Investigating various thresholds as immunohistochemistry cutoffs for observer agreement

Ali, A. et al. (2016) Investigating various thresholds as immunohistochemistry cutoffs for observer agreement. Applied Immunohistochemistry and Molecular Morphology, 25(9), pp. 599-608. (doi: 10.1097/PAI.0000000000000357) (PMID:27093449)

Abstract

Background: Clinical translation of immunohistochemistry (IHC) biomarkers requires reliable and reproducible cutoffs or thresholds for interpretation of immunostaining. Most IHC biomarker research focuses on the clinical relevance (diagnostic, prognostic, or predictive utility) of cutoffs, with less emphasis on observer agreement using these cutoffs. From the literature, we identified 3 commonly used cutoffs of 10% positive epithelial cells, 20% positive epithelial cells, and moderate to strong staining intensity (+2/+3 hereafter) to use for investigating observer agreement. Materials and Methods: A series of 36 images of microarray cores stained for 4 different IHC biomarkers, with variable staining intensity and percentage of positive cells, was used for investigating interobserver and intraobserver agreement. Seven pathologists scored the immunostaining in each image using the 3 cutoffs for positive and negative staining. Kappa ([kappa]) statistic was used to assess the strength of agreement for each cutoff. Results: The interobserver agreement between all 7 pathologists using the 3 cutoffs was reasonably good, with mean [kappa] scores of 0.64, 0.59, and 0.62, respectively, for 10%, 20%, and +2/+3 cutoffs. A good agreement was observed for experienced pathologists using the 10% cutoff, and their agreement was statistically higher than for junior pathologists (P=0.02). In addition, the mean intraobserver agreement for all 7 pathologists using the 3 cutoffs was reasonably good, with mean [kappa] scores of 0.71, 0.60, and 0.73, respectively, for 10%, 20%, and +2/+3 cutoffs. For all 3 cutoffs, a positive correlation was observed with perceived ease of interpretation (P<0.003). Finally, cytoplasmic-only staining achieved higher agreement using all 3 cutoffs than mixed staining patterns. Conclusions: All 3 cutoffs investigated achieve reasonable strength of agreement, modestly decreasing interobserver and intraobserver variability in IHC interpretation. These cutoffs have previously been used in cancer pathology, and this study provides evidence that these cutoffs can be reproducible between practicing pathologists.

• Item Type: Articles
• Status: Published
• Refereed: Yes
• Glasgow Author(s) Enlighten ID: Jamieson, Professor Nigel and Oien, Professor Karin and Derakhshan, Professor Mohammad and Bell, Dr Sarah and Bilsland, Dr Alan and Orange, Miss Clare and McKay, Mr Colin and Chang, Professor David and Ali, Dr Asif
• Authors: Ali, A., Bell, S., Bilsland, A., Slavin, J., Lynch, V., Elgoweini, M., Derakhshan, M. H., Jamieson, N. B., Chang, D., Brown, V., Denley, S., Orange, C., McKay, C., Carter, R., Oien, K. A., and Duthie, F. R.
• College/School: College of Medical Veterinary and Life Sciences > School of Cancer Sciences; College of Medical Veterinary and Life Sciences > School of Cardiovascular & Metabolic Health; College of Medical Veterinary and Life Sciences > School of Medicine, Dentistry & Nursing
• Journal Name: Applied Immunohistochemistry and Molecular Morphology
• Publisher: Lippincott, Williams & Wilkins
• ISSN: 1541-2016
• ISSN (Online): 1533-4058
• Published Online: 16 April 2016
• Copyright Holders: Copyright © 2016 Wolters Kluwer Health, Inc.
• First Published: First published in Applied Immunohistochemistry and Molecular Morphology 2017
• Publisher Policy: Reproduced in accordance with the copyright policy of the publisher