Bridging the intraductal papillary mucinous neoplasm risk stratification gap: A critical review of current and emerging biomarkers

Abstract

Intraductal papillary mucinous neoplasms (IPMNs) are the most common cystic precursor to pancreatic ductal adenocarcinoma. However, current clinical guidelines do not sufficiently distinguish indolent lesions from high-grade dysplasia or invasive carcinoma. The existing risk stratification frameworks are either sensitivity-focused (e.g., International Association of Pancreatology, European, American College of Gastroenterology) or specificity-oriented (e.g., American Gastroenterological Association). They must balance the risk of missed cancer against substantial overtreatment. Therefore, numerous diagnostic and prognostic biomarkers have been investigated, including cyst fluid carcinoembryonic antigen, cytology, next-generation sequencing panels, cytokines such as interleukin 1β and prostaglandin E₂, glycoproteins such as carbohydrate antigen 19-9 and mucin 5AC, and cell-surface markers such as Das-1. Although several markers were promising in early studies, their performance has been limited by small cohorts, surgically-enriched populations, inconsistent histopathologic endpoints, frequent pooling of high-grade dysplasia and invasive carcinoma, and heterogeneous sample platforms. Few biomarkers have undergone rigorous multicenter validation or cost-effectiveness assessment, and many require technically complex workflows that limit widespread adoption. This critical review summarizes the current evidence and delineates the methodological, biological, and translational barriers that impede accurate IPMN risk stratification. We also identify key pitfalls in the biomarker literature and outline the priorities necessary to bridge the gap between early proof-of-concept studies and clinical integration into IPMN management.

Key Words: Pancreatic cancer; Intraductal papillary mucinous neoplasms; Biomarker; Risk stratification; Cyst fluid; Liquid biopsy

Core Tip: Current guidelines for intraductal papillary mucinous neoplasm management inadequately distinguish indolent lesions from high grade dysplasia, resulting in both overtreatment and missed cancer. This evidence review critically examines established and emerging circulating and cyst fluid biomarkers, including molecular, inflammatory, and glycoprotein markers. We emphasize diagnostic performance, methodological limitations, and translational barriers. We highlight the impact of pooled histological endpoints, small surgical cohorts, assay heterogeneity, and lack of multicenter validation on the accuracy of the current studies. Finally, we outline the requirements needed to clinically integrate these biomarkers into intraductal papillary mucinous neoplasm risk stratification.

INTRODUCTION

Pancreatic cancer is highly malignant, leading to almost 500000 deaths worldwide[1]. With a 7.9%-14.6% 5-year survival rate, it is considered likely be the second leading cause of cancer-related mortality by 2032[2]. In 10%-15% of cases pancreatic cancer arises from mucinous cystic precursors including intraductal papillary mucinous neoplasm (IPMN) and mucinous cystic neoplasm (MCN)[3,4]. IPMNs are a mucin-producing epithelial neoplasm originating within the pancreatic ductal system. Due to the increasing use of high-resolution cross-sectional imaging, the frequency of IPMN detection has increased. Although most IPMNs follow a benign clinical course, they are established precancerous lesions with the potential to progress to invasive adenocarcinoma through a well-defined adenoma-carcinoma sequence. This malignant potential and the limitations of current diagnostic tools pose a significant clinical management challenge of accurately distinguishing indolent lesions from those requiring surgical intervention.

Diagnosis

Most IPMNs are discovered incidentally during radiological examinations such as computed tomography or magnetic resonance imaging[5,6]. Most lesions are asymptomatic at the time of diagnosis. However, symptomatic IPMNs can present with recurrent pancreatitis, abdominal discomfort, or pain due to mucin (MUC) overproduction and ductal obstruction. They will infrequently manifest with obstructive jaundice, weight loss, or new-onset diabetes[7]. Once a pancreatic cystic lesion is detected, the diagnostic workflow focuses on characterization, classification, and malignant potential.

Accurate classification of the IPMN subtype is fundamental because the risk of progression differs substantially between main-duct (MD), branch-duct (BD), and mixed-type lesions. MD-IPMN and mixed-type IPMNs have a higher incidence of high-grade dysplasia (HGD) or invasive cancer and trigger a lower threshold for surgical management than isolated BD-IPMN[7,8]. Imaging features that denote MD involvement (main pancreatic duct dilation > 5 mm) or a solid enhancing component usually prompt surgical referral. Most small BD-IPMNs behave indolently and are appropriate for conservative management with imaging surveillance in the absence of high-risk criteria[7-9].

Imaging is the key diagnostic approach for the initial assessment and longitudinal surveillance. Magnetic resonance imaging with magnetic resonance cholangiopancreatography is the preferred first-line modality. Contrast-enhanced computed tomography is useful for surgical planning and the detection of calcification or solid extracystic disease[10]. Endoscopic ultrasound (EUS) with or without fine-needle aspiration (FNA) can also improve the diagnosis and stratification of IPMN.

LITERATURE SEARCH

This article was conducted as a narrative critical review with a structured literature search focused on biomarkers relevant to the diagnosis and risk stratification of IPMN. The objective was not to perform a formal quantitative meta-analysis, but rather to synthesize the most clinically relevant evidence while critically appraising recurring methodological limitations that affect interpretation and translation of biomarker studies.

A literature search was performed in PubMed/MEDLINE, Scopus, and Web of Science for studies published in English from January 2000 to September 2026. Search terms combined controlled vocabulary and free-text keywords related to IPMN and biomarkers, including: “IPMN”, “intraductal papillary mucinous neoplasm”, “pancreatic cyst”, “biomarker”, “cyst fluid”, “serum”, “cyst fluid cytology”. Reference lists of key reviews, meta-analyses, and guideline papers were also screened to identify additional relevant studies.

