Toward Timely Diagnosis of Pancreatic Cancer: Revolutionizing Early Detection Through Genomics, Artificial Intelligence, and Noninvasive Biomarkers

ABSTRACT

Background

Pancreatic ductal adenocarcinoma (PDAC) remains one of the most aggressive cancers, typically diagnosed at an advanced stage due to its subtle and often absent early symptoms. Despite representing only 3% of new cancer cases, it is projected to become the second leading cause of cancer‐related deaths by 2030. Currently, early diagnosis remains a significant challenge, and survival rates remain poor due to the lack of effective screening tools.

Methods

We conducted a comprehensive literature review to explore the most recent advances in PDAC detection, focusing on novel biomarkers, liquid biopsies, artificial intelligence (AI)–enhanced imaging, and non‐invasive surveillance strategies. We examined the role of circulating tumor DNA (ctDNA), microRNAs, and volatile organic compounds (VOCs) as diagnostic tools, alongside the integration of advanced imaging modalities like MRI, EUS, and MRCP in high‐risk individuals, including those with hereditary cancer syndromes.

Results

Emerging technologies, such as AI‐driven imaging and liquid biopsy, have shown promising improvements in detecting PDAC at earlier, potentially resectable stages. Surveillance strategies for high‐risk populations, including BRCA1/2 mutation carriers and individuals with Lynch syndrome, have demonstrated increased detection of Stage I PDAC, offering a significant opportunity for curative intervention. AI and machine learning techniques are also enhancing the sensitivity and specificity of imaging, providing a new frontier in early‐stage diagnosis.

Conclusion

The integration of molecular diagnostics, advanced imaging technologies, and AI may enable a paradigm shift in PDAC detection, transitioning from late to early‐stage diagnosis and potentially improving survival rates. However, further clinical validation and standardization of these technologies are essential to ensure their widespread clinical adoption. The future of PDAC detection lies in a multimodal, personalized approach, optimizing diagnostic accuracy and early intervention for high‐risk individuals.

Abbreviations

AI

artificial intelligence

ATM

ataxia‐telangiectasia mutated

AUC

area under the curve

BMP3

bone morphogenetic protein 3

BRCA

breast cancer gene

CA 19‐9

carbohydrate antigen 19‐9

CA125

cancer antigen 125

CA242

cancer antigen 242

CAPS

Cancer of the Pancreas Screening (program)

CDKN2A

cyclin‐dependent kinase inhibitor 2A

CEA

carcinoembryonic antigen

CEH‐EUS

contrast‐enhanced harmonic endoscopic ultrasound

cfDNA

cell‐free DNA

CT

computed tomography

CTCs

circulating tumor cells

ctDNA

circulating tumor DNA

ERCP

endoscopic retrograde cholangiopancreatography

EUS

endoscopic ultrasound

EUS‐FNA

endoscopic ultrasound–fine‐needle aspiration

FDA

Food and Drug Administration

FNA

fine‐needle aspiration

GC‐IMS

gas chromatography–ion mobility spectrometry

GC‐TOF‐MS

gas chromatography–time‐of‐flight mass spectrometry

IMS

ion mobility spectrometry

IPMN

intraductal papillary mucinous neoplasm

KRAS

Kirsten rat sarcoma viral oncogene homolog

LFTs

liver function tests

LYVE1

lymphatic vessel endothelial hyaluronan receptor 1

MicroRNA

miRNA/miR

MOB

methylation on beads

MRCP

magnetic resonance cholangiopancreatography

MRI

magnetic resonance imaging

MUC5AC

mucin 5AC

OS

overall survival

PAC model

pancreatic cancer prediction model

PALB2

partner and localizer of BRCA2

PDAC

pancreatic ductal adenocarcinoma

PGV

pathogenic germline variant

PIK3CA

phosphatidylinositol‐4,5‐bisphosphate 3‐kinase catalytic subunit alpha

PMS2

postmeiotic segregation increased 2

PRSS1

protease serine 1

PTEN

phosphatase and tensin homolog

REG1A

regenerating family member 1 alpha

RT‐qPCR

reverse transcription quantitative polymerase chain reaction

STK11

serine/threonine kinase 11

TAUS

transabdominal ultrasound

TBX15

T‐box transcription factor 15

TFF‐1

trefoil factor 1

THBS2

thrombospondin 2

TP53

tumor protein p53

USPSTF

US Preventive Services Task Force

VOCs

volatile organic compounds

1. Introduction

Pancreatic ductal adenocarcinoma (PDAC) is a relatively uncommon but exceptionally lethal malignancy. Despite it representing approximately 3% of all newly diagnosed cancers annually, it is projected to become the second leading cause of cancer‐related deaths by 2030 [1, 2]. Even for patients diagnosed at a localized stage, the 5‐year overall survival (OS) rate is only 20%, while for those with distant metastases, survival drops to just 2% [3]. The high mortality rate is largely attributed to its silent progression, often reaching advanced or metastatic stages before clinical detection [4]. These statistics underscore the aggressive nature of PDAC and the profound challenges associated with its early diagnosis. Surgical resection remains the only potentially curative treatment, but only 10%–20% of patients are diagnosed at a stage amenable to surgery [1]. Even among those who undergo resection, the 5‐year OS remains low, ranging from 10% to 25% [5]. Overall, only about 12% of patients survive beyond 5 years, as most present with advanced, incurable disease at diagnosis.

Despite these sobering figures, there are signs of progress. Improved clinical management and a deeper understanding of PDAC biology have contributed to a doubling of survival rates over the past decade [6]. The diagnostic criteria for PDAC subtypes typically rely on a combination of imaging modalities such as computed tomography (CT), magnetic resonance imaging (MRI), and endoscopic ultrasound (EUS) along with blood tests and tissue biopsy. However, due to the lack of early symptoms, these diagnostics are usually performed at later stages of disease progression. Although population‐wide screening for PDAC might, in theory, aid early diagnosis, this approach is currently impractical due to the high costs and limited specificity of available diagnostic tests. Efforts to identify reliable risk factors for targeted screening have also faced limitations. Nevertheless, a promising alternative is the surveillance of individuals at elevated risk for developing PDAC [7].

This highlights the urgent need for improved early detection and intervention strategies. Individuals with hereditary cancer syndromes or pancreatic cystic lesions are at significantly higher risk for PDAC, and monitoring these high‐risk groups may enable earlier diagnosis. Detecting asymptomatic, early‐stage neoplasms could greatly improve the chances of curative treatment. This review examines current diagnostic approaches, risk stratification methods, and transformative advances in early detection through genomics, artificial intelligence (AI), and noninvasive biomarkers. Figure 1 presents a graphical overview of the major concepts and early detection strategies discussed in this review.

FIGURE 1.

Graphical abstract of manuscript.

