Some statistical concepts for assays: Medical education [Deneye dayali çalişmalarda bazi i̇statistiksel kavramlar]

Abstract

Laboratory studies are major elements in medicine for diagnosis, disease management and follow-up of the treatment response. For a good medical approach, laboratory studies should be correct, precise and sensitive as possible. On the other hand, every laboratory should take care of consistency of their results. When a laboratory director tries to decide for the most appropriate method for his/her laboratory, such as a scanning device or a kit, he/she should have the knowledge about the principles used for the method assessment. Selection of a new method has the following steps to evaluate the characteristics of the method; application, methodological and performance characteristics, market research and decision for an appropriate method through the proposed methods, respectively. The main factors that affect the new method selection are the total cost, the practical features related with application, the analytical performance and the medical benefits. Following the decision of a new method which will be used in the laboratory, the measurable characteristics of the method are determined by some experiments. In this article, some statistical concepts that are frequently used for in vitro laboratory studies are examined. In this frame, methodological characteristics such as assay specificity and assay sensitivity, predictive value, performance characteristics such as accuracy, inter-intra assay precision, recovery, linearity, inter-intra rater reliability, and finally statistical methods such as correlation coefficient, linear regression, chi-square (Mc-Nemar test), Stuart-Maxwell test, Wilcoxon test, t-test, Mann-Whitney U test, F test, Kruskal-Wallis test were explained with examples. Copyright © 2009 by Türkiye Klinikleri.