Abstract
Background/objectives:
Screening for pancreatic ductal adenocarcinoma alters the course of disease among high-risk individuals (HRIs) and is recommended by multiple societies including the International Cancer of the Pancreas Screening Consortium, American Gastroenterological Association, and American Society of Gastrointestinal Endoscopy. However, there are limited analyses on the screening rates and barriers to adherence among HRIs. This study aims to describe real-world longitudinal screening adherence of a HRI surveillance cohort and identify potential barriers to adherence.
Methods:
Patients followed by Stanford’s Benign Pancreas Clinic were identified as HRIs if they met screening criteria per the latest abovementioned screening guidelines, and were included in our study if they underwent at least 1 screening exam. Data on HRIs were obtained retrospectively from our hospital’s electronic health record system. Patient and screening characteristics were analyzed by adherence rates.
Results:
262 HRIs undergoing recommended screening were identified. The mean follow-up time per patient was 4.9 years and the mean successful annual screening rate within the entire cohort was 67%. HRIs in the lowest quartile of adherence were more likely to have more EUS rather than MRI surveillance exams compared to those who were completely adherent (p = 0.01). HRIs who were completely adherent were also an older cohort compared to those with non-complete adherence (p = 0.02) or in the lowest quartile of adherence (p = 0.01).
Conclusions:
It is difficult to achieve high adherence rates for annual pancreatic cancer screening of HRIs as recommended by the latest guidelines. Age and screening modality may be factors that contribute to adherence difficulty.
Keywords: Real-world surveillance, High-risk individuals, Pancreatic cancer screening
1. Introduction
Pancreatic cancer is a notoriously deadly disease. Despite a relatively moderate incidence rate of 13.5 per 100,000 persons per year, it is the third leading cause of cancer-related death in the United States with a 5-year survival rate of only 13% [1]. Late detection of this malignancy is a prominent contributor to its high mortality rate.
Screening for pancreatic ductal adenocarcinoma (PDAC) may alter the course of disease among high-risk individuals (HRIs). A meta-analysis of prospective studies of formal HRI surveillance programs showed that only 20% of PDAC among these programs were diagnosed at stage IV disease [2], compared to greater than 50% in the general population. Other studies suggest that PDAC detected from HRI surveillance programs are more likely to be resectable and that surgical treatment of these cancers are associated with higher R0 resection rate and lower mortality [3–9].
However, implementation of HRI surveillance programs has thus far demonstrated varying rates of screening adherence. The latest study from the largest HRI cohort in the United States, the multicenter Cancer of Pancreas Screening-5 (CAPS5) cohort, revealed that 16.1% of the cohort had a baseline study without any subsequent surveillance after a lapse of ≥1 year [10]. Meanwhile, a large HRI screening study from the Dutch Familial Pancreatic Cancer Surveillance Study Group reported that 81% of follow-up visits were performed within the designated month and only 1% lapsed a full year [6]. Prior studies have shown the negative impact of COVID-19 on pancreatic cancer surveillance [11]. Studies that explore and define barriers to screening adherence among HRI in pancreatic cancer screening programs are limited.
Currently, multiple guidelines describe recommendations for high-risk individual screening. The International Cancer of the Pancreas Screening Consortium (CAPS) [12], American Gastroenterological Association (AGA) [13], and American Society of Gastrointestinal Endoscopy (ASGE) [14] have published guidelines that each differ slightly on the criteria for high-risk individuals, most recently in 2019, 2020, and 2022, respectively. All 3 of these guidelines agree that high-risk individuals should be screened annually with ideally either magnetic resonance imaging (MRI) or endoscopic ultrasound (EUS).
This study aims to describe real-world longitudinal screening adherence, within a tertiary academic medical center, of a HRI surveillance cohort and identify potential barriers to annual surveillance adherence.
