Distinct Antemortem Profiles in Patients With Pathologically Defined Frontotemporal Dementia

Background  Clinical-pathologic studies are crucial to understanding brain-behavior relations and improving diagnostic accuracy in neurodegenerative diseases.

Objective  To establish clinical, neuropsychological, and imaging features of clinically diagnosed patients with frontotemporal dementia (FTD) that help discriminate between pathologically determined tau-positive FTD, tau-negative FTD, and frontal-variant Alzheimer disease.

Design  Retrospective clinical-pathologic survey.

Setting  Academic medical center.

Patients  Sixty-one participants with the clinical diagnosis of a frontotemporal spectrum disorder who underwent a neuropsychological evaluation and had an autopsy-confirmed disease.

Main Outcome Measures  Neuropsychological performance and high-resolution structural magnetic resonance imaging (MRI).

Results  Distinguishing features of patients with tau-positive FTD include visual perceptual-spatial difficulty and an extrapyramidal disorder significantly more often than other patients, significant cortical atrophy in the frontal and parietal regions as evidenced on MRI, and the burden of pathology is greatest in the frontal and parietal regions. Patients with tau-negative FTD are distinguished by their greater difficulties with social, language, and verbally mediated executive functions, significant cortical atrophy in the frontal and temporal regions as evidenced on MRI, and significant frontal and temporal pathology. Patients with Alzheimer disease at autopsy have significantly impaired delayed recall during episodic memory testing; atrophy that involves temporal areas, including the hippocampus, as evidenced on MRI; and widely distributed pathology including the medial temporal structures. A discriminant function analysis grouped patients on the basis of clinical and neuropsychological features with 87.5% accuracy.

Conclusion  Clinical, neuropsychological, and imaging profiles can contribute to accurate antemortem diagnosis in FTD.

Clinical-pathologic studies are crucial to understanding brain-behavior relations and improving diagnostic accuracy in neurodegenerative diseases. Elegant analyses of frontotemporal dementia (FTD) suggest that progressive nonfluent aphasia is frequently associated with Pick disease or Alzheimer disease (AD),¹ although this is not a universal finding.² Other reports³ describe semantic dementia in frontotemporal lobar degeneration with ubiquitin-positive inclusions (FTLD-U). Comparative studies^4,5 allow direct contrasts across pathologically defined conditions that may be difficult to distinguish during life. For example, patients with autopsy-proven FTD present with executive difficulty on measures such as category-naming fluency, whereas patients with AD have episodic memory difficulty.⁶

In this article, we evaluate clinical features of patients with the antemortem diagnosis of an FTD spectrum disorder that have undergone autopsy. We investigate clinical, imaging, and neuropsychological characteristics comparatively, facilitated by a classification scheme for neurodegenerative diseases based on biochemical criteria (Figure 1).^7,8 Patients with FTD are subdivided into those with the accumulation of the microtubule-associated protein tau, such as in Pick disease, corticobasal degeneration (CBD), progressive supranuclear palsy, and argyrophilic grain disease. Other pathologic entities, such as FTLD-U, are said to be tau negative because of the accumulation of tau-negative but ubiquitin- and TAR DNA-binding protein of 43 kDa (TDP-43)–immunoreactive inclusions.⁹ Less common tau-negative conditions associated with the absence of both tau and ubiquitin inclusions are dementia lacking distinctive histopathology and neuronal intermediate filament inclusion disease. Both tau-positive and tau-negative subtypes of FTD differ from AD, in which pathologic tau filaments accumulate as neurofibrillary tangles together with extracellular β-amyloid deposits.¹⁰

An autopsy registry of more than 500 patients with neurodegenerative diseases at the Center for Neurodegenerative Disease Research of the University of Pennsylvania was examined to identify patients with a clinical diagnosis of an FTD spectrum disorder who also had sufficiently detailed clinical and neuropsychological information to investigate clinical-pathologic relations. This search generated a list of 61 patients. Clinic visits occurred from January 1, 1988, through December 31, 2005, and all autopsies were performed at the University of Pennsylvania from 1995 through 2006. One case was seen clinically only once, and the data from this case were included with the initial evaluation. Demographic characteristics are summarized in Table 1.

Each medical record was scored by 2 independent investigators for the chief concern volunteered by the patient and the accompanying caregiver(s), the presence of symptoms elicited from the patient and caregiver(s) by the physician, and evidence of signs during a neurologic examination conducted by the attending neurologist (A.C., C.M.C., H.B.C., M.G., or H.I.H.). If a symptom or sign was not mentioned in the case record, it was scored as absent. Rare discrepancies between reviewers were resolved through discussion. In this study, we focused on the presence of clinical features not quantified by neuropsychological testing, such as the presence of a social disorder and the presence of a movement disorder.

