Association of Regional Cortical Network Atrophy With Progression to Dementia in Patients With Primary Progressive Aphasia

Abstract

Background and Objectives

Patients with primary progressive aphasia (PPA) have gradually progressive language deficits during the initial phase of the illness. As the underlying neurodegenerative disease progresses, patients with PPA start losing independent functioning due to the development of nonlanguage cognitive or behavioral symptoms. The timeline of this progression from the mild cognitive impairment stage to the dementia stage of PPA is variable across patients. In this study, in a sample of patients with PPA, we measured the magnitude of cortical atrophy within functional networks believed to subserve diverse cognitive and affective functions. The objective of the study was to evaluate the utility of this measure as a predictor of time to subsequent progression to dementia in PPA.

Methods

Patients with PPA with largely independent daily function were recruited through the Massachusetts General Hospital Frontotemporal Disorders Unit. All patients underwent an MRI scan at baseline. Cortical atrophy was then estimated relative to a group of amyloid-negative cognitively normal control participants. For each patient, we measured the time between the baseline visit and the subsequent visit at which dementia progression was documented or last observation. Simple and multivariable Cox regression models were used to examine the relationship between cortical atrophy and the likelihood of progression to dementia.

Results

Forty-nine patients with PPA (mean age = 66.39 ± 8.36 years, 59.2% females) and 25 controls (mean age = 67.43 ± 4.84 years, 48% females) were included in the data analysis. Greater baseline atrophy in not only the left language network (hazard ratio = 1.47, 95% CI = 1.17–1.84) but also in the frontoparietal control (1.75, 1.25–2.44), salience (1.63, 1.25–2.13), default mode (1.55, 1.19–2.01), and ventral frontotemporal (1.41, 1.16–1.71) networks was associated with a higher risk of progression to dementia. A multivariable model identified contributions of the left frontoparietal control (1.94, 1.09–3.48) and ventral frontotemporal (1.61, 1.09–2.39) networks in predicting dementia progression, with no additional variance explained by the language network (0.75, 0.43–1.31).

Discussion

These results suggest that baseline atrophy in cortical networks subserving nonlanguage cognitive and affective functions is an important predictor of progression to dementia in PPA. This measure should be included in precision medicine models of prognosis in PPA.

Primary progressive aphasia (PPA) is a clinical syndrome characterized by the progressive breakdown of language abilities.¹ Previous research has identified 3 major subtypes of PPA, each of which is characterized by a unique profile of language impairment and corresponding anatomical distribution of cortical atrophy.^2,3 Early in the course of their illness, patients with PPA typically function well in their daily lives, often making modifications to compensate for the language problem but generally performing most high-level instrumental activities of daily living relatively well.⁴ As such, during the initial phase of the illness, patients with PPA have language impairment with relatively preserved independent functioning consistent with the cognitive functional status of mild cognitive impairment (MCI).^5,6 As the disease progresses, worsening cognitive and/or behavioral symptoms emerge in executive function, memory, socioemotional processes, and motivation that lead to a loss of independent functioning as the patient's cognitive functional status progresses from the MCI stage to the mild dementia stage and beyond.^7-10 The timeline of this progression from MCI to dementia in PPA is variable across patients,^11,12 and the field currently lacks methods for prognostication. There is a critical need to develop such methods because an estimate of the amount of time during which a patient with PPA may be able to function independently would be invaluable for developing a personalized care plan, including treatment options, major life decisions, safety assessments, and support planning.

Prior work on prognostication in PPA showed that superior baseline cognitive performance (e.g., as measured by Mini-Mental State Examination [MMSE]) is associated with greater functional autonomy later in the illness,¹³ although it is likely that the MMSE is not an ideal instrument for prognostication in PPA because it was not designed for this purpose. It is also unclear how this finding relates to the time of progression to dementia. In general, it is challenging to use standard performance-based neuropsychological assessment tools in patients with PPA because of the effects of language deficits on other cognitive domains of performance, which are often assessed verbally.¹⁴

An alternative approach to predicting clinical trajectories of PPA may be possible to achieve with neuroimaging biomarkers. The development of neuroimaging measures to predict time to dementia in individuals at the MCI stage of PPA can help with working around the challenges of cognitive assessment in this population. We previously reported the utility of MRI-derived cortical thickness in the “AD signature” regions¹⁵ as quantitative neuroanatomical prognostic biomarkers for loss of independent functioning in the progression from MCI to dementia due to AD.¹⁶ The development of similar prognostication tools for the population with PPA would be valuable in clinical and research applications (e.g., treatment recommendations, financial planning, family education and counseling, and clinical trial design). In PPA, greater effect size estimates for percentage change over 12 months were observed with cortical thickness than with neuropsychological measures of language function.¹² This finding supports the role of structural MRI measures as a robust and sensitive means to assess decline in PPA.^17-19 We believe this may make structural MRI measures useful as a prognostic biomarker.

