We conducted a retrospective chart review and database search for MMSE data on subjects with either FTD or AD. Subjects were evaluated at the University of Texas Southwestern Alzheimer’s Disease Center (UTSW) and the University of Southern California Alzheimer’s Disease Research Center (USC). All of the subjects gave consent for participation in research on initial enrollment in longitudinal study at the centers, and the protocols for collecting the data were approved by the IRB at each institution.
To answer our study question, we sought subjects with two or more MMSE scores with a minimum inter-test interval of at least 9 months. At USC, subjects responded to the MMSE annually (within 1 month of first visit anniversary); at UTSW, subjects are followed annually but sometimes have additional testing as part of a parallel research protocol, such that MMSE scores may be measured less than a year apart.
While subjects were not excluded on the basis of lower limits on the MMSE, there may have been cases where the subject stopped visiting the clinic due to nursing home placement or disability so severe that the caregiver could not present the subject for further evaluation at the AD center.
Additional inclusion criteria were: known onset age of dementia, known educational level, and diagnosis of either FTD (n = 44) by published consensus criteria [1
] or autopsy-proven AD (n = 45, either by Reagan/NIA [28
] or Braak and Braak criteria [29
]. Cases with AD were excluded if they also had a secondary neuropathological diagnosis, such as infarct. We further divided the FTD sample into two subgroups: those who presented initially with marked behavioral disturbance (FTD-B) vs. with primary progressive aphasia (FTD-L) [2
We excluded those subjects with significant visual impairment whose MMSE scores would have had to be adjusted for impossible tasks, as well as those subjects who could not speak English.
Variables reflecting the annualized rate of change were created using the first and last scores for each patient. Annualized change scores for total MMSE and for each subscore < 1 point (figure copy item excepted) were calculated by first determining the change score (last minus first) and then dividing by the time in years (number of days between visits divided by 365.25 to account for leap years) between these two tests. A minimum inter-test interval of 9 months maximized the available raw scores for analysis with an average of 2.83 data points (range 2-8) for each subject. The majority of subjects, 60.7%, in the total group had more than two MMSE scores and only 22.5% had more than three MMSE scores with inter-test time intervals exceeding 9 months. Individual subscores were available for MMSEs from 43 FTD and 40 AD subjects.
MMSE subscores were calculated by grouping various items of the MMSE by domain: orientation to time (0-5 points possible), orientation to place (0-5), registration (0-3), recall (0-3 points), attention/concentration (0-5 points), language (0-8 points), and figure copy representing constructional praxis (0 or 1 point).
Statistical Analyses: Two independent samples t tests (shown as means and ± standard errors in ) were performed to compare the two groups (FTD and AD) on demographic characteristics and initial clinical measures. Analysis of covariance (ANCOVA) was used to examine group differences on the initial total MMSE score. Covariates included in the model (if p< 0.05) were education, disease duration (from onset of illness to initial MMSE), and age at initial MMSE. For the dichotomous variable (gender), a two-sample proportions test was used to compare groups.
Demographics and MMSE scores of frontotemporal degeneration (FTD) and Alzheimer’s disease (AD) groups
Comparisons of the two groups on the annualized rate of change measures were performed using ANCOVA. Possible covariates included in each model were education, baseline measurement of the variable, and disease duration (time from onset to the last measure). Covariates were included in the final model if p< 0.05; if no covariates were found significant, t tests were performed. ANCOVA models also compared the FTD-B, FTD-L, and AD groups on all annualized rate of change measures.
To explore the hypothesized slower rate of decline on constructional praxis, groups of patients with perfect performance (score of 1) at the first testing on the figure copy/praxis item of the MMSE were compared with the results at the last testing (0 or 1) using a two independent samples proportion test. Random regression modeling [30
] was performed on the all of the consecutive raw score measurements of the MMSE variables, with duration of illness, level of education, onset age, and an indicator for diagnostic group membership as possible covariates. We report the estimates of the rate of change in groups for this exploratory comparison study, as opposed to having set a critical threshold of decline.
Two-tailed analyses were chosen as a more conservative basis for group comparisons and the p value for significance was 0.05. We performed the analyses using both Statistical Package for the Social Sciences (SPSS) software, version 13.0, and SAS version 9.13.