BACKGROUND: Interventions are needed to reduce poor perinatal health. We trained community health workers (CHWs) as home visitors to address maternal/infant risks. METHODS: In a cluster randomised controlled trial in Cape Town townships, neighbourhoods were randomised within matched pairs to 1) the control, healthcare at clinics (n = 12 neighbourhoods; n = 594 women), or 2) a home visiting intervention by CBW trained in cognitive-behavioural strategies to address health risks (by the Philani Maternal, Child Health and Nutrition Programme), in addition to clinic care (n = 12 neighbourhoods; n = 644 women). Participants were assessed during pregnancy (2% refusal) and 92% were reassessed at two weeks post-birth, 88% at six months and 84% at 18 months later. We analysed 32 measures of maternal/infant well-being over the 18 month follow-up period using longitudinal random effects regressions. A binomial test for correlated outcomes evaluated overall effectiveness over time. The 18 month post-birth assessment outcomes also were examined alone and as a function of the number of home visits received. RESULTS: Benefits were found on 7 of 32 measures of outcomes, resulting in significant overall benefits for the intervention compared to the control when using the binomial test (p = 0.008); nevertheless, no effects were observed when only the 18 month outcomes were analyzed. Benefits on individual outcomes were related to the number of home visits received. Among women living with HIV, intervention mothers were more likely to implement the PMTCT regimens, use condoms during all sexual episodes (OR = 1.25; p = 0.014), have infants with healthy weight-for-age measurements (OR = 1.42; p = 0.045), height-for-age measurements (OR = 1.13, p<0.001), breastfeed exclusively for six months (OR = 3.59; p<0.001), and breastfeed longer (OR = 3.08; p<0.001). Number of visits was positively associated with infant birth weight ≥2500 grams (OR = 1.07; p = 0.012), healthy head-circumference-for-age measurements at 6 months (OR = 1.09, p = 0.017), and improved cognitive development at 18 months (OR = 1.02, p = 0.048). CONCLUSIONS: Home visits to neighbourhood mothers by CHWs may be a feasible strategy for enhancing maternal/child outcomes. However, visits likely must extend over several years for persistent benefits. TRIAL REGISTRATION: ClinicalTrials.gov NCT00996528.