Ambulatory blood pressure variability in prediction of target organ injury: the SHIP AHOY study

Abstract

Background: Ambulatory hypertension is associated with elevated left ventricular mass index (LVMI), cardiac dysfunction, and increased arterial stiffness in adolescents. Whether the addition of measures of BP variability improves the prediction of subclinical cardiovascular target organ damage (TOD) over mean BP measures is not known.

Methods: We assessed clinic and ambulatory BP (ABP), anthropometrics, and TOD in 397 adolescents. ABP means standard deviation (SD), BP, and heart rate (HR) dipping were calculated; coefficients of variability (CV) were calculated (SD/mean) to assess ABP variabilities. Measures of TOD included LVMI, left ventricular hypertrophy (LVH), LV systolic shortening, LV diastolic function (e'/a'), and pulse wave velocity. General linear models were used to determine if increased ABP variability measures were significant determinants of TOD in models containing mean ABP percentiles, age, sex, race/ethnicity, BMI z-score, and HR.

Results: Mean participant age was 15.6 ± 1.7 years (63% white, 59% male) with mean casual BP 122.6/71.6 mmHg ±12.4/10.5, and mean awake systolic ABP 124.2/72.0 ± 11.3/7.7 mmHg. In linear models, increased awake CV-DBP and HR dipping were significant determinants of LVMI. CV-HR was an independent determinant of diastolic (e'/a') but not systolic dysfunction. Using logistic regression, the combination of awake and asleep diastolic ABP variability and awake systolic ABP percentile improved the prediction of LVH.

Conclusions: Consideration of ABP variability in addition to ABP percentile may aid in identifying adolescents at risk for LVH.

Keywords: Blood pressure; Blood pressure variability; Diastolic dysfunction; Heart rate variability; Left ventricular hypertrophy.

Fig. 1

ROC curves for prediction of LVH (LVM > 38.6 g/m2.7)1 and diastolic dysfunction (e’/a ’< 10th% for normotensive participants) using same variables as in linear regression. Full model for LVH included the sex, BMI z-score, mean awake ambulatory SBP%, mean ambulatory awake CV-DBP and asleep CV-BP and HR dipping. BP model only included sex, BMI z-score and mean awake ambulatory SBP%. BP variability model included sex, BMI z-score, both CV values and HR dipping. P value for difference between full model and BP model = 0.002; between BP and BP variability model = 0.008. Full model for e’/a ’included BMI z-score, mean awake ambulatory DBP%, and asleep CV-HR. There was no significant difference between C-statistics for full model and BP or between BP and HR variability models. The figures compare BP and BPV (left panel) and BP and HRV (right panel) only and are not adjusted for covariates.