Abstract
The population of Korea is 48 million without ethnic diversity. Life expectancy at birth is 75.9 years (72.1 years for men and 79.5 years for women). The proportion of those aged 65 years and over will double in the next 20 years (from 7.3% in 2000 to 15.1% in 2020) [1]. A Korean clinical report suggested that vascular dementia (VaD) was much more prevalent than Alzheimer's disease (AD) in 1988 [2]. Several Korean populationbased epidemiological reports on dementia in the 1990s [3–5] reported the prevalence of all the dementias to range from 6.8% to 10. 8%; the prevalence of AD and VaD ranged from 4.2% to 6.5% and 1.3% to 2.5%, respectively. Alzheimer's disease has become twice as prevalent as VaD in Korea, Japan and China since the early 1990s [6]. Prior to this, in the 1980s, VaD was more prevalent than AD in these countries [7].
These differences in prevalence may, in part, be explained by the methodological differences including differing diagnostic criteria, screening methods and threshold for establishing diagnosis. Additionally, these differences may also be explained by true differences in the incidence of dementia, differential survival and mortality after the onset of the dementia, and the age structure of the society. Theoretically, prevalence is a function of the incidence and the duration of illness.
In the absence of preventable risk factors for a disease with a constant incidence, the prevalence is primarily dependent on the disease mortality rate [6]. Risk factors for VaD are thought to be more modifiable than those for AD. Known risk factors for AD are cerebrovascular disease (CVD), hypertension, arrhythmias, hyperlipidaemia, alcohol-related disorders, diabetes mellitus or smoking, which can be modified and prevented [8–11]. One prospective longtitudinal study in Hisayama, Japan reported on the changing pattern of the prevalence of AD and VaD in 1985 and 1992 [12]. It reported that the prevalence of VaD decreased in men, but not in women, while that of AD remained relatively constant. It was suggested that improved management of blood pressure helped reduce the incidence of strokes and development of less severe strokes, both preventing development of vascular dementia [12]. Also, modification of other risk factors including the antismoking movement may also directly or indirectly influence this epidemiological transition in dementia.
In this study, the prevalence of all dementias, AD and VaD, among elderly Koreans was estimated. The prevalence rates were compared with those observed in other East Asian populations with a view to find attributable or related factors for the reported secular trend in the epidemiology of dementia.
Method
This community study was conducted from December 1996 through August 1997 in Yonchon County, a typical Korean rural agricultural area, which has a relatively stable community. Yonchon is located adjacent to the demilitarized zone (DMZ) between South Korea and North Korea. Yonchon County does not have asylums and nursing homes for the aged. In 1996, there were very few institutions for dementia care in Korea, so almost all dementia sufferers were cared for at home. However, dementia sufferers living alone may have moved out of the county to their children's home if they needed more intensive care. Subjects were required to be living and to legally residing in Yonchon County on 10 December, 1996 (prevalence day) to be considered for the study.
Sampling
This study employed a two-phase design to obtain estimates of the prevalence of dementia and its subtypes. To select subjects, random multistage cluster sampling was performed on the basis of households. Yonchon County had two towns and eight rural districts, which had 54 890 inhabitants. There were 3521 households of 4862 inhabitants aged 65 years and older. The list of 4862 elderly and 3521 households was established from the resident registration records in May 1996. All Koreans legally have to register with the regional resident registration office for taxation and voting purposes. A total of 3521 households were carefully divided into 40 clusters and care was taken to keep each cluster homogenous. One cluster composed of 88 households on average (range 74–102). Sampling stage 1 consisted of a random sample of six areas out of 10. In sampling stage 2, a random sample of clusters was drawn from each selected area. Fourteen clusters out of the 24 from the six areas selected in sampling stage 1 were further selected. Sampling stage 3 consisted of a random sample of households in each sampled cluster. Households with 3 or 9 in the last digit of their serial number were eliminated. In sampling stage 4, the process concluded with a random selection of one elderly person from each household by taking one individual whose birth date was the earliest in any given month and year. Finally, 1217 elderly were selected. Only one subject in each sampled household was selected through this meticulous sampling procedure in order to avoid the sampling bias when identifying risk factors because some socio-demographic and psychosocial factors (including income, marital status, number of family members, type of residence, medical insurance, or social security status) may influence eachother within the same household.
