underwent brain MRI were slightly younger, more often male, and higher educated compared with their counterparts without MRI (Table S3). Prevalence of cerebral microbleeds was 33.4%, with 20.5% having ≥2 microbleeds. Median volume of white matter hyperintensities was 7.1 cm³ (interquartile range, 5.3–9.3).

MRI features indicative of higher postevent dementia risk were hippocampal volume (HR per 1 SD increase, 2.59 [95% CI, 1.74–3.86]), cortical thickness (HR, 2.21 [95% CI, 1.67–2.93]), Alzheimer disease cortical signature (HR, 2.13 [95% CI, 1.62–2.80]), white matter hyperintensity volume (HR, 1.56 [95% CI, 1.30–1.87]), and presence of ≥2 microbleeds (compared with no cerebral microbleeds; HR, 3.07 [95% CI, 1.65–5.69]) (Table 3). After mutual adjustment for other imaging markers, burden of white matter hyperintensities and microbleeds remained statistically significant (Table 3).

Table 3.

Prognostic Indicators of Postevent Dementia on Brain MRI

	Crude model		Adjusted for age and sex		Multivariable model*
	Subdistribution hazard ratio (95% CI)	P value	Subdistribution hazard ratio (95% CI)	P value	Subdistribution hazard ratio (95% CI)	P value
Total brain volume	1.23 (0.98–1.55)	0.08	1.21 (0.85–1.73)	0.29	1.16 (0.75–1.80)	0.50
Hippocampal volume	2.59 (1.74–3.86)	<0.001	1.79 (1.08–2.94)	0.023	1.58 (0.92–2.71)	0.10
Cortical thickness	2.21 (1.67–2.93)	<0.001	1.65 (1.17–2.33)	0.005	1.33 (0.91–1.95)	0.14
Alzheimer disease cortical signature	2.13 (1.62–2.80)	<0.001	1.57 (1.12–2.21)	0.009	N/A†
White matter hyperintensities	1.56 (1.30–1.87)	<0.001	1.36 (1.10–1.68)	0.005	1.28 (1.01–1.62)	0.044
Number of microbleeds
0	Reference		Reference		Reference
1	1.25 (0.48–3.26)	0.64	0.93 (0.35–2.44)	0.88	0.86 (0.30–2.51)	0.78
≥2	3.07 (1.65–5.69)	<0.001	2.89 (1.58–5.30)	<0.001	2.49 (1.33–4.68)	0.005

Analyses among 436 participants with transient ischemic attack and stroke with available brain MRI scan. Estimates for continuous measures reflect a change per 1 SD increase. Models including total brain volume, hippocampal volume, and white matter hyperintensities are corrected for intracranial volume. All models further included event severity (ie, transient ischemic attack, minor stroke, and major stroke). MRI indicates magnetic resonance imaging.

^*Includes age, sex, education, event severity and imaging covariables (total brain volume, hippocampal volume, white matter hyperintensities, microbleeds, and cortical thickness).

^† Not estimated due to high correlation with overall cortical thickness (r=0.975).

Recurrent Stroke

In total, 1246/1252 (99.5%) of the participants with TIA and stroke had follow‐up information on recurrent stroke. Of those, 195 (15.7%) experienced a stroke event during follow‐up, a median 3.3 years (interquartile range, 0.15–6.5) after their index event. Stroke recurrence was associated with a moderate, nonsignificant increase in dementia risk (HR, 1.31 [95% CI, 0.87–1.98]), more profound for patients after an initial TIA (HR, 2.00 [95% CI, 1.03–3.88]) rather than after first‐ever stroke (HR, 0.95 [95% CI, 0.56–1.61]).

DISCUSSION

We found in this population‐based setting that dementia risk is increased after stroke and persists over long term‐follow up, particularly after minor stroke. After major stroke, long‐term risk is attenuated due to high mortality, and accounting for competing risk of mortality substantially attenuated absolute dementia risk estimates. Dementia risk was not increased after TIA. Excess dementia risk with covert brain infarction was between that of TIA and minor stroke. Event severity, combined with clinical and imaging characteristics, was related to dementia risk after TIA and stroke.

Results of the current study provide contemporary, population‐based estimates of poststroke dementia risk, stratified by event severity, illustrating the importance of accounting for competing risk of mortality. When comparing our findings to the Oxford Vascular Study, we observed a lower cumulative incidence of dementia, even after considering the competing risk of mortality. ⁵ Inclusion of some patients with recurrent stroke in Oxford, as opposed to first‐ever stroke only in the current study, may explain part of the difference. Moreover, the dementia screening in the Oxford Vascular Study included several in‐person or telephone cognitive assessments during the first 2 years after stroke, which may detect cases of cognitive impairment that go unnoticed with longer visit‐to‐visit intervals and reliance on in‐person examination or medical record‐based follow‐up for dementia. In contrast to the findings from Oxford, aphasia did not predispose to dementia in the current study, which may reflect a more conservative approach to diagnosis in this group of patients in our study.

