[Skip to Content]
Access to paid content on this site is currently suspended due to excessive activity being detected from your IP address 54.205.87.3. Please contact the publisher to request reinstatement.
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
[Skip to Content Landing]
Download PDF
Kaplan-Meier curves showing the likelihood of reaching the end point of clinically significant decline, defined as a 5-point or greater drop on the Mini-Mental State Examination (MMSE) score, by group (based on estimated rate of progression prior to visit 1). Group 1 had calculated preprogression rates of less than 2 points per year; group 2, 2 to 4.9 points per year; and group 3, 5 or more points per year.

Kaplan-Meier curves showing the likelihood of reaching the end point of clinically significant decline, defined as a 5-point or greater drop on the Mini-Mental State Examination (MMSE) score, by group (based on estimated rate of progression prior to visit 1). Group 1 had calculated preprogression rates of less than 2 points per year; group 2, 2 to 4.9 points per year; and group 3, 5 or more points per year.

Table 1. 
Progression Characteristics of Patients With Alzheimer Disease*(N = 298)
Progression Characteristics of Patients With Alzheimer Disease*(N = 298)
Table 2. 
Pearson Correlations*
Pearson Correlations*
Table 3. 
Comparison of Patients With Slow, Intermediate, and Rapid Progression of AD by Preprogression Rate in Lost MMSE Points per Year*
Comparison of Patients With Slow, Intermediate, and Rapid Progression of AD by Preprogression Rate in Lost MMSE Points per Year*
1.
Jacobs  DSano  MMarder  K  et al Age at onset of Alzheimer's disease: relation to pattern of cognitive dysfunction and rate of decline. Neurology.1994;44:1215-1220.
2.
Chui  HLyness  SSobel  ESchneider  L Extrapyramidal signs and psychiatric symptoms predict faster cognitive decline in Alzheimer disease. Arch Neurol.1994;51:676-681.
3.
Stern  YTang  MAlbert  M  et al Predicting time to nursing home care and death in individuals with Alzheimer disease. JAMA.1997;277:806-812.
4.
Kraemer  HTinklenberg  JYesavage  J "How far" vs "how fast" in Alzheimer disease. Arch Neurol.1994;51:275-279.
5.
Salmon  DThal  LButters  NHeindel  W Longitudinal evaluation of dementia of the Alzheimer type: a comparison of three standardized mental status examinations. Neurology.1990;40:1225-1230.
6.
Morris  JEdland  SClark  C  et al The Consortium to Establish a Registry for Alzheimer's Disease (CERAD), IV: rates of cognitive change in the longitudinal assessment of probable Alzheimer's disease. Neurology.1993;43:2457-2465.
7.
Haxby  JRaffaele  KGillette  J  et al Individual trajectories of cognitive decline in patients with dementia of the Alzheimer type. J Clin Exp Neuropsychol.1992;14:575-592.
8.
Stern  RMohs  RDavidson  M  et al A longitudinal study of Alzheimer's disease: measurement, rate, and predictors of cognitive deterioration. Am J Psychiatry.1994;151:390-396.
9.
Brooks  JYesavage  J Identification of fast and slow decliners in Alzheimer disease: a different approach. Alzheimer Dis Assoc Disord.1995;9(suppl):S19-S25.
10.
Braak  HBraak  E Neuropathological staging of Alzheimer-related changes. Acta Neuropathol.1991;82:239-259.
11.
Bracco  LGallato  RGrigoletto  F  et al Factors affecting course and survival in Alzheimer's disease: a 9-year longitudinal study. Arch Neurol.1994;51:1213-1219.
12.
McKhann  GDrachman  DFolstein  MKatzmann  RPrice  DStadlan  EM Clinical diagnosis of Alzheimer's disease: report of the NINCDS-ADRDA Work Group under the auspices of the Department of Health and Human Services Task Force on Alzheimer's Disease. Neurology.1984;34:939-944.
13.
Van Belle  GUhlmann  RHughes  JLarson  E Reliability of estimates of changes in mental status test performance in senile dementia of the Alzheimer type. J Clin Epidemiol.1990;43:589-595.
14.
Crum  RAnthony  JBassett  SFolstein  M Population-based norms for the Mini-Mental State Examination by age and educational level. JAMA.1993;269:2386-2391.
15.
Clark  CSheppard  LFillenbaum  G  et al Variability in annual Mini-Mental State Examination score in patients with probable Alzheimer disease. Arch Neurol.1999;56:857-862.
16.
Nardi  ASchemper  M New residuals for Cox regression and their application to outlier screening. Biometrics.1999;55:523-529.
Original Contribution
March 2001

