Cognitive Impairment after Mild Stroke: Development and Validation of the SIGNAL 2 Risk Score

Abstract

Background:

Post stroke cognitive impairment (PSCI), an important complication of strokes, has numerous risk factors. A scale adequately classifying risk of cognitive impairment 3–6 months after mild stroke will be useful for clinicians.

Objective:

To develop a risk score based on clinical and neuroimaging variables that will be useful in identifying mild ischemic stroke patients at high risk for PSCI.

Methods:

The risk score development cohort comprised of a retrospective dataset of 209 mild stroke patients with MRI confirmed infarcts, without pre-stroke cognitive impairment, and evaluated within 6 months post-stroke for PSCI. Logistic regression identified factors predictive of PSCI and a risk score was developed based on regression coefficients. The risk score was checked for stability using 10-fold cross-validation and validated in an independent prospective cohort of 185 ischemic mild stroke patients.

Results:

Within 6 months post-stroke, 37.32% developed PSCI in the retrospective dataset. A 15-point risk score based on age, education, acute cortical infarcts, white matter hyperintensity, chronic lacunes, global cortical atrophy, and intracranial large vessel stenosis was highly predictive of PSCI with an AUC of 0.829. 10.11% with low scores, 52.69% with moderate scores, and 74.07% with high scores developed PSCI. In the prospective validation cohort, the model had an AUC of 0.776, and exhibited similar accuracy and stability statistics at both 6 and 12 months.

Conclusion:

The seven item risk score adequately identified mild stroke patients who are at an increased risk of developing PSCI.

Keywords

Cognitive impairment ischemic stroke magnetic resonance imaging prognosis risk score

INTRODUCTION

Strokes continue to be a major cause of disability worldwide [1, 2]. Aggressive acute stroke care has resulted in an increase of stroke survivors with both physical and cognitive disability [3]. Post-stroke cognitive impairment (PSCI) results in poor quality of life, slow physical recovery, and higher mortality [2 , 5]. This is particularly relevant to patients with mild strokes, as physical impairment may be non-disabling, but cognitive impairment may be significant, resulting in loss of employment and social abilities.

Existing literature suggest that at 3–6 months post-stroke, the prevalence of vascular dementia (VaD) and PSCI is 21.3% and 36.7% [6]. At 1-year post-stroke the prevalence of VaD and PSCI is reported to be 12.2% and 51.5% [7, 8]. PSCI may present with varying degrees of impairment in global cognition, executive impairment, episodic memory impairment, impairment in information processing speed, language deficits, and impairment in visuospatial function [9, 10]. Factors associated with cognitive impairment following ischemic strokes include older age, low education, infarct size and location, presence of white matter disease, and severity of vascular risk factors such as hypertension and diabetes mellitus [8 , 14]. In a meta-analysis involving 7,511 patients, the rates of post-stroke dementia at or before one year ranged from 7.4% in population-based to 41.3% in hospital-based studies [15].

A risk score for cognitive impairment applicable to stroke patients with mild physical disability has several important uses. Clinicians may use the score to identify and monitor patients at risk for dementia, so that cognitive rehabilitation and pharmacological interventions may be initiated [16]. The risk score may also be useful to select high-risk patients into clinical trials targeting PSCI. The objective of this study was to develop a risk score based on clinical and neuroimaging variables that can identify mild stroke patients who are at significant risk for developing PSCI within 6 months post-stroke.

MATERIALS AND METHODS

Risk score development dataset

Analysis was carried out using two independent datasets, with one for model development and the other for validation. The model development dataset was a retrospective dataset of patients who presented to a tertiary stroke clinic following discharge from inpatient care for acute mild ischemic stroke between 2008 and 2012. During inpatient care, patients were assessed for cognitive symptoms and underwent MRI and MR Angiography (MRA). Patients who were assessed by the clinical teams as being at risk for developing PSCI were scheduled for outpatient follow-up within 3–6 months after incident stroke. During outpatient care, patients were assessed for progression of risk factors and for cognitive status using the Mini-Mental State Exam (MMSE) [17]. Where clinically indicated or further confirmation of cognitive status was required, the Montreal Cognitive Assessment (MoCA) [18] was carried out as well.

From this database, we included patients with mild ischemic stroke, defined as having a modified Rankin scale of ≤2 with an MR-confirmed acute infarct. Exclusion criteria included 1) presence of pre-stroke cognitive impairment, 2) absence of MRI performed during acute stroke work-up, and 3) presentation to outpatient clinic outside 3–6 months.

