Influence of Alzheimer pathology on clinical diagnostic accuracy in dementia with Lewy bodies

Methods.

Results.

Results for the initial analysis of overall clinical diagnostic accuracy for the cohort of 98 autopsy-confirmed DLB subjects are as follows: 44.9% were correctly diagnosed by the clinician with DLB, 46.9% were diagnosed with AD, 6.1% with mixed dementia (AD with additional vascular component), 1% with progressive supranuclear palsy, and 1% as normal.

Results for the subsequent analyses are confined to the 90 autopsy-proven DLB subjects with a clinical diagnosis of either AD or DLB. Demographic data for the low vs high Braak stage groups are summarized in table 1. They did not differ significantly with regard to sex, education, age at onset, or age at death. There were also no significant differences with regard to Mini-Mental State Examination24 (MMSE) score at presentation or interval from last MMSE to death. However, disease duration did differ between the two cohorts (p = 0.04), with a longer mean duration of illness in the high Braak stage group. In addition, mean MMSE score at last examination before death was lower in the high Braak stage subjects (p = 0.03).

Clinical characteristics of the cohorts are presented in table 2. Overall, visual hallucinations—one of the core clinical features of the Consortium criteria1—were present in fewer than half of the patients with DLB (42%). Spontaneous EPS—another core clinical feature—were present in only 55% of the patients with DLB. Only 27% of the autopsy-confirmed patients with DLB demonstrated both visual hallucinations and spontaneous EPS at any point in the course of their illness. Almost a third (30%) of the patients with DLB never had visual hallucinations or spontaneous EPS.

After stratifying the DLB subjects into low and high Braak stage, we found that visual hallucinations were more common in those patients with low (65%) vs high (33%, p = 0.008) Braak stage. The prevalence of spontaneous EPS, on the other hand, was similar in both groups (low Braak 61% vs high Braak 53%, p = 0.51). Only 21% of the patients in the high Braak stage group had both visual hallucinations and EPS. A surprising 35% of the patients in the high Braak stage group did not show EPS or visual hallucinations at any point in the course of their disease (see table 2). Other clinical signs, such as rigidity, parkinsonian tremor, and bradykinesia, did not differ significantly in prevalence between the two Braak stage groups.

To control for the influence of sex, education, age at death, duration, rigidity, parkinsonian tremor, bradykinesia, and depression on the significant difference between the low and high Braak stage groups with regard to visual hallucinations, we entered these variables into a logistic regression analysis. In this analysis, an association between visual hallucinations and subjects in the low Braak stage group remained (OR 8.56, p < 0.01).

Overall, only 49% of the DLB study subjects were correctly diagnosed with DLB by the clinicians (table 3). When the patients with DLB were stratified by Braak stage, clinical diagnostic accuracy was higher in subjects with low (75%) compared to high (39%) Braak stages (p = 0.0039). An examination of clinical diagnostic accuracy by individual Braak stage revealed that the lower the Braak stage, the more likely the subject was to have been diagnosed with DLB; the higher the Braak stage, the more likely the patient was to have been diagnosed with AD (table 4).

Discussion.

We examined the accuracy of the clinical diagnosis of DLB and the hypothesis that superimposed AD neurofibrillary pathology affects the clinical characteristics, and thus the clinical diagnostic accuracy, of DLB. Assessment of clinical diagnostic accuracy in relation to Braak stage has not, to our knowledge, been described previously. We found that the clinical diagnostic accuracy of DLB remains low—less than 50%. In addition, we found that only 27% of our pathologically confirmed DLB cohort developed both visual hallucinations and spontaneous EPS, two of the core clinical features for the diagnosis of DLB. Clinical diagnostic accuracy was strongly influenced by Braak stage: patients with lower Braak stages were significantly more likely to have been diagnosed with DLB by the clinician.

Several studies have found relatively low clinical diagnostic accuracy for DLB5-10⇓⇓⇓⇓⇓ with sensitivity ranging from 34% to 65%.5-8,10⇓⇓⇓⇓⇓ Only one prospective study25 to date has reported high diagnostic accuracy of 83% for DLB. Differences in referral practices and patient populations may have been responsible for this discrepancy.

The prevalence of core clinical features of DLB in our cohort was low: only 42% showed visual hallucinations and only 55% demonstrated spontaneous EPS during the course of their illness. In fact, only 27% of the subjects in our sample demonstrated both visual hallucinations and EPS, two of the three core clinical features of DLB. This low prevalence of core clinical features is supported by a study, in which hallucinations could be identified in only 42% of 56 patients with DLB and concomitant AD pathology.9 It is further supported by a recent study6 of subjects with DLB in which only 50% showed visual hallucinations and only 66% developed EPS. This is in contrast, however, to one study25 in which high prevalence rates (85% visual hallucinations and 87% EPS) for these features were observed.