2. Methods
2.1. Patient population
Stanford’s Benign Pancreas Clinic has established a pancreatic cancer screening program to which patients are referred from the community by various providers including primary care physicians, oncologists, and genetic counselors. Patients followed by Stanford’s Benign Pancreas Clinic were identified as HRIs if they met screening criteria per the 2019 CAPS, 2020 AGA, and/or 2022 ASGE screening guidelines. Screening recommendations include annual screening, alternating EUS and MRI. If eligible individuals do not have a baseline screening preference, an EUS is selected for the baseline exam. If not already completed, high-risk individuals are referred to genetic counseling to complete hereditary genetic testing.
HRIs were retrospectively included in our study if they underwent at least one baseline screening exam by either EUS or MRI. Following the baseline screening exam, patients were deemed to have completed annual surveillance exams if they completed either an abdominal MRI with contrast or EUS in the calendar year following their last screening/surveillance exam. Surveillance was stopped if the patient developed pancreatic cancer, died, underwent total pancreatectomy, developed co-morbidities that would preclude pancreatic surgery (e.g. other metastatic cancer with poor prognosis, transition to hospice care, or frailty with inadequate functional status for surgery), or decided to stop surveillance of their own accord (e.g. because of advanced age).
2.2. Data collection
Data on HRIs were obtained retrospectively using chart review of our hospital’s electronic health record system (EHR) from inception of our EHR to December 31, 2024. Patients’ charts included clinical notes, imaging reports, and procedure reports from all Stanford-affiliated hospitals as well as other healthcare organizations connected to the EHR through a secure health information exchange platform called Care Everywhere. Demographic and clinical data were gathered from each HRI including age, gender, race/ethnicity, zip code, high-risk indication, personal history of other cancer, primary language, timing and type of screening/surveillance exam, and high-risk findings of either high-grade dysplasia (HGD) or PDAC. Zip code was used to determine distance from our screening center as a continuous variable using an online global positioning system map. Zip code was also used to determine national area deprivation index (ADI), a measure of socioeconomic disadvantage in which higher indices indicate greater disadvantage. ADI was determined as a continuous variable using Neighborhood Atlas, an interactive online mapping function developed and validated by the University of Wisconsin Center for Health Disparities Research [15]. This study was approved by the institutional review board of Stanford University.
2.3. Statistical analysis
Descriptive statistics were presented as percentages, means, medians, or interquartile ranges (IQRs). For each patient, screening adherence rate was calculated as the number of years in which a screening/surveillance exam was performed divided by the number of years of follow-up. Follow-up years was considered the time between index baseline exam and either death, diagnosis of pancreatic cancer, decision to stop surveillance, or December 31, 2024 when data collection stopped. If patients were lost to follow-up from either Stanford or other healthcare centers connected through Care Everywhere, it was assumed that the patient did not undergo pancreatic cancer surveillance during those years. We compared characteristics between those who adhered to annual surveillance during the duration of follow-up, those who missed at least 1 annual exam, and those in the lowest quartile of adherence rate. Continuous and categorical variables were compared using the Wilcoxon rank-sum test and Chi-Squared test, respectively. Statistical analyses were performed using the Python 3.12.3 statistical package. P-values were two-sided and p < 0.05 were considered statistically significant.
3. Results
From March 2009 to December 31, 2024, 340 HRIs were identified. Of these, 262 underwent at least one baseline screening exam and were thus included in the study. Among the 262 participants, 790 total screening/surveillance exams were performed, of which 359 (45.4%) were MRIs and 431 (54.6%) were EUS (Table 1). The mean number of exams per individual was 3 (SD 2.3). 59 (22.5%) met high-risk criteria due to family history only, while 203 (77.5%) met high-risk criteria due to a high-risk hereditary mutation. Among the cohort, 199 (76.0%) were women and the mean age at baseline screening was 59.6.
Table 1.
