To examine the effects of intensive patient and/or physician diabetes education on patient health outcomes, a controlled trial was conducted in which internal medicine residents and their 532 diabetic patients were randomly assigned to: routine care; patient education; physician education; or both patient and physician education. Patient outcome data were analyzed either by analysis of covariance on post intervention values (2-hour post-prandial plasma glucose [PPG]; body weight [BW]; blood pressure [BP]; or analysis of variance conducted on change values (fasting plasma glucose [FPG] and glycosylated hemoglobin [A1Hgb]). After patient education, significant improvements were observed in FPG, A1Hgb, BW, and systolic and diastolic BP. Physician education resulted in significant decreases in FPG, A1Hgb and BW. The combination of patient plus physician education resulted in the greatest improvements in patients' health outcomes including FPG, A1Hgb, PPG, BW and diastolic BP. Adjusted systolic BPs were not significantly different in the two groups. While these physiologic improvements were statistically and probably clinically significant, hyperglycemia and obesity still persisted. Thus, achieving optimal patient outcomes for a chronic disease like diabetes mellitus may require a greater or more effective use of resources than currently estimated.