To assess the evolution of triglyceride (TG) levels in HIV-infected patients receiving stable potent antiretroviral therapy treated with N-3 polyunsaturated fatty acids (PUFAs), a prospective double-blind randomized design for a reliable assessment of TG evolution was performed. One hundred twenty-two patients with TG levels >2 g/L and < or =10 g/L after a 4-week diet (baseline TG: 4.5 +/- 1.9 g/L) were randomized for 8 weeks to N-3 PUFAs (2 capsules containing 1 g of fish oil 3 times daily, n = 60), or placebo (1 g of paraffin oil capsules, n = 62). An 8-week open-label phase of N-3 PUFAs followed. Evaluation criteria were TG percent change at week 8, percentage of responders (normalization or > or =20% TG decrease), and safety issues. Ten patients with baseline TG levels >10 g/L were not randomized and received N-3 PUFAs as open treatment. The difference (PUFA - placebo) in TG percent change at week 8 was -24.6% (range: -40.9% to -8.4%; P = 0.0033), the median was -25.5% in the PUFA group versus 1% in the placebo group, and mean TG levels at week 8 were 3.4 +/- 1.8 g/L and 4.8 +/- 3.1 g/L, respectively. TG levels were normalized in 22.4% (PUFA) versus 6.5% (placebo) of patients (P = 0.013) with a > or =20% reduction in 58.6% (PUFA) versus 33.9% (placebo) of patients (P = 0.007). Under the open-label phase of N-3 PUFAs, the decrease in TG levels was sustained at week 16 for patients in the PUFA group (mean TG: 3.4 +/- 1.7 g/L), whereas a 21.2% decrease in TG levels occurred for patients in the placebo group (mean TG: 3.3 +/- 1.4 g/L). No significant differences were observed between groups in the occurrence of adverse events. The median TG change at week 8 was -43.6% (range: Q1-Q3; 95% CI: -66.5% to -4.6%) for patients with baseline TG levels >10 g/L. The difference in mean total cholesterol between groups (PUFA - placebo) at week 8 was -8.5% (P = 0.0117). This study demonstrated the efficacy of PUFAs to lower elevated TG levels in treated HIV-infected hypertriglyceridemic patients. N-3 PUFAs have a good safety profile.