A Method for Estimating Progression Rates in Alzheimer Disease

Author Affiliations

From the Departments of Neurology, Alzheimer's Disease Research Center (Drs Doody and Massman), and Internal Medicine, Division of Design and Analysis (Dr Dunn), Baylor College of Medicine, Houston, Tex.

Arch Neurol. 2001;58(3):449-454. doi:10.1001/archneur.58.3.449
Abstract

Background  The ability to predict progression of disease in patients with Alzheimer disease (AD) would aid clinicians, improve the validation of biomarkers, and contribute to alternative study designs for AD therapies.

Objective  To test a calculated rate of initial decline prior to the first physician visit (preprogression rate) for its ability to predict progression during subsequent follow-up.

Methods  We calculated preprogression rates for 298 patients with probable or possible AD (using the criteria of the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer's Disease and Related Disorders Associations (NINCDS-ADRDA) with a formula using expected Mini-Mental State Examination (MMSE) scores, scores at presentation, and a standardized estimate of duration. The patients are being followed up longitudinally in our Alzheimer Disease Research Center . The time to clinically meaningful deterioration, defined as an MMSE score drop of 5 or more points, was compared for patients stratified as slow, intermediate, and rapid progressors based on the preprogression rate. Cox regression analysis was used to examine the contribution of demographic variables (age, sex, ethnicity, and level of education), initial MMSE scores, estimated symptom duration, and the calculated preprogression rate to the time it took to reach the end point across the groups.

Results  Both initial MMSE (hazard ratio, 0.95 (0.002); z = 4.19; P<.001) and the calculated preprogression rate (hazard ratio, 1.06 (0.019); z = 3.16; P = .002) were significant in determining time to clinically meaningful decline during longitudinal follow-up (Cox regression analysis). Slow, intermediate, and rapid progressors (based on preprogression rates) experienced significantly different time intervals to clinically meaningful deterioration, with the slow progressors taking the longest time, the intermediate progressors in the middle, and the rapid progressors reaching the end point first (log rank χ21 = 9.81, P = .002).

Conclusion  An easily calculable rate of early disease progression can classify patients as rapid, intermediate, or slow progressors with good predictive value, even at initial presentation.

NEWLY DIAGNOSED patients with Alzheimer disease (AD) and their families frequently ask: How severe is the disease? How fast will it progress? How much longer do I have before it gets bad? The ability to predict progression rates would aid clinicians, patients, and patients' families in decision making. Models for predicting progression would also help to validate putative biomarkers, which should correlate better with progression rates than with severity if they reflect the active pathogenic process. Finally, better predictive models would affect the design of clinical trials by making it possible in some cases to use changes in individual progression rates to assess the efficacy of disease-modifying therapies.

Natural history studies as well as data on placebo groups in AD clinical trials have documented tremendous between-subject and between-group variability of measured progression rates. This reported heterogeneity likely reflects multiple phenomena, including (1) true differences in disease progression rates between patients; (2) differing properties, ie, floor and ceiling effects, of the measures selected; (3) differences in the end points selected to represent progression (cognitive decline, functional decline, nursing home placement, or death); (4) other methodological differences, such as the number of patients, duration of follow-up, and interval between visits; (5) differences in medical comorbidities; and (6) differences in patient care. Some authors have proposed that certain clinical features, such as age at onset, sex, duration of illness, and the presence or absence of extrapyramidal features are associated with the time to progression as assessed by nursing home placement, drop in Mini-Mental State Examination (MMSE) score, or death.13 Others have suggested that the initial stage of disease at the beginning of the observation period ("how far") is an important predictor of subsequent decline ("how fast").3,4 Most staging measures fail to document linear decline over the course of AD,5,6 and as a result both "bilinear"7 and "trilinear"8,9 models of decline have been proposed. However, no clear, predictive models of progression have been developed and validated for AD.

