Role of endoscopic ultrasonography or magnetic resonance imaging for screening of pancreatic cancer in low-risk individuals

Abstract

BACKGROUND

Magnetic resonance imaging (MRI) and endoscopic ultrasonography (EUS) are recommended in combination for screening pancreatic cancer in high-risk individuals. However, in clinical practice, MRI and EUS are increasingly utilized for pancreatic surveillance during routine health examinations.

AIM

To investigate the feasibility of these imaging modalities for screening in low-risk individuals.

METHODS

This retrospective study included patients at low risk for pancreatic cancer who underwent MRI or EUS at two health evaluation centers between March 2019 and December 2024. Basic characteristics, laboratory data, and imaging results were collected.

RESULTS

A total of 3364 low-risk individuals underwent pancreatic screening: 1553 (46.1%) received MRI, and 1811 underwent EUS. No significant differences were observed in age or sex distribution between the groups. In imaging screening, EUS demonstrated a higher detection rate of abnormal pancreatic lesions (12.8% vs 2.6%; P < 0.001). MRI detected more cystic lesions than did EUS (P < 0.001). EUS identified smaller nodular lesions compared to MRI (9.2 mm vs 18.0 mm; P = 0.044). The MRI group had a higher number of confirmed intraductal papillary mucinous neoplasms (P = 0.031), whereas the EUS group identified more suspected branch-duct intraductal papillary mucinous neoplasms (P < 0.001). Pancreatic adenocarcinoma was found in three patients (0.08%), with no significant difference in detection rates between EUS and MRI (0.11% vs 0.06%; P = 0.656).

CONCLUSION

In low-risk individuals, MRI and EUS offer comparable effectiveness for pancreatic cancer surveillance. The choice of imaging strategy for health evaluation depends on cost considerations and degree of invasiveness.

Key Words: Endoscopic ultrasonography; Low-risk individuals; Pancreatic cancer; Magnetic resonance imaging; Screening

Core Tip: Screening is not recommended for average-risk individuals due to the low incidence of pancreatic cancer. However, in practice, magnetic resonance imaging and endoscopic ultrasonography are increasingly used for pancreatic surveillance during health examinations. This study demonstrated the comparable effectiveness of both imaging modalities in detecting pancreatic adenocarcinoma in health evaluation centers. The implementation of broader screening strategies for low-risk individuals should be carefully considered in light of the age-related increases in risk factors, procedural invasiveness, and associated costs.

INTRODUCTION

Pancreatic cancer ranks as the sixth leading cause of cancer mortality due to its insidious onset and aggressive nature[1]. The global incidence of pancreatic cancer has increased significantly from 5.47 per 100000 in 1990 to 5.96 per 100000 in 2021[2]. Incidence and mortality follow a similar age-related trend, with gradual escalation after 30 years of age, peaking in the 65-69 years age group[3]. The highest burden of pancreatic cancer is observed in high-income Asia Pacific and North America, with incidence rates of 10.689 and 10.196, respectively[2]. Pancreatic cancer deaths are projected to increase by approximately 1.97-fold by 2060[3]. In Taiwan, the incidence of pancreatic cancer increased from 6.04 per 100000 in 2013 to 8.02 per 100000 in 2022, with adenocarcinoma being the most common type, followed by neuroendocrine tumors[4,5].

Pancreatic cancer is influenced by various risk factors, including age, male sex, smoking, obesity, diabetes, alcohol intake, chronic pancreatitis, and family history of pancreatic cancer or certain genetic syndromes[6]. In high-risk populations comprising carriers of gene mutations or first-degree relatives of patients with familial pancreatic cancer, the lifetime risk of developing pancreatic cancer is approximately 10%[7]. Surveillance programs for this group have shown promising results, with earlier stage diagnoses potentially reducing mortality rates[8]. Currently, annual imaging using magnetic resonance imaging (MRI) and endoscopic ultrasonography (EUS) in an alternating manner is recommended for high-risk individuals[9]. A microsimulation screening analysis demonstrated that annual screening led to a 49% reduction in mortality among high-risk patients[10]. Among non-invasive screening methods, carbohydrate antigen 19-9 (CA19-9) is commonly used for pancreatic cancer screening, with a sensitivity of approximately 80%; however, CA19-9 presents challenges, including false positives in inflammatory conditions and false negatives in Lewis-negative individuals[11]. Furthermore, CA19-9 lacks sensitivity for early-stage detection[12]. The protease activity-based assay using a magnetic nano sensor test, when used alongside the CA19-9 test, identifies early-stage pancreatic cancer with 85% accuracy[13].

