OBJECTIVES: To determine safety-specific, efficacy-specific and genotypic-specific dose requirements of efavirenz (EFV) in children aged 3 to less than 36 months with HIV infection. DESIGN: IMPAACT P1070 was a 24-week prospective cohort trial of EFV (as open capsules) and two nucleoside reverse transcriptase inhibitors in children with HIV infection 3 to less than 36 months without tuberculosis (Cohort 1). METHODS: CYP2B6 G516T genotype was determined, and intensive pharmacokinetics was performed at week 2. EFV dose was adjusted if outside the target area under the curve (AUC) 35-180 mug*h/ml. Pharmacokinetic and CYP2B6 G516T genotype data were used to model EFV exposures based on Food and Drug Administration (FDA)-approved doses. RESULTS: Forty-seven participants, median age 19 months, initiated the study regimen with 24 weeks median follow-up; 38 516GG/GT and 9 516TT genotypes. Initially, median EFV AUC was higher in 516TT vs. 516GG/GT (median 490 vs. 107 mug*h/ml; P = 0.0001) with all 516TT above AUC target. Following an amendment that reduced the 516TT EFV dose by 75%, pharmacokinetic modeling predicted that 83% of participants met the AUC target (31/38 516GG/GT, 8/9 516TT). In contrast, modeling using P1070 data predicted that FDA-approved doses would produce subtherapeutic AUCs in almost one-third of participants with 516GG/GT and excessive AUCs in more than 50% with 516TT genotypes. CONCLUSION: CYP2B6 G516T genotype strongly influences EFV exposures in this age group. Genotype-directed dosing yields therapeutic EFV concentrations and appears to outperform other dosing approaches.