Volume 5, Number 5, October 2003

Articles from "Statistical Methodology in Alzheimer's Disease Research II", May 2002

Pages 349-355
Julia L. Bienias, Laurel A. Beckett, David A. Bennett, Robert S. Wilson, Denis A. Evans
Design of the Chicago Health and Aging Project (CHAP)
Abstract: The design of the Chicago Health and Aging Project (CHAP) is described. CHAP is a longitudinal population study of common chronic health problems of older persons, especially of risk factors for incident Alzheimer’s disease, in a biracial neighborhood of the south side of Chicago. Special attention is given to three unusual design features of the study. One feature is that clinical evaluation for Alzheimer’s disease is confined to a stratified random sample of all participants. This feature results in substantial cost savings and substantially less bias than screening approaches but has the disadvantages of adding analytic complexity and requiring the use of indirect means to identify a disease-free cohort for the development of incident Alzheimer’s disease. The second unusual feature is efficiently combining in analyses the successive independent multiple samples that are drawn, one from each data collection cycle. The third unusual feature is entering successive age cohorts of community residents into the study as they attain 65 years of age. This has the advantages of enhancing direct investigation of the effect of age on the action of risk factors for Alzheimer’s disease and direct examination of cohort effects. The interaction of these features is described, especially as they pertain to a study in which data are collected in successive waves. The results from these waves must be combined for effective analysis of the relation among risk factors and incident disease.

Pages 357-365
Danielle J. Harvey, Laurel A. Beckett, Dan M. Mungas
Multivariate Modeling of Two Associated Cognitive Outcomes in a Longitudinal Study

Abstract: Longitudinal studies of Alzheimer’s disease provide information about cognitive decline and predictors of this decline. However, overall cognitive function is comprised of many underlying processes, each of which may respond differently over time and may be affected by different predictors. In addition to studying how these processes decline independently, one might also be interested in how the processes decline together. Multivariate growth models, an extension and modification of random effects models, provide a means of dealing with these issues and enable assessing the association between the processes of interest. This technique allows for separate random effects and predictors for each process in the same model, thereby providing simultaneous estimates of the model parameters and variability for each process. We can then determine if factors associated with decline in one process are also associated with decline in another process and the extent to which the processes differ. We provide data that include information on two underlying processes of cognitive function, namely memory and executive function, to illustrate this methodology.

Pages 367-373
Larry F. Hughes, Kyle Perkins, Benjamin D. Wright, Heather Westrick
Using a Rasch Scale to Characterize the Clinical Features of Patients with a Clinical Diagnosis of Uncertain, Probable, or Possible Alzheimer Disease at Intake

Abstract: Objective: This study examined the clinical features of patients with clinical diagnoses of probable Alzheimer disease (AD), possible AD, and uncertain. Design: Case study comparing three groups of AD patients diagnosed at their initial visit to an Alzheimer outpatient clinic. Setting: Southern Illinois University School of Medicine’s Center for Alzheimer Disease and Related Disorders (CADRD) assessment sites (20) in rural Illinois. Participants: 300 patients assessed at CADRD between January 1, 1994 and July 1, 2000. Measurements: Patients were given an extensive clinical battery consisting of physical and neurologic examination, mental status testing including the Mini-Mental State Exam (MMSE), Short Blessed Dementia (SBD) and Blessed Dementia Scale (ADL), medical history evaluation, and laboratory tests. Other data included age at visit, gender, and medical history variables. Results: Mean MMSE, SBD, and ADL scores differed significantly between groups (p’s<.01). In all three cognitive tests, the uncertain group was the least impaired while the probable AD group was the most impaired. A Rasch model indicated that only the cognitive measures were useful in discriminating between the three diagnostic groups. Conclusion: In general, probable AD patients were distinguished from possible AD patients by the severity of their dementia as measured by the MMSE, ADL and SBD as well as Hachinski-Ischemic Score (HIS) scores. A Rasch model did well at predicting group membership based upon dementia measures only. The uncertain group differed from the AD groups in age and dementia severity as measured by the MMSE, ADL and SBD. Noting differences between this and previous studies, we speculate disparity may be related to differences in population ethnicity.

Pages 375-382
Ming Ji, Chengjie Xiong, Michael Grundman
Hypothesis Testing of A Change Point During Cognitive Decline Among Alzheimer’s Disease Patients

Abstract: In this paper, we present a statistical hypothesis test for detecting a change point over the course of cognitive decline among Alzheimer’s disease patients. The model under the null hypothesis assumes a constant rate of cognitive decline over time and the model under the alternative hypothesis is a general bilinear model with an unknown change point. When the change point is unknown, however, the null distribution of the test statistics is not analytically tractable and has to be simulated by parametric bootstrap. When the alternative hypothesis that a change point exists is accepted, we propose an estimate of its location based on the Akaike’s Information Criterion. We applied our method to a data set from the Neuropsychological Database Initiative by implementing our hypothesis testing method to analyze Mini Mental Status Exam scores based on a random-slope and random-intercept model with a bilinear fixed effect. Our result shows that despite large amount of missing data, accelerated decline did occur for MMSE among AD patients. Our finding supports the clinical belief of the existence of a change point during cognitive decline among AD patients and suggests the use of change point models for the longitudinal modeling of cognitive decline in AD research.