2. Difficulties in Identifying Pancreatic Cancer in Early Stages

PDAC is difficult to diagnose because it often presents with vague, nonspecific symptoms, such as abdominal or back pain, which typically emerge in the later stages. Many patients are diagnosed only after the cancer has metastasized, with approximately 50% already having metastatic disease at diagnosis [8]. The asymptomatic nature of PDAC in its early stages presents a principal challenge in early detection. Research shows that only about 7% of PDAC are diagnosed while still localized. This contrasts starkly with earlier detection rates in other cancers such as breast (61%), colon (40%), lung (16%), ovarian (19%), and prostate (91%) [9], underscoring the uniquely insidious progression of PDAC. Anatomically, the pancreas is located deep within the abdominal cavity as a retroperitoneal organ, making it inaccessible to routine physical examination. Unlike breast or skin cancers, no visible or palpable masses are typically present. It also cannot be evaluated through digital rectal examination like prostate cancer, nor can it be visualized through intraluminal endoscopy as with colorectal lesions. When symptoms do appear, they tend to be vague such as nausea, anorexia, jaundice, weight loss, or upper abdominal pain. Pain may radiate to the back, especially when the tumor is located in the pancreatic body or tail. In contrast, tumors in the pancreatic head often present with painless jaundice and may be associated with steatorrhea and early cachexia. Ampullary tumors are more likely to cause early jaundice, potentially allowing for earlier surgical intervention. Standard laboratory blood tests are generally nonspecific. Liver function tests (LFTs) may reveal elevated enzymes due to biliary obstruction but are not diagnostic. Carbohydrate antigen 19‐9 (CA 19‐9), a commonly used tumor marker, has limited sensitivity and specificity particularly in patients with obstructive jaundice [10]. Another obstacle is the low incidence of PDAC in the general population. In the United States, the overall incidence is approximately nine cases per 100 000 people, increasing to around 68 per 100 000 among individuals over 55. The predictive accuracy of a screening test improves with higher disease prevalence [11, 12]. This rarity makes population‐wide screening impractical. For instance, screening 100 000 individuals over age 55 using a test with 98% specificity and 100% sensitivity would yield 1999 false positives and only 68 true positives. To achieve an acceptable positive predictive value, test specificity would need to exceed 99% [13]. Given the high morbidity associated with pancreatic surgery, a high false‐positive rate is clinically unacceptable. Currently, no existing diagnostic tool is suitable for widespread screening of the general population for PDAC or its precursors [11].

3. Can Pancreatic Cancer Be Detected Early?

Given the aggressive nature of pancreatic cancer, questions remain about whether early detection is truly achievable and whether it would meaningfully reduce mortality.

However, emerging evidence suggests that a critical window exists during which screening and intervention could be effective. Iacobuzio‐Donahue [14] used mathematical modeling to analyze genetic mutations across multiple lesions from individual patients. Their findings estimate that pancreatic cancer may remain in a potentially curable state for up to 10 years from the initiation of tumorigenesis. Using evolutionary biology‐based models, they proposed an average interval of 6–8 years between the development of the first malignant cell and the onset of distant metastasis. Notably, most patients are currently diagnosed during the final 2 years of this progression [15], highlighting a missed opportunity for earlier intervention. Surgical outcomes improve significantly when tumors are detected early before they grow large, invade surrounding tissues, or spread to lymph nodes. Tumor size, margin status, perineural invasion, vascular involvement, and histologic grade are key prognostic factors. Thus, early‐stage detection is generally associated with more favorable tumor characteristics and improved survival. Despite these promising observations, more robust prospective studies are needed to confirm whether early detection and screening lead to reduced morbidity and mortality in practice. A relevant corollary is colorectal cancer (CRC): the use of fecal occult blood testing, sigmoidoscopy, and colonoscopy was not widely accepted until the 1990s after large trials demonstrated their efficacy. Today, colonoscopy is the gold standard for CRC screening [16, 17, 18, 19].

4. Pancreatic Cystic Precursors: Clinical Risk and Implications for Early Detection

Pancreatic cysts, particularly intraductal papillary mucinous neoplasms (IPMNs), represent a valuable opportunity for early detection of PDAC due to their malignant potential and the feasibility of imaging‐based surveillance. Studies report that individuals with pancreatic cysts have an annual PDAC incidence of 0.95% and a 22.5‐fold increased risk compared with those without cysts [20]. IPMNs are the most common type of pancreatic cyst and can lead to two distinct forms of cancer:

1. IPMN‐derived cancer, in which the IPMN itself progresses to malignancy, and

2. IPMN‐comorbid cancer, in which conventional PDAC develops independently of the cyst.

According to a nationwide survey conducted in Japan, 90% of IPMN‐associated malignancies were of the branch duct type, while half of IPMN‐derived cancers were associated with the main pancreatic duct [21]. In main duct‐type IPMNs, the average incidence of invasive carcinoma or high‐grade dysplasia is 61.6% (range: 36%–100%). Surgical resection is strongly recommended when the main pancreatic duct diameter is ≥ 10 mm, especially in the presence of mural nodules or jaundice. In contrast, the risk of carcinoma originating from branch‐duct IPMNs has been reported as 0.2% to 3.0% annually, while the incidence of comorbid cancer is estimated at 0% to 1.1% per year. Notably, 90% of comorbid malignancies and half of IPMN‐origin cancers occur in cases involving branch‐duct IPMNs. Thus, close surveillance is crucial even in the absence of high‐risk features at the time of diagnosis [22]. The 2015 American College of Gastroenterology (ACG) guidelines suggest that follow‐up of asymptomatic pancreatic cysts may be discontinued if there is no change in cyst size or features after 5 years [23]. However, a large‐scale Japanese study of 1404 patients (median follow‐up: 6 years; 9231 patient‐years) found that 68 patients developed PDAC—38 with IPMN‐derived malignancy and 30 with comorbid cancer—corresponding to an annual incidence of 0.7% and cumulative rates of 3.3% at 5 years, 6.3% at 10 years, and 15% at 15 years [24]. These findings suggest that branch‐duct IPMNs may require extended surveillance beyond 5 years. The same study also found that the size of the cyst and main pancreatic duct correlated with the risk of IPMN‐derived cancer but not with the incidence of comorbid malignancy. Therefore, even patients without large cysts or ductal dilation may still be at risk for developing cancer, reinforcing the need for continued follow‐up. As PDAC arises from the ductal epithelium, it may induce early morphological changes referred to as “indirect findings” in the pancreatic duct. For example, even minimal dilation of the main pancreatic duct (≥ 2.5 mm) has been associated with a 6.4‐fold increased risk of PDAC [7]. These findings reinforce the role of pancreatic cyst surveillance as a practical and essential component of early detection strategies.

5. Advancements in Detection Method for Pancreatic Cancer at an Earlier Stage and Risk Factors Causing Pancreatic Cancer

Understanding genetic risk factors is critical for identifying individuals who may benefit from targeted surveillance strategies. Two primary approaches have emerged for the early detection of PDAC:

1. Imaging‐based techniques

2. Biomarker‐based strategies

Detectable and potentially curable pancreatic lesions can be identified using established imaging modalities such as EUS, MRI, and CT [13]. Meanwhile, molecular technologies targeting circulating tumor cells (CTCs), tumor‐specific proteins, mucins, and microRNAs (miRNAs) are under active investigation [25, 26, 27]. When applied to minimally or noninvasively collected specimens such as blood, saliva, or stool, these approaches have the potential to transform PDAC screening. Several genetic and environmental factors are associated with increased PDAC risk. Key germline mutations include alterations in BRCA1, BRCA2, CDKN2A/p16, ATM, PALB2, MLH1, MSH2, MSH6, PMS2, APC, and STK11. A family history of PDAC, particularly in families with two or more affected close relatives, including at least one first‐degree relative, also confers significant risk [28].

PDAC can arise in the setting of hereditary cancer syndromes, which are summarized in Table 1.

TABLE 1.

Top hereditary syndromes associated with PDAC with other hereditary syndromes and non‐genetic risk factors.

