Incidence Rates of Dementia, Alzheimer’s disease and Vascular Dementia in the Japanese American Population in Seattle, WA: The Kame Project

Abstract

There are few dementia incidence studies in representative minority populations in the U.S., and no population-based studies of Japanese American women. We identified 3,045 individuals aged 65+ with at least 1 parent of Japanese descent living in King County, WA in 1992–4, of which 1,836 were dementia-free and were examined every two years (1994–2001) to identify incident cases of all dementias, Alzheimer’s disease (AD), vascular dementia (VaD) and other dementias. Cox regression was used to examine associations with age, sex, years of education and Apolipoprotein (APOE)-ε4. Among 173 incident cases of dementia, the overall rate was 14.4/1,000/yr, with rates being slightly higher among women (15.9/1,000) than men (12.5/1,000). Rates roughly doubled every 5 years for dementia and AD; the age trend for VaD and other dementias was less consistent. Sex was not significantly related to incidence of dementia or its subtypes in adjusted models. There was a trend for an inverse association with increasing years of education. APOE-ε4 was a strong risk factor for all dementias (HR=2.89, 95% CI 1.88–4.46), AD (HR=3.27, 95% CI 2.03–5.28) and VaD (HR=3.33, 95% CI 1.34–8.27). This study is the first to report population-based incidence rates for both Japanese American men and women.

INTRODUCTION

Marked geographic differences in the relative proportions of dementia subtypes were thought to exist in the 1980s–1990s between Asia and the West, with Alzheimer’s disease (AD) accounting for the majority of cases in the West and vascular dementia (VaD) being more common in Asia¹. The Ni-Hon-Sea Project was conducted in Hiroshima, Japan (The Adult Health Study, AHS); Honolulu, Hawaii (Honolulu-Asia Aging Study, HAAS) and Seattle, Washington (Kame Project) to examine prevalence and incidence rates and risk factors for dementia and its subtypes among Japanese populations in Japan and among immigrants and U.S.-born individuals, with the goal to discover whether rates of AD increase with migration to the West². Incidence rates for dementia and its subtypes from Japanese American populations have been published from the Honolulu-Asia Aging Study, which included only men³. The Kame Project is the first population-based study of dementia among community-dwelling Japanese American men and women.

METHODS

Study population

Identification and recruitment of the baseline population has been described previously¹. A study census was conducted of all Japanese Americans in King County, WA aged 55 and over, representing 90% of all Japanese Americans of this age group as identified by the 1990 US census (1). The study is named Kame (pronounced kah-meh) after the turtle, the Japanese symbol of longevity. Ninety-six percent of the population had 2 parents of Japanese origin; the remainder had one. Of 3,062 eligible residents, 3,045 were alive and invited and 1,985 (65.2%) agreed to participate in the baseline examination (May 1992–May 1994). Of these, 149 were found to have prevalent dementia (1); the remaining 1,836 were considered dementia-free.

Procedures

Baseline evaluation

Baseline evaluations were conducted from 1992–1994 in English or Japanese (21.6%). Highly-structured interviews were administered in-person by trained interviewers to gather risk factor data from non-demented participants.

We screened for dementia using the Cognitive Abilities Screening Instrument (CASI)⁴ over the follow-up period (May 1994–Dec 2001) consisting of four incidence waves: 1994–1996; 1996–1998; 1998–2000 and 2000–2001. Figure 1 shows the numbers of participants alive and those lost to follow-up at each successive wave. At each incidence wave, participants scoring <87/100 on the CASI were invited for proxy, neurologic and neuropsychological evaluations. A cut-point of 87 was selected to provide high sensitivity; at baseline, a CASI score of 86 had a sensitivity of 0.94 and a specificity of 0.94 for dementia⁴. Proxy interviews included the Clinical Dementia Rating Scale⁵, Blessed Dementia Rating Scale⁶ and Informant Questionnaire on Cognitive Decline in the Elderly⁷. Standard neuropsychological batteries were administered by a trained psychometrist and interpreted by a geriatric neuropsychologist (JU/SM), and clinical and neurologic examinations were performed by a neurologist (JB) or a geriatrician (WC). The Consortium to Establish a Registry for Alzheimer’s Disease neuropsychological battery was used⁸, excluding the MMSE, which is embedded in the CASI. Additional tests included: the Wechsler Memory Scale-Revised, Verbal Paired Associates (immediate and delayed recall)⁹, Trailmaking Test Parts A and B¹⁰, Digit Span and Digit Symbol subtests of the Wechsler Adult Intelligence Scale—Revised¹¹ and the CES-D¹². Participants suspected of having dementia had standard laboratory evaluations and some of these had non-contrast CT brain scans.

