Mild Sede Dementia of the Alzheimer Type: 3. bngit-u&nal and Cross-Sectional Assessment Leonard Berg, MD," Deborah S. Smith, PhD,? John C. Morris, MD,"$ J. Philip Miller, AB,? Eugene H. Rubin, MD, PhD,§ Martha Storandt, PhD," and Lawrence A. Coben, MD"
Sixty-six subjects diagnosed by validated criteria as having senile dementia of the Alzheimer type (SDAT) were assessed with clinical measures commonly used to study dementia. The severity of the SDAT was mild in 24, moderate in 24, and severe in 18. The data from these three groups in a cross-sectional study were compared with results in our earlier study of 43 subjects enrolled with mild SDAT and followed longitudinally. We concluded that the crosssectional method underestimates the severity of progression as indicated by some of the clinical measures. Additionally, the 24 new subjects with mild SDAT were assessed longitudinally. This replication study confirmed ow earlier conclusions that four of six clinical measures showed similar patterns over time and were useful throughout the study, global measures being more informative than brief individual measures with ceiling or floor effects. Berg L, Smith DS, Morris JC, Miller JP, Rubin EH, Storandt M, Coben LA. Mild senile dementia of the Alzheimer type: 3. Longitudinal and cross-sectional assessment. Ann Neurol 1990;28:648-652
In an earlier paper, Berg and colleagues [l}described the longitudinal course of a sample of 43 individuals with carefully defined mild senile dementia of the Alzheimer type (SDAT). The term SDAT was chosen to refer to the disorder diagnosed by clinical and laboratory criteria without biopsy or autopsy. The clinical course of SDAT from the mild through the moderate and then to the severe stage was described by using quantitative information obtained from repeated administrations of a semistructured clinical interview designed to diagnose dementia and rate its severity. Longitudinal studies are costly and slow to produce results. Therefore, the purpose of the present study was to determine whether similar results could be obtained in a cross-sectional study examining different individuals in each stage of dementia at the same time rather than following the same individuals as they progressed through the stages of increasing impairment. A second purpose was to replicate the original longitudinal study on a new sample followed over time.
original longitudinal study, that is, mainly notices in the media. There were 24 people (I0 men and 14 women) with mild SDAT, 24 (7 men and 1 7 women) with moderate SDAT, and 18 (4 men and 14 women) with severe SDAT. Inclusionary and exclusionary criteria were identical to those in the previous study [l). The demographic characteristics of the new sample were comparable with those of the original longitudinal sample. Age at entry ranged from 67 to 85 years (mean SD 75.4 2 5.1 yr). Those with moderate and severe dementia were 4.8 yr) than those with mild SDAT slightly older (76.6 (73.2 5 5.0 yr), as was also the case in the original longitudinal study for data obtained at later stages of the disease in individuals who were first seen in the mild stage. Education 3.7 yr), and social posiranged from 3 to 20 years (11.4 tion 12) ranged from 2 to 5 (3.7 i: 1.1). In addition to inclusion in the cross-sectional sample, the 24 people with mild SDAT were enrolled in a new longitudinal study. Time intervals between assessments were chosen to match those in the original longitudinal study. Three times of assessment have now been completed for this new group.
Methods In 1984 and 1985,66 individuals with SDAT were recruited from the general community by the same means used in the
specified in earlier papers and are compatible with the criteria for "probable Alzheimer's disease" proposed by McKhann and colleagues [4}.Ours have been validated [ 5 } by demonstration of pathological changes sufficient for the diag-
From the Alzheimer's Disease Research Cenrer, the *Department of Neurology and Neuroloical Surgery (Neurology), the ?Division of
Received Feb 2 1, 1990, and in revised form May 24. Accepted for publication May 26, 1990.
Sample
*
*
*
Diagnostic and Staging Criteria Our research diagnostic criteria [3f for SDAT have been
648 Copyright 0 1990 by the American Neurological Association
Table I . Mean 2 SD of Scores on Clinical Measures in Cross-Sectionaland Longitudinal Patients Grouped by Severity Measure
Method
CDR 1
CDR 2
Sum of Boxes
C L C L C L C L C L
6.0 6.3 6.0 5.4 3.3
12.3 ? 12.2 +11.2 2 11.1 f 6.7 ? 7.5 ? 8.5 & 8.4 2 8.7 f 10.2 -t-
DS DSC SPMSQ AB
3.6 5.8 5.4 3.8 3.6
& &
It
? ? % %
?
?
1.3 1.3 2.4 3.1 1.4
1.6 2.4 2.2 3.7 5.0
CDR 3
2.2 1.6 3.2
4.9 2.6 2.5 1.4 1.6 7.3 8.7
16.3 2 1.0 17.6 -t- 0.8 15.8 i 3.9 17.9 & 4.6 11.5 ? 2.4 13.7 -t- 3.2 7.2 ? 1.8 9.9 ? 0.2 23.3 2 7.4 30.7 -t- 5.8 ~
Statistics based on 76, 32, and 58 observations in the longitudinal CDR 1, 2, and 3 stages, respectively. There were 24, 24, and 18 observations in the three groups, respectively, in the cross-sectional study.
