Background HIV prevalence among 15–24 year-olds in South Africa is amongst the highest in the world. A cluster-randomised trial began in 2012, assessing the effectiveness of a sport-based HIV prevention intervention in reducing age-disparate partnerships, multiple partnerships, and perpetration of gender-based violence.

Methods Forty-six schools in Cape Town and Port Elizabeth were randomised to either receive sessions led by trained Grassroot Soccer Coaches or standard Life-Orientation classes led by teachers. At baseline, 4485 grade-nine students (median age 15) used touchscreen mobile phones to complete a self-administered questionnaire assessing demographic/socio-economic, psychosocial, lifestyle, behavioural and relationship factors. Answers were confidential, with no personal identifying information captured. Chi-squares were used to assess crude differences by group, followed by multivariate linear and logistic regressions adjusting for age and school-level clustering.

Results The control group was slightly older than the intervention group (mean age 15.8 vs 15.5 years, p = 0.071). At baseline, more control than intervention participants reported not living with both parents (67.8% vs 64.1%, p = 0.009), ever having had sex (55.9% vs 52.6%, p = 0.032), having an age-disparate partner (5+ years difference) in the last year (14.5% vs 12.1%, p = 0.097), depressive symptoms (28.7% vs 25.7%, p = 0.024), and HIV stigma (27.4% vs 23.8%,p = 0.007). More males in the control group reported ever having perpetrated intimate-partner violence (37.6% vs 33.4%, p = 0.051) and rape (29.1% vs 23.7%, p = 0.029). These differences between trial arms lost statistical significance after adjusting for age and clustering. No differences between groups were observed for sex, socio-economic status, HIV knowledge, self-efficacy, or number of sexual partners.

Conclusions There was a small age imbalance between study groups, which appears to explain other observed baseline imbalances. Outcome analyses for the main trial should be adjusted for age. High levels of reported intimate-partner violence and rape perpetration were observed, suggesting this is a crucial area for intervention.