Alzheimer disease likely begins histopathologically many years before clinical symptoms are apparent.10 The insidious onset of symptoms makes it difficult for patients and families to be certain that an abnormal cognitive state has actually started, and often complicates the initial diagnosis. Patients therefore come to medical diagnosis at variable intervals after the first symptoms begin. Previous studies have reported average estimates of disease duration from 3 to 4.5 years at the time of first clinical presentation.4,6,11 At the time of presentation, a period of symptomatic disease has therefore existed for long enough to allow an estimate of initial rates of decline, reflecting disease activity prior to the in-clinic observation of progression (preprogression). We hypothesized that this calculated preprogression rate of decline would be predictive of subsequent disease progression (ie, rapid initial progressors would continue to decline more rapidly than slow initial progressors), but the relationship would not be linear. We therefore developed a simple procedure for estimating initial progression at the patient's first visit, and examined the ability of this measure to accurately predict the time to significant clinical decline during subsequent longitudinal observation in our Alzheimer's Disease Research Center (ADRC) at Baylor College of Medicine, Houston, Tex. We also examined the value of several previously reported variables, including age at onset, sex, level of education, duration of symptoms, and initial severity, for predicting observed progression.

PATIENTS AND METHODS
PATIENTS

Three hundred and thirty-six patients were recuited to our ADRC and diagnosed with probable (78%) or possible (22%) AD according to established criteria.12 Patients were self-referred or physician-referred and most often came for an initial diagnostic evaluation. Approximately half were from the greater Houston area, and the rest were from elsewhere in Texas or from other states. All eligible subjects who met the following criteria were analyzed for this study: complete demographic data, a standardized physician's estimate of duration prior to diagnosis, an initial-visit MMSE score of 5 or higher, at least 1 year of follow-up, and at least 3 MMSE scores. The requirements for a minimum follow-up and minimum number of MMSE scores were designed to improve the reliability of the estimated slopes of observed MMSE score decline.13 Thirty-eight subjects did not meet the eligibility criteria: 1 lacked the physician's estimate of duration; 10 had initial MMSE scores lower than 5; 18 had only 2 MMSE scores; and 8 had less than 1 year of follow-up at the time of the analysis. Ninety percent of our subjects were white; 65% were women; and the overall mean (SD) age was 70 (8.3) years. The mean education level was 13 (3.5) years, and the mean duration of symptoms prior to diagnosis in our center was 3 (1.8) years.

MEASURES

Duration of symptoms was estimated according to a standardized ADRC protocol: we interviewed the patient and all available informants, reviewed medical records to look for previous chronologies of symptoms, and asked the patient's caregiver to estimate the duration of 34 symptoms commonly associated with AD. The physician then estimated the duration of symptoms to the nearest half year after resolving any discrepant information through further questioning and by relating time frames to the patient's life events.

A preprogression rate was calculated for each patient according to the following formula: (MMSE score [expected] − MMSE score [initial]) / physician's estimate of duration [in years]). The expected MMSE score was derived from a table of age- and education-corrected population-based norms.14 The importance of using MMSE normative data rather than assuming that the premorbid MMSE should be a perfect score of 30 for every patient is illustrated by the following example: a healthy 60-year-old man with a college degree would be expected to score 29, whereas a normal 20-year-old man with 4 years of education would only be expected to score a 20.14

Based on calculated preprogression rates, patients were stratified into slow (0-1.9 MMSE points per year), intermediate (2-4.9 MMSE points per year), and rapid progressors (≥5 MMSE points per year). The cutoff points were based on literature showing that average group decline is usually in the range of 2 to approximately 4 points per year.5,6 We therefore set this range as intermediate, and defined slow and fast in relation to it. Twenty patients in whom the preprogression rate was unexpectedly less than 0 were considered a separate group for the analysis.

