Diagnostic delay in type I autoimmune pancreatitis: clinical, imaging, endoscopic and histologic predictors of timely diagnosis.

Autoimmune pancreatitis (AIP) is a rare fibroinflammatory disease of the pancreas. It was first described in 1961 by Sarles et al. as a form of primary sclerosing cholangitis, but it was not identified as a separate entity until 1995 by Yoshida et al. ( 1 , 2 ). Type 1 AIP, also known as lymphoplasmacytic sclerosing pancreatitis, is a pancreatic manifestation of IgG4-related disease (IgG4-RD) and accounts for approximately 80% of cases ( 3 , 4 ). Type 2 AIP, also known as idiopathic duct-centric pancreatitis, makes up most of the remaining cases and is often seen in patients with inflammatory bowel disease, specifically ulcerative colitis ( 3 ). AIP that does not meet type 1 or 2 criteria is labeled as AIP ‘Not Otherwise Specified’. A commonly used AIP diagnostic criteria in the United States is the Mayo HISORt criteria; which include components of histology, imaging, serology, organ involvement and response to therapy ( 5 , 6 ). There are other diagnostic criteria, including the Japanese Pancreas Society criteria, Korean diagnostic criteria, Asian diagnostic criteria, and the International Consensus Diagnostic Criteria, which also take into account imaging, serology, and histopathology ( 7 - 10 ). Based on a Japanese epidemiological survey, 60% of AIP patients present with nonspecific symptoms, such as abdominal pain and jaundice, while 40% present asymptomatically but with abnormal imaging or laboratory results ( 11 ). AIP can present as a localized mass in the pancreas on imaging, making it difficult to differentiate from pancreatic cancer. Further, characteristic serologic markers of AIP have low sensitivity and are also elevated in pancreaticobiliary cancer ( 12 - 14 ). These factors frequently delay diagnosis of AIP while malignancy is being ruled out. In addition, endoscopic ultrasound-guided fine-needle aspiration (EUS-FNA) and core biopsies are operator dependent and may lead to false negative results for AIP due to inadequate tissue and the occasional patchy involvement of the pancreas. Some patients will undergo resection of believed pancreatic and/or biliary malignancies only for histology to reveal AIP instead. As a result of these difficulties in diagnosis, one retrospective study reports a mean of 25.6 months between symptom onset and diagnosis of AIP ( 15 ). There is limited data quantifying the delays in diagnosis of AIP. There is also a limited understanding of what factors are associated with delays in the diagnosis, as such we sought to quantify the delays in diagnosis and identify associated risk factors for such delays. We present this article in accordance with the STROBE reporting checklist (available at https://tgh.amegroups.com/article/view/10.21037/tgh-25-46/rc ).

We conducted a retrospective, single-center study between 2013 and 2023 of adult patients at a tertiary referral center in Houston, TX, USA. Patients with a formal diagnosis of AIP were compared to a randomly selected group of controls with pancreatic ductal adenocarcinoma (PDAC). Data on clinical presentation, timing of diagnosis (weeks from symptom onset to diagnosis, weeks from presentation to healthcare to diagnosis), lab results, imaging, procedures, and initial treatment regimen were collected. