Article Text

Download PDFPDF

Cognitive impairment predicts poststroke death in long-term follow-up
  1. N K J Oksala1,2,
  2. H Jokinen3,4,
  3. S Melkas3,
  4. A Oksala2,
  5. T Pohjasvaara3,
  6. M Hietanen3,4,
  7. R Vataja3,
  8. M Kaste3,
  9. P J Karhunen2,
  10. T Erkinjuntti3
  1. 1
    Division of Vascular Surgery, Department of Surgery, Tampere University Hospital, Tampere, Finland
  2. 2
    School of Medicine, Forensic Medicine, University of Tampere and the Laboratory Centre Research Unit, Tampere University Hospital, Tampere, Finland
  3. 3
    Department of Neurology, Helsinki University Central Hospital, Helsinki, Finland
  4. 4
    Department of Psychology, University of Helsinki, Helsinki, Finland
  1. Correspondence to Dr N Oksala, Department of Surgery and Forensic Medicine, Medical School, University of Tampere and Tampere University Hospital, 33014 University of Tampere, Finland; niku.oksala{at}


Background: Poststroke global cognitive decline and dementia have been related to poor long-term survival. Whether deficits in specific cognitive domains are associated with long-term survival in patients with ischaemic stroke is not known in detail.

Methods: Patients with acute stroke subjected to comprehensive neuropsychological evaluation were included in the study (n = 409) and followed up for up to 12 years.

Results: In Kaplan–Meier analysis, impairments in following cognitive domains predicted poor poststroke survival (estimated years): executive functions (48.2%) (5.8 vs 10.1 years, p<0.0001), memory (59.9%) (6.8 vs 9.3 years, p = 0.009), language (28.9%) (5.3 vs 8.6 years, p = 0.004) and visuospatial/constructional abilities (55.2%) (5.6 vs 10.1 years, p<0.0001). Low Mini Mental Status Examination (MMSE) ⩽25 (30.5%) (4.4 vs 9.3 years, p<0.0001), low education (<6 years) (31.8%) (6.4 vs 8.2 years, p = 0.003) and poor modified Rankin score (39.9%) (3.9 vs 9.7 years, p<0.0001) were also related to poor survival. In Cox regression proportional hazards analyses including age, sex and years of education as covariates, deficits in executive functions (hazard ratio (HR) 1.59, p<0.0001), memory (HR 1.31, p = 0.042), language (HR 1.33, p = 0.036) and visuospatial/constructional abilities (HR 1.82, p<0.0001) were significant predictors of poor poststroke survival. Of these, executive functions (HR 1.33, p = 0.040) as well as visuospatial/constructional abilities (HR 1.53, p = 0.004) remained as significant predictors after addition of MMSE⩽25 and poor modified Rankin score as covariates. Furthermore, cognitive impairment no dementia (CIND) was also an independent predictor of poor poststroke survival (HR 1.63, p = 0.0123).

Conclusions: In patients with ischaemic stroke, cognitive impairment, particularly in executive functions, and visuospatial/constructional abilities relate to poor survival.

Statistics from

Poststroke cognitive impairment is frequent—up to 65–78% of patients demonstrate cognitive decline1 2 and 25–31% of patients demonstrate poststroke dementia.3 4 5 6 7 Further ischaemic stroke is an important factor predicting long-term survival.8

Global measures of cognition, for example Mini Mental Status Examination (MMSE)9 10 11 12 and similar short mental status tests,13 have been related to a poor survival in poststroke follow-up studies. Similarly poststroke dementia has been related to a poor survival.4 5 6 7 14

Poststroke cognitive impairment consists of decline in various cognitive domains such as memory, language, visuospatial and constructional abilities1 15 and also in executive function.16 17 However, the role of these specific cognitive impairments on poststroke long-term survival has not been studied in detail. Moreover, it is not known whether a milder cognitive impairment short of dementia is relevant for poststroke survival.

The aim of our study was to detail the effect of different cognitive deficits as well as cognitive impairment no dementia (CIND) on poststroke survival controlling for main covariates in a large well-defined ischaemic stroke cohort followed up to 12 years. We hypothesised that cognitive impairment would be related to impaired poststroke survival.



