1994 Continuing Review SPECIFIC AIMS

Our recently completed prevalence study of dementia and Alzheimer's Disease (AD) in Indianapolis and Ibadan revealed in initial data analysis a significantly lower prevalence of both dementia and AD in the Yoruba speaking community sample, age 65 years and older in Ibadan as compared to the African American combined community and nursing home sample age 65 years and older in Indianapolis. (Ibadan age adjusted prevalence for dementia 1.76%, for AD 1.11%, Indianapolis age adjusted prevalence for dementia 5.74% community only, 9.06% combined sample for AD, 3.04% community only, 5.68% combined sample). The purpose of this study is to explore possible explanations for this finding by conducting longitudinal follow- up studies of individuals diagnosed cognitively impaired and demented and by conducting a second incidence study in year three. The two communities of related ethnic origin, with different environments and cultures offer a unique opportunity for these studies. The specific aims of the proposal are: Primary 1.To conduct a second incidence study of dementia, AD and cognitive impairment in the two samples, three years after the first incidence study. Hypothesis: There will be a lower incidence of both AD and dementia and of cognitive impairment in Ibadan as compared to Indianapolis. 2.To conduct an annual follow-up study for five years of individuals identified as demented or as cognitively impaired during the initial prevalence and incidence studies. Hypothesis: There will be a slower progression of the dementing symptoms both functional impairment and cognitive decline in the Ibadan subjects as compared to the subjects in Indianapolis. A smaller proportion of cognitively impaired individuals will develop dementia in Ibadan than in Indianapolis. 3. To follow all subjects to determine age specific mortality rates in the two samples. Hypothesis: Age specific mortality rates will be higher in Ibadan than in Indianapolis in individuals with the diagnosis of cognitive impairment and dementia as compared to the mortality rates for the non- cognitively impaired group. 4. To collect blood and type for ApoE alleles in all subjects in both samples. Hypothesis:a) The presence of the ApoE-ε4 allele will be more strongly associated with AD and with cognitive impairment in Indianapolis than in Ibadan. b) There will be statistically significant interaction between ApoE genotyping and some risk factors such as smoking in the two sites. 5. To continue brain autopsy studies in both sites. Hypothesis: Plaques, tangles and other histopathological hallmarks of AD will occur in lower frequency and severity in the clinically diagnosed demented patients in Nigeria than in the African American patients. 6. To complete the data analysis of the current prevalence and incidence study of dementia. Hypothesis:a) The incidence of both dementia and cognitive impairment will be lower in Indianapolis than in Ibadan. b) Risk factors for dementia and cognitive impairment will be identified that are site specific as well as occurring in both sites.

Developmental Aims: 1. To continue to collect data on other putative risk factors. 2.To collect blood samples from the subjects in both sites which will be available to allow other genotyping or other biochemical measurements such as lipid measurements to occur as hypotheses are generated. 3.To construct an algorithm from our screening instrument which will be able to identify with a high degree of accuracy both dementia and AD in the two populations. 4.To refine the criteria for cognitive impairment based upon predicted value in determining outcome. 5.To assess the effects of education, occupation life time learning and social interaction on dementia in the two sites. Data analyses is not quite complete in Ibadan with approximately 60 clinical assessments not having full consensus diagnosis. The final analysis will be available prior to review. Our results will be shared for 1994 Continuing Review multi-site comparisons with the members of the WHO age associated dementia study (Dr. Amaducci, P.I.) and with Drs. Levy and Richards who are conducting a study of Jamaican subjects living in London (See accompanying letters).

BACKGROUND AND SIGNIFICANCE

Alzheimer's Disease is likely to be caused by a combination of aging, genetic and environmental factors. The search for risk factors for AD would be greatly expedited if populations could be identified with significantly lower or significantly higher rates of the disease. However the evidence that such populations exist so far has been scanty. While different rates of dementia and AD have been reported in different countries for example Japanese studies generally report lower prevalence rates of AD than North America and European studies, few studies have used identical methodologies, although the Ni-Hon-Sea study is now applying standardized protocols to Japanese samples (Graves and Kukull, 1994). There also remains the still not well explained differences in the prevalence of dementia, between New York and London in a study which used identical methods and criteria in both sites (Gurland et al, 1983). There are only a few studies comparing rates of dementia in African Americans to whites in the United States with two reports suggesting that in biracial community populations the prevalence of dementia is higher among blacks than among whites (Schoenberg et al, 1985; Heyman et al, 1991). We have now demonstrated significant differences in prevalence rates of both dementia and AD between the African American sample of Indianapolis and the Yoruba speaking sample of Ibadan, Nigeria. (See progress report). As genetic and environmental factors are not likely to be acting independently this finding affords us the opportunity to study the interaction of genetic and environmental factors in two populations originating from the similar geographic area and now living in two very different environments, although we recognize that intermarriage has occurred between African Americans and other American groups. The evidence that AD is at least in part genetically based has increased dramatically over the past two years. Previously familial forms of early onset AD had been linked to two loci on chromosome 21 (St. George Hyslop et al, 1987) and on chromosome 14 (Schellenberg et al, 1992, 1993). These subtypes however accounted for only a very small proportion of the overall individuals with AD. More recently considerable evidence has accumulated that the ε4 allele of ApoE gene on chromosome 19 constitutes a major susceptibility factor for the development of the familial and sporadic forms of late onset AD (Saunders et al, 1993). The risk for AD is higher and the age of onset lower for ε4 heterozygotes and especially for ε4 homozygotes (Corder et al, 1993). We have now extended this association to African Americans where earlier studies using smaller samples had failed to show an association. (See progress report). The finding of the association between ε4 allele and the risk for AD has been described by the Alzheimer's Association National Public Policy Forum as "the most exciting research breakthrough since Congress expanded its commitment to Alzheimers research." As pointed out by several senior investigators the results demonstrate more than ever the need for investigations that are community based, longitudinal in design and involve more than one ethnic group in order to understand the true risk susceptibility for and biological significance of the ApoE-ε4 allele in the etiology of AD. It should also be pointed out that in our study as in the other studies a large percentage of individuals with AD (in our study 44%) did not possess the ε4 allele and some individuals with the ε4 allele (in our study 22%) did not develop AD suggesting that other risk factors are also likely to be involved in the pathogenesis of the disease. The frequency of the ApoE-ε4 allele is reported to be high (30%) in Nigerians (Sephernia et al, 1989) yet the presence of AD appears to be low (Osuntokun et al, 1994). It should be noted that ApoE-ε4 allele predispose also to atherosclerosis which again occurs as of now in relatively low frequency in Nigeria suggesting the presence of an environmental/genetic interaction also in this disease (Ogunnowo et al, 1989). Apart from family history and age, little is known with any certainty about other risk factors for AD although associations with the disease have been reported for severe head trauma, depression, hypothyroidism and a history of Down syndrome or Parkinson's disease (Breteler et al, 1992). The relationship between education and occupation with AD has been controversial. Several cross sectional studies have found an association between AD and limited educational attainment (Zhang et al, 1990, Fratiglione et al, 1985). It has been suggested that advanced educational attainment may supply a brain 1994 Continuing Review reserve either through an acquired set of skills or repertoire or as a result of increased synaptic density in the neocortex on the basis of stimulation which delays the onset of the symptoms for four to five years (Katzman, 1993; Friedland, 1993). This hypothesis has been supported by the finding of an inverse relationship between education and parieto-temporal perfusion deficits in cerebral blood flow studies on AD subjects (Stern et al, 1992). Other prevalence studies do not support this association however (Knoefel et al, 1991). It is interesting to note that the two most recent incidence studies come to opposite conclusions i.e. increased educational attainment reduces the risk of AD in New York (Stern et al, 1994), but had no effect on incident AD in Cambridge (Paykel et al, 1994). If education is indeed associated with the occurrence of AD there remain a number of questions. Is the association between education and AD per se or some closely related factors e.g. socioeconomic status and/or early nutrition? Does the effect of education occur in childhood or is it the result of lifetime learning and continuing intellectual stimulation in the elderly? Obviously in Ibadan where only 14% of the population report any formal education in childhood, education is unlikely to prove to be a major protective factor in that population. We have broadened our screening interview questions to attempt to explore the concept of continuing intellectual activity in both populations. If stimulation of CNS is the major factor, can stimulation occur in ways other than simply intellectual stimulation? One striking feature of the life of the elderly in Ibadan is the amount of social interaction that occurs between the elderly, their large extended family and their neighbors all living in close proximity. In traditional Yoruba society, the elderly have a major responsibility for decisions affecting the family. In addition to providing a protective milieu which may mask the disability of subjects with early dementia, it is possible that this degree of social stimulation has the same effect on the brain as intellectual stimulation. One of the goals of our study will be to develop measures of social interaction which can be used in both sites. Smoking (Graves et al, 1991; Brenner et al, 1993) and the ingestion of estrogens (Graves et al, 1991) and anti inflammatory agents (NSAIDS) (Breitner et al, 1994; Schnabel, 1993) have been suggested as protective factors for the development of AD but all of these findings have been questioned in papers which have been presented recently. The association with smoking may be accounted for by the increased mortality of smokers with the ε4 allele (Welsh, personal communication), the association between estrogen ingestion and AD is lessened when adjustment is made for education level (Graves, personal communication) and one clinic based study could not confirm the association between NSAIDs and AD (Kukull et al, 1994). Our finding of different prevalence rates in the two populations could have several explanations however, in addition to the relative influence of risk factors in the two populations. It is possible that this finding is due to the increased mortality rates for demented subjects in Ibadan due to causes such as bronchopneumonia where the medical care is not as easily accessible as in Indianapolis. This hypothesis is supported by the evidence that patients from the same community, Lundby, Sweden suffering from dementia lived longer in 1957 than they did in 1948 as a result of improvements in medical care (Gruenberg, 1977). Incidence studies by avoiding the issue of differential survival provide a better means of comparison of illness rates between the two populations than cross sectional prevalence studies. They also have the advantage that they provide a measurement of change of cognitive function in individuals over time thus controlling for the possible confounding effects of education and culture on the diagnosis of dementia. Incidence studies are also likely to provide less biased risk factor data, because risk factor data obtained prior to onset of disease is less likely to be influenced by selective memory than data obtained after disease onset. Incidence studies of dementia however are difficult to perform and because of this there are few reports in the literature. All show a consistent increase in incidence rates with age except for the very old population (over 85) where the data are conflicting. Incidence rates of 2.3% for the age group 75-79 years, 4.6% for ages 80-84 years and 8.5% for ages 85-89 years have recently been reported from England (Paykel et al, 1994). Incidence studies, because of relatively low rates of new cases per year require a sufficient amount of time to reveal significant differences, however high death rates particularly in the very old can lead to considerable sample attrition. We are proposing to conduct one additional incidence study, three years after our current incidence study and five years after the initial prevalence study. There are no cross-cultural incidence studies recorded for dementia using identical methodologies. Differences in incidence rates between the two samples of either identifiable dementias or cognitive decline would provide powerful evidence that differential risk factors exist between the two populations. Another explanation for our finding of differences in prevalence rates particularly for the relative 1994 Continuing Review absence of subjects with severe dementia in Ibadan may be that the rate of progression of the symptoms of AD is different in the two countries, being slower in Ibadan than in Indianapolis. The longitudinal natural history of AD is not well defined at present. Most studies done to date have been for relatively short periods of time and have involved patients from tertiary care clinics whose populations may be most representative of patients with long and progressive illnesses. These studies have suggested that the clinical course of AD is somewhat unpredictable and influenced by factors such as extrapyramidal features, family history, behavioral pathology and illness co-morbidity (Heyman et al, 1987; Rubin et al, 1988; Mayeux et al, 1985) There have been no previous comparative cross-cultural longitudinal studies of AD. However one of the major conclusions of the WHO sponsored studies on schizophrenia was that schizophrenic patients from developing countries had better prognosis than their counterparts from industrialized societies (Jablensky and Sartorius, 1988). A major focus of this longitudinal study will be to follow the course of individuals identified as suffering from cognitive impairment using several criteria both clinical and algorithmically derived from our normative neuropsychological data. The ability to detect the earliest stages of AD is becoming increasingly important as new pharmacological therapies are becoming available. The definition of cognitive impairment remains controversial with terms such as benign senescent forgetfulness, age associated memory impairment, questionable dementia, minimal dementia and prodromal AD being used (Levy et al, 1994). Regardless of how it is defined, many of the elderly fall into this category. These individuals represent a high risk group for the development of future dementia as demonstrated by the follow-up study of Paykel et al (1994) where 10 out of 22 subjects with minimal dementia as defined by the CAMDEX developed frank dementia within 2½ years. The characteristics of the cognitively impaired individuals who eventually develop dementia are unclear however. Morris et al (1991) has shown a strong correspondence between information provided by family members on functional change and subsequent autopsy indications of AD. In our proposed study individuals diagnosed with cognitive impairment as well as dementia from the prevalence wave will be followed annually to monitor the rate of progression of symptoms to determine which of the individuals with cognitive impairment develop dementia and to attempt to identify predictive factors from these cases. One of the major limiting factors in epidemiological studies of the dementing disorders is the time, expense and the subject tolerance of the traditional two stage study involving a screening phase followed by a clinical assessment stage. In many developing countries because of a lack of medical resources, the clinical assessment stage is often impossible. If an instrument could be developed that could be used by non-physicians and that could identify not just dementia but also AD with a high degree of certainty this would represent an enormous advantage. We continue to attempt to evolve our screening instrument in the hope that it will be able to serve this purpose. This task has proven to be extremely difficult in the past but the development of the Dementia Questionnaire (Silverman et al, 1986), which can be administered by a trained research assistant and has reported to have good validity both for AD and dementia (Kawas et al, 1994) gives us hope that we may be able to accomplish this goal with the CSI"D."

