Current challenges in screening for pancreatic ductal adenocarcinoma: an appraisal of the METAPAC study

Overview

Pancreatic ductal adenocarcinoma (PDAC) remains lethal despite decades of research (1). This can be attributed to asymptomatic, early systemic dissemination and treatment resistance (2). Technical and oncological improvements to surgical care and the introduction of multiagent chemotherapy has modestly improved survival (1). While most patients present with advanced disease, those diagnosed with resectable disease achieve more favorable outcomes after surgery. Thus, growing emphasis has been placed on early detection.

This was the rationale behind the recently published METAPAC study, where the screening performance of two plasma metabolomic signatures was investigated in a cohort of 1,129 patients including 489 cases of PDAC. These signatures were developed to rule out new onset PDAC in high-risk cohorts (1% annual incidence) to be selected for regular surveillance (3). To overcome sample size constraints, the authors used an innovative enrichment strategy to enable a precise estimate of test performance under the assumption that this would generalize to the surveillance setting of interest (i.e., the study cohort was not a sample from a true screening population). The study avoids many of the limitations that have historically hindered biomarker discovery and their clinical integration and builds on robust preceding work to minimize the chances of false discovery (4-7).

The results of METAPAC must be interpreted with care because the development of effective screening programs is fraught with theoretical nuances and practical problems that can be subtle. These include general limitations of cancer screening as well as factors specific to the aggressive biology of PDAC.

Effective screening for cancers is inherently difficult

Some innate limitations of screening for cancers are evident when considering current recommendations from major societies for breast, colorectal, lung, and prostate cancer. At present, these are “best-case scenarios” where risk benefit ratios are optimal. Observational studies supporting the efficacy of these screening programs are affected by lead- and length-time biases that exaggerate benefit. For most indications, randomized trials have detected only small absolute improvements in disease-specific survival (DSS) and evidence for overall survival (OS) benefits remains inconclusive at best (8). Conversely, the physical and psychological risks associated with the cascade of diagnostics and interventions that follow a positive screening test remain underappreciated (8,9). A degree of prior skepticism regarding the potential benefits of screening is thus warranted. Readers should not approach the results of METAPAC having implicitly assumed that screening already “works” for other cancers.

Furthermore, for diseases like melanoma and lung cancer, growing discordance has been observed between rising rates of screening-related diagnosis versus stable death rates, even when accounting for improved treatment options (10,11). This suggests that the biological behavior of cancers that present clinically cannot always be extrapolated to those discovered by screening in healthy populations, even in the setting of radiologic or histologic similarity (12). Therefore, assuming the transportability of screening performance or efficacy across different populations may not be prudent and is a limitation of METAPAC’s enrichment strategy.

How early is early enough for PDAC?

Screening is effective where (I) the disease dictates patients’ prognosis; (II) effective interventions to improve outcomes are available and the chance of cure diminishes with later diagnosis; and (III) minimal harm is caused by overdiagnosis. The interplay of these factors presents a unique calculus when considering the utility of screening for PDAC when compared to other cancers.

Virtually all patients who present clinically and are diagnosed with PDAC will die of their disease. This means that (I) is nearly universally true and applicable and makes a strong case for the potential value of screening. In PDAC, (II) is particularly true of the role of surgery. Mahajan et al. note that patients with resected PDAC have a better prognosis, with 5-year OS [long-term survival (LTS)] of up to 43% in some trials. However, only a minority of patients with PDAC are able to undergo resection (3,13). Moreover, the prolonged survival noted in such cohorts of PDAC, like that cited by the authors, reflects strict patient selection in some adjuvant-setting trials and are optimistic versus the guarded prognosis observed in all-comer studies (14-17). Nonetheless, tools to increase the proportion of patients who undergo surgery could improve outcomes (3). However, this logic is further subject to nuances that place a ceiling on the magnitude of potential benefit.

First, indirect evidence suggests that clinically detectable PDACs progress rapidly and can become unresectable in a 1–2 years interval (18,19). This means that, for the population of patients with chronic pancreatitis or indeterminate pancreas lesions that the authors discuss in their conclusions, there is a limited window of opportunity for effective screening to sample patients with cancer that remains amenable to resection (19,20).

