SUBJECTS AND METHODS

Between June 1997 and November 1999, we examined 254 subjects aged 60 to 91 years old, living in the rural community of Sefuri village, Saga, Japan (the total population of the village numbered 1938 as of November 1999). All subjects were volunteers recruited from a local senior citizens' centre. The subjects comprised 69 men and 185 women, with a mean age of 73.9 (SD 6.8) years and a mean educational achievement of 8.5 (SD 1.8) years in this study. The participation rate in the study was 38% for those aged 60 years or older. Any persons with chronic intracranial illnesses including cerebrovascular disease, psychiatric disorders, or a history of head trauma were excluded. All subjects were literate, living independently without any apparent dementia, at home in this community. Cranial MRI was performed on all subjects, and a patient history, general haematology tests, and biochemistry tests were also performed to evaluate any risk factors for cerebrovascular diseases such as hypertension, diabetes mellitus, and hyperlipidaemia. Blood pressure was measured in the sitting position by the standard cuff method. For the neurological examination, the presence of a Barré sign, asymmetric deep tendon reflexes, and Babinski's sign were recorded on each side. Left ventricular hypertrophy, ST depression, or atrial fibrillation were registered on 12 lead ECG recordings. The cognitive function was assessed in a battery of neuropsychological tests—the mini mental state examination (MMSE).⁹ The mean MMSE score in this study was 26.0 (SD 3.3) points. A few subjects had lower scores in the MMSE test, and some of subjects labelled as “cognitive impairment” might have had mild dementia. However, they could still care for themselves, and no subject had any obvious dementia that fulfilled the DSM-IV diagnostic criteria.¹⁰ All subjects were stratified into two groups based on the severity of cognitive impairment, as defined by MMSE score less than 24.^11–15 Consequently, 46 (18.1%) subjects were defined as having a cognitive impairment.

Cranial MRI was performed on a 1.0 T superconducting magnet (MAGNEX a, Shimadzu, Kyoto, Japan) using the spin echo technique and fluid attenuated inversion recovery (FLAIR) sequences. Transverse T1 weighted (TR/TE 380/14 ms), T2 weighted (TR/TE 3750/110 ms), and FLAIR (TR/TI/TE 5800/1700/110 ms) images were obtained with a slice thickness of 8 mm separated by a 2 mm interscan gap. The MRI data were modified from 256×256 pixels to 512×512 pixels, and then were transferred from the MRI unit to a Macintosh computer. The area of white matter lesions and the brain volume was measured quantitatively on one slice using a computer assisted processing system (NIH Image version 1.61). White matter lesions were defined as high signal intensity areas on T2 weighted images but isointense with normal brain parenchyma on T1 weighted images. All measurements were performed on FLAIR images obtained at 2 slices above the level of the pineal body. The area of white matter lesions was quantitatively measured using a semiautomatic method to count the pixels with a given intensity. To correct for individual differences in head size on the measurement of white matter lesions (WMLs), they were calculated as a percentage of the total intracranial volume and defined as the %WMLs. Using the same slice level as that for the measurement of white matter lesions, the area of the cerebral parenchyma was quantified on T2 images and was divided by the area inside the skull to calculate the % brain value (%Brain), which was used as an index of cerebral atrophy. Cerebral infarcts were defined qualitatively as lesions with an abnormal signal in vascular distribution and no mass effects, and were defined as low signal intensity areas on the T1 weighted images and high signal intensity areas on the T2 weighted images and their sizes were 5 mm or larger. The presence or absence of infarcts and the number of infarcts in the whole brain were determined qualitatively.

All values are given as mean (SD). We used the SPSS software package (SPSS Advanced Statistics^TM version 6.1, SPSS Inc, Chicago, IL, USA) for the statistical analysis. A univariate analysis was performed with a t test for continuous variables and the χ² test for categorical variables. A linear regression was performed with Spearman's rank correlation. Multivariate analyses were done using a logistic regression model to identify any factors predictive of cognitive impairment on the entire sample of 254 subjects. We used the forward stepwise method of logistic analysis. We chose the variables for entry into the multivariate analysis based on the clinical and neuroradiological findings with values of p<0.05 considered significant after univariate testing.