PROGRESS REPORT

Specific Aims of Original Study The specific aims for the previous study were to compare the age specific prevalence and incidence of dementing disorders especially AD in a sample of both community and nursing home dwelling African American individuals age 65 years and over living in Indianapolis (N=2500) and a sample of Nigerians age 65 years and over living in Ibadan (N=2500), to conduct brain autopsy studies on individuals dying during the course of the study both normal and demented and to conduct a study of risk factors for AD in the two communities in association with the WHO multi-centered studies.

The Community Screening Phase In Indianapolis the community screening phase of the prevalence study was carried out over a period of a year in 29 contiguous census tracts, 2212 African American individuals age 65 years and older living at randomly sampled addresses were interviewed. The simple random sample of 60% of the residential addresses in the catchment area was constructed by the Indianapolis Water Company using all residential 1994 Continuing Review addresses in the tracts as the sampling frame. In the screening phase 121 (4.6%) eligible subjects were too ill to participate, 249 (9.6%) eligible subjects refused, 4915 households were ineligible due to age of residents, 383 households did not have any African American residents. Apart from the fact that they were age 65 or older and African American we were not able to get other information from those who refused or were too ill. It is possible that some of the individuals in the "too ill" group and some of those who refused had dementia but this cannot be documented. In Ibadan the household enumeration and census in Idikan wards was completed on 3489 households, interviews were completed with 2494 individuals. In Ibadan there were 41 (1.6%) eligible individuals who were too sick or refused. The distribution by age groups for the community in each site is as follows: age 65-74 Indianapolis 1307 (59%), Ibadan 1793 (72%); age 75-84 Indianapolis 704 (32%), Ibadan 474 (19%); age 85 and older Indianapolis 201 (9%), Ibadan 227 (9%). Characteristics for the samples are as follows: Table 1. DESCRIPTIVE CHARACTERISTICS OF SAMPLES Category Indianapolis Ibadan

N = 2212 N = 2494

% Male 35.1% 34.9%

% Female 64.9% 65.1%

Mean Age 73.9 72.3 (+7.0) (+7.5) Literacy 97.9% 15.2%

Mean # Years Education 9.6 0.7 (+3.1) (+2.3)

Mean Cognitive Score 30.4 27.9 (+2.7) (+3.7) Mean Cognitive Score 28.9 29.2 Adjusted for Education

Mean Relative Score 2.8 2.5 (+3.4) (+2.4) Mean Discriminant Score .030 .062 (+.13) (+.11)

Selection for the clinical assessment is based upon cut off scores derived first in the study we conducted with the Cree, which were then tested in the pilot study of this project. Selection is as follows: Discriminant score > .184 (100%), Discriminant score > .120 < .184 (50%), Discriminant score < .120 (5%), Cognitive score only < 28.5 (100%), Cognitive score only > 28.5 < 29.5 (50%), Cognitive only > 29.5 (5%). The 5% sample is weighted for 75% age > 75 years. Scores for the two sites are as follows: Table 2. SCREENING SCORES BY GROUP 1994 Continuing Review Category Indianapolis Ibadan

N = 2212 N = 2494

5% Sample n = 1783 n = 2020 80.6% 81.0%

50% Sample n = 167 n = 195 7.6% 7.8%

100% Sample n = 262 n = 279 11.8% 11.2%

Faculty Exchange between Sites Faculty exchanges have occurred between the two sites for purposes of training field workers (K. Hall, 1992, A. Ogunniyi, 1992), setting up the data management system (H. Brittain, 1992, B. Musick, 1992, S. Hui, 1992), and consensus diagnostic conferences (H. Hendrie, 1992, 1994, V. Burdine, 1993, A. Brashear, 1993, M. Farlow, 1993, B. Osuntokun, 1992, 1993, 1994, O. Gureje, 1993, A. Ogunniyi, 1992, A. Adeyinka, 1992). Dr. A. Akang, neuropathologist, spent a full year from July 1992 to July 1993 in Indianapolis training and working with Dr. B. Ghetti.

Normative Study Existing norms for the neuropsychological tests used in this study were derived from non-minority samples. The normative sample for the Consortium to Establish a Registry for Alzheimer's Disease- Neuropsychological Battery (CERAD-NB) consisted of white Americans with a mean age of 68 and two years of college (Morris et al, 1989). The influence of education and age on test performance was not discussed in the original article. The applicability of these norms outside college educated, upper-middle class whites is questionable. We conducted normative studies using the CERAD-NB and the modified CAMCOG in Indianapolis and Ibadan so that we could accurately characterize the neuropsychological performance of subjects receiving clinical assessments in the dementia prevalence study. In Indianapolis, potential African American normative subjects were recruited from churches, community centers and a large outpatient clinic. All potential subjects were living independently in the community. Though these subjects were recruited from the sampling area of the Indianapolis Study of Health and Aging, they were not participants in the dementia prevalence study. Eligibility for the normative study was determined by subject self-report to a brief health history questionnaire. Specifically, potential subjects were excluded if they reported a history of central nervous system disease or disorder (i.e., stroke, AD, Parkinson's disease, multiple sclerosis, brain tumor, seizure disorder, or head injury with greater than 30 minutes loss of consciousness) or current, active, major Axis I psychiatric disorder (i.e., schizophrenia, major depression). In addition, individuals had to report themselves as having normal memory for their age in order to be included in the study. Eighty-three African American men and women aged 65 and older were selected for the normative study by this method. Correlational analyses indicated pervasive education influences on CERAD-NB test performance, age effects were less powerful and noted primarily on the memory indices. Education-stratified normative data were produced for each CERAD-NB test. A significant Education by Age interaction was noted on the Word List Learning Test which indicated that advanced age had more deleterious effect on verbal learning among the well educated than those with limited schooling. Age- and education-stratified normative data were produced for the Word List Learning measures. No sex differences were noted on any CERAD-NB measures. Factor analysis of the CERAD-NB indicated a two factor solution (i.e., a general cognitive functioning factor and a verbal memory factor) which accounted for 67.2% of the variance. The Ibadan Study of Health and Aging involved a population survey; therefore, potential normative subjects had all been screened with the CSI-D. Subjects who were reported by an informant to be normal in cognition and activities of daily living were approached for participation in the normative study. One hundred normal, healthy, Yoruba-speaking Nigerian men and women aged 65 and older completed the Ibadan 1994 Continuing Review version of the CERAD-NB and the modified CAMCOG. Correlational analyses indicated pervasive education-influences on CERAD-NB tests performance. Gender-effects on the CERAD-NB were accounted for by education and there were essentially no age-effects. Education-stratified normative data were produced for all CERAD-NB test (Table 3). Factor analysis revealed a one factor solution which accounted for 54.7% of the variance. (See Appendix VI for reports).