Second, patterns of recurrence and mechanistic studies have established that the systemic spread of PDAC often precedes the clinical presentation of even “localized” cases (1,21). Thus, nearly all patients with resected PDAC need adjuvant systemic therapy and most still succumb to metastases due to the relatively poor efficacy of chemotherapy (17,22,23). In fact, one of the proposed merits of neoadjuvant therapy over upfront resection for surgical candidates is that it may lead to “reduced” rates of futile surgery in patients with chemo-resistant disease.

Ultimately, while early detection might improve resection rates, realistic expectations should be maintained regarding the magnitude of survival benefit that screening could deliver in the case of PDAC. When considering the authors’ interpretation of METAPAC as evidence of the biomarkers’ improved performance over carbohydrate antigen 19-9 (CA19-9), one must note that only a small subset of these patients may actually stand to have improved clinical outcomes.

Finally, based on lessons learnt from the surveillance and management of pancreatic cysts, achieving (III) is difficult and casting too wide a net in the hope of pre-empting invasive cancer is itself associated with high morbidity and costs due to diagnostics and interventions (24,25). The author’s discussion of the need to target a high-risk population with a test that maintains adequate specificity represents strong foresight. Interestingly, however, several of the challenges pertaining to PDAC that have been illustrated, including rapid progression of clinically detectable lesions and early metastases, are less applicable to precancerous lesions like pancreatic cysts. The risk stratification of such lesions under observation may represent a promising avenue in which to investigate the performance of the i- and m-metabolic signatures.

The right test for the right patients

The adequacy of the biomarkers in METAPAC to exclude PDAC in a high-risk surveillance setting has been debated. Though sensitivity is often discussed as a measure of the ability to rule out disease, specificity and pretest probability also contribute to the negative predictive value (NPV) of a test (26).

One reason for the importance given to sensitivity and specificity are that these are often perceived as pure test characteristics that are fixed once cutoffs are selected, but this is not the case in the absence of a sharp binary split between healthy and diseased status, i.e., spectrum bias. In the continuum of progression from health to terminal, incurable illness, the biology of PDAC and its measurable/testable manifestations are dynamic. As the authors state, CA19-9 has no sensitivity in non-secretors regardless of calibration and loses specificity in the presence of benign biliary obstructions. Cross-sectional imaging has poor sensitivity for occult (by definition) malignancy even though these individuals do have “cancer” at that time. These considerations raise important questions about the generalizability of the results observed in METAPAC.

One of the strengths of metabolomic signatures that the authors present is their status as a relatively “high-level” reflection of the disease state and metabolism, as opposed to “low-level” genomics or transcriptomics (3,27). However, this means that the results of METAPAC may not apply to patients without computed tomography (CT)-apparent lesions in earlier stages of cancer where metabolism may be different. This reflects a form of spectrum bias. If the test can only exclude CT-apparent localized disease, it may miss patients who are in the window of opportunity for screening and early intervention discussed previously. Instead, as the authors state, the most immediate clinical utility of these metabolomic signatures may lie not in screening asymptomatic high-risk individuals, but in triaging patients with indeterminate pancreatic lesions or precancerous masses (e.g., cystic lesions) identified on imaging.

These questions also relate to the enrichment strategy used by the authors, who should be commended for devising a feasible study which would otherwise have required an unrealistically large sample size. The authors designed METAPAC such that the anticipated frequency of PDAC was to be 20% but with a prevalence of patient risk factors in similar proportions to their target population. However, it is important to recall that most known risk factors for PDAC do not massively increase the chances of developing the disease, i.e., the relative risks of smoking, diabetes, and BMI (in 5 unit increments) are 1.8, 1.5, and 1.1 (1). New onset diabetes and familial pancreas cancer are associated with age standardized incidence ratios of 1.5 and 4.9, respectively. The baseline incidence of PDAC in the general populace is low and these are not large effect sizes. Thus, usual clinical risk factors probably do not sufficiently account for why some people develop PDAC, even if they can be used to delineate “high-risk” cohorts of individuals (often with low incidences of PDAC nonetheless) (28).