We included original studies, prospective and retrospective cohorts, case-control studies, meta-analyses, and major guideline or consensus documents that addressed: (1) Biomarker-based stratification of IPMN dysplasia or invasive disease; or (2) Methodological and translational issues relevant to biomarker implementation. Studies focused exclusively on non-IPMN cystic neoplasms without clear relevance to IPMN were excluded from the main synthesis, although selected studies were referenced when necessary for context.

Given the major heterogeneity across the field, particular attention was paid to the following study-level features during evidence appraisal: Specimen source [cyst fluid, serum, tissue, extracellular vesicles (EV)], cohort type (surgical vs surveillance), endpoint definition [low-grade dysplasia (LGD) vs HGD vs invasive carcinoma (IC); pooled vs separated endpoints], assay platform, cutoff prespecification, and validation status (single-center derivation vs external or multicenter validation). Because the literature is dominated by surgically enriched retrospective cohorts and inconsistent histopathologic terminology, emphasis in this review was placed not only on reported sensitivity and specificity, but also on the extent to which these design features may inflate apparent biomarker performance. This approach was chosen to match the central aim of the review: To identify not only promising biomarkers, but also the methodological barriers that currently limit reliable clinical integration.

RISK STRATIFICATION

IPMNs of the pancreas present a management challenge that requires balancing the lifetime risk of malignant progression against the substantial morbidity associated with pancreatic resection[11]. The goal of IPMN surveillance is to prevent invasive pancreatic cancer by identifying patients at the highest risk for surgical intervention, specifically those with HGD. The window of opportunity for surgical intervention is broad; genomic analyses suggest a period of over 3 years between the emergence of HGD to IC[12]. However, the sensitivity of EUS-FNA for detecting HGD or IC is suboptimal. As a result, clinical decision making often relies on surrogate clinical and morphological features (Table 1).

Table 1 Clinical and morphological criteria defining features in intraductal papillary mucinous neoplasm management guidelines.

	Clinic			Serology	Cytology	Morphology
	Jaundice	Acute pancreatitis	New-onset diabetes	CA19-9 (≥ 37 U/mL)	HGD/MLG	MPD > 10 mm	MPD 5-9.9 mm	Mural module > 5 mm	Mural module < 5 mm	Solid component	Cyst size	Wall enlargement	Lymphadenopathy	Growth rate
International evidence-based Kyoto guidelines	HRS	WF	WF	WF	HRS	HRS	WF	HRS	WF	HRS	WF (≥ 30 mm)	WF	WF	WF (≥ 2.5 mm/year)
European evidence-based guidelines	AI	RI	RI	RI	AI	AI	RI	AI	RI	AI	RI (≥ 40 mm)	NA	NA	RI (≥ 5 mm/year
AGA	NA	NA	NA	NA	CF	HRF (size not specified)		NA	NA	HRF	HRF (≥ 30 mm)	NA	NA	NA
ACG	HRC	HRC	NA	HRC	HRC	HRC	HRC	HRC	HRC	HRC	HRC (≥ 30 mm)	NA	NA	HRC (≥ 3 mm/year)

CA19-9: Carcinoembryonic antigen 19-9; HGD: High-grade dysplasia; MLG: Moderate-grade dysplasia; MPD: Main pancreatic duct; AGA: American Gastroenterological Association; ACG: American College of Gastroenterology; HRS: High-risk stigmata; AI: Absolute indication; NA: Not available; HRC: High-risk characteristic; WF: Worrisome feature; RI: Relative indication; CF: Concerning features; HRF: High-risk feature.

The actual risk of malignant transformation in IPMN is unknown because data are primarily derived from either surgical series, which carry a significant selection bias, or longitudinally followed cohorts of patients with presumed IPMN. In the latter underestimation of risk is inevitable as a definitive diagnosis of IPMN requires histopathological confirmation. Therefore, a subset of patients in these cohorts may not actually have IPMN.

MD-IPMN has a high risk of malignancy with an estimated risk of 40%-50%, but it has been reported as high as 72%[13]. A retrospective multicenter study found that IPMN-related mortality in patients with MD-IPMN presenting with high-risk stigmata was 43% over a median follow-up of 37 months[14]. Conversely, for BD-IPMN the reported rate of malignant transformation varies widely (6%-46%). Recent observational studies suggest that the true risk lies toward the lower end of this spectrum. In low-risk BD-IPMN (defined by the absence of main pancreatic duct dilation > 5 mm, mural nodules, or HGD/IC on cytology), meta-analyses have consistently found a malignant transformation rate between 0.61% and 0.78% per year[15-17]. This progression appears to be linear, implying a cumulative 10-year risk of 6.1%-7.8%.

These data must be interpreted with caution due to significant heterogeneity and potential publication bias in meta-analyses. Smaller studies tend to report a higher incidence of invasive cancer. Nevertheless, a large prospective single-center Japanese study of 1404 patients (9231 person-years of follow-up) reported similar findings. The risk of pancreatic ductal adenocarcinoma (PDAC) (either IPMN-derived or concomitant) was 6.6% at 10 years and 15.0% at 15 years. Notably, these rates were lower in a subanalysis of patients followed for more than 5 years (3.5% risk at 10 years). Progression of BD-IPMN (i.e. the development of worrisome features or high-risk stigmata) occurs in approximately 20% of cases, and the risk of malignancy in non-low-risk BD-IPMN is nearly 25% at 10 years[15,16].