Mortality after stroke was high, particularly after major strokes, with an 85% mortality rate at 10 years. In the Oxford Vascular Study, approximately half of all major stroke patients died within 5 years of their initial event. ⁵ This high mortality risk is reflected in the lower estimates of poststroke dementia after considering competing mortality risk. Our study showed up to 24% lower dementia risk estimates in SDH models, similar to the roughly 20% lower risk when using this type of modeling in the Oxford population. These findings emphasize the necessity of accounting for competing mortality risk to obtain accurate poststroke dementia absolute risk estimates. Owing to the prolonged, 10‐year follow‐up duration in our study, we were able to capture these competing risks of mortality in the long term. The lower excess risk of mortality with minor stroke, compared with major stroke, underlines the importance of long‐term, stringent secondary prevention also with relatively minor functional impairment, with the aim of reducing further cognitive decline as well as stroke recurrence risk. It is important to note that more recent advancements in stroke care, such as the introduction of endovascular thrombectomy and the early use of dual antiplatelet therapy following ischemic events, have been implemented in clinical practice only in recent years. ⁴⁰ , ⁴¹ These interventions were not standard practice before 2015, ⁴⁰ and their potential influence on long‐term outcomes, including the incidence of poststroke dementia, remains to be elucidated.

Our findings indicate no significant increase in dementia risk following a TIA over the study period, in contrast to a relatively small risk increase in the Oxford Vascular Study. ⁵ However, patients who experienced a stroke after their TIA were at 2‐fold increased risk of developing dementia. This highlights the critical importance of initiating secondary prophylaxis immediately after a TIA, which can reduce the risk of recurrent stroke by up to 80%. ⁴² , ⁴³ The absence of increased dementia risk in stroke patients with a recurrent stroke suggests that pre‐event factors may play a more significant role in dementia prognosis, rather than the occurrence of a second stroke. Our analyses were insufficiently powered to stratify the impact of event recurrence by stroke severity.

To our knowledge, this study is the first to contextualize dementia risk following covert brain infarction alongside symptomatic events. The observed 30% to 40% relative risk increase for dementia with covert brain infarction aligns with findings from an earlier meta‐analysis. ⁷ These results strengthen the evidence that people with covert brain infarction are at increased risk of developing dementia, with risks higher than TIA but lower than symptomatic stroke. Because covert brain infarctions in the Rotterdam Study are not routinely reported back to participants, secondary prevention efforts were likely less strict than after clinical events. This suggests that targeted prevention could further reduce the risk of clinical sequelae after covert brain infarction. Until now, the absolute risks of dementia post covert brain infarction, especially in the context of mortality risk, were undetermined. Compared with reference participants, dementia risk was notably higher after long‐term follow‐up. However, the first‐year risk is likely underestimated due to the healthy participant effect, as individuals who voluntarily undergo MRI in a study setting are generally healthier compared with those in an emergency department setting. Additionally, it is important to note that the exact timing of the covert brain infarction and subsequent follow‐up is unknown, as only the scan date is available. Larger studies are necessary to differentiate the absolute risks of dementia following covert brain infarction by clinical and imaging characteristics and in more diverse populations.

A prior study within the Rotterdam Study demonstrated that, despite patients with stroke experiencing steeper cognitive declines both in the 10 years preceding their first‐ever event and after the stroke itself, the stroke itself exerts the most profound impact on cognition. ⁴⁴ Building on these insights, the current study adds clinical and imaging factors, alongside event severity, as tools to refine individualized dementia risk stratification after stroke. This can help counsel patients, tailor follow‐up strategy, and recruit patients for preventative trials. When comparing our results with the 1 prior study investigating prognostic indicators of poststroke dementia in an unselected population, higher age, lower premorbid cognition, and cerebrovascular imaging markers are consistently associated with dementia risk after stroke and TIA. ⁵ Together with the role of APOE, these findings underline that dementia risk after stroke is driven by concomitant small vessel and neurodegenerative pathology, as well as stroke characteristics. ⁴⁴ Most cardiovascular risk factors were not significantly related to dementia from a prognostic perspective, although diabetes and hypercholesterolemia may hold some prognostic value. Although factors such as hypertension are causally linked to dementia, their high prevalence within the TIA/stroke population likely reduces their prognostic utility, as they do not differentiate risk effectively in this already high‐risk group. ⁴⁵ We observed a lower dementia risk in patients with left hemispheric strokes compared with right hemispheric strokes, whereas prior studies more specifically have shown that both the left frontotemporal lobes and the right parietal lobes are predictive of poststroke dementia. ¹⁴