The METAPAC population consisted of patients who had presented to surgical or gastroenterological care and had undifferentiated pancreas lesions on CT scan. Patients with a symptomatic and grossly visible pancreatic mass are a fundamentally different population from those undergoing screening, even if the latter are “high-risk” individuals as discussed previously. The validity of extrapolating these findings to true surveillance populations depends on the unproven assumption that the biological behavior of PDAC and its metabolomic signature are consistent across these settings. In other words, the absence of evidence for differences in known risk factors between the study cohort and population cannot be taken as evidence of absence of meaningful biological differences between the two.

The previously discussed “high-level” nature of metabolomics supports the plausibility of the biomarkers performing differently in a surveillance cohort where patients do not already have clinical disease/lesions at baseline. The authors’ group has presented some data that, within the spectrum of resectable disease and beyond, tumor burden did not significantly affect the PDAC metabolic signature, but the stability of the signature should be studied further in the same population in whom the test is aimed for deployment (29).

Relating performance metrics and decisions

Although performance characteristics are insufficient to conclude that a biomarker is clinically useful, the metrics reported in METAPAC merit discussion.

In all stages, the i-metabolic (sensitivity: 67.5%; specificity: 90.4%) and m-metabolic (sensitivity: 59.9%; specificity: 93.6%) signatures produced 43 and 55 fewer false positives than CA19-9 (sensitivity: 81.8%; specificity: 79.1%) alone but missed 36 and 78 cases of PDAC each. However, the significance of false positives and negatives is not the same in this setting. It is important to note that, though it is the current “gold standard” blood biomarker for PDAC, CA19-9 is not used to definitively diagnose or exclude PDAC (26,29). As the authors state, false positives would still undergo additional diagnostic testing to confirm cancer as well as to adequately stage the disease for treatment planning and prognostication (although the patient and resource costs of false positives should not be underestimated). However, while a false negative CA19-9 would also not usually lead to major diagnostic ramifications, the metabolomic tests under study are being proposed as tests to exclude PDAC. Missing such a critical diagnosis may be unacceptable to patients and clinicians who have opted to participate in surveillance. Thus, in either scenario of a positive or negative test in the study cohort, patients would probably need additional testing. Serial testing as part of the surveillance strategy may not be a solution to this challenge as, given the rapid progression of PDAC from detectable to unresectable stages as discussed previously, the opportunity for early, actionable detection may lapse.

The study reports that the NPVs at a 20% prevalence for the i-metabolic, m-metabolic, and CA19-9 tests were 91.7%, 90.3%, and 94.5%, i.e., similar. At the cohort prevalence these values were 78.4%, 75.3%, and 75.3%. The authors also assessed the transportability of their test to a separate cohort of patients with new-onset diabetes where 3 of 242 patients were diagnosed with PDAC. Using the same test characteristics as before, the i-metabolic, m-metabolic, and CA19-9 tests would yield NPVs of 99.6%, 99.5%, and 99.7%. At a low prevalence, it appears the signatures and CA19-9 perform similarly to exclude PDAC. The former did consistently yield superior positive predictive values (PPVs) than CA19-9 which should be acknowledged.

Similar results were observed in the subset of resectable tumors, which are considered the actionable cases. Here, at a 20% prevalence, the NPVs for the i-metabolic, m-metabolic, and CA19-9 tests were 92.2%, 90.6%, and 98.1%, i.e., in favor of CA19-9. In the cohort prevalence group, these values were 95.4%, 93.7%, and 80.6%. In the new-onset diabetes setting, these values would roughly be 99.6%, 99.5%, and 99.9%.

One additional point of concern regarding the analysis of resectable patients was the disproportionately high sensitivity of CA19-9 in this subgroup. CA19-9 is typically less than 97% sensitive for resectable PDAC and this datapoint was surprising. It would be interesting if a future publication by the authors were able to explain this outlier through features of the patients included, the methods of biomarker measurement or data handling and analysis, chance, or other factors.

