Reliability model from the in vitro durability tests of a left ventricular assist system

A reliability test of the Novacor N100PC left ventricular assist system (LVAS) with valved conduits, including a pump/drive unit with compact controller and LVAS monitor was performed. The initial test objective was to demonstrate sufficient reliability for clinical use as a long-term circulatory support system. The subsequent objective, a test to failure, was intended to provide an assessment of the durability of the design and to determine the LVAS wearout modes. Testing began in April 1993 and was performed with 12 systems on gravity-feed mock circulatory loops. The pump/drive units were submersed in body temperature saline for the duration of the test. Each of the LVAS units was operated at nominal afterloads of 75, 90, and 105 mm Hg, with test conditions varied to yield nominal pump outputs of 5.6, 7.1, and 8.3 L/min. Failure was defined as the inability of the LVAS to maintain an average pump output of 4 L/min or an average output pressure of 60 mm Hg. after 3 years, all systems remained on test, with durations of 2.3 to 3.0 years. Analysis of the testing to that date, using a constant hazard rate model, indicated a minimum demonstrated reliability of 94% at a 60% confidence level, or 86% at a 90% confidence level, for a 2 year mission time. This greatly surpasses the reliability level included in the STS-ASAIO Long-Term Mechanical Circulatory Support System Reliability Recommendation (80% reliability, 60% confidence level for al year mission time). In the subsequent test-to-failure protocol, all systems ran failure-free for at least 3 years. System failures occurring at longer durations were caused by a single common cause: wear of the energy converter's armature support bearings and shafts. The wearout mode was gradual and could be diagnosed noninvasively before failure. An analysis using a Weibull model was performed, using the test durations of those devices that failed, those that were electively removed from test for analysis of the wear mode, and those that continued on test. As of April 1998, the test results showed a reliability, at a 60% confidence level, of >99.9% for a 1 year mission time, 99.5% for a 2 year mission, and 92.0% for a 3 year mission (>99.9%, 98.3%, and 85.9% for equivalent mission times, at a 90% confidence level). Systems continue on test after as long as 4.9 years.