Response to Brachial Pulse Pressure and Cardiovascular Risk
We thank Blacher et al1 for their interest in our recent article.2 We agree that current recommendations regarding clinical management of cardiovascular risk depend on an arbitrary definition of “hypertension.” However, we note that, like pulse pressure (PP), systolic (SBP) and diastolic blood pressure are also risk factors observed in both normotensive and hypertensive individuals. Our data support the well-described linear associations of SBP and diastolic blood pressure with cardiovascular risk, beginning at levels well below current thresholds for definition of hypertension.3
Collinearity certainly affects the statistical evaluation of SBP and PP when they are included in the same model. We reported a pairwise correlation coefficient of 0.78 for these 2 measures. In addition to presenting models that contained both SBP and PP simultaneously, therefore, we also examined them in separate models. In all of the cases, and for each outcome, SBP consistently revealed better predictive utility than PP, as indicated by greater statistical significance and superior model fit, informativeness, and discrimination. Principal component analysis would be another way to analyze these variables but would be unlikely likely to provide different results.
Although the mean age of our study sample was 39 years, this large cohort still included 9184 individuals aged ≥50 years, and 2364 aged ≥60 years, a group that experienced large numbers of events during follow-up. Because this was an employed sample, we had only 176 individuals aged ≥70 years, but SBP appeared to provide greater predictive utility than PP even in this small subset.
We agree that use of a single baseline blood pressure measurement and long-term follow-up raise concern for regression dilution bias. However, the Prospective Studies Collaboration3 adjusted for regression dilution in 1000000 participants and found that SBP and diastolic blood pressure provided greater informativeness for stroke and ischemic heart disease deaths than PP. In addition, whereas the presence of regression dilution may lead to underestimation of the true association between a baseline measure and outcome, its presence should not preclude the comparison of BP components measured at the same time.
Blacher J, Safar ME, Vesin C, Rudnichi A. Brachial pulse pressure and cardiovascular risk. Hypertension. 2007; 50: e161.
Mosley WJ, Greenland P, Garside DB, Lloyd-Jones DM. Predictive utility of pulse pressure and other blood pressure measures for cardiovascular outcomes. Hypertension. 2007; 49: 1256–1264.