A Longitudinal Study of Alzheimer s Disease: Measurement, Rate, and Predictors of Cognitive Deterioration

Transcription

1 A Longitudinal Study of Alzheimer s Disease: Measurement, Rate, and Predictors of Cognitive Deterioration Robert G. Stem, M.D., Richard C. Mohs, Ph.D., Michael Davidson, M.D., James Schmeidler, Ph.D., Jeremy Silverman, Ph.D., Elise Kramer-Ginsberg, Ph.D., Terena Searcey, B.A., Linda Bierer, M.D., and Kenneth L. Davis, M.D. Objective: This study measured the annual rate ofcognitive change in patients with Alzheimer s disease and determined the effects ofclinical variables on that rate. It also compared the ability of two cognitive scales to measure change over the entire range of dementia severity. Method: The cognitive subscale of the Alzheimer s Disease Assessment Scale and the Blessed test for information memory and concentration were given to patients with Alzheimer s disease and 72 nondemented elderly comparison subjects at 6-month intervals for up to 9 months. Longitudinal changes in scores on the cognitive subscale were measured with several different methods of data analysis. Results: For the patients with Alzheimer s disease, the annual rate of change in cognitive subscale scores showed a quadratic relationship with dementia severity in which deterioration was slower for mildly and severely demented patients than for patients with moderate dementia. Gender, age at onset, and family history of dementia 1 had no effect on the rate of cognitive deterioration. The comparison group showed a slight improvement in cognitive performance over time. All data analytic methods gave similar results. The cognitive subscale of the Alzheimer s Disease Assessment Scale was more sensitive to change in both mild and severe dementia than was the Blessed test. Conclusions: These results suggest that cognitive deterioration is slow during the early and very late stages of Alzheimer s disease and more rapid during the middle stages. No clinical variables other than degree of cognitive impairment and previous rate of cognitive decline predicted rate of deterioration. These results have implications for treatment trials and attempts to identify subgroups. (Am J Psychiatry 1994; 151:39-396) P rospective longitudinal assessment of cognitive change in Alzheimer s disease is an important part of current research efforts to develop effective treatments and to investigate possible subtypes of the disorder. Clinical trials of antidementia drugs are designed to determine whether a drug can improve cognitive performance or slow the rate of cognitive decline (1). Performance-based cognitive tests are es- Presented in part at the 145th annual meeting of the American Psychiatnic Association, Washington, D.C., May 2-7, Received Feb. 11, 1993; revision received Aug. 9, 1993; accepted Aug. 2, From the Department of Psychiatry and the Department of Biomathematical Sciences, Mount Sinai School of Medicine, Mount Sinai Medical Center, New York, and the Psychiatry Service, Bronx VA Medical Center. Address reprint requests to Dr. Mohs, Psychiatry Service (1 16A), Mount Sinai School of Medicine, VA Medical Center, 13 West Kingsbridge Rd., Bronx, NY Supported by grants AG-2219 and AG-5138 from the National Institute on Aging. The authors thank Theresa Ryan, Rachel Markofsky, and the staff ofthe Alzheimer s Disease Research Center of the Mount Sinai School of Medicine for their cooperation. sential tools in assessing the efficacy of potential treatments (2). Subtypes of Alzheimer s disease might be differentiated by many factors, one of which is rate of deterioration. Some data from genetic (3), phenomenological (4), neumochemical (5), and longitudinal (6) studies suggest that patients with early onset of the disorder may be different from those with later onset. Studies have suggested that patients with a family history of dementia may have different biological (7) and phenomenological (8) profiles from those without such a family history. To determine whether a characteristic such as age at onset or family history predicts clinical course, it is necessary to follow patients over time with cognitive measures that are sensitive to change over a broad range of dementia severity. Several instruments, including the cognitive portion of the Alzheimem s Disease Assessment Scale (9), the Mini-Mental State examination (1), the Blessed test for information memory and concentration (1 1 ), and the battery from the Consortium to Establish a Registry for Alzheimer s Disease (12), are widely used in the as- 39 Am J Psychiatry 1 51 :3, March 1994

