We aim to assess if the relationship between short or long sleep duration and hypertension is present among adults from epidemiological evidence and to investigate the relationship quantitatively. We performed a comprehensive search of cross-sectional and longitudinal studies using PubMed and the Cochrane Library through February 2012. Our search was supplemented by reviewing reference lists of original and relevant reviews. After the related data were extracted by two investigators independently, pooled odds ratios (ORs) or relative risks (RRs) were estimated using a random-effects model or a fixed-effects model. Publication bias was evaluated, while sensitivity and meta-regression analyses were performed. Twenty-four adult studies met our inclusion criteria, with ages ranging from 18 to 106 years. Twenty-one studies involving 225,858 subjects were included in the meta-analysis. The pooled results from the cross-sectional studies showed that short sleep duration was associated with a greater risk for hypertension (OR, 1.21; 95% confidence interval [CI], 1.09-1.34; P<0.001), and long sleep duration also increased the risk for hypertension (OR, 1.11; 95% CI, 1.04-1.18; P=0.003). There was no evidence of publication bias. Pooled analysis from the longitudinal studies indicated a significant association between short sleep duration and hypertension (RR, 1.23; 95% CI, 1.06-1.42; P=0.005), but an insignificant relationship between long sleep duration and hypertension (RR, 1.02; 95% CI, 0.91-1.14; P=0.732). The effects of sleep duration differed by gender, location of the population, and definitions of short or long sleep duration. Meta regression analysis including seven variables did not find the sources of heterogeneity. Among adults, a U-shaped relationship between habitual sleep duration and hypertension was found at the cross-sectional level. Short sleep duration was associated with a higher risk for hypertension even longitudinally. We must pay more attention to this lifestyle factor. Copyright © 2012 Elsevier B.V. All rights reserved.