Syndrome/factor	Phenotype	Causative gene	Function of gene and mechanism	Cumulative PDAC risk	PDAC relative risk
Lynch syndrome (LS)	Nonpolyposis colorectal or endometrial cancer Other cancers: stomach, small intestine, urinary tract, and brain	MLH1, MSH2, MSH6, PMS2	Mismatch repair system	3.7% at 70 years of age [ 29 ]	8.6‐fold [ 29 ]
Familial adenomatous polyposis syndrome (FAP)	Colorectal polyposis. Increased risk of colorectal cancer, hepatoblastoma, thyroid cancer, and desmoid tumors	APC	Tumor suppressor	1.7% at 80 years of age [ 30 ]	4.5‐fold [ 31 ]
Li–Fraumeni syndrome (LFS)	Sarcoma, adrenocortical, breast, and/or brain carcinoma	TP53	Tumor suppressor DNA repair	< 5% [ 32, 33 ]	—
Hereditary breast–ovarian cancer (HBOC) syndrome	Breast and ovarian cancer	BRCA1, BRCA2, PALB2	Tumor suppressors Homologous repair	1.5%–4% at 70 years of age (specifically increased in BRCA2) [ 34, 35 ]	BRCA1: 4–6‐fold [ 34, 36 ] BRCA2: 3.5–22‐fold [ 34, 36 ] PALB2: 6‐fold [ 37, 38 ]
Familial atypical multiple mole melanoma (FAMMM)	Multiple atypical nevi (> 50) and history of melanoma Other tumors: breast, lung, and endometrium	P16/CDKN2A	Tumor suppressor Cell cycle	17% at 75 years of age [ 39 ]	13–46.6‐fold [ 40, 41 ]
Peutz–Jeghers syndrome (PJS)	Mucocutaneous pigmentation and gastrointestinal hamartomatous polyps High risk of gastrointestinal, breast, ovary, endometrial, and lung cancers	STK11/LKB1	Tumor suppressor AMPK signaling	8%–11% at 70 years of age Up to 36% lifetime risk [ 42, 43 ]	132‐fold [ 43, 44 ]
Hereditary pancreatitis (HP)	Chronic pancreatitis and recurrent acute pancreatitis	PRSS1, SPINK1, PRSS2, CTRC	Encodes cationic trypsinogen/encodes trypsinogen inhibitor	Up to 53% at 75 years of age [ 45, 46 ]	26–87‐fold [ 47 ]
Ataxia telangiectasia (AT)	Cerebellar ataxia/telangiectasias	ATM	Tumor suppressor DNA repair	< 5% [ 32, 48 ]	2.7‐fold [ 48 ]
Cystic fibrosis (CF)	Respiratory infections and pancreatic insufficiency	CFTR	Encodes transmembrane conductance regulator	< 5% [ 48 ]	5.3‐fold [ 49 ]
Familial pancreatic cancer (FPC)	Familial PC aggregation	—	—	−3 or more FDR with PC: 16%–40% cumulative risk [ 50 ] −2 FDR with PC: Up to 12% cumulative risk [ 50 ] −1 FDR with PC: up to 6% [ 51 ]	−3 FDR with PC: 32‐fold [ 51 ] −2 FDR with PC: 6‐fold [ 51 ] −1 FDR with PC: 2–5‐fold [ 52 ]
New‐onset diabetes mellitus (NODM)	Sudden onset of hyperglycemia, insulin resistance, and beta‐cell dysfunction	TCF7L2, KCNJ11	Gene mutations affect insulin secretion and resistance, leading to pancreatic dysfunction and increasing the risk of PDAC	1%–2% in 5 years, rising to 3%–5% long‐term [ 53, 54 ]	6–8‐fold [ 55, 56, 57 ]
Obesity	Excess body fat, BMI > 30 kg/m2, insulin resistance, and metabolic syndrome	FTO, MC4R, ADIPOQ	Insulin resistance, chronic inflammation, and increased adipokine production	1.5%–2% higher than for those with a normal BMI [ 58 ]	1.5–2‐fold [ 59, 60 ]
Tobacco use	Habitual use of cigarettes, cigars, or smokeless tobacco	TP53, KRAS, EGFR	Carcinogens in tobacco smoke induce mutations in key genes, promoting cancer development	2%–3% higher compared with nonsmokers [ 61 ]	1–3‐fold [ 62 ]
Excessive alcohol consumption	Regular consumption of large quantities of alcohol (more than 14 drinks per week for men, 7 for women)	ADH1B, ALDH2, CYP2E1	Alcohol‐induced DNA damage, inflammation, oxidative stress, and pancreatic fibrosis	2%–3% higher in heavy drinkers [ 63 ]	1.5–5‐fold [ 64, 65 ]

Abbreviations: ADH1B, alcohol dehydrogenase 1B; ADIPOQ, Adiponectin; ALDH2, aldehyde dehydrogenase 2; AT, ataxia telangiectasia; BMI, body mass index; CF, cystic fibrosis; CYP2E1, cytochrome P450 2E1; EGFR, epidermal growth factor receptor; FAMMM, familial atypical multiple mole melanoma; FAP, familial adenomatous polyposis; FPC, familial pancreatic cancer; FTO, fat mass and obesity‐associated gene; HBOC, hereditary breast–ovarian cancer; HP, hereditary pancreatitis; KRAS, Kirsten rat sarcoma viral oncogene homolog; LFS, Li–Fraumeni syndrome; LS, Lynch syndrome; MC4R, melanocortin 4 receptor; NODM, new‐onset diabetes mellitus; PJS, Peutz–Jeghers syndrome; TP53, tumor protein p53.

Recognizing these risk factors is essential for designing effective surveillance protocols and optimizing early diagnostic strategies.

6. Effectiveness of Surveillance in High‐Risk Individuals for Early Detection of PDAC

Several guidelines have defined eligibility for PDAC surveillance based on hereditary cancer syndromes, including Lynch syndrome, Peutz–Jeghers syndrome, familial pancreatic cancer (FPC), and BRCA1/BRCA2‐associated hereditary breast–ovarian cancer (HBOC), as well as genetic mutations, family history, or personal medical history. These criteria help identify individuals with a lifetime risk of PDAC greater than 5% [66].

The National Comprehensive Cancer Network (NCCN) Clinical Practice Guidelines [67] in Oncology for PDAC recommend against population‐based screening but support risk‐adapted surveillance in selected high‐risk individuals. According to NCCN guidance, surveillance is appropriate for carriers of STK11 (Peutz–Jeghers syndrome) and CDKN2A, as well as for individuals with BRCA1, BRCA2, PALB2, or ATM pathogenic germline variants in the presence of a family history of PDAC. The NCCN recommends MRI/magnetic resonance cholangiopancreatography (MRCP) and/or EUS as preferred surveillance modalities, initiated at 50 years of age or 10 years earlier than the youngest affected relative, with earlier initiation for certain hereditary syndromes. These recommendations align closely with CAPS, AGA, and ASGE guidance, reinforcing the role of targeted surveillance for early detection of potentially resectable PDAC [67]. While current guidelines endorse EUS as a preferred surveillance modality, it is important to briefly summarize the evidence supporting its use in high‐risk populations. EUS is a key component of PDAC surveillance in high‐risk individuals due to its high spatial resolution and ability to detect small, early‐stage lesions that may not be visible on cross‐sectional imaging. Evidence supporting EUS‐based surveillance is derived primarily from prospective cohort studies and long‐term surveillance programs, which have consistently demonstrated a stage shift toward earlier, resectable disease. In the Cancer of the Pancreas Screening (CAPS) studies and other high‐risk cohorts, surveillance using EUS and/or MRI detected the majority of PDAC at Stage I, with substantially higher resection rates and improved survival compared with symptom‐detected cases. Although randomized trials are lacking, these outcome‐based data, together with concordant international guideline recommendations, provide the current evidence base for EUS as an effective surveillance modality in selected high‐risk populations [68, 69]. These data provide the rationale for incorporating EUS into contemporary surveillance programs for individuals at increased risk of PDAC.