Table 3. SITE-SPECIFIC CERAD-NB NORMATIVE DATA BY EDUCATION (MEANS ± SD) Ibadan N=100 Indianapolis N=83

0 Years 1+ Years 0-11 Years 12+ Years

Mini-Mental Status 19.3 (±3.3) 25.5 (±3.5) 26.4 (±2.2) 28.2 (±1.7)

Boston Naming 7.2 (±2.5) 10.3 (±3.2) 11.2 (±2.5) 13.6 (±1.6)

Animal Fluency 12.4 (±3.5) 16.7 (±4.8) 12.6 (±3.2) 16.8 (±4.4)

Constructional Praxis 3.4 (±2.3) 5.7 (±1.3) 6.7 (±0.6) 7.0 (±0.2)

Word List Learning 14.0 (±2.8) 17.5 (±3.0) 12.7 (±3.0) 17.6 (±4.3)

Word List Delayed 5.4 (±1.1) 6.3 (±0.9) 3.2 (±1.7) 5.1 (±2.4)

Word List Recognition 16.2 (±2.3) 18.5 (±1.6) 17.7 (±1.9) 19.1 (±1.5)

Clinical Assessments For the clinical assessments for which there is a consensus diagnosis the classification of screening interview scores and the diagnostic outcome are shown below:

Three hundred and fifty-one individuals were clinically evaluated in Indianapolis and 388 evaluated in Ibadan. For the group with poor performance on the screening interview in Indianapolis N=261, clinical assessments were completed on 175. Individuals were not clinically evaluated in the poor performance group in Indianapolis for the following reasons: 1. Death N=10 (3.8%), 2. Too sick N=16 (6.1%), 3. Refusal N=51 (19.0%), 4. Other N=9 (3.4%). Information for the individuals who were not seen in Ibadan is not in the computerized data base and is being retrieved by chart review, we do know that refusal rates are very low. 1994 Continuing Review Table 4. CONSENSUS DIAGNOSES Diagnosis Indianapolis Ibadan

N = 351 N = 388

Dementia 65 (18.5%) 24 (6.2%) Alzheimer's type 49 17 Probable 32 15 *Possible 17 2 Vascular 10 6 Other 6 1

Cognitive Impairment 105 (29.9%) 144 (37.1%)

Not Demented 181 (51.6%) 220 (56.7%)

*Includes four individuals with considerable cerebro-vascular component to their dementia in Indianapolis and one individual in Ibadan i.e. evidence of previous stroke but primary diagnosis was considered to be AD. The CDR ratings to date of the demented subjects are as follows: Ibadan: CDR 0.5 N=3, CDR 1 N=12, CDR 2 N=5, CDR 3 N=4, CDR >3 N=0. Indianapolis: CDR 0.5 N=4, CDR 1 N=37, CDR 2 N=8, CDR 3 N=10, CDR >3 N=2.

Nursing Home Study Indianapolis The 1990 census reported 3,020 (8%) of a total of 37,891 African Americans age 65 years and older in nursing homes in Indiana. The original intention of the study was to sample 300 nursing home residents approximately ½ of the 600 African American individuals living in nursing homes in the study catchment area. However this proved to be an extremely difficult task. Many if the nursing homes in the area had administrative problems and cooperation from the staff was difficult to obtain primarily because of a lack of organization in the nursing home. The information available on the patients was often inadequate and family members were often not available. We decided therefore to focus our study on samples of African American residents of 6 homes (N=106) the homes selected having the demographic characteristics, economic characteristics and number of beds similar to the homes not in the study. As the prevalence of dementia was likely to be high, the screening instrument was not administered to the individuals. Instead each of the 106 subjects was given the full clinical assessment with neuropsychological tests, functional assessment interview with a nurse, neurological and psychiatric evaluation. There were 70 females and 36 males. Table 5. CONSENSUS DIAGNOSES Diagnosis Indianapolis

N = 106

Dementia 72 (67.9%) Alzheimer's type 51 Probable 44 *Possible 7 Vascular 10 Other 7 Undiagnosable 4

Cognitive Impairment 21 (19.8%)

Not Demented 12 (11.3%)

Undiagnosable 1 (1.0%)

*Includes two individuals with a considerable cerebro-vascular component to their dementia i.e. previous stroke but primary diagnosis 1994 Continuing Review was considered AD.

Prevalence of Age Associated Dementia and Alzheimer's Disease We still have approximately 60 subjects from Ibadan who do not have a consensus diagnosis. Using only the subjects with a consensus diagnosis we can still obtain a preliminary unbiased estimate. The estimated prevalence rates for dementia and for AD (together with 95% confidence limits) for the two community samples are shown below:

Table 6. PREVALENCE OF DEMENTIA (in %) Age Group Ibadan Indianapolis

Community Nursing Home Combined

65-74 0.65 2.24 45.83 3.02 95% confidence limits (0.30 to 1.00) (1.37 to 3.11) (25.90 to 65.76) (2.10 to 3.94) 75-84 2.06 8.10 72.22 12.69 95% confidence limits (1.18 to 2.94) (6.16 to 10.04) (57.59 to 86.86) (10.60 to 14.78) 85+ 7.53 19.70 76.32 34.39 95% confidence limits (1.94 to 13.12) (14.79 to 24.81) (62.80 to 89.94) (29.34 to 39.44)

Overall Age Adjusted 1.76 5.74 9.06 95% confidence limits (0.66 to 2.66) (4.38 to 7.10) (7.74 to 10.38)

Table 7. PREVALENCE OF ALZHEIMER'S DISEASE (in %) Age Group Ibadan Indianapolis

Community Nursing Home Combined

65-74 0.39 0.93 25.00 1.36 95% confidence limits (0.09 to 0.69) (0.45 to 1.41) (7.69 to 42.31) (0.80 to 1.92) 75-84 1.33 4.06 47.22 7.16 95% confidence limits (0.67 to 1.99) (2.79 to 5.33) (30.92 to 63.52) (5.40 to 8.82) 85+ 4.72 12.68 68.42 27.14 95% confidence limits (1.76 to 7.68) (8.64 to 16.72) (53.64 to 83.20) (22.28 to 32.00)

Overall Age Adjusted 1.11 3.04 5.68 95% confidence limits (0.50 to 1.72) (2.12 to 3.96) (4.71 to 6.65)

For both dementia and AD the rates observed in Indianapolis are significantly higher than those observed in Ibadan. The statistical analysis and assumptions used to determine prevalence rates are as follows: because a stratified random sampling plan was used to select individuals for clinical diagnosis using crude rates to calculate prevalence rates would lead to biased estimates. Commonly, when stratified random sampling is used, prevalence rates are calculated as a weighted combination of sample strata prevalence estimates, where the population stratum proportions serve as the weights (Fleiss, 1981). However as reported by Beckett et al (1992) the estimates can be unstable if the sample sizes in each stratum are not adequately large. Because of this instability, smoothed estimates were obtained using weighted logistic regression. The logits of the probability of dementia and AD were modeled using exact age, sex, and screening group as covariates. The weighted logistic regression adjusts for differential sampling fractions 1994 Continuing Review weighting each individual by one over the probability of their being selected for clinical assessment. This adjustment gives each stratum the same weight in the clinically assessed sample as it had in the screened sample. All computations were carried out in SUDANN, assuming infinite populations. Using the coefficient estimates produced by SUDANN estimated probabilities of dementia and AD were then calculated for each subject involved in the screening. Specific age group estimates were obtained by averaging the predicted values for all individuals belonging to that particular group. Because these predicted values are correlated, determining associated standard errors requires the multiplication of a series of matrices (Beckett et al, 1992). Dementia and AD prevalence rates and their corresponding standardized errors are given in Tables 6 and 7 respectively. The nursing home subjects were not included in the modelling procedure mentioned above. All individuals selected from the nursing homes were clinically assessed and the resulting observed proportion are reported in column three of Tables 6 and 7. Each of the combined age-specific rates, column four of Tables 6 and 7, for Indianapolis were computed by taking a weighted combination of the community prevalence rate and the nursing home prevalence rate. The weights were the proportion of individuals in the age group that were living in nursing homes as reported in the 1990 Census for the state of Indiana; 1.8% for age group 65-74, 7.17% for age group 75-84, and 25.94% for age group 85 and over. Because the nursing home subjects were not included in the modelling procedure, the two estimates are uncorrelated. In order to make a comparison between the sites, the overall estimates for both sites have been direct adjusted to the age distribution of African American Indiana residents observed in the 1990 Census, 60%, age 65-74, 30%, age 75-84, and 10%, age 85 and over.

Maps In Indianapolis the Indianapolis Mapping and Geographic Infrastructure System (IMAGIS), that is the result of a 9.2 million dollar consortium involving the city, utilities and Indiana University, contains a very detailed accurate map of Marion County. Data for the study have been merged onto a unique layer of the system enabling spatial analysis now and in the future for possible hypothesis testing. See Appendix V for maps of the study area showing location by age, sex and diagnostic outcome.

