Balancing malignancy risk and healthcare resources: Optimizing surveillance for branch-duct intraductal papillary mucinous neoplasm

Abstract

The rising use of cross-sectional imaging has led to increased detection of branch-duct intraductal papillary mucinous neoplasms (BD-IPMNs), posing a clinical challenge in balancing cancer prevention against the risks and costs of long-term surveillance. A recent multicenter study by Kayali et al published in World Journal of Gastroenterology, proposed data-driven cyst size thresholds (1.5 cm and 3 cm) and suggested that the presence of at least two worrisome features may enhance the prediction of malignant progression. While these evidence-based refinements offer a pragmatic step toward cost-effective care, they also highlight the inherent limitations of morphology-based surveillance. Emerging approaches including radiomics, artificial intelligence-assisted imaging analysis, and circulating biomarkers (e.g., cell-free DNA or cyst fluid molecular profiling) hold potential to improve individualized risk assessment beyond conventional imaging criteria. Moreover, the growing prevalence of incidental pancreatic cysts in aging populations necessitates the integration of life expectancy, comorbidities, and patient preferences into surveillance planning. Finally, international variability in healthcare resources calls for validation of surveillance strategies across diverse systems. Together, these considerations highlight the need to advance from size-based criteria toward multidimensional, precision-based risk stratification in BD-IPMN management.

Key Words: Branch-duct intraductal papillary mucinous neoplasm; Pancreatic cystic neoplasms; Risk stratification; Imaging surveillance; Radiomics; Molecular biomarkers

Core Tip: Recent data-driven analyses offer new insights into the surveillance of branch-duct intraductal papillary mucinous neoplasm. A multicenter study highlights refined cyst size thresholds and the value of multiple worrisome features in predicting malignancy. However, morphology-based surveillance alone cannot fully assess biological heterogeneity. Emerging approaches, including radiomics, artificial intelligence-based imaging, and fluid or blood biomarkers, may supplement current imaging criteria. Integrating these advances with patient-specific factors could enable more precise and individualized management in the future.

This editorial refers to “Optimizing branch-duct intraductal papillary mucinous neoplasms surveillance: Data-driven dimensional grouping for risk stratification and cost-effectiveness” by Kayali et al, 2026; https://doi.org/10.3748/wjg.v32.i20.114867.

---

INTRODUCTION

The recent multicenter cohort study by Kayali et al[1], published in the World Journal of Gastroenterology, provides a timely evaluation of risk stratification and cost-effectiveness in the surveillance of branch-duct intraductal papillary mucinous neoplasms (BD-IPMNs). This study addresses a key clinical challenge arising from the increased detection of incidental pancreatic cysts on cross-sectional imaging, which complicates the formulation of optimal surveillance strategies[2].

BD-IPMN is among the most commonly diagnosed pancreatic cystic neoplasms. While typically slow-growing, its potential to progress to invasive pancreatic cancer warrants careful clinical attention[3]. Given the poor prognosis of pancreatic ductal adenocarcinoma, effective surveillance aimed at the early identification of high-risk lesions is paramount in BD-IPMN management[4,5].

Current international guidelines rely heavily on imaging-based morphological features, like cyst size, mural nodules, and main pancreatic duct dilation, for risk assessment[6-8]. However, many diagnostic thresholds are derived from expert consensus rather than prospective data, resulting in inconsistencies across guidelines[9]. This evidence gap contributes to ongoing uncertainty in balancing early cancer detection against the risks and costs of prolonged surveillance.

Against this background, the study by Kayali et al[1] offers valuable evidence to refine risk stratification and enhance surveillance efficiency. This editorial examines the clinical implications of their findings, critiques the limitations of morphology-based surveillance, and considers the potential of emerging tools such as radiomics and molecular biomarkers in advancing toward more precise and cost-effective management of BD-IPMN. To illustrate this conceptual evolution in surveillance strategy, Figure 1 schematically outlines the transition from conventional morphology-based assessment toward a multidimensional precision approach.

