To test for effects on systolic and diastolic blood pressure and to provide precise estimates of their magnitude, we conducted an overview of randomized clinical trials that aimed to reduce the intake of sodium in human subjects. We excluded from pooled analyses trials with confounded designs, those that compared intake levels beyond the usual range in the population, and those without published reports. Two reviewers abstracted information in duplicate and differences were reconciled. Twenty-three trials with outcome data from an aggregate of 1,536 subjects were included. Data were pooled both separately for hypertensive and nonnotensive subjects and for all trials combined. With the use of sample size weighting, blood pressure reductions (net of controls) were 4.9±13/2.6±0.8 mm Hg (systolic and diastolic, respectively, with 95% confidence limits) in hypertensive subjects and 1.7 ±1.0/1.0 ±0.7 mm Hg in nonnotensive subjects. The combined blood pressure reductions were 2.9±0.8/1.6±0.5 mm Hg. These changes were associated with mean reduction of urinary sodium excretion ranging from 16 to 171 mmol/24 hr for individual trials. A dose-response relation across trials was found, both in nonnotensive and in hypertensive subjects. These results indicate that sodium reduction lowers mean blood pressure in both hypertensive and nonnotensive individuals for periods of at least several months. The findings are highly consistent with results of observational epidemiological studies and have implications for preventive strategies of blood pressure control. {Hypertension 1991;17[suppl I]:I-27-I-33) R andomized intervention trials, properly conducted, provide the best test of an effect of prophylactic or therapeutic treatments. 1 Although the blood pressure effects of reducing dietary sodium have been studied since early in the century, 2 randomized trials have been reported only within the past 17 years, 3 largely the past 10 years. Trials have both increased in frequency and become more varied and complex in design. These changes reflect the importance of the question for public health and biology but also pose a challenge for summarizing and interpreting results. Several reviews have been published in recent years, 4 -5 but they did not present pooled analyses of all relevant trials to the time of the reviews, and several trials have since been completed. We favor pooling as part of a review because it formalizes the objective of including all relevant trials and an explicit method of combining their results. It also furnishes a quantitative estimate of the effect, which