Screening for Pancreatic Cancer: Updated Evidence Report and Systematic Review for the US Preventive Services Task Force

Importance  Pancreatic adenocarcinoma is the third most common cause of cancer death among men and women in the United States.

Objective  To systematically review benefits and harms of screening for pancreatic adenocarcinoma to inform the US Preventive Services Task Force.

Data Sources  MEDLINE, PubMed, and the Cochrane Collaboration Registry of Controlled Trials, from January 2002 through April 27, 2018; surveillance through March 22, 2019.

Study Selection  Studies of adults with or without risk factors for pancreatic adenocarcinoma (eg, family history of pancreatic cancer, personal history of new-onset diabetes) undergoing imaging-based screening; studies of treatment for adults with screen-detected or asymptomatic pancreatic adenocarcinoma. Included study designs were randomized clinical trials, nonrandomized controlled intervention studies, diagnostic accuracy studies with a reference standard, cohort studies, and case-control studies (for evaluation of harms only). Studies consisting entirely of populations with known genetic syndromes associated with pancreatic cancer were excluded.

Data Extraction and Synthesis  Two investigators independently reviewed abstracts and full-text articles and rated included studies for quality; data were quantitatively analyzed to calculate a pooled diagnostic yield and narratively synthesized.

Main Outcomes and Measures  Mortality, morbidity, or quality of life; diagnostic accuracy of screening tests; any harm of screening or treatment.

Results  Thirteen fair-quality prospective cohort screening studies (N = 1317) conducted predominantly in populations at high familial risk for pancreatic adenocarcinoma were included. No studies reported on the effect of screening on morbidity or mortality or on the effectiveness of treatment for screen-detected pancreatic adenocarcinoma. Although no studies evaluated the diagnostic accuracy of screening tests, all 13 studies reported the diagnostic yield. Yields ranged from 0 to 75 cases per 1000 persons in studies using endoscopic ultrasound, magnetic resonance imaging, and/or computed tomography–based screening. In total, 18 cases of pancreatic adenocarcinoma were detected in 1156 adults at increased familial risk and 0 cases were detected in 161 average-risk adults. In 8 studies (n = 675) assessing procedural harms of screening, no serious harms from initial screening were reported. Two studies (n = 271) found no evidence of psychosocial harms related to screening. Evidence of surgical harms was limited.

Conclusions and Relevance  Imaging-based screening in groups at high familial risk can detect pancreatic adenocarcinoma with limited evidence of minimal harms. However, the effect of screening on morbidity and mortality in groups at high familial risk has not been studied, and no data are available in average-risk populations. There is limited evidence to assess benefits or harms of surgical intervention for screen-detected pancreatic adenocarcinoma.

Pancreatic adenocarcinoma is the third most common cause of cancer death in the United States.¹ The mean 5-year survival rate for patients with early-stage disease was 32% in 2014²; however, more than 80% of incident cases diagnosed between 2005 and 2011 were detected at advanced stages, for which 5-year survival is less than 5%.³ Screening to detect pancreatic cancers and their potential precursor lesions could improve survival if it facilitated surgical resection for early-stage disease. However, since incident pancreatic cancer is rare, with 12.6 new cases per 100 000 people in the United States in 2011-2015,⁴ identifying populations at the highest risk for pancreatic cancer is critical to developing meaningful screening or early detection programs.

In 2004, the US Preventive Services Task Force (USPSTF) recommended against routine pancreatic screening in asymptomatic adults (D recommendation).⁵ This systematic review addresses the benefits and harms associated with screening and treatment of pancreatic adenocarcinoma. It was conducted to support an updated USPSTF recommendation for screening in asymptomatic adults.

This review addressed 5 key questions (KQs) (Figure 1). Methodological details (including study selection, a list of excluded studies, and description of data analyses), as well as detailed results for each study (including descriptions of all screening programs), are available in the full evidence report⁷ at https://www.uspreventiveservicestaskforce.org/Page/Document/UpdateSummaryFinal/pancreatic-cancer-screening1.

All articles included in the previous USPSTF evidence report on screening for pancreatic cancer⁸ were evaluated for inclusion. MEDLINE, PubMed, and the Cochrane Central Register of Controlled Trials were searched from January 1, 2002 to April 27, 2018 (eMethods in the Supplement). The database searches were supplemented by scanning reference lists of existing reviews and primary studies. Ongoing surveillance was conducted through article alerts and targeted searches of high–impact-factor journals identified by the USPSTF⁶ to identify major studies published in the interim. The last surveillance was conducted on March 22, 2019, and identified no new studies.

