We targeted all residents in the town of Tajiri, Japan, aged 65 years or older (N = 3207). Of these, 1654 (51.6%) participated in the study (participant group 1). From among this population, 564 adults were randomly selected to receive magnetic resonance imaging (MRI). Finally, 497 participants (88.1%) (Participant Group 2) agreed to undergo MRI.
For male and female participants, the prevalences of dementia were 3.5% and 0.9%, respectively, in the group aged 65 to 69 years; 5.0% and 1.9%, respectively, aged 70 to 74 years; 3.9% and 11.0%, respectively, aged 75 to 79 years; 21.9% and 14.7%, respectively, aged 80 to 84 years; and 33.9% and 39.4%, respectively, aged 85 years and older. Overall, 9.2% of male participants and 8.0% of female participants received a diagnosis of dementia.
The most common dementing disease in Tajiri, Japan, was possible Alzheimer disease (AD) with cerebrovascular disease (CVD) (by means of the National Institute of Neurological Disorders and Stroke and the Association Internationale pour la Recherche et l'Enseignement en Neurosciences criteria), followed by probable AD (by means of National Institute of Neurological and Communicative Disorders and Stroke and Alzheimer's Disease and Related Disorders Association criteria). It seemed that vascular dementia (VaD) tends to be overdiagnosed if possible AD with CVD is thought to be VaD.
Meguro K, Ishii H, Yamaguchi S, Ishizaki J, Shimada M, Sato M, Hashimoto R, Shimada Y, Meguro M, Yamadori A, Sekita Y. Prevalence of Dementia and Dementing Diseases in JapanThe Tajiri Project. Arch Neurol. 2002;59(7):1109-1114. doi:10.1001/archneur.59.7.1109
Vascular dementia (VaD) has been considered to be more prevalent than Alzheimer disease in Japan. However, this might be the result of overdiagnosis stemming from some problematic diagnosis of VaD or of the frequent use of magnetic resonance imaging to detect cerebrovascular disease in older adults.
We investigated the prevalence of dementia and the ratios of dementing diseases. The effects of different criteria for VaD (DSM-IV, Alzheimer's Disease Diagnostic and Treatment Centers [ADDTC], and National Institute of Neurological Disorders and Stroke and the Association Internationale pour la Recherche et l'Enseignement en Neurosciences [NINDS-AIREN]) were considered. Hippocampal atrophy and vascular contribution to dementia were evaluated using magnetic resonance imaging findings.
We targeted all residents 65 years and older (n = 3207) in Tajiri, Japan, and examined 1654 (participant group 1). Of these, 564 (participant group 2) were randomly selected, and 497 underwent magnetic resonance imaging and diagnosis of dementing diseases.
We found the overall prevalence of dementia to be 8.5% (141/1654) in participant group 1. Of these, 21 (14.9%) had a history of stroke. Of the 113 participants who had a history of stroke independent of dementia, 18.6% (21/113) were demented. For participant group 2 (n = 497), 32 were demented. The ratio among the dementia for probable VaD based on the NINDS-AIREN criteria was 18.8% (6/32), whereas that for ischemic vascular dementia was 31.3% (10/32) according to the ADDTC criteria.
We confirmed the overall prevalence of dementia in adults 65 years and older to be 8.5%. We found that VaD was not a common disorder according to the NINDS-AIREN criteria. Rather, the condition of possible Alzheimer disease with cerebrovascular disease was more common.
FOR COUNTRIES with large populations of older adults, knowledge of the prevalence of dementia is needed for health policy planning. Most studies reported the prevalence to be greater with older age.1,2
Vascular dementia (VaD), rather than Alzheimer disease (AD), has been believed to be the most common dementing condition in Japan.3 However, the diagnostic criteria for VaD are problematic. The DSM-IV4 requires "laboratory evidence indicative of cerebrovascular diseases that are judged to be etiologically related to the disturbance" for a diagnosis of VaD. Confirmation of etiologically related cerebrovascular disease (CVD) requires sophisticated neurological knowledge.
Also, the higher prevalence of suspected "vascular lesions," which are frequently detected as high-signal intensity on T2-weighted magnetic resonance imaging (MRI) findings,5 might result in overdiagnosis. Because of the increasingly common practice of performing MRI, some patients might have received a diagnosis of VaD simply because they were demented and had CVD.
