Reproducibility of subjective immunoscoring of steroid receptors in breast cancer

OBJECTIVE: To evaluate the reproducibility of subjective immunoscoring of steroid receptors in breast cancer by inexperienced observers after brief training by one experienced observer. STUDY DESIGN: Thirteen inexperienced observers were trained for a few minutes semiquantitatively (0-4) to score tumor nuclei in two breast cancer frozen sections immunohistochemically stained each for estrogen or progesterone receptors. Then, the fractions of 0, 1, 2, 3 and 4 'stained' nuclei were estimated individually. Thereafter, the experienced observer pointed out the nuclei one by one, and the participants score these nuclei individually. From these data, histoscores were calculated. RESULTS: When comparing individual scorings (n = 1,320) with the gold standard, discrepancies occurred mainly between scores and 2 and 3, and the best concordance was achieved for the scores 0 and 4. For the histoscores, coefficients of variations for the estimated scores (7.3-33.7, mean 17.2) were systematically higher than those derived from nucleus-by- nucleus scoring (2.6-15.4, mean 10.3). CONCLUSION: Inexperienced observers can be trained quickly to arrive at fairly reproducible histoscores for immunohistochemically stained estrogen and progesterone receptors in breast cancer frozen sections. Nucleus-by-nucleus scoring seems to be more reproducible than estimating fractions of nuclei with certain positivity and should be recommended in cases of values close to the decision threshold.