Comparison of prosthetic valve hydrodynamic function: objective testing using statistical multilevel modeling.

Abstract

BACKGROUND AND AIM OF THE STUDY: The performance of novel prosthetic heart valves is assessed using in-vitro hydrodynamic function tests. The study aim was to examine the problem of objective discrimination of hydrodynamic performance to determine significant differences between valve designs, and illustrate proposed methodology using data collected from five different polyurethane tri-leaflet valve designs. METHODS: Two engineering designs were manufactured with leaflets of the same polyurethane (GE, LE); design L was manufactured using three further leaflet materials of differing material modulus (LL, L4, L5). Six valves were made in each design, each tested at five flow rates in a standard hydrodynamic test rig, with five test replications for each valve. The data were analyzed using multilevel statistical modeling methods, allowing simultaneous comparison of multiple regression lines describing valve performance. The multilevel model is hierarchical in structure, in this case with two levels of data, describing individual valves at level 2 and test replicates at level 1. In all cases, the multilevel model uses the hydrodynamic function measure of interest, e.g. mean pressure gradient or leakage, with logarithmic transformation as required as the dependent variable, Y. The independent variable, X, is, in all cases, the natural logarithm of the RMS flow measured through the valve. RESULTS: The two-design multilevel model enabled quantitative discrimination of designs GE and LE, showing that design GE had significantly better hydrodynamic function overall than design LE in this case (mean pressure gradient was estimated as 0.93 mmHg lower at low cardiac output, 14.74 mmHg lower at 9.6 l/min). The five-design multilevel model showed clearly the relatively poor hydrodynamic performance of designs L4 and L5 compared with others. The procedure was straightforward, and produced a statistical comparison among valve designs that is not easily achieved by other means. CONCLUSION: This methodology provides a useful means of objective assessment of valve function for valve developers. Variance estimates provided by the analysis also provide a basis for quality control of valve production and testing.