Statistical analysis in method comparison studies – Part one

Method comparison is commonly performed by laboratory specialists to assess the comparability of two methods. The quality of method comparison study determines the quality of the results and validity of the conclusions. The key to the successful method comparison is therefore a well-designed and carefully planned experiment. The question to be answered by the method comparison is whether two methods could be used interchangeably without affecting patient results and patient outcome. In other words, by comparing two methods we are looking for a potential bias between methods. If bias is larger than acceptable, methods are different and cannot be used interchangeably. It is important to understand why bias cannot be adequately assessed by correlation analysis and by performing t-test. It is also important to be aware of the importance of graphical presentation of the data (scatter plots and difference plots), as a first step in data analysis. Graphical presentation of the data will ensure that outliers and extreme values are detected. This article will provide some key concepts related to the design of the method comparison study, data analysis and graphical presentation. Passing-Bablok and Deming regression are going to be covered in the subsequent article (part two). Introduction One of the important aspects of the method verification is the assessment of method trueness. Method trueness can be assessed either by following the CLSI EP15-A2 standard, which defines procedure of the verification of performance for precision and trueness, or the CLSI EP09-A3 standard, which provides guidance on how to estimate the bias by comparison of measurement procedures using patient samples [1,2]. The CLSI EP09-A3 standard also defines several statistical procedures which can be used to describe and analyze the data. The choice of correct statistical procedures for data analysis and knowledge about how to interpret the results of statistical analysis is of key importance for proper assessment of the method trueness. This article provides the insight into the proper design of the method comparison study and some basic considerations about initial steps in data analysis and graphical presentation (scatter and difference plots). The following article will address statistical methods used in method comparison studies (Passing-Bablok and Deming regression). Article downloaded from acutecaretesting.org Ana-Maria Simundic: Statistical analysis in method comparison studies – Part one Page 1