In addition to hereditary syndromes, family history of PDAC alone is a well‐recognized risk factor for which surveillance may be considered, even in the absence of an identified pathogenic germline variant. Individuals from FPC kindreds, typically defined as those with two or more affected relatives including at least one first‐degree relative, are commonly included in surveillance programs using MRI and/or EUS. In contrast, individuals with a personal history of PDAC are not candidates for screening or early detection; their follow‐up represents posttreatment surveillance for recurrence, which is beyond the scope of PDAC screening and is therefore not addressed in this review [70].

Over the past two decades, there has been growing interest in targeted surveillance of individuals at elevated risk for PDAC due to inherited pathogenic variants or a strong family history of the disease [71, 72]. The CAPS program, which focuses on such high‐risk individuals, detected 10 PDAC cases among 1461 participants, 78% of which were Stage I at diagnosis through MRI and/or EUS [73]. Similarly, a 20‐year Dutch surveillance program monitored 347 carriers of germline CDKN2A mutations, primarily using MRI. Among participants who developed PDAC, 83% had resectable disease and 33.3% were diagnosed at Stage I [74]. These findings highlight the potential for earlier, surgically curable detection in genetically predisposed cohorts. Individuals with BRCA1/2‐associated HBOC represent an increasingly recognized high‐risk group and are now routinely included in PDAC surveillance programs [75].

Table 2 summarizes the recommended surveillance strategies for major high‐risk groups, detailing preferred imaging modalities, starting ages, and follow‐up intervals. These protocols are largely based on expert consensus and guidelines from leading gastroenterological and oncological societies.

TABLE 2.

Recommended surveillance for high‐risk groups.

Risk group	Recommended methods	Starting age	Frequency
BRCA1/2, PALB2 mutation + family history	MRI, EUS	50 years or 10 years before youngest case	Annually
Lynch syndrome + family history	MRI, EUS	50 years or 10 years before youngest case	Annually
Peutz–Jeghers syndrome	MRI, EUS	35 years or 10 years before youngest case	Annually
CDKN2A mutation carriers	MRI, EUS	40 years or 10 years before youngest case	Annually
ATM or TP53 mutation + family history	MRI, EUS	50 years or 10 years before youngest case	Annually
IPMN > 10 mm + family history of PDAC	MRI, EUS	At diagnosis	Every 6–12 months
BRCA1/2‐associated HBOC	MRI/MRCP, EUS	50 years or 10 years before youngest case	Annually

Abbreviations: ATM, ataxia‐telangiectasia mutated; BRCA, breast cancer gene; CDKN2A, cyclin‐dependent kinase inhibitor 2A; HBOC, hereditary breast–ovarian cancer; IPMN, intraductal papillary mucinous neoplasm; PALB2, partner and localizer of BRCA2; TP53, tumor protein p53.

Following the summary of surveillance protocols, Table 1 outlines the proposed criteria for determining eligibility for PDAC surveillance in individuals with an estimated lifetime risk greater than 5%. This distinction clarifies how surveillance should be performed (Table 1) and who should undergo it (Table 3).

TABLE 3.

Proposed criteria for eligibility for pancreatic surveillance in individuals with over 5% lifetime risk.

Classification	Pancreatic surveillance criteria	Starting surveillance age
Familial PDAC history	Familial pancreatic carcinoma syndrome involves having two or more than two family members with PDAC, where at least two of these relatives are first‐degree relatives	50 years old or 10 years young in age as compared with the most youngest affected family person.
Hereditary syndrome
Familial breast and ovarian cancerous syndrome	PGV in BRCA2, BRCA1, or PALB2	50 years old or 10 years more earlier as compared with the most youngest relative affected by PDAC.
Lynch syndrome	PGV in MLH1, MSH2, MSH6 + a first‐ or second‐degree relative having PDAC	50 years old or 10 years earlier to the most youngest relative influenced by PDAC.
Familial atypical multiple mole melanoma syndrome	PGV in CDKN2A	40 years old or 10 years younger to the most youngest relative affected by PDAC.
Peutz–Jeghers syndrome	PGV in STK11	35 years old or 10 years young in age to the most youngest relative affected by PDAC.
Others	PGV in ATM + a first‐ or second‐degree relative having PDAC PGV in TP‐53 + a first‐ or second‐degree relative having PDAC	50 years old or 10 years young in age to the youngest relative influenced by PDAC.
Hereditary genetic pancreatitis	PGV in PRSS1	40 years old or 10 years young in age to the most youngest relative affected by PDAC.
Persons having a personal history of pancreatic tumors	Intrapancreatic mucinous neoplasm (IPMN) larger than 10 mm or a genetic history of a first‐degree relative with PDAC	Not applicability
BRCA1/2‐associated HBOC	BRCA1/2 carriers with at least one first‐degree or two second‐degree relatives with pancreatic cancer should undergo annual pancreatic surveillance. Those without a family history are generally not recommended for surveillance	Starting at age 50 years or 10 years earlier than the youngest affected relative with PDAC.

Abbreviations: PDAC, pancreatic ductal adenocarcinoma; PGV, Pathogenic germline variant.

Over the past decade, multiple expert groups have emphasized the value of MRI and EUS as the preferred modalities for high‐risk surveillance due to their ability to detect small, potentially resectable lesions without ionizing radiation. The US Preventive Services Task Force (USPSTF) does not recommend population‐wide screening for PDAC in average‐risk individuals [76]. Given the disease's rarity, randomized trials comparing surveillance to no surveillance in high‐risk populations are impractical. However, selective monitoring of high‐risk groups has emerged as a viable alternative. A review of 25 surveillance studies involving more than 3000 high‐risk individuals found that EUS, MRI, and/or CT imaging can identify asymptomatic PDAC at an early stage [77]. Based on such evidence, the American Gastroenterological Association issued a 2019 clinical practice update recommending targeted surveillance for high‐risk patients [78]. Similarly, the American Society for Gastrointestinal Endoscopy released 2022 guidelines advocating surveillance in these populations using a Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach [77, 79]. Evidence also suggests that surveillance improves outcomes even after accounting for lead‐time bias. In a Dutch cohort of individuals with the CDKN2A/p16 founder mutation, surveillance‐detected PDAC had a higher 5‐year survival rate (32.4%) and a greater proportion of Stage I diagnoses (38.7%) compared with registry‐based controls [80]. However, some challenges remain. A retrospective study of 28 PDAC cases arising in 2552 high‐risk individuals enrolled in 16 international surveillance programs found that nearly half of the high‐grade dysplasia or PDAC lesions were detected between scheduled imaging visits, with a median interval of 11 months following a normal scan [81]. This underscores the need for improved surveillance sensitivity and possibly shorter follow‐up intervals in certain subgroups. The “holy grail” in PDAC surveillance is a multimodal approach integrating patient demographics, clinical evaluation, imaging, genomics, biomarkers, and AI to deliver personalized risk prediction. Serum‐based biomarkers, including circulating tumor DNA (ctDNA) and protein panels, could enhance existing EUS/MRI‐based strategies if they are cost‐effective, highly specific, and sensitive enough for early detection [82, 83, 84, 85, 86, 87]. A recent study in JAMA Internal Medicine found that CT scans may cause around 103 000 new cancer cases each year in the United States, contributing to about 5% of all cancer diagnoses annually. This shows that while CT is commonly used in diagnosing and monitoring diseases, its potential to cause cancer due to radiation exposure should be carefully considered, especially in routine surveillance [88]. Such tests could be used between scheduled imaging sessions or in conjunction with imaging to improve yield. Advances in AI are also enhancing the sensitivity and specificity of CT, MRI, and EUS for PDAC detection [89, 90]. AI tools can aid in identifying high‐risk individuals from large datasets, calculating individualized PDAC risk from electronic health records, and potentially expanding focused surveillance to broader populations [91]. Although population‐wide screening for PDAC is not cost‐effective, several modeling studies suggest that targeted surveillance may be economically reasonable in carefully selected high‐risk populations. Cost‐effectiveness analyses have demonstrated favorable incremental cost‐effectiveness ratios for surveillance using MRI and/or EUS in individuals with Peutz–Jeghers syndrome, CDKN2A mutations, or FPC, particularly when surveillance leads to detection of early‐stage, surgically resectable disease. However, most available data are derived from decision‐analytic models rather than prospective trials, and results are highly sensitive to assumptions regarding cancer incidence, test performance, surveillance intervals, and surgical outcomes. Further real‐world studies are needed to define the cost‐effectiveness of emerging multimodal surveillance strategies incorporating biomarkers and AI [92, 93].