Independence in advanced and instrumental activities of daily living requires a range of cognitive and affective functions including executive functions, language, memory, socioaffective processes, and motivation.^20,21 These functions depend on the integrity of the brain networks that subserve them, including the frontoparietal,^22,23 dorsal attention,²³ salience,²⁴ language,^25,26 default mode,²⁷ and ventral frontotemporal (also called a semantic appraisal²⁸) networks. Neurodegenerative change in the language network is a core feature of PPA,^2,3,11,19 leading to aphasia and the disruption of daily activities dependent on language. Despite the important roles of other brain networks in cognitive and affective functions supporting functional independence, and some evidence that executive functions, memory, and socioaffective functions become impaired as PPA progresses, no published studies to date have examined the potential role of the integrity of these networks in predicting clinical trajectories of PPA.

The main goal of this study was to examine the role of these large-scale brain networks subserving various cognitive and affective functions in predicting progression to dementia from the MCI stage of PPA. We hypothesized that, in patients with PPA at the MCI stage, relatively greater atrophy in the language network in the dominant hemisphere would predict progression to the dementia stage (Hypothesis 1). In addition, we hypothesized that relatively greater atrophy in bilateral frontoparietal, dorsal attention, default mode, salience, and ventral frontotemporal networks would also predict progression to dementia (Hypothesis 2). Our second goal was to examine the independent or synergistic contributions of atrophy in the language network and other networks. Greater atrophy outside the language network may lead to the earlier emergence of impairment in nonlanguage cognitive-behavioral domains, leading to a loss of independent functioning and eventual dementia. This is based on evidence demonstrating that impairments in executive functions and episodic memory were both the strongest predictors of the loss of functional independence across a symptomatic spectrum of Alzheimer disease.²⁹ Therefore, we hypothesized that in patients with PPA, relatively greater atrophy in bilateral cognitive-affective networks (i.e., frontoparietal, dorsal attention, default mode, salience, and ventral frontotemporal networks) would predict progression to dementia more so than atrophy in the language network (Hypothesis 3).

Methods

Participants

Individuals diagnosed with PPA were included in this study, all of whom were recruited through the Massachusetts General Hospital (MGH) Frontotemporal Disorders Unit PPA program from 2007 to 2019. All patients received a standard clinical evaluation comprising a structured history obtained from both patient and informant, comprehensive medical, neurologic, and psychiatric history and examinations, neuropsychological and speech-language assessments, and a clinical brain MRI scan that was visually inspected for (1) regional atrophy consistent with or not consistent with a given syndromic diagnosis and (2) other focal brain lesions or evidence of cerebrovascular disease. Clinical formulation was performed through consensus conference by our multidisciplinary team of neurologists, neuropsychologists, and speech and language pathologists, with each patient classified based on all available clinical information as experiencing MCI or dementia (cognitive functional status), cognitive-behavioral syndrome, and likely etiologic diagnosis.³⁰ All patients included in this study met the diagnostic criteria for PPA, which include the exclusionary feature of the absence of early prominent behavioral impairment; these patients therefore did not exhibit behavioral impairment at baseline. All but 1 was able to be subclassified into one of the major subtypes with a clinical imaging–supported atrophy pattern: the nonfluent/agrammatic variant, semantic variant, or logopenic variant of PPA.² Furthermore, all patients also met the following criteria: (1) baseline cognitive functional status of MCI; (2) baseline research MRI data available and at least approximately 6 months of clinical follow-up; (3) no focal brain lesions or significant cerebrovascular disease (e.g., previous strokes, cerebral hemorrhages, and meningiomas); (4) no major psychiatric illness not adequately treated; and (5) a native speaker of English. For each patient, we measured the time between the first visit and the follow-up visit at which progression to dementia was documented or last observation if the patient remained at the MCI stage of PPA. Dementia was defined as a cognitive functional status in which multidomain cognitive and/or behavioral symptoms interfere with the patient's ability to function independently at work or at usual daily activities.³¹ For a subset of our patients (n = 25/50), amyloid β (Aβ) positivity was determined by visual read according to previously published procedures³² and a summary distribution volume ratio (DVR) of frontal, lateral, and retrosplenial (FLR) regions greater than 1.2 based on ¹¹C-Pittsburgh compound B (¹¹C-PiB) PET³³ or CSF levels of Aβ1-42.