Data on demographic factors and vascular risk factors were collected from admission records. MR images were reviewed independently by a neurologist and a neuroradiologist. Acute infarcts were quantified based on the number and location of lacunar and non-lacunar infarcts. T2 sequences were used to quantify white matter hyperintensity (WMH) using the Fazekas scale [19]. Chronic lacunes were quantified by number and location. Gradient-echo sequences were used to rate microhemorrhages using the Microbleeds Anatomical Rating Scale [20]. MRA images were rated for presence and severity of intracranial large vessel stenosis [21]. T1 sequences were also assessed for global cortical atrophy (GCA) [22]. Any differences in ratings between raters were resolved by consensus.

Development of risk models

Subjects in the development dataset were classified as PSCI if there were any cognitive symptoms during the clinical visit, and scored either MMSE ≤25 or MoCA ≤22, following validated local cutoffs [23]. PSCI patients did not meet criteria for dementia. Patients who did not meet criteria for PSCI or dementia were categorized as “No Cognitive Impairment (NCI)”. We considered a wide range of factors based on existing literature on PSCI, including demographics (age, gender, education, ethnicity), vascular risk factors (diabetes mellitus, hypertension, hyperlipidemia, smoking, atrial fibrillation), genetics (apolipoprotein E), post-stroke complications (depression, cognitive impairment, physical disability), acute stroke characteristics (location, size, and number of infarcts), and pre-stroke ischemia (presence of chronic lacunes, and burden of WMH, microhemorrhages, and intracranial stenosis) [8 , 25]. As the primary aim was to develop a clinical risk score, only factors available during clinical stroke workup were considered. The operationalization and criteria for studied factors is described in the supplementary section (Supplementary Material).

Potential variables for the predictive model were identified by comparing PSCI and NCI patients and tested for significance using independent sample t-test or Wilcoxon-Mann-Whitney test for continuous data, and χ² test or Fisher’s Exact test for categorical data. Statistically significant continuous variables were converted into categorical variables based on clinical cutoffs and retested using χ² test or Fisher’s Exact test. All variables significant at univariate level were put into multivariate logistic regression models as predictor variables, with PSCI status as the outcome variable. A point system was developed based on the β coefficients from the final model.

Accuracy and stability analysis

Predictive accuracy was evaluated based on discrimination and calibration. Discrimination was assessed in receiver operating characteristic (ROC) analysis using the area under the curve (AUC), from which an optimum cutoff was identified based on a balance between sensitivity and specificity.

We categorized patients as having low, moderate, or high risk scores based on a tertile split and calculated the percentage of PSCI patients in each risk category. Calibration was assessed by plotting predicted probabilities of PSCI and actual percentages of patients who developed PSCI at every point of the risk score. The model was also assessed for stability via 10-fold cross validation.

Validation cohort

An independent cohort was used to carry out internal validation of the risk score. This was a prospectively recruited cohort of patients recruited from our stroke clinic between 2012 and 2014 following discharge from inpatient care for an acute ischemic stroke. Informant Questionnaire on Cognitive Decline in the Elderly [26] was used to determine status of pre-stroke cognition and data on Instrumental Activities of Daily Living (iADL) were collected [27].

As per the development cohort, demographic, clinical, and MRI and MRA data were collected from patient records. Both MMSE and MoCA were conducted for all subjects in this cohort. In addition, patients underwent neuropsychological assessments that looked at episodic memory (Ten Word Immediate and Delayed Recall task), attention/working memory (Digit Span Forward task), executive function (Frontal Assessment Battery and Sunderland Clock Drawing task), language (animal fluency task and fruit fluency task), and screened for depression (Patient Health Questionnaire). A subsection of these patients also consented to an additional visit 12–18 months post-stroke and the neuropsychological assessment was repeated.

Internal validation of model

The subjects in the validation cohort were similarly classified into PSCI and NCI based on the same criteria, both at 3–6 months and 12–18 months post-stroke. The strength of the model in screening for PSCI was tested at both time points via ROC analysis and calculating the accuracy and reliability statistics based on the previously identified cutoffs. These figures were compared with results obtained from the development dataset for concordance.

All statistical analyses were performed using Stata version 10.1 (StataCorp, College Station, TX, USA) statistical software. All tests were two-tailed and statistical significance was set to p < 0.05 except during stepwise elimination during model development regression, where p < 0.2 was used in order to prioritize model sensitivity. All study procedures were carried out in accordance with institutional guidelines and under approval by the institutional ethics review board. Participants provided voluntary informed consent prior to collection of any research data.

