We evaluated a new trileaflet prosthesis and a control bileaflet prosthesis in the mitral and aortic positions in 27 calves. The prototype trileaflet valve (TV1) functioned satisfactorily in the mitral position (TV1m, n = 7) but later yielded thrombogenic complications in the aortic position (TV1a, n = 4). The valve was redesigned (TV2) and retested in the mitral (TV2m n = 4) and aortic (TV2a, n = 5) positions, along with control valves (Cm, n = 4; Ca, n = 3). At necropsy, the valves were graded on a scale of 0 (no visible thrombi) to 4 (thrombi greater than 5 mm and/or obstructed leaflets). The TV1m, TV2m, and Cm animals, respectively, had implant durations of 215+/-112, 140+/-63, and 159+/-89 days and thrombus grades of 0.71+/-0.76, 0.33+/-0.58, and 1.50+/-0.58. The TV1a, TV2a, and Ca animals had implant durations of 18+/-12, 159+/-61, and 108+/-62 days and thrombus grades of 2.75+/-1.00, 0.50+/-0.58, and 0.67+/-0.58 (p < .005; TV2a vs. TV1a). Three TV1a calves died early of valve related complications. A design irregularity, undetected in the mitral position but revealed in the aortic position, caused a high early mortality in the TV1a animals. Redesigning the prosthesis eliminated valve related mortality and significantly reduced the thrombus grade. Because satisfactory performance in the mitral position does not guarantee safety or efficacy in the aortic position, site specific preclinical testing is crucial for mechanical heart valves.