Interobserver agreement for estrogen receptor immunohistochemical analysis in breast cancer: a comparison of manual and computer-assisted scoring methods

Ann Diagn Pathol. 2004 Feb;8(1):23-7. doi: 10.1016/j.anndiagpath.2003.11.004.


Differences in scoring methods for estrogen receptor (ER) immunohistochemistry may cause significant variation in the results. Scoring practices differ within the United States and internationally and include semiquantitative scoring formulas, manual estimations, and computer-assisted techniques. The goal of this study was to determine the rate of interobserver variability for manual ER scoring at our institution and compare the ER scores obtained by manual scoring with those obtained using image-analysis software (QCA, Lake Bluff, IL). In a series of 70 consecutive invasive breast cancers, ER was assayed using standard immunohistochemical techniques and the monoclonal antibody 6F11. Scoring was performed independently by three breast pathologists, and the scores were compared with those obtained using the QCA image-analysis system, using 10% nuclear staining as the cutoff for positivity. We found that 43 cases (61%) were ER positive, 25 cases (36%) were ER negative, and two cases (3%) showed ER staining of less than 10%. The consensus scores for the 70 cases showed a high level of agreement with the ER scores determined by image analysis (kappa = 0.84). Interobserver variability was low. The kappa scores for each observer showed strong agreement with the consensus score, the image-analysis score, and between the observers. Our findings show that interobserver agreement for manual scoring of ER is strong, and that manual or computer-aided scoring techniques are comparable.

Publication types

  • Comparative Study

MeSH terms

  • Breast Neoplasms / epidemiology*
  • Female
  • Humans
  • Image Processing, Computer-Assisted* / methods
  • Immunohistochemistry
  • Observer Variation
  • Receptors, Estrogen / metabolism*
  • Research Design


  • Receptors, Estrogen