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Baylor College of Medicine institutional review board (IRB #H-53685) and individual consent for this retrospective analysis was waived. PDAC was selected as a control group due to its overlapping features with AIP, such as presenting as an isolated pancreatic mass. While other entities, such as primary pancreatic lymphoma, may also share these features, their rarity made them less suitable to be a control group. A list of patients with ICD-9 code (577.1) and/or ICD-10 code (K86.1) for ‘Other chronic pancreatitis’ with or without concomitant immunosuppressive medication use was generated using the electronic medical record system. In addition, our institution’s pathology database was queried for all surgical pathology and or cytology specimens obtained from the pancreas, which were then reviewed to identify patients with pancreatic histology findings consistent with AIP (lymphoplasmacytic sclerosing pancreatitis, periductal lymphoplasmacytic infiltrate, obliterative phlebitis, or storiform fibrosis). Patients identified using these two approaches were then manually reviewed to confirm AIP diagnosis based on the HISORt criteria. Only patients with type 1 AIP were included. Finally, using the previously referenced pathology database, patients with confirmed PDAC were randomly selected as controls using a computerized algorithm. Date of ‘formal’ AIP diagnosis was established once the patient met diagnostic criteria based on the 2009 revision of HISORt criteria as follows ( 10 ): AIP was diagnosed if there was diagnostic histology (lymphoplasmacytic sclerosing pancreatitis or >10 cells/hpf on biopsy and/or >50 cells/hpf in a resection specimen of IgG4-positive plasma cells with ≥2 of the following: periductal lymphoplasmacytic infiltrate, obliterative phlebitis, or storiform fibrosis) findings on pancreatic core biopsy or resection specimen. AIP was also diagnosed if there was typical imaging (diffuse enlarged gland with delayed enhancement, with or without capsule-like rim) with any of the following: elevated IgG4 (≥2 times the upper limit of normal), other organ involvement, or compatible histology (storiform fibrosis with lymphoplasmacytic infiltrate and obliterative phlebitis). Response to steroids was used to diagnose AIP after a negative work up for malignancy and presence of 1 highly suggestive feature of AIP (IgG4 >2 times upper limit of normal, other organ involvement), or at least 2 supportive features of AIP (serum IgG4 elevation <2 times upper limit of normal, clinical/radiological evidence of other organ involvement or compatible histology). Date of PDAC diagnosis was derived from the date of biopsy or surgical resection with histology findings consistent with PDAC. Steroid response was defined as improvement or resolution in imaging findings above and/or serologic markers. Date of symptom onset was derived from patient-reported onset of symptoms related to AIP or PDAC as documented by medical provider notes. Date of presentation to health care was derived from the first presentation (either hospital admission or clinic appointment) to our medical center or other centers, as documented in the chart for symptoms or imaging findings concerning AIP or PDAC. All patients were evaluated with either endoscopic (EUS) or imaging modalities [computed tomography (CT)/magnetic resonance imaging (MRI)] or both. Patients were considered lost to follow-up if their last documented encounter with a healthcare provider within our medical system occurred before fulfilling the HISORt criteria for AIP diagnosis. Two-sample comparisons were by chi-squared tests for proportions, Student’s t -test for normally distributed variables, and Mann-Whitney U test for other variables. We used Pearson correlation to evaluate the association between the indicated continuous variables. Shapiro-Wilk test was used to determine normality of continuous variables. Due to the nature and distribution of the primary outcome (time to diagnosis), generalized negative binomial regression models were used to study the association between covariates and time to diagnosis. A two-sided P value of <0.05 was used to indicate statistical significance in all analyses. Patients that were lost to follow up were not included in analysis. STATA version 16.0 (Statacorp, College Station, TX, USA) was used for statistical analysis.