As the incidence of pancreatic cancer continues to rise, there is a growing demand for pancreatic cancer screening. Screening programs could enhance awareness of pancreatic health and encourage individuals to seek timely medical attention for symptoms. Theoretically, early diagnosis might result in improved survival, as demonstrated in high-risk groups undergoing regular surveillance. However, there are no widely recommended screening tools or protocols for low-risk individuals due to the rare incidence, and the benefits of screening remain limited and controversial because of patient burden, overdiagnosis, and overtreatment. This study assessed the efficacy of MRI or EUS for screening low-risk patients in the health examination setting.

MATERIALS AND METHODS

Study population and data collection

This retrospective study included adult patients (≥ 20 years old) at low risk of pancreatic cancer who underwent screening with abdominal MRI or EUS at two health evaluation centers in Taiwan between March 1, 2019 and December 31, 2024. Whole-abdomen MRI without contrast was performed at MacKay Memorial Hospital using either a 1.5-T scanner (Signa HDxt or Discovery MR750, GE Healthcare, MKE, United States; or Magnetom Trio Tim, Siemens Medical Solutions, Erl, Germany) or 3.0-T scanner (Intera Achieva, Philips Medical Systems, Best, Netherlands). The MRI protocol included T1-weighted gradient-echo, T2-weighted axial and coronal sequences, turbo spin-echo or a turbo spin-echo variant, and water-fat Dixon reconstruction. EUS was performed at Dianthus Health Clinics using an GF-UCT260 EVIS LUCERA (Olympus, Tokyo, Japan) ultrasound gastrovideoscope. The EUS procedure was combined with colonoscopy and esophagogastroduodenoscopy, all performed simultaneously under anesthesia.

Baseline characteristics, including age, body mass index (BMI), family history of pancreatic cancer, alcohol consumption, and smoking habits, were obtained from a questionnaire completed at examination. Clinical data included complete blood cell count, liver and renal function tests, hemoglobin A1C, CA19-9, and abdominal ultra-sound results obtained from participants on the same health check-up day. All clinical information for these patients was available in an integrated electronic medical record system that included all patient encounters. The inclusion criteria comprised patients undergoing their first screening for pancreatic cancer using MRI or EUS. The exclusion criteria included patients with a medical history of pancreatic lesions or surgery for pancreatic tumors, a family history of familial pancreatic cancer in first-degree relatives, chronic pancreatitis, or those who underwent repeated screenings during the study period. Clinically concerning pancreatic lesions were followed up via telephone contact. This study was approved by our institutional review board (approval No. 25MMHIS129e) with a waiver of informed consent, as it was a data-only retrospective review of de-identified data with no patient interactions.

Statistical analysis

Descriptive statistics for continuous variables are reported as mean ± SD. Categorical variables are described using frequency distributions and re-ported as number (%). We used the χ² or Fisher exact tests for categorical data and independent t-test or Wilcoxon rank-sum test for continuous data comparisons between study groups. For data analyses, we used SPSS for Windows software version 25.0 (Chicago, IL, United States). All P-values were two-sided, and P < 0.05 indicated statistical significance.

RESULTS

Screening characteristics

A total of 3364 low-risk individuals underwent pancreatic screening: 1553 (46.1%) received an MRI, whereas 1811 underwent an EUS examination (Table 1). No significant differences were observed in age or sex distribution between the two groups. However, individuals in the EUS group had a higher BMI compared to those in the MRI group (23.5 vs 22.5; P < 0.001). No differences were found in smoking habits, alcohol consumption, or medical history of diabetes. Regarding laboratory tests, no significant differences were determined in complete blood count, liver function, or metabolic profiles (low-density lipoprotein cholesterol and hemoglobin A1C) between the two groups, excluding creatinine levels, which were higher in the EUS group (0.84 mg/dL vs 0.79 mg/dL; P < 0.001). CA19-9 levels showed no significant differences between groups.

Table 1 Basic characteristics of the enrolled patients, n (%)/mean ± SD.

