We report a randomized trial of a revised Behavioral Family Systems Therapy for Diabetes (BFST-D) intervention. Families of 104 adolescents with diabetes were randomized to standard care (SC) or to 6 months of an educational support group (ES) or BFST-D. Family communication and problem-solving skills were assessed at 0, 6, 12, and 18 months by independent rating of videotaped family problem-solving discussions. BFST-D improved individual communication of adolescents and mothers, but not fathers. BFST-D significantly improved quality of family interaction compared to SC (10 of 12 comparisons) and ES (6 of 12 comparisons). Changes in family communication were differentially associated with changes in glycemic control, adherence, and family conflict. BFST-D improved family communication and problem solving relative to SC and modestly relative to ES.