OBJECTIVE: To investigate epidemiological aspects of blood pressure (BP) levels, and the prevalence of arterial hypertension in a representative sample of the whole elementary and high school population of the city of Belo Horizonte (state of Minas Gerais, southeastern Brazil). To describe the blood pressure levels found and the prevalence of BP levels above the 90th and 95th percentiles in the first and second measurements. To evaluate prevalence using four different normality references separately to assess the impact of choice of reference. To study the frequency at which the studentś blood pressure has been evaluated.METHODS: In an observational cross-sectional epidemiological study, we studied a sample of 1005 students, aged 6 to 18 year, randomized from an universe of 486,166 students from the 521 public and private elementary and high schools in the city. Each student was evaluated by means of a personal interview, anthropometric measurements, and two blood pressure measurements in a single visit and at a 10-minute interval.RESULTS: The systolic and diastolic blood pressure levels, by age and gender, were slightly lower than those described by the Task Force Report (1987) and by Rosner et al. (1993). The prevalence of BP levels (systolic or diastolic) over the 95th percentile were, according to the reference used: 8.7% - Task Force Report (1987); 5.5% - Rosner et al. (1993); 6.5% - Update on the Task Force Report (1996); and 9.8% - the 95th percentile from our own study. In the second measurement these prevalence rates dropped to 4.9%; 2.4%; 3.5% and 5.8%, respectively.CONCLUSIONS: We present a detailed description (mean values and standard deviations, and the 90th, 95th and 99th percentiles by age and gender) of the BP levels found in both measurements. Mean values and upper percentiles found in this study are very close to those found in the American references, which validates the current use of these references in Brazil. The prevalence of 9.0% of BP levels above the 90th percentile in two measurements made at random hints at the number of children who should be monitored for hypertension, and reinforces the importance of routine evaluation of blood pressure at every pediatric examination.