Our objective was to determine the effects of fish oil on renal function, symptoms, and serum lipids in patients with lupus nephritis. A double-blind, randomized crossover trial of fish oil versus placebo (olive oil) was done on 26 patients with confirmed systemic lupus; 21 completed the study. Intervention was fish oil or placebo, 15 g/day, for one year followed by a 10 week wash-out period, and then the reverse treatment for one year. At baseline and six month intervals, we measured platelet membrane fatty acids, indices of renal function, a disease activity index, serum lipid levels, blood pressure, serum viscosity and red cell flexibility. We found that platelet membrane phospholipids were uniformly affected by fish oil supplementation (P < 0.001) but with significant carry-over effects despite a 10 week wash-out period. Glomerular filtration rate and serum creatinine were not affected. A non-significant reduction in mean (SE) 24-hour proteinuria occurred, from 1424.1 mg (442.7) on placebo to 896.7 mg (352.2) on fish oil (P = 0.21). Fish oil lowered serum triglycerides from 1.89 (0.25) mmol/liter to 1.02 (0.11) mmol/liter (P = 0.004). VLDL cholesterol decreased markedly whether patients initially received fish oil or placebo (P = 0.004). The size of the reduction was affected by the order of treatment (P = 0.03), but parallel comparisons were significant before the crossover (P = 0.0006). With the possible exception of bleeding time, no other treatment effects were shown with fish oil. However, treatment order effects were seen in urinary IgG excretion (P = 0.03), whole blood viscosity (P < 0.0001), red cell flexibility (P = 0.004), and bleeding time (P = 0.06). In conclusion, one year of dietary supplementation with fish oil in patients with stable lupus nephritis did not improve renal function or reduce disease activity, but did alter some lipid parameters. Hitherto unreported carry-over effects and treatment order effects caused by the olive oil created a risk of type II error, and bear methodologic consideration in the design of future studies.