	MRI (n = 1553)	EUS (n = 1811)	P value
Age (year)	55.1 ± 11.0	52.9 ± 10.2	0.251
Gender (male)	976 (62.8)	1101 (60.8)	0.222
Body mass index, kg/m2	22.4 ± 6.9	23.5 ± 4.2	< 0.001
Smoking	139 (8.9)	195 (10.8)	0.073
Alcohol consumption	165 (10.6)	189 (10.4)	0.889
Diabetes	87 (5.6)	125 (6.9)	0.122
Laboratory result
Hemoglobin, g/dL	14.4 ± 1.5	14.3 ± 1.6	0.299
White blood cell, 103/μL	5.6 ± 1.5	5.5 ± 1.4	0.203
Platelet, 103/μL	248.3 ± 58.6	238.2 ± 57.1	0.595
Amylase, U/L	61.7 ± 25.8	61.8 ± 21.2	0.167
GPT, U/L	27.3 ± 20.5	27.8 ± 17.8	0.411
Total bilirubin, g/dL	0.96 ± 0.46	0.97 ± 0.53	0.270
Creatinine, mg/dL	0.79 ± 0.27	0.84 ± 0.19	< 0.001
LDL, mg/dL	124.1 ± 35.4	127.9 ± 35.1	0.317
Hba1c, %	5.72 ± 0.76	5.77 ± 0.74	0.063
CA19-9, U/mL	11.7 ± 10.1	12.4 ± 11.2	0.380

MRI: Magnetic resonance imaging; EUS: Endoscopic ultrasonography; GPT: Glutamic pyruvic transaminase; LDL: Low-density lipoprotein; Hba1c: Hemoglobin A1c; CA19-9: Carbohydrate antigen 19-9.

Characteristics of MRI and EUS in detecting pancreatic lesions

Among the abnormal lesions detected by the two imaging modalities (Table 2), 40 patients (2.6%) in the MRI group had pancreatic lesions - 35 with cystic lesions and 5 with solid lesions. No complications occurred during screening. In the EUS group, 231 patients (12.8%) had abnormal pancreatic lesions, including 182 with cystic lesions, 42 with solid lesions, and 7 with both cystic and solid lesions. The detection rates of pancreatic cystic and solid lesions were higher with EUS than that with MRI (P < 0.001). The detection rate of abnormal pancreatic lesions on abdominal ultrasound was 0.5%, compared to that of the combined MRI and EUS group at 0.8%, with no significant difference. Cystic lesions were the most common type, accounting for 21 of 22 cases (95.5%). Among confirmed neoplasms, the detection rate of adenocarcinoma was 0.06% in the MRI group and 0.11% in the EUS group, without a significant difference (P = 0.656). Similarly, the detection rate of neuroendocrine tumors did not differ between the two modalities (P = 0.656). However, for intraductal papillary mucinous neoplasm (IPMN), the MRI group showed a higher detection rate compared to that of the EUS group (0.26% vs 0%; P = 0.031). Among suspicious neoplasms, no significant differences were observed between the EUS and MRI groups in detecting main-duct IPMN, mixed-type IPMN, serous cystic neoplasm, mucinous cystic neoplasm, or solid pseudopapillary neoplasm. However, branch-duct IPMN showed a higher detection rate in the EUS group compared with the MRI group (2.3% vs 0.5%, P < 0.001).

Table 2 Abnormal pancreatic lesions detected by magnetic resonance imaging and endoscopic ultrasound during surveillance, n (%).

	MRI (n = 1553)	EUS (n = 1811)	P value
Pancreatic lesion	40 (2.6)	231 (12.8)	< 0.001
Cystic lesion	35 (2.3)	182 (10.1)	< 0.001
Solid lesion	5 (0.3)	42 (2.3)	< 0.001
Cystic and solid lesion	0 (0)	7 (0.4)	0.014
Abdominal ultrasound finding	7 (0.5)	15 (0.8)	0.176
Cystic lesion	6 (0.4)	15 (0.8)	0.105
Solid lesion	1 (0.06)	0 (0)	0.280
Proven neoplasm
Adenocarcinoma	1 (0.06)	2 (0.11)	0.656
Neuroendocrine tumor	1 (0.06)	2 (0.11)	0.656
IPMN	4 (0.26)	0 (0)	0.031
Suspicious neoplasm
Branch-duct IPMN	7 (0.5)	42 (2.3)	< 0.001
Main-duct IPMN	1 (0.06)	1 (0.06)	0.913
Mixed-type IPMN	0 (0)	2 (0.11)	0.190
Serous cystic neoplasm	1 (0.06)	2 (0.11)	0.656
Mucinous cystic neoplasm	0 (0)	1 (0.06)	0.354
Solid pseudopapillary neoplasm	0 (0)	1 (0.06)	0.354

MRI: Magnetic resonance imaging; EUS: Endoscopic ultrasonography; IPMN: Intraductal papillary mucinous neoplasm.