Phase I screening survey
In the phase 1 screening survey, eight trained university students gathered the information from direct face-to-face interview. The screening examination included interviews of both the subject and the informant (spouse, child, other relatives, or close friends). The semistructured interview consisted of demographic questions, activities of daily living [13, 14], psychological and medical symptoms, medical history of subject and the Korean version of the Psychogeriatric Assessment Scale (PAS-K) [15–18]. The Psychogeriatric Assessment Scales (PAS) is a standardized instrument, which focuses on the clinical changes seen in dementia and depression on a set of scales [17]. The PAS-K has been shown to have good psychometric properties for screening for dementia and depression and have excellent validity and reliability on several parameters in Korea [18]. Demographic questions composed of age, education, marital status, past and/or present occupation, drinking and smoking. Medical history included hypertension, diabetes mellitus, stroke, endocrine/hepatic/pulmonary/renal/cardiac disease, head trauma, seizure, Parkinson's disease, alcohol abuse, depression, other major psychiatric illness, and the use of prescribed and/or non-prescribed medications.
Phase 2 clinical diagnostic evaluation
A cut-off score of 8/9 on the PAS-K Cognitive Impairment Scale has been shown to be sensitive in identifying potential cases of dementia [18]. Using this cut-off score, 209 subjects were identified for further evaluation in phase 2. A further 114 subjects, who were identified from the caseness groups of other subscales of the PAS-K (depression scale, behavioural change scale and cognitive decline scale) were also evaluated in phase 2. Also, a randomly selected sample of 50 subjects from the group that did not meet criteria for caseness were selected for phase 2 evaluation. Thus, a total of 370 subjects underwent phase 2 clinical evaluation, and this was conducted from January 1996 to August 1997.
In the phase 2 clinical evaluation, two psychiatrists (S.GH, K. JK) with equal advanced training in neuropsychiatry and dementia research conducted standardized clinical interviews and physical and neurological examination. The clinical diagnosis of dementia was based on the DSM-III-R criteria [19]. Subjects who satisfied the criteria recommended by the National Institute of Neurological and Communication Disorders and Stroke-Alzheimer's Disease and Related Disorders Association (NINCDS-ADRDA) [20] were diagnosed as having probable or possible AD. Subjects who satisfied the criteria recommended by the National Institute of Neurological Disorders and Stroke Association Internationale pour la Rechereche et l'Enseignement en Neuroscience (NINDS-AIREN) [21] were diagnosed as having probable or possible VaD. Though a standardized method of laboratory investigation to rule out other causes of dementia could not be applied to all of them, available medical records were reviewed to find relevant data with the aid of the public health centre of Yonchon County. Subtyping of dementia was based on consensus panel discussion reviewing all the information gathered from phases 1 and 2 evaluations, including the diagnostic impression of the interviewing psychiatrist and the raw materials from the clinical interview, physical examination and neuropsychological tests. Efforts were made to designate the primary cause of dementia in each case. For subjects who seemed to have both AD and VaD processes, we defined the more pronounced as the primary diagnosis. Mattis Dementia Rating Scale, a standardized neuropsychological test battery providing an estimate of general cognitive functioning and including subtests of attention, initiation and perseveration, construction, conceptualization, as well as verbal and non-verbal short-term memory, was applied as this neuropsychological test [22, 23]. The modified Hachinski ischaemic score [24] also was determined for each dementia case for reference purposes. All VaD cases detected in this study showed greater than 4 on the modified Hachinski ischaemic score. Brain computed tomography (CT) was applied to differentiate five ambiguous cases. Though we found few patients with other dementias during our field survey in 1996, we did not define them because, at first, diagnostic criteria for dementia with Lewy bodies and frontotemporal dementia were not definitely defined in 1996 when this study was performed and secondarily, causal relationship for diagnosing alcohol-induced persisting dementia or dementia due to head injury was difficult to prove and finally most of these cases could be categorized as AD, VaD or mixed dementia.
Statistical analysis
The age and sex standardized prevalence of dementia was calculated using the direct standardized method with reference to the 1997 national census data. Ninety-five percent confidence intervals (CIs) for each prevalence estimate were derived using exact method of a binomial parameter [25]. To examine independent associations between the alleged risk factors and dementia, we calculated the prevalence odds ratio (POR) and the 95% CI using a multivariate logistic regression model. Possible confounders such as presence of cardiac disease, hypertension and diabetes and history of cerebrovascular accident (CVA) and head trauma were used as covariates to statistically eliminate their effect on dementia. The model consisted of age, sex, education and cigarette smoking (in pack-years). Pack-years were calculated by dividing the average number of cigarettes smoked per day by 20, the number of cigarettes in a pack, and multiplying the average duration (in a year) of smoking. The relationship between the linear increase in logit risk and risk factors were evaluated by using the likelihood ratio test.
