PDF
Citation
Print
INTRODUCTION
Pancreatic cancer is one of the most lethal malignancies globally, with a poor prognosis and rapidly rising incidence, both domestically and internationally [1]. In the United States, pancreatic cancer is projected to become the second leading cause of cancer-related death by 2030, surpassing colorectal cancer [2]. In South Korea, pancreatic cancer surpassed gastric cancer for the first time in 2022, becoming the fourth leading cause of cancer-related mortality [3]. According to the Korea National Cancer Registry statistics from 2021, the 5-year relative survival rate for pancreatic cancer remains low, at approximately 15.9%, highlighting the critical need for improved diagnostic strategies and early intervention [4]. Pancreatic cancer is currently the seventh leading cause of cancer mortality worldwide and is predicted to rise to the second or third leading cause in some Western countries by 2030, mainly because of aging populations and ongoing diagnostic challenges.
In contrast, intraductal papillary mucinous neoplasms (IPMNs) of the pancreas represent precancerous lesions that are more amenable to early detection and preventive intervention. Approximately 10% of all pancreatic cancers are estimated to arise from IPMN, underscoring the significance of IPMN as a potential therapeutic target for early management and cancer prevention [5,6].
IPMNs are anatomically and histologically classified into main-duct (MD), branch-duct (BD), and mixed-type (MT) subtypes and pathologically graded as low-grade dysplasia, high-grade dysplasia, or invasive carcinoma [7]. A recent meta-analysis found that MD-IPMNs and MT-IPMNs have a high likelihood of malignant transformation, with 59% (95% confidence interval, 54–64%) demonstrating high-grade dysplasia or invasive carcinoma at the time of surgical resection. Conversely, the malignant transformation rate for BD-IPMNs was only 8.2% after long-term follow-up exceeding 10 years [8]. This notable difference in malignancy risk between subtypes highlights the need for individualized risk assessment and management plans.
Molecular studies have revealed that IPMN most commonly harbors dual KRAS and GNAS mutations. Further progression to high-grade dysplasia or invasive carcinoma involves the accumulation of additional genetic alterations [9]. Liquid biopsy methods, such as circulating cell-free DNA (cfDNA) analysis and next-generation sequencing (NGS) of cyst fluid, have demonstrated promising diagnostic accuracy, with high specificity and moderate-to-high sensitivity for detecting at least one KRAS or GNAS mutation [10]. These molecular markers, combined with imaging and clinical indicators, are currently being refined into precise prognostic tools for improved patient management.
Clinical guidelines for IPMN management have been revised frequently over the past 2 decades. The Sendai guidelines in 2006 and the subsequent Fukuoka guidelines in 2012 and 2017 introduced the concepts of high-risk stigmata (HRS) and worrisome features (WF) to stratify indications for surgery [11]. However, concerns have been raised about the low specificity of these guidelines, potentially leading to unnecessary surgical resections. The 2018 European evidence-based guidelines further refined indications for surgery by introducing relative and absolute criteria and including elevated carbohydrate antigen (CA) 19-9 levels as a consideration for surgery [12]. The 2024 Kyoto international evidence-based guidelines redefined HRS and WF and stated that surveillance could be discontinued for BD-IPMN tumors ≤ 2 cm in size, if they remain stable during 5 years of follow-up [13]. Nevertheless, discrepancies between guidelines continue to complicate clinical decision-making.
Recent studies demonstrated superior predictive accuracy of artificial intelligence (AI)-based radiomics models, compared to the 2017 Fukuoka guidelines, for identifying malignant IPMN [14]. Moreover, a meta-analysis found that small BD-IPMN lesions that are stable in size during long-term follow-up have an extremely low risk of progression [15]. Unresolved clinical questions remain, including cost-effectiveness of blood- and cyst fluid-derived genetic panels and the optimal timing of surveillance discontinuation (especially in older or frail patients with a reduced life expectancy and/or high surgical risk).