All patients were assessed with the following neuropsychological tests: Wechsler Adult Intelligence Scale–Revised Digit Span subtest,¹¹ forward and backward span assessed working memory; letter fluency,¹² naming words beginning with a specified letter (FAS) in 60 seconds; Trail Making Test, Part B,¹³ drawing a line alternating between randomly arrayed numbers and letters (ie, 1-A-2-B, etc) for up to 300 seconds; Stroop Color-Word Interference Test,¹⁴ naming the font color of color names printed in a discordant color; animal fluency,¹⁵ naming as many animals as possible in 60 seconds; Boston Naming Test,¹⁶ naming 15 line drawings of objects; semantic category membership task,^17,18 judging the semantic category membership of familiar pictures or words; geometric figure copy,¹⁵ copying 4 geometric designs varying in perceptual-spatial complexity; and Verbal Serial List Learning Test,¹⁹ verbal memory for a 10-word list administered for 3 trials, followed by a recall test and then a delayed recognition test on which the 10 original words were intermixed with 10 novel words.

A subgroup of 12 patients (tau positive FTD: n = 5; tau negative FTD: n = 4; AD: n = 3) underwent imaging with high-resolution structural magnetic resonance imaging (MRI), and the results were contrasted with those of 12 age-matched controls. High-resolution T1-weighted 3-dimensional spoiled gradient echo images were acquired with a repetition time of 35 milliseconds, an echo time of 6 milliseconds, section thickness of 1.3 mm, flip angle of 30°, matrix of 128 × 256, and in-plane resolution of 0.9 × 0.9 mm by a 1.5-T MRI scanner (Horizon Echospeed; GE, Milwaukee, Wisconsin). Brain volumes were normalized to our local template of patients with early-onset dementia and age-matched controls (n = 25). A novel symmetric normalization algorithm derived the optimal average shape and appearance template directly from the data set²⁰ along with transformations from the template to each image. Each symmetric normalization transformation thus is composed of an affine and diffeomorphic component that is optimized in a 4-level multiresolution pyramid. The jacobians of the diffeomorphic maps generated by symmetric normalization are combined with a probabilistic segmentation method (Oxford Centre for Functional Magnetic Resonance Imaging of the Brain's Automated Segmentation Tool ([FAST]),²¹ which labels the brain volumes as gray matter, white matter, cerebrospinal fluid, and other with inhomogeneity correction to generate normalized, spatially varying estimates of gray matter volume for each individual. These gray matter volume images are used to perform statistical tests of atrophy and to correlate cortical volume with cognitive decline. The analysis threshold includes all voxels with gray matter. Implicit masking is used to ignore zeros, and the global calculation is omitted. Using SPM2,²² the gray matter volume images are smoothed with a 2-mm full-width at half-maximum gaussian filter to minimize individual gyral variations. The statistical threshold for atrophy images was set at P < .01 after correction for multiple comparisons with false discovery rate for both voxel-level and cluster-level analyses, and we accepted only clusters composed of 50 or more adjacent voxels. Regression analyses related cortical volume to figure copy, animal fluency, letter fluency, confrontation naming, and memory recall. The statistical threshold was set at P < .001 for voxel-level and cluster-level analyses, and we accepted only clusters that exceeded 100 adjacent voxels.

A detailed description of the pathology protocol is provided elsewhere.^7,9 Briefly, tissue obtained at autopsy was fixed in neutral-buffered formalin and 70% ethanol in 150-mmol/L sodium chloride (pH, 7.4), paraffin embedded, and cut into 6- to 10-μm-thick sections. Sections were stained with hematoxylin-eosin and thioflavine S. Immunohistochemical analysis was performed on sections of the neocortex (anterior cingulate gyrus, midfrontal gyrus, superior temporal gyrus, and angular gyrus), hippocampus, putamen, globus pallidus, cerebellum, and midbrain, including the substantia nigra pars compacta. Immunohistochemical analysis was performed with antibodies to tau, α-synuclein, ubiquitin, TDP-43, β-amyloid, and other proteins required for the diagnostic workup. The neuropathologic diagnosis of FTD, summarized in Table 1, used criteria that emphasized tau-positive and tau-negative but TDP-43– and ubiquitin-immunoreactive inclusions in gray matter, white matter, and subcortical neurons and glia.^7,9,23 Patients with frontal-variant AD (fv-AD)²⁴ were diagnosed on the basis of histopathologic and immunohistochemical analysis performed with standard and previously published protocols that used antibodies that detect phosphorylated tau (PHF1,²⁵ generously provided by Peter Davies, MD) and β-amyloid (4G8; Senetek, Maryland Heights, Missouri). Nerve cell loss, spongiosis, gliosis tau, amyloid plaque, and ubiquitin disease were assessed semiquantitatively in the cortical gray matter, white matter, and subcortical regions by experienced pathologists (M.S.F. and J.Q.T.) on a 0- to 3-point scale as absent (0), mild (1), moderate (2), or severe (3). Several cases were included in previous reports.^7,9,26,27