A single investigator reviewed the titles or abstracts of citations initially identified as of low relevance using key words relating to exclusion criteria. The remaining abstracts were dual-reviewed by 2 independent investigators. From the 2 processes, the remaining full-text articles were reviewed for consistency with prespecified inclusion criteria (eTable 1 in the Supplement). Discrepancies were resolved through consultation with a third investigator.

For key questions on screening (KQ1, KQ2, KQ3), the population of interest was adults 18 years or older with or without risk factors for pancreatic adenocarcinoma (eg, family history of pancreatic adenocarcinoma, personal history of new-onset diabetes, or other risk factors). Studies consisting entirely of persons with confirmed genetic syndromes (eg, Peutz-Jeghers syndrome, Lynch syndrome, hereditary pancreatitis, known mutations in CDKN2A, BRCA1, BRCA2, CTFR, or ATM genes) were excluded. Any imaging-based screening protocol—including endoscopic ultrasound (EUS), magnetic resonance imaging (MRI), or computed tomography (CT)—was included. Studies using biomarker-based initial screening protocols were excluded, since no biomarkers have been validated as screening tests at the time of the review.⁹

For key questions on treatment (KQ4, KQ5), the population of interest was adults with screen-detected, asymptomatic, or incidentally detected pancreatic adenocarcinoma treated with surgical resection with or without chemotherapy or radiation. Study populations with pancreatic adenocarcinoma detected clinically or symptomatically were excluded to focus the review on treatment for screen-detected cancers. Studies eligible for KQ4 needed to have a comparison group of either no treatment or delayed treatment; thus, comparative effectiveness treatment studies were excluded.

Included study designs were randomized or nonrandomized controlled intervention studies (KQ1, KQ3, KQ4, KQ5), diagnostic accuracy studies with a reference standard (KQ2), prospective cohort studies (KQ3, KQ4, KQ5), and case-control studies (KQ3, KQ5).

Outcomes of interest were pancreatic adenocarcinoma–specific morbidity or mortality, all-cause mortality, or quality of life (KQ1); measures of diagnostic accuracy, including sensitivity, predictive value, and diagnostic yield (KQ2); procedural or psychosocial harms of screening (KQ3); morbidity, mortality, or quality of life (KQ4); or any surgical harms (KQ5). For KQ2, additional outcomes of interest were pancreatic adenocarcinoma or its associated precursor lesions, including pancreatic intraepithelial neoplasia, intraductal papillary mucinous neoplasm, and mucinous cystic neoplasm.

Two investigators critically appraised all articles that met inclusion criteria based on the USPSTF design-specific quality criteria (eTable 2 in the Supplement). Each study was rated as good, fair, or poor quality. A good-quality study met all quality criteria. A fair-quality study failed to meet at least 1 criterion but had no known issue that would invalidate its results. Poor-quality studies were those with a major risk of bias and were excluded from this review. The most common reasons for poor-quality exclusion were insufficient information on patient recruitment or the screening process. Disagreements about quality rating were resolved by consensus.

One investigator completed primary data abstraction; a second investigator checked all data for accuracy and completeness.

For each KQ, data were summarized narratively using tables that included details on study design and quality, setting, population, screening program details, length of follow-up, outcomes, and reported harms.

For KQ2, data on diagnostic yield were quantitatively synthesized, as it was not possible to calculate sensitivity and specificity from the included studies. The diagnostic yield of pancreatic adenocarcinoma and 2-sided 95% confidence intervals were calculated assuming binomial distribution; for studies that detected 0 relevant findings, 1-sided 97.5% confidence intervals were calculated. After confirming that the yield of different imaging modalities was similar across studies and none visually appeared to be outliers, a pooled diagnostic yield was calculated and illustrated in forest plots to show the range of effects across studies. Diagnostic yield was calculated for initial screening and, when possible, from initial and repeated screening combined. Diagnostic yield could not be calculated for repeat screenings alone because the number of participants undergoing repeat screenings was not consistently reported across studies. All analyses were conducted in Stata version 15 (StataCorp).

A total of 19 596 abstracts were reviewed (Figure 2), including 2168 citations initially identified as of low relevance. The remaining 17 428 citations were reviewed by 2 independent investigators. From the 2 processes, the team reviewed 824 full-text articles.