The observed progression rate during follow-up was obtained by the formula (MMSE score [last] − MMSE score [initial]) / interval between follow-up visits (years). For generating Kaplan-Meier curves and performing Cox regression analysis, we chose as our clinical progression end point a drop of 5 or more points on the MMSE score from the initial clinic visit. This value was chosen to reflect clinically meaningful cognitive decline. Few patients' scores decline this much in a typical year, and patients who improve or "back-cross" on the MMSE score rarely achieve a 5-point improvement in score.15

STATISTICAL ANALYSIS

Basic statistics (ie, frequencies for categorical variables, means, and SDs for continuous variables) for the following variables were initially determined for the group as a whole: sex, ethnicity, age, education, expected MMSE score (normative), physician's estimate of symptom duration, initial-visit MMSE score, calculated preprogression rate, follow-up years, observed progression rate, and percentage of patients reaching the end point of a 5-points decline in MMSE score.

Pearson's r (with Bonferroni corrections for multiple comparisons) was used to test for linear correlations between the continuous variables. Analysis of variance was used to test whether the continuous variables of interest differed for the groups stratified into slow, intermediate, rapid, or "negative" progressors, based on the calculated preprogression rate, with the χ2 test used for the categorical variables. Cox regression analysis was used to look for predictors of the time to a 5-point decline in MMSE score. After testing the assumption of proportional hazards (Schoenfield residuals), 1-variable, 2-variable, and finally multivariable Cox equations were considered, using the following variables: age, sex, education level, physician's estimate of duration, initial-visit MMSE score, and preprogression rate. Age and education level were considered both as continuous and categorical variables (age, <65 years vs age ≥65 years; education, high school or less vs more than high school). These models were assessed with and without the small group of 20 individuals with a negative preprogression rate. The overall fit of the models was assessed by the Cox-Snell residuals, and the model accuracy and outliers were assessed by the pattern of deviance residuals.16 Finally, we constructed Kaplan-Meier curves depicting the time to 5-point or greater MMSE score drops for the estimated slow, intermediate, and fast progressors, and analyzed the separation of these groups over time with the log rank test.

RESULTS

The mean (SD) normalized MMSE score was 28 (1.4); initial MMSE score, 20 (6.3); and last MMSE score, 12 (8.7). The calculated preprogression rate for the group of 298 subjects was 3 (3.4) points per year on the MMSE score, and the mean observed progression during follow-up was 3 (3.2) points per year. At the time of the analysis, 33% of the subjects had been observed for 1 year, 28% for 2 years, 18% for 3 years, and 22% for 4 or more years, with continued follow-up ongoing. For 20 subjects, the initial-visit MMSE score was higher than the population-derived expected MMSE score, indicating that the normative score was not accurate for these subjects, and yielding a negative preprogression rate. The MMSE values and progression rates as well as the average duration, average follow-up, and percentage reaching the end point are given for the 298 patients in Table 1.

Significant correlations between the demographic and progression variables are given in Table 2. As expected, the preprogression rate correlates negatively with its components: initial MMSE score and duration of symptoms. Also as expected, the initial MMSE score correlates with the final MMSE score, and correlates negatively with the symptom duration. Observed progression rate did not correlate with the initial MMSE score, perhaps reflecting the nonlinearity of progression as assessed by this measure (confirmed by a graph of the 2 variables, not shown), and the preprogression rate did not correlate with the observed progression rate or time to event, probably for the same reason.