Clinical measures are reported as the percentage of patients impaired, and each neuropsychological measure is reported as a mean ± SD score. Statistical comparisons of neuropsychological measures were performed with parametric techniques, such as analysis of variance (ANOVA), follow-up Tukey tests, and t tests, or with nonparametric techniques, such as the Friedman ANOVA by ranks, Mann-Whitney test (clinical assessment), and Wilcoxon signed rank test (pathologic grading). All analyses were performed with SPSS statistical software, version 14.0 (SPSS Inc, Chicago, Illinois).

Table 2 indicates that more patients with tau-negative FTD had alterations in behavior and social comportment than those with tau-positive FTD (z = 2.23; P = .03) or fv-AD (z = 2.82; P < .001). This problem was evident at the initial visit more often in patients with tau-negative (z = 3.24; P < .001) and tau-positive FTD (z = 1.92; P = .05) compared with those with fv-AD and at the last examination in patients with tau-negative FTD compared with those with tau-positive FTD (z = 2.00; P = .04) and fv-AD (z = 2.28; P = .02). Patients with tau-positive FTD had an extrapyramidal disorder more often than those with tau-negative FTD (z = 2.19; P = .03).

The tau-positive group obtained a lower score than the tau-negative (P < .001) and fv-AD (P < .001) groups for figure copy (F_2,41 = 8.43; P < .001). A significant ANOVA for letter fluency performance (F_2,16 = 14.00; P < .001) was due to fewer responses in patients with tau-negative FTD than those with tau-positive FTD (P < .001) or fv-AD (P < .001). A difference for confrontation naming (F_2,39 = 2.39; P = .05) was due to lower performance by patients with tau-negative FTD compared with tau-positive FTD (P = .04). An ANOVA for animal fluency (F_2,39 = 2.65; P = .08) was borderline statistically significant. There were trends for the tau-positive group to produce lower scores on animal fluency than patients with fv-AD (P = .08). On the delayed recall condition of the verbal memory test (F_2,42 = 3.54; P = .04), patients with fv-AD recalled fewer words than those with tau-negative FTD (P = .03).

Table 2 also provides the results of a discriminant function analysis. We identified all patients who performed all significant measures: output on the letter fluency test, performance on the figure copy test, performance on the delayed recall memory test, and alterations in social comportment. These measures correctly classify 87.5% of patients into their respective disease groups (Wilkes λ = 0.64; χ²₈ = 21.20; P < .001).

Figure 2 illustrates the distribution of significant cortical atrophy in tau-positive, tau-negative, and fv-AD patient groups. The anatomical location of peak voxels in the atrophic clusters is summarized in Table 3. Patients with tau-positive FTD showed significant atrophy in bilateral frontal and parietal regions that was more prominent in the right hemisphere. Those with tau-negative FTD demonstrated significant atrophy in bilateral frontal and temporal distributions. Patients with fv-AD showed significant atrophy in bilateral temporal and parietal regions, including the hippocampus, with some extension into the frontal cortex.

Figure 2 and Table 3 also provide correlations between neuropsychological performance and cortical volume. Performance on figure copy, the measure that was difficult for patients with tau-positive FTD, correlated with cortical volume in the parietal regions. Performance on letter fluency and animal fluency, measures that proved difficult for patients with tau-negative FTD, correlated with cortical volume in the lateral frontal and ventral temporal regions. Naming correlated with bilateral temporal cortex volume. Performance on memory recall, the measure that was difficult for patients with fv-AD, correlated with lateral and medial temporal volume, including the hippocampus.

The anatomical distribution of pathology in the groups of patients is summarized in Table 4. The tau-positive group had abundant tau inclusions but few or no amyloid plaques or other disease-specific lesions, particularly in the frontal, parietal, lateral temporal, hippocampal, and basal ganglia regions. For the tau-negative group, greater amounts of ubiquitin compared with tau and amyloid were noted in the frontal, parietal, temporal, and hippocampal regions. The disease protein TDP-43 colocalizes with ubiquitin in these regions in patients with FTLD-U. In the fv-AD group, all 3 types of pathology were equally severe in all areas of the brain that were sampled.