RESULTS

The volume of %WMLs ranged from 0.37% to 29.7% of the intracranial area with a mean of 5.61 (4.44)%. Figure 1 shows examples of the quantitative measurement of %WMLs of axial FLAIR density MRI images. The volume of the %Brain ranged from 66.1% to 94.6% of the intracranial volume with a mean of 82.3 (6.4)%. Definite brain infarcts were seen in 54 subjects (21.3%) with a mean number of 1.65 (1.01) infarcts. The MRI and clinical findings were analyzed to detect any factors influencing the cognitive function of the subjects.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Example of quantitative measurement of % WMLs of axial FLAIR-density (TR/TI/TE 5800/1700/110 ms) MRI images. The percentages of patients were 56.7%, 33.5%, and 9.8% in each of the three categories represented in (A), (B), and (C), respectively. (A) An 80 year old subject with %WMLs=0.9% (corresponding to grade 0 of Fazekas's classification). (B) A 66 year old subject with %WMLs=7.8% (corresponding to grade 1 of Fazekas's classification). (C) An 81 year old subject with %WMLs=23.6% (corresponding to grades 2 and 3 of Fazekas's classification).

Table 1 shows the comparisons of the independent variables between the subjects with and without cognitive impairment as defined by the MMSE cognitive scale. Subjects with cognitive impairment, who scored below 24 points on the MMSE scale, were significantly older (78.0 years v 72.9 years, p<0.0001), less educated (7.4 years v 8.7 years, p<0.0001), more likely to have cerebral infarction on MRI (35% v 18%, p=0.013), a history of hypertension (52% v 33%, p=0.014), lower packed cell volume (0.36 v 0.39, p=0.0012), and higher systolic blood pressure (159 mm Hg v 145 mm Hg). The mean %WMLs volume for subjects with cognitive impairment was 8.25 (4.40)% compared with 5.03 (4.25)% for the subjects without cognitive impairment (p<0.0001). The mean %Brain volume for the subjects with cognitive impairment was 79.0 (5.8)% compared with 83.1 (6.3)% for the subjects without cognitive impairment (p<0.0001).

Table 2 shows the results of a multivariate analysis between cognitive impairment and the MRI and clinical findings. The logistic analysis included factors which were significant based on a univariate analysis as variables in the models. Within the multivariate models cognitive impairment was related to the %WMLs (odds ratio (OR) 1.575, 95% confidence interval (95%CI) 1.123–2.208), to the %Brain (OR 0.761, 95%CI 0.587–0.987), and to the educational level (OR 0.682, 95%CI 0.544–0.855). A statistical analysis was done for an increase in educational background by 1 year, as well as for the %WMLs and the %Brain by 5.0%.

A multivariate analysis of the %WMLs and the %Brain showed an influence on cognitive impairment, and a significant correlation was noted between these parameters and the cognitive function, as shown in figures 2 and 3.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Correlation between %WMLs (white matter lesions) and cognitive function (MMSE score). A linear regression was performed with Spearman's rank correlation and the regression line is shown. Coefficient of association=−0.297, p value<0.0001.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Correlation between %Brain and cognitive function (MMSE score). A linear regression was performed with Spearman's rank correlation and the regression line is shown. Coefficient of association=0.258, p value=0.0002.

DISCUSSION

In the present study, we used a quantitative MRI method to investigate the factors affecting the cognitive function of elderly people who were leading normal lives at home, and concluded that white matter lesions, cerebral atrophy, and educational background had an important influence on cognitive impairment. Several studies of the general population have previously focused on the prevalence of dementia and the factors affecting its onset.^{1–6, 16, 17} However, most of these reports tend to focus on particular lesions including WMLs and silent cerebral infarction, and thus study the frequencies and risk factors of such lesions, or relate such lesions to cognitive impairment. Investigating the MRI findings in relation to cognitive impairment in elderly subjects is considered to be important because such data can be used for both prophylaxis and early detection. But only a few reports in the past have attempted to discuss a wide range of factors which influence cognitive troubles in the general population.^{7, 8} Among these studies, Kuller et al⁷ performed a large scale resident study and reported that factors influencing cognitive impairment were cerebral atrophy, white matter lesions, and cerebral infarction on MRI, ApoE4 genotype, age, educational background, and race. They qualitatively assessed abnormalities on MRI images,¹⁸ whereas we analyzed the quantitative data. They also studied elderly subjects under various conditions,¹⁹ whereas we excluded subjects with intracranial lesions, cerebrovascular accidents, or dementia, and instead studied a relatively homogeneous group of healthy elderly subjects.