Figure 1.

Flowchart of Kame Project non-demented participants from baseline through four incidence waves, 1992–2001.

Dementia and subtype diagnoses were made by consensus committee (JB, WC, SM, JU, ARB, NZ), employing the Diagnostic and Statistical Manual IV (DSM-IV)¹³ blinded to CASI scores. The DSM-IV criterion of “impairment in social or occupational function… and decline from a previous level of functioning” was judged by changes in job performance, household responsibilities, hobbies, community and driving ability or personal activities. The NINCDS-ADRDA criteria¹⁴ were used to classify probable/possible AD and the NINDS-AIREN criteria¹⁵ for probable/possible VaD. Possible AD and possible VaD were not mutually exclusive, and some participants received both diagnoses. We report here incidence rates for all dementias (DSM-IV), probable and possible AD, probable and possible VaD, and other dementias.

Between 1994–1996, the cohort was invited to a blood draw to genotype the Apolipoprotein E gene¹⁶; 65.6% participated. Other variables used here (age, sex and years of education) were self-reported from the risk factor questionnaire.

All participants gave written informed consent and the study was approved by the University of Washington and University of South Florida IRBs.

Statistical Analysis

Incidence rates for dementia, AD and VaD were calculated using person-years. Crude rates were determined by dividing the number of new cases by the number of person-years at risk in 5-year age strata beginning at age 65 and ending with age 95+ (expressed per 1,000/year). Age was used as the time scale, such that for each non-demented participant, the number of person-years contributed was the difference between the age at study entry (left-truncation) and the age at last biennial examination. For demented participants, the age at which dementia occurred was the midpoint between the latest exam in which they were considered non-demented and the first exam at which they received a dementia diagnosis. 95% confidence intervals (CI) for incidence rates were derived assuming a Poisson distribution for the number of cases within each age stratum.

Multiple imputation techniques were used to adjust for participants who became demented at the first biennial visit but had missing values on dementia status at baseline due to not being sampled, or who were sampled but did not come in for the baseline diagnostic examination (n=68). The imputations were based on the probability of dementia given CASI and baseline age¹⁷. The imputations account for any differences in the numerators and denominators in the tables.

Age, sex, level of education (continuous and categorical, analyzed in this population as <8, 9–11, 12 and ≥13 years) and APOE genotype (presence or absence of an ε4 allele) were examined using Cox proportional hazards regression models with age as the time scale left-truncated at baseline age. Hazard Ratios (HR) are reported with 95% CI and P-values. All analyses including multiple imputation macros were conducted using SAS 9.3 (Cary, NC).

RESULTS

From the non-demented cohort (N=1,836), 454 screened below 87 on the CASI during the follow-up period and received full clinical evaluations; 173 of these met criteria for incident dementia. Figure 1 shows the number of incident cases at each wave and those lost to follow-up (1992–2001).

Table 1 depicts the total number of incident cases arising during the follow-up period (1994–2001). As participants could meet either probable or possible NINCDS/ADRDA and probable or possible NINDS-AIREN criteria (e.g., probable AD by NINCDS/ADRDA and possible VaD by NINCDS/AIREN, or possible AD and possible VaD), the numbers shown are not mutually exclusive. There were 17 other dementias, 1 due to vascular disease, 2 to Parkinson’s disease, 2 to head trauma, 5 due to other general medical conditions, 1 substance-induced dementia, and 4 due to multiple etiologies, and 2 due to unknown causes.

TABLE 1.

Number of Incident Cases in the Kame Project, King County, WA from May 1994–Dec 2001, by Outcome Type (not Mutually Exclusive).