Overall, 41 AIP patients were identified ( Figure 1 ), and compared to 41 patients with PDAC. All patients had type 1 AIP. Patients with AIP were significantly younger at 57.8±16.4 years compared to 66.0±13.8 years for the PDAC group (P=0.005). The predominant race was Caucasian in both the AIP (n=29, 71%) and the PDAC (n=31, 76%) groups, respectively. Other variables, including alcohol intake, diabetes prevalence, number of prior pancreatitis events, and current tobacco use were comparable between groups ( Table 1 ). Flow diagram of included patients with autoimmune pancreatitis. AIP, autoimmune pancreatitis; ICD, International Classification of Diseases. *, statistically significant difference (P<0.05). AIP, autoimmune pancreatitis; BMI, body mass index; IBD, inflammatory bowel disease; PDAC, pancreatic ductal adenocarcinoma; SD, standard deviation; UC, ulcerative colitis. Diagnosis of AIP was confirmed with histology in 19 (46.3%) patients, with the remainder diagnosed with a combination of clinical, imaging and laboratory findings with or without a positive steroid trial. All patients (41, 100%) with PDAC were diagnosed by histology. AIP patients had significantly longer delays in formal diagnosis and therapy from the onset of symptoms compared to controls with PDAC (79.1±141.2 vs. 5.2±5.43 weeks, P<0.001). After presentation to healthcare, patients with AIP had significantly longer delays in formal diagnosis and therapy compared to controls with PDAC (38.1±89.9 vs. 1.7±3.2 weeks, P<0.001) ( Figure 2 ). Timing to diagnosis of AIP from presentation to healthcare was significantly shorter in the last 5 years of the study compared to the initial 5 years (13 vs. 28 weeks, P=0.02). Boxplot comparison of AIP and PDAC patients. (A) Weeks from symptom onset to formal diagnosis (8 outliers censored for graphing purposes). (B) Weeks from presentation to healthcare to formal diagnosis. AIP, autoimmune pancreatitis; PDAC, pancreatic ductal adenocarcinoma. Female sex was associated with a significantly longer delay in diagnosis from onset of symptoms to formal diagnosis (136.6±202.2 vs. 42.3±63.5 weeks, P=0.004). Patients who never consumed alcohol were diagnosed later than those who had >2 standard drinks per day (103.7±168.46 vs. 31.6±39.2 weeks, P=0.005). Patients with diabetes were diagnosed earlier with AIP compared to non-diabetics (34.8±33.9 vs. 110.4±177.2 weeks, P=0.02). No other baseline characteristics were associated with delays including race, ethnicity, body mass index, tobacco, or other prior medical history, including family history of autoimmune disorders or comorbidities. Presenting with an episode of acute pancreatitis was associated with a significant delay in obtaining a final diagnosis (133.3±212.0 vs. 50.9±76.1 weeks, P=0.03). Jaundice was associated with a significantly decreased time from onset of symptoms to formal diagnosis (27.1±33.0 vs. 93.7±156.3 weeks, P=0.01). No other symptom correlations were identified including presenting with abdominal pain, weight loss, back pain, acute pancreatitis, or new onset diabetes ( Tables 2,3 ). Additionally, no correlation between time to diagnosis and total bilirubin, IgG4 level, fasting blood glucose levels, HbA1c, lipase or CA 19-9 was identified, however the statistical power of these comparisons was limited given the extent of missing lab values ( Table S1 ). *, statistically significant difference (P<0.05). AIP, autoimmune pancreatitis; SD, standard deviation. *, statistically significant difference (P<0.05). AIP, autoimmune pancreatitis; n/a, not available; SD, standard deviation. A total of 35 (85%) and 27 (66%) AIP patients underwent CT and MRI imaging, respectively. Presence of a mass of the pancreas on any imaging or endoscopic modality (30/41, 73.2%) was associated with a significantly shorter time to diagnosis from presentation to healthcare (17.9±31.9 vs. 92.9±157.6 weeks, P=0.04). Obtaining a CT scan did not decrease the time to diagnosis of AIP. Location of pancreatic mass, pancreatic duct dilation or atrophy, and imaging findings of chronic pancreatitis were not associated with a statistically significant difference in time to diagnosis of AIP ( Table 4 ). In contrast, the presence of fatty infiltration of the pancreas (fatty pancreas) on CT imaging was associated with a significantly longer delay in diagnosis (194.1±163.9 vs. 33.7±60.3 weeks, P<0.001). *, statistically significant difference (P<0.05). AIP, autoimmune pancreatitis; CT, computed tomography; SD, standard deviation. Patients with MR imaging were diagnosed with AIP earlier than those without (45.4±86.0 vs. 143.9±199.1 weeks, P=0.006). The presence of a dilated main bile duct on MRI was associated with a decreased time to AIP diagnosis (11.1±14.9 vs. 55.3±95.5 weeks, P=0.003). Parenchymal pancreatic atrophy and chronic pancreatitis on MRI were associated with delays in AIP diagnosis ( Table 5 ). No other MRI findings, including edematous pancreas, pancreatic capsular prominence, or lymphadenopathy were associated with timing of AIP diagnosis. *, statistically significant difference (P<0.05). AIP, autoimmune pancreatitis; MRCP, magnetic resonance cholangiopancreatography; MRI, magnetic resonance imaging; n/a, not available; SD, standard deviation. A total of 37 AIP patients (90%) underwent EUS, and 18 (44%) underwent ERCP. In the subset of patients who underwent EUS, a finding of a pancreatic mass or ‘mass-like’ area on EUS or its location was not associated with timing of diagnosis. Other EUS findings, such as common bile duct or pancreatic duct dilation, or endosonographic stigmata of chronic pancreatitis had no association with time to AIP diagnosis. On ERCP, presence of bile duct dilatation, was associated with shorter time to AIP diagnosis ( Table 6 ). *, statistically significant difference (P<0.05). AIP, autoimmune pancreatitis; CBD, common bile duct; ERCP, endoscopic retrograde cholangiopancreatography; EUS, endoscopic ultrasound; PD, pancreatic duct; SD, standard deviation. Obtaining EUS-guided or surgical biopsies was associated with a significantly shorter time to formal diagnosis (49.7±79.9 vs. 221.6±262.9 weeks, P=0.04). Seven (17%) patients with AIP underwent surgical interventions (2 Whipple, 1 distal pancreatectomy, 4 exploratory laparotomy with biopsy) for evaluation of pancreatic mass. One of these patients also had the additional indication of recurrent pancreatitis. All 7 patients were found to have AIP instead of malignancy on surgical pathology. One patient was found to also incidentally have a pancreatic mucinous cystic neoplasm in addition to AIP on pathologic evaluation. Listwise deletion was used for missing data as they appeared missing completely at random. Missing data not already outlined on respective tables were outlined in Table S1 .

In this 10-year retrospective analysis, we collected extensive data points to quantify delays in diagnosis of AIP and the predictors of delays in 41 type 1 AIP patients compared to 41 randomly selected PDAC patients. We found that patients with AIP had significant delays in diagnosis when compared to PDAC controls (delay to diagnosis from symptom onset: 79.1 vs. 5.2 weeks, P<0.001, delay to diagnosis from presentation to healthcare: 38 vs. 1.7 weeks, P<0.001). We identified several factors associated with delays in diagnosis including female gender, never consuming alcohol, no history of diabetes, presenting with acute pancreatitis, fatty infiltration of the pancreas on CT, and parenchymal pancreatic atrophy or chronic pancreatitis on MRI. Studies evaluating delays to diagnosis suggest a potential gender bias in diagnosing some gastrointestinal disorders, with females experiencing delays compared to males ( 16 ). Differences between male and female presentations in AIP have been recently reported. A multicenter retrospective analysis of 735 European patients with AIP found significant differences between males and females with type 1 AIP: males were older, and more often had smoked and/or consumed alcohol. Males more frequently presented with jaundice, and less often with abdominal pain. Pancreatic exocrine insufficiency was more prevalent in males. Finally, radiologically, parenchymal enlargement occurred less in males, but narrowing of the main pancreatic duct more frequently ( 17 , 18 ). Additionally, epidemiological studies indicate that AIP has a male-to-female ratio of approximately 3:1, which lowers the pretest probability of AIP in women ( 11 ). Coupled with the higher prevalence of functional gastrointestinal disorders in women compared to men (46.5% vs. 34.2%), this disparity may influence clinicians to deprioritize AIP as a diagnosis in female patients and instead consider functional gastrointestinal disorders ( 19 ). Beyond gender, behavioral factors may influence the AIP diagnostic timeline. For instance, individuals who abstain from alcohol and lack a history of diabetes may experience a delay in diagnosis due to requiring less frequent medical care than those with formal diagnosis of diabetes or alcohol use disorder. In theory, more frequent contact with healthcare provides more opportunities for symptoms to be evaluated and vice