We additionally included a group of cognitively normal individuals as control participants, also recruited at MGH. These control participants underwent a neurologic and cognitive assessment to confirm the absence of a medical history of neurologic or psychiatric conditions, structured interview of the participant and an informant by a neurologist, neurologic examination, and neuropsychological test battery (Uniform Data Set 3.0),³⁴ and were determined to be clinically normal with the clinical dementia rating scale (0). All controls had normal brain structure based on MRI and low cerebral amyloid based on quantitative analysis of ¹¹C-PiB PET data (FLR DVR <1.2). This control sample was used as a reference for quantifying the magnitude of atrophy in our patients with PPA.

Neuroimaging Data Acquisition and Preprocessing

MRI data were collected from all participants on a Siemens 3-Tesla MAGNETOM Tim Trio scanner using a 12-channel phased-array head coil. Structural MRI data were acquired using a T1-weighted Magnetization-Prepared Rapid Gradient-Echo (MPRAGE) sequence with the following parameters: repetition time = 2,530 ms, echo time = 3.48 ms, flip angle = 7°, number of interleaved sagittal slices = 176, field of view = 256 mm, voxel size = 1 mm isotropic. Each participant's MPRAGE data underwent intensity normalization, skull stripping, and an automated segmentation of cerebral white matter to locate the gray-white boundary through FreeSurfer v6.0, which is documented and freely available for download online.³⁵ Defects in the surface topology were corrected,³⁶ and the gray/white boundary was deformed outward using an algorithm designed to obtain an explicit representation of the pial surface. All cortical surface derivatives were visually inspected for technical accuracy and were manually edited when necessary. Cortical thickness was calculated as the closest distance from the gray/white boundary to the gray/CSF boundary at each vertex on the tessellated surface.

We defined our large-scale functional cortical networks of interest using an established parcellation of the cerebral cortex derived from a sample of 1,000 healthy young adults.³⁷ Specifically, based on the 17-network solution of this cortical parcellation, we defined the following networks bilaterally: language (original network IDs: 14, 17), default mode (15, 16), frontoparietal control (11, 12, 13), dorsal attention (5, 6), salience (7, 8), ventral frontotemporal (9, 10), and visual association (1) (Figure 1). It should be noted, however, that this is not the only approach to defining these networks and that they can be subdivided or, in some cases, combined depending on the research question of interest. These network masks, originally defined in a template surface space (fsaverage), were then registered to each participant's native surface space. The mean cortical thickness within each network mask was then calculated for every participant by averaging thickness estimates at all vertices falling within its boundaries. In addition, we also registered all participants' cortical thickness data to fsaverage space and smoothed them geodesically with full-width-half-maximum of 10 mm for a vertex-wise analysis (see further).

Figure 1. Functional Cortical Networks of Interest.

These networks were defined using a parcellation of the cerebral cortex originally developed by Yeo et al.³⁶ In the original nomenclature, the ventral frontotemporal network shown here was defined as a “limbic” network. In this work, we used the former to refer to this network, given that limbic tissue is part of other networks (e.g., default mode, salience) as well.⁴⁵ Network parcellation masks are displayed on a semi-inflated surface template (fsaverage).

