Inter- and intra-observational variability in immunohistochemistry: a multicentre analysis of diffuse large B-cell...

Although many immunohistochemical (IHC) cancer biomarkers have been identified, very few have translated into routine clinical practice, primarily because of technical and observational inconsistencies between studies. However, despite the obvious need to address such variability, very few studies have done so.

Using bcl-6, CD10, MUM1, GCET1 and FOXP1 antibody staining on diffuse large B-cell lymphoma cases (n = 138) as a model, we employed Cronbach α analysis to quantify interobserver and intraobserver variability between four independent observers (two per institution), scoring two tissue microarrays (TMAs) stained at both institutions using differing staining procedures. The overall concordance between all observations irrespective of staining procedure or TMA source was high (average α = 0.951), with the highest level being reached for CD10 staining (average α = 0.967) and the lowest for bcl-6 (average α = 0.924). Interslide and interinstitutional reproducibility were similarly high (average α = 0.952 and average α = 0.934, respectively). Interobserver/intrainstitutional and interobserver/interinstitutional comparisons showed lower levels of concordance (average α = 0.870 and average α = 0.877, respectively), and intraobserver/interinstitutional comparisons showed the lowest levels of concordance (average α = 0.810), particularly for bcl-6 staining (α = 0.658).

This study suggests that most variability in IHC studies between centres results from inherent limitations of the biomarkers investigated rather than procedural or observational differences.