Results
Of the 1217 eligible subjects, 1037 (85.2%) completed the phase 1 screening interview; the remaining subjects were lost due to death (n = 24), not being at home for three attempted visits (n = 104), a home move (n = 32) or refusal (n = 20). There were no significant differences between respondents and non-respondents in age, sex and area. Table 1 shows the demographic characteristics of these phase 1 subjects. Out of 370 subjects selected for phase 2 clinical evaluation, 333 subjects (90%) completed the diagnostic procedures and clinical evaluations. The reasons for non-response in phase 2 were refusal (n = 5), death (n = 8), long-term absence (n = 5) and not being at home for three attempted visits (n = 19). Subjects who did not respond to phase 2 clinical interview were not significantly different from responders in terms of age, sex and education. None of the randomly sampled 50 subjects who did not meet for caseness on the PAS-K had dementia. The mean interval (standard deviation) between completion of phase 1 and phase 2 was 93 ± 62 days.
Sociodemographic characteristics of the study population in a community-based epidemiological survey in the elderly, YonchonCounty, South Korea
Prevalence of dementia
The age and sex-standardized prevalence for all dementias, AD, and VaD are presented in Table 2. The phase 1 study identified 206 cases as having questionable or definite cognitive impairment. The numbers of subjects who were diagnosed as having all dementias, AD and VaD in the phase 2 study were 74, 45 and 26, respectively. The standardized prevalence for all dementias, AD and VaD was 6.8%, 4.2% and 2.4%, respectively. The prevalence of all dementias and of AD increased steeply with advancing age, but VaD showed a two-tailed or bellshaped distribution. Table 2 shows that men and women have different prevalence patterns with age. The prevalence was similar in men and women until the age of 80 years, but after the age of 80 years women had a higher prevalence. The prevalence of VaD was highest in those under the age of 80 years and lower in those over the age of 80 years. The AD/VaD ratio in women was 3.3 (= 5.3/1.6), but men had a higher prevalence of VaD than AD (AD/VaD = 0.7 [= 2.4/3.5]). Three cases with mixed dementia had a similar clinical course to AD associated with focal neurological symptoms and evidence of relevant ischaemic changes confirmed by brain CT scan.
Age-standardized, sex-specific prevalence of all dementias, Alzheimer's disease, and vascular dementia per 100 population aged 65 and older in Korea †
Effects of risk factors on dementia
To explore the separate associations of age, sex, education and smoking with the prevalence of all dementias, AD and VaD, we calculated the odds ratio by multiple logistic regression analysis. As shown in Table 3, older age and lower education were significantly associated with increased frequency of AD, but sex and cigarette smoking had no significant association. Lower levels of education (i.e. less than 6 years) and smoking longer than 30 pack-years were significantly associated with increased frequency of VaD, but sex and older age had no significant association. Older age, lower education and smoking longer than 30 pack-years were significantly associated with increased frequency of all dementias, but sex had no significant association.
Effects of age, sex, education, and smoking on the risk of Alzheimer's disease, vascular dementia and all dementias after controlling for the presence of cardiac disease, and history of cerebrovascular disease and head trauma
Discussion
We identified the prevalence of all dementias, AD and VaD in a rural community of Korea through a two-phase method with diagnosis based on DSM-III-R diagnostic features of dementia, NINCDS-ADRDA criteria for AD, and NINDS-AIREN criteria for VaD. In addition, we explored independent associations of age, sex, education and smoking with dementia after controlling for the possible confounding variables such as heart disease, cerebrovascular disease and head trauma.
Sensitivity (SE) and specificity (SP) of the Cognitive Impairment Subscale of the PAS-K were excellent (SE = 0.90 and SP = 0.84) in an eariler validation study [8]. To reduce false negatives in this study, two experienced psychiatrists performed face-to-face clinical diagnostic evaluation for all subjects entering phase 2, a random sample of 50 subjects who did not meet the criteria for caseness in phase 1. Phase 2 clinical evaluation was performed in 32.1% (333/1037) of those who completed phase 1 evaluation.
