Autoimmune pancreatitis (AIP) is a distinctive form of chronic pancreatitis that can mimic pancreatic carcinoma. In the past, AIP accounted for up to 27% of Whipple resections performed for suspected adenocarcinoma. More recently, with increased awareness of AIP and reports of its steroid responsiveness, tru-cut needle biopsies are increasingly used as an aid in preoperative diagnosis. We noticed a distinctive patchy distribution to the pathologic abnormalities in many cases of resected AIP that could potentially interfere with preoperative diagnosis by needle biopsy. We studied 39 pancreatic resections with AIP, defined by the following triad of features: (1) lymphoplasmacytic infiltrates around ducts, (2) acinar lymphoplasmacytic inflammation with atrophy and fibrosis, and (3) obliterative phlebitis. Criteria for inclusion in the study included either submission of the entire resection specimen (n=21) or extensive histologic sampling (n=18) defined as submission of > or =10 sections. We reviewed all hematoxylin and eosin-stained sections and (1) mapped areas of sparing and involvement by AIP, (2) classified the AIP as lobulocentric, ductocentric, or mixed, and (3) tabulated numbers of immunoglobulin (Ig) G4+ plasma cells in areas of involvement and sparing. To be included as an area of sparing, both duct and acinar parenchyma had to be free of lymphoplasmacytic inflammation, and the focus had to be at least 0.5 cm in diameter. Our results demonstrate a high prevalence of patchiness in AIP. Thirty-two (82%) specimens had areas of sparing (mean of 22% of each specimen spared, range 0.8% to 80%). The largest focus of uninvolved pancreas varied from 0.5 to 8.8 cm(2) (mean: 1.8 cm(2)). In the remaining 7 (18%) cases, the changes of AIP were diffuse, with involvement of the entire submitted specimen. Number of IgG4+ plasma cells correlated highly with areas of involvement versus sparing by AIP; there were > or =5 IgG4+ plasma cells/20x field in 34 of 35 (97%) involved foci, but in only 1 of 26 (4%) histologically uninvolved foci (P<0.001). Classification as lobulocentric AIP (n=11), ductocentric AIP (n=15), and mixed AIP (n=12) did not correlate with extent of patchiness (P=0.92) or with the volume of spared parenchyma (P=1.0). These results demonstrate patchy involvement by AIP in a majority of resected pancreata. In specimens containing large areas of uninvolved parenchyma, this raises the potential for underdiagnosis by tru-cut biopsy. In patients with radiologic and serologic features (eg, elevated serum IgG4 level) suspicious for AIP, this potential pitfall in pathologic diagnosis should be considered before proceeding to surgery. IgG4 immunostaining of apparently negative biopsies may be helpful, but only in a small minority of cases.