Memory impairments are among the earliest and most devastating clinical symptoms of dementia (Brandt & Munro, 2002
; Paulsen et al., 1995
; Salmon et al., 1989
). Consistent with previous studies (Butters, Delis, & Lucas, 1995
; Kramer et al., 1989
; Noe et al., 2004
; Pillon et al., 1993
), our findings indicate that different degenerative dementias have distinct patterns of memory impairments, probably reflecting different pathological substrates. AD patients displayed more severe impairments in learning and memory of both items' names and locations, regardless of mode of testing (immediate recall, delayed recall, yes/no recognition), or modality of stimulus presentation (visual or auditory) than PD and HD patients. PD patients needed fewer trials to learn the location of items than HD patients, but otherwise performed comparably with HD patients. Thus, the subcortical dementias (HD and PD) were more similar to one another in pattern of memory performance than they were to the prototypical cortical dementia (AD). Indeed, based on their characteristics of learning and memory, an HD patient was more likely to be misclassified as a PD patient than an AD patient, and a PD patient was more likely to be misclassified as having HD than AD. Our findings therefore provide further support for the cortical–subcortical dichotomy (at least at the level of the cognitive syndrome).
Specific measures of episodic memory, such as delayed free recall, were more effective than others in distinguishing cortical and subcortical patterns of performances. In addition, delayed recall of names of items, presented either verbally or visually, differentiated all three dementias better than did delayed recall of locations. Previous studies have demonstrated the deterioration of organization and content of semantic memory in AD patients (Chertkow & Bub, 1990
; Rohrer, Salmon, Wixted, & Paulsen, 1999
; Salmon, Heindel, & Lange, 1999
). The storage of new information into episodic memory is directly associated with, and dependent upon, the recovery of information from semantic memory (Steven, Takashi, & Roberto, 2007
; Tulving, 1995
). This may help explain why AD patients perform more poorly on delayed recall of items compared with delayed recall of locations. Moreover, memory of items presented visually may depend more critically on the inferior temporal lobe, where AD pathology abounds (Arnold, Hyman, Flory, Damasio, & Van Hoesen, 1991
), than does memory for spatial location. Why the same episodic memory measure distinguishes between the two subcortical dementias is less clear and warrants further investigation. Finally, the ROC curves revealed that the recognition index (a score that reflects the ability to identify target items and reject distractor items) also had discriminative power. This measure did not, however, remain in the final model in the stepwise discriminant analysis probably because it shared a considerable variance with other independent variables.
The superiority of delayed recall over recognition accuracy in distinguishing primarily cortical from primarily subcortical diseases has been debated. Several studies support the effectiveness of recognition accuracy in distinguishing between patients with cortical and subcortical dysfunction (Delis, Kramer, Kaplan, & Ober, 2000
; Massman, Delis, & Butters, 1993
). Whereas AD, PD, and HD patients typically perform poorly on delayed free recall, only HD or PD patients exhibit substantial improvement on measures of recognition discrimination (Butters et al., 1995
; Delis et al., 1991
; Deweer et al., 1994
; Glosser, Friedman, Grugan, Lee, & Grossman, 1998
). On the other hand, several studies have shown that both recall and recognition are compromised in HD and PD patients (Brandt & Munro, 2002
; Brandt et al., 2005
; Montoya et al., 2006
; Solomon et al., 2007
; Zizak et al., 2005
) and the best overall discriminator between groups with cortical and subcortical dementia is delayed recall (Noe et al., 2004
; Paulsen et al., 1995
; Welsh, Butters, Hughes, Mohs, & Heyman, 1991
). Our results suggest that both delayed recall and recognition differentiate distinct dementias, but delayed recall has greater discriminative power.