7. Methods for Detecting Pancreatic Cancer at Early Stages

Several modalities have been explored to detect PDAC at an early stage. These approaches span imaging‐based techniques, liquid biopsies, and advanced computational models, each offering distinct advantages and limitations. Figure 2 provides a schematic overview of the major diagnostic strategies currently used or under investigation.

FIGURE 2.

Different methods for detecting PDAC early.

To complement this visual representation, Table 4 summarizes the diagnostic performance of these methods, including sensitivity, specificity, key strengths, and notable limitations. Presenting the data in this format enables direct comparison and helps highlight the gaps that must be addressed to achieve reliable early detection.

TABLE 4.

Diagnostic methods for early PDAC detection.

Method	Modality	Sensitivity	Specificity	Strengths	Limitations
CA 19‐9	Blood	~80%	~70%–90%	Widely used, cost‐effective	Low sensitivity in early‐stage PDAC
EUS	Imaging	~86%	~90%	High resolution; allows FNA	Operator dependent; invasive
ctDNA	Liquid biopsy (blood)	65%–75%	~90%	Detects mutations; dynamic monitoring	Low abundance in early disease
miRNA panel	Blood	~85%	~85%–90%	Stable, noninvasive; early‐stage detection	Heterogeneity in methods, needs standardization
Radiomics + AI	CT/MRI	~85%–90%	~90%	Detects subclinical lesions; automated	Requires validation; cost, data complexity

Abbreviations: AI, artificial intelligence; CA 19‐9, carbohydrate antigen 19‐9; CT, computed tomography; ctDNA, circulating tumor DNA; EUS, endoscopic ultrasound; FNA, fine‐needle aspiration; miRNA, microRNA; MRI, magnetic resonance imaging; PDAC, pancreatic ductal adenocarcinoma.

8. Imaging Techniques

This section focuses on the diagnostic performance and limitations of imaging modalities for early‐stage PDAC.

Building on earlier discussion, this section focuses on the diagnostic performance of imaging modalities in early‐stage PDAC. The USPSTF has assessed imaging modalities such as CT, MRI, and EUS in clinical trials primarily among high‐risk individuals with hereditary syndromes or a strong family history of PDAC [76]. The diagnostic performance of imaging varies considerably in early‐stage PDAC. A multicenter retrospective study from Japan evaluated 200 cases diagnosed at Stage 0 or Stage I, finding that only 20% of patients presented with symptoms at diagnosis [94]. Reported diagnostic accuracies were transabdominal ultrasound (TAUS): 67.5%, CT: 98.0%, MRI: 86.5%, and EUS: 86.5%. In early‐stage disease, CT, EUS, and MRCP often reveal irregular narrowing and dilation of the main pancreatic duct. Abbreviated or accelerated MRI protocols are being explored as a strategy to improve the feasibility of pancreatic imaging, particularly in surveillance and longitudinal assessment of high‐risk individuals. These protocols reduce acquisition time and cost by focusing on a limited number of high‐yield sequences that capture pancreatic parenchymal morphology and ductal abnormalities while avoiding ionizing radiation. Preliminary studies suggest that abbreviated MRI may retain acceptable sensitivity for detecting clinically relevant pancreatic lesions; however, prospective validation and standardized protocols are needed before routine clinical implementation [95, 96]. However, CT and TAUS have limitations, as they largely depend on indirect findings such as ductal dilatation or changes in glandular contour. Notably, CT detected localized fatty changes in the pancreatic parenchyma in 42% of Stage 0 cases and 41.8% of Stage I cases highlighting the importance of recognizing such subtle features [94]. EUS offers distinct advantages, including the ability to detect small solid lesions without ionizing radiation or contrast agents, and to obtain cytopathological samples via fine‐needle aspiration. More recently, contrast‐enhanced harmonic EUS (CEH‐EUS) has emerged as a minimally invasive method to visualize microvascular perfusion in pancreatobiliary diseases. In a prospective single‐center study, CEH‐EUS demonstrated significantly higher sensitivity (94.5% vs. 83.1%) and diagnostic accuracy (84.1% vs. 78.6%) than conventional EUS [18]. A meta‐analysis further reported a pooled sensitivity of 93% (95% CI: 0.91–0.95) and specificity of 80% (95% CI: 0.75–0.85) for CEH‐EUS in diagnosing PDAC [97]. Contrast‐enhanced CT remains the gold standard for pancreatic tumor imaging, with reported sensitivity of up to 90%, specificity of up to 99%, and minimal interobserver variability [98]. Endoscopic evaluations most often involve EUS‐guided fine‐needle aspiration (EUS‐FNA), while endoscopic retrograde cholangiopancreatography (ERCP) is generally reserved for specific diagnostic or therapeutic indications [11, 12, 13]. Beyond conventional visual assessment, advanced computational approaches have been developed to improve early detection.

9. Radiomics

Radiomics enables quantitative extraction of imaging features that may improve early PDAC detection [99]. Using CT or MRI scans, radiomic data can be computationally processed and combined with clinical information to build diagnostic and prognostic models. When integrated with AI and machine learning, these models have the potential to enhance early cancer detection [100, 101, 102, 103, 104]. Recent studies have shown that radiomics coupled with machine learning can detect PDAC up to 2 years before a clinical diagnosis is established [105]. This is particularly valuable because conventional imaging often struggles to identify early‐stage disease. However, widespread CT‐based surveillance is currently impractical due to cost, radiation exposure, and the need for specialized analysis. One example of innovation in this area is the FELIX Project at Johns Hopkins, which uses deep learning algorithms trained to identify subtle pancreatic abnormalities on CT scans that may indicate early neoplasia [106]. These developments suggest that AI‐powered radiomics could become an important tool in presymptomatic PDAC detection, especially in high‐risk populations.

10. Liquid Biopsy

Liquid biopsy approaches include ctDNA, CTCs, and noncoding RNAs, each with distinct advantages and limitations.

Liquid biopsy represents an ideal strategy for early PDAC detection, as it allows minimally invasive or noninvasive sampling most often through blood collection. Among the promising technologies is methylation on beads (MOB), a nanotechnology platform capable of isolating and analyzing trace amounts of DNA from serum. In one study, Yi et al. used MOB to assess methylation differences in circulating DNA from 42 PDAC patients. They reported: BNC1 promoter methylation: 79% sensitivity, 92% specificity; ADAMTS1 methylation: 48% sensitivity, 92% specificity. When combined, these markers achieved 81% sensitivity (95% CI: 69%–93%) and 85% specificity (95% CI: 71%–99%) [107].

11. CTCs and Cell‐Free DNA (cfDNA)

CTCs and ctDNA are among the most promising liquid biopsy biomarkers for PDAC. While initially studied for their prognostic value in advanced disease, increasing evidence shows that CTCs can also be detected in early‐stage PDAC and even in preinvasive lesions. This highlights their potential as noninvasive indicators of tumor presence before clinical or radiologic signs appear [108, 109]. Advances in microfluidics‐based enrichment and imaging flow cytometry have improved the ability to isolate and characterize these rare cells from peripheral blood. However, challenges remain in standardizing detection methods and improving yield, particularly in asymptomatic or high‐risk individuals.

cfDNA (circulating free DNA) consists of short DNA fragments circulating in plasma, with the tumor‐derived fraction known as ctDNA [110].