Pages 383-390
Kathleen A. Lane, Sujuan Gao, Siu L. Hui, Jill R. Murrell, Kathleen S. Hall, Hugh C. Hendrie
Apolipoprotein E and Mortality in African-Americans and Yoruba
Abstract: The literature on the association between apolipoprotein E (ApoE) and mortality across ethnic and age groups has been inconsistent. No studies have looked at this association in developing countries. We used data from the Indianapolis-Ibadan Dementia study to examine this association between APOE and mortality in 354 African-Americans from Indianapolis and 968 Yoruba from Ibadan, Nigeria. Participants were followed up to 9.5 years for Indianapolis and 8.7 years for Ibadan. Subjects from both sites were divided into 2 groups based upon age at baseline. A Cox proportional hazards regression model adjusting for age at baseline, education, hypertension, smoking history and gender in addition to time-dependent covariates of cancer, diabetes, heart disease, stroke, and dementia was fit for each cohort and age group. Having ApoE e4 alleles significantly increased mortality risk in Indianapolis subjects under age 75 ( hazard ratio: 2.00; 95% CI: 1.19 – 3.35; p=0.0089). No association was found in Indianapolis subjects 75 and older (hazard ratio: 0.71; 95% CI: 0.45 - 1.10; p=0.1238), Ibadan subjects under 75 (hazard ratio: 1.04; 95% CI: 0.78 to 1.40; p=0.7782), or Ibadan subjects over 75 (hazard ratio: 1.21; 95% CI: 0.83 to 1.75; p=0.3274).

Pages 391-398
Marta S. Mendiondo, J. Wesson Ashford, Richard J. Kryscio, Frederick A. Schmitt
Designing a Brief Alzheimer Screen (BAS)
Abstract: Context: With advances in the treatment of Alzheimer’s disease (AD), clinical focus has shifted to early patient identification. Memory recall tests and category fluency distinguish normal individuals from early AD patients. Objective: Develop a brief test for general practitioners to screen for AD. Design: Examination of items from the MMSE and category fluency. Setting and Participants: A Brief Alzheimer Screen (BAS) was developed from cognitive assessments on 406 normal subjects and 342 mild AD patients CERAD (Consortium to Establish a Registry for AD) dataset. The derived measure was then applied to a second validation sample. Main Outcome Measure: Logistic regression was used to derive a predictive equation, which was then applied to two validation samples to estimate sensitivity and specificity. Results: The resulting logistic model for discriminating between mild AD and controls included: recall of 3 words, number of animals named in 30 seconds, date, and spelling of WORLD backwards, (p<0.001 for each) accounting for 77% of the variance. When applied to the validation samples, sensitivity and specificity were over 99% and 87%, respectively. Conclusions: These data support the use of the BAS as a potential screen of patients over 60 years of age.

Pages 399-407
Jianzhao Shen, Paul Crane, Sujuan Gao
A Latent Variable Model Approach for Assembling and Scoring Screening Tests for Dementia
Abstract: In dementia studies, the diagnosis of dementia often relies on results of screening tests aimed at measuring various dimensions of cognitive functions. The current practice of scoring a screening test involves simply summing the correct responses from each item. However, this method may be imprecise and inefficient in the predictive power of the score for dementia. We propose a latent variable model approach for the scoring and item selection of such tests. We model the item responses to be random variables based on latent variables. We also model the disease outcomes to be a function of the latent variables. Maximum likelihood estimates are obtained by maximizing the joint likelihood functions of disease and the item responses over a specified distribution function for the latent variables. Variances of model parameters are estimated using a nonparametric bootstrap method. We illustrate the approach using a screening test for dementia from a community-based study.

Pages 409-418
Chengjie Xiong, J. Philip Miller, John C. Morris
Testing Correlation of Cognitive Decline at Adjacent Stages of Dementia
Abstract: This paper studies the correlation of cognitive progression for subjects whose dementia has made a transition from a milder stage of severity to the next stage of impairment. We model the progression of cognitive decline at adjacent stages of dementia by using a general linear mixed model. We also propose a three-step procedure to detect the best configuration of the covariance matrices for the random components in the model. After the best configuration of covariance matrices for the random components in the model is determined, we then recommend another two-step process to test whether there exists a significant correlation between the rate of cognitive decline before the transition, the cognitive status at the transition time, and the rate of cognitive decline after the transition. In addition, we present asymptotic confidence interval estimates for the correlations associated with a transition. This method is applied to several composite psychometric factor scores in the longitudinal database from the Alzheimer’s Disease Research Center (ADRC) at Washington University in St. Louis.

RETURN TO INDEX

top