Medial temporal lobe atrophy on MRI in dementia with Lewy bodies

Methods.

MTA rating.

A standardized scale15 was used to rate left and right MTA from hard copies of T1-weighted coronal images. This scale rates atrophy as 0, absent; 1, minimal; 2, mild; 3, moderate; and 4, severe based on the width of the surrounding CSF spaces (temporal horn and choroid fissure) and the height of the hippocampal formation (which includes the hippocampus proper, subiculum, and parahippocampal and dentate gyri). Examples of rating are shown in the figure. All scans were rated blindly by consensus between three experienced raters (J.O., A.G., P.S.). FIGURE

• Download figure
• Open in new tab
• Download powerpoint

Figure. T1-weighted MR coronal images orientated perpendicular to the long axis of the hippocampus (slice thickness, 5 mm) and showing medial temporal lobe atrophy (MTA) rating scores. (A) MTA score, 1; minimal atrophy. (B) MTA score, 2; mild atrophy. (C) MTA score, 3; moderate atrophy. (D) MTA score, 4; severe atrophy.

Results.

Demographic and clinical characteristics of individuals are summarized in table 1. Groups were comparable for age, sex, length of history, and years of education. As would be expected, CAMCOG and MMSE scores were significantly lower in all dementia groups compared with control subjects (p < 0.001). The DLB group was significantly more impaired than VaD patients on MMSE (13.5 versus 18.0; p < 0.05) and CAMCOG (46.1 versus 61.7; p < 0.05).

Rating of MTA was symmetric (i.e., same score on both left and right) in most individuals (n = 94/104, 90%). For the purposes of analysis, left and right scores were summed to give an overall combined MTA score, ranging from 0 to 8, as shown in table 2. There were no differences in MTA scores between men and women in any group.

All dementia groups had significantly higher MTA scores compared with control subjects (p < 0.001). MTA scores were significantly lower in DLB compared with AD patients (mean MTA scores, 3.6 versus 5.2; Mann–Whitney U test; p = 0.002), and a trend was seen for lower scores in DLB compared with VaD patients (mean MTA scores, 3.6 versus 4.4; Mann–Whitney U test; p = 0.07).

All but two control subjects had an MTA rating of 1 (minimal) for each side. This suggested a combined score of greater than 2 would represent the optimum cutoff for distinguishing elderly control subjects from those with dementia. Using this cutoff, MTA was present in 100% AD patients, 87.5% VaD patients, 62% DLB patients, and 4% control subjects as shown in table 3. MTA was significantly more common in AD compared with DLB patients (p < 0.001). Table 4 shows the sensitivity, specificity, and positive and negative predictive values for the absence of MTA to separate DLB from AD and VaD.

There was a trend for patients with NINDS/ADRDA-probable VaD (n = 14) to have lower MTA scores (i.e., less atrophy) than those with possible VaD (n = 10, 3.9 versus 5.1, Mann–Whitney U test, p = 0.059). Patients with possible VaD had MTA scores of a similar magnitude compared with those with AD (5.2 versus 5.1) and significantly more atrophy than those with DLB (5.1 versus 3.8, Mann–Whitney U test, p = 0.017). In contrast, those with probable VaD had significantly less atrophy than AD (3.9 versus 5.2, Mann–Whitney U test, p = 0.037) but not DLB patients (3.9 versus 3.8). Patients with probable and possible VaD had similar MMSE and CAMCOG scores, but possible VaD patients were older (81 versus 74 years; p = 0.009).

MTA scores correlated with age in DLB patients (r = 0.58, p = 0.002) but not other groups. In all dementia patients MTA scores correlated with the CAMCOG memory subscale (combined memory score: r = −0.34, p = 0.003) but not with other subscales or total CAMCOG (r = −0.15, p = 0.184) or MMSE scores.

Comparing the clinical profiles of atrophic and nonatrophic DLB patients, those with atrophy were older at onset (mean age, 75.8 versus 66.7 years; p = 0.001) and had greater impairment of recent memory (p = 0.007). There were no differences with respect to duration of illness, total CAMCOG score, or other subscales, and psychiatric or neurologic presentation including parkinsonism.

Discussion.

This study examined MTA in a large number of patients with DLB in conjunction with normal control subjects and those with AD and VaD. We found that although MTA was significantly more common in all types of dementias compared with control subjects, patients with DLB had significantly less atrophy than those with AD, and there was a trend toward less atrophy in DLB compared with VaD.