Comparing MRI and EUS of pancreatic lesion

No difference was detected in cystic or solid lesion detection among the abnormal pancreatic findings between the EUS and MRI groups (Table 3). For cystic lesions, the mean size showed no difference between the EUS and MRI groups (6.2 cm vs 6.4 cm; P = 0.503), with a higher number of cysts in the MRI group (P < 0.001). For solid lesions, EUS detected smaller lesions than that of the MRI group (9.2 mm vs 18.0 mm; P = 0.044).

Table 3 Characteristics of pancreatic lesions on magnetic resonance imaging and endoscopic ultrasound, n (%).

	MRI (n = 40)	EUS (n = 231)	P value
Cystic lesion	35 (87.5)	182 (78.8)	0.203
Size (mm)	6.4 ± 2.2	6.2 ± 5.0	0.503
Number
1	21	171	< 0.001
2	7	6
3	4	3
≥ 4	3	2
Solid lesion	5 (12.5)	42 (18.2)	0.381
Size (mm)	18.0 ± 11.8	9.2 ± 6.9	0.044
Number
1	5	41	0.727
2	0	1
Cystic and solid lesion	0 (0)	7 (3.0)	0.265

MRI: Magnetic resonance imaging; EUS: Endoscopic ultrasonography.

Characteristics of proven pancreatic neoplasm

In the EUS group, one adenocarcinoma measuring 1.2 cm was located in the body of the pancreas in a 62-year-old male patient, whereas another measuring 4 cm was found in the tail of the pancreas in a 66-year-old female. Regarding the two patients with neuroendocrine tumors identified by EUS, one measured 8.9 mm in a 37-year-old man and another measured 7.9 mm in a 69-year-old man. In the MRI group, adenocarcinoma occurred in a 66-year-old man with a tumor size of 2.5 cm in the head of the pancreas. A neuroendocrine tumor was noted in a 68-year-old female with a tumor size of 8.0 mm. Abdominal ultrasound detected only one patient (16.7%) with a 2.5 cm solid lesion in the head of the pancreas among the six patients with solid lesions found by EUS or MRI. All cases of adenocarcinoma and neuroendocrine tumors were histologically confirmed following imaging-based screening. As for the four patients with IPMN, all were diagnosed via MRI.

DISCUSSION

Currently, there are no pancreatic cancer screening guidelines for the general population due to its relatively low lifetime risk of 1.6%[14], although theoretical benefits to screening low-risk individuals exist. This is the first study to evaluate the feasibility of using EUS or MRI for screening during health evaluations of low-risk individuals. In this study, EUS demonstrated a higher detection rate of abnormal pancreatic cystic or solid lesions compared to that of the MRI. The most common cystic lesion identified on EUS was branch-duct IPMN. EUS was more effective at detecting smaller solid lesions, whereas MRI identified more cystic lesions. The detection rate for adenocarcinoma was 0.08%, with no significant difference between the two imaging modalities.

The mean age at screening in this low-risk study was over 50 years. Pancreatic cancer screening is recommended for high-risk individuals starting at age 50, at age 40 for specific mutation carriers, or at age 35 for individuals with Peutz-Jeghers syndrome[9]. No differences were found in basic characteristics between these two screening tools in our study, excluding higher BMI and creatinine levels in the EUS group. This finding is consistent with those of previous studies showing that higher BMI is associated with increased creatinine clearance[15]. A prospective cohort study revealed that overweight individuals (BMI ≥ 25 kg/m²) have an increased risk of pancreatic cancer[16]; however, the mean BMI in the EUS group was < 25 kg/m².

MRI, particularly when combined with magnetic resonance cholangiopancreatography, demonstrates high sensitivity for detecting pancreatic abnormalities, especially cystic lesions. It detects small lesions in high-risk populations when dedicated protocols are used[17]. Imaging at 3T improves temporal and spatial resolution for evaluating small focal pancreatic lesions, and preliminary results suggest that the signal-to-noise ratio can be as much as twice as high as at 1.5T[18]. Contrast-enhanced MRI increases the T1 signal intensity of normal pancreatic parenchyma, aiding in differentiation from pancreatic masses such as neuroendocrine tumors[19]. Recently, the Pancreatic Cancer Early Detection Consortium published recommendations for a standardized MRI protocol for individuals with elevated risk of pancreatic cancer[20].