RESULTS

For the development dataset, we identified 243 patients with MRI-confirmed mild acute infarcts. We excluded 34 patients (six presented outside of 3 to 6 months and 28 had incomplete clinical or investigative data). A total of 209 subjects were analyzed in the development dataset. The mean age was 61.67 years (SD 12.46 years) and the mean education was 4.59 years (SD 4.46 years). 32.06% of subjects were female, and 82.78% were of Chinese descent, with the remaining 17.22% being a mix of Malay, Indian, Eurasian, and others minority ethnic races. The mean interval between acute infarct and follow-up evaluation was 3.66 months (SD 3.24 months), by which time 78 subjects (37.32% ) were found to be PSCI.

In PSCI subjects, age at incident stroke was higher, education tended to be lower, there was a higher proportion of females, and hypertension and atrial fibrillation were more prevalent. With regards to neuroimaging characteristics, GCA, WMH, cortical non-lacunar infarcts, chronic lacunes, and intracranial stenosis were all worse in PSCI subjects (Table 1). In multivariate logistic regression for demographic and clinical variables, age (β= 1.06, 95% CI: 0.50–1.62; p < 0.001) and education (β= 1.77, 95% CI: 1.00–2.54; p < 0.001) remained significant after reverse stepwise regression. For the neuroimaging variables, WMH (β= 0.60, 95% CI: 0.18–1.03; p = 0.005), GCA (β= 0.40, 95% CI: –0.14–0.94; p = 0.147), non-lacunar acute infarcts (β= 0.39, 95% CI: –0.02–0.80; p = 0.061), chronic lacunes (β= 0.58, 95% CI: –0.12–1.28; p = 0.106), and intracranial stenosis (β= 1.23, 95% CI: 0.57–1.90; p < 0.001) all survived stepwise regression. A scoring system was derived from the β coefficients (Table 2).

A 15-point risk score was developed and had an AUC of 0.829 (95% CI: 0.77–0.88). The risk score was given the acronym SIGNAL₂ (Stenosis, Infarct type, Global Cortical Atrophy, Number of years of education, Age, Leuokoariosis/White Matter Hyperintensity, Lacune count). A cutoff of≥7 was 73.21% accurate, and had a sensitivity of 82.05% , specificity of 67.94% , positive predictive value (PPV) of 60.38% , and negative predictive value (NPV) of 86.41% . Prevalence of PSCI across the tertiles of the scale were 10.11% at the lowest tertile, 52.69% at the medium tertile, and 74.07% at the highest tertile. In 10-fold cross validation, the iterations remained relatively stable, ranging from 0.761–0.880. Calibration demonstrated that the risk scale closely follows statistical prediction models (Fig. 1).

For the validation cohort, 185 subjects were prospectively recruited. Compared to the development dataset, these subjects were younger [mean age 59.60 years (SD 11.18 years)] and more educated [mean years of education 9.01 years (SD 3.17 years)], while the gender (31.46% females) and race (84.27% Chinese) ratios were comparable. The mean NIHSS score at time of acute stroke was 3.7 (SD 2.1). Stroke subtypes based on the TOAST classification were; 35.3% Small Vessel Occlusion subtype, 31.1% Large Artery Atherosclerosis Subtype, 25.5% Cardioembolic subtype, and 8.1% undefined etiology.

35 (18.92% ) were identified as PSCI at 3–6 months post-stroke [mean interval time 3.13 months (SD 2.38 months)]. PSCI subjects were older; less educated, were more likely to have hypertension and atrial fibrillation, and had more GCA and WMH than NCI subjects. Unlike the development dataset, subjects did not differ in terms of gender, infarcts, chronic lacunes, and intracranial stenosis (Table 3). The neuropsychological assessments corroborated cognitive classifications, with PSCI subjects performing consistently worse than NCI subjects (Supplementary Table 2).

In spite of baseline differences between the development and validation data, the SIGNAL₂ risk score was still relatively reliable in the validation cohort with an AUC of 0.7755 (95% CI 0.700–0.851). A cutoff of≥7 had an accuracy of 76.76% , sensitivity of 54.29% , specificity of 82.00% , PPV of 41.30% , and NPV of 88.49% . 9.91% of lower score tertile subjects, 30.00% of middle tertile subjects, and 75.00% of upper tertile subjects were PSCI.