Risk Factors for Dementia Our risk factors analysis is just beginning. The following represents some preliminary results for subjects who were clinically evaluated. Information on risk factors was obtained from relative informants at the screening (Stage I) level. A multiple logistic regression analysis was used to study the relationship between dementia and various risk factors: the dementia rates were adjusted for potential confounding by age, sex and age by sex interaction.

Table 8. RISK FACTORS ANALYSIS FOR DEMENTIA Ibadan N=313 Indianapolis N=347

Odds Ratio p Odds Ratio p Head injury with loss of 0.98 0.98 0.61 0.35 consciousness

History of fracture 2.14 0.26 1.67 0.41

Heart attack 1.71 0.33 0.83 0.67

Angina 1.60 0.48 0.98 0.96

Hypertension 1.88 0.26 0.88 0.68

Diabetes 3.67 0.25 0.84 0.62

Stroke 5.87 0.14 1.68 0.19 1994 Continuing Review Smoking 0.66 0.49 0.46 0.01*

Alcohol 1.82 0.29 0.59 0.09

Moderate Alcohol 1.06 0.93 0.44 0.01*

Heavy Alcohol 5.49 0.08 0.58 0.21

Parkinson's Disease 3.08 0.05* 1.75 0.63

Living with others 0.33 0.06 1.60 0.13

Cancer ------0.63 0.37

Electroconvulsive therapy ------1.29 0.75

The putative protective effect of smoking in AD has been reported by other investigators. The possible protective effect of moderate alcohol consumption on cognitive function deserves further investigation as some recent evidence supports this (Christian et al, 1993) although heavy drinking has been reported as a risk factor for dementia in many studies. Parkinson's disease appears to be a risk factor in Nigerians, but not in African Americans in Indianapolis. Living with others, which could be a measure of psychosocial stimulation appears protective (although not significantly) in Ibadan but not in Indianapolis.

Genetic Study ApoE-ε4 We genotyped for ApoE blood samples from the first 85 subjects (51 female, 34 male) from Indianapolis who were adjudged to be normal (32 female, 22 male) or suffering from AD (19 female, 12 male). No subjects with the diagnosis of cognitive impairment were included. There were no significant differences in gender between the AD patients and control subjects. The AD patients ( mean age 82.7 ± 5.7 years) were significantly older than the control subjects (mean age 78.2 ± 6.1 years). The frequencies of the ApoE alleles in control subjects were ε2 10.2%, ε3 76.8% and ε 13.0%. The corresponding frequencies in AD patients were ε2 4.8%, ε3 54.8% and ε4 40.3%. The frequency of the ε4 allele among AD patients was significantly higher than that among the control group (p<.0001). The frequency of the ε2 allele was lower in AD patients than in control subjects but this difference did not reach significant levels (p=.22). 22.6% of the AD patients were homozygous for ε4 as compared with only one control subject (1.9%). The percentage of AD patients with either one or two copies of ApoE-ε4 was significantly higher than in controls (Chi square=13.8, p=.001). Since the AD patients were significantly older than the controls, a multiple logistical table controlling for age was applied but the association between ApoE- ε4 and AD remained highly significant (p=.0032). Odds ratios (OR) for AD were developed from this model for one or two copies of ApoE-ε4. For one copy the OR was 4.2 (1.27 to 13.9, 95% confidence intervals) and for two copies the OR was 39.6 (3.43 to 459.0, 95 % confidence intervals). This study extends the association that the observed allelic association is biologically relevant. (See Appendix VI for report). The funding for the pilot study came from Washington University AD Center (Schellenberg, P.I.) and from the Alzheimer's Disease Core Center in Indianapolis, and from the Albert E. Sterne fund.

Radiological Evaluations As part of the clinical assessments 53 individuals had head CT scans in Indianapolis and 44 individuals had head CT scans in Ibadan. In Indianapolis 15 of the head CT scans were performed at other institutions as part of a dementia work up. In addition to reviewing the images for diagnostic purposes a more formal comparative analysis of brain atrophy and its relationship to cognitive scores in the two sites is underway using the following procedure: Multiple axial non-contrasted CT images of the brain were obtained at both sites. Ten mm. thick contiguous images were obtained at 15 degrees to the canthomeatal line. The images were reviewed for any pathologic findings. Following this, the images were evaluated as per the data form. Using a cursor, the distances of various structures were measured (bi-caudate ratio, bi-frontal ratio and pre-pontine cistern ration). Using a small region of interest cursor, density measurements were obtained 1994 Continuing Review in Hounsfield Units (frontal and temporal density). Visual determinations of atrophy were made (vertex grade, cerebral atrophy and ventricular size). Utilizing a cursor, a tracing was made of the margins of various structures. When the tracing was closed to create a defined area, the CT machine calculated the area in square cm. (brain, frontal horn, temporal horn, fourth ventricle, third ventricle, sylvian fissure, suprasellar cistern and lateral ventricular). Lastly, any miscellaneous findings, including infarct, were listed. Prior to the study Dr. Adeyinka, Ibadan radiologist, visited Indianapolis for one week to train with Dr. Smith in using these techniques. Within site reliability studies reanalyzing 10% of the brain images are underway. Between site reliability is more difficult due to difference in equipment. Data analysis for this study is not yet complete.

Autopsies Obtaining consent for autopsies has proved to be a very delicate task for our African American population in Indianapolis and for the residents of the Idikan ward in Ibadan. Our difficulties with the African American population in Indianapolis mirrors the problems encountered with this population throughout the United States (Ballard et al, 1993). We are attempting to resolve it in Indianapolis by enrolling as many African Americans as possible regardless of diagnosis into our NIA supported Alzheimer's Core Center where we have a special educational program directed towards the necessity of obtaining autopsies for Alzheimers research. So far we have enrolled 77 African Americans in our Alzheimer's Core, 37 male, 40 female, 29 with dementia, 17 with cognitive impairment, 31 normal subjects. So far 6 individuals have given preliminary agreement to autopsies but this educational process is in its earliest stages. In Idikan wards, the problem is that 80% of the population are Moslem. It is the Moslem custom that individuals be buried within 24 hours of death. The prospect of obtaining autopsies is primarily with those individuals who die in the hospital. So far one brain has been harvested in this manner. This procedure has however worked well in a previous study conducted by Dr. Osuntokun and Dr. Masters (Lancet, 1994). As examination of Alzheimer's like changes in elderly brains particularly in normal brains is an important focus of our study we are proposing to extend our brain autopsy studies in both sites to African American and African subjects who die in the University Hospitals of Ibadan and of Indianapolis and who have agreed to an autopsy regardless of whether or not these individuals are part of the community study. As part of this study, every attempt will be made to evaluate the individual's cognitive status prior to death providing sufficient notice can be given.

Jamaican Pilot Study We have been awarded a supplement to our current grant for the year July 1, 1994 to June 30, 1995 to conduct a pilot study with the Jamaican residents of African origin. In this pilot study our current screening instruments and clinical assessment methods will be translated, harmonized and pretested for use in the Jamaican population and standardization studies will be conducted to ensure that the instruments and diagnostic criteria are being used reliably and in an equivalent manner by the Jamaican investigators. The study will be conducted on a random sample stratified by age and sex of 85 community dwelling individuals and 15 known demented individuals to evaluate sensitivity and specificity of the screening instrument and the clinical assessment procedures.

Tracking Study Participants Indianapolis: In Indianapolis two methods are being employed to track individuals. The first method is a telephone call from one of the field interviewers to confirm address and telephone number of participants. Contacts were made from February 1994 through April 1994. The telephone contact also included a five minute interview asking about health care utilization (Questionnaire is in Appendix VII). The participation and briefly some results of the telephone interview are shown in the table below. Secondly, each month notes are sent to participants who have birthdays during the month. 1994 Continuing Review The envelope is stamped as follows: "forwarding address requested" to obtain forwarding addresses if the individual has moved. Table 9. INDIANAPOLIS HEALTH UTILIZATION SURVEY

Completed the interview: 1768 (80.2%) Have health insurance: 97.7%

Confirmed address but no interview: 68 (3.0%) Have a source of health care: 97.4%

Not reached (quit after 3 tries): 212 (9.6%) Saw M.D. in last year: 88.9%

Moved to nursing home: 29 (1.3%) Mean M.D. visits in last year: 4.1 ± 3.9

Refused: 13 (.5%) Self rate health excellent and good: 52.9%

Deceased: 86 (1.3%) Feel health needs are met: 95.4%

The data for the utilization study were analyzed and reported in a Masters thesis by D. Mosley. (See Appendix VI). He and K. Hall are revising the report to submit for publication. Ibadan: In Ibadan the field interviewers visit the area on a regular basis. The Idikan wards are well established political and social entities with concomitant formal and informal information networks. Interviewers track movement of study participants using these information networks. The field interviewers also stay in contact with the staff of clinicians and visiting nurses of the clinic located in the geographic center of the area. In Idikan to date 64 deaths have been reported.