A priori predictions that associate poor neuropsychological performance with a particular pathology in specific anatomical regions were evaluated with t tests. On the basis of the neuropsychological and imaging studies described herein, we examined whether poor performance on the figure copy test was associated with tau pathology in the frontal and parietal regions in the tau-positive group, whether reduced letter fluency output was associated with TDP-43 and ubiquitin inclusions in the frontal and temporal regions in the tau-negative group, and whether reduced memory recall was associated with amyloid plaques that involved the entorhinal and temporal cortices in the fv-AD group. An anatomical variable was constructed from the 6 pathologic anatomy regions of interest in Table 4. Participants were assigned to a low pathology group if their pathology rating was 0 or 1 or to a high pathology group if the rating were 2 or 3 for each type of pathology (ie, tau, ubiquitin, or amyloid).

For the tau-positive group, lower scores on the figure copy test were associated with greater tau pathology in the frontal (P < .001) and parietal (P < .001) cortices. Among the patients with fv-AD, reduced memory recall was associated with greater amyloid burden in the temporal (P = .02) and entorhinal cortices (P = .04). However, reduced letter fluency did not correlate statistically with frontal and temporal TDP-43 and ubiquitin lesions in the tau-negative group.

Our observations show distinguishing patterns of clinical, neuropsychological, and imaging impairments in histopathologically defined subgroups of patients with the clinical diagnosis of FTD. Distinguishing features of tau-positive FTD include visuospatial and extrapyramidal motor difficulties. Tau-negative FTD is distinguished by social limitations and language and executive difficulty. Patients with amyloid pathology characteristic of fv-AD are distinguished by episodic memory difficulty. These impairments correlate with specific profiles of cortical atrophy on MRI and appear to forecast a particular anatomical distribution for each type of pathology. A pattern of antemortem clinical deficit thus may be statistically associated with a specific histopathologic diagnosis at autopsy.

Patients with tau-positive FTD are significantly more impaired than those with tau-negative FTD or fv-AD in their visual perceptual–spatial performance. Visual perceptual–spatial difficulty has been described in earlier studies of patients with a pathologically confirmed tau-positive disorder.^7,27,29 Patients with tau-positive FTD have significant antemortem cortical atrophy in the frontal and parietal regions that are known to contribute to performing constructional tasks that involve visual perceptual–spatial material,^30-32 and performance on this task is correlated with parietal atrophy on imaging studies. These characteristics may be in part due to the large number of patients with CBD in the tau-positive group.²⁷ The relatively small number of patients in the current study limits statistical power and therefore cannot reveal the true extent and distribution of frontal-parietal atrophy associated with the visual constructional deficit on the MRI studies of patients with tau-positive FTD. The visual perceptual–spatial deficits in this group nevertheless are associated with denser tau-positive pathology in the frontal and parietal cortex.

Two-thirds of the patients in the tau-positive group also had a movement disorder, even though these patients did not necessarily present with a motor complaint.^27,33 These patients have the clinical diagnosis of CBD and progressive supranuclear palsy, accompanied by significant tau pathology in the basal ganglia. Other clinical-pathologic series have also described a motor disorder in their tau-positive cohort.^1,5,34

Additional clinical difficulties may be present in patients with tau-positive FTD, although these do not discriminate these patients from other patient groups.⁷ Patients with tau-positive FTD have memory complaints, but their memory deficit tends to be relatively modest. Previous work has demonstrated medial temporal disease in patients with tauopathy,³⁵ and some patients have such prominent memory difficulty that they are misdiagnosed clinically as having AD.³⁶ Patients with tau-positive FTD may also have language impairments related to frontal and temporal disease. The restricted nature of the language examination limited the ability to identify qualitatively distinct impairments such as impaired grammatical comprehension that may distinguish patients with tau-positive FTD from other autopsy-defined groups.^1,27,37