Diagnosis	Number of Cases
All Dementias	173
Probable AD		79
Possible AD		56
(Probable + possible AD)			135
Probable VaD		32
Possible VaD		20
(Probable + Possible VaD)			52
Other (DSM-IV)		17
Vascular				1
General medical conditions				9
Parkinson’s					2
Head trauma					2
Other					5
Substance-induced				1
Multiple Etiologies				4
Head trauma					2
Other					2
Other, unknown				2

Table 2 shows baseline characteristics of the participants who did and did not become demented during follow-up. Women who became cases were older than men who became cases (p=0.008). Cohort participants who remained unaffected had more education than those who became cases (1.6 years more education for men and 0.8 years more education for women) (p<0.0001). About 20% of the cohort had 1 or 2 APOE-ε4 alleles, with 38.5% of incident cases and 18.8% of unaffecteds possessing an APOE-ε4 allele, with no differences by sex (p=0.36).

TABLE 2.

Baseline Characteristics of 173 Incident Dementia Participants and 1,663 Unaffected Participants Enrolled in the Kame Project, King County, WA, May 1992–Dec 2001.

	All Dementias			Unaffected			pa
	Men	Women	Total (%)	Men	Women	Total (%)
Number (%)	65 (37.6%)	108 (62.4%)	173	743 (44.7%)	920 (55.3%)	1,663
Mean Age (SD)	75.6 (7.0)	078.0 (7.4)	77.1 (7.4)	71.2 (4.8)	71.5 (5.8)	71.4 (5.4)	< 0.0001
Age group
65–69	12	10	22 (12.7%)	314	399	713 (42.9%)
70–74	22	32	54 (31.2%)	257	315	572 (34.4%)
75–79	19	26	45 (26.0%)	128	121	249 (15.0%)
80–84	7	18	25 (14.5%)	34	52	86 (5.2%)
85–89	1	14	15 (8.7%)	7	16	23 (1.4%)
90–94	2	4	6 (3.5%)	3	12	15 (0.9%)
95+	2	4	6 (3.5%)	0	5	5 (0.3%)	< 0.0001
Average years of education (SD)	12.0 (3.0)	11.6 (2.9)	11.7 (2.9)	13.6 (3.0)	12.4 (2.6)	13.0 (2.9)	< 0.0001
Education group
≤8 years	8	21	29 (16.8%)	36	73	109 (6.6%)
9–11 years	9	16	25 (14.5%)	66	129	195 (11.7%)
12 years	29	45	74 (42.8%)	278	437	715 (43.0%)
≥13 years	19	24	43 (24.9%)	363	276	639 (38.4%)	< 0.0001
APOE-ε4
1 or 2 alleles	14	21	35 (38.5%)	91	101	192 (18.8%)
0 alleles	31	25	56 (61.5%)	408	420	828 (81.2%)	< 0.0001

^ap-values for comparison between “All dementias” and “unaffected”.

We imputed incidence rates for 68 individuals who were sampled at baseline for the prevalence phase, but who did not come in for clinical evaluation until the first incidence wave. Of these 68, 18 were found to have incident dementia at the first biennial who had missing data from baseline. Five imputations were conducted, and based on these, 5, 7, 7, 3 and 5 of the 18 subjects were imputed to have had dementia at baseline. We used the first imputation in our incidence analyses to take these cases into account.

Age and sex-specific incidence density rates and 95% CIs are shown in Table 3 for all dementias, AD, VaD, and other dementias. AD was the most common form of incident dementia in this population. The overall incidence rate for all dementias was 14.4/1,000 (95% CI: 12.3, 16.4); for AD, 11.3 (95% CI: 9.4, 13.2) and for VaD, 4.4 (95% CI: 3.2, 5.6). For women, the rates were 15.9 (95% CI: 13.0, 18.8); 13.2 (95% CI: 10.5, 15.9), and 5.3 (95% CI: 3.6, 7.0) for all dementias, AD and VaD; for men, the corresponding rates were 12.5 (95% CI: 9.5, 15.5), 8.8 (95% CI: 6.3, 11.3), and 3.3 (95% CI: 1.8, 4.9), respectively.

TABLE 3.

Age- and Sex-Specific Incidence Density Rates per 1,000 Person-Years in the Kame Project, King County, WA, May 1992–Dec 2001 After Multiple Imputation to Account for 68 Subjects who were Sampled at Baseline but not Clinically Evaluated.

