We evaluated the risk of sampling errors in specimens of biopsy size, which may be caused by heterogeneous overexpression of Her2/neu in gastric cancer (GC). The study cohort comprised 454 gastrectomy patients with adenocarcinoma of the stomach or esophago-gastric junction. Tissue micro-arrays (TMAs) served as ‘biopsy procedure’ and were generated from formalin-fixed and paraffin-embedded tissue: five tissue cylinders were collected randomly from each tumor, rendering 2230 core cylinders. These were compared with 454 whole tissue sections obtained from the same paraffin blocks. Her2/neu expression and gene amplification were analyzed by immunohistochemistry and in situ hybridization. The Her2/neu status was determined according to GC scoring system by two independent observers. In whole tissue sections, 37 (8.1%; observer 1) and 38 (8.4%; observer 2) of the GCs, and in the corresponding TMAs, 28 (6.3%; observer 1) and 28 (6.3%; observer 2) of the GCs were classified as Her2/neu-positive (kappa value 98.5% and 96.2%; P < 0001). Comparison of whole tissue sections with corresponding TMAs showed a false-negative rate of 24% and a false-positive rate of 3% for TMAs. Assessment of the Her2/neu status in tissue biopsies carries a significant risk of sampling errors, thereby rendering patients unsuitable for treatment with trastuzumab.