These fragments often harbor tumor‐specific mutations, such as oncogenic KRAS alterations present in over 90% of PDAC cases that can be detected using highly sensitive techniques like digital droplet PCR and next‐generation sequencing [111, 112]. Detection of these mutations in plasma shows high concordance with tumor tissue genotyping, providing a minimally invasive means of molecular profiling [113]. Despite these advances, early‐stage disease presents a challenge because CTC and ctDNA concentrations are typically very low. This necessitates ultrasensitive assays and careful interpretation to avoid false negatives or positives [114]. Tumor heterogeneity and background noise from non‐tumor circulating free DNA further complicate analysis. Integrating liquid biopsy data with advanced imaging and serum biomarkers can improve diagnostic accuracy [115]. Recent applications of AI and machine learning have enhanced interpretation by identifying subtle genomic patterns and temporal changes that may be missed by conventional analysis [116, 117]. Multimodal diagnostic approaches combining genomics, AI, and non‐invasive biomarkers hold strong potential to shift PDAC detection to earlier, more treatable stages. Large‐scale prospective studies are underway to validate CTC and ctDNA applications for population screening, risk stratification, and high‐risk surveillance.

12. ctDNA

Although no cost‐effective population‐wide screening tool currently exists, ctDNA testing offers the advantage of repeated, longitudinal monitoring using small blood volumes. Several studies have demonstrated the diagnostic potential of ctDNA. In a 2017 study using digital droplet PCR in 52 PDAC patients, ctDNA achieved 65% sensitivity and 75% specificity, slightly lower than CA 19‐9 (79% sensitivity, 93% specificity) and EUS‐guided biopsy (73% sensitivity, 88% specificity) [118, 119]. Importantly, ctDNA detection both before and after surgery correlates with poorer progression‐free and OS, suggesting prognostic utility. Emerging methods such as DNA methylation analysis within ctDNA further enhance early detection. Techniques like MOB can discriminate malignant from benign pancreatic conditions [120]. Additionally, PDAC often has shorter ctDNA fragment sizes and higher concentrations than healthy individuals [121]. KRAS mutations, detected in plasma ctDNA since 1994, remain central molecular markers, and combining KRAS mutation detection with protein biomarkers increases diagnostic accuracy [122]. When KRAS mutation detection was combined with four protein biomarkers, diagnostic sensitivity rose to 64% and specificity to 99.5% [123]. Moreover, methylation profiles within ctDNA can help differentiate chronic pancreatitis from PDAC [124]. Key challenges include the very low abundance of ctDNA in early disease, requiring ultrasensitive techniques such as ultradeep next‐generation sequencing or enhanced PCR‐based assays. Biological confounders such as tumor heterogeneity and nontumor circulating free DNA from clonal hematopoiesis necessitate rigorous assay standardization to minimize false results. Longitudinal monitoring is another promising application. Serial ctDNA measurements can enable early detection of recurrence postsurgery and real‐time assessment of treatment response—capabilities not always achievable with imaging [125]. Multiple ongoing clinical trials are evaluating ctDNA for screening and surveillance in high‐risk populations, and commercial platforms (digital droplet PCR, Idylla, COBAS z480, and BEAMing) are being compared for accuracy and feasibility [126]. Integration of ctDNA analysis with advanced imaging and serum biomarkers can enhance sensitivity and specificity. The application of AI to recognize subtle genomic and epigenomic patterns in ctDNA datasets may further improve early PDAC diagnosis. Although cost‐effectiveness and scalability remain considerations, continued technological progress may make ctDNA‐based screening a transformative tool in shifting PDAC detection to earlier, more curable stages. Despite promising diagnostic accuracy, sensitivity in very early‐stage disease remains a key limitation.

13. miRNAs

Several circulating miRNA signatures have demonstrated strong diagnostic potential for PDAC, including early‐stage disease. Circulating free RNAs (cfRNAs) encompass various RNA species, but miRNAs have attracted particular attention because they are resistant to RNase degradation and reflect tumor‐specific biological processes. Most studies employ reverse transcription quantitative polymerase chain reaction (RT‐qPCR) for detection, with serum being the most common sample type. Several specific miRNA signatures have demonstrated diagnostic potential. In one study, Shi et al. identified miR‐1246, miR‐205‐5p, and miR‐191‐5p as promising serum biomarkers, achieving 91.5% accuracy in distinguishing PDAC from chronic pancreatitis and healthy controls, including early‐stage disease [111]. Similarly, Mato Prado et al. reported a bile‐based miRNA profile capable of differentiating malignant from benign pancreaticobiliary diseases [112]. This study also proposed a dual‐miRNA serum signature miR‐125b‐5p and miR‐194‐5p that effectively distinguished PDAC from cholangiocarcinoma [112]. A recent meta‐analysis involving over 4326 patients across 46 studies reported a pooled sensitivity of 0.79 (0.76–0.82) and specificity of 0.74 (0.68–0.79) and an AUC of 0.81 (0.77–0.84) for miRNA in PDAC diagnosis, further supporting their diagnostic utility. The results of this meta‐analysis showed that circulating miRNAs yielded a high diagnostic accuracy for PDAC. More importantly, they also exhibited a satisfactory diagnostic performance for early‐stage PDAC, meeting the urgent need for an ideal biomarker for early‐stage PDAC in clinical settings [127]. Machine learning approaches have further improved miRNA‐based diagnostics. Algorithms such as support vector machines and neural networks can detect subtle expression patterns, outperforming conventional threshold‐based methods. These AI models integrate complex miRNA profiles from large datasets to differentiate PDAC from benign conditions and other malignancies with high precision. Despite these promising results, clinical translation faces challenges, including heterogeneity in miRNA panel selection, lack of standardized preanalytical protocols, and variability between patient populations. Large, diverse, and prospective validation studies are needed. Ultimately, incorporating miRNAs into AI‐powered, multimodal diagnostic platforms alongside ctDNA, imaging, and protein biomarkers may enhance accuracy and enable earlier detection of PDAC in routine clinical practice.

14. Emerging Biomarkers and Innovative Approaches for Early Pancreatic Cancer Detection

Recent research has identified novel biomarkers across various biological fluids and experimental platforms that may significantly enhance the early detection of PDAC. These emerging strategies offer noninvasive and potentially more accurate alternatives to current diagnostic tools [128].

Several biological fluids such as blood, cystic fluid, urine, pancreatic juice, and stool have been studied as sources of tumor‐derived genetic, epigenetic, or protein markers. Additionally, AI‐based models and multimodal machine learning tools are being developed to interpret these biomarkers in a clinically meaningful way [129].

Table 5 summarizes selected novel biomarkers currently under investigation for early PDAC detection, detailing their source, biological target or mechanism, and reported diagnostic performance in terms of sensitivity and specificity.

TABLE 5.

Novel biomarkers under investigation for early detection.