In short, these data indicate that while aiming for a highly specific test to avoid submitting cancer-free patients to unnecessary workup (high PPV) was valid, the authors’ aim to more reliably exclude cancer (NPV) in the target surveillance population is an extremely ambitious goal. However, the value of the high specificity of these metabolomic signatures should not be understated. At a surveillance prevalence of 1%, the PPVs of the i-metabolic and m-metabolic signatures would be about 6.6% and 8.6% (versus 3.8% with CA19.9). PPVs of 6.6–8.6% are quite comparable to those achieved by mammography and reflect good performance for a screening test. However, to then determine the clinical value of the test to include, rather than exclude, pancreatic cancer in a high-risk surveillance cohort, one would need to consider the multitude of factors risks, benefits, and constraints discussed thus far.

Future avenues

The authors of METAPAC adhered to high standards of biomarker research and designed an informative experiment that was otherwise infeasible. They attempted to validate biomarkers to exclude PDAC in high-risk patients while maintaining low false-positive rates. Overall, the test has better performance characteristics than CA19-9 alone and represents a step towards the early detection of PDAC and risk stratification of pancreatic masses. The credibility of the test may be enhanced if its stability across the phases of PDAC progression were established and if decision curve analysis were to demonstrate its utility (30). However, given that successful early detection, by definition, increases the measured survival duration of patients (since T0 is brought forward in time, i.e., lead time bias), the only practical but definitive way to prove clinical benefit remains a pragmatic randomised trial.

Acknowledgments

None.

Footnote

Provenance and Peer Review: This article was commissioned by the editorial office, Translational Gastroenterology and Hepatology. The article has undergone external peer review.