Other criteria were developed in an attempt to overcome this problem. The Alzheimer's Disease Diagnostic and Treatment Centers (ADDTC) criteria6 require evidence of 2 or more strokes or a single stroke with a clear temporal relationship to the onset of dementia. The criteria of the National Institute of Neurological Disorders and Stroke and the Association Internationale pour la Recherche et l'Enseignement en Neurosciences (NINDS-AIREN)7 state that a causal relationship between CVD and dementia is required for a diagnosis of VaD and propose that the conditions under which patients with dementia and concomitant CVD may be considered to have possible AD with CVD. We suspect that many cases previously overdiagnosed as VaD in Japan might better have been diagnosed as possible AD with CVD.
Pohjasvaara et al8 found that the different criteria for VaD were not interchangeable. Major discriminating factors between the criteria were the requirements of focal neurological signs, unequal distribution of cortical dysfunctions, and evidence of relevant CVD based on neuroimaging findings.
Since we suspected that the previously reported high prevalence of VaD in Japan was due to overdiagnosis, we sought to confirm this low interchangeability of the criteria by means of a community-based study. Therefore, we examined MRI findings in older adults in a community for the prevalence of dementia and the ratios of dementing diseases, with a focus on VaD.
Based on the Tajiri Project, a community-based study on stroke, dementia, and bed-confinement prevention in Tajiri, a typical agricultural area in northern Japan,2,9- 11 we targeted all residents 65 years and older (N = 3207). We did not use a screening test design; the survey included Clinical Dementia Rating (CDR) assessments,12 dementia diagnosis, and results of a neuropsychological examination (not reported herein). The survey requested agreement for all components.
We included 1654 participants (51.6%) in the study. The participants/population for each age group were 404/1066 (65-69 years), 560/856 (70-74 years), 348/607 (75-79 years), 212/397 (80-84 years), and 130/281 (≥85 years). The reasons for refusal to participate were mainly psychological (25.6%) and physical (14.6%). Each age group satisfied the statistically sufficient number for the confidence interval to be 95%, provided that the prevalence of dementia was 10%. Home interviews were performed for 87 participants. Fifteen participants were reported by families to be confined to bed; 8 of these participants, however, were able to maintain a sitting position without support. A medical history of stroke was present in 113 (6.8%) of the 1654 participants.
From among the population of 1654, 564 adults were randomly selected to undergo MRI. The cost of all MRIs was officially paid for by the town government (about $300 per person). Finally, 497 participants agreed to undergo MRI. Figure 1 illustrates the study design. Written informed consent was obtained from all healthy participants and from the family members of patients with dementia.
Using MRI (1.5 T; SIERRA, GE-YMS, Japan), we evaluated hippocampal atrophy and CVD lesions, since both are important for the diagnosis of dementia.
For hippocampal assessment, the T1-weighted (repetition time/echo times, 400/14 ms) plane along the long axis of the hippocampus showing the measurement point described by Jobst et al13 was selected.14 Using the semiaxial plane, we measured the minimum width of the hippocampus13 and divided by the brain width on the same plane (defined here as the hippocampal width [HippW]).14 The HippW (given as a percentage) was measured by 2 board-certified neurologists independent of this study. Each neurologist made 2 assessments. The interreader and intrareader reproducibilities were calculated as follows:
For the interreader reproducibility, HippW1 and HippW2 were assessed by 2 neurologists. For the intrareader reproducibility, HippW1 and HippW2 were assessed by the same neurologist. Both results were better than 95%.
For the CVD evaluation, the combined axial T1-weighted and T2-weighted (repetition time/echo time, 3000/90 ms) images were used. Lesions were considered to be CVD when they showed low intensity on the T1-weighted MRI and high intensity on the T2-weighted MRI at the same location. We operationally classified CVD into the following 3 categories: (1) those smaller than 4 mm in size (defined here as état criblè), (2) those 4 to 8 mm (small CVD), and (3) those larger than 8 mm (large CVD).15 We counted the CVD numbers and assessed the moderate and large CVD distributions. After independent assessment by the 2 neurologists, the final evaluation was made by a third senior neurologist (K.M.).
We calculated the prevalence of dementia for both sexes in each age group. For the effect of educational level, participants were classified into the following 3 groups according to the old Japanese system: 6, 8, and 10 or more years of schooling. Dementia was diagnosed by means of the DSM-IV criteria. Severity of CDR was decided by a clinical team consisting of a psychiatrist, neurologists, and public health nurses. They assessed the subjects' mental status by asking them about their daily lives and family situations. The patients with dementia exhibited a CDR rating of 1 or greater. Using the 113 participants who had a history of stroke, the prevalence of poststroke dementia was also calculated. For MRI measures, we evaluated the HippW of each CDR and age group and the prevalence of CVD.