White matter lesions have been extensively investigated regarding cognitive impairment. Many studies^{1, 7, 20–24} have used qualitative methods for evaluating the size or severity of white matter lesions, but quantitative data have also been increasingly reported.^25–30 Because qualitative methods provide rather obscure criteria for evaluating the severity of white matter lesions, we employed a quantitative method. In our preparatory research using the qualitative method, the relation between WMLs and cognitive impairment has already been described.³¹ Regarding the role of white matter lesions in cognitive impairment, an association between such lesions and cognitive impairment in normal elderly subjects has been reported,^{24–27, 32} although the absence of any relation with cognitive function has also been reported.^{22, 23, 33–35} DeCarli et al²⁶ counted the number of pixels on the MR images to quantitatively define the volume of WMLs in their study of normal subjects. They evaluated the subjects by whole brain volume, and therefore, the %WMLs was smaller than in our results, and they defined 0.5% or less to be normal. Their results show the volume of WMLs to be significantly related to cognitive impairment, which reflects the level of frontal lobe function. Wahlund et al²⁸ quantitatively analyzed the WMLs of normal subjects and made a comparison with the MMSE results. They reported that the MMSE scores and volumes of WMLs were closely correlated. The above mentioned diversity in the results of research on the influence of WMLs on cognitive functions is partly due to the differences in the research methods and the contents of psychological tests, in addition to differences between subjects.

Our present results demonstrated no significant relation between brain infarction and cognitive impairment. The cerebral infarction shown on MRI was mostly silent and small in size in this study. Some reports^{13, 36–40} have already well described the relation between silent brain infarctions and cognitive impairment. Regarding the role of silent cerebral infarctions in cognitive impairment, no association between such lesions and cognitive impairment in normal elderly subjects has previously been reported,^{7, 41} although the absence of any relation with cognitive function has also been reported.^{8, 42, 43} In our previous study,⁴⁴ in which we used CT, of about 200 community dwelling elderly subjects, we found that problems of mental functions were influenced by aging, cerebral infarction, and WMLs. The absence of any association between cerebral infarction and cognitive impairment in the present study may be attributable to the following factors: (1) MRI detected even very small infarcts and (2) those with known cerebrovascular disease were excluded, so the subjects rarely had infarcts that affected critical regions.

A substantial degree of cerebral atrophy corresponds to the development of cognitive impairment, according to the findings of many studies on degenerative dementia.^45–48 Some other studies on normal subjects^{7, 49, 50} have made similar conclusions. Previous studies have also quantified the extent of cerebral atrophy.^{46–48, 51–55} The cerebral atrophy that occurs in degenerative dementia is thought to differ from that in normal elderly people both regarding its aetiology and its clinical significance. The present study showed that cerebral atrophy had a slight influence on cognitive impairment. However, further studies are required to both clarify the mechanism by which cerebral atrophy plays a part in cognitive impairment in independent elderly people, as well as to identify the factors responsible for cerebral atrophy.

Educational background is thought to affect the results of the MMSE,^{7, 56} which was used in this study. In fact, a strong influence of education was found in our study. The MMSE is probably the most widely used rating scale for a simple assessment of cognitive function in elderly subjects.⁹ A score of 23 or less is commonly used to reflect cognitive impairment when the issue is case detection.^11–15 However, the MMSE score dose not reflect the cognitive function alone. Uhlmann et al⁵⁷ have suggested that the cut off point of the MMSE should be adjusted according to the subject's educational background. On the other hand, some authors^{53, 55, 58} have stated that there is an inverse relation between educational background and cerebral atrophy which prevents cognitive impairment, although others have reported^59–61 that low educational background is related to the risk of Alzheimer's disease. The educational background is greatly dependent on the environment during childhood, and therefore some researchers also consider that the childhood environment may influence future cognitive impairment. For the prevention and early detection of dementia, we therefore should discuss how to improve the educational level of the general population.

In conclusion, the present cross sectional study showed cognitive impairment in non-demented elderly subjects to be related to the educational level, the presence of WMLs, and brain atrophy. In the future, longitudinal studies are expected to clarify the risk factors or early findings of cognitive decline that can potentially be used for the early detection and prevention of dementia.