Long-Term Outcomes and Risk of Pancreatic Cancer in Intraductal Papillary Mucinous Neoplasms

This cohort study investigates the prevalence of intraductal papillary mucinous neoplasms, their association with pancreatic cancer risk, and proportion with malignant transformation.


Introduction
Incidental detection of pancreatic cystic lesions (PCLs) has increased in recent years primarily due to improved resolution and widespread use of cross-sectional imaging. 1,24][5] Current international consensus guidelines 4 for management of IPMNs recommend image-based surveillance with the aim to detect clinical and imaging features of advanced neoplasia (high-grade dysplasia or pancreatic cancer).
Although a minority of IPMNs warrant surgical resection, our existing knowledge about the prevalence, natural history, and risk of malignancy in IPMNs is largely derived from surgical series 6,7 or from patients treated at tertiary centers, [8][9][10] populations which are not representative of the population burden of the disease. 2,10,11There is a critical need to define the prevalence and natural history of IPMNs in a population-based cohort to inform surveillance guidelines.
A 2019 systematic review and meta-analysis 12 estimated the pooled population prevalence of PCLs at approximately 8%.The reported prevalence varies across studies depending on the imaging modality used, geographical location, and age of the cohort studied, and most studies include all PCLs not specifically IPMNs.Furthermore, to our knowledge, there are no population-based estimates of the burden of pancreatic cancer (PC) in individuals with IPMNs or the proportion of PCs that develop from or adjacent to an IPMN.This study uses a population-based cohort to assess the prevalence and associated long-term outcomes and PC risk in IPMN.In addition, we investigated the proportion of PCs that arose from IPMNs (IPMN-PCs) and compared their outcomes with those of non-IPMN PCs.

Methods
This cohort study was approved by the Mayo Clinic Foundation Institutional Review Board and Olmsted Medical Center Institutional Review Board (IRB).The IRB approved a waiver of the requirement to obtain informed consent in accordance with 45 CFR §46.116 as justified by the investigator and a waiver of Health Insurance Portability and Accountability Act of 1996 (HIPAA) authorization in accordance with applicable HIPAA regulations.The study is reported following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.
The Rochester Epidemiology Project (REP) is a medical records linkage system that provides longitudinal, population-based medical data for residents of Olmsted County, Minnesota.The REP database captures more than 95% of the medical care delivered to county residents across all community health care institutions and has served as a valuable resource facilitating populationbased research in a variety of diseases. 13,14From the REP database, we created 2 population cohorts to address our aims: the computed tomography (CT) cohort to estimate the population prevalence and natural history of IPMNs and the PC cohort to investigate the proportion of PCs that arose from malignant transformation of IPMNs and compare clinical outcomes with those of non-IPMN PCs.

CT Cohort
Using the REP database and procedure codes for abdominal imaging, we identified all unique Olmsted County residents between January 1, 2000, and December 31, 2015, aged 50 years or older who underwent a contrast-enhanced abdomen CT scan (eTable 1 in Supplement 1).We divided the (branch duct, mixed, or main duct) based on our a priori definitions (eAppendix 3 in Supplement 1). 9,15PCLs that met IPMN criteria were further classified as Fukuoka high-risk (F-HR), worrisome (F-W), or negative (F-N) based on current international consensus guidelines. 4If patients with IPMNs had subsequent cross-sectional imaging of the abdomen (contrast CT or magnetic resonance imaging [MRI]) performed more than 6 months from the date of index imaging, the most recent cross-sectional imaging and clinical data were also recorded.

PC Cohort
Using the REP database, we identified all patients aged 18 years or older with confirmed PC in Olmsted County between January 1, 2000, and December 31.2019.These patients were initially identified using diagnostic codes (eTable 2 in Supplement 1).Then, each patient electronic medical record was manually reviewed to ensure accurate diagnosis and determine patients who had pathology-confirmed pancreatic ductal adenocarcinoma or cancer that was clinically treated as PC based on imaging and laboratory testing.Patients who were not Olmsted County residents for 1 year or more were excluded.For each unique patient, we collected relevant demographic, clinical, and staging 16 data (eAppendix 4 in Supplement 1).
The cross-sectional imaging at the time of PC diagnosis was reviewed by the study radiologist (N.T.).Each patient was classified using established a priori imaging criteria as definitive IPMN-PC, probable IPMN-PC, possible IPMN-PC, or non-IPMN PC (definitions of classifications used are in eAppendix 5 in Supplement 1).For patients in the PC cohort, surgical pathology was reviewed by a gastrointestinal pathologist (R.P.G.) to confirm histology and determine if the PC arose from an IPMN.