The Helsinki Stroke Aging Memory (SAM) cohort comprised a consecutive series of all Finnish (caucasian) patients with suspected stroke admitted to Helsinki University Central Hospital (n = 1622) between 1 December 1993 and 30 March 1995, as described in detail previously.18 19 20 Patients aged 55–85 years living in Helsinki were included. Patients without ischaemic stroke (n = 175), presenting with intracerebral (n = 229) or subarachnoid haemorrhage (n = 69), were excluded. A total of 642 patients fulfilled the inclusion criteria and were invited to a baseline visit 3 months later. Of these, 71 died (11.1%) before the 3-month follow-up, 82 refused (12.8%), and three were lost (0.5%) due to undefined causes. Finally, 486 (85.1%; 246 men, 240 women) of the living patients were included in the final cohort.21 Of these, 409 (84.2%) participated in neuropsychological examination. Patients with severe aphasia and thus unable to participate neuropsychological examination were excluded together with those with severe sight or hearing disabilities, or reduced level of consciousness.

The study was approved by the Ethics committee of the Department of Clinical Neurosciences, Helsinki University Central Hospital, Finland. The study was explained to the patients, and informed consent was obtained.

Clinical evaluation

A detailed medical and neurological history was taken.19 In the basic cohort of 486 patients, there were 388 (79.8%) patients with first-ever and 98 patients (20.2%) with recurrent stroke. Exclusion of those with recurrent stroke did not have any effect on study results, and so the final study was conducted including patients with recurrent stroke. Stroke severity was assessed using the modified Rankin score at 3 months.22

Neuropsychological analyses

A comprehensive neuropsychological assessment was administered 3 months after the index stroke as detailed previously.23 24 25 There were no statistically significant dfferences in gender, age, education, MMSE or modified Rankin score between these and the original 486 subjects.

In brief, global cognitive function was measured with the Mini-Mental State Examination.26 Score ⩽25 indicated cognitive impairment. Executive functions including attention were evaluated with the Trail making test,27 the Stroop colour naming test28 and the Digit span subtest of the Wechsler Memory Scale,29 the modified Wisconsin Card Sorting Test and the verbal fluency test.30 Data were available in 365 and missing in 44 subjects (10.8%). The assessment of memory functions included the Logical memory and Visual reproduction subtests of the Wechsler Memory Scale—revised31 as well as the Fuld Object Memory Evaluation.32 Data were available in 379 and were missing in 30 subjects (7.3%). Language was evaluated with the short version of the Token test33 the Boston naming test34 and the verbal fluency test.30 Moreover, the overall evaluation of speech functions was based on the Boston Diagnostic Aphasia Examination.35 Data were available in 395 and missing in 14 subjects (3.4%). Visuospatial and constructional abilities were assessed with the Block design subtest of the Wechsler Adult Intelligence Scale—revised36 and the clock test (recognising and setting time), and by copying a triangle, a flag, a three-dimensional cube and a Greek cross.37 Data were available in 391 and missing in 18 subjects (4.4%). Abnormality (impaired vs not impaired) in each domain was judged with the use of normative data from a random healthy Finnish population (2 SD or, if more than one test was used, 1 SD below the level of the norm on several tests indicated abnormality).38 The normal values were evaluated in different age groups.

Cognitive impairment without dementia, that is cognitive impairment no dementia (CIND),39 was defined as cognitive impairment in any of the above-mentioned domains after the exclusion of patients with dementia. Of the patients with MMSE⩽25 (n = 138), 94/138 (68.1%) had dementia at the 3-month follow-up. Further, of these, 62/94 (66.0%) had had a prestroke cognitive decline. Assessment of prestroke cognitive decline involved interviews with the patient and a knowledgeable informant, using structured questions. Dementia was defined according to Diagnostic and Statistical Manual of Mental Disorders 3rd edn criteria. From the basic cohort of 486 patients, we first excluded those who had not been tested for dementia (n = 35, 7.2%) and those who had dementia (n = 115, 23.7%). Further we investigated whether patients had impairment in any of the assessed domains. Fifty-two patients were excluded at this point because of missing values in the assessed domains, at the same time as none of the existing values indicated impairment. Finally, there were 212 patients with CIND and 72 patients with no cognitive impairment.

