Differentiating between Metastatic Carcinoma of Breast Origin (MCBO) and Primary Lung Carcinoma: A Search for the Ideal Immunopanel.
Kathy Kawaguchi, Yi-Fang Liu, Paul Chadwick, Patrick Wagner, Stefano Monni, Edi Brogi, Sandra J Shin. Weill Cornell Medical College, New York; Memorial Sloan Kettering Cancer Center, NY
Background: Differentiating between MCBO and primary lung carcinomas is currently a diagnostic quandary as there are no immunostains that are highly sensitive and specific for breast. A commonly employed immunopanel including CK7, CK20, ER, PR and GCDFP-15 in addition to pertinent negative stains specific for other sites is often utilized but commonly proves insufficient. Furthermore, the significance of “focal positivity” using any one stain is questionable. The purpose of our study was to identify immunostains either alone or as a group that would better distinguish between these two entities.
Design: Tissue microarrays containing duplicate 0.6 mm diameter cores of 109 metastatic breast ca and 102 primary lung ca cases were constructed. 4 µ-thick sections were used to stain with antibodies against CK 7 (Dako), CK20 (Dako), ER (Novocastra), PR (Novocastra), Androgen Receptor (AR) (Biogenex), Mammoglobin (Dako), Napsin (IBL), GATA-3 (Santa Cruz Biotech), and TTF-1 (NeoMarkers). For each case, percentage of tumor cells stained as well as intensity of staining was recorded. An H-score was calculated (range 0-300) and the data statistically analyzed.
Results: Recursion partition analysis was employed to identify which immunostain or combination of immunostains could be effective in differentiating between MCBO and primary lung carcinomas. TTF-1 was determined to be the best classifier. Carcinomas with TTF-1 H-scores <30 were identified as MCBO samples and those with H-scores >=30 as primary lung carcinomas. This classifier was trained on a data set with 43 MCBO samples and 39 primary carcinomas, and its misclassification error was 14.5%. It was selected among more complex models as having the lowest cross-validation misclassification error (16.8%). When applied to an independent cohort of similar size, its misclassification error was 15.9%. Further investigation showed that GATA-3 and Napsin could also be employed effectively, with H-scores of above or below 52.2 and 55 in MCBO, respectively.
Conclusions: TTF-1, GATA-3, and Napsin proved to be most effective in distinguishing between MCBO and primary lung carcinomas among the immunostains studied. A recursion partition analysis of H-scores showed that these three immunostains were nearly equally proficient in separating between the two entities.
Tuesday, March 1, 2011 1:45 PM
Platform Session: Section B, Tuesday Afternoon