Health-related quality of life (HRQoL) is used to quantify patients' burden of disease.1 This measure focuses on the patient rather than the disease and is helpful to compare health status between different types of disease. HRQoL also allows direct comparisons of benefits from a variety of interventions on a variety of diseases.1,2 Recently, the US Food and Drug Administration announced to the pharmaceutical industry that HRQoL could be a possible endpoint in clinical trials.3 Therefore, HRQoL has begun to be used as an outcome measure in clinical trials on various diseases, including stroke and dementia.4,5

HRQoL was worse in people with Alzheimer's disease (AD) compared to those with mild cognitive impairment or normal cognition, but there was no difference between mild cognitive impairment and normal cognition.6 HRQoL of stroke survivors was lower than that of healthy elderly people living in the community.7 However, HRQoL has not been evaluated and reported in stroke victims with cognitive impairment but no dementia.

The utility weight of EQ-5D (EQ-5D_index) in a representative sample of the Korean population was recently reported as a feasible and valid tool for measuring HRQoL in stroke survivors.8

The objectives of the present study were to compare patients with post-stroke cognitive impairment with no dementia (PSCIND) to those with post-stroke dementia (PSD) and healthy people with normal cognition with respect to HRQoL, measured by EQ-5D_index, and to investigate the individual effect of each cognitive domain on EQ-5D_index.

A series of 1,528 patients who visited the stroke prevention clinic at Seoul National University Bundang Hospital, underwent brain magnetic resonance imaging (MRI), and underwent the 60-min neuropsychological assessment protocol of the Korean version of Vascular Cognitive Impairment Harmonization Standards (K-VCIHS-NP) and EQ-5D between May 2007 and March 2009, were identified using the clinical data warehousing system embedded in our institution's electronic medical records system. This study protocol was reviewed and approved by your Institutional Review Board, and that informed consent was waived.

Of those 1,528, the PSD and PSCIND groups were selected from those who 1) had been hospitalized at our institution for acute stroke management and 2) were impaired in at least one of the frontal executive, language, visuospatial, or memory domains on the K-VCIHS-NP, and 3) in whom the K-VCIHS-NP might have been administered at least 90 days after stroke onset. Among them, those who did not meet the criteria of dementia were classified as the PSCIND group and those who met the criteria formed the PSD group. Dementia was diagnosed according to the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition9 criteria based on clinical interviews, cognitive functions assessed using the K-VCIHS-NP, and social functioning status assessed on the scale of Instrumental Activities of Daily Living. Patients who had depression were excluded. Stroke (ischemic or hemorrhagic) was defined using the World Health Organization's definition of "rapidly developed clinical signs of focal or global disturbance of cerebral function, lasting more than 24 hours or leading to death, with no apparent cause other than a vascular origin."

The control group was chosen from individuals who visited the stroke prevention clinic worrying about a risk of stroke, but who had no history of stroke and normal cognition on the 60-minutes protocol. Those who had moderate to severe white matter hyperintensities on the fluid-attenuated inversion recovery sequences of brain MRI, psychiatric illness, or other neurological disorders were excluded. We also excluded those who were not evaluated with EQ-5D. A flow chart for the identification of study subjects is displayed in Figure 1.

We selected cases and controls in the age band between 50 years (the lowest in PSD group) and 82 (the highest in control group). Each PSCIND patient (N=50) was matched to one PSD patient and one control according to age (±3 years) and sex, blinded to other clinical information.

Of 1,528 patients who visited the stroke prevention clinic and underwent brain MRI and the K-VCIHS-NP during the study period, 586 had a history of stroke and 942 had not (Figure 1). Among those 586 with stroke, we identified 70 PSD and 126 PSCIND patients, after excluding those who had depression or in whom the EQ-5D was not available. Among 942 without stroke, 222 were suitable for the control group. Lastly, 150 subjects, 50 for each group matched for age (±3 years) and sex, were enrolled in the study.

The mean age was 69 years and 72.0% of subjects were male. The clinical characteristics of the study subjects are presented in Table 1. The level of education was significantly different among the three groups. The PSD and PSCIND groups had higher systolic blood pressure, were more likely to be current smokers, and were less likely to have hyperlipidemia compared to the control group. Three-month mRS score was different between the PSD and PSCIND groups and significantly correlated with the EQ-5D_index at 3 months of stroke onset in the PSD group (-0.38, P=0.007), but not in the PSCIND group (-0.06, P=0.70).

Comparisons between the 3 groups were made with respect to the z-scores of the individual neuropsychological tests, adjusted for education (Table 2). The z-scores of all the cognitive tests of the PSD group were significantly lower than those of the PSCIND and control groups, and those of the PSCIND group were lower than those of the control group. The most frequently impaired cognitive domain was memory, and the most frequently impaired frontal executive function test was the Korean Trail Making Test B (K-TMT-B), in both the PSCIND and PSD groups. With respect to Instrumental Activities of Daily Living scores, the PSCIND group was not different from the control group, while the PSD group had significantly lower scores than the other groups. The frequency of pre-stroke cognitive decline evaluated using Korean-IQCODE (IQCODE ≥3.6) was 35% (137 of 389) among stroke survivors.

