Trastuzumab-containing therapy is a standard of care for human epidermal growth factor receptor-2 (HER2)-positive breast cancer. In pre-clinical models, a wide range of molecular mechanisms have been associated with reduced sensitivity to trastuzumab in vitro. These include expression of the truncated HER2 receptor fragment p95HER2, activating mutation of the gene encoding the class 1A catalytic subunit of phosphatidylinositol 3-kinase (PIK3CA), loss of phosphatase and tensin homolog (PTEN), activation of other downstream signal transducers, prevention of cell cycle arrest, increased signaling through alternative (HER or non-HER) tyrosine kinase receptors, and resistance to antibody-dependent cellular cytotoxicity. However, the clinical significance of these mechanisms as determinants of trastuzumab efficacy in vivo has been unclear. Here, we review clinical studies of potential predictive biomarkers of trastuzumab efficacy in HER2-positive breast cancer and consider whether evaluation of such markers might inform patient selection for therapy. We find that clinical evidence relating to potential predictive biomarkers is mostly limited to small, retrospective studies, many of which have yielded conflicting findings. Some trends are evident in the retrospective data and in biomarker analyses from randomized clinical trials, particularly relating to activation of the phosphatidylinositol 3-kinase pathway, but none is sufficiently strong to form a basis for patient selection. This may be explained by the fact that multiple mechanisms of action determine the clinical efficacy of trastuzumab. In the absence of novel, validated biomarkers of efficacy, trastuzumab eligibility should continue to be based on evaluation of HER2 status according to standard methods.