BACKGROUND: The reliability of immunohistochemically-determined biomarkers using tissue microarrays (TMAs) of clinical specimens has long been open to question. Heterogeneity related to tumor biology might compromise determination of accurate biomarker expression in tumors, especially in small core biopsies. We evaluated the reliability of immunohistochemical staining scoring in small core biopsies using 11 biomarkers in non-small-cell lung cancer (NSCLC). MATERIALS AND METHODS: Four 1-mm tumor cores from 178 NSCLCs, 2 representing peripheral areas close to the border of normal lung tissue and 2 representing central areas, were examined. The biomarkers analyzed included p63, p40, cytokeratin 1/5/10/14, cytokeratin 7, thyroid transcription factor-1, napsin A, cyclin-D1, p53, Ki-67, integrin beta-1, and thymidylate synthase. RESULTS: Using a random intercept logistic regression model, immunohistochemical marker expression of TMAs had a moderate to high reliability measured using the intraclass correlation coefficient (0.67-0.99) between cores within the same tumor. Reliability was dependent on the selected biomarker, with lineage-specific biomarkers being less heterogeneously expressed than functional biomarkers. Expressions of most biomarkers showed no significant difference between central versus peripheral tumor cores. CONCLUSION: Our results demonstrated that biomarkers involved in clinical tumor classification (cell lineage markers) of NSCLC can be adequately assessed using 1 or 2 biopsy samples. However, the optimal number of cores required for biomarkers with functional properties varied from 1 to > 4 cores. The results indicate that the optimal number of biopsies required to compensate for potential biomarker heterogeneity should be determined individually for each biomarker used in clinical settings.