We studied the retention of 111Indium-labeled canine endothelial cells on 32 grafts (16 dogs). Canine endothelial cells were harvested from the external jugular veins, grown in culture, and labeled with 111Indium oxine; 10(6) factor VIII positive cells were inoculated on fibronectin-coated, 4 mmID Hytrel grafts and cultured 18 hours to reach confluence. An autologous seeded graft was interposed in each of the common carotid arteries and exposed to flow for six hours. 111Indium label was measured pre- and postperfusion and corrected for decay. Twenty-five grafts from 13 dogs were available for study. Scanning electron microscopic planimetry was used to determine percent surface coverage by six mutually exclusive surface characteristics: endothelial cells, bare graft, white blood cells on graft, white blood cells on endothelium, white blood cells under endothelium, and thrombus. 111Indium retention was compared with percent coverage by scanning electron microscopy using regression analysis. 111Indium labeling projected an erroneous retention of 41% at zero percent coverage (r = 0.67; p less than 0.01). Multiple regression analysis revealed an equivalent distribution of 111Indium label over nonendothelial portions of the flow surface and indicated a leak rate into the circulation of 25.6% of the initial 111Indium label over six hours. We conclude that: 1) 111Indium labeling data usually overestimates endothelial cell retention; 2) an average of 4.67%/hour is lost into the general circulation; 3) 111Indium label can be found equally on surfaces of thrombus, white blood cells, and hydrophilic Hytrel graft; and 4) 111Indium labeling is not a reliable method for in vivo studies of endothelial cell retention.