Objective  To analyze the bidirectional longitudinal relationship between blood pressure and brachial-ankle pulse wave velocity (baPWV) in non-hypertensive populations.

Methods Analyze the data of individuals who underwent check-ups in Shanghai Health and Medical Center at least twice from 2015 to 2022. Use logistic regression analysis to assess the impact of baseline baPWV on patients with newly diagnosed hypertension, and use receiver operating characteristic curve (ROC) analysis and Delong test to compare the predictive performance of different prediction models. Use a linear mixed model to analyze the bidirectional longitudinal relationship between mean arterial pressure, systolic blood pressure, diastolic blood pressure, pulse pressure and baPWV.

Results The data of 1 873 non-hypertensive individuals who met the criteria were collected, including 1 384 males (73.9%) and 489 females (26.1%), aged between 23 and 86 years, with an average age of (47.3±8.5) years, and the follow-up time interval was (3.4±0.2) years. Grouped by tertiles of baseline baPWV. There were 628 cases in the low percentile group, 623 cases in the middle percentile group, and 622 cases in the high percentile group. Logistic regression analysis suggested that baPWV was an independent risk factor for the incidence of hypertension. Taking the low percentile group as a reference, the OR(95%CI) of hypertension incidence in the middle and high percentile groups was 1.897(95%CI: 1.075 to 3.347) and 3.189(95%CI: 1.851 to 5.495), respectively, with P＜0.05. In the study of predicting new-onset hypertension by baseline baPWV, the Akaike's information condition (AIC) value of the model with both systolic and diastolic blood pressure included in the baseline blood pressure index was the lowest, and the area under the ROC curve was the largest, (0.810, 95%CI 0.780 to 0.839). Delong test analysis showed that the difference between the model with baseline systolic and diastolic blood pressure and the model without baseline blood pressure and the model with baseline diastolic blood pressure was statistically significant (P＜0.05). Linear mixed model analysis was performed on 1 821 patients who had never taken antihypertensive drugs. After adjusting the influencing factors such as baseline value and age, the baseline baPWV had an effect on follow-up systolic blood pressure, follow-up diastolic blood pressure, follow-up mean arterial pressure and follow-up pulse pressure, β values were (0.181, 95%CI: 0.139 to 0.223), 0.137 (95%CI: 0.089 to 0.185), 0.151 (95%CI: 0.103 to 0.199) and 0.111 (95%CI: 0.060 to 0.161) respectively; and the baseline systolic blood pressure, baseline diastolic blood pressure, baseline mean arterial pressure and baseline pulse pressure also had an impact on follow-up baPWV, β values were 0.056 (95%CI 0.025 to 0.087), 0.040 (95%CI: 0.010 to 0.071), 0.054 (95%CI: 0.022 to 0.085), and 0.031 (95%CI: 0.002 to 0.060) respectively. Grouped by gender or age, the interaction between systolic blood pressure and baPWV persisted in the subgroup analysis (P＜0.05).

Conclusions In non-hypertensive individuals, the baPWV is longitudinally correlated with systolic blood pressure. When predicting new-onset hypertension with baPWV, the model with baseline systolic and diastolic blood pressure may have a better prediction effect.