DETERMINING THE TROUGH-TO-PEAK RATIO IN PARALLEL-GROUP TRIALS

We explored how in parallel-group trials interindividual variability, correction for placebo effects, and smoothing of blood pressure profil es can be handled in measuring the trough-to-peak ratio in 244 individ uals with isolated systolic hypertension (greater than or equal to 60 years) enrolled in the placebo-controlled Systolic Hypertension in Eur ope Trial. Net treatment effects were computed by subtracting the mean changes from baseline during placebo (n=133) from those during active treatment (n=111). At entry, systolic/diastolic pressures averaged 17 6/86 mm Hg in the clinic and 149/80 mm Hg on 24-hour ambulatory monito ring. With corrections applied for baseline and placebo, nitrendipine (10 to 40 mg/d), with the possible addition of enalapril (5 to 20 mg/d ) and/or hydrochlorothiazide (12.5 to 25 mg/d), reduced (P<.001) these blood pressure values by 16.6/7.3 and 9.8/4.7 mm Hg, respectively. Th e net trough-to-peak ratios were first determined from blood pressure profiles (12 hours) with I-hour precision, synchronized by the morning and evening doses of the double-blind medication. According to the us ual approach, disregarding interindividual variability, the systolic/d iastolic net trough-to-peak ratios were 0.46/0.40 in the morning and 0 .77/0.99 in the evening. In individual subjects, the baseline-adjusted trough-to-peak ratios were nonnormally distributed. We therefore used a nonparametric technique to calculate the net trough-to-peak ratios from the results in individual subjects. In the morning, these ratios averaged 0.25 systolic (95% confidence interval, 0.09 to 0.41) and 0.1 5 diastolic (95% confidence interval, 0.00 to 0.31) and in the evening , 0.19 and 0.36 (95% confidence intervals, 0.00 to 0.38 and 0.14 to 0. 56), respectively. When the blood pressure profiles were smoothed by s ubstituting the 1-hour averages by moving or fixed 2-hour averages or by Fourier modeling, the trough-to-peak ratios remained unchanged afte r-the morning dose (0.20/0.13, 0.20/0.14, and 0.16/0.21, respectively) but tended to increase in the evening (0.32/0.38, 0.28/0.40, and 0.48 /0.49). In conclusion, the parallel-group analysis proposed makes it p ossible for one to correct the trough-to-peak ratio for baseline as we ll as placebo, to account for interindividual variability, and to calc ulate a confidence interval for the net trough-to-peak ratio. Accounti ng for interindividual variability reduces the trough-to-peak ratio. S moothing affects the individualized net trough-to-peak ratios in an un predictable way and should therefore be avoided.