Age	All Dementias			Alzheimer’s Disease			Vascular Dementia			Other Dementias
Men	N	P-Y	IDR† (95% CI)	N	P-Y	IDR† (95% CI)	N	P-Y	IDR† (95% CI)	N	P-Y	IDR† (95% CI)
65–69	1	642	1.6 (0.0–4.6)	0	643		0	643		1	642	1.6 (0.0–4.6)
70–74	7	1822	3.8 (0.1–6.6)	3	1828	1.6 (0.0–3.5)	3	1824	1.6 (0.0–3.5)	3	1825	1.6 (0.0–0.35)
75–79	27	1628	16.8 (10.7–22.8)	20	1631	12.3 (7.1–17.5)	8	1643	4.7 (1.4–8.1)	4	1652	2.4 (0.1–4.8)
80–84	15	772	19.4 (9.9–28.9)	11	780	14.1 (6.0–22.2)	5	785	5.9 (0.4–1.4)	2	792	2.5 (0.0–6.0)
85–89	6	164	34.1 (6.2–62.0)	6	167	36.0 (8.5–63.5)	1	172	4.7 (0.0–16.2)	0	177
90–94	2	25	96.8 (0.0–224.7)	2	27	75.1 (0.0–171.9)	1	28	28.0 (0.0–96.2)	0	31
95+	3	18	168.9 (48.0, 289.7)	3	18	168.9 (48.0, 289.7)	0	21		0	21
Overall	61	5,071	12.5 (9.5, 15.5)	45	5,094	8.8 (6.3, 11.3)	18	5,116	3.3 (1.8, 4.9)	10	5,139	1.9 (0.7–3.1)
Women
65–69	0	810		0	811		0	811		0	811
70–74	14	2363	5.9 (2.9–8.9)	10	2371	4.2 (1.7–6.8)	4	2373	1.9 (0.1–3.6)	2	2377	0.8 (0.0–2.0)
75–79	26	2004	13.1 (8.2–18.0)	23	2015	11.4 (6.9–15.9)	11	2030	5.3 (2.2–8.5)	1	2047	0.5 (0.0–1.4)
80–84	27	898	30.1 (19.3–40.9)	22	910	24.6 (14.7–34.5)	10	926	10.8 (4.2–17.4)	2	940	2.1 (0.0–5.1)
85–89	19	353	53.3 (31.0–75.5)	18	358	49.7 (28.5–70.9)	7	368	19.0 (5.4–32.6)	1	381	2.6 (0.0–7.7)
90–94	9	101	89.1 (37.7, 140.6)	7	104	67.6 (21.9, 113.4)	1	113	8.9 (0.0, 26.0)	1	118	8.4 (0.0, 24.8)
95+	8	38	212.9 (105.5, 320.2)	7	40	176.5 (72.9, 280.0)	2	51	39.6 (0.0, 92.2)	0	54
Overall	103	6,567	15.9 (13.0, 18.8)	87	6.609	13.2 (10.5, 15.9)	35	6,672	5.3 (3.6, 7.0)	7	6727	1.0 (0.3–1.8)
Total
65–69	1	1,452	0.69 (0–2.0)	0	1,454	0	0	1,454	0	1	1452	0.7 (0–2.0)
70–74	21	4,185	5.0 (2.9–7.1)	13	4,199	3.1 (1.4–4.7)	7	4,197	1.8 (0.5–3.1)	5	4201	1.2 (0.2–2.2)
75–79	53	3,632	14.7 (10.9–18.5)	43	3,646	11.8 (8.4–15.2)	19	3,673	5.1 (2.8–7.4)	5	3699	1.4 (0.2–2.5)
80–84	42	1,670	25.2 (17.9–32.4)	33	1,690	19.8 (13.2–26.3)	15	1,711	8.5 (4.2–12.9)	4	1732	2.3 (0.1–4.6)
85–89	25	517	47.2 (29.5–64.9)	24	525	45.4 (28.3–62.4)	8	540	14.5 (4.4–24.5)	1	558	1.8 (0–5.3)
90–94	11	126	90.8 (42.1–139.6)	9	131	69.2 (27.7–110.6)	2	141	12.8 (0.0–32.3)	1	149	6.7 (0–19.7)
95+	11	56	198.7 (115.7–281.8)	10	58	174.1 (91.9–256.3)	2	72	28.0 (0.0–65.5)	0	75
Overall	164	11,638	14.4 (12.3, 16.5)	132	11,703	11.3 (9.4, 13.2)	53	11,788	4.4 (3.2, 5.6)	17	11,866	1.4 (0.8–2.1)

The incidence rates for all dementia and AD increased with age in both sexes, approximately doubling every 5 years. A less consistent age trend was evident for VaD. Women had higher incidence rates than men for all dementia and AD in most age groups. Women aged 80–89 had higher rates of vascular dementia compared with men. Rates for ‘other dementias’ were based on very small sample sizes.