Risk stratification is guided by international consensus frameworks (Table 1) and guideline-defined indications for resection are expressed in most major documents as high-risk stigmata and worrisome features (or as absolute and relative indications for surgery in European guidelines, respectively). High-risk stigmata that prompts resection in fit patients include obstructive jaundice attributable to the cystic lesion, an enhancing mural nodule (commonly using 5 mm as a practical threshold), markedly dilated main pancreatic duct (commonly ≥ 10 mm), and cytology positive for HGD or IC[18,19]. Worrisome features, such as cyst diameter thresholds (often 30-40 mm), main duct dilation in an intermediate range (e.g., 5.0-9.9 mm), new-onset diabetes, rising serum carbohydrate antigen (CA) 19-9, non-enhancing mural nodules, or rapid cyst growth, indicate a close follow-up of EUS ± FNA rather than immediate resection unless EUS findings or cytology substantiate HGD[8,18,19].

If the lesions are not indicated for resection, they are subjected to continuous surveillance. Surveillance strategies differ between frameworks. The American Gastroenterological Association (AGA) guideline proposed a relatively conservative approach in 2015 and recommended less frequent imaging and discontinuation of surveillance after several years of stability. The European, International (Kyoto), and American College of Gastroenterology guidelines favor prolonged (often lifelong) surveillance for unresected IPMNs reflecting the documented risk of late malignant transformation[8,18,19].

Surveillance intervals are typically stratified by cyst size, growth rate, and presence of worrisome features. For example, it is recommended that small BD-IPMNs without concerning features are imaged at 12-24 month intervals, whereas larger or borderline lesions are imaged at 6-12 month intervals with earlier EUS evaluation if needed[8,18,19]. Selection of an appropriate algorithm should be individualized according to local expertise, patient age and comorbidities, and the patient’s preferences through a shared decision making process.

Undertreatment and overtreatment risk

Although multiple guidelines exist, all result in either missed high-grade lesions or overtreatment. The current IPMN guidelines vary in sensitivity and specificity (Table 2). The International, European, and American College of Gastroenterology (plus most American College of Radiology-based pathways) are highly sensitive but with low specificity. These guidelines recommend surgery for most MD-IPMN and mixed-type IPMN. For BD-IPMN with high-risk stigmata or worrisome features, surgery is also recommended. van Huijgevoort et al[20] observed that both the International Association of Pancreatology (IAP) and European guidelines detected 94%-96% of advanced neoplasia but did so with very low specificity (17%-24%). However, the AGA guideline detected only 27% of advanced neoplasia but spared the majority of low-grade lesions from surgery. A large meta-analysis similarly found pooled sensitivities of 0.67 (Fukuoka) and 0.59 (AGA) with specificities of 0.64 and 0.77, respectively, underscoring that the Fukuoka/IAP and European guidelines prioritize not missing cancer, while the AGA prioritizes avoiding unnecessary surgery[21,22].

Table 2 Real world clinical accuracy and key tradeoffs for different intraductal papillary mucinous neoplasm management guidelines.

Guideline family	Sensitivity	Specificity	Key trade-offs	Ref.
Kyoto 2024 (IAP)	100%	20%	Pros: Almost no missed HGD/IC	[18]
			Cons: Very high overtreatment; many surgeries for low-grade IPMN; increases morbidity and costs
IAP/Fukuoka 2017 (≥ 1 WF or HRS)	90%-98%	16%-45%	Pros: Captures most HGD/IC; strong safety margin	[20]
			Cons: Over-calls risk; many unnecessary resections; psychological burden and repeated EUS/imaging for WF
European Evidence-Based 2018	94%-96%	24%	Pros: Good detection of high-risk disease; harmonized with European practice	[20]
			Cons: Low specificity → high surgical rates and overtreatment risk
AGA 2015	27%-62%	77%-91%	Pros: Few unnecessary surgeries; high specificity avoids morbidity and cost	[20]
			Cons: Substantial risk of missing HGD/IC (up to 45% may be missed in some cohorts); unsafe for high-risk IPMN
ACG 2018	Moderate-high (IAP-like)	Low-moderate	Pros: Good capture of advanced disease; cyst-type-specific guidance; clinically usable algorithms	[19]
			Cons: Still low specificity; many patients undergo EUS-FNA or resection unnecessarily
ACR 2017 (imaging-driven)	High (IAP-like)	Low-moderate	Pros: Easy to apply in imaging-first workflows; high sensitivity	[20]
			Cons: Overdetection leads to frequent EUS and surgeries; less nuanced than GI-society guidelines

IAP: International Association of Pancreatology; HGD/IC: High-grade dysplasia/invasive carcinoma; IPMN: Intraductal papillary mucinous neoplasm; HRS: High-risk stigmata; EUS: Endoscopic ultrasound; WF: Worrisome feature; AGA: American Gastroenterological Association; ACG: American College of Gastroenterology; FNA: Fine-needle aspiration; ACR: American College of Radiology; GI: Gastrointestinal.

A significant proportion of patients who undergo surgical resection for lesions represent cases not classified as advanced neoplasia. In a large multicenter American study involving 106 institutions, 478 pancreatic resections were performed for IPMN over a single year (2014). Notably, HGD or IC were identified in less than one-quarter of the surgical specimens (23%)[23]. Similarly, an international retrospective study of 292 patients reported that HGD or IC were present in only 36.6% of resected cases with an additional 5.5% of patients with IPMN and concomitant PDAC[24].