89 of the subjects from the prospective cohort returned for a second assessment at the 12–18 month mark [mean interval time 16.35 months (SD 4.60 months)], 17 (19.10% ) of whom were classified PSCI at this point. As with the 3–6 month mark, neuropsychological assessment scores were also in keeping with PSCI status (Supplementary Table 2).

The SIGNAL₂ risk score performed similarly for PSCI at 12–18 months with an AUC of 0.783 (95% CI 0.667–0.899). A cutoff of≥7 was 76.40% accurate, with sensitivity 64.71% , specificity 79.17% , PPV 41.30% , and NPV 88.49% . 6.25% of subjects at the lower score tertile here were PSCI, compared to 31.58% at the middle tertile and 66.67% at the upper tertile.

DISCUSSION

We demonstrate that the 15-point SIGNAL₂ risk score, comprising clinical and neuroimaging variables can sufficiently identify patients at risk of developing PSCI 3–6 months post-stroke. PSCI was more prevalent in patients classified as having higher risk. The score was shown to perform comparably well in both development and internal validation.

The variables in the risk score have been previously demonstrated to be strongly associated with cognitive impairment among stroke patients. Older age has been associated with greater burden of vascular injury, neurodegenerative pathologies, and an odds ratio of 3.5–9.4 for PSCI [8]. Cortical infarcts being associated with increased risk of PSCI is consistent with earlier studies reporting higher risk of PSCI with infarcts in the middle cerebral artery territory [15]. Frontal cortical infarcts with lesions in the left ventrolateral region and right superior medial region have been shown to result in slow reaction time and a decreased number of correct responses to targets [28]. Our findings of the number of cortical infarcts significantly influencing the risk of PSCI is consistent with earlier reports [29]. Infarcts in the cortical region are more likely to result in disruption of frontal-subcortical circuits which play an important role in cognitive performance. WMH has been demonstrated to be a major risk factor for PSCI [12], and in deep locations has been demonstrated to result in impaired executive function, slowed processing speed, and visuospatial difficulties likely due to frontal-subcortical circuitry disruption [30]. Other recent studies have also demonstrated intracranial stenosis detected on MR angiography and CT angiography to be correlated with increased risk of progressive cognitive decline likely as a result of chronic cerebral hypoperfusion [31].

A risk score index for PSCI is important in providing an effective yet simple method to identify stroke patients who are at risk for PSCI. Identification of stroke patients who are at risk for PSCI would allow clinicians to triage patients who need more regular follow-up, so that appropriate management including acetylcholinesterase inhibitors and cognitive rehabilitation may be initiated at appropriate times. Aggressive optimization of vascular risk factors for this group of patients will also be important to reduce risk of recurrent strokes and worsening cognitive impairment. A PSCI risk score is also of value in research studies to identify high-risk stroke patients to target preventive strategies to reduce development of PSCI. It should also be recognized that the negative predictive value of the score was high (86.41% in the development cohort and 88.49% in the validation cohort) and thus the score may be useful in identifying stroke patients who are unlikely to develop cognitive impairment.

The strengths of this study include the use of separate development and validation datasets to develop the risk score. The confirmation of stroke diagnosis with MRI for all subjects and the use of neuropsychological tests to support the diagnosis of PSCI are also strengths. Also, the prospective design of the validation cohort allowed for more robust validation of the risk score. Despite the retrospective nature of the development dataset, the fact that the risk score works well for its intended use in both prospective and retrospective data demonstrates the strength and reliability of the scale, largely due to the already-known relevance that the scale components have in PSCI. The score was also shown to be useful in longer timeframes of 12–18 months, demonstrating that the scale takes into account concrete and stable factors leading to PSCI, as opposed to transient or secondary factors. However due to lack of biomarkers for Alzheimer’s disease (AD) or for cerebral perfusion, it was not possible to be certain about the etiology for cognitive impairment post-stroke. This may be related to both vascular disease and AD and future prospective biomarker supported studies will be required to clarify the usefulness of the score in predicting AD dementia or VaD following strokes.

Limitations include how the model is only applicable to patients who have undergone MRI and MRA at the time of acute stroke. It is the hope that MRI and MRA will be performed increasingly frequently in stroke centers worldwide, hence increasing the usefulness of this scale over time. Additionally modifications of the scale to be more applicable for cohorts who undergo CT and CT angiography should also be performed. As the population in this study was predominantly Asian, generalizability to non-Asian populations will need to be performed with caution. Another limitation is the lack of differentiation for VaD, since this analysis mainly focused on PSCI. The usefulness of this score in predicting both PSCI and VaD will need to be ascertained with additional longitudinal data, along with the effects of treatment of cognition or vascular risk factors on the risk score. Therefore, the risk score can be used to accurately identify stroke patients who are at an increased risk of developing PSCI within 3–6 months post-stroke. The availability of this tool would allow further studies to validate its usefulness in predicting PSCI at different time points from the acute stroke event.

