BACKGROUND: Neonatal BCG vaccination is part of routine vaccination schedules in many developing countries; vaccination at school age has not been assessed in trials in low-income and middle-income countries. Catch-up BCG vaccination of school-age children who missed neonatal BCG vaccination could be indicated if it confers protection and is cost-effective. We did a cluster-randomised trial (BCG REVAC) to estimate the effectiveness (efficacy given in routine settings) of school-age vaccination. METHODS: We assessed the effectiveness of BCG vaccination in school-age children (aged 7-14 years) with unknown tuberculin status who did not receive neonatal BCG vaccination (subpopulation of the BCG REVAC cluster-randomised trial), between July, 1997, and June, 2006, in Salvador, Brazil, and between January, 1999, and December, 2007, in Manaus, Brazil. 763 schools were randomly assigned into BCG vaccination group or a not-vaccinated control group. Neither allocation nor intervention was concealed. Incidence of tuberculosis was the primary outcome. Cases were identified via the Brazilian Tuberculosis Control Programme. Study staff were masked to vaccination status when identified cases were linked to the study population. We estimated cost-effectiveness in Salvador by comparison of the cost for vaccination to prevent one case of tuberculosis (censored at 9 years) with the average cost of treating one case of tuberculosis. Analysis of all included children was by intention to treat. For calculation of the incidence rate we used generalised estimating equations and correlated observations over time. FINDINGS: We randomly assigned 20,622 children from 385 schools to the BCG vaccination group and 18,507 children from 365 schools to the control group. The crude incidence of tuberculosis was 54·9 (95% CI 45·3-66·7) per 100,000 person-years in the BCG vaccination group and 72·7 (62·8-86·8) per 100,000 person-years in the control group. The overall vaccine effectiveness of a first BCG vaccination at school age was 25% (3-43%). In Salvador, where vaccine effectiveness was 34% (8-53%), vaccination of 381 children would prevent one case of tuberculosis and was cheaper than treatment. The frequency of adverse events was very low with only one axillary lymphadenitis and one ulcer greater than 1 cm in 11,980 BCG vaccinations. INTERPRETATION: Vaccination of school-age children without previous tuberculin testing can reduce the incidence of tuberculosis and could reduce the costs of tuberculosis control. Restriction of BCG vaccination to the first year of life is not in the best interests of the public nor of programmes for tuberculosis control. FUNDING: UK Department for International Development, National Health Foundation.