Overtreatment imposes both clinical and economic burdens. Pancreatic resection carries a substantial risk of mortality and morbidity. Studies have shown that the risk of perioperative mortality is 2%[11,25] while overall morbidity is 30%[11] with non-trivial rates of endocrine and exocrine insufficiency and a measurable negative impact on quality of life. Pancreatectomy is costly. A recent review of pancreatic cancer-related expenditures in Europe estimated mean surgery costs around €30800 per patient (excluding subsequent chemotherapy)[26]. Cost-effectiveness modelling of guideline-based pancreatic cyst surveillance demonstrated that current management strategies are not cost-effective and may increase mortality because of overtreatment of low-grade cysts[27].

The clinical cost of missed high-grade lesions is also substantial. Multiple cohorts demonstrated that patients with HGD undergoing resection have survival outcomes close to those with low-grade IPMN and dramatically better than patients with IC. In a Finnish nationwide series, the 10-year survival was 72.5% for low-grade disease, 66.7% for HGD, and 23.1% for invasive IPMN with pancreatic cancer-specific mortalities of 5.0%, 9.1%, and 71.8%, respectively[28]. Similarly, a South Korean study showed that invasive IPMN had a worse 5-year cumulative recurrence risk than LGD or HGD (LGD vs HGD vs invasive: 0.7% vs 4.3% vs 37.6%, P < 0.001) and a worse 5-year survival rate (89.0% vs 84.0% vs 48.4%, P < 0.001)[29]. Non-invasive IPMN with HGD had significantly better outcomes than IPMN-associated IC with a 5-year survival rate of 70% for HGD compared with 30% for IPMN-related PDAC[30]. These data indicate that accurate identification and resection of HGD IPMN is crucial for favorable clinical outcomes.

BIOMARKERS

Biomarkers are a well-established tool for diagnosis and risk stratification across a wide range of pathologies, particularly cancers. Accordingly, there has been considerable interest in identifying and validating biomarkers that can aid in the diagnosis and stratification of IPMNs.

Diagnostic biomarkers

Currently, the diagnostic biomarkers from cystic fluid or serum are well established. The most common biomarker to differentiate mucinous from non-mucinous pancreatic cysts is carcinoembryonic antigen (CEA) with a cutoff of 192 ng/mL (or 200 ng/mL in some series). Values above this threshold have a moderate sensitivity and a good specificity for mucinous cysts but do not reliably predict HGD or IC[31,32]. A meta-analysis showed a pooled sensitivity of 55%-60% and specificity exceeding 85%, indicating that CEA is most effective when used in combination with imaging and cytology rather than as a standalone marker[33].

In the last decade, intracystic glucose has been used as an alternative to CEA for identifying mucinous cystic lesions. Evidence from two large meta-analyses published in 2021 supports its superior diagnostic performance. McCarty et al[34] demonstrated that glucose levels had a sensitivity of 91% and specificity of 86% in differentiating mucinous from non-mucinous cysts, outperforming CEA. Similarly, Faias et al[35] analyzed only resected cases with histological confirmation of lesion type and reported that a low glucose threshold (< 50 mg/dL) provided a sensitivity of 90%, specificity of 82%, and a higher accuracy compared with CEA[35].

Amylase levels are also analyzed from cystic fluid and are especially useful in distinguishing between IPMN or pseudocysts and MCN or serous cystadenoma. Levels below 250 U/L indicate MCN, levels from 250-25000 U/L indicate typical IPMN, and very high amylase levels (> 10000 U/L) suggest pseudocyst[36,37].

Additionally, mutations in KRAS and GNAS are the most robust molecular biomarkers for identifying the IPMN/mucinous lineage. GNAS codon 201 mutations are strongly associated with IPMN and are relatively accurate with a pooled sensitivity of 44% [95% confidence interval (CI): 34%-54%] and specificity of 100% (95%CI: 99%-100%)[33]. KRAS mutations are also common across mucinous pancreatic precursor lesions, and the presence of KRAS mutations is strongly indicative of IPMN with a pooled sensitivity of 61% (95%CI: 53%-69%) and specificity of 99% (95%CI: 98%-100%) to identify mucinous cysts[33]. Detection of either mutation in cyst fluid greatly increases the confidence of a mucinous IPMN diagnosis.

Stratification and prognostic biomarkers

Prognostic biomarkers are less well established and usually do not provide definitive answers. They are typically used as additional measures. Several are currently used in the clinic while others are under investigation.

CA19-9: CA19-9, also known as Sialyl Lewis-a, is a cell surface glycoprotein complex. Structurally, it is a tetrasaccharide carbohydrate with a transmembrane protein skeleton and extensively glycosylated extracellular oligosaccharide chains. Normally, CA19-9 is produced by ductal cells in the pancreas and biliary system and by epithelial cells in the stomach, colon, uterus, and salivary glands. CA19-9 is overexpressed in PDAC and other malignancies such as biliary, hepatocellular, gastrointestinal, urological, pulmonary, gynecological, thyroid, and salivary gland cancers[38]. Serum levels of CA19-9 are used as a supportive diagnostic tool for PDAC diagnosis with a standardized cutoff value of 37 U/mL[39,40].