Statistical Analysis
All data for our study were stored and managed with a research electronic data capture tool. 17ntinuous data are summarized as mean (SD) unless otherwise noted.Categorical data are presented as frequency (percentage).CIs for binomial proportions were calculated using the Agresti-Coull method.CIs for incidence estimates were calculated using exact methods with the Poisson distribution.We used the 2-sample t test and Pearson χ 2 test to compare independent groups for continuous and discrete variables, respectively.The Stuart-Maxwell test for marginal homogeneity was used to test for differences in discrete variables between baseline and follow-up images.2B).Patients with an IPMN compared with patients without an  The prevalence of IPMNs using computed tomography is presented A, in each 4-year period of the study and B, in age group by decade.

JAMA Network Open | Gastroenterology and Hepatology
Outcomes   Of 74 patients with PC who underwent surgical resection, 14 individuals (18.9%) were confirmed to have IPMN-PC after pathology expert review.4][5] We used contrast-enhanced CT scans given that this is the most widely used cross-sectional abdominal imaging modality in the US 1 and is less impacted by selection bias compared with more pancreasspecific imaging modalities, such as MRI or endoscopic ultrasound.Previous studies 5,10,19,20 using CT imaging reported a PCL prevalence of 1.2% to 5.4%.Our study identified a higher prevalence, likely associated with image review by experts and limiting of the study population to patients aged 50 years and older.Laffan et al, 10 in a CT-based study in the US, described a PCL prevalence in individuals aged 50 to 89 years ranging from 1.5% from 8.7%, while our study found a prevalence of presumed or suspected IPMNs ranging from 4.9% to 21.2% for the same age group.This reported difference Multiple studies 8,9,15 in tertiary centers have found that IPMNs were associated with increased PC risk and that risk was higher in the presence of F-W or F-HR features.Attempts to quantify the risk of malignant transformation of PCLs to PC using large administrative databases have found that overall risk was relatively low. 21,22Specifically, IPMN-PC was more frequently earlier stage at diagnosis and had overall improved survival compared with non-IPMN PC. 23,24 Our population-based study's findings agree with these prior observations.The improved prognosis of IPMN-PC has been attributed to an earlier stage of diagnosis and a greater proportion of nonmetastatic disease, as observed in our study.Moreover, the survival benefit in IPMN-PC has been primarily observed in node-negative disease, further highlighting outcomes associated with stage shift.Although several studies 8,9 describe the prevalence of PC in patients with IPMN, the true prevalence in the general population has not been previously reported, to our knowledge.This study provides estimates of the prevalence of IPMN-PC in a population-based PC cohort and addresses a critical gap in our understanding of IPMN-PC epidemiology.Using our CT cohort, we also found that the risk of developing PC was relatively low.This is highlighted further in our study given that patients with F-N IPMNs did not have different rates of PC compared with patients without IPMNs.This finding highlights previous work published by our group finding that F-HR IPMNs were the primary contributors to the risk of developing PC while F-W and F-N IPMNs had a low 5-year PC risk. 15though the risk of IPMN-PC is has been extensively described, 8,9 our population-based study further demonstrates that most IPMNs did not progress in Fukuoka stage and did not transform into PC, a similar message expressed by the current American Gastroenterological Association pancreatic cyst guidelines, published in 2015, 25 and studies published in 2022 26 and 2016. 27

