Abstract
Background:
The language profile of behavioral variant frontotemporal dementia (bvFTD) remains to be fully defined.
Objective:
We aimed to quantify the extent of language deficits in this patient group.
Methods:
We assessed a cohort of patients with bvFTD (n = 24) in relation to patients with semantic variant primary progressive aphasia (svPPA; n = 14), nonfluent variant primary progressive aphasia (nfvPPA; n = 18), and healthy age-matched individuals (n = 24) cross-sectionally and longitudinally using a comprehensive battery of language and general neuropsychological tests. Neuroanatomical associations of language performance were assessed using voxel-based morphometry of patients’ brain magnetic resonance images.
Results:
Relative to healthy controls, and after accounting for nonverbal executive performance, patients with bvFTD showed deficits of noun and verb naming and single word comprehension, diminished spontaneous propositional speech, and deterioration in naming performance over time. Within the bvFTD group, patients with MAPT mutations had more severe impairments of noun naming and single word comprehension than patients with C9orf72 mutations. Overall the bvFTD group had less severe language deficits than patients with PPA, but showed a language profile that was qualitatively similar to svPPA. Neuroanatomical correlates of naming and word comprehension performance in bvFTD were identified predominantly in inferior frontal and antero-inferior temporal cortices within the dominant hemispheric language network.
Conclusions:
bvFTD is associated with a language profile including verbal semantic impairment that warrants further evaluation as a novel biomarker.
Keywords
INTRODUCTION
Among the phenotypes of frontotemporal lobar degeneration, the behavioral variant of frontotemporal dementia (bvFTD) is the most clinically, anatomically, and pathologically diverse [1–5]. While patients typically present with social disintegration and personality change [5, 6], there is substantial phenotypic overlap with other entities, in particular the primary progressive aphasias (PPA), even at relatively early stages [1, 5]. Previous studies including data on language functions in bvFTD [6–22] are summarized in Table 1. Deficits in confrontation naming [7–9, 22], comprehension of single words [7, 18] and sentences [6, 16], and more generalized semantic and language impairment [12, 20] have been described in bvFTD. Regional frontotemporal atrophy in bvFTD often overlaps brain networks canonically concerned with language [2, 3] and available neuroanatomical evidence has implicated distributed frontal, temporal, and parietal circuitry in the genesis of language deficits in this syndrome [6, 19]. Taken together, this evidence suggests that bvFTD may lead to language dysfunction, particularly where there is a requirement for processing verbal associations, searching the verbal lexicon, or planning propositional utterances. However, direct head-to-head comparisons between bvFTD and primary progressive aphasia syndromes with simultaneous, comprehensive language assessment and neuroanatomical correlation have been undertaken only infrequently [8, 9] (see Table 1). Language deficits in bvFTD remain incompletely defined and may go unrecognized.
Here we addressed this issue with a comprehensive assessment of language and general neuropsychological functions in a well-characterized cohort of patients with bvFTD. The language profile of bvFTD was determined after taking into account general (nonverbal) executive performance, an important potentially confounding factor in this group of patients. The bvFTD cohort was compared to patient cohorts with canonical syndromic variants of primary progressive aphasia, assessed both cross-sectionally and longitudinally, in order to examine the relative salience and chart the progression of any language deficits. Neuroanatomical associations of language impairments were assessed using voxel-based morphometry of patients’ brain magnetic resonance images. Building on the cumulative evidence of previous work, we hypothesized that patients with bvFTD have deficits particularly affecting language functions such as word retrieval and propositional speech that are likely to engage executive processes. We further hypothesized that these deficits particularly implicate anterior subregions of the distributed dominant hemisphere language network.