Incidence Study 1994-1995 Phase I: Individuals are included in the incidence study on the basis of their scores on the screening interview and clinical assessments in the prevalence study. It is important in the study of new cases to exclude individuals who are known to already have had the disease or had a high probability of having had it at the time of the prevalence study. All of the individuals who scored in the good performance group in the prevalence study will be included (1.0% in Indianapolis; 2% in Ibadan, of this group were demented). All of the individuals who received clinical assessments in the prevalence study and were diagnosed as not demented and not cognitively impaired and therefore free of Alzheimer's disease will be included in the incidence study. Individuals who were selected for clinical assessments from the poor performance, 100% sample group, and the intermediate performance, 50% sample group, and who were not clinically evaluated (due to refusal or illness) will be included in the incidence study but the data for this group will be analyzed separately because they may have already had the disease at the time of the prevalence study. In Indianapolis as of June 6, 1994, 227 interviews have been completed with 5 (2.1%) refusals. The study is underway in Ibadan so far with no refusals, with approximately 150 interviews. Phase II: The clinical assessments will be given to a subset of subjects screened based upon the categories used in the prevalence study and within categories, change from the initial prevalence screening interview in the discriminant score or change in cognitive score (for those without an informant interview). For those with change in the score, interview scores will be grouped into three categories, as follows: no or little change, intermediate change, and great change. All individuals in the group with great change will be eligible for clinical assessment, of those in the intermediate change 50% group will have clinicals and in the no change group 5% will be sampled for clinical assessment. In addition because of the particular difficulties in distinguishing age associated cognitive decline from dementia in the very old clinical assessments will be offered to all study participants age 90 years and older. All patients diagnosed with Alzheimer's disease and other dementias are being seen for follow up after approximately two years from the first diagnostic assessment. The purpose of this follow-up is to document the course of the dementing process and cognitive impairment. The instruments for the community interview and clinical assessments are essentially the same as were used in the prevalence study. In the community interview, minor modifications were made to eliminate redundant items, and some neurological assessment items were added. (See Appendix VII). The protocol for clinical assessments did not change. The protocol for following up cases of Alzheimer's and other 1994 Continuing Review dementia is in "Methods" section.

Changes in Specific Aims The major change in our proposed specific aims from our previous submission is the inclusion in the study of genotyping specifically for ApoE. In Indianapolis we will not follow up the nursing home sample except for those subjects with AD whose family will agree to enter our Alzheimer's Core and will permit autopsy studies. No funding is requested for the autopsy study in this grant. This will be covered by our Alzheimer's Disease Core Center. Of course subjects from the community sample who are admitted to nursing homes will be followed. We recognize that we have not completed the incidence study before applying for renewal of funds. However we are applying at this time in order to ensure that the essential interaction between members of our team and the individuals in the study population be continuous. A break in continuity could make it extremely difficult to start up again this type of longitudinal study.

METHODS

Research Design This longitudinal study will conduct annual clinical assessments of all patients diagnosed, either at baseline prevalence or the 1994 incidence study, as demented or cognitively impaired. In year three an incidence study will be carried out following the two stage design of an initial screening interview and a second stage clinical assessment of a subset of those screened. Annual telephone or home visit contacts will be made, birthday cards sent, and formal and informal information networks will be employed to track study subjects.

Populations Ibadan: Idikan Wards (population about 50,000) are within the municipality of Ibadan. The Idikan wards are political and administrative units of the more ancient parts of Ibadan (a city founded by itinerant warriors at the end of the 18th century). The population is fairly stable and is comprised mainly of the Yorubas a negroid predominantly town-dwelling people about 35 million in population domiciled in South- western parts of Nigeria and the adjacent contiguous eastern region of Benin formerly Dahomey. Some members of the communities of Idikan wards do go away to the farms, usually for a few days. For the study, a census listing was conducted enumerating 3489 households, in Idikan wards of those 2494 individuals age 65 and older were interviewed, 41 (1.6%) refused. Because of the expected low prevalence of dementia in Nigerians in Ibadan and because missing a few cases would seriously influence the assessment of prevalence ratios and incidence rates, it was desirable to carry out a total population survey by door to door screening in a geographically defined population, as was done in the USA's Copiah Study (Schoenberg et al, 1985), in a previous study of neurological disorders in Nigerians (Osuntokun et al, 1987), in the studies of the age associated dementias in Northern Manitoba Cree Indians and in the recent Italian Study, (Rocca et al, 1990). Also, as pointed out in the WHO's (1989) protocol, sampling an entire village or a well defined area offers advantages in terms of clear structure, low refusal rate and homogeneity. The representativeness of these populations at the national level would of course be questionable but this problem is also inherent in other forms of sampling. Dr. Ross, together with Drs. Denis Evans and Marcia Baldereschi made a WHO-sponsored visit to Ibadan in December 1990 to meet with Dr. Osuntokun and his colleagues to assess the facilities for the research. (See Appendix IV for report). For individuals with no recorded dates of birth, date of birth was estimated from a table of "historical" landmarks well known to the population, a well-tested long-standing practice in Nigeria for assessing ages of adults. In a previous study Ogunniyi and Osuntokun (1993) estimated the ages of 57 elderly individuals using this method. When the estimates were then compared to accurate birth records the estimates proved to be accurate within three year for 94% of cases. During the course of the original study social unrest and political instability in Nigeria have been widely reported by news media. In spite of some general strikes and the annulment by the military of a Presidential election that was widely believed to be fair and held on June 12, 1993, the installation of a puppet civilian interim government on August 27, 1993 and a coup d'etat on November 17, 1993 the Nigerian 1994 Continuing Review research team has persisted and successfully completed the field work and clinical assessments for the study. The successful completion of work demonstrates the resourcefulness of the team in Ibadan to work under difficult conditions. The political future of Nigeria is unclear but the track record of the project in Ibadan is clear and we are confident that the longitudinal study will succeed. (See maps in Appendix V). Indianapolis: The geographic target area used in the study consists of 29 contiguous tracts in the middle of Indianapolis (see maps in Appendix V), and was selected on the basis of demographic and economic factors. For those census tracts African Americans represented 80% of the population in the 1990 U.S. Census, and represent over two thirds of the African American population age 65 years and older in Indianapolis. Age, sex, minority distribution and socioeconomic status of the residents of these tracts are the same relative to Marion county as well as to the entire state of Indiana (Validated by Indiana state demographer Dr. J. McKibben). The 1990 U.S. Census total population for the study area was 82,387 with 71,003 (86.1%) African American and a total of 32,954 households. The age distribution for the 2,212 African Americans age 65 years and older in the study sample is similar to that of the population: age 65 to 74 census N=5815 (57.3%), sample N=1307 (59%), age 75 to 84 census N=3223 (31.7%) sample N=704 (32.0%), age 85 and older census N=1099 (10.8%) sample N=210 (9%). The interviewers in Indianapolis are native residents who grew up in the study area and have extensive family and social ties in the community. They have had responsible jobs (four had previously retired) and prior to joining the study earned respect and high stature in this community. We are sure that these qualities in the interviewers produced the relatively low refusal rate (9.6%) for this study. These interviewers will continue working with the study to track participants. They have been trained by the Central Indiana Council on Aging as liaisons to the various social and health service programs in the area. The tracking of participants and linking them with needed services are crucial parts of the planned longitudinal study. The study has forged ties with community leaders, neighborhood associations and other community groups, these ties will be carefully nurtured to insure continued grass roots support.

Research Plan The community samples for the initial prevalence study for the two sites are grouped as follows: Group Indianapolis Ibadan 1. Total community sample 2212 2494 2. Clinically diagnosed dementia 65 24 3. Clinically diagnosed cognitive impairment 105 144 4. Clinically diagnosed not demented 181 220 5. Possible cases without clinical assessment 86 NK (at this time) 6. Borderline and Good performance on screening interview (no clinical) 1956 2225 The proposed study will follow these groups as follows: Groups 2 and 3: Clinically Diagnosed Dementia and Cognitive Impairment Annual follow-up clinical assessments will be done on all individuals who have been diagnosed as demented and cognitively impaired. The clinical assessment protocol currently in use for follow-ups will be employed. Groups 4 and 6: Clinically Diagnosed Not Demented; Borderline and Good Performance on Screening Interview with No Clinical Assessment

Year 1 -Telephone or home visit contact Year 4 -Telephone or home visit contact Year 2 -Telephone or home visit contact Year 5 -Telephone or home visit contact Year 3 -Community survey for incidence study -Clinical assessments for incidence study

All individuals who are not diagnosed with dementia or cognitive impairment will be recontacted in year one and two by telephone in Indianapolis (except for those without telephones who will be visited) and 1994 Continuing Review by home visit in Ibadan to confirm their address and identify deceased subjects. This contact will also include a brief questionnaire on health related questions, similar to the health utilization survey conducted in Indianapolis in 1994. In Indianapolis individuals who move to nursing homes will be followed. The Central Indiana Council on Aging receives data on all nursing home admissions and has agreed to share information with the project. In year three, 1997, community screening interviews will be carried out to identify new cases of dementia for the incidence study. A subset of individuals who are screened will receive clinical assessment for diagnosis as described under "Incidence Study" in the "Progress Report" section of this proposal. Individuals newly diagnosed with AD will be offered treatment in the Alzheimer's Core Center and will have annual follow-up assessments, other dementias and cognitively impaired will also have annual follow-up assessments in year 4 and 5. Group 5: Possible Cases without Clinical Assessment

Year 1 -telephone or home visit contact Year 4 -telephone contact Year 2 -telephone or home visit contact Year 5 -telephone contact Year 3 -community screening -clinical assessments

The group of individuals who scored in the poor performance group and were not clinically assessed in the initial prevalence study will be invited to participate in the study for tracking purposes in years 1 and 2 with the telephone (home visit) contact. In year 3 they will be invited to have the screening interview. The data for this group will be analyzed separately from the data for groups 4 and 6 because it's possible that individuals in this group might have had dementia either at the time of the initial prevalence study or subsequently. We will attempt to track these individuals over the course of the entire study.

Mortality In Indianapolis 86 individuals have been reported as deceased and in Ibadan 64 have been reported. In Indianapolis Death Certificates have been requested from the Indiana State Board of Health. In Ibadan where death certificates are not routinely used the interviewers record the date of death and the symptoms and medical conditions known to the family at the time of death. In Indianapolis autopsies are not routinely done in elderly; family physicians usually report known medical conditions and may determine cause of death from symptoms reported by the family. Due to these customary practices the specific cause of death may be somewhat questionable in both sites therefore crude age specific death rates will be used.