SD = standard deviation; CDR Scale Cognitive; SPMSQ
=
= Clinical Dementia Rating; C = cross-sectional;L = longitudinal; DS Short Portable Mental Status Questionnaire; AB = Aphasia Battery.
nosis of Alzheimer’s disease [6] in a consecutive series of 5 1 autopsies through December 3 1, 1789, following diagnosis of SDAT by these criteria. The severity of dementia was staged by the Washington University Clinical Dementia Rating (CDR), a quantitative global assessment of dementia {7-9}. The CDR was derived from a semistructured clinical interview with the patient and an appropriate collateral source. The clinician rated the patient’s cognitive function in each of six categories: memory, orientation, judgment and problem solving, function in community affairs, function in home and hobbies, and function in personal care. Impairment was rated in each category on a five-point scale (none = 0, questionable = 0.5, mild = 1, moderate = 2, severe = 3). From those six ratings, the global CDR was established by a simple published algorithm {7, 91 that results in a rating of no dementia (CDR 0), or questionable (CDR 0.5), mild (CDR l),moderate (CDR 2), or severe dementia (CDR 3). Using the current samples, interrater reliability on the CDR was good [lo] (Kendall’s tau B = 0.71; kappa = 0.74, weighted kappa = 0.87). (The present study used the version of the CDR published in 1784 [8]. The currently recommended CDR {7] differs slightly by elimination of certain ambiguities.)
Clinical Measures A global measure of dementia severity more quantitative than the CDR was provided by summing the ratings in each of the six cognitive domains (“Boxes”)from which the CDR was derived (“Sum of Boxes” El], maximum impairment = 6 x 3 = 18). The clinical interview included several well-known scales. The Blessed Dementia Scale (DS) is a behavioral checklist 111)derived exclusively from reports of the collateral source and rates the patient on performance of everyday activities, changes in personal care, and changes in personality, interests, and drive. The Blessed Dementia Scale Cognitive (DSC) is that portion of the DS that assesses cognitive function in everyday activities and personal care. The Short Portable Mental Status Questionnaire (SPMSQ) of Pfeiffer 1121 is a 10-item questionnaire directed to the patient. An Aphasia Battery (AB) [13] based on ver-
=
Dementia Scale; DSC
=
Dementia
bal tasks from the Boston Diagnostic Aphasia Evaluation 1143 was used to assess language function. The Face-Hand Test (FHT) was scored according to the method of Zarit and associates {15]. These six clinical measures (Sum of Boxes, DS, DSC, SPMSQ, AB, and FHT) were obtained at each assessment. A battery of psychometric tests { 16) was also administered at each clinical assessment, but psychometric results were not known to the clinician and did not enter into clinical ratings. All testing procedures and the menas of obtaining informed consent from both patients and responsible relatives were approved in advance by the Human Studies Committee of Washington University.
Analyses A two-way multivariate analysis of variance was used to compare data obtained from the longitudinal and cross-sectional methods. Independent variables were stage of dementia (mild, moderate, or severe) and method (longitudinal vs. cross-sectional). Dependent variables are listed in Table 1. The Face-Hand Test was not included because of extensive missing data. We have found it difficult to standardize the administration and scoring of this procedure, and no longer include it as part of the clinical assessment. Data included in this analysis were from the first time of assessment of the cross-sectional sample of patients recruited in 1984 and 1985, and from all times of assessment for the original longitudinal sample of patients if they were seen for clinical assessment and assigned a CDR and Sum of Boxes; that is, individuals from the original longitudinal sample were represented multiple times as they progressed through the stages of the disease. This analysis is comparable with that reported by Botwinick and his colleagues [l?) for the psychometric test data obtained from these same samples of patients. Missing values on selected dependent variables were handled in the following ways to reduce the number of patients who would have to be eliminated from the multivariate analysis. A maximum value was assigned if the individual was unable to respond (SPMSQ and AB). Maximum SPMSQ values were assigned for 1 of the 32 CDR 2 and for 10 of the
Berg et al: Clinical Assessment of SDAT 649
58 CDR 3 observations in the longitudinal group. Eight of the longitudinal CDR 3 observations were also adjusted to maximum values for the AB. None of the observations on cross-sectional patients was missing SPMSQ or AB. For the DS and DSC, cell medians ( 5 of 232 observations) were used if a collateral source was not available. The choice of cell medians was deemed appropriate in that the values were missing because of unavailability of collateral source rather than a patient’s inability to respond. Nonparametric analyses using both rank and normal scores provided the same results as those from the analysis listed above. The rank transformation approach involves the replacement of raw scores with either rank or normal scores, followed by the usual F tests [18}. Only the statistics from the parametric analysis are reported. All computations were performed with Statistical Analysis System (SAS Institute, Inc, Cary, NC) 6.03 C191.