MATERIALS AND METHODS
Participants
Twenty-four patients with behavioral variant frontotemporal dementia (bvFTD), 14 patients with semantic variant primary progressive aphasia (svPPA), and 18 patients with nonfluent variant primary progressive aphasia (nfvPPA) were recruited via a tertiary cognitive disorders clinic as part of a cross-sectional and longitudinal cognitive and neuroimaging study of frontotemporal lobar degeneration. All patients fulfilled current consensus criteria for a probable or definite syndromic diagnosis [1, 23] corroborated by general neuropsychological assessment. Genetic screening of the patient cohort revealed pathogenic mutations in 11 cases (five C9orf72, four bvFTD, one nfvPPA; six MAPT, all bvFTD). Volumetric brain MRI showed compatible profiles of regional brain atrophy in each of the syndromic groups; none of the patients had a significant intercurrent burden of cerebrovascular disease. Twenty-four healthy individuals with no history of neurological or psychiatric illness age-matched to the patient cohort also participated. Demographic, clinical, and background neuropsychological data for all participant groups are summarized in Table 2.
All participants gave informed consent, and ethical approval for the study was granted by the National Hospital for Neurology and Neurosurgery and the University College London Hospital Research Ethics Committees, in line with Declaration of Helsinki guidelines.
Assessment of language
Language tests administered to participants covered seven core domains of language processing: speech input (an adapted version of the Psycholinguistic Assessment of Language Processing in Aphasia subtest 3, PALPA3 minimal pairs discrimination) [24], speech repetition (word and sentence repetition) [25], single word comprehension (word–picture matching using the British Picture Vocabulary Scale [26]; concrete and abstract words from the synonyms comprehension test) [27], sentence comprehension (adapted from the PALPA55 picture-sentence matching task) [24], lexical retrieval (noun naming, Graded Naming Test [28]; a novel verb naming test using pictured actions, further details in Supplementary Materials), reading (Graded Nonword Reading test [29], National Adult Reading Test [30]; Schonell Graded Word Reading Test) [31] and spelling (Graded Difficulty Spelling test) [32] (see Table 3). Further details of the test procedures are provided in Supplementary Material, and the language battery used here has been described previously [33]. In addition, participants’ spontaneous propositional speech was assessed by asking them to describe their last holiday; participants were encouraged to talk for up to three minutes, using prompts if necessary. This procedure was designed to be as open-ended as possible, in order to encompass the anticipated very wide range of fluency and general language competence across participant groups; prompts were intended to ensure that the language sample obtained for each participant was as complete as possible (examples of prompts included, “Where did you go?”, “How long were you there for?”, “How did you get there?”, and “What did you do there?”). Both the participants’ responses and any examiner prompts were recorded for offline analysis.
Longitudinal neuropsychological assessments between one and three years apart were conducted for a subset of the bvFTD (n = 16; mean (standard deviation) test interval 570 (212) days), svPPA (n = 9; 518 (236) days), nfvPPA (n = 10; 409 (110) days), and healthy control (n = 15; 471 (190) days) groups. The mean test interval did not differ significantly between participant groups.
Analysis of behavioral data
Propositional speech recordings were first processed to extract for each participant the total number of words, number of words normalized for number of prompts required from the examiner, and median word frequency (British National Corpus, spoken portion, http://www.natcorp.ox.ac.uk/). The proportions of nouns and verbs produced by each participant were calculated by dividing the number of words in each category by total number of words produced, in order to control for overall utterance length.
All behavioral data were analyzed using Stata® v12. Demographic characteristics and general neuropsychological data were compared between groups using independent samples t-tests for continuous variables and chi square tests for dichotomous variables. Language performance was compared between groups separately for each graded difficulty language test and for propositional speech variables using a linear regression model incorporating test score as the variable of interest with covariates of gender and WASI Matrices score (an index of general nonverbal executive function and surrogate of disease severity). In order to take account of near-ceiling performance by healthy controls, pass/fail variables were compared between groups separately using chi-square tests. Post hoc subgroup analyses were conducted to compare bvFTD patients with genetic mutations with other bvFTD cases using independent samples t-tests for continuous variables and chi square tests for dichotomous variables. Neuropsychological performance profiles were constructed for each patient group by transforming raw group mean scores on each test to a z-score relative to the healthy older control group mean wherefeasible. Longitudinal language data were analyzed using Wilcoxon matched-pairs signed-ranks tests. For all behavioral comparisons, a threshold of p < 0.05 was accepted as the criterion for statistical significance.