Footnotes

ACKNOWLEDGMENTS

This research was supported by the National Neuroscience Institute, Singapore.

Authors’ disclosures available online ().

Appendix

References

Balami

, Chen

, Grunwald

, Buchan

(2011) Neurological complications of acute ischaemic stroke. Lancet Neurol 10, 357–371.

Gottesman

, Hillis

(2010) Predictors and assessment of cognitive dysfunction resulting from ischaemic stroke. Lancet Neurol 9, 895–905.

Lee

, Shafe

, Cowie

(2011) UK stroke incidence, mortality and cardiovascular risk management 1999-2008: Time-trend analysis from the General Practice Research Database. BMJ Open 1, e000269.

Melkas

, Oksala

, Jokinen

, Pohjasvaara

, Vataja

, Oksala

, Katse

, Karhunen

, Erkinjuntti

(2009) Poststroke dementia predicts poor survival in long-term follow-up: Influence of prestroke cognitive decline and previous stroke. J Neurol Neurosurg Psychiatr 80, 865–870.

Oksala

, Jokinen

, Melkas

, Oksala

, Pohjasvaara

, Hietanen

, Vataja

, Kaste

, Karhunen

, Erkinjuntti

(2009) Cognitive impairment predicts poststroke death in long-term follow-up. J Neurol Neurosurg Psychiatr 80, 1230–1235.

Sachdev

, Brodaty

, Valenzuela

, Lorentz

, Looi

, Berman

, Ross

, Wen

, Zagami

(2006) Clinical determinants of dementia and mild cognitive impairment following ischaemic stroke: The Sydney Stroke Study. Dement Geriatr Cogn Disord 21, 275–283.

Makin

, Turpin

, Dennis

, Wardlaw

(2013) Cognitive impairment after lacunar stroke: Systematic review and meta-analysis of incidence, prevalence and comparison with other stroke subtypes. J Neurol Neurosurg Psychiatr 84, 893–900.

Rasquin

, Lodder

, Ponds

, Winkens

, Jolles

, Verhey

(2004) Cognitive functioning after stroke: A one-year follow-up study. Dement Geriatr Cogn Disord 18, 138–144.

O’Brien

, Erkinjuntti

, Reisberg

, Roman

, Sawada

, Pantoni

, Bowler

, Ballard

, DeCarli

, Gorelick

, Rockwood

, Burns

, Gauthier

, DeKosky

(2003) Vascular cognitive impairment. Lancet Neurol 2, 89–98.

10.

Reitz

, Luchsinger

, Tang

, Manly

, Mayeux

(2006) Stroke and memory performance in elderly persons without dementia. Arch Neurol 63, 571–576.

11.

Jokinen

, Kalska

, Mantyla

, Ylikoski

, Hietanen

, Pohjasvaara

, Kaste

, Erkinjuntii

(2005) White matter hyperintensities as a predictor of neuropsychological deficits post-stroke. J Neurol Neurosurg Psychiatr 76, 1229–1233.

12.

Kandiah

, Wiryasaputra

, Narasimhalu

, Karandikar

, Marmin

, Chua

, Sitoh

(2011) Frontal subcortical ischemia is crucial for post stroke cognitive impairment. J Neurol Sci 309, 92–95.

13.

Pohjasvaara

, Mantyla

, Salonen

, Aronen

, Ylikoski

, Hietanen

, Kaste

, Erkinjuntti

(2000) How complex interactions of ischemic brain infarcts, white matter lesions, and atrophy relate to poststroke dementia. Arch Neurol 57, 1295–1300.

14.

Rasquin

, Verhey

, van Oostenbrugge

, Lousberg

, Lodder

(2004) Demographic and CT scan features related to cognitive impairment in the first year after stroke. J Neurol Neurosurg Psychiatr 75, 1562–1567.

15.

Pendlebury

, Rothwell

(2009) Prevalence, incidence, and factors associated with pre-stroke and post-stroke dementia: A systematic review and meta-analysis. Lancet Neurol 8, 1006–1018.

16.