Overall response rates in this study were satisfactory (phase 1: 85.2%; phase 2: 90%), and the refusal rates (phase 1: 1.8%; phase 2: 1.4%) were very low. Care was taken to avoid misdiagnosis of dementia due to low levels of education or intellectual disability.
Prevalence of dementia
The rate of AD and all dementias increased exponentially with advancing age. The standardized prevalence of dementia estimated in this study was 6.8%. This figure is similar to that reported from Hong Kong and Japan, as well as in Japanese Brazilians [26–29], but lower than a recent urban Seoul report [30]. Age-specific prevalence rates for dementia estimated in this study are similar to those reported by a number of earlier studies in Asia, Europe and North America [26–29, 31–38].
The reported prevalence of AD and VaD varies enormously throughout the world. These differences in prevalence may, in part, be explained by methodological differences including differing diagnostic criteria, screening methods and threshold for establishing caseness. Additionally, these differences may also be explained by true differences in the incidence of dementia, differential survival and mortality after the onset of the dementia, and the age structure of the given society [6].
The epidemiological transition in the prevalence of dementia whereupon the prevalence of AD has become twice that of VaD since the early 1990s has been observed in the Far-East [6]. Prior to this, in the 1980s, VaD was more prevalent than AD [7]. Many of the risk factors for VaD are potentially preventable, whereas those for AD are more difficult to modify. This epidemiological transition in the prevalence of dementia may be attributable to the modification of risk factors for VaD, although there is no clear evidence of a causal relationship for this [6, 27, 35]. The prevalence of VaD in this study did not increase with age and it was more prevalent in men. This has been observed in Tokyo and Hisayama in Japan [12, 27, 39]. The annual mortality rate of VaD was 25.7% in a Chinese study and its relative risk of contributing to death was the highest from all causes of death [40]. Recent reports on the prevalence of VaD from Europe and Japan show an increase in the prevalence of VaD with advancing age [27, 37, 38]. The prevalence of VaD in those over the age of 80 years may be low if those over the age of 80 years participated less due to higher rates of disability and immobility. However, this is unlikely as we visited subjects at home. Yonchon County dose not have asylums for the aged or nursing homes. Therefore, it is possible that the lower prevalence of VaD in those over 80 years may be due to a higher mortality rate of VaD (especially in the older age group) rather than lower incidence of VaD because of improved health care provision in Korea. However, trends toward less severe strokes are likely in the near future as a consequence of adopting a less traditional lifestyle, improved blood pressure management, acquisition of healthier lifestyles and improved medical care throughout Korea [6].
Factors related to higher mortality rate of VaD in Korea
Over half the elderly Korean population is illiterate [30]. Sixty-one per cent of all subjects had never attended school and 89% had less than 6 years of formal education in this study population. There was a highly significant difference between men and women for ever having smoked despite no difference in levels of education. Most of the never-smoked group were women (86 men, 382 women), while most of the ‘more-than-30-pack-year group’ were men (172 men, 34 women) (Table 1). Both smoking and alcohol drinking are more likely to be observed in men rather than in women in Asia [31, 41, 42, 43]. Education itself may be a proxy for factors associated with socioeconomic status and health behaviours. Long-term smoking, after controlling for other confounding variables, was the strongest risk factor for VaD in this study. Cobb mentioned ‘low educational attainment is associated with an increased risk of non-AD dementia, perhaps because of deleterious smoking habits’ [44]. Long-term heavy smoking may cause cerebrovascular disease leading to vascular dementia. Previously observed ‘higher prevalence of VaD compared to AD in the Far-East’ may be partly related to higher prevalence of smoking in men than in women, because men have higher rates of stroke than women in Korea, Japan and China [45–47]. Moreover, life expectancies in Korean men and women were 72.1 years and 79.5 years respectively in 2000 [1]. It has been previously reported that differential survival (between smokers and never-smokers) may be a source of selection bias in epidemiological studies [48]. Cross-sectional studies can define associations and relationships between hypothesized exposures and outcomes but they cannot always establish temporal relationships between exposures and disease or prove cause and effect. We hope longitudinal, prospective incidence and mortality studies will test the hypothesis that epidemiological transition in the prevalence of dementia is causally related to modifiable risk factors of dementia (i.e. long-term smoking in men).
Acknowledgement
This study was partly supported by (2001) BK 21 Project for Medicine, Dentistry and Pharmacy.