For the patients with dementia (diagnosed by means of the DSM-IV criteria), we analyzed the dementing diseases using the following criteria: (1) DSM-IV for dementia of the Alzheimer type (DAT), (2) DSM-IV for VaD, (3) National Institute of Neurological and Communicative Disorders and Stroke and Alzheimer's Disease and Related Disorders Association16 for probable AD, (4) NINDS-AIREN for possible AD with CVD, (5) NINDS-AIREN for probable VaD, and (6) ADDTC for probable ischemic vascular dementia (IVD).
Based on a previous study,8 the presence of focal neurological signs for the diagnosis of VaD included at least 1 of the following: hemianopia, lower facial weakness, dysarthria, motor or sensory hemisyndrome, hemiplegic gait, or a positive Babinski sign.
Since the poor interchangeability of the criteria for VaD was reported,8 we used only the probable criteria. Patients received a diagnosis of possible AD with CVD by means of the NINDS-AIREN criteria, provided that the vascular effect on dementia was considered to be too ambiguous to diagnose as VaD.
The Hachinski Ischemic Scale (HIS)17 was also used to evaluate the possible effect of differential diagnosis. For MRI measures, regional CVD distribution in dementia was also evaluated.
Figure 2 illustrates the prevalence of dementia. Overall, 64 (9.2%) of 694 male participants and 77 (8.0%) of 960 female participants received a diagnosis of dementia; 141 (8.5%) of 1654 participants had dementia.
For the effect of educational level, the prevalence of dementia was 15.5% (38/245), 7.7% (89/1157), and 5.6% (14/252) for the groups with 6, 8, and 10 or more years of schooling, respectively.
Dementia was present in 21 (18.6%) of the 113 participants who had a history of stroke, who constituted (14.9%) of the 141 participants with dementia. These 21 participants met the NINDS-AIREN criteria for possible VaD.
Table 1 gives the HippW value of each CDR group. Owing to incomplete MRI findings in 5 cases, 492 participants underwent analysis. A significant CDR effect was noted.
For the prevalence of état criblè, the mean total numbers (bilateral sides) ± SE were 4.1 ± 0.3 (CDR, 0), 3.8 ± 0.5 (CDR, 0.5), and 6.6 ± 1.1 (CDR, ≥1). Results of 1-way analysis of variance (ANOVA) showed no significant CDR effect with a significant covariant effect of age (F = 15.635; P<.001).
For small CVD, the numbers were 0.4 ± 0.1 (CDR, 0), 0.6 ± 0.1 (CDR, 0.5), and 1.1 ± 0.2 (CDR, ≥1). We found a significant CDR effect (F = 6.296; P = .002), with no effect of age. For large CVD, the numbers were 0.2 ± 0.04 (CDR, 0), 0.4 ± 0.1 (CDR, 0.5), and 0.7 ± 0.1 (CDR, ≥1). We also found a significant CDR effect (F = 7.225; P = .001), with no effect of age.
Thirty-two participants received a diagnosis of dementia. Figure 3 shows the dementing diseases. The ratios of probable AD, possible AD with CVD, and probable VaD are shown. For these 3 diagnoses, the 2 neurologists were in perfect agreement. The prevalence of VaD was not as common as previously believed. The most common condition was possible AD with CVD. Other conditions included frontotemporal dementia,18 dementia with Lewy bodies,19 hypoxic encephalopathy, posttraumatic dementia, and alcoholic dementia.
Although the diagnoses by the 2 neurologists of the 3 conditions described in the previous paragraph were in agreement, the condition of possible AD with CVD was heterogeneous. As given in Table 2, this condition can be classified into 3 subgroups.
Patients in this subgroup had probable AD with concomitant, nonstrategic CVD. They met the DSM-VI criteria for DAT, but did not meet the ADDTC criteria for IVD. The 2 neurologists completely agreed.
Patients in this group did not meet the ADDTC criteria for probable IVD, and the 2 neurologists did not completely agree on the DSM-IV–based diagnosis of DAT or VaD. Thus, both possibilities existed (described as DAT/VaD).
