MPS 14-08 Predictors of Visit-to-Visit Blood Pressure Variability (BPV) in Patients with Hypertension

Abstract

Objective: Blood pressure variability (BPV) is becoming an important consideration in the prediction of cardiovascular risk. Through a retrospective post hoc analysis, we explored potential risk factors for high visit-to-visit BPV, in hypertensive patients. Design and Method: Five studies (47,558 BPV-evaluable patients) with individual subject data, ≥1 active comparator, and hypertension treatment duration ≥12 weeks, were included. BPV was assessed using coefficient of variation (CV) of systolic BP across visits from 12-weeks. Individual trial and meta-analyses were performed for CV-based methodologies. Across studies, spread of BPV followed a highly skewed distribution; therefore, parametric procedures were applied to transform data into a symmetric distribution. Covariates considered included: age, gender, ethnicity, body mass index (BMI); history of stroke; smoking; baseline systolic and diastolic BP, blood glucose, eGFR; any antihypertensive medication; study treatment and study. Covariates were initially screened using hierarchical cluster analysis to address multicollinearity among factors; factors succeeded from the initial screen were fitted to a regression model in their original forms with no interactions, the residuals from this model were then used to detect proper functional forms and interactions. A backward step-down procedure for regression analysis was implemented to select risk factors for BPV. Results: Patient characteristics were largely consistent across 5-studies (64.2% male; mean [±SD] age 64.8[±8.4] years). After transformation, significant (P < 0.001) predictors of BPV included systolic and diastolic BP at baseline (both linear and quadratic terms), prior stroke, and the interaction terms for age and BMI with smoking. Conclusions: This large-scale analysis of five controlled trials highlights that data need transforming with specific statistical protocols prior to analyzing predictors of BPV. Following transformation, a number of known risk factors were significant predictors of BPV.