Results Cross-sectional Versus Longitudinal Methods Mean scores for the longitudinal and cross-sectional samples at each CDR (severity) level are shown in Table 1. The multivariate analysis of variance revealed significant main effects of both method (F 15,222) = 3.89, p = 0.002) and CDR (F 110, 444} = 125.60, p = 0.0001). The CDR-by-X method interaction was of major interest. Does the course (progression) of the disease look the same in a cross-sectional study as it does in longitudinal design? This interaction was significant in the multivariate analysis ( F [lo, 4441 = 1.89, p = 0.04). Univariate analyses revealed significant interactions on two variables, Sum of Boxes (F 12, 2261 = 3.69, p = 0.03) and Aphasia Battery (F 12, 2261 = 6.14, p = 0.003). Post hoc comparisons revealed that the interaction was due to the significantly poorer performance of the severely demented longitudinal patients compared with the severely demented cross-sectional group for both measures ( t s 1741 = 4.39, p < 0.0001, and 3.47, p = 0.0006, respectively). This explains the significant main effect of method obtained in the multivariate analysis. The significant main effect of CDR was expected, of course. Performance on all measures declined with increased severity. Replication of Longitudinal Study The procedures for displaying data were used in accordance with those in the previous study [11. Percentile scores (75th, SOth, and 25th) were calculated for the clinical measures at each time of testing. (The FaceHand Test is included for completeness.) The problem of missing data during follow-up was handled by first providing data only on those patients who were tested and then assigning maximum scores (greatest impairment) to those patients no longer avdable for assessment. Rationale for this procedure was provided else-
Table 2. Number of Patients Alive and Numbws Rated (Second Longitudinal StudylOriginal Longitudinal Study) Months after Entry
Alive CDR Sum of boxes
5 measures
0
15
34
24/43 24/43 24/43 24/43
2014 3 2014 1 19/38 17/33
19/37 19136 18/33 15/28
CDR = Clinical Dementia Rating.
where 111. For each time of testing, Table 2 provides the numbers of patients still alive and rated on CDR and Sum of Boxes, and the mean numbers of patients rated on the other five measures (BDS, BDSC, SPMSQ, AB, and FHT). Generally, when patients were missing one measure, they were missing all measures at a given time of testing. Table 3 provides percentile scores measured in tested patients and adjusted percentile scores based on the assignment of maximal scores to untested (missing) patients. The range of possible scores for each measure is given in brackets. In each instance, larger scores mean more impairment. The changes in the percentile scores over time depict the progressive nature of the impairment in SDAT with increasing numbers of patients reaching ceiling or maximum impairment. Repeated measures analyses of variance compared each of the six clinical measures (adjusted and nonadjusted scores) between the original study and the current study across the three time points. None of these analyses demonstrated any difference between studies nor any interactions between study and time period (all p > 0.10). Table 3 also reports the semiinterquartile range for each measure at each time of assessment. This robust measure of dispersion, equal to half the difference between the 75th and 25th percentile scores, indicates for most measures that variability was small at entry but increased at subsequent times of assessment.
Discussion The comparison of the longitudinal and cross-sectional methods leads to the conclusion that the latter may underestimate the severity of the progression. This conclusion is logical because, compared with deterioration in a longitudinal study, only the less deteriorated patients, especially at CDR 3, are likely to be enrolled in a cross-sectional study. Extensive efforts are invested in a longitudinal study to retain even the most severely demented individuals. Based on the longitudinal versus cross-sectional analyses of psychometric data, Botwinick and associates [17) concluded that “relatively little seems to be lost in carrying out cross-sectional investigations rather
650 Annals of Neurology Vol 28 No 5 November 1990
Table 3. Measured (and Adjusted) Percentile Scores and Semiinterpartile Ranges for 24 SDAT Patients, Mild at Entry Months after Entry Measures
Percentiles
15
34
7 6
12(14.8) 10(10.5)
5
(37) 3(3.9) 13.5(24.9) 8.5(12) 5.9(6.6) 3.8(7.1) 7.5(15.6) 4.5(6.5) 3.0(3.6) 2.3(6) 10(10) 8(7.5) 4(7) 3(1.5) 9.6(35) 2.2(5.9) l(1.2) 4.3u6.9) 16(16)
17.3(18) 13(16) 10.3(11) 3.5(3.5) 14.5(28) 12(14) 7.5(8.8) 3.5(9.6) 9(17) 7(8.8) 4.8(5.3) 2.1(5.9) lO(10) W0) 5(8) 2.5(1) 18.3(35) 8(21.4) l(4.7) 8.7(15.2) 16(16) 406) 0.5(2.5) 7.8(6.8)
0
~
Sum of Boxes [0-l8y
DS C0-281” DSC [0-171”
SPMSQ [0-101”
AB 10-351”
FHT E0-161”
75th 50th 25th SIR 75th 50th 25th SIR 75th 50th 25th SIR 75th 50th 25th SIR 75th 50th 25th SIR 75th
50th 25th SIR
1 7.9 5.5 4.6 1.6 4.4 3 2.5 0.9 7.8 5.5 4 1.9 5.2 2.5 0.7 2.2 15 2 1 7
3(3.3) 6.5(6.4)
~
“Values withm the brackets represent the range of possible scores for each measure, from no impairment (0) to maximal impairment. SDAT = senile dementia of the Alzheimer type; SIR = semiinterquartile range; DS = Dementia Scale; DSC = Dementia Scale Cognitive; SPMSQ = Short Portable Mental Status Questionnaire; AB = Aphasia Battery; FHT = Face-Hand Test.