Stratification by preprogression rates classified 123 patients as slow, 110 as intermediate, and 65 as rapid progressors, and 20 as negative progressors. Since the clinical significance of this last group (comprising patients with higher MMSE scores at presentation with dementia than would be predicted if they were not demented) is unknown, the analysis was performed both with and without this group. Data for slow, intermediate, and rapid progressors is given in Table 3. Analysis of variance testing for differences in age, length of follow-up, and χ2 analysis of sex percentages did not reveal differences across the stratified groups, but education level did differ across the preprogression categories (N = 298; F3,294 = 5.18, P = .002). Negative preprogressors had a mean (SD) 11 (3.6) years of education vs 14 (3.2) years for slow, 13 (3.8) for intermediate, and 12 (3.1) years for rapid progressors. Because analysis of variance assumptions were not met for initial MMSE scores across preprogression categories, this variable was examined by the Kruskal-Wallis test and showed significant differences (χ23 = 14.51, P = .002), with a gradient of lower scores with more rapid preprogression rates, as would be expected. The physician's estimate of duration also differed in the Kruskal-Wallis test, with the shortest duration in the most rapidly progressing subjects (χ23 = 65.99, P<.001). The results were the same with and without the 20 subjects with negative preprogression rates, and were the same when we used a MMSE of 30 (instead of the normalized score) in the preprogression formula.

Rapid progressors, intermediate progressors, slow progressors, and patients with negative preprogression rates were equally likely to reach the end point of a 5-point or greater MMSE score drop (χ23 = 4.30, P = .23). Seventy-three percent of the patients in this study experienced this degree of decline in the course of follow-up, reflecting the fact that we achieved sufficient follow-up for patients in all groups to detect significant progression. Individual Cox regression analyses examining age (continuous or categorical), sex, education level (continuous or categorical), duration, calculated preprogression rate, and MMSE score at first visit, individually, as predictors of time to event showed that only the preprogression rate (hazard ratio [SE], 1.06 [0.019]; z = 3.16; P = .002; confidence interval [CI], 1.021-1.094) and initial MMSE score (hazard ratio [SE], 0.96 [0.011]; z = −3.87; P<.001; CI, 0.936-0.979) were significant.

When age, education level, disease duration, and sex were added to preprogression rate in the Cox model, preprogression rate remained significant, but none of the other variables reached significance. The results were similar with and without the negative preprogressors. When these variables were added one at a time to the model using initial visit MMSE scores, both initial visit MMSE score (hazard ratio [SE], 0.95 [0.011]; z = −4.25; P<.001; CI, 0.930-0.974) and duration (hazard ratio [SE], 0.92 [0.038]; z = −2.05; P = .04; CI, 0.847-0.996) were simultaneously significant, but none of the other variables yielded P values close to significant. When the 20 negative preprogressors were omitted, only initial visit MMSE score remained significant in the Cox model that included initial visit MMSE scores. Models using multiple (>2) variables in addition to either preprogression or initial-visit MMSE score did not yield any other significant variables. The model of choice is therefore either preprogression alone or initial MMSE score with duration if the negative preprogressors were excluded. When the analysis was repeated on the probable AD group alone (n = 233), preprogression and intial MMSE score, but not duration, were significant. The power was too low to draw any conclusions for the possible AD group (n = 65). Findings of the Cox-Snell residuals suggest that the model using both initial-visit MMSE score and duration may fit slightly better than the preprogression rate, but the pattern for deviance residuals did not distinguish between the 2 models (data not shown).

Kaplan-Meier curves comparing the time to event for the slow, intermediate, and rapid preprogressors are shown in Figure 1. The log rank test demonstrated that the time to event was different when the 3 groups (χ22 = 20.17, P<.001) vs the 4 groups (creating a separate group for the 20 patients with a negative preprogression rate) were considered (data not shown). The mean (SD) time to a 5-point MMSE score drop (end point) was 2.3 (1.42) years for the slow progressors, 1.8 (1.13) years for the intermediate progressors, and 1.6 (0.94) years for the rapid progressors. The greater variability in the intermediate group may indicate multiple factors influencing progression in this group. Additional Kaplan-Meier curves comparing the time to event for those achieving equal to or above the median initial MMSE score (21 points) with those below it were also generated (not shown), and the log rank test demonstrated significant differences (χ21 = 9.81, P = .002). A post hoc analysis comparing the relationship between MMSE score severity group (5-10, 11-19, and 20-30 MMSE points) and subsequent observed progression rate (< 0, 0-1.9, 2-4.9, ≥5 MMSE points/y) was negative (χ23 = 10.26, P = .11).