Biomarker	Source	Target/mechanism	Sensitivity	Specificity
KRAS mutation	ctDNA	Oncogenic driver mutation	~70%	~90%
Methylated TBX15	Cystic fluid	Epigenetic changes	~80%	~88%
miR‐1246, miR‐205‐5p	Serum	Oncogenic miRNA expression	~85%–90%	~85%
THBS2 + CA 19‐9	Blood	Protein biomarker panel	87%	90%–95%
LYVE1, REG1A, TFF‐1	Urine	Secreted protein panel	~85%	~88%

Abbreviations: CA 19‐9, carbohydrate antigen 19‐9; ctDNA, circulating tumor DNA; KRAS, Kirsten rat sarcoma viral oncogene homolog; LYVE1, lymphatic vessel endothelial hyaluronan receptor 1; miR, microRNA; REG1A, regenerating family member 1 alpha; TBX15, T‐box transcription factor 15; TFF‐1, trefoil factor 1; THBS2, thrombospondin 2.

Building upon these findings, broader biomarker discovery efforts have categorized fluid‐based strategies and computational modalities according to their detection source and technological platform.

Table 6 summarizes these fluid‐based biomarker strategies alongside computational diagnostic innovations, highlighting the diversity of approaches under evaluation for early PDAC detection.

TABLE 6.

Emerging biomarker strategies and diagnostic modalities for early detection of pancreatic cancer.

Biomarker strategies based on biological fluids
Cystic fluid	Monoclonal antibody which are Das 1 KRAS, GNAS, TP53, PIK3CA, and PTEN gene mutation Methylated TBX15 along with BMP3
Pancreatic juice	Methylated C13 or f18 and FER1L4, along with BMP3
Blood	CA 19‐9 (inside a multi marker panel) Thrombospondin 2 (THBS2) Cancer‐SEEK: A proteins panel along with mutations in cfDNA Plasma methylated DNA markers
Urine	Protein panel with LYVE 1, REG1A, and TFF‐1
Stool	Proteobacterial along with Firmicutes microbial dominance in earlier stages of PDAC
Other novel modalities
End and PAC model	Scores calculation based on differences in body weight along with blood glucose levels and age at onset in individuals having new onset diabetes mellitus
Comp‐Cyst	Very supervised machine learning module integrating with clinical plus imaging features and genetic along with biochemical markers from cyst fluid
EUS‐based artificial intelligence models	Detecting advanced neoplasia inside pancreatic cyst Separating PDAC from pancreatitis

Abbreviations: cfDNA, cell‐free DNA; EUS, endoscopic ultrasound; PAC model, pancreatic cancer prediction model; PDAC, pancreatic ductal adenocarcinoma; REG1A, regenerating family member 1 alpha; TFF‐1, trefoil factor 1; THBS2, thrombospondin 2.

In addition to these strategies, several experimental tools are under active development to identify volatile organic compounds (VOCs) released by tumors. Technologies such as gas chromatography–ion mobility spectrometry (GC‐IMS) and field asymmetric waveform ion mobility spectrometry (FAIMS) have shown potential in analyzing urine or breath samples for PDAC‐specific VOC signatures. Nanotechnology‐based biosensors, such as ultra–pH‐sensitive fluorescent nanoprobes, are also emerging as high‐specificity diagnostic tools [130].

These modalities when combined with advanced computational models may eventually form the basis of cost‐effective, highly sensitive, and noninvasive screening programs for high‐risk populations. However, extensive clinical validation and standardization are required before routine clinical adoption.

15. Serological Biomarkers

Serum‐based biomarkers remain the most clinically accessible tools, though their utility in early detection is limited. Despite many proposed candidates, no serological biomarker has yet achieved sufficient accuracy for routine early diagnosis of PDAC [131, 132]. To date, CA 19‐9 remains the only FDA‐approved serum marker for PDAC. In clinical practice, panels combining multiple markers such as CA 19‐9, carcinoembryonic antigen (CEA), CA 125, and CA 242 have shown improved diagnostic accuracy. Among these, CA 19‐9 offers the highest sensitivity (~80%) while CA 242 provides the greatest specificity (~90%) [133, 134, 135]. Despite its widespread use, CA 19‐9 has notable limitations as a biomarker for PDAC. One significant issue is its susceptibility to false positives, particularly in patients with obstructive jaundice, which can elevate CA 19‐9 levels independent of PDAC. Other benign conditions, such as cholangitis, pancreatitis, and cirrhosis, can also lead to raised levels of CA 19‐9, complicating its specificity [136, 137]. Consequently, while CA 19‐9 remains valuable as a screening and diagnostic tool, it is important to interpret the results in conjunction with other clinical findings and diagnostic modalities to avoid misdiagnosis. This highlights the need for multimodal approaches, such as the combination of CA 19‐9 with imaging techniques or other biomarkers, to enhance diagnostic accuracy and minimize the risk of false positives. When used together, these four markers achieve a sensitivity of 90.4% and a specificity of 93.8%, outperforming any single marker [134]. In recent years, several novel protein biomarkers have emerged as potential complements to traditional serological tests. For instance, thrombospondin 2 (THBS2), macrophage inhibitory cytokine‐1 (MIC‐1), and osteopontin have shown promising diagnostic value, particularly when combined with CA 19‐9. In one study, the combination of CA 19‐9 and THBS2 achieved an area under the receiver operating characteristic curve of 0.97 for distinguishing early‐stage PDAC from healthy individuals. Additionally, advances in AI and machine learning now enable integration of large‐scale biomarker data including CA 19‐9 levels, novel protein markers, and genomic features into predictive models. These AI‐enhanced approaches have demonstrated superior accuracy in differentiating early PDAC from benign pancreatic conditions, offering a pathway toward more precise and timely diagnosis [138]. Ongoing research is now focusing on integrating serological markers with non‐invasive genomic and proteomic signatures such as ctDNA, exosomal RNA, or AI‐assisted pattern recognition to further improve early detection accuracy. Recent advancements in cfDNA fragmentomics have demonstrated significant potential in the early detection of PDAC. A study by Yin et al. [139] developed a machine learning model utilizing cfDNA fragmentomic features, such as copy‐number variations, fragmentation size ratios, and methylation patterns, to accurately distinguish early‐stage PDAC from healthy controls. The model achieved remarkable performance, with an area under the curve (AUC) of 0.992 in the training cohort and 0.987 in the validation cohort. This promising approach could complement existing surveillance methods by offering a noninvasive, highly sensitive tool for early PDAC detection, particularly in high‐risk populations [139]. These emerging technologies, including cfDNA fragmentomics, represent promising advances in the quest for more accurate, noninvasive, and early detection methods for PDAC, which could significantly improve patient outcomes, especially in high‐risk populations.

16. Alternative Biofluids for Liquid Biopsy in Pancreatic Cancer

Beyond blood‐based diagnostics, alternative biofluids including pancreatic juice, stool, and saliva are under investigation for their potential to enable early detection of PDAC. These fluids may harbor tumor‐specific genetic and epigenetic alterations that serve as promising biomarkers.

16.1. Pancreatic Juice Analysis

Pancreatic juice contains cfDNA and RNA shed from epithelial cells lining the pancreatic ducts. Genetic analyses of pancreatic juice from patients with PDAC have frequently identified mutations in KRAS, TP53, and SMAD4, which can help differentiate malignant lesions from benign or precancerous conditions [140, 141]. Additionally, promoter methylation of APC and histamine receptor H2 (HRH2) genes has been observed more frequently in pancreatic juice from cancer patients, supporting their potential role as diagnostic biomarkers [142]. Notably, Japanese researchers have demonstrated that exosomal miRNAs, particularly miR‐21 and miR‐155, can effectively distinguish PDAC from chronic pancreatitis, with AUC values of 0.90 and 0.89, respectively. When combined with cytological analysis, miRNA profiling further enhances diagnostic accuracy [143]. Compared with plasma, pancreatic juice may contain a higher concentration of tumor‐derived nucleic acids, potentially offering greater sensitivity for detecting low‐grade dysplasia or early‐stage PDAC. However, collection of pancreatic juice typically requires ERCP. ERCP is now primarily reserved for therapeutic purposes, such as biliary drainage, due to the associated risk of complications, including perforation, pancreatitis, and infection. These risks have led to a more cautious approach in its use for diagnostic purposes, with alternative imaging techniques like MRCP becoming more common, which limits its feasibility for routine screening [144]. Emerging studies are also investigating extracellular vesicles and long noncoding RNAs (lncRNAs) in pancreatic juice as novel biomarker candidates, further expanding its diagnostic potential.