Figure 1 Conceptual evolution from morphology-based surveillance to multidimensional precision surveillance in branch-duct intraductal papillary mucinous neoplasms. The figure illustrates a stepwise transition from conventional morphology-based assessment to an integrated, precision surveillance framework. In the left panel, conventional surveillance relies on cross-sectional imaging modalities, including computed tomography and magnetic resonance imaging, where risk stratification is primarily based on static morphological features such as cyst size, main pancreatic duct dilation, and mural nodules. These features are typically interpreted using predefined threshold criteria (e.g., 1.5 cm to 3 cm), with the accumulation of multiple worrisome features associated with increased malignancy risk. However, as shown in the middle panel, this morphology-based approach has important limitations. Tumor biological heterogeneity cannot be adequately captured by imaging alone, leading to radiologic-pathologic discordance. In addition, reliance on static snapshots fails to reflect the dynamic nature of tumor evolution over time. Collectively, these limitations restrict accurate characterization of tumor biology and risk. In contrast, the right panel presents a multidimensional precision surveillance paradigm that integrates diverse data sources, including radiomics, molecular biomarkers (e.g., circulating tumor DNA), artificial intelligence-based imaging analysis, and patient-specific clinical factors. Through multimodal data integration, this approach enables more robust risk prediction and supports individualized surveillance and management strategies. The bottom schematic depicts the continuum of disease progression from normal pancreas to branch-duct intraductal papillary mucinous neoplasms, dysplasia, and invasive pancreatic cancer. It highlights how the incorporation of molecular and artificial intelligence-assisted tools may facilitate earlier detection and intervention within this progression spectrum. CT: Computed tomography; MRI: Magnetic resonance imaging; ctDNA: Circulating tumor DNA; AI: Artificial intelligence; BD-IPMN: Branch-duct intraductal papillary mucinous neoplasm.

DATA-DRIVEN REFINEMENT OF MORPHOLOGICAL RISK STRATIFICATION

Current surveillance strategies for BD-IPMN are predominantly predicated on imaging-derived morphological criteria[6,10]. International guidelines stratify malignancy risk using features such as cyst diameter, mural nodules, main pancreatic duct caliber, and other so-called worrisome features[6-8]. However, many of these thresholds originate from expert consensus rather than large-scale prospective datasets, contributing to variability across guidelines regarding optimal risk stratification criteria[11].

The recent multicenter study by Kayali et al[1] provides empirical support for refining this framework by identifying two data-driven size thresholds approximately 1.5 cm and 3.0 cm that delineate risk categories with distinct clinical implications. While the 3.0 cm threshold aligns with the “worrisome feature” definition in the Fukuoka guidelines, the novel 1.5 cm cutoff provides additional granularity for early risk stratification, pending further validation in diverse populations. This lower threshold may improve risk sensitivity but could also lead to over-surveillance in low-risk populations. These thresholds serve as a pragmatic bridge between existing protocols such as the American Gastroenterological Association’s multi-feature approach and European clinical-radiologic criteria and evidence-based practice[6-8]. Notably, the study emphasizes that size alone should not be an absolute criterion; the coexistence of multiple worrisome features often confers greater predictive value than isolated size abnormalities[12].

Despite their clinical appeal, the debate over 1.5 cm and 3.0 cm thresholds persists due to the inherent radiologic-biologic discordance of BD-IPMN. Significant overlap in malignant potential across size categories suggests that diameter alone cannot fully discriminate underlying biological behavior[10,12]. Reliance on rigid cutoffs risks oversimplifying a heterogeneous disease where small cysts may harbor aggressive molecular alterations, while larger lesions may remain indolent for years[12,13].

Overall, this study represents an incremental step toward evidence-based refinement rather than a paradigm shift. As the detection of incidental pancreatic cysts rises, particularly in aging populations, surveillance strategies must balance early malignancy detection against the risks of unnecessary imaging, invasive procedures, and escalating healthcare costs[9,14,15].