Thirteen unique prospective cohort screening studies reported in 24 articles^11-34 and with results for 1317 people (Table 1) were included. All screening populations except 1 small comparison group in 1 study were exclusively persons at elevated familial risk for pancreatic adenocarcinoma, with or without confirmed genetic mutations or syndromes. No studies reported the effect of screening for pancreatic adenocarcinoma on cancer morbidity, mortality, or all-cause mortality (KQ1). All 13 studies reported diagnostic yield of screening tests for pancreatic adenocarcinoma (N = 1317, KQ2); 9 of these studies (18 articles)^11-20,24-31 reported on the procedural harms (n = 675) or psychological harms (n = 271) of screening (KQ3). No studies on the effectiveness of treatment for screen-detected pancreatic adenocarcinoma (KQ4) met inclusion criteria. Six studies (12 articles)^{12-14,16,18-21,23,27,29,34} reported on the harms of treatment of screen-detected pancreatic adenocarcinoma (n = 32, KQ5). The studies were conducted in the United States, Canada, and Europe; all included studies were of fair quality.

Key Question 1. Does screening for pancreatic adenocarcinoma improve cancer morbidity or mortality or all-cause mortality?

Key Question 1a. Does screening effectiveness vary by clinically relevant subpopulations (eg, by age group, family history of pancreatic cancer, personal history of new-onset diabetes, or other risk factors)?

No studies met inclusion criteria for KQ1.

Key Question 2. What is the diagnostic accuracy of screening tests for pancreatic adenocarcinoma?

Thirteen studies reported in 24 articles (n = 1317) met inclusion criteria for KQ2 (Table 1).^11-34 Screening programs used EUS, CT, and/or MRI screening alone or in combination with another screening modality. Studies evaluating more than 1 type of screening reported abnormal results and yield of pancreatic adenocarcinoma by type of test. Follow-up time after initial screening ranged from 12 to 60 months. All studies reported final pathology determined using fine-needle aspiration biopsy, surgery, or both.

Nine studies (n = 885) evaluated EUS-based screening, with yields of pancreatic adenocarcinoma ranging from 0 (97.5% CI, 0.0-16.9) to 68.2 (95% CI, 14.3-186.6) cases per 1000 persons.^{13,14,16-18,23,27,31,33} In 2 studies reporting CT findings (n = 294),^18,23 the yield of CT for pancreatic adenocarcinoma ranged from 0 (97.5% CI, 0.0-16.9) to 12.8 (95% CI, 0.3-69.4) per 1000. Eight studies reported MRI screening results (n = 849), with yields of pancreatic ductal adenocarcinoma ranging from 0 (97.5% CI, 0.0-16.9) to 75.0 (95% CI, 15.7-203.9) cases per 1000 persons.^{16,22-24,27,31,32}

In total, 18 cases of pancreatic adenocarcinoma were detected among 1156 screened persons at increased familial risk: 9 on initial screening (yield, 7.8 per 1000 persons [95% CI, 3.6-14.7]), 8 on repeated screening or during surveillance of abnormal screening results (yield, 15.6 per 1000 persons [95% CI, 9.3-24.5]), and 1 at an unspecified time point (Figure 3). Twelve of 18 cases (66.6%) were detected at stage I or II or classified as resectable, whereas 6 (33.3%) were detected at stage III or IV. One study with 161 screened average-risk adults found no cases of pancreatic adenocarcinoma.¹⁸ Screen-positive results, biopsy rates, and follow-up of screen-negative results were inconsistently reported, prohibiting calculation of diagnostic accuracy.

Eleven of the 13 studies reported the number of precursor lesions, including pancreatic intraepithelial neoplasia, intraductal papillary mucinous neoplasm, and other nonmalignant pancreatic lesions in addition to pancreatic adenocarcinoma. In total, the screening programs identified a total of 38 individuals with intraductal papillary mucinous neoplasm (n = 5), pancreatic intraepithelial neoplasia (n = 13), or both intraductal papillary mucinous neoplasm and pancreatic intraepithelial neoplasia (n = 20). These findings are not considered false-positives because they often serve as indications for surgical resection, the individuals are enrolled in surveillance programs to monitor lesion progression, or both.

Key Question 3. What are the harms of screening for pancreatic adenocarcinoma?

Nine studies met the inclusion criteria for KQ3 (Table 1). Eight of these studies reported on procedural harms from screening (n = 675).^{11,13,14,16-18,27,31} No serious harms from initial screening were reported. One study (n = 216) reported prevalence of 25.5% for mild pain after EUS. Adverse events related to anesthesia were reported in 13 people (6.0%).¹⁷ No harms were reported in 2 studies of 240 people screened with MRI or magnetic resonance cholangiopancreatography (MRCP),^11,27 while 1 person reported a mild reaction to contrast in 1 study of CT screening (n = 78).¹⁸ Of 150 individuals who underwent follow-up testing with endoscopic retrograde cholangiopancreatography (ERCP) across 2 studies,^17,18 15 people (10%) reported acute pancreatitis, 9 of which required hospitalization. One of these studies (n = 24 receiving ERCP) found 2 cases of acute pancreatitis, 1 requiring hospitalization¹⁷; the other study (n = 126 receiving ERCP)¹⁸ found 8 cases (6.3%) of pancreatitis requiring hospitalization (mean hospital stay, 8.25 days) and 5 cases not requiring hospitalization.