COMMENT

Progression rates of AD are variable between subjects, and a graph of cognitive decline vs time does not yield a linear function over the course of the disease. Nonetheless, we have shown that progression rates in AD show some consistency in that patients who begin with progression rates that are more rapid than average (≥5 MMSE points per year) continue to experience clinically significant decline sooner than patients who begin at slow (≤2 points per year) or average rates (2-4.9 points per year). We used a calculated rate of progression, reflecting disease activity prior to the initial visit (the preprogression rate) as well as data obtained by longitudinal follow-up of up to 10 years of 298 patients in our ADRC. Several of our results support this conclusion of consistency in progression: (1) Kaplan-Meier curves graphing the time to significant clinical decline that show continued separation of the slow, intermediate, and rapid preprogressing patients; (2) standardized estimates of disease duration were shorter for those classified as rapid preprogressors, and longest for the slow preprogressors, with the intermediate preprogressors showing times in the middle; and (3) the fact that initial visit MMSE scores were lowest for patients who were rapid preprogressors, with the expected gradient for the intermediate and slow groups.

Our data support previous reports that AD-associated drop in MMSE scores over time is nonlinear59: there was no simple correlation in our study between preprogression rates and observed progression rates, despite the fact that the average (SD) preprogression rate (3 [3.4] MMSE points per year) was similar to the observed progression rate for the group (3 [3.2] points per year). We did observe a relationship between preprogression and the observed time to significant clinical deterioration, defined as the time to a 5-point drop on the MMSE score. Selection of the time to clinically significant worsening as our primary outcome measure and use of a survival analysis allowed us to discover the importance of preprogression rates to subsequent deterioration on the MMSE score, even though decline on this measure is not linear.

Our model differs from that of Stern et al3 in several ways. First, their model seeks to predict mean progression rates (and variances), whereas ours is an attempt to predict the interval before clinical decline. Their predictors were derived from one population and applied to another, while ours were derived from a single population. Theirs is a multicenter study while ours is a single-center study, although the number of patients included in the 2 studies is comparable. The 2 models are compatible but lead to different information: after applying their approach, a patient can be given a mean time and range before reaching nursing home placement or death. After applying our method, the patient might be told whether he/she is declining at a rapid or a slow rate, and how long it might be before clinically meaningful deterioration occurs, of a nature to be reflected on the MMSE score.

Our results also show that, for modeling purposes, knowing the duration of symptoms in addition to the initial visit MMSE score (2 of the 3 components of preprogression rate) is as good as knowing the preprogression rate. However, combining these variables into the simple formula that we used to calculate preprogression rates makes the data points more useful in a clinical setting, and the predicted time to clinical deterioration was similar with both methods. The use of the preprogression rate was problematic for 20 subjects who had higher MMSE scores when they presented with dementia than the population-based norms would have predicted for their best normal scores. Most of these individuals had relatively low (less than high school) educational levels, suggesting that the normative MMSE values we used for this study14 might not be as accurate for subjects with lower educational levels. However, when these 20 negative preprogressors were excluded from the analysis, or when we substituted an MMSE score of 30 in place of the normalized score, the predictive value of the preprogression rate remained significant.

In conclusion, our data suggest that it is reasonable to calculate a patient's rate of progression prior to presentation and to use this information to predict whether subsequent clinical decline will occur at a shorter-than-average, average, or longer-than-average interval. These findings are limited because we used only the MMSE score as a measure of decline. Future studies will investigate whether the MMSE-based preprogression rate is predictive of decline on other measures. Also, we have not yet observed all of the subjects throughout the duration of their disease, so our findings apply only to individuals observed for at least a year and up to 10 years. We do not know whether the predictive utility of the preprogression rate is equally strong in the first, second, or third years, etc. Better understanding of the preprogression rate will likely lead to models of progression in AD that can be used for clinical prognostication, validation of putative biological markers, and for designing clinical trials of disease-modifying therapies where patients' preprogression rates may serve as controls for their postintervention rates of decline.