than the more difficult, costly and time-consuming longitudinal studies.” The difference between the results in that analysis and the present one probably is accountable to the fact that patients reached ceiling or floor on the psychometric measures more readily than on several of the clinical measures. The results of the second longitudinal study replicate the main conclusions of the original study. Progressive impairment on all clinical measures constitutes the hallmark of the deterioration in SDAT. As in the original study, four of the clinical measures used (Sum of Boxes, DS, DSC, and AB) revealed remarkably similar patterns over time. They were useful throughout the course of the studies, and ceiling effects were observed only when the patients were profoundly demented. In contrast, there was wide variability in performance initially on the FHT, and many of the subjects rapidly reached ceiling on the SPMSQ. This replication study confirms the earlier conclusion El} that a global measure of dementia such as the Sum of Boxes appears more useful for following the natural history of SDAT than brief individual measures that have ceiling or floor effects. An alternate strategy would be to form a composite of multiple individual measures. Finally, it is reassuring that the results are similar in cohorts recruited 5 years apart.
Our longitudinal findings are robust and were similar in two samples of patients recruited 5 years apart. Cross-sectional data can approximate longitudinal data; however, they are not identical and must be interpreted with caution. Cross-sectional studies may underestimate the magnitude of deterioration in severely demented individuals. Addendum To date (9/4/90)among 64 autopsies on patients diagnosed as SDAT, the diagnosis has been confirmed in 62. Two patients have been found to have a non-Alzheimer dementia. Supported in part by Grant MH 31054 from the National Institute of Mental Health, and Grants AG 03991 and AG 05681 from the National Institute on Aging. We thank the patients and their families, Elizabeth Grant, PhD, for help with data management, and Patti Nacci for expertly preparing the manuscript.
References 1. Berg L, Miller JP, Storandt M, et al. Mild senile dementia of the
Alzheimer type: 2. Longitudinal assessment. Ann Neurol 1988;23 :4?7-484 2. Hollingshead AB. Two Factor Index of Social Position. New Haven, CT:AB Hollingshead, 1957
Berg et al: Clinical Assessment of SDAT 651
3. Berg L, Hughes CP, Coben LA, et al. Mild senile dementia of
4. 5.
6. 7. 8. 9. 10.
11.
Alzheimer type: research diagnostic criteria, recruitment, and description of a study population. J Neurol Neurosurg Psychiatry 1982;45:962-968 McKhann G, Drachman D, Folstein M, et al. Clinical diagnosis of Alzheimer’s disease. Neurology 1984;34:939-944 MorrisJC, McKeel D W Jr, Fulling K, et al. Validation of clinical diagnostic criteria for Alzheimer’s disease. Ann Neurul 1988; 24117-22 Khachaturian 2s. Diagnosis of Alzheimer’s disease. Arch Neurol 1985;42:1097- 1105 Hughes CP, Berg L, Danziger WL, et al. A new clinical scale for the staging of dementia. Br J Psychiatry 1982;140:566-572 Berg L. Clinical dementia rating. Br J Psychiatry 1984;145:33‘) Berg L. Clinical dementia rating. Psychopharmacol Bull 1988; 241637-639 Burke WJ, Miller JP, Rubin EH, et al. Reliability of the Washington University Clinical Dementia Rating (CDR). Arch Neurol 1988;45:31-32 Blessed G, Tomlinson BE, Roth M. The association betwren quantitative measures of dementia and of senile change in the cerebral grey matter of elderly subjects. Br J Psychiatry 1968;114:797-811
12. Pfeiffer E. A short portable mental status questionnaire for the assessment of organic brain deficit in elderly patients. J Am Geriatr Soc 1975;23:433-44 1 13. Faber-Langendoen K, Morris JC, Knesevich JW, et al. Aphasia in senile dementia of the Alzheimer type. Ann Neurol 1988; 231365-370 14. Goodglass H, Kaplan E. The Assessment of Aphasia and Related Disorders. 2nd ed. Phladelphia: Lea and Febiger, 1983, Appendix 1-28 15. Zarit SH, Miller NE, Kahn RL.. Brain function, intellectual impairment and education in the aged. J Am Geriatr Soc 1978; 26:58-67 16. Storandt M, Botwinick J, Danziger WL, et al. Psychometric differentiation of mild senile dementia of the Alzheimer type. Arch Neurol 1984;41:497-499 17. Botwinick J, Storandt M, Berg L, Boland S. Subject attrition and testability in research. Senile dementia of the Alzheimer type. Arch Neurol 1988;45:493-496 18. Conover WJ. Iman RL. Rank transformations as a bridge between parametric and nonparametric statistics. The American Statistician 1981;35:124-129 19. SAS Institute Inc. SASSTAT User’s Guide. Version 6, 4th ed. Cary, NC: SAS Institute, 1989
652 Annals of Neurology Vol 28 No 5 November 1990