When Cox models were constructed, neither sex nor education was associated with any of the outcomes, although there was a trend for education to be inversely associated with dementia and AD (all dementia HR=0.94, 95% CI 0.87–1.02; AD HR=0.92, 95% CI 0.84–1.01; VaD HR=1.07, 95% CI 0.92–1.25). The presence of one or two APOE-ε4 alleles was the sole statistically significant predictor of all dementias (HR= 2.89, 95% CI: 1.88, 4.46), AD (HR=3.27, 95% CI: 2.03, 5.28), and VaD (HR=3.33, 95% CI: 1.34, 8.27), adjusting for sex and years of education.

DISCUSSION

In this population-based study of Japanese Americans living in King County, WA, the overall incidence rate for all dementias was 14.4/1,000 (95% CI: 12.3, 16.5) in 11,638 person-years, with women having an incidence rate of 15.9/1,000 (95% CI: 13.0, 18.8) and men 12.5/1,000 (95% CI: 9.5, 15.5). The overall incidence rate for AD was 11.3/1,000 (13.2/1,000 for women and 8.8/1,000 for men), and for VaD, 4.4/1,000 (5.3/1,000 for women and 3.3/1,000 for men). This is the first time that rates for Japanese Americans can be compared by sex within one population, the only other study of Japanese Americans being the Honolulu-Asia Aging Study, which was conducted among only men.

Comparison with other large population-based studies

Age and sex-specific incidence rates for dementia in the Kame Project were generally close to those found in other studies. Our rates were higher in the oldest old (85+) than those in the Rotterdam Study¹⁸, but the small number of cases in this age stratum may obscure true differences. Women aged 85+ in Kame had higher rates than those from the Framingham Study¹⁹, but among older men, rates were similar. Incidence rates from France (PAQUID)²⁰ were similar to those in Kame for all dementias and AD, except at ages 70–74 and 80–84 where PAQUID’s rates were somewhat higher. Differences in cognitive screening tests, sampling procedures, case definitions and other methodologic applications may hamper the comparison of rates between studies. Our rates were consistent with a meta-analysis of 23 studies²¹. An unexpected finding in our study was the apparent higher rate of vascular dementia seen in women: 5.3 (95% CI: 3.6, 7.0) vs. 3.3 (95% CI: 1.8, 4.9) in men. In Hiroshima²², comparable rates averaged out to be about half that seen in Kame women.

We could not conduct a meta-analysis comparing incidence rates from Kame with those from Hiroshima²², Honolulu³ and ACT²³ in Seattle, WA due to minor differences in methodologic applications at the four sites, such as using different CASI cut-points and using different clinical criteria in the published incidence papers. Descriptively, age-specific rates for AD were similar in the Kame Project and Hiroshima²² and in Caucasians studied in ACT²³. For men, HAAS reported lower rates for AD than the other studies³.

Incidence rates and sex

Whether women are at higher age-specific risk of dementia/AD than men is still not resolved. In this study, no differences were found. One meta-analysis²⁴ found that women are about 50% more likely to develop AD, but no difference by sex was found when the outcome was overall incidence of dementia. Jorm²¹ reported that while the difference in risk by sex is small, women aged 85+ tend to have higher incidence rates than men. In a study focusing on people aged 90+²⁵, there were no differences by sex. The AHS²² and ACT²³ also found no differences by sex.

Incidence rates and education

Education has been shown in many studies to be associated with incident dementia and AD²⁶ but about half show no association, and in Kame, education was not statistically significant in the fully-adjusted models. However, in models not adjusted for APOE-ε4, education was inversely associated with the risk of dementia and AD but not VaD. The fact that education lost statistical significance when APOE-ε4 was added to the model may be related to lower statistical power due to reduced sample size (not all participants donated blood). Consistent with this view, the point estimate did not change for AD and increased only slightly for all dementias. Education can be considered in the context of other early-life exposures²⁷, such as perinatal conditions, early-life brain development, body growth and socioeconomic conditions, environmental enrichment and neural reserve as one of many variables that may preserve brain function. Sharp and Gatz²⁶ extended this model to variables that occur during adulthood such as healthy lifestyle and cognitive stimulation and their associations with reduced risk for dementia in late-life. In addition, several studies that consider both intelligence (IQ) and education simultaneously show that including IQ in the predictive model nullifies the significant association with education^27,28. In representative autopsy studies, risk factors for disease expression such as education and head circumference are not related to neuropathology^29,30. In different populations, variables related to achieved education in childhood and/or adulthood may be surrogates for years of formal education. In the Kame Project for example, verbal IQ from the National Adult Reading Test-Revised was correlated with education 0.42 (p=0.0001)³¹. The direct measurement of years of education may have different meaning in different cultures and cohorts²⁶.