Limitations
We acknowledge several limitations in our study.The REP primarily represents a homogenous White population, and inferences from this data may not apply to a population that is more diverse.Most individuals in the CT cohort did not have pathology-confirmed IPMN.We used a priori imaging-based definitions to reduce risk of PCL misclassification. 9,15This is reflective of clinical management given that most IPMNs in clinical practice are presumed or suspected based on imaging and are not biopsy proven.The CT cohort was constructed randomly from a population-based sample of individuals who had undergone cross-sectional imaging.While this potentially introduced selection bias, it provided population-level data and improved on our current epidemiologic knowledge of IPMNs, which is primarily derived from hospital-based referral cohorts.While most CT scans in this population sample were performed for indications unrelated to the pancreas, we did not use indication for imaging as an inclusion or exclusion criterion; therefore, it is possible that some patients in this random sample underwent imaging for a known pancreatic cyst or suspected pancreas cancer.The alternative of an unbiased population-based imaging study to estimate incidental IPMN prevalence would be cost and effort prohibitive and expose study participants to the risk of contrast-enhanced imaging.Another possible limitation of our study was the use of CT instead of MRI.We selected CT imaging given that it is the most widely used imaging modality while recognizing that MRI is better at identifying small IPMNs.However, CT is generally able to detect IPMNs that are clinically relevant, and using MRI as the imaging modality could potentially limit the sample size.A similar population-based study using MRIs may be considered in the future and will be particularly relevant if assessing the incidence of IPMNs.A small subset of the IPMN population in this study underwent long-term imaging follow-up.While this limited the study sample size for assessing long-term outcomes associated with IPMNs, our study provides data on population-level outcomes in contrast to the more widely available referral center-based outcomes in this disease.
The small number of PC events in the CT cohort resulted in wide CIs, and inferences based on asymptotic theory may be unreliable.Additionally, to detect IPMN-PCs, we combined imaging and

Conclusions
This cohort study provides population estimates exploring the association between IPMNs and PCs.
In the general population, IPMNs on CT were present in approximately 1 of 10 patients aged 50 years or older, of which more than 80% were branch-duct IPMNs without high-risk or worrisome features.
Among IPMNs that were F-N at baseline, fewer than 10% developed worrisome or high-risk features

( 5 .
6%) who progressed within Fukuoka criteria.Of 86 patients who were F-N at baseline, 6 patients (6.9%) progressed to F-W (4 patients) or F-HR (2 patients).Patients who progressed from F-N to F-HR did so within 5 years of the index CT scan.No F-W IPMNs progressed to F-HR.In 231 patients with IPMNs, there were 4 PC diagnoses during a median (IQR) follow-up of 12.0 (8.1-15.3)years and 1507 total person-years (2 of 5 patients in F-HR and 2 of 187 patients in F-N).In 1883 patients without IPMNs, there were 17 PC diagnoses during a median (IQR) follow-up of 11.7 (8.4-16.4)years and 15 656 total person-years.Patients with F-N IMPNs who developed PC during surveillance progressed to F-HR, and they both developed PC within 5 years of the index CT scan.The PC incidence rates per 100 person-years for F-HR, F-W, F-N, and

Figure 3 .
Figure 3. Incidence of Pancreatic Cancer

Figure 5 .
Figure 5. Adjusted Survival in Pancreatic Cancer (PC) With and Without Intraductal Papillary Mucinous Neoplasm (IPMN) 100 Outcomes and Risk of Pancreatic Cancer With Intraductal Papillary Mucinous Neoplasms Clinical and demographic data were entered in the study database (eAppendix 2 in Supplement 1).Race was identified from the electronic health record by study team members (A.C. and J.L.).Race options in the databases were Black, Asian, White, other, and unknown.Information regarding race was collected given that this was a population study and the demographics of the population studied may not be generalizable to other populations.CT images were manually reviewed (J.D.L.F. and A.N) for each study individual.If new pancreatic findings or discrepancies from what was described in the original CT scan report were identified, images were reviewed by a second reviewer (H.N. or N.T.) to ensure accuracy.If a PCL was present, it was categorized as an IPMN JAMA Network Open.2023;6(10):e2337799. doi:10.1001/jamanetworkopen.2023.37799(Reprinted) October 17, 2023 2/11 Downloaded From: https://jamanetwork.com/ on 10/20/2023 16-year study into 4-year periods and used a computer-generated, stratified random sampling of a total 2500 unique patients: 625 patients from each 4-year period.See exclusion criteria in eAppendix 1 in Supplement 1.