Sixty-six subjects diagnosed by validated criteria as having senile dementia of the Alzheimer type (SDAT) were assessed with clinical measures commonly used to study dementia. The severity of the SDAT was mild in 24, moderate in 24, and severe in 18. The data from these three groups in a cross-sectional study were compared with results in our earlier study of 43 subjects enrolled with mild SDAT and followed longitudinally. We concluded that the crosssectional method underestimates the severity of progression as indicated by some of the clinical measures. Additionally, the 24 new subjects with mild SDAT were assessed longitudinally. This replication study confirmed ow earlier conclusions that four of six clinical measures showed similar patterns over time and were useful throughout the study, global measures being more informative than brief individual measures with ceiling or floor effects. Berg L, Smith DS, Morris JC, Miller JP, Rubin EH, Storandt M, Coben LA. Mild senile dementia of the Alzheimer type: 3. Longitudinal and cross-sectional assessment. Ann Neurol 1990;28:648-652
In an earlier paper, Berg and colleagues [l}described the longitudinal course of a sample of 43 individuals with carefully defined mild senile dementia of the Alzheimer type (SDAT). The term SDAT was chosen to refer to the disorder diagnosed by clinical and laboratory criteria without biopsy or autopsy. The clinical course of SDAT from the mild through the moderate and then to the severe stage was described by using quantitative information obtained from repeated administrations of a semistructured clinical interview designed to diagnose dementia and rate its severity. Longitudinal studies are costly and slow to produce results. Therefore, the purpose of the present study was to determine whether similar results could be obtained in a cross-sectional study examining different individuals in each stage of dementia at the same time rather than following the same individuals as they progressed through the stages of increasing impairment. A second purpose was to replicate the original longitudinal study on a new sample followed over time.
original longitudinal study, that is, mainly notices in the media. There were 24 people (I0 men and 14 women) with mild SDAT, 24 (7 men and 1 7 women) with moderate SDAT, and 18 (4 men and 14 women) with severe SDAT. Inclusionary and exclusionary criteria were identical to those in the previous study [l). The demographic characteristics of the new sample were comparable with those of the original longitudinal sample. Age at entry ranged from 67 to 85 years (mean SD 75.4 2 5.1 yr). Those with moderate and severe dementia were 4.8 yr) than those with mild SDAT slightly older (76.6 (73.2 5 5.0 yr), as was also the case in the original longitudinal study for data obtained at later stages of the disease in individuals who were first seen in the mild stage. Education 3.7 yr), and social posiranged from 3 to 20 years (11.4 tion 12) ranged from 2 to 5 (3.7 i: 1.1). In addition to inclusion in the cross-sectional sample, the 24 people with mild SDAT were enrolled in a new longitudinal study. Time intervals between assessments were chosen to match those in the original longitudinal study. Three times of assessment have now been completed for this new group.
Methods In 1984 and 1985,66 individuals with SDAT were recruited from the general community by the same means used in the
specified in earlier papers and are compatible with the criteria for "probable Alzheimer's disease" proposed by McKhann and colleagues [4}.Ours have been validated [ 5 } by demonstration of pathological changes sufficient for the diag-
From the Alzheimer's Disease Research Cenrer, the *Department of Neurology and Neuroloical Surgery (Neurology), the ?Division of
Received Feb 2 1, 1990, and in revised form May 24. Accepted for publication May 26, 1990.
Sample
*
*
*
Diagnostic and Staging Criteria Our research diagnostic criteria [3f for SDAT have been
648 Copyright 0 1990 by the American Neurological Association
Table I . Mean 2 SD of Scores on Clinical Measures in Cross-Sectionaland Longitudinal Patients Grouped by Severity Measure
Method
CDR 1
CDR 2
Sum of Boxes
C L C L C L C L C L
6.0 6.3 6.0 5.4 3.3
12.3 ? 12.2 +11.2 2 11.1 f 6.7 ? 7.5 ? 8.5 & 8.4 2 8.7 f 10.2 -t-
DS DSC SPMSQ AB
3.6 5.8 5.4 3.8 3.6
& &
It
? ? % %
?
?
1.3 1.3 2.4 3.1 1.4
1.6 2.4 2.2 3.7 5.0
CDR 3
2.2 1.6 3.2
4.9 2.6 2.5 1.4 1.6 7.3 8.7
16.3 2 1.0 17.6 -t- 0.8 15.8 i 3.9 17.9 & 4.6 11.5 ? 2.4 13.7 -t- 3.2 7.2 ? 1.8 9.9 ? 0.2 23.3 2 7.4 30.7 -t- 5.8 ~
Statistics based on 76, 32, and 58 observations in the longitudinal CDR 1, 2, and 3 stages, respectively. There were 24, 24, and 18 observations in the three groups, respectively, in the cross-sectional study.