2 STERN, MOHS, DAVIDSON, ET AL. TABLE 1. Characteristics of Patients With Change in Cognitive Functioning Alzheimer s Disease and Nondemented Elderly Comparison Subjects in a Longitudinal Study of Patients With Alzheimer s Disease (N=111) Nondemented Elderly Subjects (N=72) Variable Mean SD Median Range Mean SD Median Range Age at onset of dementia (years) Education (years)a Age at first assessment (years) Cognitive subscaleb score at first assessmentc Number of months of follow-up per subjectc Number of follow-up visits per subjecte asignificant difference between groups (t=3.18, df=1 81, pczo.1). bf the Alzheimer s Disease Assessment Scale. csignificant difference between groups (t=7.4, df=1 81, p<.1) S dsignincant difference between groups (t=s.s1, df=181, p<.1). esignificant difference between groups (t=6.23, df=181, p<.1). sessment of cognitive status in dementia. Each of these tests has been shown to be reliable and to differentiate patients with dementia from nondemented persons. Each could be used as an outcome measure in a clinical trial or to investigate subgroup differences in clinical progression of the disorder. Despite the widespread use of these instruments, however, relatively few studies have examined how cognitive functioning, as measured by these scales, changes over time in patients with Alzheimer s disease. For the Blessed test there is some cvidence that the rate of cognitive decline depends on dementia severity (13, 14), although this result may be due in part to ceiling effects (14). The present study investigated longitudinal change as measured by the cognitive subscale of the Alzheimer s Disease Assessment Scale in a large, well-characterized group of patients with Alzheimer s disease and a nondemented comparison group. The cognitive subscale of the Alzheimer s Disease Assessment Scale assesses memory, language, praxis, and orientation and has been used extensively in treatment trials of antidementia drugs in the nited States (15), Europe (16), and Japan (1 7). Rate of change was examined as a function of dementia severity and as a function of four clinical variables: gender, age at onset, family history of dementia, and prior rate of cognitive decline. Finally, the performance on the cognitive subscale and on the Blessed test of the nondemented elderly subjects and the subjects with very severe dementia was compared to determine the relative sensitivity of the two scales in these groups at the opposite ends of the severity continuum. METHOD The patients with Alzheirner s disease and the elderly comparison group are participants in a large prospective study on the natural history of Alzheimer s disease that has been described previously ( 14, 18, 19). Briefly, the patients met the criteria of the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer s Disease and Related Disorders Association for probable Alzheimer s disease (2) and had no current or past psychiatric disorder. Patients and nondemented elderly subjects were included in the current analysis only if they had been evaluated on at least two different occasions 12 or more months apart. Table I presents demographic and baseline data on the patients with Alzheimer s disease (67 male and 44 female) and the 72 nondemented persons (48 male and 24 female). There were 65 patients with onset before age 65 and 41 with onset after age 65; for five patients there was insufficient cvidence to estimate age at onset. Family histories were obtained for 82 patients by methods described previously (21). Twenty-nine patients had a first-degree relative with an Alzheimer s-disease-like illness, and 53 had no family history of Alzheimer s disease; for the remaining 29 patients, adequate family history data were not available. As table 1 indicates, the patients with Alzheimer s disease and the nondemented comparison subjects did not differ significantly in age at first assessment, but the comparison subjects were slightly better educated, had lower scores on the cognitive subscale of the Alzheimer s Disease Assessment Scale at baseline, had more follow-up visits per subject, and had a longer duration of follow-up. The patients and nondemented elderly subjects were assessed on the cognitive subscale of the Alzheimer s Disease Assessment Scale, scored (no impairment) to 7 (most severely impaired), and the Blessed test, scored (no impairment) to 33 (most severely impaired), at 6-month intervals whenever possible. The cognitive subscale consists of 1 1 items assessing memory, language, orientation, and praxis and takes approximately 3 minutes to administer. The Blessed test was adapted for American use as described earlier ( 14). These tests were given as part of a larger assessment battery that included