Patients in this group met the ADDTC criteria for probable IVD. There was complete agreement by the 2 neurologists, despite the fact that the NINDS-AIREN criteria for VaD were not satisfied. No complete agreement was obtained for the DSM-IV criteria, as with subgroup 2.
The mean ± SE HIS values were 2.2 ± 0.8, 4.1 ± 0.5, and 6.7 ± 0.8 for the groups with probable AD, possible AD with CVD, and probable VaD, respectively. Results of 1-way ANOVA disclosed a significant group effect (F = 7.664; P = .003), with the post hoc test showing a higher VaD score compared with the probable AD score (P = .004).
The HippWs (left-right mean ± SE) were 8.4% ± 1.0%, 6.0% ± 0.6%, and 8.7% ± 1.0% for the groups with probable AD, possible AD with CVD, and probable VaD, respectively. Results of 1-way ANOVA disclosed a significant group effect (F = 3.782; P = .04) with no effect of age, with the post hoc test showing a smaller tendency of possible AD with CVD compared with that of probable VaD (P = .08).
As for the regional CVD distribution for dementia with CVD, 6 patients with a diagnosis of VaD (by means of NINDS-AIREN criteria) met the ADDTC criteria for IVD and the DSM-IV criteria for VaD. The 2 neurologists were in perfect accordance as to the VaD diagnosis for these patients. The patients tended to have CVD in bilateral basal ganglia regions, except for the putamen, or at least unilateral large cortical CVD. For the groups with AD and CVD, mostly nonstrategic areas were affected by CVD.
Some methodological issues should be mentioned first. Although the number of patients in participant group 1 was sufficient for the confidence interval to be 95%, the response rate was relatively low. For refusals to participate due to psychological reasons, we considered that asking test questions to older adults might be culturally inappropriate. For refusals owing to physical reasons, although we attempted to perform home interviews, the refusal rate was greater than 50%. Therefore, the low prevalence of VaD might have resulted from the families of the most severely affected residents refusing to participate in the survey. Also, the number of patients in participant group 2 was small due to economic limitations. Among participant group 2, 32 people (6.4%) received a diagnosis of dementia, the prevalence being lower than that among participant group 1. This was due to the older subjects' difficulty in undergoing MRI. Although these 32 people were from a randomly selected sample, we should be cautious about the findings of such a relatively small sample.
A poor interchangeability of the VaD criteria was reported.8 We have not herein reported the κ values because the number of patients was relatively small; because we used only the probable criteria; and because probable AD and probable VaD were in perfect accordance. For MRI methods, precise volumetry is recommended.20 However, the MRI equipment used in this study did not allow us to perform volumetry. We instead measured the HippW that was available for clinical study.21 Despite these limitations, we consider that the results yielded some information on dementia in Japan.
The high prevalence of dementia in Tajiri was previously discussed.2 Surveys conducted in communities that have higher rates of out-migration to nursing homes have a lower prevalence. On the contrary, surveys in communities such as Tajiri and East Boston, Mass,1 which have lower out-migration owing to strong family support systems, tend to have a higher prevalence. Strong family support exists in Tajiri, where 3 generations often live in the same house.2 This tendency does not differ between demented and nondemented older adults. Previously, we estimated the overall prevalence of dementia among adults 65 years and older to be 8.0%.2 We herein identified it to be 8.5%.
The increasing prevalence of dementia with age easily leads to the misconception that dementia is inevitable with aging.22 However, a meta-analysis of epidemiological studies23 concluded that dementia is better conceptualized as an "age-related" rather than an "aging-related" disorder. Results of neuropsychological investigations of healthy older adults11 and longitudinal24 studies on screening test performances during a 5-year period support that idea. However, further investigation is needed for the oldest old population or for the longitudinal incidence.
The prevalence of dementia was high in those with lower levels of education. Although the effect of education has not been fully analyzed compared with that of age, analysis of all such effects has disclosed the prevalence to be high in subjects with lower levels of education.25 Mortimer and Graves26 suggested that education could induce dendritic growth and that more highly educated people are protected to some degree against AD. However, Filley et al27 found no protective effect of education. Moreover, educational level may merely be a marker of other socioeconomic determinants such as nutrition. Further investigation is needed to clarify this point.