Patients with tau-negative FTD are more impaired than other patient groups on language-mediated tasks,³⁸ including confrontation naming and category-naming fluency tasks. The latter task additionally involves a prominent executive component. Those with tau-negative FTD also have significant social difficulties. However, patients with tau-negative FTD do not have prominent motor or visual perceptual–spatial deficits.^7,38 These clinical features may be in part due to the large number of patients with FTLD-U in the tau-negative series. TPD-43 and ubiquitin pathology is a marker of the tau-negative variants of FTD, and most of these patients in this series had FTLD-U, since dementia lacking distinctive histopathology and neuronal intermediate filament inclusion disease is far less common.^7,9,39 TDP-43 and ubiquitin pathology is densest in the frontal and temporal neocortex, and tasks that involve language and executive functioning depend on these brain regions.^40-43 Patients with tau-negative FTD have frontal and temporal MRI atrophy, and in this study their performance on category-naming fluency tasks correlated with frontal and temporal atrophy. Language, executive, and social deficits are consistent with the frontal and temporal disease evident in these patients.^44,45 Antemortem identification of patients with tau-negative FTD is more important than ever, in view of the recent discovery of TDP-43 as the disease protein in FTLD-U, which now can serve as a target for drug discovery.⁹ Despite these associations, we may not have observed a significant correlation between clinical and neuropathologic disease in patients with tau-negative FTD in part because we averaged over several TDP-43 subtypes.^9,46 TDP-43 associated with neurites is related to language difficulty and prominent temporal lobe disease, whereas TDP-43 with neuronal intranuclear inclusions is more strongly associated with limited executive functioning and prominent frontal lobe disease.⁴⁷

It is important to distinguish patients with tau-positive disease or tau-negative disease from patients with fv-AD because 15% to 30% of patients with the clinical diagnosis of FTD have AD at autopsy.^{1,3-5,7,48,49} As a group, patients with fv-AD have relatively modest memory complaints.^7,24 Patients with AD pathology in the present series had significant memory difficulty, but this is evident only for delayed recall. Their recall impairment is less than patients with a clinical diagnosis of AD,¹⁹ and patients with fv-AD do not differ from those with tau-positive and tau-negative FTD in learning and recognition performance. Quantitative memory assessment thus is important for identifying fv-AD. An additional finding that can help differentiate patients with fv-AD clinically is their performance on quantitative language, visual, and executive measures, in which those with AD are less impaired than patients with tau-positive or tau-negative FTD.^50,51 Antemortem MRI suggests lateral and medial temporal atrophy and memory recall correlated with medial temporal pathology in fv-AD. A previous study⁶ of patients with pathologically defined disease dissociated episodic memory deficits in AD from executive difficulty in an undifferentiated group of patients with FTD. We found multiple double dissociations differentiating patients with fv-AD from those with tau-positive or tau-negative disease, but these findings depend on a quantitative neuropsychological evaluation.

A discriminant function analysis successfully used 4 clinical features, visual perceptual–spatial functioning, category-naming fluency, social complaints, and memory recall, to distinguish among the tau-positive, tau-negative, and fv-AD groups. One reason why the present study can distinguish among the groups, unlike other work,⁴ may be the use of quantitative cognitive and imaging assessments. Antemortem diagnostic accuracy of individuals with a neurodegenerative disease is crucial to the goal of effective treatment for these patients.

Several caveats should be kept in mind. Since we sought to distinguish among the 3 pathologically defined groups, the size of each cohort was relatively small. We attempted to maximize group size by ascertaining a smaller number of measures in a larger number of patients, but this limited the scope of the neuropsychological battery. Likewise, we urge caution in interpreting the imaging data because of the small number of available images in patients with autopsy-confirmed patients. With these limitations, we conclude that neuropsychological and imaging measures can contribute to identifying the specific histopathologic disease in patients with clinically diagnosed FTD.

Correspondence: Murray Grossman, MD, Department of Neurology, 2 Gibson, University of Pennsylvania School of Medicine, 3400 Spruce St, Philadelphia, PA 19104-4283 (mgrossma@mail.med.upenn.edu).

Accepted for Publication: March 10, 2007.

Author Contributions: Dr Grossman had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Grossman, Libon, Farmer, Lee, and Trojanowski. Acquisition of data: Grossman, Massimo, Wood, Moore, Farmer, Clark, Coslett, and Hurtig. Analysis and interpretation of data: Grossman, Libon, Forman, Massimo, Anderson, Chatterjee, Lee, and Trojanowski. Drafting of the manuscript: Grossman, Libon, Anderson, and Farmer. Critical revision of the manuscript for important intellectual content: Grossman, Libon, Forman, Massimo, Wood, Moore, Chatterjee, Clark, Coslett, Hurtig, Lee, and Trojanowski. Statistical analysis: Grossman and Libon. Obtained funding: Grossman, Lee, and Trojanowski. Administrative, technical, and material support: Grossman, Forman, Massimo, Wood, Moore, Farmer, Chatterjee, Lee, and Trojanowski. Study supervision: Grossman, Massimo, Clark, and Coslett.

Financial Disclosure: None reported.

Funding/Support: This work was supported in part by grants AG17586, AG15116, NS44266, AG09215, AG10124, AG19724, and AG23501 from the National Institutes of Health and by the Dana Foundation.