the entire Alzheimer s Disease Assessment Scale along with other neuropsychological and clinical measures. Patients and comparison subjects were dropped from the follow-up study only if they withdrew consent, died, became untestable, or moved away so that they could not be tested. Of the subjects who were included in the longitudinal analysis, some missed one or more assessment visits because of illness, travel, or other factors, and the total number of visits for each subject was sometimes less than would be expected from the length of follow-up. There is no generally accepted best way to calculate longitudinal change in naturalistic studies, since patients are followed for different lengths of time, and rate of change may be nonlinear. A variety of plausible methods have been used previously. They differ in the extent to which they utilize all data points and make assumptions about the shape of the longitudinal change function. To ensure that the conclusions presented in this article would not depend on the choice of a single analytic method, we used several different methods to calculate longitudinal change in cognition from scores on the cognitive subscale of the Alzheimer s Disease Assessment Scale. The relationship of baseline and clinical variables to change according to all of the methods was then examined. Further information on some of these methods has been presented (14). First, the annual rate of cognitive change was estimated by dividing the difference between the first and last available scores on the cognitive subscale by the number of years between the two assessments. This unrestricted two-point estimate is quite simple but assumes a linear change over time, does not utilize all data points, and may be markedly affected by ceiling and floor effects. Second, we used a restricted two-point estimate, which was identical to the unrestricted two-point estimate except that a subject s first rating of 7, as well as all subsequent ratings of that particular subject, on the cognitive sub- Am ] Psychiatry 1 51 :3, March

3 LONGITDINAL STDY OF ALZHEIMER S DISEASE TABLE 2. Annual Rate of Cognitive Change in Patients With Alzheimer s Disease and Nondemented Elderly Comparison Subjects as Calculated by Seven Data Analytic Methods From Scores on the Cognitive Subscale of the Alzheimer s Disease Assessment Scale Patients With Alzheimer s Disease (N=1 1 1 ) Nondemented Elderly Subjects (N=72) Change Score Number Number Change Score of Effect of Method Scores Mean SD Median Mode Size Scores Mean SD Median Mode Multiple 12-month intervals First 12-month interval Multiple 6-month intervals S First 6-month interval nrestricted two-point estimate I I I S Restricted two-point estimate S Least squares FIGRE 1. Absolute and Cumulative Frequencies of Year Change Scores on the Cognitive Subscale of the Alzheimer s Disease Assessment Scale for Patients With Alzheimer s Disease (Multiple 12-Month Interval Method of Calculation) B. E C z a I- -J (=) YEAR Ln.NGE IN COGNITIVE SBSCALE SCORE scale were excluded to minimize ceiling artifacts. Third, the least squares method was used to estimate the annual rate of cognitive change by calculating the slope of the linear regression of all cognitive subscale scores on years from the baseline assessment for each subject; a subject s first score of 7 and all subsequent ratings of that particular subject were excluded. This method uses all intermediate data points but assumes linearity. Next, with the multiple 12-month interval method, the change in score on the cognitive subscale between pairs of visits separated by 12 months was calculated by subtracting the earlier (baseline) score from the later one. When a subject generated more than one pair of cognitive subscale scores 12 months apart, that subject contributed more than one value for annual rate of cognitive change to the analysis. Pairs of scores with a baseline score greater than 63 were excluded to avoid a ceiling artifact. The multiple 12-month interval method uses all data points and does not assume linearity over the entire follow-up period. However, it poses some potential statistical difficulties owing to the fact that subjects contribute different numbers of observations depending on the length of follow-up. To have an analysis of measured, fixed intervals with each patient contributing a single observation, we also used the first 12-month interval method, in which the change in cognitive scale score during the first 12-month interval was calculated for each patient. To determine whether the estimates based on fixed intervals were affected by the time between assessments, we also calculated the