An outcome-driven threshold for pulse pressure amplification

Figures

• 

  Fig. 1. Cumulative incidence of the cardiovascular and coronary endpoints by tertiles of pulse pressure amplification. Cumulative incidence of the cardiovascular (A) and coronary (B) endpoints was derived by Cox regression with adjustment for cohort, sex and age. P values denote the overall significance of the difference between the PPA categories. Vertical lines denote the SE. Tabulated data are the number of participants at risk at 5-year intervals

• 

  Fig. 2. Threshold and calibration of pulse pressure amplification (PPA) in the discovery sample including 3945 participants. Hazard ratios (HRs) express the risk at each PPA level relative to the average risk in the whole discovery sample for cardiovascular (A) and coronary endpoints (B). The upper confidence limit crosses unity, denoted by the vertical line, at 1.28 and 1.26 for cardiovascular and coronary endpoints, signifying decreased risk. The PPA levels yielding equivalent 5-year risks compared with a systolic blood pressure of 140 mmHg were 1.28 and 1.29 for the cardiovascular (C) and coronary (D) endpoints, respectively. Model calibration for the cardiovascular (E) and coronary (F) endpoints demonstrated that across PPA quintiles the predicted risk was similar compared with the overoptimism-corrected Kaplan–Meier estimates. All analyses were multivariable adjusted for cohort (random effect), sex, age, body mass index, heart rate, smoking and drinking, total-to-HDL serum cholesterol ratio, estimated glomerular filtration rate, antihypertensive drug intake, history of cardiovascular disease, and diabetes

• 

  Fig. 3. Hazard ratios expressing the risk of cardiovascular and coronary endpoints per the 1.3 pulse pressure amplification threshold by subgroups. Hazard ratios, given with 95% confidence interval accounted for cohort (random effect), sex, age, body mass index, heart rate, smoking and drinking, the total-to-HDL serum cholesterol ratio, the estimated glomerular filtration rate, antihypertensive drug intake, history of cardiovascular disease and diabetes. Adjustment for sex, age and antihypertensive treatment was omitted, if these variables defined the strata. Squares represent the point estimates of the hazard ratios (HR). Horizontal lines denote 95% confidence interval. NLE/NLR vs NHE/NHR refer to the number of participants with an event/number of participants at risk dichotomized by the threshold of pulse pressure amplification (<1.3 vs ≥1.3, respectively)