Discrepancies in Observed and Predicted Longitudinal Change in Central Hemodynamic Measures: The Framingham Heart Study

Community-based studies have evaluated cross-sectional age relations of aortic stiffness measures, which are not often recapitulated in longitudinal studies. We examined baseline and longitudinal change in aortic stiffness in 5491 participants (mean age, 49.5±14.5 years; 54% women) who attended 2 sequential examinations (6.0±0.6 years apart) of the Framingham Heart Study. Cross-sectional relations of central hemodynamics (mean arterial pressure, central pulse pressure, carotid-femoral pulse wave velocity, and characteristic impedance) with age and risk factors were assessed at visits 1 and 2 (models 1 and 2). We used model 1 coefficients (M1), visit 1 risk factor levels (R1), and age at each visit (A1, A2) to estimate values at visits 1 (M1R1A1) and 2 (M1R1A2). While using model 1 coefficients, we accounted for age and risk factor level (R2) changes to predict values at visit 2 (M1R2A2). Using model 2 coefficients (M2) and visit 2 age and risk factor levels, we predicted visit 2 values (M2R2A2). We calculated predicted change 3 ways: delta1=M1R1A2-M1R1A1, delta2=M1R2A2-M1R1A1, and delta3=M2R2A2-M1R1A1. Delta1 values were biased and correlated poorly with actual changes (r=-0.02-0.14). For mean arterial pressure, delta1=1.9±0.8 mm Hg (r=0.14), observed change=-3.3±10.3 mm Hg, and discrepancy=5.2±10.2 mm Hg (P<0.0001). For characteristic impedance, delta1=7.2±14.7 dyne×sec/cm5 (r=0.07), observed change=20.5±68.2 dyne×sec/cm5, and discrepancy=-13.3±68.7 dyne×sec/cm5 (P<0.0001). Delta2 values were moderately correlated with change (r=0.17-0.54) but remained biased whereas delta3 values were moderately correlated with change with no bias. Projected change in hemodynamic measures extrapolated from cross-sectional age relations may differ substantially from actual change, particularly for variables with nonlinear age relations.