Improving the reproducibility of diagnosing micrometastases and isolated tumor cells.

BACKGROUND: The latest edition of the tumor-lymph node-metastasis (TNM) classification of malignant tumors distinguishes between isolated tumor cells (pN0) and micrometastases (pN1mi). The reproducibility of these categories has not been assessed previously.
METHODS: Digital images from 50 cases with low-volume lymph node involvement from axillary sentinel lymph nodes were circulated twice for evaluation (Evaluation Rounds 1 and 2) among the members of the European Working Group for Breast Screening Pathology, and the members were asked to categorize lesions as micrometastasis, isolated tumor cells, or something else and to classify each case into a pathologic lymph node (pN) category of the pathologic TNM system. Methods for improving the low reproducibility of the categorizations were discussed between the two evaluation rounds. kappa Statistics were used for the assessment of interobserver variability.
RESULTS: The kappa value for the consistency of categorizing low-volume lymph node load into micrometastasis, isolated tumor cells, or neither of those changed from 0.39 to 0.49 between Evaluation Rounds 1 and 2, but it was slightly lower for the pN categories (0.35 and 0.44, respectively). Interpretation of the definitions of isolated tumor cells (especially with respect to their localization within the lymph node), lack of guidance on how to measure them if they were multiple, and lack of any definitions for multiple simultaneous foci of lymph node involvement were listed among the causes of discordant diagnoses.
CONCLUSIONS: The results of the current study indicated that the definitions available have minor contradictions and do not permit a reproducible distinction between micrometastases and isolated tumor cells. Refinement of these definitions, therefore, is required. One refinement that may improve reproducibility is suggested in this report. (c) 2004 American Cancer Society.