Back to top
Article Information

Accepted for publication January 2, 2000.

Drs Doody, Massman, and Dunn were supported by Alzheimer's Disease Research Center grant AGO8664 from the National Institute on Aging, Bethesda, Md. This work was also supported by the B. and L. Martin fund and the Fyfe Foundation, both in Houston, Tex.

We would like to thank Elaine Teoh for her assistance in preparing the manuscript, and Peggy Lynn for her assistance in an earlier version of the analysis.

Reprints: Rachelle Smith Doody, MD, PhD, Baylor College of Medicine, Department of Neurology, 6550 Fannin St, Suite 1801, Houston, TX 77030.

References
1.
Jacobs  DSano  MMarder  K  et al Age at onset of Alzheimer's disease: relation to pattern of cognitive dysfunction and rate of decline. Neurology.1994;44:1215-1220.
2.
Chui  HLyness  SSobel  ESchneider  L Extrapyramidal signs and psychiatric symptoms predict faster cognitive decline in Alzheimer disease. Arch Neurol.1994;51:676-681.
3.
Stern  YTang  MAlbert  M  et al Predicting time to nursing home care and death in individuals with Alzheimer disease. JAMA.1997;277:806-812.
4.
Kraemer  HTinklenberg  JYesavage  J "How far" vs "how fast" in Alzheimer disease. Arch Neurol.1994;51:275-279.
5.
Salmon  DThal  LButters  NHeindel  W Longitudinal evaluation of dementia of the Alzheimer type: a comparison of three standardized mental status examinations. Neurology.1990;40:1225-1230.
6.
Morris  JEdland  SClark  C  et al The Consortium to Establish a Registry for Alzheimer's Disease (CERAD), IV: rates of cognitive change in the longitudinal assessment of probable Alzheimer's disease. Neurology.1993;43:2457-2465.
7.
Haxby  JRaffaele  KGillette  J  et al Individual trajectories of cognitive decline in patients with dementia of the Alzheimer type. J Clin Exp Neuropsychol.1992;14:575-592.
8.
Stern  RMohs  RDavidson  M  et al A longitudinal study of Alzheimer's disease: measurement, rate, and predictors of cognitive deterioration. Am J Psychiatry.1994;151:390-396.
9.
Brooks  JYesavage  J Identification of fast and slow decliners in Alzheimer disease: a different approach. Alzheimer Dis Assoc Disord.1995;9(suppl):S19-S25.
10.
Braak  HBraak  E Neuropathological staging of Alzheimer-related changes. Acta Neuropathol.1991;82:239-259.
11.
Bracco  LGallato  RGrigoletto  F  et al Factors affecting course and survival in Alzheimer's disease: a 9-year longitudinal study. Arch Neurol.1994;51:1213-1219.
12.
McKhann  GDrachman  DFolstein  MKatzmann  RPrice  DStadlan  EM Clinical diagnosis of Alzheimer's disease: report of the NINCDS-ADRDA Work Group under the auspices of the Department of Health and Human Services Task Force on Alzheimer's Disease. Neurology.1984;34:939-944.
13.
Van Belle  GUhlmann  RHughes  JLarson  E Reliability of estimates of changes in mental status test performance in senile dementia of the Alzheimer type. J Clin Epidemiol.1990;43:589-595.
14.
Crum  RAnthony  JBassett  SFolstein  M Population-based norms for the Mini-Mental State Examination by age and educational level. JAMA.1993;269:2386-2391.
15.
Clark  CSheppard  LFillenbaum  G  et al Variability in annual Mini-Mental State Examination score in patients with probable Alzheimer disease. Arch Neurol.1999;56:857-862.
16.
Nardi  ASchemper  M New residuals for Cox regression and their application to outlier screening. Biometrics.1999;55:523-529.
×