versa. Finally, the presence of a fatty pancreas and/or an atrophic pancreas likely limit or attenuate the classic imaging findings of AIP, possibly driving clinicians away from the diagnosis when attention is not brought to the possibility of AIP by a radiologist. Unfortunately, autoimmune gastrointestinal conditions like AIP and IgG4-related cholangitis often have overlapping features with malignancy, leading to diagnostic errors and delays, as seen in 17% of our patients who underwent surgical interventions for presumed cancer. Similarly, one study found that 34% of 53 patients underwent surgery for suspected malignancy but were later diagnosed with IgG4 cholangitis or AIP, with a resulting median follow-up of 58 months ( 20 ). Further complicating matters, some autoimmune gastrointestinal conditions, such as autoimmune hepatitis—like AIP—can also initially present asymptomatic (25–34%), with symptoms developing over 2–120 months (2±2.46 years) ( 21 , 22 ). Even when patients present to the emergency department with gastrointestinal symptoms of IgG4-RD (typically abdominal pain or jaundice), the average diagnostic delay is 8 months (range, 1–130 months), shorter for pancreatobiliary involvement (4 months) and longer for retroperitoneal fibrosis (27 months) ( 23 ). Delays were significantly longer (17 months, range, 1–160 months) when patients were initially referred to outpatient clinics, a timeline that closely mirrors our experience as a tertiary referral center. Interestingly, we also identified several factors associated with a more expeditious diagnosis of AIP, including the presence of pancreatic mass on cross sectional imaging, biliary ductal dilation on MRI or ERCP, and obtaining EUS-guided or surgical biopsies. The presence of a pancreatic mass and obtaining EUS or surgical biopsies are closely correlated, since a diagnosis of mass on imaging often prompts referral for biopsy, and consequently a histologic diagnosis of AIP can be obtained where tissue is evaluated for classic AIP features, including periductal lymphoplasmacytic infiltrate, obliterative phlebitis, or storiform fibrosis. The presence of biliary ductal dilation on either MR imaging or ERCP was also associated with earlier diagnosis, likely due to patients presenting with jaundice from biliary obstruction from AIP and ultimately undergoing more invasive procedures to rule out malignancy, ultimately leading to a more expeditious diagnosis. Finally, we found that patients presenting in the last 5 years of the study were diagnosed significantly faster compared to those presenting in the first 5 years. We suspect this is related to a number of factors, including increased awareness of AIP as well as wider availability and increased yield of EUS-guided biopsies. Our study has a few inherent limitations, particularly being a retrospective study at a single tertiary referral center. One consideration is referral bias, where patients with complex presentations and diagnostic ambiguity were more likely to be referred to our center, potentially increasing observed delays compared to a community practice. Time of symptom onset was identified from initial physician documentation in the electronic medical record, which is subject to recollection bias by the patient. However, our analysis still demonstrated that even when the more objective timing from presentation to healthcare to diagnosis was analyzed, significant delays in diagnosis of AIP remained in comparison to PDAC. A priori power calculation was not performed due to this study’s retrospective design. Given the rarity of AIP, our relatively small sample size may limit generalizability and power to detect other less significant associations with delays in AIP diagnosis. It is important to point out that our findings only apply in patients with type 1 AIP, since that was the only type included in our cohort. Finally, while only 46% of AIP cases had histologic confirmation, all patients met HISORt diagnostic criteria. This reflects real world practice, but nevertheless, misclassification remains a possibility particularly in patients diagnosed based on predominantly clinical and radiological features.

AIP remains a challenging diagnosis, with extensive delays in formal diagnosis averaging 1.5 years from symptom onset. We identified demographic characteristics, symptoms, imaging findings, and diagnostic modalities associated with either a delayed or shortened time to AIP diagnosis. These findings may allow clinicians to identify patients with AIP more readily and in a timely manner.

The article’s supplementary files as