Survival data and causes of death

Long-term survival data and causes of death at 21 September 2006 were obtained from Statistics Finland. The mean (SD) follow-up time was 7.5 (4.0) years with a range between 0.3 and 12.8 years.

Statistical analysis

SPSS/WIN (version 12.0, SPSS, Chicago) software was used. The χ2 test was used to test dichotomous variables and Student t test to test continuous variables. The effect of different neuropsychological domains (executive functions, memory, language and visuospatial/constructional abilities), CIND and copredictors such as global cognitive function, that is MMSE (MMSE⩽25), education (low education <6 years), and poor modified Rankin score (3–5 vs 0–2) on long-term survival were first analysed using Kaplan–Meier log rank analysis. The median estimated years of survival are presented. The cumulative hazard function was also used, and according to these analyses, the proportional hazards assumption was met for each parameter included in further models. In Cox-regression proportional hazards survival analysis, in the forced entry model (Model 1) only the demographic background variables were used as covariates (age, sex, years of education). Another model (Model 2) was constructed adding MMSE⩽25 and poor modified Rankin score (3–5 vs 0–2) as clinical covariates. Statistical significance was set at p<0.05.


Characteristics of the study sample

The basic demographics of the cohort are as follows: age 71.2 (7.7) years; female sex 49.4%; MMSE⩽25 28.6%; poor modified Rankin score 39.9%.

The demographics with respect to presence of deficits in cognitive domains are presented in table 1. Patients with executive, language or visuospatial/constructional deficit (p<0.0001) were older than those without it. Women were overpresented in patients with language (p = 0.009) or visuospatial/constructional deficits (p = 0.004). Low education was more prevalent in patients with deficits in executive functions (p = 0.019), memory (p = 0.023), language (p = 0.017) and visuospatial/constructional abilities (p = 0.002). Poor modified Rankin score was significantly more prevalent in those with any of the cognitive deficits (p<0.0001) (table 1). MMSE⩽25 was frequent in all patients with deficits in any of cognitive domains (p<0.0001) (table 1).

Table 1

Association of cognitive deficits with demographic and clinical parameters (Stroke Aging Memory cohort)

The cognitive deficits were highly inter-related: 77.3% of patients with executive deficit had also a deficit in visuospatial/constructional abilities, and 73.9% had a memory deficit. On the other hand, the least frequent combination was executive and language deficit (30.1%).

Effect of cognitive impairment on survival

In Kaplan–Meier logrank analysis, deficits in executive functions (n = 176, 48.2%) (5.8, CI 4.7 to 7.0 vs 10.1, CI 8.7 to 11.4 years, p<0.0001), memory (n = 227, 59.9%) (6.8, CI 5.3 to 8.3 vs 9.3, CI 7.9 to 10.7 years, p = 0.009), language (n = 114, 28.9%) (5.3, CI 3.8 to 6.9 vs 8.6, CI 7.7 to 9.4 years, p = 0.004) and visuospatial/constructional abilities (n = 216, 55.2%) (5.6, CI 4.6 to 6.7 vs 10.1, CI 8.5 to 11.7 years, p<0.0001) predicted poor survival (fig 1).

Figure 1

Effect of different cognitive deficits (groups with given deficits are marked with a + or − sign), Mini-Mental State Examination (MMSE) and cognitive impairment not dementia (CIND) on overall poststroke survival (endpoint: all cause death) in Stroke Aging Memory cohort. Kaplan–Meier log rank analysis.

Furthermore, low education (<6 years) (n = 151, 31.8%) (6.4, CI 5.4 to 7.3 vs 8.2, CI 7.1 to 9.1 years, p = 0.003), MMSE ⩽25 (n = 138, 30.5%) (4.4, CI 3.7 to 5.1 vs 9.3, CI 8.3 to 10.4 years, p<0.0001) and poor modified Rankin score (n = 194, 39.9%) (3.9, CI 4.2 to 5.6 vs 9.7, CI 8.5 to 10.9 years, p<0.0001) predicted poor poststroke survival (fig 1).

