Objectives To assess whether MRI can exclude axillary lymph node metastasis, potentially replacing sentinel lymph node biopsy (SLNB), and consequently eliminating the risk of SLNB-associated morbidity. Methods PubMed, Cochrane, Medline and Embase databases were searched for relevant publications up to July 2014. Studies were selected based on predefined inclusion and exclusion criteria and independently assessed by two reviewers using a standardised extraction form. Results Sixteen eligible studies were selected from 1,372 publications identified by the search. A dedicated axillary protocol [sensitivity 84.7 %, negative predictive value (NPV) 95.0 %] was superior to a standard protocol covering both the breast and axilla simultaneously (sensitivity 82.0 %, NPV 82.6 %). Dynamic, contrast-enhanced MRI had a lower median sensitivity (60.0 %) and NPV (80.0 %) compared to non-enhanced T1w/T2w sequences (88.4, 94.7 %), diffusionweighted imaging (84.2, 90.6 %) and ultrasmall superparamagnetic iron oxide (USPIO)-enhanced T2*w sequences (83.0, 95.9 %). The most promising results seem to be achievable when using non-enhanced T1w/T2w and USPIO-enhanced T2*w sequences in combination with a dedicated axillary protocol (sensitivity 84.7 % and NPV 95.0 %). Conclusions The diagnostic performance of some MRI protocols for excluding axillary lymph node metastases approaches the NPV needed to replace SLNB. However, current observations are based on studies with heterogeneous study designs and limited populations. Main Messages• Some axillary MRI protocols approach the NPVof an SLNB procedure.• Dedicated axillary MRI is more accurate than protocols also covering the breast.• T1w/T2w protocols combined with USPIO-enhanced sequences are the most promising sequences.