Instruments and Clinical Evaluation Screening Instrument: The Community Screening Instrument for Dementia (CSI"D") was developed during the course of the study of Cree in Manitoba and has now been translated into Yoruba as well as Cree and English. This is a hybrid instrument consisting of items from the following widely used dementia instruments: CAMDEX (Roth et al, 1986), MMSE (Folstein et al, 1975), Dementia Rating Scale (Blessed et al, 1968), Comprehensive Assessment and Referral Evaluation (CARE) (Gurland et al, 1977), and the East Boston Memory Test (Evans et al, 1989). Items were selected which would in short form test memory, abstract thinking, judgement, other disturbance of higher cortical function (aphasia, apraxia, agnosia, constructional difficulty) personality changes and functioning at work and in social relationships. These items are consistent with DSM-III-R and ICD-10 criteria for dementia. It has two parts, a cognitive and risk factor section consisting of 39 items which takes approximately 20 minutes to administer and an informant section consisting of 30 items which takes approximately 15 minutes to administer. (See Appendix VII). High reliability can be obtained with this instrument using lay interviewers after an appropriate, two week training course. In the community prevalence study in Indianapolis sensitivity was 92.0%, and specificity was 84.0%; and in Ibadan sensitivity was 76% with specificity 82%. These estimates were obtained from a weighted average of the stratum-specific rates, i.e. applying these rates to the entire sample. Although we missed only 2% of demented subjects in the "good performance" group, similar to reports by others (McDowel and Lindsay, 1994), we plan to refine the discriminatory score to optimize the correct classification of subjects. Both cognitive and informant sections attained excellent comparability in both the Cree and English-speaking 1994 Continuing Review populations in Manitoba and in pilot studies with the African Americans in Indianapolis and the Yorubans in Ibadan, although the cognitive score is significantly correlated with education in all populations. Appendix VI contains a full description of the psychometric properties of the CSI"D" as well as its performance as a screening instrument for dementia. Questions regarding risk factors consistent with the World Health Organization protocol for the Age Associated Dementias have been added to the screening instrument.

Informant Interview for Clinical Evaluation: A structured interview with an informant (See Appendix VII) was developed during the course of the Manitoba and Indianapolis/Ibadan study and follows the format of the clinical assessment being used in the WHO study. It was based primarily on the CAMDEX but has been modified so that it includes information pertinent to the diagnosis of dementia and the dementia subtypes. After training sessions in Indianapolis and Ibadan, a very high degree of inter-rate reliability has been obtained with a Kappa of one for almost all items.

Neuropsychological Assessment: A neuropsychological assessment was derived from the CAMDEX neuropsychological evaluation, the CAMCOG which contains the MMSE and a series of well-standardized neuropsychological tests used by CERAD and WHO. It has been adapted for use in Ibadan and again has a high degree of reliability after suitable training. (See Appendix VII).

Physical and Neurological Examination: Are recorded according to the CERAD form (See Appendix VII).

Laboratory Investigation: Laboratory studies are ordered when deemed clinically appropriate to arrive at diagnoses.

Radiological Investigation: CT scans of the head without contrast are ordered at both sites when deemed clinically appropriate and as far as possible for all subjects with dementia. In addition to their use for diagnostic purposes a more structured evaluation of brain atrophy is being conducted, as described in the Progress Report.

Research Blood Samples: As far as possible research blood samples for genotyping will be taken from every individual clinical assessment during the clinical assessment phase. In addition an attempt will be made to collect blood spots on filter paper on all subjects in the study.

Quality Control and Reliability Screening interview data are reviewed by a supervisor before being sent to data entry. If there are ambiguities or inconsistencies the supervisor goes over the instrument with the interviewer, who if necessary may recontact the subject to get clarification. Data are entered by data entry person #1 (E. Breen Indianapolis, B. Olaeye Ibadan) re-entered to verify by person #2 (K. Jessup Indianapolis, L. Amusa Ibadan). Screening data are periodically analyzed for interviewer comparisons. Prior to each phase the interviewers are trained to maintain high reliability. Interviewer comparison analysis is also periodically done for the Informant Interview for the Clinical Assessment and neuropsychological tests. Formal training via case review, item by item, is carried out every six months with the nurses and psychometricians.

Diagnostic Criteria Criteria for Dementia: Both DSM-III-R criteria and ICD-10 criteria for dementia are recorded separately and together. To meet criteria for dementia for our WHO related study both sets of criteria have to be satisfied. ICD-10 criteria from our experience tends to be stricter than DSM-III-R criteria perhaps because of the requirement in ICD-10 that a decline in both memory and in judgement and reasoning be identified that separately interfere with social and occupational functioning. Severity of dementia is recorded mild/moderate/severe using both ICD-10 and DSM-III-R criteria. In addition CDR ratings are made (Berg, 1984).

Criteria for Cognitive Impairment: There is as yet no generally recognized criteria for cognitive impairment. 1994 Continuing Review To be consistent with other WHO sites, we chose a very broad clinically derived definition of cognitive impairment in which clinician makes a judgement of cognitive impairment based upon historical information and information about activities of daily living from the informant and from cognitive testing. If in the clinicians judgement there is evidence of some cognitive impairment not sufficient to meet criteria for dementia a diagnosis of cognitive impairment is made. In addition an etiologically based subclassification is made based upon available clinical data as follows: Pro-dromal AD: some evidence of cognitive decline usually decline in memory function with no other obvious cause for this decline. Post stroke: cognitive decline clearly followed vascular event and was not progressive except in the context of another cerebro-vascular event. Alcoholism: cognitive decline occurring in the context of continued drinking. Sensory handicap: cognitive impairment occurring with severe sensory handicap blindness/deafness where it is difficult to distinguish the effects of the handicap vis a vis brain disease. Depression: cognitive impairment in association with major depression disorder. Severe physical illness: cognitive impairment occurring in the context of severe physical illness e.g. cancer, heart disease. Often these illnesses are life threatening. Static Cognitive Defects: Cognitive defect present but appears to be lifelong. No evidence of decline e.g. mental retardation, learning problems. n.b. This is a very broad definition of cognitive impairment, broader than the criteria used in other epidemiological studies (such as minimal dementia or questionable dementia). It was chosen deliberately so that a large number of individuals meeting this definition could be followed as so little is known about the earliest stages of dementia. From the data collected from these individuals, other less encompassing criteria can be applied for example the newly proposed criteria for IPA and WHO Aging and Associated Cognitive Decline (Levy) or AD prodrome (CERAD). The relative merits of the different criteria in predicting outcome can therefore be assessed. From our normative data on the CERAD battery now available for Indianapolis and from Ibadan statistical definitions of cognitive impairment are also being applied. (See Appendix VI).

Criteria for AD: The diagnosis of probable/possible DAT is made using the criteria proposed by NINCDS/ADRDA.

Criteria for vascular dementia: The diagnosis of vascular dementia is made using ICD-10 criteria.

Secondary dementias: Secondary dementia will be diagnosed on the basis of the medical history and of specific clinical and laboratory findings. ICD-10 criteria are applied.

Unspecified dementia: Unspecified dementia will be diagnosed when the diagnostic criteria for dementia are satisfied and the subject cannot be assigned to a specific dementing disorder. However, every effort should be made to reach a specific diagnosis.

Clinical Assessment Processes All clinical assessments are made blind to the screening status. Informant interviews are conducted by a specially trained nurse. Neuropsychological assessment of the subject is performed by a technician. Information is reviewed by physicians who then confirm essential data by a short interview with informant and subject. Physical and neurological examination is performed by the physician. The physician also writes a short clinical summary excluding the provisional diagnosis on each subject. On the basis of the examination laboratory studies and a head CT scan may be ordered (for all subjects with suspected dementia unless these studies have already been conducted). Every attempt has been made to ensure diagnostic consistency across sites. All physicians participated in the WHO sponsored diagnostic standardization studies (Baldereschi et al, 1994). The diagnostic process is as follows: 1.After reviewing all clinical data, the examining physician makes a preliminary diagnosis. 2.A consensus conference on diagnosis is conducted at each site where the examining physician, the nurse, the neuropsychologist and at least one other physician (In Indianapolis Dr. 1994 Continuing Review Hendrie always participates, in Ibadan Dr. Osuntokun always participates) review all the data and arrive at a consensus diagnosis. 3.The clinical data from one site is reviewed blind to the site diagnosis at the other site by a visiting physician from the other site and the visiting physician records the diagnosis. 4.A consensus conference involving physicians from both sites review cases with discrepant diagnosis between sites and arrive at a final consensus diagnosis.

Follow-up Clinical Assessment Process The procedures for the follow-up clinical assessments for already diagnosed cognitively impaired or demented individuals are essentially similar to that described above. A large number (30%) of the clinical assessments have had to be conducted at home due to the reluctance of the elderly and often very frail individuals to come to the clinic even though transportation is provided for them. With home visits essentially the same procedures are carried out. The nurse and technician visit the subject at home first and collect the data as above. The physician reviews the data then makes a home visit to conduct a physical and neurological exam, collect blood for lab tests and order a CT scan if necessary. We anticipate that this large percentage of home visits for clinical assessments will continue or highly increase in our proposed study. When head CT scans are ordered, transportation is provided to take the subject and a family member to the hospital.