To account for potential confounders, and to estimate the predictive value of different factors, a multivariate Cox regression proportional hazards analysis utilising two models (table 2) was used. In Cox proportional hazards regression analysis (Model 1, including age, sex and years of education), deficits in executive functions (hazard ratio (HR) 1.65, CI 1.27 to 2.13, p<0.001), memory (HR 1.31, CI 1.01 to 1.70, p = 0.042), language (HR 1.33, CI 1.02 to 1.73, p = 0.036) and visuospatial/constructional abilities (HR 1.82, CI 1.40 to 2.36, p<0.0001) were significant predictors of poor poststroke survival (table 2). Deficits in executive functions (HR 1.38, CI 1.05 to 1.81, p = 0.021) and visuospatial/constructional abilities (HR 1.53, CI 1.14 to 2.04, p = 0.004) remained as significant predictors after addition of MMSE⩽25 and poor modified Rankin score as covariates (Model 2) (table 2). MMSE⩽25 (HR 1.32 to 1.46) and poor modified Rankin score (HR 1.54 to 1.61) predicted poor survival in all the models.

Table 2

Cox regression analysis on the association of multiple risk factors and cognitive deficits with poor long-term survival (all cause death endpoint) among patients with ischaemic stroke (Stroke Aging Memory cohort)

To further explore the role of cognitive impairment, we analysed the effect of cognitive impairment no dementia (CIND) on poststroke survival. In this population (n = 284), CIND (n = 212, 74.6%) (7.9, CI 7.4 to 8.5 vs 9.7, CI 8.8 to 10.6, p = 0.003) was a predictor of poor poststroke survival when compared with patients with no cognitive impairment in any of the assessed domains. In Cox proportional hazard regression analysis with age, sex and years of education as covariates, CIND remained an independent predictor of survival (HR 1.63, CI 1.12 to 2.39, p = 0.012).

Association of neuropsychological domains and causes of death

Of all the deaths (n = 347), 30.5% were attributable to cardiac causes, 33.7% due to brain-related causes (bleeding, infarct, dementia), 14.0% due to cancer, 4.3% due to infectious causes, 3.9% due to traumatic causes and 12.5% due to other causes.

According to Kaplan–Meier log rank test, the probability of dying due to cardiac causes alone showed no association with deficits in any of the cognitive domains. Death due to brain-related causes on the other hand was more likely in those with deficits in executive functions (p = 0.019), language (p = 0.001) and visuospatial/constructional abilities (p<0.0001). No associations between cognitive domains and infectious, traumatic or other causes of death were found.


The present study showed that poor poststroke survival relates to deficits in cognitive functions, that is executive, memory, visuospatial/constructional functions and language. Global cognitive impairment was also related to poor survival. Deficits in executive and visuospatial/constructional functions remained as independent predictors after controlling for global cognitive impairment and severity of stroke. Furthermore, cognitive impairment no dementia (CIND) was indepdendently related to poor poststroke survival. These findings on CIND are in line with previous studies.39 40 41

In the present study, proportion of patients with deficits in different cognitive domains varied between 29 and 60%, the lowest proportion found in language and the highest in memory. Studies on the effect of cognitive domains on poststroke survival are limited. Deficits in domains like verbal, visuomotor and memory performance and reaction time have previously been associated with shortened survival at general population level.42 43

Prestroke,6 7 any poststroke4 5 6 and new-onset poststroke7 dementia has been consistently related to poor survival. This is in line with our finding of the association of global MMSE with survival. In a population-based study 3 months after stroke,9 the proportion of cognitively impaired subjects (MMSE<25) was found to be 38%, which is considerably higher than in the present study of hospitalised patients. However, similar prevalence was detected in a study of hospitalised patients.11 We consider our cohort representative of hospitalised patients with stroke. Our cohort with up to 12 years’ follow-up showed a significant effect of low MMSE on long-term survival. In line with our results, cognitive impairment as measured by MMSE was associated with impaired survival up to 4 years’ follow-up.9 10 11 12 However, two short 1-year studies demonstrate that MMSE is not an independent predictor of survival.44 45 A single study utilising a short 10-item mental status questionnaire showed that cognitive impairment is an independent predictor of 6-year survival.46

In a previous study,47 severe aphasia was independently associated with poor survival, which is in line with our finding of the association of deficit in language domain and survival. In earlier studies, CIND has been related to impaired survival in studies of community-based populations where stroke has not been regarded.39 40 41 Our results of CIND, a potentially reversible condition, as an independent predictor of poor poststroke survival in a specified poststroke population is of clinical significance. Recognition of CIND might alert the clinicians to schedule more strict and accurate poststroke follow-up.

