Systematic review and meta-analysis of cytokeratin 19-based one-step nucleic acid amplification versus histopathology for sentinel lymph node assessment in breast cancer

Background One‐step nucleic acid amplification (OSNA) is a new rapid assay for detecting breast cancer metastases during surgery, saving a second procedure for patients requiring an axillary clearance. Many centres in the UK and abroad have adopted OSNA in place of routine histopathology, despite no published meta‐analysis. The aim of this systematic review and meta‐analysis was to determine whether intraoperative OSNA for lymph node assessment is comparable to routine histopathology in the detection of clinically relevant metastases. Methods PubMed, Embase, Web of Knowledge and regional databases were searched for relevant studies published before December 2012. Included studies compared OSNA and standard histology using fresh lymph nodes that were assessed in a clearly defined systematic manner in accordance with the index study. Results Twelve eligible studies were identified that included 5057 lymph nodes from 2192 patients. Although meta‐analysis using a random‐effects model showed a similar overall proportion of macrometastases detected (429 of 3234 versus 432 of 3234; odds ratio 0·99, 95 per cent confidence interval 0·86 to 1·15), analysis of concordance showed that the pooled positive predictive value for detecting macrometastases was 0·79. This suggests that up to 21 per cent of patients found to have macrometastases using OSNA would have an axillary clearance when histology would have classified the deposits as non‐macrometastases. Furthermore, analysis of data from the index publication showed that the range of cytokeratin 19 titres for tumours of a given volume is too wide to predict tumour size. Conclusion OSNA has an unacceptably low positive predictive value, leading to axillary clearances that would not be recommended if standard histology had been used to assess the sentinel node. Positive predictive value for macrometastasis is low