Statistical aspects of the design and analysis of studies to compare haemodialysis membranes

Abstract

Some aspects of the statistical design and analysis of short-term studies of haemodialysis membranes are reviewed. The potential benefits of crossover studies comparing more than two treatments are emphasized. Designs based on Latin squares are advocated and the importance of avoiding carryover effects by ensuring adequate time between successive treatments is stressed. Studies of this type will usually record data serially within each period, so appropriate graphical techniques and the use of summary measures for the analysis of this type of data are considered. An example of the analysis of a study, using MINITAB, is included. Some issues of sample size estimation are discussed, and finally some more advanced designs and methods of analysis are briefly considered.