The major strengths of this study include the recruitment of a large number of community-based elderly patients with late-life dementia, with comparable age, sex, education, and duration of illness. MR images were acquired using a standardized protocol, and atrophy was assessed blind to the diagnosis by experienced raters using an established scale. Although volumetric measurement of specific medial temporal lobe structures may have provided a more detailed analysis of atrophy, one of the aims of this study was to evaluate the usefulness of estimating MTA in general clinical settings in which volumetric analysis is impractical because of the time required to analyze each scan. Moreover, there is good agreement between volumetric measurements and visual ratings.23 A potential criticism of this study would be the reliance on clinical, rather than pathologic, diagnoses. However, clinical diagnoses were made after detailed assessments and in accordance with standardized criteria. In the case of DLB, we have shown that the clinical criteria identify correctly 90% of patients judged by autopsy.24

In this study DLB was associated with less MTA than AD. This finding was not due to differences in age or dementia severity between groups because, if anything, DLB patients had greater cognitive impairment than other groups. This finding is consistent with a volumetric MRI study by Hashimoto et al.13 and a postmortem study by Lippa et al.25 They found that patients with pathologically “pure” DLB had significantly larger medial temporal lobes than either those with AD or mixed AD/DLB pathology. Other studies26-28 have also found less extensive neuronal loss and AD-type pathology in the medial temporal cortex of patients with DLB compared with those with AD.

When compared with control subjects, all dementia groups including DLB had more extensive MTA. Nonspecific atrophic processes common to all dementias may account for this finding. An alternative explanation could relate to the presence of concurrent AD pathology in DLB and VaD patients with MTA. As reported by Lippa et al.,25 a greatly reduced medial temporal lobe mitigates against a diagnosis of DLB unless there is co-existing AD pathology. Of interest, in our pilot study12 both cases of DLB with temporal atrophy to come to autopsy had evidence of concurrent temporal lobe AD pathology.

It is generally accepted that a majority of patients with DLB have co-existing AD-type pathology, although neurofibrillary tangles are usually absent or scarce and the exact frequency of these changes depends on the pathologic criteria used.1,29 Conversely, approximately one-third of DLB patients can be classified as pathologically “pure,” and in this study a similar proportion of patients had normal MTA scores. The observed correlation between MTA severity and increasing age in DLB patients could also be explained by an increased probability of AD-type pathology with advancing age. Interestingly, the only clinical features that distinguished nonatrophic and atrophic DLB patients were age at onset and memory impairment. Neuropathologic examination is necessary to clarify the relationship, if any, between MTA on MRI in DLB and concurrent AD pathology.

The pattern of MTA in VaD was intermediate between AD and DLB. Laakso et al.14 also found hippocampal atrophy was significantly greater in VaD patients than in control subjects but less than in AD patients. In contrast, O’Brien et al.11 found temporal lobe atrophy was uncommon in VaD patients, although only seven VaD patients were included. A number of factors may account for these discrepancies. The studies used different approaches to assessing atrophy, and the location of vascular pathology may have varied (only one patient in our study had a strategically located infarct that could account directly for the presence of MTA). Perhaps more likely, variations may be due to differences in the selection of patients and the diagnostic criteria used. Interestingly, in this study the level of diagnostic confidence in VaD appeared to be associated with differing levels of MTA, with patients with possible NINDS-AIREN VaD having a similar pattern of MTA to AD. These findings could imply a diagnosis of possible VaD is less effective at excluding patients with concurrent AD pathology, especially in older patients in whom mixed dementias are more likely. The proportion of mixed-dementia patients misclassified as VaD by a diagnosis of possible VaD (NINDS-AIREN) is at least 29%.30 As for DLB, MTA in VaD may prove to be a marker of AD pathology whether this is occurring alone or in combination with other disease processes.

Previous studies5,8,9,15,31 have usually examined differences in MTA between AD and healthy elderly control subjects and, as in this study, have consistently demonstrated differences between AD and normal control subjects. Relatively few studies11,13,32,33 have investigated the usefulness of MRI in distinguishing AD from other types of dementias. The findings of this study indicate that the presence of MTA is a sensitive marker of AD, although the specificity of MTA to exclude other types of dementias is considerably lower. Conversely, the absence of MTA was highly specific for DLB, although again the sensitivity was lower. Overall these results suggest the diagnostic value of estimating MTA in patients with dementia depends on the question the test is aimed to resolve: The presence of MTA is good at confirming the presence of AD (but not specific), and the absence of MTA is suggestive of a diagnosis of DLB (but not sensitive). Of course, in common with other investigations, the value of estimating MTA in routine clinical practice needs to be seen in the context of the overall clinical assessment, as in the case of VaD, in which evidence of infarction is central to ascertaining the diagnosis. Additional studies should determine whether the diagnostic utility of MRI can be improved by specific volumetric analysis or by examining longitudinal changes.

Finally, atrophy of medial temporal lobe structures, notably the hippocampus and amygdala, has been associated consistently with memory impairment in AD.15,34-36 In this study MTA correlated specifically with memory impairment in all dementia patients, suggesting atrophy in this region is central to understanding memory impairment in DLB and VaD, as well as AD. This implies a common pathophysiology of amnesia irrespective of diagnosis.