Characteristics and screening rates of high-risk individuals categorized by adherence patterns.
| All HRIs | HRIs surveilled annually | HRIs who missed >1 exam | P-valuea | HRIs in lowest quartile of adherence | P-valueb | |
|---|---|---|---|---|---|---|
| No. of participants (%) | 262 (100) | 76 (29.0) | 186 (71.0) | 61 (23.3) | ||
| No. of completed screening/surveillance exams (%) | 790 (100) | 242 (30.6) | 548 (69.4) | 99 (12.5) | ||
| MRI | 359 (45.4) | 111 (45.9) | 248 (45.3) | 0.94 | 27 (27.3) | 0.01 |
| EUS | 431 (54.6) | 131 (54.1) | 300 (54.7) | 72 (72.7) | ||
| No. of missed screening/surveillance exams (%) | 475 (100) | 0 | 475 (100) | 266 (56.0) | ||
| MRI | 261 (54.9) | 0 | 261 (54.9) | 145 (54.5) | ||
| EUS | 215 (45.3) | 0 | 215 (45.3) | 121 (45.4) | ||
| Mean age at baseline exam (SD) | 59.6 (9.3) | 61.8 (9.5) | 58.7 (9.1) | 0.02 | 57.9 (9.2) | 0.01 |
| Female (%) | 199 (76.0) | 55 (72.4) | 144 (77.4) | 0.43 | 53 (86.0) | 0.06 |
| Race/ethnicity (%) | ||||||
| Non-Hispanic White | 169 (64.5) | 48 (63.2) | 121 (65.1) | 0.89 | 38 (62.3) | 0.42 |
| Hispanic | 35 (13.4) | 11 (14.5) | 24 (12.9) | 8 (13.1) | ||
| Asian | 34 (13.0) | 10 (13.2) | 24 (12.9) | 7 (11.5) | ||
| Black | 4 (1.5) | 2 (2.6) | 2 (1.1) | 0 | ||
| Other/Unknown | 21 (8.0) | 5 (6.6) | 15 (8.1) | 8 (13.1) | ||
| High-risk indication (%) | ||||||
| Family history only | 59 (22.5) | 18 (23.7) | 41 (22.0) | 0.88 | 14 (23.0) | 1.0 |
| Hereditary mutation | 203 (77.5) | 58 (76.3) | 145 (78.0) | 47 (77.0) | ||
| History of other cancer prior to screening (%) | 118 (45.0) | 33 (43.4) | 85 (45.7) | 0.84 | 30 (49.2) | 0.62 |
| Non-English speaking (%) | 15 (5.7) | 4 (5.3) | 11 (5.9) | 1.0 | 4 (6.6) | 1.0 |
| Distance from screening center, mi | ||||||
| Mean (SD) | 58.0 (78.3) | 66.5 (98.3) | 54.5 (68.5) | 0.93 | 56.8 (71.3) | 0.68 |
| Median (IQR) | 30.4 (16.1, 59.0) | 30.2 (12.7, 57.4) | 30.6 (16.2, 60.4) | 31.7 (18.2, 65.6) | ||
| Area Deprivation Index | ||||||
| Mean (SD) | 7.2 (12.4) | 6.1 (11.7) | 7.6 (12.7) | 0.36 | 7.6 (13.4) | 0.53 |
| Median (IQR) | 2 (1, 6) | 2 (1, 4) | 2 (1, 3) | 2 (1, 6.5) | ||
| Mean follow-up years (SD) | 4.9 (2.9) | 3.2 (2.7) | 5.5 (2.7) | 6.1 (2.7) | ||
| Mean no. screening exams per patient (SD) | 3.0 (2.3) | 3.2 (2.7) | 2.9 (2.1) | 1.6 (1.0) | ||
| Mean screening adherence rate (SD) | 0.67 (0.3) | 1 (0) | 0.53 (0.2) | 0.28 (0.1) | ||
| No. patients with high-risk lesions (%) | ||||||
| High-grade dysplasia | 2 (0.8) | 2 (2.6) | 0 | 0 | ||
| PDAC | 2 (0.8) | 1 (1.3) | 1 (0.5) | 0 |
Abbreviations: HRI, high-risk individuals; y, years; mi, miles; SD, standard deviation; IQR, interquartile range; PDAC, pancreatic ductal adenocarcinoma.
