BACKGROUND: Lactoferrin (LF) is a broad-spectrum antimicrobial and immunomodulatory milk glycoprotein. OBJECTIVE: To determine the effect of bovine LF on the prevention of the first episode of late-onset sepsis in Peruvian infants. METHODS: We conducted a pilot randomized placebo-controlled double blind study in infants with a birth weight (BW) less than 2500g in 3 Neonatal Units in Lima. Patients were randomized to receive bovine LF 200mg/kg/d or placebo for 4 weeks. RESULTS: One hundred and ninety neonates with a BW of 1591 +/- 408 g and a gestational age of 32.1 +/- 2.6 weeks were enrolled. Overall, 33 clinically defined first late-onset sepsis events occurred. The cumulative sepsis incidence in the LF group was 12/95 (12.6%) versus 21/95 (22.1%) in the placebo group, and 20% (8/40) versus 37.5% (15/40) for infants less than or equal to 1500 g. The hazard ratio of LF, after adjustment by BW, was 0.507 (95% CI: 0.249-1.034). There were 4 episodes of culture-proven sepsis in the LF group versus 4 in the placebo group. Considering that children did not received the intervention until the start of oral or tube feeding, we ran a secondary exploratory analysis using time since the start of the treatment; in this model, LF achieved significance. There were no serious adverse events attributable to the intervention. CONCLUSIONS: Overall sepsis occurred less frequently in the LF group than in the control group. Although the primary outcome did not reach statistical significance, the confidence interval is suggestive of an effect that justifies a larger trial.