Poggesi

, Salvadori

, Valenti

, Nannucci

, Ciolli

, Pescini

, Pasi

, Fierini

, Donnini

, Marini

, Chiti

, Rinnoci

, Inzitari

, Pantoni

(2014) The Florence VAS-COG clinic: A model for the care of patients with cognitive and behavioral disturbances consequent to cerebrovascular diseases. J Alzheimers Dis 42 Suppl 4 S453–S461.

17.

Godefroy

, Fickl

, Roussel

, Auribault

, Bugnicourt

, Lamy

, Canaple

, Petitnicolas

(2011) Is the Montreal Cognitive Assessment superior to the Mini-Mental State Examination to detect poststroke cognitive impairment? A study with neuropsychological evaluation. Stroke 42, 1712–1716.

18.

Pendlebury

, Mariz

, Bull

, Mehta

, Rothwell

(2012) MoCA, ACE-R, and MMSE versus the National Institute of Neurological Disorders and Stroke-Canadian Stroke Network Vascular Cognitive Impairment Harmonization Standards Neuropsychological Battery after TIA and stroke. Stroke 43, 464–469.

19.

Fazekas

, Chawluk

, Alavi

, Hurtig

, Zimmerman

(1987) MR signal abnormalities at 1.5 T in Alzheimer’s dementia and normal aging. Am J Roentgenol 149, 351–356.

20.

Gregoire

, Chaudhary

, Brown

, Yousry

, Kallis

, Jager

, Werring

(2009) The Microbleed Anatomical Rating Scale (MARS): Reliability of a tool to map brain microbleeds. Neurology 73, 1759–1766.

21.

Houkin

, Nakayama

, Kuroda

, Nonaka

, Shonai

, Yoshimoto

(2005) Novel magnetic resonance angiography stage grading for moyamoya disease. Cerebrovasc Dis 20, 347–354.

22.

Pasquier

, Leys

, Weerts

, Mourner-Vehier

, Barkhof

, Scheltens

(1996) Inter- and intraobserver reproducibility of cerebral atrophy assessment on MRI scans with hemispheric infarcts. Eur Neurol 36, 268–272.

23.

Dong

, Sharma

, Chan

, Venketasubramanian

, Teoh

, Seet

RCS

, Tanicala

, Chan

, Chen

(2010) The Montreal Cognitive Assessment (MoCA) is superior to the Mini-Mental State Examination (MMSE) for the detection of vascular cognitive impairment after acute stroke. J Neurol Sci 299, 15–18.

24.

Jacquin

, Binquet

, Rouaud

, Graule-Petot

, Daubail

, Osseby

, Bonithon-Kopp

, Giroud

, Bejot

(2014) Post-stroke cognitive impairment: High prevalence and determining factors in a cohort of mild stroke. J Alzheimer Dis 40, 1029–1038.

25.

Verdelho

, Madureira

, Ferro

, Basile

, Chabriat

, Erkinjuntti

, Fazekas

, Hennerici

, O’Brien

, Pantoni

, Salvadori

, Scheltens

, Visser

, Wahlund

, Waldemar

, Wallin

, Inzitari

(2007) Differential impact of cerebral white matter changes, diabetes, hypertension and stroke on cognitive performance among non-disabled elderly, The LADIS study. J Neurol Neurosurg Psychiatr 78, 1325–1330.

26.

Jorm

(1994) A short form of the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE): Development and cross-validation. Psycholol Med 24, 145–153.

27.

Lawton

, Brody

(1969) Assessment of older people: Self-maintaining and instrumental activities of daily living. Gerontologist 9, 179–186.

28.

Alexander

, Stuss

, Shallice

, Picton

, Gillingham

(2005) Impaired concentration due to frontal lobe damage from two distinct lesion sites. Neurology 65, 572–579.

29.

Saczynski

, Sigurdsson

, Jonsdottir

, Eiriksdottir

, Jonsson

, Garcia

, Kjartansson

, Lopez

, Van Buchem

, Gudnason

, Launer

(2009) Cerebral infarcts and cognitive performance: Importance of location and number of infarcts. Stroke 40, 677–682.

30.

Delano-Wood

, Abeles

, Sacco

, Wierenga

, Horne

, Bozoki

(2008) Regional white matter pathology in mild cognitive impairment: Differential influence of lesion type on neuropsychological functioning. Stroke 39, 794–799.

31.

Zhu

, Wang

, Li

, Deng

, Zhou

(2014) Intracranial artery stenosis and progression from mild cognitive impairment to Alzheimer disease. Neurology 82, 842–849.