To quantify the magnitude of cortical atrophy within each functional cortical network in each patient with PPA, we calculated network-wise W scores. W scores are analogous to Z scores adjusted for specific covariates of no interest, which in this study were age and sex.^38,39 Separately for each network in each hemisphere, we performed a multiple regression analysis using cortical thickness data obtained from a group of Aβ- controls, which resulted in β coefficient values for age and sex and individual values of residuals. Using these parameters, we then computed W scores for each network, each hemisphere, and each patient with the following formula:

where T_ijk = the observed mean cortical thickness of network i, hemisphere j, and patient k, , the predicted mean cortical thickness of network i, hemisphere j, and patient k based on age and sex of this patient and β coefficients obtained from Aβ-controls, and SD_ij = the SD of the individual residuals obtained from Aβ- controls for network i in hemisphere j. For our vertex-wise analysis, we computed a W score at each vertex point following the same definition, resulting in a surface W score map per hemisphere for each patient with PPA. Because W scores in this study were calculated using cortical thickness, more negative values indicate greater cortical atrophy relative to what would be expected solely based on age and sex.

Statistical Analysis

To test our first and second hypotheses regarding the relationship between network atrophy and the likelihood of progression from MCI to dementia, we constructed a simple Cox regression model separately for each network in each hemisphere. Predictor variables were network-wise mean W scores, whereas the outcome variable was the time (months) between the baseline visit and progression to dementia or last observation. To test our third hypothesis regarding independent and synergistic contributions of networks to dementia, we constructed a multivariable Cox regression model based on the results of simple models and by examining the Pearson product moment correlation of predictor variables across all patients (see Results). All statistical analyses were performed using SPSS Statistics v27 (IBM Corp.). Prior to model fitting, all network-level predictor variables were inverted so that more positive W scores would indicate greater atrophy. Statistical significance was assessed at p < 0.05. We additionally denoted results surviving correction for multiple comparisons through Bonferroni correction. Specifically, for simple Cox regression models, we adopted a Bonferroni-corrected threshold of p < 0.003, corresponding to α = 0.05 divided by the number of models (7 networks per hemisphere and 3 covariates of no interest). For multivariable models, a Bonferroni-corrected threshold of p < 0.025 was used to correct for 2 models, each corresponding to a hemisphere.

Standard Protocol Approvals, Registrations, and Patient Consents

The study design and protocol were approved by the Mass General Brigham Institutional Review Boards for human research. Each participant and their informant gave written informed consent in accordance with the Mass General Brigham Human Subjects Research Committee guidelines.

Data Availability

Data not provided in the article because of space limitations may be shared (anonymized) at the request of any qualified investigator for purposes of replicating procedures and results.

Results

One patient was excluded from data analyses because of suboptimal quality of cortical surface reconstruction of the research MRI scan, leaving the final patient sample of 49 patients with PPA. Full demographic and clinical characteristics of these patients are summarized in Table 1. Sixty-two percentage of patients reported as being White and non-Hispanic, whereas information about race and ethnicity was not available in the remainder of the patients. The clinical follow-up interval after baseline assessment ranged from 5 to 79 months (M = 30.39 ± 18.47 months). At some point during their follow-up, 39 of 49 patients (79.6%) were diagnosed with dementia; the remaining 10 patients had a cognitive functional status of MCI at last follow-up, which ranged from 5 to 50 months (M = 31.5 ± 16.2). Demographic and clinical characteristics related to PPA subtype classification are reported in eResults and eTable 1 (links.lww.com/WNL/C397). A group of 25 amyloid-negative cognitively normal individuals were additionally recruited as controls (M_age = 67.43 ± 4.84, 13M/12F), whose data were used to calculate for each patient the magnitude of cortical atrophy expressed as W scores.

Table 1.

Demographic and Baseline Clinical Features of Patients With PPA

Topography of Baseline Cortical Atrophy

To characterize the spatial topography of cortical atrophy at baseline in our sample, we calculated the mean vertex-wise, whole-cortex W score map across all patients with PPA. This analysis revealed a characteristic pattern of cortical atrophy affecting the perisylvian region with marked left lateralization, consistent with previous studies (Figure 2A). To characterize this distributed topography of atrophy in terms of functional networks, we investigated the percentage of all vertices with values beyond a W score threshold of −0.5 belonging to each of the 7 networks of interest. This analysis showed that subtle atrophy was also present in nonlanguage functional networks, including the ventral frontotemporal, salience, and frontoparietal control networks (Figure 2B). The topography of baseline cortical atrophy separately for each major PPA subtype is shown in eFigure 1 (links.lww.com/WNL/C397).

Figure 2. Spatial Topography of Cortical Atrophy in Primary Progressive Aphasia.

