Statistical analysis of heart valve outcomes

Statistics is a rapidly growing field utilizing thousands of data analysis functions, though a relatively few are used in the majority of cardiac surgical studies. Dr. Wechsler recently published a list of the statistical terms which are important for cardiothoracic surgeons to be familiar with. These can be grouped into 3 categories: General concepts; Summary and test statistics; and Special techniques. The last category includes three techniques which are widely used to describe the long-term results of heart valve series: Hazard functions, Kaplan-Meier event-free curves, and "Actual" event-free curves. Some conclusions from a review of these techniques are: (1) an event with a non-constant hazard should not be described using "Linearized" rates (the Weibull distribution adds a parameter to accommodate changing hazard); (2) for non-fatal events, "Actual" analysis provides the true event probability, which the Kaplan-Meier overestimates. These methods are used in the analysis of published heart valve series to conclude that: (1) the wide scatter in complications rates with the same valve models implicates factors other than the valve; (2) the earliest porcine series had low mean ages and low durability compared with later series.