We did not consider fluctuating cognition—the third core clinical feature of DLB—in our analyses, because this feature has been the most difficult to define and report reliably,8,10,26,27⇓⇓⇓ but this did not preclude the clinician from using fluctuating cognition—or any of the additional supporting features of DLB (e.g., repeated falls, syncope, neuroleptic sensitivity)—in the clinical assessment. However, recognizing that clinicians often cannot assess the presence of fluctuating cognition with any consistency, and that interrater reliability studies6,8⇓ suggest that identifying fluctuations remains problematic, we believed that we were not able to reliably ascertain the prevalence of this feature in our cohort and therefore chose not to include it in our analyses. On the other hand, as assessed by the kappa statistic, hallucinations and EPS appear to be some of the most reliable symptoms in interrater reliability studies in DLB. For example, kappa scores of 0.9 for EPS and 0.87 for visual hallucinations but only 0.57 for fluctuations have been reported6 in subjects with DLB. Because the purpose of the current study was to examine the hypothesis that superimposed AD neurofibrillary pathology affects the clinical characteristics, and thus the clinical diagnostic accuracy, of DLB, we thought it imperative to analyze those clinical characteristics that have been reliably operationalized. Although the Newcastle group25,28⇓ has reported prevalence rates for fluctuating cognition of up to 85%, which may have enhanced their ability to diagnose DLB in life, others6 suggest that less than a third of patients with DLB develop all three core clinical features. In light of the low prevalence of core clinical features in patients with DLB, it is not surprising that clinical diagnostic accuracy for this entity is low. In our study, clinical diagnostic accuracy of DLB improved when core features were present.

Consistent with previous studies,6,28⇓ we found that the subjects with pure DLB showed a higher prevalence of core symptoms than those subjects with admixed AD pathology. In our sample, only 33% of patients with DLB and AD pathology developed visual hallucinations, as opposed to 65% of patients with predominantly DLB pathology. This was in keeping with a report8 that found that 42% of patients with admixed AD pathology, as opposed to 70% of patients with pure DLBD, had hallucinations.

There are several possible explanations for these findings. First, it may be that patients with DLB with high Braak stages, and thus more extensive AD neurofibrillary pathology, show a pattern of clinical and behavioral change so typical of AD that the core clinical features of DLB are essentially masked, despite the added presence of Lewy bodies. Second, it is possible that the extent of Lewy bodies or α-synuclein pathology varies and is less extensive in patients with DLB with high Braak stages. Novel antibodies29 that recognize pathologic but not normal α-synuclein may allow appropriate quantification of the burden of Lewy bodies and neuritic pathology to address this question. Third, it may be that differential cholinergic and other transmitter deficits contribute to the occurrence of EPS and visual hallucinations in these patients. A comparison of levels of midfrontal choline acetyltransferase activity between the two cohorts in our study, however, failed to reveal any differences (p = 0.69, data not presented).

Several potential limitations and issues related to this study should be addressed. First, the subjects who comprised this study were followed prospectively and assessed on an annual basis. However, it was not until 1991 that the diagnosis of DLB existed formally on our DWS, necessitating that the records of a subset of the subjects (approximately one third) be reassessed in a blinded fashion. We do not feel, however, that this altered the outcome of these results in any way, because there were no significant differences with regard to diagnostic accuracy in our prospectively (46%) vs retrospectively (53%) diagnosed patients with DLB.

Second, we recognize that H-E and ubiquitin staining to detect Lewy bodies is less sensitive than α-synuclein staining. Whereas the latter method may have identified more Lewy bodies, it would not have changed any of the classifications in our study, because all subjects already met criteria for DLB,1 and the purpose of this investigation was to assess the influence of concomitant tangles on clinical characteristics and clinical diagnostic accuracy.

Finally, in our study, the duration of illness of the high Braak stage patients was significantly longer than that of the low Braak stage subjects, suggesting several interesting possibilities. It may be that because the subjects in the high Braak stage group live longer, there is a greater opportunity to develop additional pathologic features. Alternatively, it may be that patients in the low Braak stage group demonstrate more hallucinations and spontaneous EPS—symptoms that, in themselves, have been reported to be negative predictors for survival in patients with dementia.30

This study examined the hypothesis that concomitant AD tangle pathology confounds the clinical characteristics—and thus the clinical diagnostic accuracy—of DLB. We found that subjects with DLB with high tangle load in the limbic and neocortical cortex are less likely to express the clinical features of DLB, making the recognition of DLB, and differentiation from AD, more difficult. The current study emphasizes the need for better diagnostic tools for DLB, including options such as neuroimaging or biomarkers, to ensure improved recognition and potential intervention for patients with DLB.