Brain image acquisition and analysis
Volumetric brain MR images were acquired for all patients in a 3.0 T Siemen’s Trio MRI scanner using a 32-channel phased array head-coil and a T1-weighted sagittal 3D magnetization rapid gradient echo sequence (TE = 2.9 ms, TR = 900 ms, TI = 2200 ms), with dimensions of 256×256×208, and voxel size of 1.1×1.1×1.1 mm3.
In order to conduct a voxel-based morphometry (VBM) analysis of patients’ neuroanatomical data, brain images were first pre-processed and normalized to NMI space using SPM12 software (http://www.fil.ion.ucl.ac.uk/spm/software/spm12/) and the DARTEL toolbox with default settings for all parameters [34, 35] running under Matlab® R2012a. Images were smoothed using a 6 mm full-width at half-maximum (FWHM) Gaussian kernel. To control for individual differences in head size, total intracranial volume (TIV) was calculated for each participant by summing grey matter, white matter, and cerebrospinal fluid volumes following segmentation. A study-specific template brain image was created by warping all native space whole-brain images to the final DARTEL template and calculating the average of these images.
Voxel intensity (an index of brain volume) was modeled separately in each patient group as a function of language performance for each task on which bvFTD patients showed a deficit in the behavioral analysis. Age, gender, TIV, and WASI Matrices score were incorporated as covariates of no interest in all models. An explicit brain mask was applied, whereby a voxel was included in the analysis if grey matter intensity at that voxel was >0.1 in >70% of the participants [36]. Statistical parametric maps (SPMs) were assessed at two significance criteria: thresholded at p < 0.05 after family-wise error (FWE) correction for multiple voxel-wise comparisons over the whole brain; and thresholded at p < 0.05FWE after small volume correction within subregions of the left hemisphere language network pre-specified in our prior anatomical hypotheses. These anatomical regions comprised inferior frontal gyrus, posterior superior temporal cortex and anterior temporal lobe, derived from the Juelich Histological and Oxford/Harvard defined brain regions in FSL v3.12 [37, 38] and edited in MRICron® to conform to our customized group template image [39].
RESULTS
Behavioral data
Results of group comparisons for background neuropsychological tasks are presented in Table 2 and language tasks in Table 3. Individual raw performance data are presented in Supplementary Fig. 1
Participant groups did not differ significantly in age, handedness, or educational attainment and patient groups did not differ in symptom duration. Males were over-represented in the bvFTD group relative to each of the other groups and gender was accordingly incorporated as a covariate of no interest in analyses of language variables. Patient groups showed widespread general (extra-linguistic) neuropsychological deficits and the svPPA and nfvPPA groups showed, as anticipated, specific syndromic language profiles relative to the healthy control group. The bvFTD group showed significantly worse recognition memory for words and significantly better verbal and nonverbal working memory and arithmetic performance than the nfvPPA group; and significantly worse verbal working memory performance than the svPPA group.
Relative to the healthy control group, the bvFTD group overall showed impairments of noun naming, verb naming, and concrete single word comprehension (concrete synonyms). Patients with bvFTD did not show deficits of abstract single word comprehension, sentence comprehension (whether considered overall or separately by PALPA55 grammatical construction categories), or any other language domains. The bvFTD group performed significantly better than the svPPA group on tests of noun and verb naming, single word comprehension (concrete and abstract synonyms), sentence comprehension and spelling; and significantly better than the nfvPPA group on tests of nonword repetition and reading. Performance profiles across tests (based on transformed z-scores for each of the patient groups relative to the healthy control group) are presented in Fig. 1.
A more detailed analysis of the bvFTD group (summarized in Supplementary Table 1) revealed two subgroups stratified by performance on language tasks: a more severe subgroup of 10 patients performing >2 standard deviations below the healthy control group mean on tests of both noun naming and single word comprehension and a less severe subgroup comprising the remaining 14 bvFTD cases. The more severe subgroup was significantly older, had significantly lower MMSE scores and significantly shorter symptom duration than the less severe subgroup. However, there were no consistent profiles of regional brain atrophy on MRI for either subgroup: both subgroups represented a variety of atrophy profiles and in particular, each subgroup contained only a single patient with focal, asymmetric temporal lobe atrophy (see Supplementary Table 1).