Two main implications, both theoretical and clinical, can be drawn from the present study. First, our findings provide insight into the nature of memory impairments in different degenerative diseases that can result in a dementia syndrome. Prominent disease-specific memory impairments that can be captured by brief memory tests add to the existent evidence that there are probably distinct underlying neurocognitive mechanisms that contribute to the differential impairment in the learning of new information. Second, our findings are unlikely to be useful for differential diagnosis; the distinctions among AD, HD, and PD are profound, and the diseases are not usually confused with each other. However, having HD or PD does not preclude the possibility that a patient could at some point also develop AD. It is well documented in neuropathological studies that coincident AD pathology may be present in PD patients (Boller, Mizutani, Roessmann, & Gambetti, 1980
; Braak et al., 1996
; Emre, 2003
). In a patient with a primary subcortical disease (PD and HD), a severe verbal-learning impairment may be indicative of superimposed AD pathology.
A major finding of the present study concerns the effectiveness of both the HVLT-R and the HB in differentiating AD, HD, and PD patients. In the present study, the inclusion of three variables from the HVLT-R and HB resulted in an 80% correct classification of the AD, HD, and PD patients, and four measures from the same tasks produced a 91% correct classification of diseases as primarily cortical or subcortical. Kramer and colleagues (1989)
reported 76% classification accuracy of AD, HD, and PD patients matched for their levels of immediate recall performances using nine measures of learning characteristics (e.g., rates of forgetting) and error types (intrusions) from the CVLT. Massman and colleagues (1993)
showed that the CVLT recognition discriminability index resulted in a correct classification rate of 90% between patients with AD and those with HD, whereas Delis and colleagues (1991)
reported around 85% accuracy in classifying the same patient groups. However, a more recent study by Delis and colleagues (2005)
showed that traditional recall conditions of the CVLT-II (both short and long delayed recall conditions, with and without cues) failed to differentiate AD and HD patients matched for dementia rating scale scores. The Buschke Selective Reminding Test produced a 68% accuracy of group classification of AD and HD subjects (Paulsen et al., 1995
). Finally, Tröster and colleagues (1993)
reported a 79% classification accuracy of AD and HD patients using indices of the WMS-R. Thus, the HVLT-R and the HB were as good as or better than other memory tests that require more time to administer in distinguishing different dementias.
Strengths of the present study are the large sample sizes, as well as the inclusion of three different dementias. The former allows us to conduct the appropriate discriminant analysis and the latter to test the hypothesis that the pattern of memory impairment is disease-specific as opposed to only reflecting differences in cortical and subcortical degenerative processes. A conceptual and methodological issue that remains is how best to match groups of patients with different dementias (Brandt & Munro, 2002
). The failure to match groups on demographic characteristics, such as age and education, and especially on dementia severity, or the application of different matching procedures may result in heterogeneous patient samples and contribute to the discrepant findings (Graham, Emery, & Hodges, 2004
). In our study, we defined dementia severity based on MMSE scores. Although it has received criticism, MMSE is still recommended and used as the primary instrument for tracking/screening for dementia and describing dementia severity (Dubois et al., 2007
). In addition, we followed two different methodological approaches that resulted in consistent findings. First, in the discriminant analysis, we included all patients, but controlled statistically for demographic differences and dementia severity. Second, in the ROC curves analyses, we selected subgroups matched for dementia severity and education. Both methods revealed not only group differences, but also distinct patterns of memory performances that predict group membership with very high accuracy. Thus, we conclude that our findings that AD, HD, and PD patients display different patterns of memory impairment are robust.
An important caveat when interpreting our findings is the diagnostic accuracy of the AD and PD diseases. Although we applied the standard clinical criteria to diagnose AD and PD, our patients had no autopsy that would pathologically confirm their diagnosis. Consequently, some AD patients might ultimately have a different type of dementing disease and some PD patients, especially those with major memory problems, might also have AD pathology. In addition, our HD and PD patients had various degrees of cognitive impairment, as can be inferred from their MMSE scores and their ratings on the functional scales. We opted against classifying patients with HD and PD as having mild cognitive impairment, dementia, or neither, because (at one level) they all have cognitive impairments in more than one domain and functional impairments and might therefore all have some degree of dementia. We thus consider such distinctions problematic (Brandt & Munro, 2002
In summary, AD, HD, and PD patients display differences in patterns of memory performances of such magnitude that can predict their disease-defining group membership. Delayed memory recall provides the highest discrimination accuracy especially when comparing primarily cortical with subcortical dementias. Finally, both the HVLT-R and the HB can accurately discriminate the three diseases.