SD = standard deviation; CDR Scale Cognitive; SPMSQ
=
= Clinical Dementia Rating; C = cross-sectional;L = longitudinal; DS Short Portable Mental Status Questionnaire; AB = Aphasia Battery.
nosis of Alzheimer’s disease [6] in a consecutive series of 5 1 autopsies through December 3 1, 1789, following diagnosis of SDAT by these criteria. The severity of dementia was staged by the Washington University Clinical Dementia Rating (CDR), a quantitative global assessment of dementia {7-9}. The CDR was derived from a semistructured clinical interview with the patient and an appropriate collateral source. The clinician rated the patient’s cognitive function in each of six categories: memory, orientation, judgment and problem solving, function in community affairs, function in home and hobbies, and function in personal care. Impairment was rated in each category on a five-point scale (none = 0, questionable = 0.5, mild = 1, moderate = 2, severe = 3). From those six ratings, the global CDR was established by a simple published algorithm {7, 91 that results in a rating of no dementia (CDR 0), or questionable (CDR 0.5), mild (CDR l),moderate (CDR 2), or severe dementia (CDR 3). Using the current samples, interrater reliability on the CDR was good [lo] (Kendall’s tau B = 0.71; kappa = 0.74, weighted kappa = 0.87). (The present study used the version of the CDR published in 1784 [8]. The currently recommended CDR {7] differs slightly by elimination of certain ambiguities.)
Clinical Measures A global measure of dementia severity more quantitative than the CDR was provided by summing the ratings in each of the six cognitive domains (“Boxes”)from which the CDR was derived (“Sum of Boxes” El], maximum impairment = 6 x 3 = 18). The clinical interview included several well-known scales. The Blessed Dementia Scale (DS) is a behavioral checklist 111)derived exclusively from reports of the collateral source and rates the patient on performance of everyday activities, changes in personal care, and changes in personality, interests, and drive. The Blessed Dementia Scale Cognitive (DSC) is that portion of the DS that assesses cognitive function in everyday activities and personal care. The Short Portable Mental Status Questionnaire (SPMSQ) of Pfeiffer 1121 is a 10-item questionnaire directed to the patient. An Aphasia Battery (AB) [13] based on ver-
=
Dementia Scale; DSC
=
Dementia
bal tasks from the Boston Diagnostic Aphasia Evaluation 1143 was used to assess language function. The Face-Hand Test (FHT) was scored according to the method of Zarit and associates {15]. These six clinical measures (Sum of Boxes, DS, DSC, SPMSQ, AB, and FHT) were obtained at each assessment. A battery of psychometric tests { 16) was also administered at each clinical assessment, but psychometric results were not known to the clinician and did not enter into clinical ratings. All testing procedures and the menas of obtaining informed consent from both patients and responsible relatives were approved in advance by the Human Studies Committee of Washington University.
Analyses A two-way multivariate analysis of variance was used to compare data obtained from the longitudinal and cross-sectional methods. Independent variables were stage of dementia (mild, moderate, or severe) and method (longitudinal vs. cross-sectional). Dependent variables are listed in Table 1. The Face-Hand Test was not included because of extensive missing data. We have found it difficult to standardize the administration and scoring of this procedure, and no longer include it as part of the clinical assessment. Data included in this analysis were from the first time of assessment of the cross-sectional sample of patients recruited in 1984 and 1985, and from all times of assessment for the original longitudinal sample of patients if they were seen for clinical assessment and assigned a CDR and Sum of Boxes; that is, individuals from the original longitudinal sample were represented multiple times as they progressed through the stages of the disease. This analysis is comparable with that reported by Botwinick and his colleagues [l?) for the psychometric test data obtained from these same samples of patients. Missing values on selected dependent variables were handled in the following ways to reduce the number of patients who would have to be eliminated from the multivariate analysis. A maximum value was assigned if the individual was unable to respond (SPMSQ and AB). Maximum SPMSQ values were assigned for 1 of the 32 CDR 2 and for 10 of the
Berg et al: Clinical Assessment of SDAT 649
58 CDR 3 observations in the longitudinal group. Eight of the longitudinal CDR 3 observations were also adjusted to maximum values for the AB. None of the observations on cross-sectional patients was missing SPMSQ or AB. For the DS and DSC, cell medians ( 5 of 232 observations) were used if a collateral source was not available. The choice of cell medians was deemed appropriate in that the values were missing because of unavailability of collateral source rather than a patient’s inability to respond. Nonparametric analyses using both rank and normal scores provided the same results as those from the analysis listed above. The rank transformation approach involves the replacement of raw scores with either rank or normal scores, followed by the usual F tests [18}. Only the statistics from the parametric analysis are reported. All computations were performed with Statistical Analysis System (SAS Institute, Inc, Cary, NC) 6.03 C191.