Peer Review File: Available at https://tgh.amegroups.com/article/view/10.21037/tgh-2026-0049/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tgh.amegroups.com/article/view/10.21037/tgh-2026-0049/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Stoop TF, Javed AA, Oba A, et al. Pancreatic cancer. Lancet 2025;405:1182-202. Erratum in: Lancet 2025;405:1742.
2. Javed AA, Mahmud O, Fatimi AS, et al. Predictors for Long-Term Survival After Resection of Pancreatic Ductal Adenocarcinoma: A Systematic Review and Meta-Analysis. Ann Surg Oncol 2024;31:4673-87. [Crossref] [PubMed]
3. Mahajan UM, Oehrle B, Goni E, et al. Validation of two plasma multimetabolite signatures for patients at risk of or with suspected pancreatic ductal adenocarcinoma (METAPAC): a prospective, multicentre, investigator-masked, enrichment design, phase 4 diagnostic study. Lancet Gastroenterol Hepatol 2025;10:634-47. [Crossref] [PubMed]
4. Benjamini Y, Hochberg Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. J R Stat Soc Series B Stat Methodol 1995;57:289-300.
5. LaLonde A, Nissen SE. Multiplicity Control in Clinical Trials. NEJM Evid 2025;4:EVIDctw2400393.
6. Unger JM, Mazza GL, Elsaid MI, et al. When to adjust for multiplicity in cancer clinical trials. J Natl Cancer Inst Monogr 2025;2025:3-9. [Crossref] [PubMed]
7. Srivastava S, Wagner PD. The Early Detection Research Network: A National Infrastructure to Support the Discovery, Development, and Validation of Cancer Biomarkers. Cancer Epidemiol Biomarkers Prev 2020;29:2401-10. [Crossref] [PubMed]
8. Bretthauer M, Wieszczy P, Løberg M, et al. Estimated Lifetime Gained With Cancer Screening Tests: A Meta-Analysis of Randomized Clinical Trials. JAMA Intern Med 2023;183:1196-203. [Crossref] [PubMed]
9. Lucas AL, Kastrinos F. Screening for Pancreatic Cancer. JAMA 2019;322:407-8. [Crossref] [PubMed]
10. Goldsmith SM. The Rapid Rise in Cutaneous Melanoma Diagnoses. N Engl J Med 2021;384:e54. [Crossref] [PubMed]
11. Gao W, Wen CP, Wu A, et al. Association of Computed Tomographic Screening Promotion With Lung Cancer Overdiagnosis Among Asian Women. JAMA Intern Med 2022;182:283-90. [Crossref] [PubMed]
12. Cancer screening: How testing can find more disease but not make us healthier with Gil Welch - YouTube. [cited 2026 Mar 4]. Available online: https://www.youtube.com/watch?v=cmoaJMq-DL8
13. Conroy T, Castan F, Lopez A, et al. Five-Year Outcomes of FOLFIRINOX vs Gemcitabine as Adjuvant Therapy for Pancreatic Cancer: A Randomized Clinical Trial. JAMA Oncol 2022;8:1571-8. [Crossref] [PubMed]
14. Conroy T, Hammel P, Hebbar M, et al. FOLFIRINOX or Gemcitabine as Adjuvant Therapy for Pancreatic Cancer. N Engl J Med 2018;379:2395-406. [Crossref] [PubMed]
15. Sohal DPS, Duong M, Ahmad SA, et al. Efficacy of Perioperative Chemotherapy for Resectable Pancreatic Adenocarcinoma: A Phase 2 Randomized Clinical Trial. JAMA Oncol 2021;7:421-7. [Crossref] [PubMed]
16. Mailankody S, Prasad V. Overall Survival in Cancer Drug Trials as a New Surrogate End Point for Overall Survival in the Real World. JAMA Oncol 2017;3:889-90. [Crossref] [PubMed]
17. Javed AA, Rompen IF, van Goor IWJM, et al. Poor Prognostic Factors in Long-Term Survivors of Resected Pancreatic Ductal Adenocarcinoma: An International, Multicenter Cohort Study. Ann Surg 2024; Epub ahead of print. [Crossref]
18. Gyawali B, Booth CM. Treatment of metastatic pancreatic cancer: 25 years of innovation with little progress for patients. Lancet Oncol 2024;25:167-70. [Crossref] [PubMed]
19. Yu J, Blackford AL, Dal Molin M, et al. Time to progression of pancreatic ductal adenocarcinoma from low-to-high tumour stages. Gut 2015;64:1783-9. [Crossref] [PubMed]
20. Lennon AM, Hruban RH, Klein AP. Screening for Pancreatic Cancer-Is There Hope? JAMA Intern Med 2019;179:1313-5. [Crossref] [PubMed]
21. Kleeff J, Korc M, Apte M, et al. Pancreatic cancer. Nat Rev Dis Primers 2016;2:16022. [Crossref] [PubMed]
22. Javed AA, Habib A, Mahmud O, et al. Prognostic factors in localized pancreatic ductal adenocarcinoma after neoadjuvant therapy and resection: a systematic review and meta-analysis. J Natl Cancer Inst 2025;117:840-67. [Crossref] [PubMed]
23. Rompen IF, Levine J, Habib JR, et al. Progression of Site-specific Recurrence of Pancreatic Cancer and Implications for Treatment. Ann Surg 2024;280:317-24. [Crossref] [PubMed]
24. Habib JR, Fatimi AS, Mahmud O, et al. Adjuvant therapy after resection of intraductal papillary mucinous neoplasm-derived pancreatic cancer: A systematic review and meta-analysis. Cancer Treat Rev 2025;138:102969. [Crossref] [PubMed]
25. Mahmud O, Fatimi AS, Grewal M, et al. Risk of pancreatic cancer and high-grade dysplasia in resected main-duct and mixed-type intraductal papillary mucinous neoplasms: A prevalence meta-analysis. Eur J Surg Oncol 2025;51:109742. [Crossref] [PubMed]
26. Clinicians’ Misunderstanding of Probabilities Makes Them Like Backwards Probabilities Such As Sensitivity, Specificity, and Type I Error – Statistical Thinking. [cited 2026 Mar 4]. Available online: https://www.fharrell.com/post/backwards-probs/
27. Mannucci A, Cavestro GM, Goel A. A DEF perspective on the METAPAC study. Lancet Gastroenterol Hepatol 2025;10:879. [Crossref] [PubMed]
28. Blackford AL, Canto MI, Dbouk M, et al. Pancreatic Cancer Surveillance and Survival of High-Risk Individuals. JAMA Oncol 2024;10:1087-96. [Crossref] [PubMed]
29. Mayerle J, Kalthoff H, Reszka R, et al. Metabolic biomarker signature to differentiate pancreatic ductal adenocarcinoma from chronic pancreatitis. Gut 2018;67:128-37. [Crossref] [PubMed]
30. Decision curve analysis for quantifying the additional benefit of a new marker – Statistical Thinking. [cited 2026 Mar 4]. Available online: https://www.fharrell.com/post/addmarkerdca/