Vascular health, diabetes, APOE and dementia: the Aging, Demographics, and Memory Study

The ADAMS sample design and selection procedures are summarized below and described in detail elsewhere [3]. The ADAMS sample was drawn from the larger Health and Retirement Study (HRS), an ongoing prospective cohort study of individuals that was designed to investigate the social, health and economic implications of aging [26]. The HRS began in 1992 and was expanded in 1998 to become a representative sample of all persons in the contiguous USA older than 50 years of age. Data were collected from community living and institutionalized adults primarily through telephone interviews, with an overall response rate of 81% [27]. The ADAMS sample was derived from a stratified random subsample of 1,770 individuals aged 70 years or older from five cognitive strata based on participants' cognitive test scores or proxy-reported cognition from the most recent HRS interview (in 2000 or 2002). The three highest cognitive strata were further stratified by age (70 to 79 years vs. 80 years and older) and sex to ensure adequate numbers in each subgroup. A detailed description of the ADAMS sample, non-participants and the population weights used to ensure population representativeness has been published previously [3]. ADAMS participants did not differ significantly from nonparticipants in age, sex, or education, but were more likely to be African-American (P < 0.05), to have been a self-respondent at the prior HRS wave (P < 0.05), and to have scored in the normal range on the HRS cognitive test in the prior wave (P < 0.05) [3]. ADAMS initial assessments were conducted between July 2001 and December 2003, after a mean period of 13.3 months (standard deviation (SD) 6.9) from the last HRS interview. Participants were aged 71 years or older at the initial ADAMS assessment. A total of 856 individuals (56% of the nondeceased target sample) participated in all phases of the dementia assessment (Figure 1) [3].

The institutional review boards at Duke University Medical Center and the University of Michigan approved the study procedures, and all study participants or their surrogates provided informed consent.

Full details of the ADAMS assessment and diagnostic procedures are described elsewhere [3]. A nurse and a neuropsychology technician conducted an assessment of all participants at their residence. The following information about the participant was collected from a knowledgeable informant: chronological history of cognitive symptoms, current neuropsychiatric symptoms, measures of severity of cognitive and functional impairment, and family history of memory problems. During the assessment, the participant completed a battery of neuropsychological measures, a measure of depressive symptoms, a standardized neurological examination, blood pressure measurements, collection of buccal DNA samples for APOE genotyping (presence of the ε4 allele), and a seven-minute videotaped segment covering portions of the cognitive status and neurological examinations. Medical record releases were also sought to obtain relevant neuroimaging and laboratory results from participants' physicians. A consensus expert panel of neurologists, neuropsychologists, geriatric psychiatrists, and internists reviewed all information for all participants collected during the in-home assessment first without and then with medical records and assigned final diagnoses. The consensus panel used clinical judgment to assign the final diagnosis based on currently accepted standardized international criteria for dementia (Diagnostic and Statistical Manual of Mental Disorders (DSM)-III-R and DSM-IV) [28, 29], AD (National Institute of Neurological and Communicative Disorders and Stroke - Alzheimer's Disease and Related Disorders Association (NINCDS-ADRDA)) [30], and VaD (National Institute of Neurological Disorders and Stroke - Association Internationale pour la Recherché et l'Enseignement en Neurosciences (NINDS-AIREN)) [31]. The consensus panel also made an estimate of the year of dementia onset based on the available cognitive and additional data. Non-demented participants were also further divided into two categories: normal cognitive function and cognitive impairment without dementia (defined as mild cognitive or functional impairment that did not meet the criteria for dementia, or performance that was both below expectation and at least 1.5 SDs below published norms on any neuropsychological test) [17]. It should be noted that the distribution of diagnoses becomes nationally representative only after population weights are applied [3].

The presence of hypertension, diabetes, heart disease, and stroke was established based upon self- or proxy-report of these conditions at the HRS interview prior to the ADAMS assessment. During the HRS interview, each respondent was asked whether a doctor had ever told them that they had hypertension ('high blood pressure or hypertension?'), diabetes ('diabetes or high blood sugar?'), heart disease ('heart attack, coronary heart disease, angina, congestive heart failure, or other heart problems?'), or a stroke ('stroke?'). In addition to self-report of hypertension, blood pressure was directly measured during the ADAMS assessment and individuals who had systolic blood pressure of 140 mmhg or above, or diastolic blood pressure of 90 mmhg or above on two successive measurements were categorized as hypertensive. Linked Medicare claims data from 1991 were also used to confirm presence of stroke. The ADAMS subject or informant was asked to provide a list of all current medications used, and to provide the bottles for each of these medications for the nurse to confirm usage. We used these medication data to assign a 'treated' or 'untreated' status based on the current use of one or more relevant medication for each condition (hypertension, diabetes, heart disease, and stroke).

The probability that an individual participated in ADAMS was modelled using logistic regression as a function of covariates derived from the HRS, such as age, sex, education, marital status, HRS cognition scores, residential status, and health conditions, to develop nonresponse adjustments to the sample selection weights. Population sample weights were then constructed to take into account the probabilities of selection in the stratified sample design and to adjust for differential nonparticipation [3]. Multivariable logistic regression models were then used to determine the relation of vascular health (hypertension, heart disease and stroke), diabetes and APOE to all-cause dementia and the common dementia subtypes (AD and VaD). We adjusted for age, sex, education, ethnicity, and HRS wave in basic models, and also corrected for all of the other variables in fully adjusted models. Adjustment for stroke was not incorporated in the fully adjusted VaD models and we did not examine the association between stroke and VaD because cerebrovascular disease is a central diagnostic criterion for VaD [31]. In a series of secondary analyses we examined the associations between vascular conditions, diabetes and dementia separately by treatment status and examined interactions with APOE genotype. In a sensitivity analysis we also excluded those with cognitive impairment without dementia from the non-demented control group. In a second sensitivity analysis we restricted conditions to those reported in the HRS before the estimated onset of dementia in order to help establish the temporal relation of observed associations. All analyses were conducted using Stata SE version 9.2 (StataCorp, College Station, TX, USA).