Results Cross-sectional Versus Longitudinal Methods Mean scores for the longitudinal and cross-sectional samples at each CDR (severity) level are shown in Table 1. The multivariate analysis of variance revealed significant main effects of both method (F 15,222) = 3.89, p = 0.002) and CDR (F 110, 444} = 125.60, p = 0.0001). The CDR-by-X method interaction was of major interest. Does the course (progression) of the disease look the same in a cross-sectional study as it does in longitudinal design? This interaction was significant in the multivariate analysis ( F [lo, 4441 = 1.89, p = 0.04). Univariate analyses revealed significant interactions on two variables, Sum of Boxes (F 12, 2261 = 3.69, p = 0.03) and Aphasia Battery (F 12, 2261 = 6.14, p = 0.003). Post hoc comparisons revealed that the interaction was due to the significantly poorer performance of the severely demented longitudinal patients compared with the severely demented cross-sectional group for both measures ( t s 1741 = 4.39, p < 0.0001, and 3.47, p = 0.0006, respectively). This explains the significant main effect of method obtained in the multivariate analysis. The significant main effect of CDR was expected, of course. Performance on all measures declined with increased severity. Replication of Longitudinal Study The procedures for displaying data were used in accordance with those in the previous study [11. Percentile scores (75th, SOth, and 25th) were calculated for the clinical measures at each time of testing. (The FaceHand Test is included for completeness.) The problem of missing data during follow-up was handled by first providing data only on those patients who were tested and then assigning maximum scores (greatest impairment) to those patients no longer avdable for assessment. Rationale for this procedure was provided else-
Table 2. Number of Patients Alive and Numbws Rated (Second Longitudinal StudylOriginal Longitudinal Study) Months after Entry
Alive CDR Sum of boxes
5 measures
0
15
34
24/43 24/43 24/43 24/43
2014 3 2014 1 19/38 17/33
19/37 19136 18/33 15/28
CDR = Clinical Dementia Rating.
where 111. For each time of testing, Table 2 provides the numbers of patients still alive and rated on CDR and Sum of Boxes, and the mean numbers of patients rated on the other five measures (BDS, BDSC, SPMSQ, AB, and FHT). Generally, when patients were missing one measure, they were missing all measures at a given time of testing. Table 3 provides percentile scores measured in tested patients and adjusted percentile scores based on the assignment of maximal scores to untested (missing) patients. The range of possible scores for each measure is given in brackets. In each instance, larger scores mean more impairment. The changes in the percentile scores over time depict the progressive nature of the impairment in SDAT with increasing numbers of patients reaching ceiling or maximum impairment. Repeated measures analyses of variance compared each of the six clinical measures (adjusted and nonadjusted scores) between the original study and the current study across the three time points. None of these analyses demonstrated any difference between studies nor any interactions between study and time period (all p > 0.10). Table 3 also reports the semiinterquartile range for each measure at each time of assessment. This robust measure of dispersion, equal to half the difference between the 75th and 25th percentile scores, indicates for most measures that variability was small at entry but increased at subsequent times of assessment.
Discussion The comparison of the longitudinal and cross-sectional methods leads to the conclusion that the latter may underestimate the severity of the progression. This conclusion is logical because, compared with deterioration in a longitudinal study, only the less deteriorated patients, especially at CDR 3, are likely to be enrolled in a cross-sectional study. Extensive efforts are invested in a longitudinal study to retain even the most severely demented individuals. Based on the longitudinal versus cross-sectional analyses of psychometric data, Botwinick and associates [17) concluded that “relatively little seems to be lost in carrying out cross-sectional investigations rather
650 Annals of Neurology Vol 28 No 5 November 1990
Table 3. Measured (and Adjusted) Percentile Scores and Semiinterpartile Ranges for 24 SDAT Patients, Mild at Entry Months after Entry Measures
Percentiles
15
34
7 6
12(14.8) 10(10.5)
5
(37) 3(3.9) 13.5(24.9) 8.5(12) 5.9(6.6) 3.8(7.1) 7.5(15.6) 4.5(6.5) 3.0(3.6) 2.3(6) 10(10) 8(7.5) 4(7) 3(1.5) 9.6(35) 2.2(5.9) l(1.2) 4.3u6.9) 16(16)
17.3(18) 13(16) 10.3(11) 3.5(3.5) 14.5(28) 12(14) 7.5(8.8) 3.5(9.6) 9(17) 7(8.8) 4.8(5.3) 2.1(5.9) lO(10) W0) 5(8) 2.5(1) 18.3(35) 8(21.4) l(4.7) 8.7(15.2) 16(16) 406) 0.5(2.5) 7.8(6.8)
0
~
Sum of Boxes [0-l8y
DS C0-281” DSC [0-171”
SPMSQ [0-101”
AB 10-351”
FHT E0-161”
75th 50th 25th SIR 75th 50th 25th SIR 75th 50th 25th SIR 75th 50th 25th SIR 75th 50th 25th SIR 75th
50th 25th SIR
1 7.9 5.5 4.6 1.6 4.4 3 2.5 0.9 7.8 5.5 4 1.9 5.2 2.5 0.7 2.2 15 2 1 7
3(3.3) 6.5(6.4)
~
“Values withm the brackets represent the range of possible scores for each measure, from no impairment (0) to maximal impairment. SDAT = senile dementia of the Alzheimer type; SIR = semiinterquartile range; DS = Dementia Scale; DSC = Dementia Scale Cognitive; SPMSQ = Short Portable Mental Status Questionnaire; AB = Aphasia Battery; FHT = Face-Hand Test.