Correlations between EQ-5D_index and neuropsychological tests in the three groups are presented in Table 3, with and without adjustment for 3-month mRS in stroke survivors. In the PSCIND group, EQ-5D_index correlated only with visuospatial function before adjustment, but with none after adjustment. In the PSD group, EQ-5D_index correlated with Korean Mini-Mental Status Examination and visuospatial function with and without adjustment, and with Digit Symbol Coding-correct and language, with adjustment.

There was a significant difference in the mean EQ-5D_index among the three groups, and pairwise comparisons between the PSCIND and control groups and between the PSCIND and PSD groups showed similar results even after adjustment for education and depression (Figure 2, Table 4).

This study finds that the HRQoL of PSCIND is intermediate between that of PSD and normal cognition and that no particular cognitive domain is more influential on the HRQoL of PSCIND. To the best of our knowledge, this study is the first to directly compare PSD, PSCIND, and normal cognition with respect to HRQoL.

The frequency of cognitive impairment in post-stroke (66.4%) was higher than that among previous studies.19-21 A recent report from a multicenter prospective stroke cohort in Korea appeared to be similar with our study in frequency of cognitive impairment after stroke (62.6%) even though it was a bit of difference in study setting such as cutoff value of cognitive impairment (<10 percentile).11 The higher frequency of cognitive impairment in our study might be partially explained by discrepancy with pre-stroke cognitive dysfunction (7.2%) in recent study.11

A previous study reported that the mean EQ-5D_index was 0.60 in all types of dementia, which was similar to that in the PSD group of the present study, 0.61.17 In another study on AD, the mean EQ-5D_index score was 0.62 when rated by proxy, but 0.86 when self-rated.22 The HRQoL may be lower in those with PSD than in those with AD; this needs to be confirmed in another study. For determinants of HRQoL in dementia, there is a clear and consistent association with depression (the more severe the depression, the lower the HRQoL), but there have been inconsistent reports on other factors, such as age, gender, education, ethnicity, and activity limitation.23,24 Subjects who had a history of depression (previously diagnosed and treated) was excluded, despite our findings might be partially biased by the inclusion of healthier controls. With high frequency of depression in PSD evaluated by a geriatric depression scale, depression was included as confounder in final model, but the impact of PSCIND on HRQoL was not different from the unadjusted result in statistical significance. Low HRQoL in PSD could be explained by medical comorbidities such as diabetes or coronary heart disease, which are known to be related to HRQoL25,26 and more prevalent in PSD than in AD.

The mean EQ-5D_index of PSCIND (0.86) was analogous to that of chronic obstructive pulmonary disease (≒0.81) in Korea.27 The reported mean EQ-5D_index of minor stroke (mRS≤3) was between 0.6 and 0.7, which was similar to that of PSD and lower than that of PSCIND.28 A few previous studies on stroke survivors have shown no influence of vascular cognitive impairment on HRQoL.7,29 From a viewpoint of HRQoL after stroke, the contribution of mRS to HRQoL should be emphasized.7 We found that 3-month mRS was associated with 3-month EQ-5D_index in PSD, but the difference in EQ-5D_index between PSD and PSCIND was still significant after adjusting for 3-month mRS. A positive association between cognition and HRQoL has been reported among functionally independent stroke survivors.30 We cautiously suggest that cognitive impairment may further impact the HRQoL in stroke victims.

EQ-5D_index of PSCIND was significantly lower than that of normal cognition in this study. This result is different from previous ones in which there was no significant difference between patients with mild cognitive impairment and controls.4,31 Between the PSCIND and control group, a difference existed not only in cognition but also in stroke history, which had an impact on HRQoL in Koreans,32 and both of them might lead to a difference in EQ-5D_index. A further study comparing those with PSCIND and stroke survivors with normal cognition can address the question of whether cognitive impairment that is not enough to be diagnosed as dementia, truly impacts HRQoL in stroke survivors.

EQ-5D_index in the PSCIND group did not correlate with impairment in any specific cognitive domain. A relationship between poor TMT-B performance and poor perceived quality of life has been reported in stroke patients.33 This study showed that impairment in the K-TMT-B, one of the executive function tests, was observed most frequently in both PCIND and PSD groups, but the K-TMT-B scores did not correlate with EQ-5D_index. However, it should be noted that 48% in the PSD group and 20% in the PSCIND group were unable to complete the K-TMT-B; this was the worst outcome among all cognitive tests.

This study has some limitations. First, the eligible stroke patients were not representative of the general stroke population in Korea, or even hospitalized stroke patients in our institution. Inevitably, patients who were unable to complete neuropsychological tests, did not undergo MRI, and were not followed up at the outpatient clinic after discharge were excluded from this study. Second, as mentioned above, this study did not include stroke survivors with normal cognition. A further study including all cognitive spectrums of stroke survivors is therefore warranted.