Post hoc analyses of the genetic subgroups within the bvFTD group (summarized in Table 4) revealed that the MAPT mutation subgroup performed significantly worse than the C9orf72 mutation subgroup on noun naming and single word comprehension (British Picture Vocabulary Scale). The C9orf72 mutation subgroup performed significantly worse than the MAPT mutation subgroup on working memory measures (Table 4); no other significant neuropsychological differences between the genetic subgroups were identified.
In the propositional speech analysis, the bvFTD group did not differ from healthy controls in total number or mean frequency of words produced but produced significantly fewer words on average between prompts than the healthy control group. Median word frequency score in the bvFTD group was significantly lower than in both the svPPA and nfvPPA groups. There were no significant differences between groups in the proportions of nouns and verbs produced.
Results of the longitudinal analysis of language data in the participant groups are summarized in Table 5. The bvFTD group showed significant deterioration in noun naming between time-points, while the nfvPPA group showed a significant interval decline in sentence comprehension and the svPPA group showed a significant decline in single word comprehension. The subgroup of bvFTD patients less severely affected at baseline showed a longitudinal decline in naming (Supplementary Table 1), indicating this effect was not restricted to more severely affected patients with bvFTD.
Voxel-based morphometry data
Neuroanatomical associations with language deficits in the bvFTD group (noun naming and concrete single word comprehension) are compared with the svPPA and nfvPPA groups in Table 6; statistical parametric maps of associated regional grey matter atrophy in the bvFTD group are presented in Fig. 2
In the bvFTD group, worse noun naming performance was associated with decreased grey matter volume in right superior parietal cortex and left anterior fusiform gyrus (p < 0.05FWE for multiple voxel-wise comparisons over the whole brain), the cluster extending toward the left temporal pole. Worse noun naming performance in the nfvPPA group was associated with decreased grey matter in left middle temporal gyrus/superior temporal sulcus (p < 0.05FWE within pre-specified anatomical region of interest); no significant associations between noun naming and grey matter volume were identified in the svPPA group. In the bvFTD group, worse single word comprehension performance was associated with decreased grey matter volume in left inferior frontal gyrus and inferior frontal sulcus (p < 0.05FWE within pre-specified anatomical region of interest, see Fig. 2 –though note that SPMs are presented at p < 0.001 uncorrected for display purposes); no significant associations ofsingle word comprehension were identified in the svPPA or nfvPPA groups. No significant grey matter associations of reduced propositional speech output were identified. Reverse contrasts of each language variable did not yield significant grey matter associations in any group.
DISCUSSION
Here we have shown that language deficits accompany bvFTD even after taking general executive performance and disease severity factors into account. These deficits were particularly prominent in the domains of single word comprehension (as indexed using the synonyms test) and lexical retrieval (as indexed using the Graded Naming Test). While naming is a multi-component cognitive process, the naming deficit demonstrated in the bvFTD group here may be at least partly semantically based. The present findings corroborate previous evidence for impairments of naming and verbal semantic functions in bvFTD [7–9, 22] and suggest that a more specific, primary verbal semantic deficit may be a core linguistic feature of the bvFTD syndrome. In line with this, a profile analysis (Fig. 1) revealed a qualitatively similar pattern of deficits in the bvFTD and svPPA groups here. While the patient groups were selected according to current consensus criteria for the respective syndromes, this resonates with clinical experience and previous neuropsychological work suggesting convergence in the behavioral and cognitive profiles of bvFTD and svPPA [33, 41]. It is unlikely the findings are attributable simply to misclassification of svPPA cases since the most lexically impaired patients in the bvFTD cohort considered as a subgroup did not show a profile of regional brain atrophy compatible with svPPA. It is of interest that decline in naming performance over time was a signal of disease evolution in the bvFTD group but not in the PPA groups here. While this apparent discrepancy may be at least partly attributable to the relatively small size of the present PPA groups and floor effects in the svPPA group, our data suggest that naming as a general index of language function may be a candidate biomarker in bvFTD. This is pertinent given the current paucity of biomarkers in bvFTD and the difficulties surrounding measurement of the complex social and emotional behaviors that typically dominate the clinical picture in this syndrome.