Our results showed possible AD with CVD to be the most common dementing condition in Japan, contrary to the previous assumption that VaD is the most common dementing condition in Japan. As described earlier, the different criteria for VaD are not interchangeable.8 The biggest reason is the difficulty in identifying the type of CVD responsible for dementia. Participants with AD and CVD in subgroups 1 and 2 did not meet the ADDTC criteria for IVD, but met the DSM-IV criteria for DAT, since the vascular effect on dementia was considered to be too ambiguous to diagnose as IVD.
Confirmation of this etiologically related CVD requires sophisticated neurological knowledge. With recent developments of MRI technology, even small signal abnormalities can be easily detected. Some are apparently état criblè or perivascular space dilatation. A nonspecialist might easily misdiagnose this condition as VaD simply because the patient was demented and showed MRI abnormalities. The lower prevalence of VaD was supported by the analysis of poststroke dementia. These participants met the NINDS-AIREN criteria for possible VaD. However, we could not rule out the possibility that some residents who had refused to participate owing to physical reasons might be affected by VaD.
Another possibility is that the prevention of CVD has at least been partially successful in Japan, subsequently resulting in a lower prevalence of VaD compared with those of previous reports. Kiyohara et al28 compared 2 surveys of residents of Hisayama, Japan, conducted in 1985 and 1992. The overall prevalence of dementia was decreased from 5.4% to 3.3% for men and 7.5% to 6.3% for women. The prevalence of AD was the same for both sexes. A longitudinal incidence study is planned for participant group 1.
The results indicate that état criblè was related to aging but not to dementia. Instead, the numbers of small or large CVD had some effect on dementia. The results of CVD distributions indicated that bilateral involvement of the basal ganglia region, except for the putamen, or at least unilateral cortical CVD, is associated with dementia. Global neural disconnection based on even subcortical vascular lesions might be associated with VaD, such as in AD as a disconnection syndrome.29 Global cortical glucose use was related to cognitive decline.30
Although the number of patients with dementia was small, we found that the HIS and the HippW were effective for distinguishing VaD or possible AD with CVD from other dementing conditions. The HIS was effective for separating VaD from probable AD. The HippW was smallest in the group with possible AD with CVD compared with probable VaD. This finding suggests that hippocampal atrophy was caused by AD traits and vascular lesions. The hippocampus is vulnerable to degenerative and general conditions.31
Two extreme concepts regarding CVD and dementia are possible. One is that there is no concept for VaD. As a matter of course, large infarctions can affect 2 or more cognitive domains, thus resulting in satisfaction of the criteria of VaD. However, small infarctions should not be considered as primary dementing conditions. What was previously considered as VaD is actually additional CVD in background AD pathologic characteristics resulting in the progression of dementia pathogenesis. Therefore, all VaD should be categorized as possible AD with CVD. The Nun study32 found that the vascular contribution to dementia was not primary but additive to the background AD pathologic traits.
An alternative is that the vascular factor should be considered as primary. The effectiveness of the HIS found in this study supports that concept. Central executive dysfunction could be the hallmark symptom of VaD,33 as memory impairment is of AD. Hachinski and Bowler34 proposed the concept of vascular cognitive impairment, whereas Erkinjuntti et al35 presented criteria for subcortical VaD. A further investigation based on extensive neuropsychological examination is needed to clarify this subject.
Accepted for publication March 13, 2002.
Author contributions: Study concept and design (Drs Meguro, Ishii, Yamadori, and Sekita and Messrs Sato and Hashimoto); acquisition of data (Drs Meguro, Ishii, Yamaguchi, and Shimada and Messrs Ishizaki, Sato, Hashimoto, Shimada, and Meguro); analysis and interpretation of data (Drs Meguro and Ishii and Messrs Ishizaki, Sato, and Hashimoto); drafting of the manuscript (Dr Meguro); critical revision of the manuscript for important intellectual content (Drs Meguro, Ishii, Yamaguchi, Shimada, Yamadori, and Sekita and Messrs Ishizaki, Sato, Hashimoto, Shimada, and Meguro); statistical expertise (Dr Meguro); obtained funding (Dr Meguro); administrative, technical, and material support (Drs Meguro, Ishii, and Yamaguchi and Messrs Ishizaki, Sato, and Hashimoto); study supervision (Drs Meguro, Yamadori, and Sekita).
Corresponding author and reprints: Kenichi Meguro, MD, PhD, Division of Neuropsychology, Department of Disability Medicine, Tohoku University Graduate School of Medicine, 2-1, Seiryo-machi, Aoba-ku, Sendai 980-8575, Japan (e-mail: firstname.lastname@example.org).