Aims Determining the site of origin of a metastatic neuroendocrine tumour (NET) can be challenging and has important prognostic and therapeutic implications. An immunohistochemical (IHC) panel consisting of TTF1, CDX2, PAX8/PAX6 and Islet1 is often employed. However, there can be a significant IHC overlap among different primary sites. Herein, we sought to determine the utility of including Special AT-rich sequence binding protein-2 (SATB2) in the IHC panel that is used for determining the site of origin of a metastatic NET. Methods Paraffin tissue microarrays consisting of 137 primary NETs (26 lung, 22 jejunoileal, 8 appendix, 5 stomach, 4 duodenum, 17 rectum and 55 pancreas) were stained for SATB2, in addition to the well-described lineage-associated markers, such as TTF1, CDX2, PAX6 and Islet1. Additionally, a tissue microarray consisting of 21 metastatic NETs (1 lung, 1 stomach, 8 jejunoileal and 11 pancreas) was stained for TTF1, CDX2, SATB2 and Islet1. The results were recorded as no staining, weak staining and moderate to strong staining. Results All appendiceal NETs and majority (88%) of the rectal NETs were positive for SATB2. All primary foregut NETs (stomach, pancreas, duodenum and lung) were negative for SATB2, except for one pulmonary NET with weak staining. However, among the metastatic tumours, 5 of 11 pancreatic NETs, 1 stomach NET, 1 lung NET and 2 of 8 jejunoileal NETs showed weak staining. Receiver operating characteristic analysis incorporating sensitivity and specificity data of IHC panel, considering moderate to strong staining as truly positive cases, showed that inclusion of SATB2 to the previously described NET IHC panel outperformed the panel without SATB2, raising the specificity for pancreaticoduodenal NETs from 81.2% to 100%, with a positive predictive value (PPV) of 100% and negative predictive value (NPV) of 82.22% (p<0.0001); for appendiceal NETs the specificity changed from 99.1% to 98.5% and sensitivity increased from 11.8% to 80%, with a PPV and NPV of 66.67% and 99.26%, respectively (p<0.0001); and for rectal NETs the specificity increased from 97.6% to 99.3% and sensitivity raised from 7.1% to 66.7%, with a PPV and NPV of 80% and 98.53%, respectively (p<0.0001). Conclusions SATB2 stain is useful in differentiatingIslet1/PAX6 positive pancreatic and rectal NETs, as rectal NETs are typically moderately to strongly positive for SATB2 and pancreatic NETs are usually negative or weakly positive for SATB2. Moderate to strong staining for SATB2 is suggestive of an appendiceal or a rectal primary. SATB2 may complement the panel of CDX2, TTF1 and Islet1 in determining the site of origin of an NET in a metastatic setting.