Human epidermal growth factor receptor 2 (HER2) is a central predictive biomarker in breast cancer. Inaccurate HER2 results in different laboratories could be as high as 20%. However, this statement is based on data generated more than 13 years ago and may not reflect the standards of modern diagnostic pathology. We compared central and local HER2 testing in a total of 1581 HER2-positive tumors from five clinical trials. We evaluated the clinical relevance for pathological complete response (pCR) and disease-free survival in a subgroup of 677 tumors, which received an anti-HER2 therapy. Over the period of 12 years, the discordance rate for HER2 decreased from 52.4 (GeparTrio) to 8.4% (GeparSepto). Discordance rates were significantly higher in hormone receptor (HR)-positive tumors (26.6%), compared to HR-negative tumors (16.3%, P<0.0001), which could be explained by a different distribution of HER2 mRNA levels in HR-positive and HR-negative tumors. pCR rates were significantly lower in discordant tumors (13.7%) compared to concordant tumors (32.2%, GeparQuattro and GeparQuinto, P<0.001). In survival analysis, tumors with discordant HER2 testing had a reduced overall survival (OS) in the HR-negative group (P=0.019) and a trend for improved OS in the HR-positive group (P=0.125). The performance of local HER2 testing was considerably improved over time and has reached a 92% concordance, which shows that quality initiatives in diagnostic pathology are working. Tumors with discordant HER2 testing had a reduced therapy response and different survival rates.