(A) Colored vertices on the cortical surface maps indicate areas where patients with PPA, on average, showed relatively greater atrophy than Aβ- controls at a vertex-wise threshold of W < −0.5. (B) Bar graph depicts the percentage of all vertices showing W < −0.5 at the group level falling within the boundaries of each functional network of interest. Aβ = amyloid β; LAN = language network; DM = default mode network; FPC = frontoparietal control network; DA = dorsal attention network; SAL = salience network; VFT = ventral frontotemporal network; VIS = visual network.

Neuroanatomical Predictors of Progression to Dementia

Simple Cox regression models of age, sex, and education indicated that these variables did not significantly predict time to dementia. We constructed a series of simple Cox regression models to test the effect of cortical atrophy in different functional networks at baseline on subsequent progression to dementia in all patients with PPA. These models demonstrated that the magnitude of baseline cortical atrophy in the language network and other heteromodal association and paralimbic (i.e., frontoparietal control, salience, default mode, and ventral frontotemporal) networks in the left hemisphere were predictive of the rate of progression to dementia in PPA (Table 2 and Figure 3). The salience network and the homologous language network in the right hemisphere were also significant predictors of progression to dementia, although to a weaker extent and did not survive correction for multiple comparisons (Bonferroni-corrected p < 0.003).

Table 2.

Results of Simple Cox Regression Analyses to Predict Time of Progression to Dementia in Primary Progressive Aphasia

Figure 3. Hazard Ratios Obtained From Simple Cox Regression Analyses to Predict Time of Progression to Dementia in Primary Progressive Aphasia.

Only the hazard ratios (HRs) that were identified as statistically significant (at uncorrected p ≤ 0.05) are shown; for full results, see Table 2.

To examine partially independent or synergistic contributions of atrophy in different networks to progression to dementia, we constructed a multivariable Cox regression model. Our focus was the left frontoparietal control network, which we found to be the strongest predictor of progression to dementia in all patients with PPA (hazard ratio = 1.75 for each 1 SD greater magnitude of atrophy, 95% CI = [1.25–2.44]). More specifically, for every 1 SD of greater magnitude of atrophy in the frontoparietal control network, the patient's risk of progression was 1.75 times greater. To examine collinearity between the frontoparietal control network and the rest of the networks of interest, we computed the Pearson product moment correlation in mean W scores across patients between these networks in the left hemisphere. This analysis showed that, while generally a high degree of collinearity was observed, atrophy in the left frontoparietal control network was least correlated with that in the left ventral frontotemporal network: r(49) = 0.329, p ≤ 0.021 (eTable 2, links.lww.com/WNL/C397). Based on this observation, we constructed a multivariable Cox regression model with atrophy in the left frontoparietal control network, ventral frontotemporal network, and language network as predictors, while controlling for the effect of age, sex, and education. This analysis revealed significant effects of the frontoparietal control network and the ventral frontotemporal network, but not of the language network (Table 3). The corresponding model constructed using data from the right hemisphere yielded significant effects neither at the level of omnibus tests nor individual predictors. This suggests that the left frontoparietal control and ventral frontotemporal networks contribute uniquely to the prediction of the rate of progression to dementia, but that the left language network does not uniquely contribute to this prediction. Univariate survival curves for each of these networks are plotted in Figure 4. For the interested reader, we report the estimates of raw regional cortical thickness based on an anatomical parcellation of the cerebral cortex⁴⁰ (eTable 3 in the Supplement) and the results of simple Cox regression analyses to predict the time of progression to dementia using these data (eTable 4 in the Supplement).

Table 3.

Results of Multivariable Cox Regression Analyses to Predict Time of Progression to Dementia in PPA Based on Network-Specific Cortical Atrophy

Figure 4. Univariate Survival Curves Showing the Cumulative Proportion of Patients Who Have Progressed to Dementia as a Function of the Level of Atrophy.

Patients with PPA with relatively greater atrophy (i.e., cortical thickness that is more than 1 SD thinner than the group mean), represented as a solid black line in each plot above, showed more rapid progression to dementia than those with relatively lesser atrophy (i.e., cortical thickness that is more than 1 SD thicker than the group mean).

