Utility of clinical criteria in differentiating frontotemporal lobar degeneration (FTLD) from AD

Methods.

This study consisted of a retrospective review of the clinic charts of patients in whom a specific diagnosis of AD or FTLD was confirmed neuropathologically (except in one case of familial dementia, in which an affected family member had FTLD confirmed at postmortem examination). Charts were reviewed from three subspecialty AD centers: The University of California at San Francisco, The University of California at Los Angeles, and The University of California at Davis. All patients with a pathologic diagnosis of FTLD at these centers were identified, and a group of patients with a pathologic diagnosis of AD was selected to match the FTLD group based on Mini-Mental State Examination (MMSE) score obtained at the time of the initial assessment.30 Clinical assessments were based on the results of the initial evaluation at the center where the data were collected. At each center, this included a neurologic history and examination and a neuropsychological evaluation in every case in which the patient was cooperative enough for reliable data to be obtained (although the specific assessments in the neuropsychological assessment varied between centers and across time).

Before the chart review, a checklist consisting of demographic, behavioral, and cognitive variables was compiled based on the most recently published diagnostic criteria20 (table 1). Only those items for which sufficient information was available to allow accurate grading were included. For instance, although the diagnostic criteria for semantic dementia include surface dyslexia, semantic paraphasias, and associative agnosia, such information was rarely described as present or absent in our clinical charts. Therefore, these items were excluded from analysis. An unblinded physician rater (K.M.R.) reviewed each chart and documented the presence or absence of the clinical features in each patient. In evaluating the charts for each item on the checklist, the qualitative description provided in the published diagnostic criteria was used as a guide to rating. Any clinical information in the charts that could contribute to this evaluation was considered, including historical information provided by the caregivers, clinical impressions from examining physicians and neuropsycho-logists, and any results from standard behavioral measures such as the Neuropsychiatric Inventory (NPI).13 If there was no information in the chart that gave insight into a particular variable, then it was considered unavailable and treated as missing data. Because of the variability in the neuropsychological tests administered, no raw neuropsychological data were directly analyzed except for the MMSE, which had been administered to every patient. Although the patient groups were, by definition, equated based on overall MMSE score, the analysis of performance on specific measures in the MMSE, chosen a priori based on previous studies, allowed us to assess the hypothesis that the pattern of impairment on this test can be diagnostically useful.21

Because an unblinded rater reviewed data retrospectively, a second physician rater (H.J.R.) independently reviewed a subgroup of 15 charts, and the correlation between the two ratings was analyzed with a κ statistic. Both rating clinicians had at least 1 year of experience assessing dementia at a subspecialty medical center.

In each case, the initial clinical diagnosis and the last clinical impression before death were recorded. In addition, an attempt was made, based on the available clinical data, to identify the specific syndrome (FTD, semantic dementia, or PA) in each patient with FTLD. The latter analysis should be interpreted with caution. In most cases, the clinical evaluations took place before the time that these terms were being used commonly, even in subspecialty dementia settings. In particular, specific data that would help to make a diagnosis of semantic dementia, as discussed above, were not always available.

The cases seen in this study were diagnosed neuropathologically over 15 years. All of the AD cases came from the University of California at Davis and met the Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) neuropathologic criteria.31 For FTLD, cases were considered positive if they had either classical Pick’s disease, dementia lacking distinctive histology, frontal lobe dementia of the non-AD type, or FTLD with motor neuron disease, as described in the Lund–Manchester Criteria.19 No FTLD case met research criteria for AD, although several cases showed scattered plaques.

The data were analyzed as the proportion of patients in each group who were considered to have that feature, and the results were compared across groups using χ² analysis. Odds ratios representing the relative odds of having FTLD vs AD at autopsy were also calculated for each variable by entering that variable into a logistic regression analysis as the sole regressor. Continuous variables were compared across groups using t-tests. In addition, the variables where a significant difference was observed between the two groups by χ² analysis were entered into a stepwise logistic regression analysis to determine their ability to discriminate FTLD patients from those with AD, and to identify the most useful of the variables. For stepwise logistic regression, variables for which greater than 30% of the data were missing were excluded, as were variables in which a large difference existed between groups in terms of the percentage of potential observations that were missing (a difference greater than 20%). Most of the clinical criteria involve observations of abnormal behavioral or language features. Thus, for logistic regression, a feature not commented on was assumed to be not present in that patient. All statistics were performed using SPSS version 10.0.5 (SPSS Inc., Chicago, IL).

Results.

Discussion.