Genotyping Blood collection, shipment, and handling: For the purposes of ApoE genotyping, blood samples will be drawn from all subjects seen in clinical assessment including demented individuals, individuals with cognitive impairment and normal subjects. In addition the collection of blood spots on filter paper will be attempted for all other individuals in the community samples who agree to this process. It is estimated that about 1200 individuals from each site will agree to blood spot analysis in addition to approximately 2/3 of the individuals in the clinical assessment who will agree to donate blood samples, approximately 600 individuals in each site. The NIA supported AD core has funded a pilot project to Apoe genotype approximately 100 individuals in each site during the year 1994-1995. Two 10 ml EDTA (purple top) and one 10 ml sodium heparin (green top) vacutainers will be collected from each individual. The EDTA blood tubes will be used for DNA extraction and the heparin tubes for lymphoblast transformation. Lymphoblast cell lines will be established only from patients with AD in Ibadan and from familial cases of AD in Indianapolis. The heparin blood tubes from other subjects will be stored at -70°C as an additional source of DNA. The blood samples for lymphoblast transformation will be collected as part of the NIA-sponsored Alzheimer Disease National Cell Repository (Conneally PM, PI; Tischfield JA, Co-I) which is housed in the Department of Medical and Molecular Genetics. The present application is for funds for salaries and supplies needed for DNA extractions and for ApoE genotyping. These funds will not be used for setting-up lymphoblast cell lines. Blood samples from the Indianapolis site will be drawn by nursing and medical staff during the clinical assessment stage in Indianapolis. Blood samples from the Ibadan site will be collected by Dr. BO Osuntokun and his staff at University College Hospital, Ibadan. These samples (usually in groups of 20 or more) will be shipped, in approved blood shipping containers, to Indianapolis by DHL. Each blood sample will be labeled with the name of the donor, date of birth, local ID number, and the collection date. Each blood sample will be accompanied by a data sheet (copy enclosed) that gives additional subject information, including dementia status. The collection of blood spots on filter paper for ApoE genotyping will also be evaluated. Blood samples from both sites will be delivered to the supervisor (Halter C) of the Alzheimer Disease Repository in Medical and Molecular Genetics. This department houses the national repositories for a number of hereditary diseases, including Huntington Disease and AD (Conneally PM, Director). The Indianapolis collection team will inform the AD Repository Supervisor when to expect the samples, but it may not be possible for the Ibadan team to do this. When the samples are received, the supervisor will verify them against the data sheets, assign unique ID numbers, and then log the samples into protected files in the MEGADATS database on a VAX 8800 mainframe computer, along with all the identification material on the data sheets and the blood tubes. Copies of the data sheets will be given to the Indiana Alzheimer Disease 1994 Continuing Review Center for their records. The Repository Supervisor will take the blood samples to the Molecular Genetics Services section of Medical and Molecular Genetics. This section, which houses the DNA and Lymphocyte Repositories and the DNA Diagnostics Laboratory, is located in a separate building on campus. The samples for extraction are given DNA numbers and those for immortalization are assigned transformation numbers. These numbers, printed as barcodes, are placed on the blood tubes and on all vessels and paperwork used during sample processing. To maintain patient confidentiality, names, dates of birth, or any other identifiers are not used at any stage of sample processing. DNA extractions, PCR amplifications, and other laboratory procedures are carried out in the DNA Diagnostics Laboratory under the supervision of A Sahota and ME Hodes. Lymphoblast transformations are carried out in the Lymphocyte Repository under the direction of D Fugman and JA Tischfield. Rigorous quality control procedures are in operation in both laboratories. The DNA Diagnostics Laboratory is accredited in Molecular Pathology by the College of American Pathologists and is registered under the Clinical Laboratories Improvement Amendments (CLIA '88). All blood samples are considered potentially infectious and proper safety precautions are observed during handling. These include wearing disposable gloves, eye protection, and laboratory attire. Disposable plasticware is soaked in sodium hypochlorite solution, placed in biological containment bags, autoclaved, and then delivered to a central disposal facility. All cell culture is done in vertical laminar flow hoods that are certified on a yearly basis. All employees are offered free hepatitis B vaccination. An Epstein-Barr virus antibody titre is determined before an employee works in the Lymphocyte Repository. The Molecular Genetics Services laboratories operate in compliance with the Indiana University Blood Borne Pathogens Exposure Control Plan, which is designed to meet all OSHA regulations for handling biohazardous materials.

DNA Extraction: Establishment of Cell Lines, ApoE Genotyping: (Because of space restriction the well established methodology for DNA extraction, and for establishment of cell lines and for ApoE genotyping are included in Appendix III).

Data analysis: The gel photographs will be examined independently by at least two technical personnel before a genotype is assigned. If there are any ambiguities, the genotyping will be repeated on a new PCR product. The technical staff will be working with DNA numbers only and will not have knowledge of the clinical status of the individuals being genotyped. The clinical status of the individuals will then be obtained from the AD Repository Supervisor. Allele frequencies will be calculated from the raw data for the dementia cases, cognitively impaired individuals, and controls from each of the two population groups. Genotype frequencies will be determined to detect any deviations from Hardy-Weinberg equilibrium. Logistic regression will be used to investigate the association between dementia and various ApoE genotypes. This will allow us to estimate the odds ratio while controlling for confounding factors such as age. The modelling approach also allows the investigation of the effect of gene dose, e.g. zero, one, or two copies of ApoE-ε4. Furthermore the hypothesized protective effect of ApoE-ε2 can be examined. These calculations will establish whether specific ApoE alleles and genotypes are associated with AD and other dementias in the two population groups. If a single ApoE genotype is found to be associated with AD in the two groups, but the disease prevalence is different in the two groups, this will suggest the contribution of other genetic and/or environmental risk factors in the pathogenesis of AD.

Autopsy The Neuropathology Laboratory has studied adult onset dementias, including AD. A productive collaboration between Drs. Ghetti, Hendrie and Farlow goes back over ten years. Most of the techniques in the proposed Specific Aim have been extensively used by the investigators. These studies will be funded through our Alzheimer's Disease Core Center. (Because of space limitation, the description of these techniques is included in Appendix III).

Statistical Considerations The statistical analyses are described in expected chronological order for the Primary Aims followed 1994 Continuing Review by the Secondary Aims.

Primary Aim Six: Since data collection of the first incidence study in the current grant period will not be complete until the end of the grant period, final data cleaning and analyses of the wealth of data are proposed for the first three years of the continuation until data collection is completed for the second incidence study. The findings will guide us to the best analyses for the second incidence study proposed in year three of the continuation. The primary task is to estimate incidence of dementia and AD. One cannot obtain a simple estimate because of the biased verification deliberately built into the study design. Recall that all individuals in the "poor performance" group in the screening test are clinically evaluated, but only 50% of the "borderline" and 5% of the "good performance" groups have clinical evaluation, which is treated as the gold standard. In the first incidence study only those individuals not diagnosed as demented or cognitively impaired at baseline will be given a repeat screening test. Based on the repeat screening test results, individuals will be clinically evaluated according to the same criteria as at baseline. In addition, individuals who have a "significant decline" in their performance on the screening test will also be clinically evaluated. Since we have as yet no empirical longitudinal data on the screening test, we will examine the distribution of change in scores based on the first 200 subjects in the first incidence study. Adjustments to the sampling proportions will be made as more data are cumulated. To estimate the two year incidence one may then calculate the properly weighted average of the various sampling strata (based on age, screening score, etc.). Because of the low incidence within each stratum, one can get more reliable estimates using a modelling procedure--logistic regression with inverse sampling fracture as weights in the computational procedure (Beckett et al, 1992). This procedure will be similar to the one we used for estimating prevalence (See Progress Report, "Prevalence" section). We have found the software package SUDANN to perform quite reliably. The incidence data may be more prone to biases than the prevalence. We have checked that each stratum at baseline, the subjects who have been clinically evaluated have similar demographic characteristics on performance on screening as those who have not been clinically evaluated, even though some individuals have refused or were too sick for the clinical evaluation. With an intervening two years, the ill health and frailty of some aging subjects may introduce more bias into the selection of subjects for clinical evaluation. These issues will be examined in depth. If, for example, refusal rate is related to baseline screening score the absence of clinical evaluation (gold standard) in the incidence study must be treated as non-ignorable missing data (Little and Rubin, 1987). One can then model the probability of missing value in a variable as a function of relevant covariates and incorporate this in the overall estimation procedure to obtain an unbiased estimate of the incidence. Results from these analyses may be used to modify the sampling design for clinical evaluations in the proposed second incidence study in order to obtain the most efficient estimates. Another set of analyses involve the risk factors. We will first examine the reliability of self-reports of various diseases by examining answers at two time points by the subject, correlation between reports by the subject and his/her informants and checking these against medical history obtained by nurses and doctors at clinical evaluations. Depending on findings from self-report, we can run additional laboratory tests in blood samples for corroboration of a risk factor. For example, if history of cardiovascular disease is positively associated with AD, we can compare the cholesterol level (or other biochemical values) between demented and non-demented individuals.

Primary Aim 1: To estimate the incidence of dementia and AD over a five year period, we will extend the method for estimating two year incidence as described in analyses under Aim Six. Individuals who are not demented at the two year follow-up will be eligible for the five year incidence study. The incidence of dementia and AD for year three to year six will then be estimated in a similar to the first two year incidence. The cumulative five year incidence of AD will be Prob (AD at second follow-up given no AD at first follow-up) x [1 - Prob(AD at first follow-up) given no AD at baseline]. The variance of the estimated five year incidence can be estimated from the variance of the two conditional probabilities using Greenwood's formula as in standard life-table analysis (Lawless, 1982). Sample size: Based on our own prevalence data and on the preliminary incidence rates in Cambridge, U.K., we can expect in five years, 25 new cases in the age 65-74 group, 40 in the age 75-84 1994 Continuing Review group and 30 in those age 85 and over with a standard deviation of ±4 assuming a Poisson distribution. With a total of 95 new cases, the incidence rate can be estimated with a standard error about 0.25 of the point estimate. Age-specific rates would be much less precisely estimated if only subgroups are used for such estimates. However using the modelling approach allows us to smooth the estimates, effectively "borrowing strength" from the neighboring subsamples. Thus the precision of age-specific rates is only slightly inflated from the overall rates.

