Intraperson variability in both blood pressure (BP) and sodium excretion dilutes associations and leads to underestimates of the dose-response relation. The authors applied statistical correction techniques to data from the Trials of Hypertension Prevention (TOHP), Phase I, carried out 1987-1990. Men and women with high normal diastolic BP (80-89 mmHg) were randomized to sodium reduction (n = 327) or a usual care comparison group (n = 417). Regression estimates of the effects of change in sodium and sodium/potassium ratio (Na/K ratio) on blood pressure change in the pooled sample were corrected for both the within-person variance of the excretion measures and the within-person covariance with blood pressure using a multivariate error correction. The estimated cross-sectional reliability was 0.36 for square root(Na) and 0.42 for square root(Na/K ratio) and that for change was 0.31 and 0.28, respectively. Corrected coefficients suggested a decrease of 4.4 mmHg in systolic BP (95% confidence interval (CI) 0.1-8.8) and 2.8 mmHg in diastolic BP (95% CI -0.2 to 5.8) per 100 mmol/24 hour reduction in sodium, and of 3.4 mmHg in systolic BP (95% CI 0.8-6.1) and 1.7 mmHg in diastolic BP (95% CI 0.0-3.5) per unit decrease in Na/K. These results are comparable with those from the Intersalt Study, and suggest that the true effect of sodium change on blood pressure change in normotensives over 18 months is underestimated by more than half in uncorrected data.