Discussion

Most patients with PPA, when diagnosed in a timely fashion, exhibit mild symptoms of language dysfunction, mild impairments on testing attributable to language dysfunction, and if advanced or instrumental activities of daily living are compromised, they are attributable to language impairment. Although this may affect a patient's independence in their occupation if it is heavily dependent on language (e.g., teaching), patients with early-stage PPA are often largely functionally independent, consistent with a cognitive functional status of MCI. Yet unfortunately, most patients will decline progressively not only in language skills but also in other aspects of cognition and behavior such that they lose independence—that is, sooner or later, patients with PPA almost inevitably progress to dementia. The time that this progression to dementia may take varies widely between patients, and at present, there are no tools to estimate this critical element of prognosis.

In this study of patients at the MCI stage of PPA, we investigated whether atrophy in cortical networks that subserve many facets of cognition and behavior—and are therefore key to the maintenance of functional independence—predicts progression to dementia. Our analytical framework examining cortical atrophy at the level of functional networks is consistent with a “network model” of neurodegeneration including PPA,^30,41 demonstrating selective vulnerability of networks and its effect on specific cognitive and behavioral domains. In this study, we showed that the magnitude of atrophy in the large-scale language network taken as a whole predicted progression to dementia. Of importance, we found that atrophy in cortical networks subserving nonlanguage cognitive and affective functions was a better predictor of progression to dementia than atrophy in the language network itself. Specifically, relatively greater atrophy in the frontoparietal control and ventral frontotemporal networks were the most important predictors of progression to dementia in the current sample of patients with PPA, and these networks exhibited partially independent and additive effects predicting time to dementia. Thus, this imaging biomarker has potential clinical and research applications as a prognostication tool.

The spatial topography of baseline cortical atrophy across patients was highly consistent with previous reports, revealing a unique anatomical distribution of atrophy predominantly affecting the perisylvian cortical regions of the language network.^2,3,19,42,43 While the magnitude of cortical atrophy was most prominent and consistent across patients within the left language network, more subtle and variable atrophy was also observable in nonlanguage networks involving other heteromodal association and paralimbic areas. When examined as individual predictors, each of these functional networks along with the language network predicted progression to dementia across all patients. However, our multivariable model revealed that atrophy in the frontoparietal control and ventral frontotemporal networks have partially statistically independent contributions to the prediction of the rate of progression to dementia, whereas atrophy in the language network explains no additional variance in the data. This is likely because each patient has relatively prominent language network atrophy—and therefore, this variable has less dynamic range as a predictor—but there is greater variability in whether patients have subtle atrophy in other networks with some patients having relatively intact cortical tissue in these nonlanguage networks. In our clinical experience, some patients with PPA at the MCI stage have subtle executive or memory dysfunction or behavioral symptomatology, although those, by definition, are very mild and do not affect daily function. Nevertheless, these symptoms are referable to early dysfunction of networks that may include frontoparietal control and ventral frontotemporal networks. Thus, when measured at the MCI stage of PPA, atrophy in the frontoparietal control and ventral frontotemporal networks could be useful as early neuroanatomical predictors of the emergence of impairment in nonlanguage cognitive and affective functions, leading eventually to a loss of independent functioning and dementia.

The frontoparietal control network includes the dorsolateral prefrontal cortex along the middle frontal gyrus, inferior parietal lobule, dorsomedial prefrontal cortex, and dorsal precuneus.^22,23,36 In some nomenclatures, cortical functional networks with a similar topography have been variably referred to as the “central executive” network.⁴⁴ The frontoparietal control network is critical for executive functions, guiding adaptive behavior according to current task demands and desired outcomes while also exerting top-down modulation of processing in other brain regions.^22,45

The ventral frontotemporal network consists of a smaller set of regions that includes the anterior temporal lobe and the subgenual anterior cingulate cortex extending into medial orbitofrontal cortex. This network is also variably known as the “limbic” network³⁷ or the “semantic appraisal” network.²⁸ The ventral frontotemporal network is hypothesized to play a crucial role in the regulation of the body's internal milieu through coordination with subcortical (e.g., hypothalamus) and brainstem nuclei.⁴⁶ This network has also been linked to socioaffective processes, including hedonic evaluations of positive and negative valence that guide processing of social and emotional concepts²⁸ and affiliation with social others.⁴⁷ Current evidence suggests that the 2 functional networks discussed here support unique and broad aspects of cognitive and affective processes. Loss of structural integrity in these networks may be one factor that accelerates cognitive-behavioral symptoms in nonlanguage domains observed as PPA progresses into one of several dementia phenotypes with not only language but also executive, memory, and/or socioaffective dysfunction.^8,9

