Improved statistical characterization of prosthetic heart valve hydrodynamics using a performance index and regression analysis.

BACKGROUND AND AIMS OF THE STUDY: The ISO 5840 Standard (Cardiovascular implants - Cardiac valves) currently requires a minimum of three test samples per size for hydrodynamic testing. Typically, the only statistical analysis performed is a descriptive analysis, with the mean (+/- SE) given for each size and cardiac output (CO). The study aim was to develop better statistical methods, incorporating regression analysis of a performance index, equal to the effective orifice area divided by the tissue annulus area. The analysis is performed on the full dataset, with size and CO as independent variables. METHODS: Hydrodynamic data of Ionescu-Shiley pericardial valves from a published study were used to compare the two analysis methods. Three samples each of size 19, 23 and 27 mm valves were tested at COs of 4.2, 5.6, 7.0 and 8.4 l/min. Descriptive statistics were performed for each size and CO. Regression analysis was also performed on the full dataset. Confidence intervals (CI) were calculated for each statistical method and compared. RESULTS: The regression equation that best fitted the data was: Performance Index (PI) = -1.63 + (0.011 x CO) + (0.167 x size) - (0.0036 x size2). All four parameter estimates were significantly different from zero (p <0.02). The SE of the mean was 0.015 for COs of 4.2 or 8.4 l/min, and 0.013 for COs of 5.6 or 7.0 l/min, less than that of nine of 12 of the individual descriptive analysis. CI for the regression analysis were substantially tighter, averaging one-third the width of those of the descriptive statistics. CONCLUSION: The tighter CI resulting from the regression analysis allows a better comparison of the PI to an objective performance criterion. Such methods should be considered for inclusion in the new version of the ISO 5840 standard for prosthetic heart valves.