This review aims to re-examine IPMNs from the perspective of their precancerous nature by summarizing recent epidemiologic data and molecular mechanisms; comparing diagnostic and treatment recommendations across major international guidelines; evaluating evidence for imaging and molecular-based risk prediction and surveillance strategies; and suggesting future research directions. The goal is to provide practical evidence to support clinicians in making balanced decisions that avoid overtreatment and undertreatment.
MAIN SUBJECTS
Epidemiology and molecular pathogenesis
With the widespread adoption of abdominal computed tomography (CT) and magnetic resonance imaging (MRI) and the worldwide aging population, incidental detection of pancreatic cystic lesions has increased markedly. Asymptomatic pancreatic cysts are detected in an estimated 11–18% of abdominal imaging studies, with IPMN accounting for up to half of these lesions [16,17].
The incidence of IPMN increases with increasing age, exceeding 20% in adults aged ≥ 70 years. Additional risk factors for IPMN include female sex, abdominal obesity, and the presence of renal cysts. The natural history and risk of malignant transformation differ considerably according to IPMN subtype. MD-IPMN and MT-IPMN have been reported to harbor high-grade dysplasia or invasive carcinoma at the time of resection in 40–60% of cases, whereas BD-IPMN tumors have a considerably lower malignancy rate. Long-term observational studies have reported that BD-IPMNs without initial high-risk features have a cumulative malignant transformation rate of approximately 8%, but approximately 25% of these tumors develop imaging-defined WF over time, underscoring the importance of continued surveillance [18].
The molecular pathogenesis of IPMN involves the sequential accumulation of genetic alterations. Early alterations commonly involve dual KRAS (70–95%) and GNAS (45–80%) mutations. Subsequent genetic events, including inactivation of RNF43, PIK3CA, and STK11, drive progression toward high-grade dysplasia, while additional alterations, such as TP53 and SMAD4 mutations, promote progression to invasive carcinoma. Molecular diagnostic tools, such as cfDNA analysis and digital droplet polymerase chain reaction analysis of cyst fluid, have good diagnostic sensitivity (approximately 79%) and very good specificity (98%) for detecting KRAS or GNAS mutations [9]. Additionally, PancreaSeq Genomic Classifier, a comprehensive NGS panel that simultaneously analyzes 74 genes, has remarkable diagnostic accuracy. It has a 95% sensitivity and 100% specificity, outperforming conventional imaging and clinical assessments [19].
Recent advances in spatial transcriptomics have highlighted subtype-specific molecular signatures: NK6 homeobox 2 acts as a critical differentiation regulator in gastric-type IPMN, correlating with a relatively indolent disease course, whereas pancreatobiliary-type IPMN shows enhanced malignancy potential through increased cell-cycle activity, nuclear factor kappa B (NF-κB) pathway activation, and pancreatic ductal adenocarcinoma (PDAC)-like transcriptional features [20]. Furthermore, a recent digital spatial proteomics study involving 187 IPMN cases characterized subtype-specific landscapes of the nucleotide‐binding oligomerization domain–like receptor pyrin domain containing 3 (NLRP3) inflammasome and identified the interleukin (IL)-18/IL-18 binding protein ratio as an independent prognostic marker for invasive IPMN, thus emphasizing the pivotal role of the immune microenvironment in determining progression risk [21].
Thus, current evidence indicates that the risk of malignant transformation is driven by molecular alterations beginning with KRAS-GNAS mutations and further influenced by subtype-specific tumor microenvironments. An integrative approach combining imaging characteristics with genomic, transcriptomic, and immunologic biomarkers is essential to effectively stratify risk, minimize unnecessary surgical intervention, and facilitate timely prevention of progression to invasive carcinoma.
Major international/society guidelines: key differences in diagnostic and management recommendations
Clinicians managing IPMN commonly refer to four main sets of guidelines: the 2017 Fukuoka International Association of Pancreatology consensus guidelines; the 2015 American Gastroenterological Association (AGA) guidelines; the 2018 European evidence-based guidelines on pancreatic cystic neoplasms (PCN); and the 2024 Kyoto international evidence-based guidelines. Each guideline is based on different objectives and methodologies and exhibits notable differences in recommendations for surgical indications and surveillance strategies.
