摘要Traditional clinical subtype classifications(such as amnestic and non-amnestic mild cognitive impairment)rely on subjective interpretations of overlapping patterns of performance on cognitive tests,which may lead to unreliable categorization.A more precise and objective classification of mild cognitive impairment subtypes can be achieved through data-driven clustering techniques.However,because previous studies have not restricted their cohorts to patients who have mild cognitive impairment with the pathology of Alzheimer's disease,the nature of cognitive variability and its impact on disease progression in a strictly defined biomarker-positive preclinical Alzheimer's disease cohort remains unknown.We examined cognitive heterogeneity among participants with mild cognitive impairment due to Alzheimer's disease and evaluated its prognostic utility.Neuropsychological test data from 389 patients with mild cognitive impairment in whom the cerebrospinal fluid biomarker confirmed Alzheimer's disease were obtained from the Alzheimer's Disease Neuroimaging Initiative cohorts.Principal component analysis and model-based clustering were used to identify cognitive profiles,which were then validated through a 100-time bootstrap analysis.Pairwise comparisons tested for differences between the identified subgroups in participant characteristics,scores on cognitive and clinical outcomes,levels of cerebrospinal fluid biomarkers,and magnetic resonance imaging-derived brain volumes.Longitudinal analyses evaluated differences in rate of change of magnetic resonance imaging volumetric measurements and clinical outcomes over 48 months.Survival analysis assessed risk for conversion to dementia.Alpha-synuclein levels and white matter hyperintensity volumes were considered for sensitivity analysis.Two distinct cognitive profiles were identified:a"typical"group(56.04％of the sample)that demonstrated relatively poorer scores on memory testing than non-memory tests,and an"atypical"group(43.96％of the sample)with smaller differences between memory and non-memory measures,indicating a more uniform pattern of impairment across cognitive domains.While the groups had comparable levels of overall cognitive impairment and cerebrospinal fluid biomarkers of Alzheimer's disease,the typical group displayed accelerated atrophy rates every 6 months across multiple brain regions(hippocampus:29.02 mm3,standard error[SE]=10.13,P=0.005;whole brain:1799.85 mm3,SE=781.57,P=0.023;entorhinal cortex:22.26 mm3,SE=11.15,P=0.048;fusiform gyrus:66.24 mm3,SE=28.53,P=0.021).Survival analysis revealed markedly higher dementia conversion risk(hazard ratio:1.70,95％confidence interval:1.27,2.27,P＜0.001)and shorter progression time in the typical group.These findings persisted after controlling for comorbid pathologies.In conclusion,this data-driven approach identified two distinct cognitive subtypes of mild cognitive impairment due to Alzheimer's disease that differed in their rates of clinical decline and neurodegeneration.These findings could be used to improve prognostic models and inform clinical trial stratification.