Comparing groups that were surveilled annually vs missed at least 1 exam.
Comparing groups that were surveilled annually vs lowest quartile.
The cohort was followed for a total of 1,273 person-years. The mean follow-up time per patient was 4.9 years (SD 2.9) and the mean successful annual screening rate within the cohort was 67% (SD 28%).
A total of 76 (29.0%) patients were completely adherent to the recommendation for annual surveillance during every year of follow-up. In this “completely adherent” group, the mean years of follow-up was 3.2 (SD 2.7). The remaining 186 (71.0%) patients missed at least one annual surveillance exam during their years of follow-up. The mean follow-up years in this group was 5.5 (SD 2.7) and the mean annual adherence rate was 53.0%. Comparing these two groups, the “completely adherent” group consisted of older patients compared to the group with non-complete adherence (mean age 61.8 versus 58.7 years) (p = 0.02) (Table 1). There were otherwise no significant differences by type of exam performed, gender, race/ethnicity, high-risk indication, personal history of other cancer, language barrier, distance from screening center, or zip code-based national area deprivation index (ADI).
There were 61 (23.3%) patients identified as being in the lowest quartile of adherence. These patients had a mean follow-up time of 6.1 years (SD 2.7). The mean annual adherence rate was 27.8%. Comparing to the “completely adherent” group, the lowest quartile group was significantly more likely to undergo EUS (72.7% versus 54.1%) rather than MRI (27.3% versus 45.9%) compared to the complete annual adherence group (p = 0.01) (Table 1). The “completely adherent” group was also again older than the group with the lowest quartile of adherence (mean age 61.8 vs 57.9 years) (p = 0.01). Other than these two characteristics, these two groups had no significant differences by gender, race/ethnicity, high-risk indication, personal history of other cancer, language barrier, distance from screening center, or zip code-based national area deprivation index (ADI).
To query whether longer mean follow-up time could explain the low adherence group, a linear regression analysis was performed to model screening adherence rates by adherence category while controlling for follow-up time. Linear regression showed that follow-up time did not significantly impact screening adherence rate and could not account for the difference among the categories.
Two patients were diagnosed with high-grade dysplasia during surveillance. One patient with HGD was initially referred for a baseline screening exam because of an extensive family history of pancreatic cancer. He completed every annual surveillance exam and after 6 years of follow up, he was found to have a mixed-type intraductal papillary mucinous neoplasm which developed into high-grade dysplasia at the age of 72. The 2nd patient with HGD was initially referred with a strong family history of pancreatic cancer and recent acute pancreatitis episode. After 6 years of follow-up, she developed a main-duct intraductal papillary mucinous neoplasm which developed into high-grade dysplasia at the age of 78. In that time, she was adherent to 5 of her 6 annual surveillance exams.
Two patients were found to have PDAC. One patient’s pancreatic cancer was found at the time of baseline screening EUS and found to be stage IV disease at that time. The baseline screening exam was performed at age 75 upon referral to the pancreas clinic from a genetic counselor after genetic testing found him to have a BRCA2 mutation. Although referred for screening due to recent identification for being BRCA2 mutation carrier, the patient endorsed unintentional weight loss and worsening type 2 diabetes mellitus. The second patient’s pancreatic cancer was also found at baseline screening EUS at stage 1A. The patient was 70-years-old at baseline screening, and was referred for EUS based on a recent diagnosis of Lynch Syndrome.
4. Discussion
Data from this single-center study show that it is difficult to achieve high adherence rates for annual pancreatic cancer screening of high-risk individuals as recommended by the latest CAPS, AGA, and ASGE guidelines. It is simultaneously difficult to determine modifiable barriers to non-adherence. In our analysis involving nine demographic and clinical characteristics, there were only two significant differences between the group with complete annual adherence and the other two, less adherent, groups – screening modality and age. We hypothesize that reluctance to undergo sedation and more flexible schedules in older, possibly retired, patients may be contributing factors, but ultimately we need descriptive, interview-based studies to delineate the barriers to annual surveillance in these less adherent groups.
