AbstractBackgroundAlzheimer’s disease (AD) is a complex disease conditioned by environmental and genetic factors. Over the last few years, Genome‐Wide Association Studies (GWAS) have gained increasing interest representing the basis of a personalized medicine that aims to facilitate interventions at the preclinical stages. In most studies, the integration of the high‐dimensional GWAS data is accomplished via polygenic risk scores (PRSs) representing a measure of the genetic risk associated with a given phenotype. For the first time, this study aimed to explore the predictive value of a PRS developed to predict conversion to AD in patients with mild cognitive impairment (MCI) (AD‐PRS) (Bellenguez et al., 2022) on cognitive decline once the AD dementia stage has been establishedMethodThe modulatory effect of genetics on cognitive decline was studied using a sample of 3921 subjects with AD dementia (clinical dementia rating (CDR) ≥1) evaluated at Ace Alzheimer Center Barcelona. Cognitive decline, operationalized as changes in Mini Mental State Examination (MMSE) scores over time, was modeled using multivariate regression models considering clinical and sociodemographic data, and the AD‐PRS. Interaction effects between the AD‐PRS and the other covariates were explored, and subsequently, mediation analyses were applied. Afterward, Machine Learning techniques were used to select the single nucleotide polymorphisms (SNPs) forming the AD‐PRS with the highest predictive value on cognitive decline.ResultMultivariate regression models showed a significant interaction of the AD‐PRS with years of education (coef = .0006; p‐value = .005). Increased scores of AD‐PRS combined with higher formal education were found to be associated with greater cognitive decline. Furthermore, Machine Learning models identified 18 SNPs related to cognitive decline in AD, demonstrating that genetic information improves the predictive performance of the models.ConclusionThis is the first study that demonstrates the predictive capacity of the AD‐PRS on cognitive decline in AD dementia. Years of education were shown to modulate, but not mediate, the effect of the AD‐PRS on cognitive progression. Additionally, a subset of SNPs comprising the AD‐PRS with the highest predictive value of cognitive decline in AD was identified. Altogether, these results provide new insights into the genetics of the disease.