This analysis constitutes the first direct assessment of the utility of the most recently published clinical research criteria in differentiating patients with FTLD from patients with AD using a series of autopsy-verified cases. In this study, the clinical criteria efficiently separated patients with FTLD from those with AD. In general, behavioral features, including disorders of social conduct, alterations in diet, mental rigidity and stereotyped thinking or behavior were present in the two groups in significantly different proportions, whereas features characterizing speech and language, such as anomia, were not. Five clinical features, including the presence of social conduct disorders, hyperorality, and akinesia, as well as absence of amnesia and absence of a perceptual disorder, allowed the identification of 93% of the FTLD and 97% of the AD patients. These findings stand in contrast to a recent evaluation of the ability of the NINCDS-ADRDA criteria for AD to differentiate AD from FTLD. In that analysis, the clinical criteria identified most cases of AD, but incorrectly classified many FTLD cases as AD.28 The current results, combined with the results of this previous study, demonstrate that specific clinical criteria are necessary to differentiate FTLD from AD.

The demographic data available were consistent with previous observations in these groups of patients. Previous studies suggested that FTLD tends to present at a younger age than AD,32,33⇓ and estimates of the prevalence of a family history of FTLD in patients with FTLD have been close to 50%.33,34⇓ The prevalence of a family history of 38% in AD in our group may seem surprising, as estimates of familial AD range around 10%. However, the consensus criteria only require one first-degree relative with a similar disorder to satisfy this requirement. The increasing prevalence of AD with age in the population means that many individuals will have such a relative, even though there may be only a small contribution from inheritance, or none at all. One previous comparison of FTLD and AD found the prevalence of this type of family history to be 30% in AD.34

Early disorders of social conduct and alterations in diet were not detected in this group of patients with AD, but were present in FTLD. This is consistent with a recent study that demonstrated that “stereotypic and altered eating behavior” and “loss of social awareness” reliably differentiated AD from FTLD.26 Prior studies using the NPI to assess behavioral abnormalities also suggested that behavioral features—specifically disinhibition, euphoria, and apathy—were valuable features for differentiating AD from FTLD.15 These studies were limited to an assessment of behavioral features and did not include neuropathologic information. Similarly, neuropsychological studies have suggested that impairment in frontal lobe tests or loss of semantic knowledge, with relative preservation of memory, can be useful in the discrimination FTLD from AD,10 but these studies did not take behavioral features into account. The currently published criteria for the diagnosis of FTLD use both cognitive and behavioral features in their description of the various FTLD syndromes. Furthermore, although the retrospective separation of the patients into groups with semantic dementia, PA, and FTD should be interpreted cautiously, our analysis suggests that behavioral abnormalities are quite frequently seen in semantic dementia, confirming the results of previous studies,23,26,27⇓⇓ and that behavioral abnormalities can also be seen with PA.

Our results suggest that an inclusive approach, taking into account all the variables outlined above, is likely to be the most reliable in differentiating FTLD from AD. A previous study, using brain imaging as the gold standard for diagnosis, also found that a combination of behavioral and cognitive features was most useful in this differentiation. Specifically, hyperorality, stereotyped behavior, progressive reduction of speech, and sparing of spatial orientation and praxis differentiated FTLD from AD in that study.21 Our study confirmed the importance of dietary changes and visuospatial performance but also highlighted the utility of changes in social conduct. In addition, our results emphasize the utility of physical findings, specifically parkinsonian features when they are present, in identifying patients with FTLD.

Lastly, our study reinforced but qualified the utility of memory disturbances in the evaluation of dementia. For this study, severe amnesia was defined as absent if any evidence from the caregivers or the examiners observations or neuropsychological testing suggested sparing of recent episodic memory. Severe amnesia, despite being highly characteristic of patients with AD, was present in many patients with FTLD as well (43%). This latter finding is consistent with recent data suggesting that significant hippocampal and entorhinal cortical volume loss can be detected in the temporal variant of FTLD.35,36⇓ Episodic memory impairment in FTLD may also derive from alterations in attention and working memory. In addition, deficits in verbal processing abilities and word retrieval may contribute both to decreased memory performance in daily life and to the impression on the part of the caregiver (and the patient) that memory is impaired. Our analysis suggests that although it may be possible to differentiate memory impairment related to these factors from true amnesia, caregivers of patients with FTLD are less likely to make this distinction and will thus endorse memory difficulties that, on the surface, seem similar to those seen in patients with AD. In addition, as indicated in the consensus criteria, neuropsychological testing may not always be helpful in excluding severe amnesia. Thus, although the absence of severe amnesia at the initial presentation makes AD less likely, its presence by no means rules out a diagnosis of FTLD.

The chief limitations of this study were that it was retrospective and unblinded, and the fact that, in some cases, the charts offered limited evidence regarding some of the behavioral or psychological features. Most of our data relied on subjective clinical impressions made by clinicians. In addition, specific features that would have been important for to confirm the diagnosis of semantic dementia were rarely mentioned. Similarly, certain features such as emotional blunting and disorders of personal conduct, as well as the degree and type of memory impairment, can be difficult to measure and are dependent on the observations of perceptive caregivers and clinicians. Thus, it is possible that our analysis may have over- or underemphasized the ability of these features to differentiate the two diseases. However, these retrospective data stress the need for a more careful prospective approach to diagnostic evaluation, where clinical features are formally operationalized and rated as objectively as possible and neuropathology is used as the gold standard for diagnosis.