16.2. Stool‐Based Biomarkers for Pancreatic Cancer Detection

The possibility that tumor‐derived genetic material may enter the gastrointestinal tract and be detected in stool has gained increasing attention in recent years. While this strategy has not yet been clinically implemented for PDAC, several studies have investigated its feasibility [145]. A notable precedent is Cologuard, an FDA‐approved stool DNA test for CRC. This noninvasive, at‐home test combines assays for KRAS mutations, methylation of NDRG4 and BMP3, and immunochemical detection of hemoglobin [146]. Although designed for CRC, Cologuard demonstrates that genetic and epigenetic alterations arising along the gastrointestinal tract can be reliably identified in stool samples. Currently, no stool‐based screening kit exists for PDAC; however, the success and regulatory approval of Cologuard suggest that similar tools could be feasible in the near future [11]. In one study, KRAS mutations were detected in 88% of stool samples from PDAC patients, compared with only 19.6% in individuals without the disease [147]. Furthermore, stool‐based detection of miR‐181b and miR‐210 has shown the potential to differentiate PDAC patients from healthy controls, underscoring the diagnostic promise of this approach [147]. It is hypothesized that pancreatic tumor DNA reaches the intestinal lumen via bile drainage or exfoliation into the duodenum, enabling its presence in fecal matter. However, the low abundance and fragmentation of tumor DNA in stool combined with degradation by digestive enzymes and microbial contamination pose technical challenges for consistent detection. To overcome these barriers, emerging research is investigating additional molecular targets, including DNA methylation markers and extracellular vesicle‐associated cargo, as promising avenues to improve the sensitivity and specificity of stool‐based diagnostics.

16.3. Saliva‐Based Biomarkers for Pancreatic Cancer Detection

Saliva presents a noninvasive and easily accessible medium for detecting cancer‐specific biomarkers. In rodent models of PDAC, exosome‐like vesicles have been shown to transport tumor‐derived molecules into the salivary glands, supporting the biological plausibility of saliva‐based diagnostics [148, 149]. Tumor‐derived exosomes and systemic inflammatory signals are believed to alter the composition of saliva, facilitating the presence of pancreatic cancer biomarkers in this biofluid. Several salivary mRNA markers including KRAS, ACRV1, and MBD3L2 have been identified as potential discriminators between PDAC patients and healthy individuals. Notably, combining these markers has improved diagnostic accuracy, even in patients with coexisting pancreatitis, which often complicates early diagnosis [150, 151]. In addition to mRNA, salivary metabolites and alterations in oral microbiota have also shown promise as diagnostic tools for early detection of PDAC [152, 153]. Moreover, salivary miRNAs, such as miR‐31 initially studied in oral cancers, have demonstrated relevance in pancreatic cancer detection as well [154, 155, 156, 157]. However, factors such as variability in salivary flow, oral hygiene, and coexisting oral diseases may affect biomarker stability and reliability, highlighting the need for further standardization. Although numerous studies support the utility of saliva in liquid biopsy approaches, integration of these findings into clinical practice remains limited. Further validation and standardization are needed before saliva‐based testing can be adopted as a routine diagnostic tool for PDAC.

17. Emerging Technologies for Pancreatic Cancer Detection

Emerging diagnostic approaches are exploring unique biological and chemical signatures produced by tumors. For example, PDAC releases distinct VOCs that can be detected in urine, breath, and fecal samples. Interestingly, trained canines have been shown to recognize these VOCs, highlighting their potential as diagnostic biomarkers [158, 159]. In addition to liquid biopsies and genetic screenings, noninvasive diagnostic methods such as breath testing have gained traction in early cancer detection. One notable example is the COBRA1 study [160], which developed a breath test to detect CRC by identifying VOCs. The study found that 14 VOCs could accurately distinguish CRC patients from healthy individuals, achieving an AUC of 0.87. This method represents a promising step toward non‐invasive early cancer detection. Similar techniques could potentially be adapted for PDAC, offering a novel, noninvasive approach to monitor high‐risk individuals for early‐stage PDAC [160]. Going forward, FAIMS is a sensitive analytical technology capable of detecting VOCs. In a study by Nissinen et al. [161], ion mobility spectrometry (IMS) successfully differentiated urine samples from PDAC patients and those with premalignant pancreatic lesions, achieving 85% sensitivity and 75% specificity. However, the diagnostic accuracy of IMS alone remains insufficient for clinical implementation. More recently, the same group reported that combining GC‐IMS with gas chromatography–time‐of‐flight mass spectrometry (GC‐TOF‐MS) improved differentiation between PDAC and healthy controls, achieving an AUC exceeding 0.85 [162]. In parallel, nanotechnology is rapidly emerging as a transformative field in cancer diagnostics. Nanomaterials such as fluorescent nanoparticles have been developed to label and detect biological molecules with high sensitivity and specificity [163]. For example, ultra–pH‐sensitive fluorescent nanoprobes remain inactive during systemic circulation but become activated in the acidic extracellular environment typical of tumors. These nanoprobes have demonstrated high tumor specificity and imaging efficacy in several cancer models, including PDAC [164].

The integration of advanced nanotechnology with computational approaches, such as AI and machine learning, holds great promise for enhancing the sensitivity and precision of early PDAC detection. Nevertheless, challenges including standardization of techniques, cost‐effectiveness, and extensive clinical validation must be addressed before these promising technologies can be routinely implemented in clinical practice.

18. Conclusion

The early detection of PDAC remains a critical challenge, given its asymptomatic nature and late‐stage diagnosis. Recent advances in genomics, AI, and noninvasive biomarkers, such as liquid biopsies and AI‐enhanced imaging, have significantly improved the ability to detect PDAC at earlier, more treatable stages. Emerging technologies like breath testing for VOCs and advanced imaging techniques offer promising noninvasive alternatives for monitoring high‐risk populations. Furthermore, targeted surveillance strategies based on genetic predispositions, such as BRCA1/2 and Lynch syndrome, can identify individuals at higher risk, enabling earlier and potentially curative interventions. Moving forward, integrating these multimodal diagnostic approaches and ensuring their clinical validation will be essential in transforming PDAC detection into a more precise and accessible practice. Continued collaboration across multidisciplinary fields will be pivotal to achieving widespread implementation and improving patient outcomes.

Funding

This study/work was supported by 1.3.5 Project for Disciplines of Excellence—Clinical Research Incubation Project, West China Hospital, Sichuan University (2021HXFH001); 1.3.5 Project for Artificial Intelligence, West China Hospital, Sichuan University (ZYAI24030); National Natural Science Foundation of China for Young Scientists Fund (82203782); Sichuan Natural Science Foundation (2024NSFSC0742 and 2024NSFSC1949); Sichuan Science and Technology Program (2024YFFK0384 and 2024YFFK0385); Sichuan University‐Sui Ning School‐Local Cooperation Project (2022CDSN‐18); the Postdoctoral Research Fund of West China Hospital, Sichuan University (2024HXBH134); and (Qilu) Clinical Research of Sichuan Anticancer Association (XH2023‐032 and XH2023‐502).

Ethics Statement

No patients were involved in the design of the study, and no ethical approval from an institutional review board was required.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

All data generated or analyzed during this study are included in the manuscript.