Assessing the differential diagnosis between autoimmune pancreatitis (AIP) and pancreatic adenocarcinoma is like sailing between two rocks. Rock number 1 is to perform a surgical pancreatic resection in patients with AIP; rock number 2 consists of giving steroids to patients with resectable pancreatic adenocarcinoma thought to have AIP. Both attitudes are clearly diagnostic failures responsible for major patient prejudice. In this issue of Gut (see page 1704) in the study of Moon et al,1 the authors propose a diagnostic challenge with steroids during 2 weeks in patients with atypical AIP (ie, not completely corresponding to published diagnostic criteria for AIP) carefully selected in a tertiary centre. This diagnostic challenge was already included in the Mayo criteria for AIP since in the acronym HISORt (see the Box for the meaning), Rt means “Response to steroid treatment”.2 Moon et al show that all the 15 patients who had a clinical and radiological response within 15 days of treatment had confirmed AIP. On the contrary, the 7 patients without response had pancreatic adenocarcinoma. Although the study was adequately designed and the results nicely presented, the message might not be devoid of danger, bringing the sailing boat close to …