Comparison of subjective cognitive decline and polygenic risk score in the prediction of all‐cause dementia, Alzheimer’s disease and vascular dementia

Abstract

AbstractBackgroundSubjective cognitive decline (SCD) and polygenic risk scores (PRS) have both been shown to be associated with the risk of developing dementia. However, it is of interest whether they can improve the established and well‐validated cardiovascular risk factors aging and dementia (CAIDE) model and how their predictive abilities compare.MethodsSCD and PRS were assessed and determined in a large, population‐based cohort study. In this study, the CAIDE model was applied to a sub‐sample of 5,360 study participants and evaluated for the outcomes of all‐cause dementia, Alzheimer’s disease (AD) and vascular dementia (VD) by calculating Akaike’s information criterion (AIC) and the area under the curve (AUC). The improvement of the CAIDE model by SCD and PRS was further examined using the net reclassification improvement (NRI) method and integrated discrimination improvement (IDI). Besides evaluations in the total sample, predictive abilities were evaluated in the mid‐life and late‐life sub‐cohort.ResultsDuring 17 years of follow‐up, 410 participants were diagnosed with dementia, including 139 AD and 152 VD diagnoses. Overall, the CAIDE model showed high discriminative ability for all outcomes reaching AUCs of 0.785, 0.793 and 0.789 for all‐cause dementia, AD, and VD, respectively. Adding information on SCD to the CAIDE model led to a significant increase in NRI for all‐cause dementia (4.4%, p = 0.0401) and VD (7.7%, p = 0.0095). In contrast, prediction models for AD further improved when PRS was added to the model (NRI, 8.4%, p = 0.0257). The latter became more substantial when applied to the mid‐life sub‐group (50‐64 years), achieving an NRI of 19.6% (p = 0.0075). When APOE ε4 carrier status was included in CAIDE model 2, the discriminative ability of the basic CAIDE model increased, and SCD and PRS did not further improve the prediction.ConclusionWhile PRS significantly improved the prediction of the CAIDE model for AD, information on SCD increased its discriminative ability for all‐cause dementia and VD. In contrast to PRS, information on SCD can be assessed more efficiently, and thus models including SCD can be more easily transferred to the clinical setting. Nevertheless, the two variables seem negligible if APOE ε4 carrier status is available.