In cystic lesions, serum CA19-9 alone is not effective for diagnosis of malignant pancreatic cystic neoplasms. However, levels above 37 U/mL are generally accepted as a strong indicator for malignancy[32]. A meta-analysis of 15 studies showed that serum levels of CA19-9 above 37 U/mL show a pooled sensitivity 40% (95%CI: 36%-44%) and specificity 89% (95%CI: 87%-91%) for discrimination of LGD and HGD/invasive IPMN[41]. Cystic fluid levels of CA19-9 do not generally provide strong stratification efficacy. A study showed sensitivity and specificity of 64% and 68%, respectively[42]; however, that study lacked robust group stratification and compared levels across different cystic pancreatic lesions. The prognostic utility of serum CA19-9 is similarly constrained by heterogeneity in cutoff definitions, retrospective study designs, and limited sensitivity for HGD, with many studies enriched for IC, thereby inflating apparent discriminatory performance.

Cyst fluid cytology: Cyst fluid cytology obtained by EUS-FNA is incorporated into all major IPMN guidelines to confirm HGD or IC. In all guidelines positive cytology is an indication for surgical resection. Positive cyst fluid cytology has high specificity. However, there is a lower sensitivity for distinguishing high-grade IPMN or IC from benign and low-grade IPMN. A meta-analysis from 2019 showed a pooled sensitivity of 57% and specificity of 84%[6] while a newer report showed a sensitivity of 67% and specificity of 94% (but from a small sample size: 24 HGD/IC; 17 benign/LGD)[43]. Notably, both studies compared preoperative cytology with definitive histopathological findings in the resected specimen, implying a potential patient selection bias.

In clinical practice EUS-FNA cytology of cystic lesions is frequently non-diagnostic due to low cellular yield. A prospective study[44] reported adequate cellularity for evaluation in only 44 out of 128 cases. Although EUS-FNA is a safe method with low morbidity and mortality (2.66% and 0.19%, respectively[45]) and positive cytology provides high specificity when malignant cells are identified, its real-world utility is often limited by inadequate sampling and the requirement of an experienced cytopathologist for interpretation[46-48].

To address these limitations EUS-guided through-the-needle biopsy (TTNB) was developed. By utilizing microforceps through a 19-G needle, TTNB directly samples the cyst wall or solid components within the cyst. This approach is useful for lesion characterization and risk stratification[49]. Although TTNB provides a higher diagnostic yield than molecular analysis (71% vs 54%), it was outperformed by genetic testing in diagnostic accuracy for HGD/IC (sensitivity 64% and specificity 81% for TTNB vs 89% and 88% for genetics[50]). Despite its potential, TTNB has not been widely adopted due to the technical expertise required and a risk of adverse events ranging from 8.6% to 11.1%. Additionally, the patchy distribution of dysplasia within IPMN remains a concern as a biopsy showing LGD may not be representative of the entire lesion[51].

Although cyst fluid cytology remains clinically valuable when positive for high-grade atypia or malignancy, its overall role in IPMN risk stratification is limited by low sensitivity, operator- and interpreter-dependent variability, and the fact that most performance data come from surgically selected cohorts rather than broader surveillance populations.

Promising next generation biomarkers

Progressor/high-risk mutations: Progressor/high-risk mutations offer an additional layer for IPMN stratification (Table 3). It is generally well accepted that drive mutations (KRAS and GNAS in the case of IPMN) occur early in tumorigenesis. However, they are not sufficient for full malignant transformation[52,53]. Progressor mutations were identified in tumor suppressor genes such as TP53, SMAD4, CDKN2A, and TGFBR2 as well as in mammalian target of rapamycin pathway genes such as PTEN, PIK3CA, and AKT1[54]. However, there are limited well-designed studies with large cohorts evaluating the stratification accuracy of these mutations.

Table 3 Clinical performance and key limitations of biomarkers relevant for intraductal papillary mucinous neoplasm risk stratification.

Biomarker	Study design/cohort	Source	Endpoint definition	Sensitivity, %	Specificity, %	Key limitations	Ref.
NGS mutation panel (KRAS/GNAS + progressors)	Retrospective	Cyst fluid	Mucinous cyst advanced neoplasia	85	96	Retrospective design; complex NGS workflow; variant interpretation burden	[54]
NGS mutation panel (IPMN-specific)	Prospective	Cyst fluid	Mucinous cyst advanced neoplasia	84	92	Requires NGS infrastructure; not included in later meta-analysis	[54]
TP53 + PI3K pathway mutations	Retrospective	Cyst fluid	IPMN with advanced neoplasia	88	95	Limited cohort size; mutation prevalence dependent	[55]
Das-1 mAb	Retrospective	Tissue	High risk and malignant IPMN	85	95	Early cohort; limited external validation	[58]
Das-1 mAb	Retrospective	Cyst fluid	High risk and malignant IPMN	89	100	Small sample size	[58]
Das-1 mAb	Retrospective	Cyst fluid	High risk and malignant PCL	88	99	Included MCNs as well as IPMNs	[59]
Das-1 mAb (EV-based)	Case-control	Serum EVs	LGD vs HGD/invasive IPMN	82	40	Markedly reduced specificity in blood	[60]
Das-1 mAb (EV-based, HGD only)	Case-control	Serum EVs	LGD vs HGD IPMN only	100	30	Poor specificity; limited clinical utility	[60]
IL-1β	Retrospective	Cyst fluid	Low risk vs high risk lesions	79	95	Potential confounding by inflammation	[65]
IL-1β	Larger retrospective cohort	Cyst fluid	Low/moderate IMPN vs HGD/invasive IPMN	64.3	83.8	Lower sensitivity at higher cutoff	[66]
IL-1β + CA19-9	Retrospective	Cyst fluid + serum	Low/moderate IMPN vs HGD/invasive IPMN	27.5	100	Sensitivity too low for screening	[66]
PGE2	Retrospective	Cyst fluid	LGD vs HGD/invasive IPMN	63	79	Weak stand-alone performance	[69]
PGE2 + CEA	Retrospective	Cyst fluid	LGD vs HGD/invasive IPMN	78	100	Requires multi-marker testing	[69]
PGE2	Larger cohort	Cyst fluid	Low/moderate IMPN vs HGD/invasive IPMN	60.0	78.7	Modest predictive value	[66]
PGE2 + IL-1β	Retrospective	Cyst fluid	Low/moderate IMPN vs HGD/invasive IPMN	42.9	89.0	Limited incremental benefit	[66]
MUC5AC (EV-based)	Case-control	Serum EVs	LGD vs HGD/invasive IPMN	45	100	Single study; low sensitivity	[60]
MUC5AC (EV-based, HGD only)	Case-control	Serum EVs	LGD vs HGD/invasive IPMN	9	100	Extremely poor sensitivity	[60]