Neuroanatomical correlates of naming and word comprehension in bvFTD were identified in a distributed, predominantly left-lateralized and anterior network of cortical regions, including anterior and inferior temporal and inferior frontal cortices. These areas are canonical components of the language network and have been previously implicated in word retrieval and control processes both in the healthy brain and in lesion studies [42, 43]. Similar prefrontal cortical correlates of naming performance have been identified in previous work in bvFTD [9]. While the nondominant parietal correlate of naming performance shown here appears at first more surprising, strength of activation in this region during a semantic decision task has been correlated with off-line naming performance in the healthy brain [44], and its engagement here may reflect cross-modal integrative mechanisms during the picture naming task or semantic task load in these cognitively impaired individuals [42]. It is noteworthy that the bvFTD group also had reduced spontaneous generation of propositional language. While a direct neuroanatomical correlate of propositional speech output was not identified here, this is likely to be grounded in a similar anterior dominant hemispheric network, based on evidence in the healthybrain [45].
Differential involvement of these networks may also account for the stratification of language profiles between the MAPT and C9orf72 genetic mutation subgroups in this study. Though case numbers were not sufficient for direct neuroanatomical correlation, these mutation subgroups have been shown previously to have distinct neuroanatomical profiles, with relatively focal involvement of anterior temporal and inferior frontal cortices in association with MAPT mutations and involvement of a distributed thalamo-cerebello-cortical network in association with C9orf72 mutations [4, 46]. The more severe naming and semantic deficits in the MAPT subgroup compared with the C9orf72 subgroup would be in line with these neuroanatomical signatures and also with previously documented cases of patients with MAPT mutations and a clinical phenotype overlapping bvFTD and svPPA [47].
Taken together, the present findings suggest that involvement of distributed cortical networks mediating verbal semantic processing may underpin the language profile of bvFTD and further suggest that language impairment may be a relevant clinical issue in these patients. Case numbers in this study were relatively small: future work should address language functions prospectively and systematically in larger bvFTD cohorts. Like all work of this kind, the present study was potentially susceptible to patient floor-performance and healthy control ceiling-performance effects associated with conventional neuropsychological tests of language function: this issue will only be fully addressed through development of new graded difficulty tests that can capture the very wide range of performance across target groups in the relevant language domains.
It is worth noting that bvFTD represents a diverse clinicopathological spectrum, and there are inevitably certain limitations in averaging performances at a group level [48]. It will be of particular interest in future work to assess patients with progressive supranuclear palsy and corticobasal degeneration who may exhibit prominent verbal adynamia and speech production deficits [49]; and patients with defined genetic mutations, as such cases potentially illustrate the molecular phenotype of specific brain network disintegration [50].
The potential role of language indices as biomarkers in bvFTD should be further assessed longitudinally over longer periods of follow-up and particularly in genetic mutation carriers at presymptomatic and earliest clinical disease stages. The identification of language deficits in bvFTD complements previous work delineating behavioral features in PPA syndromes [49] and supports the concept of these syndromes as network-based proteinopathies that may transcend conventional syndromic boundaries [5, 50]. Our findings underline the potential for substantial syndromic overlap within the frontotemporal lobar degeneration spectrum, with implications for current diagnostic formulations.
Footnotes
ACKNOWLEDGMENTS
We gratefully thank Dr. Susie Henley for her kind assistance with the neuroimaging analyses. We would also like to thank all patients and healthy volunteers for participating in our research.
The Dementia Research Centre is supported by Alzheimer’s Research UK, the Brain Research Trust and the Wolfson Foundation. This work was funded by the Wellcome Trust, the UK Medical Research Council and the NIHR Queen Square Dementia Biomedical Research Unit. CJDH holds an MRC PhD Studentship. SJC is supported by grants from ESRC/NIHR (ES/K006711/1), EPSRC (EP/M006093/1) and an Alzheimer’s Research UK Senior Research Fellowship. JDR is funded by a National Institute for Health Research Rare Disease Translational Research Collaboration Fellowship. JDW receives salary support from the Wellcome Trust (Wellcome Trust Senior Clinical Fellowship (091673/Z/10/Z).