Incidence rates and APOE

The frequency of APOE-ε4 in Kame is similar to that of other Japanese populations³². Table 4 displays the genotypes and isoform frequencies for APOE in Kame. Kame genotypes were just within Hardy-Weinberg equilibrium (P =0.053).

TABLE 4.

Apolipoprotein E Genotypes in the Kame Project (Men and Women), King County, WA, May 1994–May 1996.

Genotype	Kame Project
	Number	Percent
ε2ε2	6	0.50
ε2ε3	107	8.9
ε3ε3	849	70.5
ε2ε4	8	0.66
ε3ε4	223	18.5
ε4ε4	12	1.0
Frequency
ε2	0.053
ε3	0.841
ε4	0.106

The presence of one or more ε4 alleles strongly predicted vascular dementia as well as AD, and with similar magnitude (HR=3.33, 95% CI: 1.34, 8.27 for VaD and 3.27, 95% CI: 2.03, 5.28 for AD). Studies of the association between APOE-ε4 and VaD are equivocal. A meta-analysis of Chinese Han showed a pooled Odds Ratio (OR) of 2.07 (95% CI: 1.69, 2.53) for VaD³³. However, another study of older Japanese conducted in Hisayama, Japan found no association between ε4 and VaD (RR=1.08, p=0.86)³⁴. In HAAS, no association was found between APOE-ε4 and incident VaD (RR=1.27, 95% CI: 0.33, 4.83)³. Among Caucasian populations, the Canadian Study of Health and Aging found an association between ε4 and incident VaD (OR=3.13, 95% CI: 1.76, 5.55)³⁵ and the Cache County study³⁶ also reported a significant HR for 2 alleles (HR=4.4, 95% CI: 1.6, 12.5), with the HR for 1 allele approaching significance (p=0.08). In the Rotterdam Study³⁷, there was also a strong tendency for ε4 to be associated with clinically diagnosed VaD [(ε2ε4: HR=3.2 (95% CI: 1.0, 10.8); ε3ε4: HR=1.8 (95% CI: 0.9, 3.4) and ε4ε4: HR=8.6 (95% CI: 2.9, 25.2)]. We speculate that since the clinical diagnosis of VaD in our study included possible as well as probable VaD, VaD patients could have comorbid AD, strengthening the association with APOE-ε4.

Strengths and limitations

The Kame Project is representative of the Japanese American population of King County aged 65 and over and is the first population-based study of dementia to include Japanese American women. Our study had a high cut-point on the CASI for screening for incident dementia (<87/100), so it is likely that most of the incident cases that occurred during the follow-up period were captured. We asked the cohort for blood during the first incidence wave, and 39.5% refused. Participants in the study who refused blood donation and were missing APOE genotypes (n=725) were older (mean= 72.9 years) (p<.001), more likely to be women (p=<0.001), and somewhat less educated (mean=12.3 years) (p<0.001) than those who agreed. Therefore, those with complete APOE data in our study may have fewer risk factors for dementia, and analyses using APOE may be biased toward the null. A limitation is that we have no autopsy results and the associations with risk factors depend on the accuracy of the clinical diagnosis of dementia subtypes. The likelihood of misclassification is minimized by following strict, clinical diagnostic criteria that are standardly used in large epidemiologic studies and adjudicating all cases by consensus committee and monitoring dementia cases annually for disease progression. Further analyses will delineate Asian vs. Western dietary and cultural patterns that may be associated differentially with incident dementia in this cohort; we have previously published, among others, associations examining acculturation and early-life exposures^38–41, estrogen and dietary soy isoflavones⁴², alcohol consumption⁴³ and consumption of fruit and vegetable juices⁴⁴.