20/2023 PC Cohort Prevalence and Outcomes of IPMN-Associated PC
Outcomes and Risk of Pancreatic Cancer With Intraductal Papillary Mucinous Neoplasms were similar after adjusting for age and sex.This cohort of 212 patients with non-F-HR IPMNs consisted primarily of patients with IPMNs less than 10 mm in diameter (145 patients [68.4%]), and IPMNs between 10 and 20 mm (57 patients [26.9%]) and 20 mm or greater (10 patients [4.7%]) were in the minority.JAMA Network Open | Gastroenterology and Hepatology JAMA Network Open.2023;6(10):e2337799. doi:10.1001/jamanetworkopen.2023.37799(Reprinted) October 17, 2023 5/11 Downloaded From: https://jamanetwork.com/ on 10/ Combining patients with pathology-confirmed IPMN-PC and those without pathology confirmation but either definitive or probable IPMN-PC based on imaging, the prevalence of IPMN-PC was 31 patients (9.8%; 95% CI, 7.0%-13.7%).This prevalence of IPMN-PC decreased to 21 patients (6.7%; 95% CI, 4.4%-10.0%)ifonlydefinitiveIPMN-PCwas included from the imaging group.Including only individuals with cross-sectional imaging, surgical pathology, or both, 31 patients with IPMN-PC were older (mean [SD] age, 76.9 [9.2] vs 71.3[12.5]years;P=.02),morelikelytoDiscussionThis cohort study found that the estimated population prevalence of IPMNs on CT in individuals aged 50 years or older was 10.9% and this prevalence increased with age.Most of these were branchduct IPMNs, and fewer than 20% of IPMNs had worrisome or high-risk features.During extended follow-up, development of IPMN-PC was infrequent and the PC risk in patients with IPMNs without worrisome or high-risk features was similar to that of individuals without IPMNs.We further identified that approximately 10% of PCs developed in the background of an IPMN.Compared with patients with non-IPMN PC, individuals with IPMN-PC were less likely to have metastatic disease at diagnosis and had improved overall survival.

JAMA Network Open | Gastroenterology and Hepatology Outcomes
and Risk of Pancreatic Cancer With Intraductal Papillary Mucinous Neoplasms Imaging criteria of IPMN-PCs have not been definitively established and may impact the accuracy of our prevalence estimates.However, these imaging criteria for IPMN-PCs were defined a priori in consultation with an expert pancreatic radiologist (N.T.) to avoid the inherent bias of limiting the study to surgically confirmed PC.To further address this limitation, we have included estimates in the subgroup of patients with surgically confirmed PCs.
on follow-up.PC development in IPMN was a rare event overall.Moreover, in F-N IPMNs, the incidence of PC was not significantly different from that among patients without IPMNs, a key finding that if validated in a larger population could challenge the relevance of extended surveillance in this population of patients with F-N IPMNs who make up most of the IPMN population.We further found that malignant IPMNs comprised approximately 10% of PC diagnoses and that IPMN-PC was earlier stage at diagnosis and associated with improved survival regardless of stage at presentation compared with non-IPMN PC.Results of this study provide insights regarding IPMN prevalence and the associated PC risk and may inform future population surveillance strategies and guidelines for patients with IPMNs.reportedreceiving grants from the Centene Charitable Foundation and National Cancer Institute during the conduct of the study and research support from Exact Sciences outside the submitted work and having patents pending for Exact Sciences.Mayo Clinic and Exact Sciences have an intellectual property development agreement, and Dr Majumder is listed as an inventor under this agreement and could share future royalties as an employee of Mayo Clinic.No other disclosures were reported.The use of research electronic data capture for data collection was made possible by the Mayo Clinic Center for Translational Science Activities through grant UL1TR002377 from the National Center for Advancing Translational Sciences, a component of the National Institutes of Health.This study used resources of the Rochester Epidemiology Project (REP) medical records-linkage system, which is supported by grant AG 058738 from the National Institute on Aging, the Mayo Clinic Research Committee, and fees paid annually by REP users.Dr Majumder was supported by a gift to the Mayo Clinic for the Pancreatic Cancer Early Detection Research Program from the Centene Charitable Foundation and by grant U01 CA210138 from the National Cancer Institute.The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.The content of this manuscript is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute or the National Institutes of Health.Computed Tomography Cohort Exclusion Criteria eAppendix 2. Clinical and Demographic Data of Computed Tomography Cohort eAppendix 3. Categories of Suspected and Presumed Intraductal Papillary Mucinous Neoplasm Based on A Priori Definitions eAppendix 4. Clinical and Demographic Data of Pancreatic Cancer Cohort eAppendix 5. Pancreatic Cancer Imaging Classification Using A Priori Definitions to Establish Malignant Transformation of Intraductal Papillary Mucinous Neoplasm eTable 1.Current Procedural Terminology Codes of Abdominal Computed Tomography (CT) Imaging Used to Develop CT Cohort eTable 2. Index Codes Used to Collect Pancreatic Cancer Cohort eTable 3. Demographics and Clinical Characteristics of Computed Tomography Cohort eTable 4. Demographics and Clinical Characteristics of Pancreatic Cancer Cohort Disclaimer: