We studied a total of 2,546 individuals (1,386 men and 1160 women) with a mean age of 60.8 ± 10.3 years (range 27–95). All participants voluntarily underwent the brain health check-up at the Shimane Health Science Center between April 2004 and July 2015. The assessment included medical history, neurological examination by an experienced neurologist, blood pressure, neuropsychological testing, and MRI scans of the head. The criteria for subject exclusion were as follows: any history of neurological or psychiatric conditions, such as cerebrovascular diseases including transient ischemic attack, dementia, depression, or other psychiatric diseases, and missing data. All individuals provided informed consent to participate this study, which was approved by the institutional ethics committee.
During the heath check-up, seated brachial artery systolic and diastolic BP were measured after 15 min of rest. PP was defined as the difference between systolic and diastolic BP. Hypertension was defined as having a systolic BP > 140 mmHg, a diastolic BP > 90 mmHg, or a history of hypertension with antihypertensive therapy. Diabetes mellitus was defined by fasting glucose level exceeding 126 mg/dl, random glucose level exceeding 200 mg/dl, an HbA1c level exceeding 6.5% and/or a medical history of self-reported history of diabetes or treatment with oral antidiabetic drugs or insulin. Dyslipidemia was defined by serum triglyceride level exceeding 150 mg/dl, high-density lipoprotein cholesterol level below 40 mg/dl, or a medical history of dyslipidemia.
Brain infarction was defined as a focal hyperintense lesion ≥3 mm in diameter on T2-weighted images. Fluid-attenuated inversion recovery images were used to differentiate infarcts from enlarged perivascular spaces. Cerebral microbleeds (CMBs) were defined as homogenous round foci of signal loss on gradient-echo T2*-weighted images that were 2-10 mm in diameter. Periventricular hyperintensities (PVHs) and white matter hyperintensities (WMHs) were evaluated based on their distinct subcortical distributions on fluid-attenuated inversion recovery images. Periventricular hyperintensity (PVH) and subcortical white matter hyperintensity (SWML) were evaluated separately based on their distinct subcortical distributions on the fluid-attenuated inversion recovery image, because PVH was observed adjacent to the ventricles and SWML was observed separately from the ventricles. PVH was graded on a scale of 0 – 4, as described previously . SWML was graded on a scale of 0 – 3 according to the Fazekas grading scheme . For statistical purposes, PVH and SWML grades were dichotomized; we defined PVH grades 0 – 2 as ‘PVH–’ and grades 3 – 4 as ‘PVH+’; similarly, SWML grades 0 – 1 were defined as ‘SWML–’, and grades 2 – 3 were termed ‘SWML+’. CMBs were identified as 2–10 mm diameter rounded hypointense lesions on T2*-weighted images. All MRI findings were evaluated separately by an experienced neurologist and a radiologist who were blinded to the patients’ profiles. When their opinions were inconsistent, a second neurologist was consulted. An interrater study for evaluating MRI lesions was performed blindly by two independent raters.
Cognitive function evaluation
General cognitive function was assessed using Okabe’s Intelligence Scale (Okabe’s test) , which is a shortened and modified Wechsler Adult Intelligence Scale-Revised for the Japanese aged population and includes orientation, semantic memory, calculation, forward and backward digit span, and paired association memory. The test scores a total of 60 points, and its reliability has been previously validated . There was a significant correlation between Okabe’s test and the Mini Mental State Examination (MMSE) [8,9]. The Kohs’ block design test (Kohs’ test) is a popular bedside screening test for constructional function and cognitive function. The subjects were shown cards with a variety of colored designs and were asked to reproduce them using a set of colored blocks, yielding an intelligence quotient . This test assessed the visuospatial ability in addition to executive function. Frontal function was estimated using the frontal assessment battery (FAB) . Affective functions were evaluated using the Self-rating Depression Scale (SDS)  and the Japanese version of the Apathy Scale .
We divided all subjects into the high and low PP groups with a cut-off value of 65 mmHg . To avoid possible confounding effects caused by grouping with PP values, we used propensity score matching in this study. To obtain the propensity score, we assigned these two groups to the explanatory variable and performed logistic regression analysis after correcting for covariates such as age, sex, and medical histories of hypertension, diabetes mellitus, and dyslipidemia. Propensity score matching was performed using the following algorithm: 1:1 ratio nearest-neighbor match with ±0.01 caliper and no replacement. Physical examination results, silent brain lesions, and cognitive functions were compared between the high and low PP groups before and after propensity matching using a two-sided Student’s t-test for continuous variables and χ2 analyses for categorical variables. To evaluate whether the association between PP and cognitive function was affected by systolic BP (SBP), mediation analysis was performed using Sobel. All statistical analyses were performed using IBM SPSS Statistics ver. 22 (SPSS, Inc.). Differences were considered significant at P < 0.05.