Visit-to-visit blood pressure variability and the risk of stroke in the Netherlands: A population-based cohort study

Abstract

Background: Apart from blood pressure level itself, variation in blood pressure has been implicated in the development of stroke in subgroups at high cardiovascular risk. We determined the association between visit-to-visit blood pressure variability and stroke risk in the general population, taking into account the size and direction of variation and several time intervals prior to stroke diagnosis.

Methods and findings: From 1990 to 2016, we included 9,958 stroke-free participants of the population-based Rotterdam Study in the Netherlands. This is a prospective cohort study including participants aged 45 years and older. Systolic blood pressure (SBP) variability was calculated as absolute SBP difference divided by mean SBP over 2 sequential visits (median 4.6 years apart). Directional SBP variability was defined as SBP difference over 2 visits divided by mean SBP. Using time-varying Cox proportional hazards models adjusted for age, sex, mean SBP, and cardiovascular risk factors, hazard ratios (HRs) for stroke up to January 2016 were estimated per SD increase and in tertiles of variability. We also conducted analyses with 3-, 6-, and 9-year intervals between variability measurement and stroke assessment. These analyses were repeated for diastolic blood pressure (DBP). The mean age of the study population was 67.4 ± 8.2 years and 5,776 (58.0%) were women. During a median follow-up of 10.1 years, 971 (9.8%) participants had a stroke, including 641 ischemic, 89 hemorrhagic, and 241 unspecified strokes. SBP variability was associated with an increased risk of hemorrhagic stroke (HR per SD 1.27, 95% CI 1.05-1.54, p = 0.02) and unspecified stroke (HR per SD 1.21, 95% CI 1.09-1.34, p < 0.001). The associations were stronger for all stroke subtypes with longer time intervals; the HR for any stroke was 1.29 (95% CI 1.21-1.36, p < 0.001) at 3 years, 1.47 (95% CI 1.35-1.59, p < 0.001) at 6 years, and 1.38 (95%CI 1.24-1.51, p < 0.001) at 9 years. For DBP variability, we found an association with unspecified stroke risk. Both the rise and fall of SBP and the fall of DBP were associated with an increased risk for unspecified stroke. Limitations of the study include that, due to an average interval of 4 years between visits, our findings may not be generalizable to blood pressure variability over shorter periods.

Conclusions: In this population-based study, we found that visit-to-visit blood pressure variation was associated with an increased risk of unspecified and hemorrhagic stroke, independent of direction of variation or mean blood pressure.

Fig 1. Schematic presentation of the analyses with and without a lag period.

(A) Analyses without lag periods, in which blood pressure (BP) variability is measured over 2 visits with immediate follow-up for stroke incidence thereafter. Blood pressure variability is then updated when the participant visits the research center again. (B) Analyses with lag periods. After blood pressure variability measurement, a lag period of 3, 6, or 9 years is incorporated before follow-up for incident stroke to account for reversed causation.

Fig 2. The association between blood pressure variability and incident stroke using different lag periods.

The blue dots and green squares represent the hazard ratio of incident stroke per SD increase in systolic blood pressure (SBP) variability and diastolic blood pressure (DBP) variability, respectively. The error bars show the 95% confidence intervals around the estimates. The lag 0 models were adjusted for age, sex, mean SBP or DBP, education, BMI, smoking, alcohol consumption, cholesterol, high-density lipoprotein cholesterol, lipid-lowering medication use, blood-pressure-lowering medication use, anti-thrombotic medication use, type 2 diabetes mellitus, atrial fibrillation, prior coronary heart disease, and type of blood pressure measuring device. Lag 3, lag 6, and lag 9 analyses were adjusted for age, sex, mean SBP or DBP, and propensity score based on all other covariates. Lag 0 represents no time interval between blood pressure variability measurement and follow-up for incident stroke, lag 3 represents a time interval of 3 years between blood pressure variability measurement and follow-up for incident stroke, and so on.

Fig 3. Stratified analyses of the association between systolic blood pressure (BP) variability and incident any stroke using different lag periods.

The black squares represent the hazard ratio (HR) of incident any stroke per SD increase in systolic blood pressure variability. The error bars are the 95% confidence intervals (CIs) around the hazard ratios. The models were adjusted for age, sex, mean systolic blood pressure, and propensity score.