Resection criteria
Table 1 summarizes the criteria for surgical resection of IPMN, as outlined in the four main international/society guidelines [11-13,22,23]. Of the four sets of resection criteria, the Fukuoka algorithm is the most aggressive, as it prioritizes sensitivity. The AGA guidelines for resection are more conservative, aiming to minimize costs and complications. The European PCN guidelines occupy an intermediate position through their absolute and relative indications approach. The Kyoto 2024 guidelines refine the Fukuoka criteria based on more recent meta-analyses and reduce overtreatment and surveillance burden for small, stable BD-IPMN tumors.
Surveillance interval and discontinuation criteria
Surveillance recommendations for tumors that do not meet the criteria for resection vary between the international/society guidelines (Table 1). The 2017 Fukuoka guidelines recommend surveillance every 3–6 months via endoscopic ultrasonography (EUS) or every 6–12 months via MRI, depending on the lesion size and associated risk factors, with lifelong surveillance as the general principle [11]. The AGA guidelines suggest MRI surveillance 1 year after diagnosis, followed by imaging every 2 years if no risk factors are present. Discontinuation of surveillance is permitted after 5 years if no structural changes occur, with reassessment only if significant changes or high-risk features develop [5,23]. In contrast, the 2018 European PCN guidelines advocate MRI surveillance at 6-month intervals for the first year, followed by annual monitoring indefinitely, as long as the patient remains fit for surgery [12]. The 2024 Kyoto guidelines introduce two notable options for small (≤ 2 cm) BD-IPMN lesions if they remain unchanged with no WF during 5 years of surveillance: discontinue surveillance or continue monitoring at least annually (mainly to detect the emergence of PDAC). Additionally, the Kyoto guidelines recommend postoperative surveillance every 6 months for the first 3 years, followed by annual imaging with CT or MRI and EUS [13].
Diagnostic tools and application of biomarkers
Table 2 shows the general use of diagnostic tools and biomarkers for risk assessment of IPMN. The Kyoto 2024 guidelines are notable for systematically incorporating molecular markers and EUS-fine needle aspiration cytology into their risk stratification model, which was based on a systematic literature review and GRADE methodology.
Practical points for clinical application
Regarding surgical decision-making, employing a “≥ 2 features” strategy for defining IPMN that warrant surgery can reduce overtreatment, which is especially relevant for older patients with significant comorbidities and high surgical risk. The AGA guidelines recommend surgery if ≥ 2 high-risk features are present, whereas the European PCN guidelines recommend surgery if any absolute indication is present, if ≥ 1 relative indications are present in patients without comorbidities, or if ≥ 2 relative indications are present in patients with significant comorbidities (Table 1). However, a main pancreatic duct (MPD) diameter ≥ 10 mm or obstructive jaundice warrants immediate surgical intervention according to all four major guidelines, leaving little room for variation.
The debate surrounding surveillance duration–discontinuation of monitoring after 5 stable years (AGA) versus lifelong surveillance (European PCN)–has moved toward a middle ground with the Kyoto guidelines, which propose both options for some lesions. The discontinuation option for small BD-IPMN lesions is worthwhile discussing with patients, as prolonged surveillance is associated with significant psychological and financial burden.
The issue of biomarker utilization has also evolved. Despite the cautious stance on cost-effectiveness from the AGA and European PCN, the 2024 Kyoto guidelines grade the assessment of KRAS and GNAS mutations in cyst fluid as evidence level B, heralding the likely future expansion of molecular diagnostics in clinical practice.
As the major international/society guidelines differ considerably in definitions of lesion risk, criteria for surgery, and surveillance intervals, clear identification of the reference guidelines being used is crucial when designing multicenter research. Cross-referencing criteria, such as HRS versus absolute surgical indications, should be explicitly mapped for comparability.