NGS: Next generation sequencing; IPMN: Intraductal papillary mucinous neoplasm; PI3K: Phosphoinositide 3-kinase; mAb: Monoclonal antibody; PCL: Plasma cell leukemia; MCN: Mucinous cystic neoplasm; EV: Extracellular vesicles; LGD: Low-grade dysplasia; HGD: High-grade dysplasia; IL: Interleukin; CA19-9: Carcinoembryonic antigen 19-9; PGE₂: Prostaglandin E2; CEA: Carcinoembryonic antigen; MUC5AC: Mucin 5AC.

The most thorough analysis used a 22-gene next generation sequencing panel (PanreaSeq) to evaluate the presence of various mutations in cystic lesions from 97 patients in a retrospective study and from 1889 patients in a prospective study. The presence of KRAS/NRAS/HRAS/BRAF/GNAS mutations along with mutations in TP53, SMAD4, PTEN, PIK3CA, or AKT1 showed 86% sensitivity and 96% specificity for identifying mucinous cyst with advanced neoplasia in the retrospective study. Moreover, the panel of the same mutations showed 84% sensitivity and 92% specificity for the prospective study[54]. Additionally, TP53 mutations concurrent with alterations in the phosphoinositide 3-kinase pathway (PIK3CA or PTEN mutations) had a sensitivity of 88% and a specificity of 95% for the identification of IPMNs with advanced neoplasia[55].

In a recent meta-analysis mutations in CDKN2A, PIK3CA, SMAD4, and TP53 showed excellent specificity (95%-98%) for the identification of advanced IPMN (HGD or PDAC). However, individual sensitivity ranged from 9% to 42%[33]. The aforementioned prospective study was not included in this meta-analysis because it was published after the inclusion deadline.

Despite promising results, interpretation of progressor mutation panels is limited by frequent use of pooled HGD/invasive endpoints, which may overestimate performance for detecting pre-invasive disease, and by reliance on surgically enriched cohorts that reduce generalizability to surveillance populations. In addition, while specificity is high, low sensitivity and variability in sequencing platforms and reporting pipelines constrain their standalone clinical utility, and prospective multicenter validation remains limited. Additionally, such panels require extensive workflows and accurate variant calling and reporting.

Das-1 monoclonal antibody: A monoclonal antibody (mAb), Das-1, was developed almost 40 years ago against a colonic epithelial protein and was shown to specifically recognize colonic epithelium without recognizing any other tissue type in the gastrointestinal tract[56]. However, it does strongly stain metaplastic and malignant tissues arising from the esophagus, stomach, and pancreas[57]. IPMN was one of the tumor types in which this antibody was extensively studied. Initially, the antibody appeared to recognize a 40 kDa colon-specific tropomyosin isoform or a tropomyosin-related protein; However, the actual antigen is 3’-sulfated LewisA/C antigen, chemically 3’-O-sulfated β-D-galactopyranosyl-(1→3)-N-acetyl-D-glucosamine(3’-Sulfo-Galβ(13GlcNAc)(34910746).

In IPMN, initial reports showed very good performance for distinguishing HGD and malignant IPMN from LGD with a sensitivity of 85% (95%CI: 73%-93%) and specificity of 95% (95%CI: 83%-99%). Interestingly, the accuracy of the Das-1 antibody to detect HGD and invasive IPMN was even higher when analyzing cystic fluid samples, reaching a sensitivity of 89% (95%CI: 65%-99%) and specificity of 100% (95%CI: 66%-100%)[58]. Consistent results were observed when the Das-1 antibody was used in a validation cohort (sensitivity of 88% and specificity of 99%). However, this follow-up study had one subtle change from the original study: High-grade pancreatic cystic neoplasms containing samples from MCNs as well as IPMNs were used[59].

Another study evaluated the performance of the Das-1 antibody, analyzing biomarkers in EV secreted into the bloodstream. Here the performance of Das-1 for distinguishing LGD and HGD/invasive IPMN reached 82% sensitivity and 40% specificity[60]. However, detailed analysis of low-grade and high-grade IPMN, excluding invasive IPMN, showed a sensitivity of 100% and a specificity of 30%[60].

While Das-1-based assays have demonstrated high sensitivity for advanced neoplasia, their reported performance is frequently influenced by studies that combine HGD with IC, as well as by reliance on surgically enriched cohorts and limited multicenter validation, raising concerns about their ability to specifically detect pre-invasive high-grade disease.