As far as we know, our study is thus far the largest poststroke cohort with a long follow-up and a detailed neuropsychological examination. It is also the only study controlling for main covariates including severity of stroke. The present study showed that impairment in major cognitive domains relates to poor poststroke survival. When controlling for the effect of global cognitive decline (MMSE⩽25) and severity of stroke, only executive functions as well as visuospatial and constructional abilities remained as independent correlates of poor survival. Executive functions, including planning, initiation, sequencing and monitoring of complex goal-directed behaviour, rely heavily on the integrity of the prefrontal cortices and their connecting pathways with the subcortical structures.48 The frontal-lobe functional system is also involved with visuospatial and constructional abilities which require intact attention and organisational skills together with the more posteriorly mediated visuoperceptual functions.

Cerebral small-vessel disease is reflected by white-matter lesions seen in MRI,49 50 which have also been related to recurrent stroke51 52 53 54 55 and poor survival mostly due to cardiovascular causes.51 53 56 White-matter lesions on the other hand associate with general cognitive function and with global functional decline50 and executive dysfunction related to the frontal-lobe system.50 57 58 59 60 We have previously demonstrated the association of white-matter lesions and cognitive deficits in the present cohort.61 In addition, recently, we have demonstrated in the present cohort that of the patients, 28% had mild, 18% had moderate and 54% had severe white-matter lesions, and that severe lesions predicted poor poststroke survival.62 Together, these results suggest that there may be a complex interplay of cerebral small-vessel pathology and a related disruption of the frontal lobe functional system behind impaired poststroke survival.

A potential weakness of our study is the possibility of selection bias, as the cohort was formed 3 months after the index stroke. This may limit generalisation of the results. Therefore, we retrospectively obtained additional data on stroke-related deaths in Helsinki University Hospital district during the collection of the cohort from an independent organisation (Statistics Finland). In these retrospective data, up to 64% of stroke-related deaths occurred in women. While the proportions of both sexes in the present study were equal, this suggests that some women may have died before hospital assessment at 3 months. Due to exclusion of patients, the true survival rate may be underestimated. In addition, since patients with severe aphasia could not be assessed with comprehensive neuropsychological examination, there is a possibility for survival bias, since these patients are prone to impaired survival resulting in underestimation of the effect of language. The strengths of our study are that the cohort is a consecutive one, that the neuropsychological and clinical characteristics of the patients were strictly evaluated, and also that the severity of stroke at admission was quantitated according to the modified Rankin score. Also, the survival data are comprehensive with negligible amount of unresolved deaths.

In summary, cognitive impairment already at less severe stages without dementia is related to poor long-term survival poststroke. Deficits in executive functions and visuospatial and constructional abilities, in particular, predict poor outcome independently of global cognitive decline and severity of stroke.


We thank E Oksala and Statistics Finland for their expert assistance in the data collection.



  • Funding This study was supported by grants from the Maire Taponen Foundation; the Paavo Nurmi Foundation; The Finnish Angiologic Association; the Medical Council of the Academy of Finland (Helsinki); the Clinical Research Institute, Helsinki University Central Hospital; the Yrjö Jahnsson Foundation (Helsinki); the Finnish Cultural Foundation and the Elli and Elvi Oksanen Fund of the Pirkanmaa Fund under the auspices of the Finnish Cultural Foundation (Tampere); the Medical Research Fund of Tampere University Hospital; the Finnish Medical Foundation; and the Finnish Foundation for Cardiovascular Research (Helsinki).

  • Competing interests None.

  • Ethics approval Ethics approval was provided by the Department of Clinical Neurosciences, Helsinki University Central Hospital, Finland.

  • Patient consent Obtained.

  • Provenance and Peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.