Primary Aim 2: To describe and compare the decline in cognitive function among the cognitively impaired and dementia between Indianapolis and Ibadan. The subgroups of cognitively impaired and demented will have annual neuropsychological evaluation for five years. To model the decline in this subpopulation, a hierarchical model for longitudinal data will be built using an empirical Bayes approach (Hui and Berger, 1983). To facilitate the interpretation of the effects of various factors on such decline, each individual's overall score (or subscales in different analysis) will be reduced to the mean and slope, which are orthogonal. These summary measures will then be used as outcome measures in weighted regression analysis in which potential factors are included as predictors and the weights are the inverse of the individuals' sum of intra- and inter-subject variances. To investigate the difference in decline between Indianapolis and Ibadan, the study site is the major covariate of interest in the prediction of individual slopes. Other covariates of interest include baseline age, gender and baseline screening score. Because of the complexity of the analysis, an alternative approach using generalized estimation equations (Zeger et al, 1988) will also be used to explore the stability of the findings. Logistic regression will be used to investigate whether those individuals who were cognitively impaired at baseline progress to dementia in larger proportion in Indianapolis than in Ibadan. The progression at year three (and again the five year cumulative progression) will be the binary outcome in the logistic model while age, sex baseline screening score, etc., can be included as covariates in addition to the major variable of study site so any imbalance at baseline can be statistically controlled. Sample Size: There are 170 demented and cognitively impaired individuals in the Indianapolis community sample and 168 in Ibadan. These sample sizes give 80% power to detect 0.3 s.d. difference in mean between the two populations. Although we have no preliminary data on the variance of the change in scores, 0.3 s.d. is about the magnitude of differences of clinical importance. There will be missing data and drop-outs of these individuals. However, the statistical procedure can handle missing-at-random data. Should there be biased drop-outs, procedures for informative censoring will be used to correct for the bias. For comparing the proportion of "cognitively impaired" (105 in Indianapolis and 144 in Ibadan) who progress to "demented," a conservative power estimation based our difference in two proportions (without covariates) shows that thee is 80% power (at 5% significance level) to detect the following differences: Larger proportion Smaller proportion .20 .07 .40 .22 .60 .41 Inclusion of covariates to reduce randomness should allow us to detect smaller differences.

Primary Aim 3: Mortality: The primary goal is to investigate whether differential mortality rates exist between Indianapolis and between demented and non-demented individuals. Since we have ascertained the actual death dates, standard survival analysis procedures can be used. For example, Kaplan-Meyer curves can be used to compare graphically the survival of cognitively impaired individuals between Indianapolis and Ibadan while log-rank tests or similar statistics can be used to formally test the statistical significance of the observed difference. To control for confounding factors, covariates such as age and screening score will be included with study site in Cox regression to examine the difference in mortality. Sample Size: With 86 confirmed deaths in the Indianapolis sample after less than two years and 64 reported deaths in Ibadan, we anticipate >200 deaths in Indianapolis and >150 in Ibadan after five years. Ignoring time to death for a dichotomous predictor, it will have 80% power to detect odds ratios of two or more for a wide range of prevalences. Power will be much less in subgroup analyses. However, the logistic regression allows using the entire data set to investigate main effects and interactions between predictors. 1994 Continuing Review

Primary Aim 4: The statistical analyses are described in the "Methods" section on ApoE.

Primary Aim 5: Brain autopsies: When enough brain autopsies have been performed, analyses similar to those for CT scans will be performed. The measurements on neuropathology will be compared between Indianapolis and Ibadan for normal, cognitively impaired and demented subjects. To deal with this problem of multiple measurements, we will first group them into subsets of measurements on biological basis such as overall measures of plaques and tangles. Within each subset, we will try to obtain a single summary measure using techniques such as principal components. Multiple regression analyses will then be used to investigate differences in measurements between Indianapolis and Ibadan while controlling for age, sex, dementia, etc.

Analysis for Developmental Aims Developmental Aims 1 and 5: To explore risk factors for AD, new measurements will be developed to measure "education" beyond schooling and "social stimulation." Reliability and validity of such measures will be investigated in small pilot studies. Psychometric properties of these instruments will be examined, with input from experts from subject areas. The role of education in AD is particularly difficult to study because most cognitive testing is education-dependent. It is difficult to determine whether lack of education leads to AD or just makes the subjects perform less well on the tests. Have longitudinal measures, i.e. more than one test score over time, will allow us to examine the relationship between education and change in test scores with relatively less bias than cross-sectional associations.

Developmental Aim 3: One of our original goals is to develop a screening test instrument which can be refined to approximate the diagnosis of dementia and AD. Our first attempt will be to use the screening test's individual items or subscores in multiple logistic regression to predict dementia/alzheimer. In order for the screening instrument to perform satisfactorily as a surrogate diagnostic tool, we have set a requirement that its sensitivity and the specificity must be over 90%. To evaluate the performance of the fitted logistic regression model on the same data set from which the model was derived, we will use jackknife/bootstrapping procedures for model evaluation so that the performance will not be overestimated.

Developmental Aim 4: Objective definition of "cognitive impairment": As explained in the "Diagnostic Criteria" section of the "Methods" section the clinical definition of cognitively impairment that we have adopted needs to be refined so that more objective criteria can be applied by other research teams. Perhaps, the more important group of "cognitive impairment" are those subjects who progress into dementia. We will compare the distribution of the decline in neuropsychological test performance between those who were cognitively impaired and those who were normal at baseline. We will explore the relationship between the objective neuropsychological test scores and the overall evaluation by clinicians. The original group of "cognitive impairment" will then be subtyped by a set of objective criteria with the aid of clustering algorithms.

Data Management The database has been programmed using FoxPro. It is a relational database management system consisting of a set of tables for each of the study instruments, including among others the community screening surveys, the clinical assessments (initial and follow-up), the neuropsychological examinations, CT scans and blood laboratory results. The programmer works directly and closely with all investigators on the various components of the study. The database is used as an integral part of the overall tools to keep track of households contacted and outcome of visits (eligibility of occupants, refusals, etc.). The data is used in ongoing processes, e.g. monitoring progress of interviews, generating lists of subjects for clinical evaluations, interim contacts, repeat surveys, etc. In the Alzheimer Clinic, the database is used to help clinical staff complete all clinical measurements, schedule patients for follow-up and to remind physicians to keep up with the preliminary, local consensus and two-site consensus diagnoses. The data entry system for the screening survey and clinical evaluation has also been installed in Ibadan. About every month, a diskette of data is sent via either travelling investigators or DHL to 1994 Continuing Review Indianapolis. After editing checks, questionable data are resolved by the field interviewers and their supervisors in Ibadan and Indianapolis. The programmer then generates a list of patients selected according to the sampling scheme for clinical evaluation. The list is given to the supervisor (by FAX to Ibadan) for arranging individuals' clinical visits. 1994 Continuing Review HUMAN SUBJECTS

Human subjects will participate in the study in Indianapolis and Ibadan. (1) Detailed descriptions including numbers, age, sex and diagnostic groups of the populations to be included in the study appear in the Progress Report subheading "The Community Screening Phase" and "Clinical Assessments" and in the Methods section subheading "Populations." In Indianapolis all subjects by design are African American and in both sites 65% are women. Because the proposed study is longitudinal and focuses on dementia some demented subjects will be included. Informed consent for demented individuals is obtained from the legal guardian or family member. (2) Standard clinical laboratory tests and head CT scans will be carried out as needed for clinical diagnosis. In addition, patients who are evaluated in the research clinic, regardless of diagnosis will be invited to donate a blood sample (20ml) for genotyping research studies. Community dwelling participants will be invited to give blood droplet samples on filter paper for genotyping studies. (3) Informed consent for the community screening interview is obtained by field interviewers who explain the form to subjects and give them ample time to read it if they wish and answer possible questions. In Ibadan where 85% of the subjects are not literate the interviewers explain the consent form and subjects sign or place their mark on the consent form. Informed consent for the clinical assessment is administered by nurses in Indianapolis and physicians in Ibadan, the consent form is verbally explained, subjects are given ample time to read the form, and the subject is given the opportunity to ask questions before they sign. Consent for autopsy, a highly sensitive issue is described in the Progress Report under "Autopsy." The Institutional Review Boards in Indianapolis and Ibadan have approved the consent forms and procedures (Indianapolis IRB# 9011-13 approved 11/26/90, Ibadan IRB approval 08/13/90). (4) Potential risks are minimal. In the community screening phase (N=5,106) there is a brief cognitive assessment that is relatively non-threatening. The interview with the family member has also proven to be acceptable. The clinical assessment consists of standard clinical evaluations to identify dementia and determine diagnosis. If a new case of dementia is identified, follow-up services are offered at the Indiana University Outpatient Department, and University of Ibadan Hospital or local clinic. There would be great benefit to the patient and family if the patient is found to be suffering from a treatable disease. If the clinical assessment identifies new medical conditions in any study subject, appropriate follow-up will be carried out. In Indianapolis the primary care physician is consulted, if there is no primary care physician, follow-up will be provided through Wishard Hospital (C. Callahan). In Ibadan medical follow-up services will be provided through University of Ibadan Hospital. The risk of collecting blood for the tube or by pin prick and filter paper are pain, bruising and rarely infection. (5) Confidentiality will be protected by keeping patient data in confidential locked files and arranging the computerized files according to I.D. numbers rather than names. Results of the genotyping will be kept confidential and will not be revealed to study subjects. Publications and reports will use aggregate data and will not report individual names. Data for both sites are in secure files in project computers located in study offices. (6) The larger part of this study is of a documentary nature to identify environmental risk factors for AD, in this aspect of the study the risks are extremely small. Risks to subjects are minimal relating to collection of blood samples (bruising, pain, rarely infection). DNA will be isolated from the blood samples to examine the association between ApoE genotypes and AD. This association has been observed in the Caucasian population and in preliminary studies in these two populations but the proposed further investigations are needed to find out whether it applies to the African Americans and Nigerians. The aim of the present study is to determine the role of ApoE genotypes in the pathogenesis of AD in our community based study. Alzheimer;s disease represents a large problem for older people and the proposal study offers the possibility of increasing our understanding of the disease.