annual rate of cognitive change by doubling the changes measured over 6-month intervals, using the same methods as described for the multiple 12-month interval and the first 12-month interval calculations to yield those for the multiple 6-month intervals and the first 6-month interval. RESLTS Table 2 presents the annual mates of cognitive change C) for the patients with Alzheimem s disease and the nondemented elderly subjects that were obtained by the seven different methods of calculation. By all methods, the patients with Alzheimer s disease showed substantial worsening over time, while the nondemented subjects improved slightly. For the patients with Alzheimer s disease, the unrestricted two-point estimate was the smallest, owing to the fact that it uses scores even after the patient has reached the ceiling value of the? scale. score With all of the on the cognitive methods, subscale the of estimated change the Alzheimer s in Disease Assessment Scale was approximately points per year. Also shown in table 2 is the effect size (mean change score divided by standard deviation) for the patients with Alzheimer s disease according to each method of measuring. When longitudinal change is used as a dependent measure, as in a drug treatment study, effect size is useful for estimating the sample size needed to detect effects of a specified size. Of particular interest are estimates based on 6- and 12-month change scores, since intervention studies usually last either 6 or 12 months; rarely do they involve variable follow-up periods, as is the case for the unrestricted two-point estimate, the restricted two-point estimate, and the least squares methods. As has been noted with the Blessed test (14), the effect size for measuring longitudinal change increases substantially when change is measured over 12 rather than 6 months. Figure 1 presents a histogram and cumulative frequency plot of the annual rate of cognitive change as measured by the multiple 12-month interval method. The distribution of measured change scores on the cognitive subscale is skewed to the right. On 7% of the measured 12-month intervals, scores actually improved. We used multiple regression analysis to assess the melation of annual rate of cognitive change to baseline cognitive subscale score according to each of the ana- 392 AmJ Psychiatry 151:3, March 1994

4 STERN, MOHS, DAVIDSON, El AL. FIGRE 2. Mean 1-Year Change Scores, Expressed as a Function of Baseline Scores, on the Cognitive Subscale of the Alzheimer s Disease Assessment Scale for Patients With Alzheimer s Diseasea Zu -J C, z <, < > BASELINE COGNITIVE SBSCALE SCORE athe month change scores graphed in figure 1 were grouped according to their baseline scores. Mean change scores from each baseline score between 3 and 67 are represented by black dots. Tnangles represent predicted values calculated from the quadratic equation fitted by the least squares method to the 253 original data points: annual rate of cognitive change= x baseline scone x baseline score squared (R=.42, F=27.46, df=2, 25, p<.1 ). Straight lines connecting the estimated points approximate the actual quadratic equation. FIGRE 3. Predicted Scores on the Cognitive Subscale of the AIzheimer s Disease Assessment Scale Based on the Multiple Regression Equation Obtained by Fitting Data From 72 Patients With Alzheimer s Disease Who Had Three or More 5Sa Cl) -I, Cl) I- z 8 I- 7 6O 5 4O 3 2#{149} 1#{149} ----5o ----4o 3 2O J FOLLOW-P TIME (months) Baseline Score lytic methods. Linear, quadratic, and cubic orthogonal polynomials were calculated, with the Bonferroni correction for multiple comparisons. For the patients with Alzheimer s disease, there was a significant quadratic component, relative to higher-order polynomial effects, between baseline scores on the cognitive subscale of the Alzheimer s Disease Assessment Scale and annual rate of cognitive change according to all assessment methods (p<o.oosin all cases), except for the first 6-month interval method, which showed a nonsignificant quadratic trend (p<o.lo). Only the unrestricted two-point estimate had a significant linear component of association between baseline score and annual rate of cognitive change (F=1.49, df=1, 49, p<.1), and none of the methods yielded a significant cubic component. For all methods, the quadratic regression equations described an inverted--shaped curve, with cognitive change slow in mildly impaired patients, more rapid in moderately demented patients, and slow again in severely affected patients. Figure 2 presents the results for the multiple 12-month interval method graphically. For each baseline cognitive subscale score between 3 (the lowest obtamed score) and 63, the mean of all 12-month change scores (from the total of 253) starting from that baseline was calculated. The amount of cognitive change cxpected in the 12 months following a specific baseline assessment varied from as little as 2.53 points if the baseline cognitive subscale score was 1 to as much as if the baseline score was 4. For the nondeathe best-fitting equation was score,= X1 -.S392X2 + ( S2X X,2)T, where score,, is the predicted score on the cognitive subscale for patient i at time t, X is the cognitive subscale score obtained at time for patient i, and T is the time, in months, since the first score was obtained (R=.89, F=161., df=s, 218, p<.1). Predicted scores are shown for patients whose scores of 1, 2, 3, 4, or SO at their initial evaluation (time ) are displayed on the right side of the figure. mented subjects theme was a highly significant negative linear correlation between cognitive status and annual rate of cognitive change by all approaches (p<o.ooi in all cases) with the exception of the least squares method. After Bonferroni correction, in all cases the polynomial quadratic or cubic components were not significant. To determine the extent to which individual patients followed over time showed a pattern of change like that shown in figure 2, we modeled the longitudinal cognitive subscale data for a subset of 72 patients, all of whom had scores that had been obtained at three or more visits. Stepwise multiple regression analysis was used to predict the cognitive subscale scores obtained after the first visit (time ) for each patient as a function of the patient s first-visit score (t=1s.8, df=222, p<.1, 53% ofvariance), the square ofthe first score (t=-3.12, df=221, p<o.oo2, 2% of variance), the time in months since the first visit (t=14.6, df=22, p<ooo1, 21.3% ofvariance), the First Score by Time interaction (t=4.43, df=219, p<o.oo1, 2% of variance), and the Square of the First Score by Time interaction (t=-2.2, df=21 8, p<o.os,.4% of variance). Terms involving the square of time did not significantly improve the fit of the model. Figure 3 presents the cognitive subscale scores pmedicted by the multiple regression for hypothetical patients with scores of 1, 2, 3, 4, and SO at their first evaluation. Because few of the 72 patients used to de- AmJ Psychiatry 151:3, March

5 LONGITDINAL STDY OF ALZHEIMER S DISEASE velop the model were tested for more than 2 years, the predictions are shown for 24 months. As expected from the grouped data in the multiple 12-month interval calculations, the predicted changes for individual patients were less for those whose initial scores indicated very mild or very severe disorder than they were for those with moderately severe disorder. The predicted 24- month changes for patients with initial cognitive subscale scores of 1, 2, 3, 4, and SO were 5.14, 12.21, 15.92, 16.28, and points, respectively. While the predicted change scores obtained by fitting individual patients data over time were not the same as those obtamed from the fit of data averaged by baseline score in the multiple 12-month interval method, the pattern of change as a function of starting value was the same. Independent of the computation method used to investigate clinical predictors of annual rate of cognitive change, the annual mate did not significantly vary with gender (male versus female), age at onset (before age 65 versus after age 65), or family history (dementia versus no dementia). We also tested the hypothesis that cognitive change during one interval would predict cognitive change during another interval for the same individual. For this analysis we used the subset of 48 patients with Alzheimem s disease who had been tested at entry into the study and 6, 12, and 1 8 months later without reaching the maximum score on the cognitive subscale. Two annual change scores, 12-month minus entry and 18- month minus 6-month, were calculated; these change scores did not use any of the same data points that would tend to produce a negative correlation due to regression to the mean (22). The correlation of the 12- month minus entry change score with the 1 8-month minus 6-month change score was.56 (df=48, p<o.oo1), indicating that there was some consistency over time in the rate of change in score on the cognitive subscale of the Alzheimem s Disease Assessment Scale. The relative sensitivity of the cognitive subscale of the Alzheimer s Disease Assessment Scale and the Blessed test in severe dementia was examined with respect to 31 of the patients who at entry into the study had not reached the maximum score on either scale but who later reached