A statistical approach to the quantitative comparison of pulsatile flow in vitro data of prosthetic heart valve...

In vitro evaluation and animal testing are fundamental steps in the assessment of prosthetic heart valves before their use in clinical practice. Valve testing under pulsatile conditions is the best in vitro simulation of cardiac valve function, and the ensuing results should support the surgeon's decision concerning the time of valve replacement. However, limits in hydraulic reproduction of cardiac function and differences in protocol implementation of in vitro testing lead to difficulties in obtaining reliable and comparable results. Debate among researchers and standardizing bodies about discrepancies in results becomes critical in light of the European CE certification of implantable medical devices. An interlaboratory environment has been created at the Biomedical Engineering unit of the Istituto Superiore di Sanita in Rome, which uses differing test apparatus to evaluate prosthetic heart valves in vitro, in order to define significant measurement parameters and procedures that produce comparative data.

Two prosthetic valves-a tilting disc and a bileaflet valve-each sized 29 mm, were tested in the aortic position on two different pulse duplicators (PDs), namely the Dynatek MP1 and a system developed at the University of Sheffield. The common protocol adopted was the FDA Interlaboratory Comparison Testing Protocol. Original software was used to manage all test phases, thus minimizing operator-dependent variability in both systems and imposing strict control of experimental conditions. Statistical analysis performed on the data followed two approaches: (i) separate fitting of the two regression equations of pressure drop-flow rate relationship obtained for each valve on both PDs, and (ii) application of a multiple regression model, to fit a single regression equation of pressure drop-flow rate relationship, using data obtained from both PDs for each valve. In addition, an additional independent (dummy) variable was introduced.

Using this approach, the valve parameter range was obtained and, by imposing strict control of the experimental set-up, the coincidence of the two valve power laws, estimated by each of the two PDs, was studied.