Some patients with PPA are able to live a high quality of life without symptoms beyond language dysfunction for many years (e.g., up to 10 years), whereas others lose the ability to perform daily responsibilities much sooner.^11,12 The difficulty in predicting the course of illness has been reported as a major source of caregiver distress,⁴⁸ and greater caregiver burden is associated with loss of independence in instrumental activities of daily living.⁴⁹ The imaging biomarker identified here may allow the clinician to estimate the duration of the MCI stage, which may be useful in planning psychoeducational support programs for patients and caregivers⁵⁰ and potentially for determining goals in speech-language therapy. Because speech therapy in PPA is most useful when a patient is in the earlier (MCI) stage of the disease,^51,52 neurodegeneration in networks beyond the language network may also be helpful in planning the introduction of compensatory strategies based on a patient's likelihood of progression to dementia and how soon this progression might be likely to occur.

This study has some limitations. Despite current evidence pointing to the utility of baseline cortical atrophy in nonlanguage functional networks in predicting dementia progression, it remains unclear to what extent this is specific to PPA or generalizable to other neurodegenerative syndromes. Future work should investigate the relationship between network-specific atrophy and subsequent dementia in patients representing a wider syndromic spectrum of Alzheimer disease or frontotemporal lobar degeneration. Second, while the present sample was heterogeneous, the size of each PPA subtype subsample was modest. Some PPA subtypes were comparable in size with those examined in previous MRI studies.¹⁹ However, because of the small number of patients who remained at the MCI stage of PPA in each subtype, we were limited in our ability to analyze subgroup differences in this study. Future studies with increased statistical power would be useful in examining the contribution of different networks to dementia within each PPA subtype.

Notwithstanding these limitations, the present findings support further translational research to bring relatively straightforward computational analyses of structural brain MRI into clinical practice for the purposes of not only diagnosis but also prognostication. A patient with PPA with relatively preserved cortical thickness in the frontoparietal control network and the ventral frontotemporal network may be more likely to remain in the MCI stage for longer, maintaining relatively independent daily function and potentially the capacity to capitalize on therapeutic strategies to compensate for language impairment. This simple measure based on a structural MRI scan could aid the development of a personalized prognostic and monitoring plan tailored to each patient, which would in turn help clinicians choose appropriate treatments aiming to maximize the well-being of the patient and their families.

Glossary

¹¹C-PiB

¹¹C-Pittsburgh compound B

DVR

distribution volume ratio

FLR

frontal, lateral, and retrosplenial

MCI

mild cognitive impairment

MGH

Massachusetts General Hospital

MMSE

Mini-Mental State Examination

PPA

primary progressive aphasia

Appendix. Authors

Study Funding

This research was supported by NIH grants R01 DC014296, R21 DC019567, R21 AG073744, K23 DC016912, P50 AG005134, and P30 AG062421 and by the Tommy Rickles Chair in Primary Progressive Aphasia Research. This research was performed in part at the Athinoula A. Martinos Center for Biomedical Imaging at the MGH, using resources provided by the Center for Functional Neuroimaging Technologies, P41EB015896, a P41 Biotechnology Resource Grant supported by the National Institute of Biomedical Imaging and Bioengineering (NIBIB), NIH. This work also involved the use of instrumentation supported by the NIH Shared Instrumentation Grant Program and/or High-End Instrumentation Grant Program; specifically, grant number(s) S10RR021110, S10RR023043, and S10RR023401.

Disclosure

Y. Katsumi, M. Quimby, D. Hochberg, A. Jones, M. Brickhouse, and M. C. Eldaief report no disclosures; B. C. Dickerson was funded by NIH grants No.R01DC014296, R21DC019567, R21AG073744, P50AG005134, and P30AG062421 and the Tommy Rickles Chair in Primary Progressive Aphasia Research at the Massachusetts General Hospital; A. Touroutoglou was funded by NIH grant No.K23DC016912. Go to Neurology.org/N for full disclosures.