Several limitations should be noted. The retrospective, multicenter design may introduce selection bias and protocol heterogeneity. The generalizability of the proposed thresholds to broader populations and their integration into existing guidelines require external validation. Furthermore, whether these refined thresholds improve patient-centered outcomes, rather than merely statistical discrimination, remains uncertain. These considerations suggest that while refined morphological criteria represent progress, imaging features alone may be insufficient to capture the full biological heterogeneity of BD-IPMN.

Cysts with similar radiologic appearances may exhibit markedly different clinical courses, indicating that radiology-derived features do not always reflect underlying neoplastic behavior[16]. Integrating biomarkers, advanced imaging analytics, and molecular profiling into surveillance algorithms may therefore improve risk discrimination and support personalized management[17,18].

LIMITATIONS OF MORPHOLOGY-BASED SURVEILLANCE

Morphology-based risk stratification remains the cornerstone of BD-IPMN surveillance. However, increasing evidence highlights a radiologic-biologic discordance wherein imaging often fails to capture the molecular diversity and clinical trajectory of these lesions[13,16]. Current surveillance frameworks rely on features such as cyst diameter, mural nodules, and main duct dilatation to estimate malignancy risk[19]. Yet, lesions with similar radiologic appearances can follow different clinical courses, and isolated imaging abnormalities often lack sufficient specificity for malignant progression[20]. This limitation complicates precise risk identification and underscores the need for complementary tools that reflect underlying tumor biology.

Biological heterogeneity of pancreatic cystic neoplasms

A key constraint of morphology-based surveillance is the substantial biological heterogeneity among IPMNs[21]. Radiologically similar lesions may harbor distinct molecular profiles and clinical courses. While many BD-IPMNs remain indolent for years, others may progress to high-grade dysplasia or invasive carcinoma despite relatively modest imaging abnormalities.

Molecular studies have identified recurrent driver alterations most notably in KRAS, GNAS, TP53, and SMAD4 that correlate with neoplastic progression[13]. These genetic alterations may precede overt morphologic change, suggesting that structural imaging criteria alone may underestimate the aggressiveness of certain lesions.

Furthermore, conventional imaging parameters such as cyst diameter or mural nodules are indirect surrogates for tumor biology[13,16]. Although larger cysts generally confer higher malignancy risk, size alone lacks the precision to reliably distinguish indolent from aggressive phenotypes. Relying solely on morphology may lead to unnecessary surveillance or delayed diagnosis. Integrating molecular markers with imaging findings could refine risk stratification and guide clinical decisions more precisely[22].

Dynamic evolution of BD-IPMN

A further limitation pertains to the dynamic natural history of BD-IPMN. These cysts often evolve slowly over many years[12]. During surveillance, cysts may gradually enlarge, develop new imaging features, or remain stable for prolonged periods. As a result, risk estimates based on a single time point may not accurately reflect long-term malignant potential.

Longitudinal studies suggest that temporal changes during follow-up such as accelerated growth, emergence of new worrisome features, or progressive duct dilatation, may provide stronger prognostic information than baseline morphology alone[4]. However, many current guidelines rely on static thresholds that inadequately account for this temporal dimension[6-8].

These observations emphasize the need to augment conventional imaging assessment with integrated approaches that capture longitudinal trends, molecular profiles, and clinical variables[17,22]. Such multidimensional strategies hold promise for improving risk discrimination while reducing unnecessary surveillance and invasive procedures.

EMERGING APPROACHES FOR PRECISION RISK PREDICTION

Given the inherent constraints of morphology-based surveillance, increasing attention has focused on novel technologies capable of refining risk prediction for BD-IPMN[17,19,23]. Advances in quantitative imaging analysis, computational modeling, and molecular diagnostics now offer opportunities to assess biological information beyond conventional radiologic assessment.

Radiomics and artificial intelligence-assisted imaging

Radiomics enables extraction of high-dimensional, sub-visual features from routine medical imaging[19]. Unlike qualitative radiological assessment, it quantifies textural, morphological, and intensity-based patterns that may reflect underlying tumor biology. Several studies have demonstrated that radiomic signatures derived from computed tomography or magnetic resonance imaging (MRI) can help distinguish benign cysts from those with high-grade dysplasia or invasive carcinoma[23].

