Background: Immunohistochemistry (IHC) and fluorescent in situ hybridization (FISH) are reliable ways to identify amplification of the HER-2/neu (HER2, symbol ERBB2) gene, but each technique requires a high-quality tissue sample, which may not be available. We investigated whether serum concentrations of the HER2 extracellular domain (ECD) can be used as an alternative to tissue HER2 status in metastatic breast cancer, and we defined an optimal decision level of serum HER2 for prediction of tissue HER2 status. Methods: We determined HER2 expression in 195 metastatic breast cancer patients by IHC and performed FISH analysis on tumors for which IHC staining was graded as 2؉. We measured serum HER2 by immunoassay and used receiver operating characteristics (ROC) curve analysis to determine optimal serum HER2 ECD concentrations for differentiation between positive and negative HER2 status. Results: IHC results were 0/1؉ for 30 (15%) of the patients, 2؉ for 89 (46%), and 3؉ for 76 (39%). FISH revealed HER2 amplification in 19 (21%) of the IHC 2؉ tumors. Mean (SE) serum HER2 ECD was 22.2 (5.1) g/L in the tissue HER2-negative group, significantly lower than the concentration of 363 (96) g/L in the tissue HER2-positive group (P <0.0001). ROC curve analysis showed 95% specificity and 62% sensitivity for tissue HER2 positivity at 37 g/L of serum HER2. Conclusion: To use serum HER2 concentration as an alternative to direct determination of tissue HER2 status, we suggest 37 g/L as a cutoff for predicting positive tissue HER2 with 95% specificity. Sensitivity, however, is low.