Imaging and molecular-based risk prediction and surveillance strategies
The primary goal in managing pancreatic IPMN is to precisely identify the minority of tumors at high risk for invasive carcinoma, while simultaneously eliminating unnecessary surgical interventions and prolonged surveillance in the majority of cases. Evidence over the past 3–4 years supports the transition to precision stratification, incorporating multilayered data from imaging morphology, AI-driven radiomics, cyst fluid NGS, liquid biopsy (cfDNA), and serum or cyst fluid protein/enzyme biomarkers. Integrated models and economic evaluations are increasingly being applied to refine risk stratification.
Traditional imaging features, such as mural nodules, MPD dilation, and acute pancreatitis, remain the strongest short-term predictors of high-risk IPMN. A 2023 meta-analysis of 9 cohorts (2,214 surgical cases) identified enhancing mural nodules ≥ 5 mm and MPD diameter ≥ 10 mm as strong risk factors for invasive carcinoma, with odds ratios of 18.7 and 11.9, respectively [24].
Resolving the clinical dilemma of when to surgically intervene versus continuing surveillance hinges on effectively combining morphologic indicators derived from imaging with molecular information obtained from blood and cyst fluid samples. A 2024 multicenter study (n = 3,336; 22,339 person-years follow-up) demonstrated that morphologic variables defined in the 2024 Kyoto guidelines accurately predicted pancreatic cancer occurrence during longer-term follow-up (from 6 months to 10 years). MPD diameter 5–9.9 mm, annual growth ≥ 2.5 mm, and history of acute pancreatitis increased long-term cancer risk by 3.5–5.7-fold, whereas HRS (e.g., enhancing mural nodule ≥ 5 mm, MPD ≥ 10 mm) increased the likelihood of cancer diagnosis within 6 months to nearly 50%. Importantly, the number of WF was directly associated with prognosis, with lesions containing 3–4 WF approaching a 50% cumulative cancer risk over 10 years. This finding provides a robust rationale for shifting from a binary (present/absent) to a quantitative, weighted-scoring approach [17].
Imaging techniques continue to evolve. High-resolution EUS significantly improves specificity (up to 99%) by detecting subtle nodules < 5 mm and minor wall thickness changes. Deep-learning radiomics can discriminate malignant IPMN based on texture patterns alone, with area under the receiver operating characteristic curve (AUC) values as high as 0.93. However, widespread adoption of this technology awaits the development of standardized equipment and protocols [25,26].
From a molecular perspective, dual KRAS and GNAS mutations remain the strongest initial predictive events. Cyst fluid NGS panel (PancreaSeq-GC) predicts pathologic high-grade dysplasia or invasive carcinoma with a 95% sensitivity and 100% specificity, thus serving as a valuable adjunctive tool for ambiguous BD-IPMN cases. Additionally, circulating cfDNA methylation signatures demonstrate high diagnostic performance (AUC = 0.89), offering potential noninvasive guidance for long-term surveillance, although cost-effectiveness and reimbursement issues remain unresolved [27,28].
Surveillance strategies translate risk stratification into clinical practice. The revised Kyoto guidelines introduced an option to discontinue surveillance for BD-IPMN ≤ 2 cm that remain stable for 5 years based on evidence showing extremely low cumulative cancer risks at 10 and 15 years (2.7% and 6.1%, respectively). Conversely, lesions with ≥ 3 WF or MPD diameter 7–9 mm warrant intensive monitoring (3–6-month intervals with CT/MRI and annual EUS-cytology/NGS), as their cumulative 10-year cancer risk exceeds 50% [17].
Nevertheless, challenges regarding risk prediction and surveillance strategies remain. First, neither imaging WF nor current molecular markers sufficiently predict synchronous PDAC, necessitating novel screening strategies targeting this specific risk. Second, high-cost technologies, such as NGS and EUS-based confocal imaging, have variable practical applicability (depending on national healthcare financing and equipment availability), highlighting the need for more comprehensive cost-effectiveness analyses. Lastly, AI-based radiomics models currently lack extensive external validation across diverse hospitals and equipment, which will be required for standardized image acquisition and regulatory approval before routine clinical integration.