Artificial intelligence (AI), particularly machine and deep learning, further augments risk stratification capabilities in pancreatic cyst evaluation. Different AI approaches show promise for specific clinical tasks. Deep learning convolutional neural networks excel at analyzing complex imaging patterns, demonstrating high accuracy in characterizing subtle mural nodules and septations that may be overlooked by visual assessment[24]. In contrast, traditional machine learning models (e.g., random forests, support vector machines) are effective for integrating heterogeneous data sources combining radiomic features with clinical parameters (e.g., cyst size, patient age) and molecular biomarkers to generate comprehensive risk stratification scores[23,25].

A notable example is the study by Cheng et al[25], which developed a machine learning-based radiomics model integrating MRI and multi-omics data, achieving high diagnostic performance (area under the curve > 0.90) for predicting malignancy in pancreatic cystic lesions, outperforming conventional clinical and radiological assessment. Collectively, early evidence indicates that AI-assisted diagnostics may offer incremental value over conventional assessment in classifying pancreatic cysts and predicting malignancy.

Technical heterogeneity and challenges for clinical translation

Nonetheless, the current enthusiasm for radiomics and AI must be tempered by a critical assessment of the significant technical hurdles that impede their clinical application. A primary barrier is the substantial technical heterogeneity across studies and institutions, which poses a major threat to the reproducibility and generalizability of these models. This heterogeneity arises from multiple sources: Variability in image acquisition parameters (e.g., slice thickness, contrast timing) across different scanners, differences among scanner vendors and imaging protocols, inconsistent image preprocessing and segmentation methods, and the absence of standardized feature definitions and robust cross-platform harmonization techniques[26].

Other obstacles hinder clinical implementation. Insufficient external validation remains a significant concern, as most models are based on retrospective, single-center cohorts with restricted generalizability. Cost-effectiveness and resource limitations present additional challenges, particularly in healthcare systems with restricted access to advanced imaging analytics, molecular testing, and computational infrastructure. Regulatory and workflow integration barriers, such as the lack of standardized reporting systems, ambiguous clinical decision thresholds, and limited interpretability of many AI models, hinder their incorporation into standard clinical workflows[27,28].

Together, these factors increase the risk of model overfitting to specific local datasets and reduce performance in real-world, multi-center settings. Therefore, while promising, most current studies remain exploratory, limited by retrospective design, small sample sizes, and lack of external validation[23]. Future efforts must prioritize standardized imaging protocols, advanced harmonization methods (e.g., ComBat), and large-scale prospective multicenter validation to ensure the reliability and clinical utility of AI-driven models for BD-IPMN risk stratification.

Circulating and cyst fluid biomarkers

Beyond imaging, molecular biomarkers from cyst fluid or blood are being actively investigated for risk stratification[29]. Cyst fluid genetic alterations particularly in KRAS, GNAS, TP53, and SMAD4 have been correlated with IPMN progression and may assist in distinguishing indolent from high-risk lesions[30]. Emerging evidence supports integrating genomic and epigenomic markers for improved precision. Next-generation sequencing panels now enable concurrent assessment of key mutations alongside epigenetic alterations like DNA methylation patterns in cyst fluid[31]. This combined genomic-epigenomic approach enhances diagnostic accuracy by capturing complementary aspects of tumor biology, with certain methylation signatures strongly associated with high-grade dysplasia.

However, cyst fluid acquisition via endoscopic ultrasound-guided fine-needle aspiration (EUS-FNA) is invasive and carries risks such as pancreatitis, bleeding, and infection particularly challenging for small cysts (< 1.5 cm), where risks may outweigh benefits.

Circulating biomarkers, including circulating tumor DNA (ctDNA) and microRNAs, offer a non-invasive alternative for monitoring disease biology[17]. However, their utility in early-stage BD-IPMN is limited by low sensitivity, as premalignant lesions shed minimal ctDNA, often below current detection thresholds[32]. Additional hurdles include pre-analytical variability, lack of assay standardization, cost, and limited accessibility, which currently restrict widespread adoption.

Advancing biomarker strategies to overcome limitations

Despite these limitations, ongoing research is developing strategies to mitigate these “pain points”. In cyst fluid analysis, the field is shifting from single gene assays toward integrated multi-omics profiling (e.g., combining mutational and epigenetic markers)[31], maximizing diagnostic yield per procedure and potentially justifying invasive sampling in select cases.

For liquid biopsies, strategies focus on improving sensitivity to detect low-abundance analytes. Epigenetic analyses such as DNA methylation profiling are promising due to greater abundance and stability in blood than somatic mutations[33]. Multi-omics approaches that integrate genomic, epigenomic, and proteomic data within machine-learning frameworks are also being developed to better distinguish low-risk from high-risk lesions[25,31,34].

These advances reflect a broader shift toward multidimensional risk assessment, combining molecular data with imaging and clinical progression to guide individualized management.

PATIENT-CENTERED AND SYSTEM-LEVEL CONSIDERATIONS IN SURVEILLANCE

Optimizing BD-IPMN surveillance must incorporate patient-specific factors and healthcare system realities. With rising detection of incidental cysts, surveillance should extend beyond morphology to include clinical context, life expectancy, and patient preferences[9,10].

Aging populations and incidental pancreatic cysts

Global aging has increased incidental cyst detection in older adults, many with comorbidities or limited life expectancy[2,10]. Since many small BD-IPMNs remain indolent[12], long-term surveillance benefits in frail or elderly patients must be balanced against risks from repeated imaging and invasive procedures[9]. Surveillance intensity should be tailored based on age, comorbidities, functional status, surgical candidacy, and patient choice.

For example, in patients ≥ 80 years with significant comorbidities, competing mortality risks may outweigh surveillance benefits, and discontinuation may be appropriate when life expectancy is limited or surgery is not feasible. This is particularly relevant when patients are unlikely to benefit from early cancer detection due to limited survival or poor operative tolerance. Conversely, younger, healthier patients may benefit from continued monitoring even with low-risk features[10].

Variability in global healthcare resources

Disparities in healthcare resources significantly affect the applicability of surveillance guidelines[35,36]. While advanced tools like molecular profiling and radiomics improve accuracy, they require cost, infrastructure, and expertise that may be unavailable in resource-limited settings[9,22,37]. Patient decisions are often influenced by financial burden, access to specialized care, and willingness to undergo invasive procedures (e.g., EUS-FNA).

A structured shared decision-making process is essential, incorporating medical risks, patient values, and real-world constraints[38]. Effective clinician-patient communication should convey the probabilistic malignancy risk in a clear and understandable manner and discuss surveillance benefits, limitations, and costs, using simplified risk categories and personalized plans to align with patient values and healthcare realities[39]. Such approaches may improve patient adherence and ensure that surveillance strategies remain both feasible and acceptable in diverse clinical settings. Future strategies must balance precision with affordability and accessibility.

Clinical implications

BD-IPMN management should integrate cyst features, longitudinal changes, patient age, comorbidities, and surgical risk[8]. Individualized approaches may reduce unnecessary monitoring in low-risk patients while focusing resources on high-risk cases. As radiomic and biomarker tools evolve, their judicious incorporation can further improve surveillance precision and efficiency[15,40].

CONCLUSION

The escalating detection of BD-IPMNs underscores the need for surveillance that balances early cancer detection with efficient resource use. Recent evidence, including data-driven size thresholds and the significance of multiple worrisome features, refines current risk stratification. Still, morphology-based approaches remain limited in capturing biological heterogeneity and lesion dynamics. Clinical decisions must also consider patient factors such as age and comorbidities and acknowledge healthcare resource disparities. Future strategies should adopt a multidimensional framework, integrating conventional imaging with emerging tools like radiomics, AI, and molecular biomarkers. When prospectively validated across diverse populations and settings, such integrated approaches can enable more precise, individualized, and sustainable BD-IPMN management.