BACKGROUND: During early postnatal development, the nervous system accretes docosahexaenoic acid (DHA; 22:6n-3), a highly unsaturated n-3 (omega-3) fatty acid (FA) used in the formation of neural cell membranes. DHA, which is present in human breast milk, may also be biosynthesized from n-3 FAs such as 18:3n-3 [alpha-linolenic acid (ALA)] or 20:5n-3 [eicosapentaenoic acid (EPA)]. An important concern is to what extent these precursors can supply DHA to the developing infant. OBJECTIVE: We analyzed measurements of fractional percentages of plasma (2)H(5)-ALA and (13)C-U-EPA directed toward the synthesis of labeled 22:6n-3 in 11 newborn infants by using compartmental modeling procedures. DESIGN: One-week-old infants received doses of (2)H(5)-ALA and (13)C-U-EPA ethyl esters enterally. We drew blood from the infants periodically and analyzed the plasma for endogenous and labeled n-3 FAs. From the time-course concentrations of the labeled FAs, we determined rate constant coefficients, fractional synthetic rates, and plasma turnover rates of n-3 FAs. RESULTS: In infants, approximately 0.04% of the (2)H(5)-ALA dose converted to plasma (2)H(5)-EPA. Plasma (2)H(5)-EPA and (2)H(5)-22:5n-3 [docosapentaenoic acid (DPA)] efficiently converted to (2)H(5)-DPA and (2)H(5)-DHA, respectively. The percentage of plasma (13)C-U-EPA directed toward the synthesis of (13)C-DHA was lower than the percentage of plasma (2)H(5)-EPA that originated from (2)H(5)-ALA. CONCLUSIONS: Endogenously synthesized EPA was efficiently converted to DHA. In comparison, preformed EPA was less efficiently used for DHA biosynthesis, which suggests a differential metabolism of endogenous EPA compared with exogenous EPA. However, on a per mole basis, preformed EPA was 3.6 times more effective toward DHA synthesis than was ALA. Newborns required an intake of approximately 5 mg preformed DHA. kg(-1) x d(-1) to maintain plasma DHA homeostasis.