EUS can detect smaller pancreatic tumors, particularly those less than 3 cm, which may not be visible on MRI[21,22]. A systematic review evaluating the impact of pancreatic cancer screening on life expectancy, as determined in model-based studies, showed that while screening could be detrimental in the general population, annual pancreatic cancer screening in high-risk populations may increase life expectancy more than one-time screening in the general population (life expectancy gain: 20-260 days vs 2-110 days)[23].

Pancreatic cystic lesions are often diagnosed incidentally, with an incidence that may reach up to 49% in the general population. The majority are benign, with only approximately 3% of detected pancreatic cysts being potentially malignant[24]. Previous studies have shown that pancreatic cysts < 15 mm at diagnosis have a very low risk of malignant transformation[25]. The mean cystic lesion size in the low-risk individuals in our study was 6.2-6.4 mm, suggesting that follow-up in the future is adequate. A recommended follow-up imaging interval is 24 months for patients with simple pancreatic cysts measuring < 2 cm[26].

This study revealed different detection rates of abnormal lesions between the two imaging modalities but similar detection rates for adenocarcinoma. The higher number of confirmed IPMN cases in the MRI group and the greater frequency of suspected branch-duct IPMN in the EUS group may reflect the hospital-based setting of the MRI cohort, where additional diagnostic evaluations were more readily performed through in-hospital transfers. One study demonstrated no significant difference between MRI and EUS in classifying pancreatic cystic lesions or predicting malignancy when images were retrospectively reviewed[27]. A systematic review of 2112 high-risk individuals who underwent imaging found that the weighted pooled proportion of focal pancreatic abnormalities detected by EUS was significantly higher than that detected by MRI (P = 0.006)[28]. MRI has important limitations regarding the timely detection of solid lesions[29], which is consistent with our findings that EUS detected smaller nodular lesions than MRI.

The advantages of MRI for low-risk individuals include its non-invasive nature and absence of radiation exposure, making it a safer option for repeated imaging if necessary. However, MRI is often more cumbersome to use, as patients must remain motionless to obtain accurate images, and it involves high costs and decreased instrument availability. EUS is invasive and may require sedation, which can deter its use as a routine screening tool. Both imaging modalities can lead to false positives and unnecessary follow-up procedures, and they are limited by the low disease prevalence. The detection rate of pancreatic adenocarcinoma in this study was 0.08%, and all cases were in early stages. This finding is comparable to the incidence rate of pancreatic cancer in Taiwan, reported as 8.02 per 100000 population in 2022[5]. A review article demonstrated that the low incidence of pancreatic cancer in the general population leads to a positive predictive value of only 0.5% in screening test scenarios; thus, there would be very low certainty that a person with a positive test result was correctly diagnosed as having pancreatic cancer[30]. Currently, data on pancreatic cancer screening in strictly low-risk general populations using EUS or MRI are limited. However, the reported sensitivities and specificities in high-risk groups may help guide expectations for low-risk populations (Supplementary Table 1). A recent review reported that EUS has a sensitivity of 81%-100% and a specificity of 73%-100% for pancreatic adenocarcinoma[21], whereas MRI demonstrates sensitivities of 84%-93% and specificities of 82%-93%[31].

Limitation

This study had some limitations. First, as a retrospective study based on data from health evaluation centers, the outcomes of abnormal pancreatic lesions could not be traced for all patients. Furthermore, patients with small lesions without worrisome features did not undergo additional follow-up imaging or pathological confirmation. Consequently, sensitivity, specificity, and positive and negative predictive values could not be assessed. This study did not directly compare EUS and MRI in the same low-risk individuals but instead analyzed groups with similar characteristics and a large sample size. Second, the MRIs in this study were performed without contrast enhancement and without magnetic resonance cholangiopancreatography, and some were conducted using a 1.5T scanner. These factors may have reduced the detection rate of smaller abnormal lesions. Finally, genetic mutations are not typically screened in patients without a family history, which may affect the detection rate of pancreatic cancer in the general population. A prospective study from Taiwan demonstrated that when genetic testing followed by MRI screening was performed for 303 patients with a family history, 7.9% had the trypsinogen gene mutation, and 6.3% of at-risk individuals were diagnosed with pancreatic cancer[32].

CONCLUSION

Highly sensitive MRI and EUS are currently reserved for high-risk groups. However, expanding screening to low-risk individuals may become justified as technology advances, risk factors increase, and cost-effectiveness improves. Both MRI and EUS offer comparable effectiveness for pancreatic cancer surveillance in low-risk populations.