Posterior cortical atrophy and logopenic variant primary progressive aphasia are more specific for Alzheimer’s disease pathology than probable Alzheimer’s disease

Abstract

Accurately diagnosing Alzheimer’s disease is challenging, as 15–30% of individuals diagnosed clinically lack evidence of Alzheimer’s disease neuropathological changes (ADNC) at autopsy. While the typical amnestic presentation of Alzheimer’s disease may have limited specificity due to coexisting age-related pathologies, atypical presentations such as logopenic variant primary progressive aphasia (lvPPA) or posterior cortical atrophy (PCA) are often associated with ADNC. This study evaluated how well clinical diagnoses of amnestic and atypical Alzheimer’s disease predict ADNC and how co-pathologies contribute to clinical-pathological discordance. To evaluate how clinical diagnosis predicts ADNC, we calculated positive predictive values (PPVs) and negative predictive values (NPVs) for clinical diagnoses of PCA, lvPPA, amnestic Alzheimer’s disease or non-Alzheimer’s disease dementia for ADNC, using four neuropathological thresholds defined by combinations of the Consortium to Establish a Registry for Alzheimer’s disease (CERAD) neuritic plaque frequency (moderate or frequent) and Braak stage (III–VI or V–VI). Analyses were replicated in vivo using amyloid- and tau-PET positivity (A+T+) as a surrogate for Alzheimer’s disease pathology. We also quantified the frequency of co-pathologies and assessed their contribution to false-positive Alzheimer’s disease diagnoses. This study included 6363 autopsy-confirmed cases from the National Alzheimer’s Coordinating Center (NACC) database (mean age: 80.0 ± 11.4 years; 46.0% female) and 76 participants from the Translational Biomarkers in Aging and Dementia (TRIAD) cohort (mean age: 67.9 ± 9.94 years; 55.3% female). The PPV of probable Alzheimer’s disease dementia ranged from 58 to 80%, while PCA and lvPPA showed higher PPVs (81–96% and 80–92%, respectively) for ADNC. The NPV of probable Alzheimer’s disease dementia ranged from 66 to 83%, compared with 38–58% for PCA and lvPPA. In TRIAD, individuals with PCA or lvPPA had 100% PPV for Alzheimer’s disease biomarker abnormality, whereas amnestic Alzheimer’s disease dementia showed a PPV of 78.1%. Among Alzheimer’s disease dementia cases, 20.2% were ADNC-negative, with a higher frequency of false-positive diagnoses in late-onset compared with early-onset cases (23.8% versus 6.9%). Among ADNC-negative Alzheimer’s disease dementia cases, vascular pathology was the most common pathology (early-onset: 93%; late-onset: 98%). In the late-onset group, limbic-predominant age-related TDP-43 encephalopathy occurred in 31% of the cases, followed by Lewy body pathology (28.8%) and primary age-related tauopathy (24.3%). Visuospatial and language variants of clinical Alzheimer’s disease offer excellent predictive value for underlying Alzheimer’s disease pathology, both in vivo and at autopsy, outperforming the amnestic presentation classically associated with Alzheimer’s disease. Clinical-pathological discordance in amnestic Alzheimer’s disease dementia is frequently associated with coexisting age-related and vascular pathologies, highlighting the complexity of interpreting clinical Alzheimer’s disease diagnoses in the absence of ADNC.