Interleukin 1β: Interleukin-1 beta (IL-1β) is another promising biomarker for IPMN stratification. IL-1β plays a protumorigenic role in pancreatic cancer through enhancing nuclear factor kappa B-mediated inflammation, promoting KRAS-driven tumorigenesis, and stimulating pancreatic stellate cells to secrete large amount of extracellular matrix[61,62]. Although no mechanistic studies have evaluated the role of IL-1β in IPMN, it is reasonable to expect a similar protumorigenic role through amplifying KRAS and GNAS-driven signaling and tumor microenvironment signaling[63,64].

Due to its paracrine/autocrine signaling nature, IL-1β levels were analyzed in IPMN cyst fluid. IL-1β is one of the most significantly upregulated cytokines in cystic fluid of high-grade lesions in patients with IPMN. A cutoff value of 1.26 pg/mL showed 95% specificity and 79% sensitivity for the identification of high-risk IPMN (high-grade or invasive)[65]. A larger follow-up study using a cutoff value of 20 pg/mL yielded 83.8% specificity and 64.3% sensitivity for predicting high-risk IPMN[66]. The combination of IL-1β and CA19-9 yielded 100% specificity but only 27.5% sensitivity.

Although IL-1β has shown potential in identifying high-risk cysts, its reported performance is derived largely from small, single-center surgical cohorts with variable assay platforms and limited validation in surveillance populations, raising concerns about generalizability.

Prostaglandin E₂: Prostaglandin E₂ (PGE₂) is tightly interconnected with IL-1β and follows a similar trend (elevated PGE₂ levels are associated with high-grade IPMN). IL-1β stimulates PGE₂ synthesis through several mechanisms such as cyclooxygenase 2 upregulation, PGE synthase upregulation, and phospholipase A2 activation[67,68]. Several studies showed upregulation of PGE₂ in HGD/invasive IPMN[69,70]. However, PGE₂ was not a robust marker on its own. A threshold of 1.1 pg/μL yielded a sensitivity of 63% and a specificity of 79% for the detection of HGD/invasive IPMN. When combined with CEA > 192 ng/mL, the sensitivity and specificity were 78% and 100%, respectively, for the detection of HGD/invasive IPMN[69].

In a follow-up study, the performance of PGE₂ with IL-1β and CA19-9 was analyzed in a larger cohort of patients. PGE₂ alone demonstrated comparable sensitivity (60.0%) and specificity (78.7%) for the identification of HGD/invasive IPMN. The combination of PGE₂ and IL-1β yielded a lower sensitivity (42.9%) and higher specificity (89.0%)[66]. Interestingly, PGE₂ synthase, IL-1β and MUC4 mRNA levels could accurately detect HGD/invasive IPMN[71]. Interpretation of PGE2 as a stratification biomarker is further complicated by assay variability, small cohort sizes, and frequent reliance on pooled HGD/invasive endpoints, which may obscure its true performance in detecting pre-invasive disease.

MUC5AC: MUC5AC is a secreted, gel-forming glycoprotein normally expressed on the surfaces of respiratory and gastrointestinal tract epithelial tissue. It is significantly overexpressed in various malignancies in which it is associated with an advanced stage and metastatic spread[72]. MUC5AC is abundantly expressed in IPMN and could be used for risk stratification[73,74]. However, only one study has analyzed the expression of MUC5AC as a stratification marker in IPMN. EVs from the serum of patients with HGD/invasive IPMN had significantly increased levels of MUC5AC compared with patients with LGD. Additionally, MUC5AC-positive EVs showed 100% specificity and 45% sensitivity for identification of HGD/invasive IPMN. However, deconvolution of HGD and invasive IPMN showed that MUC5AC-positive) detected HGD IPMN with 100% specificity, but the sensitivity was only 9%[60].

Although MUC5AC expression is biologically linked to mucinous differentiation and has shown potential in risk stratification, its clinical utility is limited by heterogeneous assay methodologies, inconsistent threshold definitions, and a predominance of retrospective surgical cohorts without standardized separation of LGD, HGD, and IC.

ANALYSIS AND REPORTING LIMITATIONS

Pooled analysis of LGD IPMN and HGD/invasive IPMN

Most studies have conducted a pooled analysis with two groups: Low risk (usually LGD and sometimes with intermediate grade); and high risk (pooled HGD and IC). These groups are beneficial for meta-analysis, power analysis, and overall accuracy metrics. However, these groups blur biologically and clinically important distinctions. HGD and IC are not the same disease. Patients with HGD who undergo resection have similar survival outcomes as patients with LGD and significantly better outcomes than patients with invasive IPMN, indicating the benefit for detecting IPMN before progression to the invasive stage. When HGD and ICs are pooled, the contribution of the biomarker performance is much greater from the IC samples than from the HGD samples. For example, the reported predictive performance of MUC5AC and the Das-1 antibody detected in serum EVs differed in the pooled and separated analyses. In the pooled analyses comparing LGD with HGD/IC, MUC5AC demonstrated a sensitivity of 45% and a specificity of 100%, whereas the Das-1 mAb showed a sensitivity of 82% and specificity of 40%. However, when the analysis separated HGD and IC and compared LGD and HGD alone, the MUC5AC sensitivity fell to 9% (with specificity retained at 100%) and the Das-1 mAb achieved 100% sensitivity but only 30% specificity[60]. To identify true high-accuracy biomarkers, it is crucial to analyze LGD, HGD, and IC separately.

Terminological inconsistency in histological endpoints

The terminology used to describe IPMN grading is highly inconsistent in the literature with several key terms used interchangeably despite representing biologically and clinically distinct entities. HGD is sometimes labeled as carcinoma in situ; malignant IPMN can be defined as encompassing both HGD and IPMN-associated PDAC or restricted to IC alone. Advanced neoplasia is inconsistently applied to HGD only, to HGD combined with invasive IPMN, or HGD with concomitant PDAC. The term benign is frequently used to describe low-grade IPMN despite the premalignant nature of the lesion. Moreover, multiple reports have labeled both HGD and IC as malignancy, amplifying transformation rates and mixing the biology and clinical outcomes. The inconsistent terminology leads to unclear clinical outcomes.

