Between August 1986 and March 1993, 124 patients (102 men; mean age of 59 years) underwent myocardial revascularization with the use of at least one free internal mammary artery (FIMA). This group represents 4.5% of the 2725 coronary bypasses performed during the same period. Seventy-six patients (61%) had suffered from at least one previous myocardial infarction. Forty-five patients (36%) had unstable angina; three-vessel disease was found in 100 cases (80.5%) and a left ventricular ejection fraction lower than 0.4 in 22 (17.7%). There were 18 (14.5%) redo procedures and 90 (72.5%) bilateral internal mammary artery (IMA) grafts. The reasons for using a FIMA were: too short an internal mammary artery pedicle in 83 patients, IMA injury at harvesting in 30 patients and post-bypass ischaemia in areas grafted with pedicled IMA (PIMA) in 11 patients. Cardiopulmonary bypass, moderate hypothermia (30 degrees C) and crystalloid anterograde and retrograde cardioplegia were used in all cases. Sixty-seven FIMA grafts were anastomosed directly to the ascending aorta; 57 were sutured via a saphenous hood using a running suture of polypropylene 7/0 and three were anastomosed end-to-end to a PIMA graft. FIMA grafts were directed to the left anterior descending (34%), the circumflex (37%) and the right coronary artery (29%). In total, 179 anastomoses were constructed using 127 FIMA, 136 using PIMA and 158 using saphenous veins (3.8 anastomoses per patient). Hospital mortality and postoperative myocardial infarction rates were 5.6% (seven patients) and 3.2% (four patients), respectively. Cardiac-related mortality was 3.2% (four patients); three of these four patients had been operated on for evolving infarction and one underwent a redo procedure. Four of the 117 survivors died later on; in two, it was cardiac-related and a result of global heart failure at 9 and 12 months. Of the 113 remaining patients, 106 are symptom free after a mean follow-up of 28.2 (range 3-84) months. Fifty-nine patients (50.4%) were restudied by angiography at a mean interval of 15 months. Patency rates of FIMA anastomosed either directly to the aorta or via a saphenous hood were 82.8 or 89.7%, respectively. Patency rates of FIMA directed to the left anterior descending, the circumflex and the right coronary artery were 85.7, 88 and 83.3%, respectively. Global FIMA patency was 86.4%, while global PIMA patency was 100%. The FIMA mid-term patency rates compare unfavourably with those of PIMA: FIMA should therefore be restricted to the cases where PIMA or other pedicled arterial grafts are unavailable.