Improvement of sex education in schools is a key part of the UK government's strategy to reduce teenage pregnancy in England. We examined the effectiveness of one form of peer-led sex education in a school-based randomised trial of over 8000 pupils. 29 schools were randomised to either peer-led sex education (intervention) or to continue their usual teacher-led sex education (control). In intervention schools, peer educators aged 16-17 years delivered three sessions of sex education to 13-14 year-old pupils from the same schools. Primary outcome was unprotected (without condom) first heterosexual intercourse by age 16 years. Analysis was by intention to treat. By age 16 years, significantly fewer girls reported intercourse in the peer-led arm than in the control arm, but proportions were similar for boys. The proportions of pupils reporting unprotected first sex did not differ for girls (8.4% intervention vs 8.3% control) or for boys (6.2% vs 4.7%). Stratified estimates of the difference between arms were -0.4% (95% CI -3.7% to 2.8%, p=0.79) for girls and -1.4% (-4.4% to 1.6%, p=0.36) for boys. At follow-up (mean age 16.0 years [SD 0.32]), girls in the intervention arm reported fewer unintended pregnancies, although the difference was borderline (2.3% vs 3.3%, p=0.07). Girls and boys were more satisfied with peer-led than teacher-led sex education, but 57% of girls and 32% of boys wanted sex education in single-sex groups. Peer-led sex education was effective in some ways, but broader strategies are needed to improve young people's sexual health. The role of single-sex sessions should be investigated further.