Background: Given the long preclinical disease course of Alzheimer disease (AD) pathology, novel treatments may be more efficacious if administered before the emergence of dementia. Thus, accurate prediction of who will develop AD dementia is of key importance in selecting individuals for trials of treatment and may become crucial for future selection of patients for therapy.
Methods: As part of the Australian Imaging, Biomarkers and Lifestyle Flagship Study of Ageing, 901 individuals who did not have dementia were recruited. We assigned individuals according to Petersen criteria and Winblad criteria for Mild Cognitive Impairment (MCI) at baseline. We then stratified individuals with amnestic MCI into 2 groups according to the severity of their memory impairment on baseline neuropsychological assessment. Incident diagnosis of AD dementia was established by consensus of an expert panel at 36 months.
Results: At 36 months, 725 (80.5%) participants were followed up, 54 (7.4%) of whom developed AD dementia. Subjects with amnestic MCI according to Petersen criteria were more likely to develop AD dementia [positive predictive value; PPV, 24.1%; 95% confidence interval (CI), 18.4-30.6] than healthy controls (PPV, 1.0%; 95% CI, 0.3-2.3). Winblad criteria were also effective, with multiple domain amnestic MCI being most accurate at predicting AD dementia (PPV, 47.3%; 95% CI, 33.7-61.2). Finally, more severe amnestic impairment below the median was useful for predicting the development of AD dementia in single domain amnestic MCI (PPV, 28.1%; 95% CI, 17.0-41.5) and in multiple domain amnestic MCI (PPV, 65.7%; 95% CI, 47.8-80.9).
Conclusions: Memory impairment per se, impairment in multiple cognitive domains and severity of memory impairment were all associated with greater risk of developing AD dementia in this sample. Characterizing the severity of memory impairment may provide prognostic stratification within Petersen or Winblad taxonomies of amnestic MCI.