Future research directions
Future research should focus on expanding precise risk stratification methodologies and developing fully personalized surveillance and intervention plans. For example, the 2024 Kyoto nomogram and algorithms based on the cumulative number of WF were developed and validated mainly with data from Japan [17]. These multimodal models (integrating imaging, clinical, and molecular variables) must be externally validated in international prospective cohorts encompassing diverse populations and clinical environments.
Although recent single-centre studies suggest that AI-radiomics models trained on texture- and shape-based features from CT, MRI, and EUS images can distinguish malignant from benign IPMN with very high apparent accuracy (development-set AUC, 0.93–0.98), early external validations show more modest performance (AUC 0.80–0.89), and real-world prospective efficacy remains under investigation [29-31]. Therefore, standardized image acquisition protocols across devices and vendors, along with harmonized data processing pipelines, are essential before integrating such tools into routine clinical practice.
In the molecular domain, the clinical and economic impacts of the cyst fluid PancreaSeq-GC NGS panel (sensitivity, 95%; specificity, 100%) on surgical decision-making and surveillance intervals should be reassessed in multicenter real-world settings [32]. Furthermore, noninvasive molecular surveillance strategies, such as blood-based cfDNA methylome and proteomics multiplex panels (e.g., the DAYBREAK study), require validation in large-scale cohorts to demonstrate cost-effectiveness [19].
Regarding surveillance de-escalation, accumulating long-term follow-up data suggest that discontinuing surveillance in older adults with stable BD-IPMN (and no WF or HRS) over a 5-year period is safe [33]. This should be confirmed in randomized controlled trials incorporating patient age, lesion size, and molecular characteristics. For high-risk surgical patients, initial studies report high technical success rates (> 90%) and low severe complication rates (< 10%) for EUS-guided local therapies (e.g., radiofrequency ablation, chemo-photothermal ablation). However, controlled studies are required to determine their long-term cancer-preventive effects [34].
Clinical trials targeting subtype-specific tumor immune microenvironments, such as the NLRP3-inflammasome and PD-L1 pathways, are also needed. Additionally, metabolic and genomic signatures capable of detecting synchronous PDAC (which cannot be explained by existing morphologic and molecular indicators) require further investigation.
CONCLUSION
IPMN represents a precancerous lesion that can precede pancreatic cancer, yet not all lesions have equal malignant potential. Over the past 2 decades, epidemiologic, molecular, and imaging research, combined with iterative updates of international/society guidelines, have improved our understanding of various issues: differences in malignancy risk between IPMN subtypes; multistage molecular progression, beginning with KRAS and GNAS mutations; hierarchical significance of morphologic and clinical indicators of cancer risk (e.g., mural nodules, MPD dilation, faster growth rate, CA 19-9); and the emerging clinical value of advanced technologies, including NGS, AI-radiomics, and liquid biopsy. The 2024 Kyoto guidelines introduced the first evidence-based roadmap for reducing overtreatment by integrating WF number and cyst fluid NGS/cytology results into risk assessment, while also allowing discontinuation of surveillance for small (≤ 2 cm) BD-IPMN lesions that remain stable for 5 years.
The challenge for clinicians is to incorporate this evolving knowledge into clinical practice and research design through “risk-adaptive” protocols. Specifically, this includes externally validating multidimensional nomograms that integrate imaging-based WF, molecular biomarkers, and AI-based scores, embedding them into routine clinical workflows; significantly reducing surveillance intensity for patients with ultra-low-risk BD-IPMN; and establishing early surgical intervention or alternative treatments, such as EUS-guided local therapies, for patients with multiple WF or molecular-positive high-risk lesions, through clinical trials. Concurrently, efforts must continue to identify novel metabolic and genomic signatures for detecting synchronous PDAC arising independently from existing IPMN and to develop immune-preventive strategies targeting subtype-specific tumor immune microenvironments.
In summary, the key to IPMN management involves achieving a balance between overtreatment and undertreatment. By systematically applying imaging assessment, molecular diagnostics, and risk stratification, tailored surveillance and intervention strategies can be developed. The ultimate goal is to develop precision onco-prevention models that effectively reduce progression to invasive pancreatic cancer, while minimizing burdens on patients and healthcare resources.
XML Download