Psychosocial harms were assessed in 2 studies, which assessed distress and cancer worry before and after screening. Distress levels remained in normal ranges at all time points in both studies (n = 271).^28,30 In the 1 study assessing cancer worry,³⁰ worry declined steadily over time (Cancer Worry Scale score, 14.4 at baseline and 12.1 at 3 years; difference, 2.3 points [P < .01]; with scores above 12 indicating severe worry levels), indicating a possible benefit to screening. In the other study,²⁸ perceived cancer risk remained stable between prescreening and 3 months’ follow-up.

Key Question 4. Does treatment of screen-detected or asymptomatic pancreatic adenocarcinoma improve cancer mortality, all-cause mortality, or quality of life?

No studies met inclusion criteria for KQ4.

Key Question 5. What are the harms of treatment of screen-detected pancreatic adenocarcinoma?

Harms of surgical treatment were limited, assessed in 6 studies (n = 32).^{13,14,16,18,23,27} Among the 32 people, a total of 7 (25%) experienced a harm from surgery, including diabetes (n = 3), pancreatic fistula (n = 2), stricture of hepaticojejunal anastomosis with cholangitis (n = 1), and other postoperative complications not further specified (n = 1). However, only 3 of the 6 studies assessed harms in all participants, limiting conclusions for this question.

The findings of this evidence review are summarized in Table 2. All included studies represent new evidence since the previous evidence review, which did not identify any studies of screening for pancreatic adenocarcinoma.⁸ A broader summary of the previous and new evidence is provided in Table 3. No studies evaluating mortality and morbidity as an effect of screening met inclusion criteria. There was limited evidence that imaging-based screening can detect pancreatic adenocarcinoma and its precursor lesions in individuals at high familial risk, and limited evidence that screening is associated with minimal to no psychological or procedural harms.

Collectively, the included studies suggest that imaging-based screening in populations at increased familial risk can identify pancreatic adenocarcinoma and may result in stage shift toward earlier stage at detection. A robust body of observational data clearly suggests a survival benefit associated with earlier stage at detection, and surgical resection of early-stage adenocarcinoma further enhances survival.^7,34 However, in the absence of longer-term follow-up data, it is unclear if the available evidence represents a true clinical benefit, different spectrum of disease, or lead-time bias. There was also little evidence to inform sensitivity, specificity, predictive value, or false-positives of screening tests. Similarly, pancreatic surgery is associated with postoperative complication rates of 20% to 50%,⁷ but evidence on the harms of surgery for screen-detected pancreatic adenocarcinoma was very limited in this review.

Detection of pancreatic adenocarcinoma precursor lesions (intraductal papillary mucinous neoplasms or pancreatic intraepithelial neoplasia) was also observed. The detection and removal of precursor lesions may prevent pancreatic adenocarcinoma and could represent a promising way forward for screening. However, in the absence of clear evidence about progression of precursor lesions and assessment of lead time bias, overdiagnosis and harms associated with treatment of precursor lesions remain possibilities. As such, it is unclear if detection and management of precursor lesions results in a decrease in pancreatic adenocarcinoma incidence, morbidity, or mortality.

The applicability of this body of evidence is limited to populations at known elevated risk for pancreatic adenocarcinoma based on family history, noting that the study populations in the included body of evidence were enriched with people with known genetic mutations or syndromes. The implications of these results to other at-risk populations are unknown, including people with new-onset diabetes, smoking history, or chronic pancreatitis.

Identification and risk assessment for people at the highest risk is critical for improving screening programs.³⁵ Only about 10% of pancreatic adenocarcinoma cases have a familial basis; of those, only about 20% are currently attributed to inherited genetic mutations.^36,37 The body of evidence in pancreatic adenocarcinoma would be strengthened with the addition of controlled trials that include screening and usual care groups of people at increased risk for pancreatic adenocarcinoma and the demonstration of improved morbidity or mortality. In the absence of such evidence, research is needed on how to best evaluate the health outcomes of screening using rigorous observational studies and statistical methods. Given the low incidence and high severity of pancreatic adenocarcinoma coupled with the potential survival benefits of early intervention, approaches to identifying individuals at the highest risk and using less invasive screening tests are warranted. More research is also needed on the progression rates of precursor lesions to pancreatic adenocarcinoma, and health outcomes and harms in people with these lesions, as well as incidentally detected cancers. Continued understanding of the harms of screening and treatment, including those associated with the detection of precursor lesions, is also needed.

