Main demographic, clinical and biomarker characteristics of the data set (Table 1). The percentages of the clusters sum to 74% of the sample, since a 26% (143 participants) have uncertain classification to any of the clusters (Spplementary Table 1). Health, how healthy the participant feels from 0 to 4 with 4 being worst; CMC, cardiovascular and metabolic conditions; CU, cognitively unimpaired; WMH/TIV × 100 (WMH in FLAIR MRI as a fraction of total intracranial volume; amyloid load, PiB SUVR-Pittsburgh compound B (PiB) PET SUVR] and mad, median absolute deviation. The four domains and their average refer to cognitive functionality, they are Z-values and higher is better cognitive score. For WMH/TIV × 100, amyloid load and the cognitive variables (Global, Memory, Executive, Language and Visuospatial), the data is presented as: estimated fitted value at the 60 years of age (estimated annual change after the age of 60 years). The values are estimated with mixed-effect models: Cluster 1 is the reference group for the models of WMH and PIB (reference group for the cognitive models are cluster 1, CU and female), single asterisk (*) signifies significant difference from 0 in intercept or slope of the reference group, whereas double asterisks (**) signify that a group is significantly different from the reference group in intercept or slope (all tests have a 0.05 confidence level and are corrected with the Satterthwaite’s method and corrected for multiple comparisons with the Holm–Šidák method).