the maximum on at least one of them. Ten (32%) of these patients reached the maximum on both scales at the same visit, and the remaining 21 patients (68%) reached the maximum score on the Blessed test before reaching the maximum on the cognitive subscale. None of the patients reached the cognitive subscale maximum first. Twenty-five of the 31 patients reached the maximum score on both tests during follow-up. For these patients, the mean time between their reaching the maximum on the Blessed test and reaching the maximum on the cognitive subscale was 7.68 months (range=-3). The relative sensitivity of the two scales to very mild cognitive impairment was evaluated among 74 elderly comparison subjects for whom we had both cognitive subscale and Blessed test scores at study entry. Forty-six comparison subjects (62%) made no errors on the Blessed test and one or more errors on the cognitive subscale; 28 (38%) made one or more errors on both scales. No comparison subject made no errors on either scale, and no comparison subject scored on the cognitive subscale and higher than on the Blessed test. The mean cognitive subscale score for the 46 comparison subjects who made no errors on the Blessed test was 3.59 (SD=1.87, range=1-9). DISCSSION The annual mate of cognitive change among the patients with Alzheimer s disease was strongly dependent on their baseline scores on the cognitive subscale of the Alzheimem s Disease Assessment Scale; measured change was slow during intervals of mild cognitive impaimment, was most rapid during intervals of moderate cognitive impairment, and was again slower during intervals of severe dementia. The measured mate of change was not a function of gender, age at onset of disease, or presence of a family history of dementia, but there was evidence that the mate of change was somewhat consistent over time for a given patient. The cognitive subscale was more sensitive both to mild cognitive deficits associated with normal aging and to change in severely demented patients than was the Blessed test. The observed relation between rate of change and severity might be due to the specific composition of the cognitive subscale of the Alzheimem s Disease Assessment Scale, or it could accurately reflect the correlation of neuropathologic and symptomatic change in Alzheimcm s disease. Only a few studies using other instruments have investigated longitudinal change in Alzheimcm s disease symptoms. Several (13, 23), but not all (6, 24) longitudinal studies using the Blessed test found that measured change decreased with increasing severity. We recently published Blessed test data obtained in the same group used to investigate the cognitive subscale of the Alzheimem s Disease Assessment Scale (14); we found that when maximum scores were eliminated to remove ceiling artifacts, there was only a modest decrease in the rate of change for severely demented patients. However, the comparison of the relative sensitivities of the cognitive subscale and the Blessed test presented here indicates that the Blessed test covers a much narrower range of severity than does the cognitive subscale. Thus, it is quite possible that no strong relation between mate of change and severity was observed with the Blessed test because of its limited mange. Recent longitudinal studies using the Mini-Mental State examination (1) for patients with Alzheimer s disease have found Mini-Mental State annual change scores markedly affected by baseline score (25, 26), with one study (25) showing an inverted- function similar to that observed with the cognitive subscale of the Alzheimcm s Disease Assessment Scale in the present study. Since the Mini-Mental State examination is different from the cognitive subscale of the Alzheimem s Disease Assessment Scale in many ways, this gives greater credibility to the hypothesis that the inverted- function me- 394 AmJ Psychiatry 151:3, March 1994

6 STERN, MOHS, DAVIDSON, El AL. flects the relation between symptomatic and neuropathologic change. The nature of the relation between neurodegenerative changes and cognitive decline in Alzheimer s disease has been a subject of intense inquiry at least since the studies of Blessed et al. (1 1 ), who correlated neuropathologic measures (senile plaques and neurofibmillamy tangles) with cognitive and behavioral measures. Their data showed a roughly linear relationship but were also consistent with the possibility that theme may be a significant amount of neuropathologic change with very little behavioral consequence (1 1 ). This might occur because of redundancy in the nervous system that enables the brain to compensate for mild to moderate