Data‐driven comparison of neurodegeneration patterns between autosomal dominant and sporadic late‐onset Alzheimer’s disease

Abstract

AbstractBackgroundAutosomal dominant Alzheimer’s disease (ADAD) and sporadic late‐onset AD (LOAD) have similar biomarker profiles, supporting a shared pathophysiology. While both ADAD and LOAD are known to be heterogeneous, there are few studies that have compared the heterogeneity of neurodegeneration as indexed by volumetric MRI in LOAD to that of ADAD. Here we apply a data‐driven model of brain atrophy subtypes from LOAD to characterize atrophy within ADAD.MethodWe used cross‐sectional MRI from ADNI (LOAD) and the Estudio de la Enfermaded de Alzheimer en Jalisciences (EEAJ; ADAD). We used 13 cortical and subcortical Freesurfer‐based regions from 1,912 ADNI participants to train a Subtype and Stage Inference (SuStaIn) model of brain atrophy progression within LOAD. We built a three‐subtype model and applied it to subtype and stage 61 EEAJ participants: 47 carriers of the A431E mutation in PSEN1 and 14 carriers of the V717I mutation in APP. To harmonize the two datasets, we removed the effects of age, sex, education years, ICV and study (ADNI vs. EEAJ flag) from each regional volumetric biomarker prior to analysis. We retained 47 of the 61 EEAJ participants with an estimate SuStaIn stage greater than zero (in any subtype) for subsequent analyses.ResultThe progression patterns for each of the three data‐driven brain atrophy subtypes (typical, cortical, subcortical) are presented in Figures 1 A‐C. These are consistent with previous findings in LOAD. We found no difference in the percentage of participants in each subtype between LOAD, ADAD‐PSEN1 and ADAD‐APP groups (χ2 = 1.5, p = 0.84; Figure 1D). Within ADAD, those assigned the typical subtype had a slightly lower average stage of progression than those in the other two subtypes (t = ‐2.0, p = 0.049; Figure 2A). The triangle plot shows the probability of subtype membership is mostly clustered around the typical and cortical subtypes (Figure 2B), while the estimated stages increase dramatically after disease onset, as expected (Figure 2C).ConclusionWe found a comparable distribution of LOAD‐based atrophy subtypes within both the PSEN1 and APP carriers groups in ADAD, showing that heterogeneous neurodegeneration is another commonality between ADAD and LOAD.