This review had several limitations. First, it excluded studies with populations solely comprising people with known genetic mutations or syndromes. As such, it should not be interpreted as an estimate of the yield of screening in people with known genetic mutations or syndromes. Second, the review intentionally included only those treatment studies conducted with screen-detected or asymptomatic populations. Third, it did not systematically assess the extensive literature showing survival benefits of surgery for early-stage pancreatic adenocarcinoma and the significant morbidities that can occur during the postoperative period. Fourth, the limited data about harms reported in the included evidence should not be interpreted to suggest that surgical treatment is without risks but rather that the magnitude of these potential harms is not well studied among people with screen-detected disease.

Imaging-based screening in groups at high familial risk can detect pancreatic adenocarcinoma with limited evidence of minimal harms. However, the effect of screening on morbidity and mortality in groups at high familial risk has not been studied, and no data are available in average-risk populations. There is limited evidence to assess benefits or harms of surgical intervention for screen-detected pancreatic adenocarcinoma.

Corresponding Author: Nora B. Henrikson, PhD, MPH, Kaiser Permanente Washington Health Research Institute, 1730 Minor Ave, Ste 1600, Seattle, WA 98101 (Nora.B.Henrikson@kp.org).

Accepted for Publication: April 24, 2019.

Author Contributions: Dr Henrikson had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: All authors.

Acquisition, analysis, or interpretation of data: Henrikson, Aiello Bowles, Blasi, Morrison, Nguyen, Pillarisetty.

Drafting of the manuscript: Henrikson, Aiello Bowles, Blasi, Morrison, Nguyen.

Critical revision of the manuscript for important intellectual content: Henrikson, Aiello Bowles, Blasi, Nguyen, Pillarisetty, Lin.

Statistical analysis: Henrikson, Aiello Bowles.

Obtained funding: Henrikson, Lin.

Administrative, technical, or material support: Aiello Bowles, Blasi, Morrison, Nguyen.

Supervision: Henrikson, Lin.

Conflict of Interest Disclosures: Dr Pillarisetty reported receiving grants from Merck and Ipsen and receiving personal fees from Merck. No other authors reported disclosures.

Funding/Support: This research was funded under contract HHSA2902015000071, Task Order 5, from the Agency for Healthcare Research and Quality (AHRQ), US Department of Health and Human Services, under a contract to support the US Preventive Services Task Force (USPSTF).

Role of the Funder/Sponsor: Investigators worked with USPSTF members and AHRQ staff to develop the scope, analytic framework, and key questions for this review. AHRQ had no role in study selection, quality assessment, or synthesis. AHRQ staff provided project oversight, reviewed the report and manuscript to ensure that the analysis met methodological standards, and distributed the draft for peer review. Otherwise, AHRQ had no role in the conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript findings. The opinions expressed in this document are those of the authors and do not reflect the official position of AHRQ or the US Department of Health and Human Services.

Additional Contributions: We gratefully acknowledge the following individuals for their contributions to this project: Howard Tracer, MD (AHRQ); current and former members of the US Preventive Services Task Force who contributed to topic deliberations; the AHRQ staff; and Evidence-based Practice Center staff members Todd Hannon, MLS, Gabrielle Gundersen, MPH, Vina Graham, and Stephen Perry. USPSTF members, expert consultants, peer reviewers, and federal partner reviewers did not receive financial compensation for their contributions.

Additional Information: A draft version of this evidence report underwent external peer review from 5 content experts (Marcia Irene Canto, MD, Johns Hopkins University School of Medicine; Teresa Brentnall, MD, University of Washington Medical Center; Paul Pinsky, PhD, National Cancer Institute; Philip Prorok, PhD, National Cancer Institute; and Theresa P. Yeo, PhD, Thomas Jefferson University Hospital) and 2 federal partners (the National Cancer Institute and the National Institute of Diabetes and Digestive and Kidney Diseases). Comments from reviewers were presented to the USPSTF during its deliberation of the evidence and were considered in preparing the final evidence review.

Editorial Disclaimer: This evidence report is presented as a document in support of the accompanying USPSTF Recommendation Statement. It did not undergo additional peer review after submission to JAMA.