Our data otherwise showed no difference among our three groups by other analyzed characteristics. Even clinical characteristics including personal history of other cancer or strong family history of pancreatic cancer, which anecdotally are often found to be motivators for cancer screening, did not prove to be a contributor to adherence disparities. Area deprivation index (ADI), the most validated zip code-based measure of socioeconomic disadvantage, has previously been linked to disparities in therapy delivery in pancreatic cancer, but had no association with lower adherence in our analysis [16]. In our cohort, we diagnosed 2 PDAC patients, one at stage IV and the other at stage IA. Both patients were referred for baseline screening due to a recently discovered hereditary mutation. Both patients underwent their baseline screening in their 70s. Though a small sample size, this finding from our cohort suggests the importance of early cascade testing.
Our study has several strengths. While there are many analyses of much larger HRI cohorts, to our knowledge this study is one of a few that focuses on real-world adherence and barriers to pancreatic cancer HRI screening guidelines [4,6,10,11,17]. Our patient population is also quite diverse compared to the larger HRI pancreas cohorts, with a particularly strong representation of Asian and Hispanic HRIs. Our cohort in fact has more HRI Asians who have undergone screening and the same number of Hispanics compared to the largest multi-center HRI cohort from the United States [9]. We assessed non-English speaking as a potential barrier but found that it was not a significant factor in adherence. It is otherwise unclear whether there are other aspects of our cohort’s diversity that influence adherence patterns.
There are limitations to our study. Our cohort is relatively small compared to larger, multicenter cohorts that have previously been mentioned. Our small sample size limits the conclusions that can be made from our analysis. Patients who were lost to follow-up from both Stanford hospitals and Care Everywhere-connected healthcare centers were assumed to have not undergone pancreatic cancer surveillance during those years. This assumption might exaggerate our low adherence rates. Though our study shows low overall adherence rates, we were not able to identify a definite modifiable barrier associated with lower adherence. There are potentially important modifiable behavioral, financial, healthcare system, and other clinical/demographic barriers that were not analyzed in this study; scheduling difficulties, the lack of an automated system for reminding patients, competing risks, financial and transportation barriers may be important factors for obtaining screening that were not assessed. We attempted to explore socioeconomic disadvantage as a potential barrier using the area deprivation index (ADI). Unfortunately, we found that our population is quite homogenous with regards to ADI and thus we likely did not have a broad enough distribution to adequately assess this factor. Additionally, while our study attempts to identify barriers to adherence, our analysis does not provide guidance on how to improve surveillance. EHR reminders for primary care physicians, automated patient portal reminders, and education for general practitioners are possible actionable methods for improving adherence that will need to be studied in the future. Future studies will focus on delineating barriers to annual surveillance in our less adherent groups through interviews and implementing methods for improving adherence.
Footnotes
Disclosure statement
No authors have any conflicts of interest.