Sample inconsistencies

Inconsistency in sample processing and analysis affects both serum and cyst fluid-based studies. Fluid characteristics, including volume, viscosity, cellularity, mucin content, and blood contamination, differ substantially between cysts and across centers. Sampling and pre-analytical procedures vary widely with studies employing different needle types, aspiration techniques, minimum volume thresholds, and handling protocols as well as inconsistent time to freezing, number of freeze-thaw cycles, and the use of preservatives. Analytical assays are also conducted using non-standardized platforms ranging from custom laboratory-developed enzyme-linked immunosorbent assays to diverse commercial kits and next-generation sequencing panels. Each of these methods have distinct limits of detection and cutoff definitions.

Lack of multicenter validation and small cohorts

Another limitation of the current biomarker literature is the lack of robust multicenter validation. Most reports originate from single institutions that are typically high-volume academic centers with specialized EUS operators, optimized cyst fluid handling protocols, and in-house molecular assays. As a result biomarker performance partially reflects center-specific expertise, technical workflows, and patient selection patterns rather than generalizable biological signal. Without external validation the reported sensitivity/specificity estimates may not be relevant in routine practice in which cyst fluid quality, assay platforms, and clinical decision making vary substantially.

This challenge is compounded by modest sample sizes. When total cohorts are large, the number of patients with HGD or IC is often small, especially in prospective surveillance cohorts, leading to wide CIs, unstable cutoff selection, and a high risk of overfitting. Small surgical cohorts are also prone to verification bias because only lesions deemed suspicious by guideline criteria undergo resection, contributing to histological endpoints. The result is an inflated impression of biomarker accuracy in high-risk, preselected populations and limited insight into performance for the much larger pool of indolent BD-IPMNs.

FUTURE RESEARCH ROADMAP

To advance IPMN biomarker development from exploratory studies to clinically actionable tools, future research must adopt a more standardized and translationally oriented framework. First, prospective multicenter cohort studies with adequate representation of both surgical and surveillance populations are essential to ensure generalizability and reduce selection bias. These studies should incorporate harmonized histopathologic endpoint definitions, with consistent separation of LGD, HGD, and IC.

Second, standardization of biospecimen collection and processing is critical. Pre-analytical variables - including cyst fluid acquisition technique, sample volume, handling time, storage conditions, and assay platforms - should be uniformly reported and controlled to enable reproducibility across centers.

Third, future studies should prioritize head-to-head comparison of candidate biomarkers and multimarker panels, rather than isolated evaluation, to determine incremental value over existing clinical, imaging, and cytologic criteria. This includes direct comparison within the same patient cohorts using prespecified thresholds and validation datasets.

Finally, biomarker development should extend beyond diagnostic accuracy to include clinical utility and implementation endpoints, such as impact on surgical decision making, surveillance intervals, reduction of unnecessary resections, and cost-effectiveness. Integration with existing clinical workflows, particularly magnetic resonance imaging and EUS-FNA-based evaluation, should be explicitly addressed to facilitate guideline incorporation and real-world adoption.

Text box: Proposed reporting standard for future IPMN risk stratification biomarker studies. Prespecified clinical question: Diagnosis of mucinous lineage, detection of HGD, detection of IC, or surveillance escalation. Standardized endpoint definitions: LGD, HGD, and IC reported separately; pooled endpoints justified explicitly. Representative cohorts: Surgical and surveillance populations both included whenever feasible, with clear recruitment pathways. Transparent pre-analytics: Needle type, fluid volume, blood contamination, storage conditions, freeze-thaw cycles, and assay platform. Prespecified cutoff/model: Avoid post hoc threshold optimization without validation. External validation: Preferably multicenter, with inter-laboratory reproducibility. Incremental value analysis: Performance beyond guideline criteria, imaging, cytology, and CA19-9. Clinical utility endpoints: Effect on resection decisions, surveillance intervals, missed HGD, and unnecessary surgery. Health-economic evaluation: Cost, workflow complexity, and feasibility in non-tertiary settings.

CONCLUSION

Despite technological progress in molecular profiling, proteomics, and cyst fluid analytics, no existing biomarkers or clinical guidelines reliably address the central challenge in IPMN management: Identifying HGD before progression to IC while minimizing unnecessary surgery. The optimistic performance reported for many emerging biomarkers is constrained by several methodological limitations, including the pooling of HGD with IC endpoints, terminological inconsistency across studies, substantial pre-analytical and assay-related variability, and the absence of multicenter, prospectively collected, adequately powered cohorts.

Furthermore, translation into clinical practice faces significant barriers. Many assays require specialized platforms and high technical expertise. Most are also associated with significant cost, yet almost no studies have incorporated health-economic modeling to justify adoption. The limited number of HGD events in surveillance cohorts leads to unstable performance metrics and contributes to persistent verification bias, yielding a limited number of biomarkers to supplement risk stratification guidelines.

Meaningful progress will require the harmonization of histological endpoints, standardized sample handling, and prospective multicenter validation with adequate representation of BD-IPMN and surveillance populations. Longitudinal studies with hard outcomes, biologically informed marker selection, and integration of biomarker development with cost-effectiveness and decision-analytic modeling should guide future investigations.