“…To examine the presence of a U-shaped association (nonlinear) between PP with cognitive decline, PP was categorized into quartiles and entered the model as a categorical variable. [19][20][21] Cognitive decline was indexed by CDR and MMSE changes, respectively. Putative confounding factors, including baseline age, sex, education, DM, hyperlipidemia, smoking, alcohol, AF or CHF, recurrent stroke, VBR, MTLA, lacunes, 22 antihypertensive therapy on discharge, and admission NIHSS score were first examined in a series of exploratory univariable models, and only variables with P<.10 in univariable models were then entered into multivariate regression models.…”