References
- [1].Siegel RL, Kratzer TB, Giaquinto AN, Sung H, Jemal A. Cancer statistics, 2025. CA Cancer J Clin 2025. Jan-Feb;75(1):10–45. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [2].Corral JE, Mareth KF, Riegert-Johnson DL, Das A, Wallace MB. Diagnostic yield from screening asymptomatic individuals at high risk for pancreatic cancer: a meta-analysis of cohort studies. Clin Gastroenterol Hepatol 2019. Jan;17(1): 41–53. [DOI] [PubMed] [Google Scholar]
- [3].Wood LD, Canto MI, Jaffee EM, Simeone DM. Pancreatic cancer: pathogenesis, screening, diagnosis, and treatment. Gastroenterology 2022. Aug;163(2): 386–402.e1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].Vasen H, Ibrahim I, Ponce CG, Slater EP, Matthäi E, Carrato A, et al. Benefit of surveillance for pancreatic cancer in high-risk individuals: outcome of long-term prospective follow-up studies from three European expert centers. J Clin Oncol 2016. Jun 10;34(17):2010–9. [DOI] [PubMed] [Google Scholar]
- [5].Canto MI, Almario JA, Schulick RD, Yeo CJ, Klein A, Blackford A, et al. Risk of neoplastic progression in individuals at high risk for pancreatic cancer undergoing long-term surveillance. Gastroenterology 2018. Sep;155(3): 740–751.e2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [6].Overbeek KA, Levink IJM, Koopmann BDM, Harinck F, Konings ICAW, Ausems MGEM, et al. Long-term yield of pancreatic cancer surveillance in high-risk individuals. Gut 2022. Jun;71(6):1152–60. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [7].Overbeek KA, Goggins MG, Dbouk M, Levink IJM, Koopmann BDM, Chuidian M, et al. Timeline of development of pancreatic cancer and implications for successful early detection in high-risk individuals. Gastroenterology 2022. Mar;162(3):772–785.e4. [DOI] [PubMed] [Google Scholar]
- [8].Canto MI, Kerdsirichairat T, Yeo CJ, Hruban RH, Shin EJ, Almario JA, et al. Surgical outcomes after pancreatic resection of screening-detected lesions in individuals at high risk for developing pancreatic cancer. J Gastrointest Surg 2020. May;24(5):1101–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [9].Paiella S, Salvia R, De Pastena M, Pollini T, Casetti L, Landoni L, et al. Screening/surveillance programs for pancreatic cancer in familial high-risk individuals: a systematic review and proportion meta-analysis of screening results. Pancreatology 2018. Jun;18(4):420–8. [DOI] [PubMed] [Google Scholar]
- [10].Dbouk M, Katona BW, Brand RE, Chak A, Syngal S, Farrell JJ, et al. The multicenter cancer of pancreas screening study: impact on stage and survival. J Clin Oncol 2022. Oct 1;40(28):3257–66. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [11].Katona BW, Mahmud N, Dbouk M, Ahmad N, Chhoda A, Dudley B, et al. COVID-19 related pancreatic cancer surveillance disruptions amongst high-risk individuals. Pancreatology 2021. Apr 20;21(6):1048–51. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [12].Goggins M, Overbeek KA, Brand R, Syngal S, Del Chiaro M, Bartsch DK, et al. Management of patients with increased risk for familial pancreatic cancer: updated recommendations from the International Cancer of the Pancreas Screening (CAPS) Consortium. Gut 2020. Jan;69(1):7–17. 10.1136/gutjnl-2019-319352. Epub 2019 Oct 31. Erratum in: Gut. 2020 Jun;69(6):e3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [13].Aslanian HR, Lee JH, Canto MI. AGA clinical practice update on pancreas cancer screening in high-risk individuals: expert review. Gastroenterology 2020. Jul;159(1):358–62. [DOI] [PubMed] [Google Scholar]
- [14].Sawhney MS, Calderwood AH, Thosani NC, Rebbeck TR, Wani S, Canto MI, et al. ASGE guideline on screening for pancreatic cancer in individuals with genetic susceptibility: summary and recommendations. Gastrointest Endosc 2022. May;95(5):817–26. [DOI] [PubMed] [Google Scholar]
- [15].Kind AJH, Buckingham WR. Making neighborhood-disadvantage metrics accessible - the neighborhood Atlas. N Engl J Med 2018. Jun 28;378(26): 2456–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [16].Mora J, Krepline AN, Aldakkak M, Christians KK, George B, Hall WA, et al. Adjuvant therapy rates and overall survival in patients with localized pancreatic cancer from high Area Deprivation Index neighborhoods. Am J Surg 2021. Jul;222(1):10–7. [DOI] [PubMed] [Google Scholar]
- [17].Klatte DCF, Boekestijn B, Wasser MNJM, Feshtali Shahbazi S, Ibrahim IS, Mieog JSD, et al. Pancreatic cancer surveillance in carriers of a germline CDKN2A pathogenic variant: yield and outcomes of a 20-year prospective follow-up. J Clin Oncol 2022. Oct 1;40(28):3267–77. [DOI] [PubMed] [Google Scholar]