Quantifying Quality: A Review and Scale Proposal

Abstract

Interlaboratory comparisons of subjective evaluations are difficult because of definition ambiguities and inconsistencies in the sign and numerical scales commonly used. With regards to the few objective evaluations reviewed, they have yet to reach a level of readily availability either because of instrumentation, costs, complexity, or stringent quality requirements of the sections. It also was found that the statistical analyses of results are infrequent although very useful and independent of the complexity of the compared numerical scales. It is proposed a quantifying scale with a numerical value of −1 for negative results. The reaction of the control is used as the intensity standard for each test with a value of +1 positive when less than half of the target elements or cells stained less or as intensely as the control. A +2 positive value is assigned when more than half of the target elements or cells are stained as intensely or more than the control. Because the staining intensity of the positive control is a defining element of the scale, for immunohistochemistry, it is recommended to use the reaction of internal controls or normal tissues (with stable epitope amounts) as the standard over that of tumor tissues (with variable and usually above normal epitope amounts). When comparing evaluated results, either between epitopes or laboratories, the scores for immunohistochemistry tests do not characterize their manifestation potential unless multiplied by the antibodies working dilution factors (which are independent of the Ig concentrations); when this correction was performed, some not statistically different results became significantly different even for epitopes, whose antibodies had similar dilution factors. The statistical comparison of results with “n” different procedures should be the subject of blind evaluations by as many reviewers as possible and each procedure should be assigned P = l/n as its theoretical frequency to be compared with the frequency obtained by the evaluators (observed) using the χ² test. Parametric tests such as F (analysis of variance) and the Student t should be always preferred to nonparametric tests to characterize interobserver variability and significance between paired or clustered results. This review also includes guidelines on how to develop a meaningful quality assurance program to be used as the basis for the general performance improvement work of the histology laboratory. (The J Histotechnol 28:89, 2005)

Submitted December 29, 2004; accepted with revisions April 15, 2005

Keywords:

• signs and numerical subjective scales
• objective quantifying scales
• controls for IHC
• epitope comparative manifestation potential
• correction of IHC scores
• antibodies dilution factors
• theoretical frequency for multiple variable tests
• parametricte tests
• QA and PI programs guidelines