than the more difficult, costly and time-consuming longitudinal studies.” The difference between the results in that analysis and the present one probably is accountable to the fact that patients reached ceiling or floor on the psychometric measures more readily than on several of the clinical measures. The results of the second longitudinal study replicate the main conclusions of the original study. Progressive impairment on all clinical measures constitutes the hallmark of the deterioration in SDAT. As in the original study, four of the clinical measures used (Sum of Boxes, DS, DSC, and AB) revealed remarkably similar patterns over time. They were useful throughout the course of the studies, and ceiling effects were observed only when the patients were profoundly demented. In contrast, there was wide variability in performance initially on the FHT, and many of the subjects rapidly reached ceiling on the SPMSQ. This replication study confirms the earlier conclusion El} that a global measure of dementia such as the Sum of Boxes appears more useful for following the natural history of SDAT than brief individual measures that have ceiling or floor effects. An alternate strategy would be to form a composite of multiple individual measures. Finally, it is reassuring that the results are similar in cohorts recruited 5 years apart.
Our longitudinal findings are robust and were similar in two samples of patients recruited 5 years apart. Cross-sectional data can approximate longitudinal data; however, they are not identical and must be interpreted with caution. Cross-sectional studies may underestimate the magnitude of deterioration in severely demented individuals. Addendum To date (9/4/90)among 64 autopsies on patients diagnosed as SDAT, the diagnosis has been confirmed in 62. Two patients have been found to have a non-Alzheimer dementia. Supported in part by Grant MH 31054 from the National Institute of Mental Health, and Grants AG 03991 and AG 05681 from the National Institute on Aging. We thank the patients and their families, Elizabeth Grant, PhD, for help with data management, and Patti Nacci for expertly preparing the manuscript.
References 1. Berg L, Miller JP, Storandt M, et al. Mild senile dementia of the
Alzheimer type: 2. Longitudinal assessment. Ann Neurol 1988;23 :4?7-484 2. Hollingshead AB. Two Factor Index of Social Position. New Haven, CT:AB Hollingshead, 1957
Berg et al: Clinical Assessment of SDAT 651
3. Berg L, Hughes CP, Coben LA, et al. Mild senile dementia of
4. 5.
6. 7. 8. 9. 10.
11.
Alzheimer type: research diagnostic criteria, recruitment, and description of a study population. J Neurol Neurosurg Psychiatry 1982;45:962-968 McKhann G, Drachman D, Folstein M, et al. Clinical diagnosis of Alzheimer’s disease. Neurology 1984;34:939-944 MorrisJC, McKeel D W Jr, Fulling K, et al. Validation of clinical diagnostic criteria for Alzheimer’s disease. Ann Neurul 1988; 24117-22 Khachaturian 2s. Diagnosis of Alzheimer’s disease. Arch Neurol 1985;42:1097- 1105 Hughes CP, Berg L, Danziger WL, et al. A new clinical scale for the staging of dementia. Br J Psychiatry 1982;140:566-572 Berg L. Clinical dementia rating. Br J Psychiatry 1984;145:33‘) Berg L. Clinical dementia rating. Psychopharmacol Bull 1988; 241637-639 Burke WJ, Miller JP, Rubin EH, et al. Reliability of the Washington University Clinical Dementia Rating (CDR). Arch Neurol 1988;45:31-32 Blessed G, Tomlinson BE, Roth M. The association betwren quantitative measures of dementia and of senile change in the cerebral grey matter of elderly subjects. Br J Psychiatry 1968;114:797-811
12. Pfeiffer E. A short portable mental status questionnaire for the assessment of organic brain deficit in elderly patients. J Am Geriatr Soc 1975;23:433-44 1 13. Faber-Langendoen K, Morris JC, Knesevich JW, et al. Aphasia in senile dementia of the Alzheimer type. Ann Neurol 1988; 231365-370 14. Goodglass H, Kaplan E. The Assessment of Aphasia and Related Disorders. 2nd ed. Phladelphia: Lea and Febiger, 1983, Appendix 1-28 15. Zarit SH, Miller NE, Kahn RL.. Brain function, intellectual impairment and education in the aged. J Am Geriatr Soc 1978; 26:58-67 16. Storandt M, Botwinick J, Danziger WL, et al. Psychometric differentiation of mild senile dementia of the Alzheimer type. Arch Neurol 1984;41:497-499 17. Botwinick J, Storandt M, Berg L, Boland S. Subject attrition and testability in research. Senile dementia of the Alzheimer type. Arch Neurol 1988;45:493-496 18. Conover WJ. Iman RL. Rank transformations as a bridge between parametric and nonparametric statistics. The American Statistician 1981;35:124-129 19. SAS Institute Inc. SASSTAT User’s Guide. Version 6, 4th ed. Cary, NC: SAS Institute, 1989
652 Annals of Neurology Vol 28 No 5 November 1990