degrees of neuronal damage. The present results are consistent with the view that after a certain degree of neuronal damage has occurred with little worsening of symptoms, each additional loss produces substantial cognitive deterioration. The slowing of the mate of change in severe dementia probably results from the fact that severely demented patients have so little remaining cognitive capacity that little additional deterioration can be observed. However, it is possible that the relation between symptomatic change and baseline severity observed in this study is a consequence of the specific items included in the cognitive subscale of the Alzheimcm s Disease Assessment Scale. In the absence of direct in vivo measures of neurodegeneration, we cannot definitively identify the relation between cognitive decline and neuronal deterioration. The effects observed in this study were not due to the particular method used to analyze the longitudinal data. While the exact estimates of rate of change did vary slightly according to the different methods of data analysis, the relation of mate of change to baseline score was qualitatively similar with all methods. We previously reported a similar result when these different analytic methods were used to calculate annual mate of cognitive change on the Blessed test (14). Also in agreement with our own previous data (14) and with most other studies (13, 23, 24, 26, 27), we found no effect of gender, age at onset, or family history on rate of change. We did, however, find that measured cognitive change was at least somewhat consistent over time for the same individual. One previous longitudinal study with the Blessed test (27) found no correlation of change in year 1 with change in year 2; that study used the 12-month data point in calculating both change scores, however, and thus may have underestimated the true correlation because of regression to the mean (22). The present results support the notion that some patients do deteriorate more rapidly than others, but at present we have no independent predictors of rate of disease progression. These results have important implications for the design of clinical treatment trials that use the cognitive subscale of the Alzheimem s Disease Assessment Scale or a similar instrument for the outcome variables. First, they suggest that theme is no need to stratify treatment groups on the basis of gender, age at onset, or family history, since these variables do not affect rate of change. Second, they indicate that matching of treatment groups on the basis of severity of cognitive impaimment is absolutely critical in treatment studies, since the expected change during the trial will depend heavily on baseline severity. Third, they indicate that the ability of instruments like the cognitive subscale of the Alzheimem s Disease Assessment Scale to detect treatment effects might be greater for patients with mild and moderate dementia than for patients with very mild and very severe dementia. Finally, the effect sizes obtained in this study for 6-month versus 12- month change scores confirm previously reported findings with the Blessed test (14) and the Mini-Mental State examination (28) that the reliability of measured change scores is greater when patients are followed for longer periods of time. Over the 12-month follow-up the effect size on the cognitive subscale of the Alzheimcm s Disease Assessment Scale was slightly greater than 1, while over the 6-month follow-up the effect size was substantially less than 1. A consequence is that the statistical power of clinical treatment trials is heavily dependent on the length of the trials. Trials lasting 1 year (active treatment versus placebo) will be able to detect a specified treatment effect (e.g., slows progression by one-half) with sample sizes less than onehalf those needed to detect the same drug effect in a 6-month study. Exact power calculations have been presented previously for the Blessed test (14). Studies using the cognitive subscale of the Alzheimer s Disease Assessment Scale, the Blessed test, or other cognitive performance measures will have to weigh the trade-off between conducting a long-term study with fewer patients or a short-term study with many more patients (14). REFERENCES 1. Ad Hoc FDA Dementia Assessment Task Force: Meeting report: antidementia drug assessment symposium. Neurobiol Aging 1991; 12: Guidelines for the Clinical Evaluation of Antidementia Drugs. Rockville, Md, Food and Drug Administration, Department of Health and Human Services, Heston LL, Mastri AR, Anderson E, White J: Dementia of the Alzheimer type: clinical genetics, natural