Our current study is strengthened by the large population‐based sample with detailed information on pre‐event prognostic indicators including clinical and imaging characteristics and by the inclusion of participants with cerebrovascular events across the spectrum of severity, including covert brain infarction. Some limitations also need to be acknowledged when interpreting our results. First, we did not examine patients in person in the acute phase after TIA or stroke, which may have led to some underestimation of stroke severity if not all symptoms were recorded in medical notes. Second, half of patients with stroke did not revisit our study center for in‐person dementia assessment, and telephone interview of patients and caregivers was not available. For these patients, we had to rely on clinical, primary care, and nursing home records for dementia diagnosis, which may have led to underdiagnosis of dementia if cognitive deficits went unnoticed or unreported in routine care. Third, the follow‐up methodology of the Rotterdam Study does not permit interval diagnoses of mild cognitive impairment between research visits. Therefore, we were unable to assess the incidence of mild cognitive impairment after stroke, as this would lead to substantial attrition and unreliable prognostic estimates. This area warrants further investigation. Fourth, information on prognostic indicators was obtained from the research center visit before the TIA or stroke event. Although this provided valuable information on premorbid health status, substantial time intervals in some cases between measurement and stroke event may have caused misclassification. Furthermore, for the prognostic indicator activities of daily living, we observed 35% missing data. However, analyses comparing complete cases with imputed data showed no significant differences. Fourth, brain imaging data at time of diagnosis were not available to determine infarct size and location. Finally, Rotterdam Study participants are predominantly White, and generalizability to populations from other ethnicities is uncertain. Populations from other ethnicities may have a higher prevalence of risk factors such as uncontrolled hypertension and diabetes, alongside inequities in access to care, which could contribute to more severe strokes and an increased risk of poststroke dementia. ⁴⁶ , ⁴⁷ Moreover, the incidence of intracerebral hemorrhage is higher in Asian populations, ⁴⁸ which may coincide with higher prevalence of cerebral amyloid angiopathy and higher dementia risk.

CONCLUSIONS

In conclusion, dementia risk is increased after stroke but not after TIA with contemporary secondary preventive treatment. Excess dementia risk persists over long‐term follow‐up for minor stroke, and competing risk of death attenuates the risk after major stroke. Several clinical and imaging indicators hold potential for personalized risk estimation of dementia in patients following a cerebrovascular event.

Sources of Funding

The Rotterdam Study is supported financially by the Erasmus Medical Center, Rotterdam, The Netherlands; the Organization for Scientific Research; the Netherlands Organization for Health Research and Development (ZonMw); the Research Institute for Diseases in the Elderly; the Netherlands Genomics Initiative; the Ministry of Education, Culture, and Science; the Ministry of Health, Welfare, and Sports; the European Commission (DG XII); and the Municipality of Rotterdam. This study is part of ABOARD, which is a public–private partnership receiving funding from ZonMW (#73305095007) and Health~Holland, Topsector Life Sciences & Health (PPP‐allowance; #LSHM20106). More than 30 partners participate in ABOARD. ABOARD also receives funding from Edwin Bouw Fonds and Gieskes‐Strijbisfonds. Furthermore, this study was supported by a Alzheimer Nederland (grant number: WE.03‐2023‐16; PI Wolters), and funding from the Netherlands Organisation for Health Research and Development (ZonMw) and Alzheimer Nederland in the context of the Memorabel project support for the Netherlands Consortium of Dementia Cohorts (NCDC) (project number: 73305095005).

Dr Wolters receives research funding from the Netherlands Organisation for Health Research and Development (Veni‐09150162010108 and BIRD‐NL‐10510032120005), the Dutch Heart Foundation (CVON2018‐28), the Alzheimer’s Association (AARF‐22‐924 982), the Erasmus Trust Fund, and the Cure Alzheimer’s Fund, all paid to the institution.

Meike Vernooij is recipient of TAP‐dementia, a ZonMw funded project (#10510032120003) in the context of the Dutch National Dementia Strategy.

The funding sources that contributed to the Rotterdam Study or to authors had no role in study design, data collection, data analysis, data interpretation, or writing of the report and in the decision to submit the paper for publication.

Disclosures

None.

Supporting information

Tables S1–S3

Figures S1–S5

STROBE Checklist