history, and associated conditions. Arch Gen Psychiatry 1981; 38: Seltzer B, Sherwin I: A comparison of clinical features in earlyand late-onset primary degenerative dementia: one entity or two? Arch Neurol 1983; 4: S. Bowen DM, Allen SJ, Benton JS, Goodhardt MJ, Haan EA, Palmer AM, Sims NR, Smith CCT, Spillane JA, Esiri MM, Neary D, Snowdon JS, Wilcock GK, Davison AN: Biochemical assessment of serotonergic and cholinergic dysfunction and cerebral atrophy in Alzheimer s disease. J Neurochem 1983; 41: Huff FJ, Growdon JH, Corkin 5, Rosen lj: Age at onset and rate of progression of Alzheimer s disease. J Am Geriatr Soc 1987; 35: Zubenko GS, Huff FJ, Beyer J, Auerbach J, leply I: Familial risk of dementia associated with a biologic subtype of Alzheimer s disease. Arch Gen Psychiatry 1988; 45: Luchins DJ, Cohen D, Hanrahan P, Eisdorfer C, Paveza G, Ashford JW, Gorelick P, Hirschman R, Freels 5, Levy P, Semla I, Shaw H: Are there clinical differences between familial and non- AmJ Psychiatry 151:3, March

7 LONGITDINAL STDY OF ALZHEIMER S DISEASE familial Alzheimer s disease? Am J Psychiatry 1992; 149: Rosen WG, Mohs RC, Davis KL: A new rating scale for Alzheimer s disease. Am J Psychiatry 1984; 141: Folstein MF, Folstein SE, McHugh PR: Mini-Mental State : a practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 1975; 12: Blessed G, lomlinson BE, Roth M: The association between quantitative measures of dementia and of senile change in the cerebral grey matter of elderly subjects. Br J Psychiatry 1968; 114: Morris JC, Heyman A, Mohs RC, Hughes JP, van Belle G, Fillenbaum G, Mellits ED, Clark C: The Consortium to Establish a Registry for Alzheimer s Disease (CERAD), part I: clinical and neuropsychological assessment of Alzheimer s disease. Neurology 1989; 39: Thai, Grundman M, Klauber MR: Dementia: characteristics of a referral population and factors associated with progression. Neurology 1988; 38: Stern RG, Mohs RC, Bierer LM, Silverman JM, Schmeidler J, Davidson M, Davis KL: Deterioration on the Blessed test in ALtheimer s disease: longitudinal data and their implications for clinical trials and identification of subgroups. Psychiatry Res 1992; 42: Davis KL, Thai, Gamzu ER, Davis CS, Woolson RF, Gracon SI, Drachman DA, Schneider LS, Whitehouse PJ, Hoover IM, Morris JC, Kawas CH, Knopman DS, Earl NL, Kumar V, Doody RS, lacrine Collaborative Study Group: A double-blind, placebo-controiled multicenter study of tacrine for Alzheimer s disease. N EngI J Med 1992; 327: Clinical Research Working Group From the Pharmaceutical Industry on Dementia: Recommendations for clinical drug trials in dementia. Dementia 199; 1: Mohs RC: Assessment of cognitive symptoms in clinical studies of antidementia drugs. Japanese J Geriatric Psychiatry 1991; 2: Kramer-Ginsberg E, Mohs RC, Aryan M, Lobel D, Silverman J, Davidson M, Davis KL: Clinical predictors of course for Alzheimer patients in a longitudinal study: a preliminary report. Psychopharmacol Bull 1988; 24: Green CR, Mohs RC, Schmeidler J, Aryan M, Davis KL: Functional decline in Alzheimer s disease: a longitudinal study. J Am Geniatr Soc 1993; 41: McKhann G, Drachman D, Folstein M, Katzman R, Price D, Stadlan 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 ALtheimer s Disease. Neurology 1984; 34: Silverman JM, Breitner JCS, Mohs RC, Davis KL: Reliability of the family history method in genetic studies of Alzheimer s disease and related dementias. Am J Psychiatry 1986; 143: Cohen J, Cohen P: Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences. Hillsdale, NJ, Lawrence Erlbaum Associates, Katzman R, Brown I, Thai, Fuld PA, Aronson M, Butters N, Klauber MR, Wiederholt W, Pay M, Renbing X, Ooi WL, Hofstetter R, Terry RD: Comparison of rate of annual change of mental status score in four independent studies of patients with Alzheimer s disease. Ann Neurol 1988; 24: Ortof E, Crystal HA: Rate of progression of Alzheimer s disease. J Am Geriatr Soc 1989; 37: Morris JC, Edland S, Clark C, Galasko D, Koss E, Mohs R, van Belle G, Fillenbaum G, Heyman A: CERAD longitudinal assessment of probable Alzheimer s disease, I: rates of cognitive change. Neurology (in press) 26. Burns A, Jacoby R, Levy R: Progression ofcognitive impairment in Alzheimer s disease. J Am Geriatr Soc 1991; 39: Salmon DP, Thai, Butters N, Heindel WC: Longitudinal evaluation of dementia of the Alzheimer type: a comparison of 3 standardized mental status examinations. Neurology 199; 4: van Belle G, hlmann RF, Hughes JP, Larson EB: Reliability of estimates of changes in mental status test performance in senile dementia of the Alzheimer type. J Clin Epidemiol 199; 43: AmJ Psychiatry 151:3, March 1994