MUC5AC belongs to a family of large, heavily glycosylated proteins that form the gel-like mucus protecting the linings of the stomach, airways and other organs. In a healthy adult pancreas, MUC5AC is essentially absent from the ductal cells that give rise to most pancreatic cancers. But during malignant transformation, something striking happens: the cells switch on the MUC5AC gene and begin flooding their surroundings with this secreted mucin. The change appears early, in premalignant lesions such as pancreatic intraepithelial neoplasia (PanIN), intraductal papillary mucinous neoplasms (IPMN) and mucinous cystic neoplasms (MCN), and it persists as tumors progress to invasive carcinoma. That pattern has long suggested that MUC5AC could serve as a molecular flag for disease, but individual studies had produced a confusing range of accuracy estimates, leaving the field without a clear consensus on its true diagnostic value.

To resolve that uncertainty, the research team conducted a rigorously protocol-driven synthesis, registered in advance with PROSPERO under ID CRD42023495046. They searched PubMed, Embase, Web of Science and Google Scholar for studies evaluating MUC5AC as a diagnostic test in pancreatic disease, then assessed methodological quality using the QUADAS-2 framework, the standard tool for appraising diagnostic accuracy research. The final dataset was enormous by biomarker-study standards: 55 studies encompassing 2,804 pancreatic cancer cases, 1,088 precursor lesions, 1,048 patients with benign pancreatic diseases such as chronic pancreatitis, and 1,022 healthy controls. Rather than combining results with naive averaging, the authors applied a bivariate random effects model, the gold-standard statistical approach for diagnostic meta-analysis, which simultaneously models sensitivity and specificity while accounting for the threshold effects and between-study heterogeneity that plague simpler pooling methods.

The headline results are striking. When pathologists measured MUC5AC directly in tissue samples, the biomarker distinguished pancreatic cancer from healthy pancreas with an area under the receiver operating characteristic curve (AUC) of 0.94, on a scale where 1.0 represents perfect discrimination. Sensitivity came in at 0.70, meaning the test caught roughly seven in ten cancers, while specificity reached an exceptional 0.98, meaning false alarms were vanishingly rare. Against benign pancreatic diseases, a far harder comparison because inflammatory conditions can mimic cancer in almost every other assay, tissue MUC5AC still achieved an AUC of 0.94, with sensitivity of 0.65 and specificity of 0.89. For a disease in which distinguishing chronic pancreatitis from malignancy remains a daily clinical dilemma, those numbers represent a genuinely meaningful advance.

Even more intriguing are the findings for precursor lesions. Tissue-based MUC5AC separated precancerous lesions from healthy pancreas with an AUC of 0.95, sensitivity of 0.79 and specificity of 0.98, and it performed spectacularly against benign disease, reaching a sensitivity of 0.97 and specificity of 0.93. In practical terms, this means that nearly every precursor lesion in the analyzed studies expressed MUC5AC, while almost none of the benign tissues did. Because the precursor lesions analyzed include the mucinous cysts and ductal changes that clinicians routinely encounter in pancreatic imaging and biopsy, the result suggests MUC5AC staining could help resolve one of the most common and consequential questions in pancreatic pathology: whether a suspicious lesion is truly at risk of becoming cancer or merely an innocent bystander.

The meta-analysis also delivered an important negative result that is just as scientifically informative. MUC5AC completely failed to separate invasive pancreatic cancer from its own precursors, producing an inverted AUC of 0.29 with specificity collapsing to 0.08. The reason is biologically coherent: because MUC5AC is already switched on during the precursor stage, virtually both the cancer and the lesion it came from stain positive, leaving no discriminative signal between them. For clinicians, this defines the biomarker’s proper role with unusual clarity. MUC5AC is not a tool for staging progression within the neoplastic sequence; it is a tool for identifying the neoplastic sequence itself, flagging tissue that has entered the malignant pathway regardless of how far along it has traveled.

Perhaps the most clinically consequential finding concerns early-stage disease. When the authors stratified cancers by stage, MUC5AC differentiated early-stage pancreatic cancer from benign disease with an AUC of 0.82, sensitivity of 0.61 and specificity of 0.85, while late-stage cancers yielded an almost identical AUC of 0.82. The equivalence matters because most existing biomarkers, including the widely used CA 19-9, perform substantially worse in early disease, precisely when detection would change survival outcomes. A marker whose accuracy does not degrade in the resectable, potentially curable window is exactly what the pancreatic cancer field has been searching for, and the pooled data suggest MUC5AC holds that property in tissue.

The story is more nuanced when the biomarker leaves the tissue and enters the bloodstream. Blood-based MUC5AC assays, typically using enzyme-linked immunosorbent assays to detect the secreted protein circulating in patient serum, showed a lower but still moderate ability to distinguish pancreatic cancer from benign disease, with an AUC of 0.83, sensitivity of 0.63 and specificity of 0.91. That performance falls short of the tissue results, which is unsurprising given that circulating mucins are diluted, fragmented and modified in the blood. Even so, an AUC above 0.80 from a simple blood test would be competitive with existing serological markers, and the authors are careful to frame this as a promising but immature avenue. Their conclusion explicitly calls for more studies to establish whether MUC5AC can mature into a non-invasive diagnostic biomarker, rather than overclaiming clinical readiness.

The methodological transparency of the analysis strengthens confidence in these conclusions. The team examined sources of heterogeneity through subgroup and meta-regression analyses, ran influence analyses to test whether removing very low-quality studies changed the pooled estimates, and formally assessed publication bias using Deeks’ funnel plots, the appropriate test for diagnostic accuracy syntheses. Threshold effect assessments checked whether apparent differences between studies simply reflected different cutoff values for calling a test positive. The work also benefited from an unusually international collaboration, with search strategy expertise contributed by Paul A. Bain of Harvard’s Countway Library and quality assessment support from Javier Marrugo of the University of Sherbrooke, and funding from the Swedish Research Council, the Knut and Alice Wallenbergs Foundation and the National Natural Science Foundation of China.

What comes next will determine whether MUC5AC moves from pathology slides into routine clinical practice. The immediate, most realistic application is as an immunohistochemical adjunct on biopsy and surgical specimens, helping pathologists classify ambiguous pancreatic lesions and identify patients whose mucinous cysts warrant surveillance or surgery. The longer-term ambition, a blood test built on circulating MUC5AC, will require larger prospective studies, standardized assay platforms and head-to-head comparisons against CA 19-9 and emerging markers. But the sheer consistency of the signal across 55 studies, spanning four comparisons and thousands of patients, gives the finding a weight few biomarker candidates ever achieve. For a cancer that kills most of its victims within five years of diagnosis, a molecular beacon that lights up both the tumor and the road leading to it is a discovery worth watching closely.

Subject of Research: Diagnostic accuracy of the MUC5AC biomarker for pancreatic cancer and precursor lesions

Article Title: MUC5AC in pancreatic cancer and precursors: a systematic review and meta-analysis

Article References: Bakry, N. S., Farag, R. S., Meng, X.-Y., & Wang, Q.-L. (2026). MUC5AC in pancreatic cancer and precursors: a systematic review and meta-analysis. BMC Cancer. https://doi.org/10.1186/s12885-026-16968-2

Image Credits: AI Generated

DOI: 10.1186/s12885-026-16968-2

Keywords: pancreatic cancer, MUC5AC, biomarker, diagnostic accuracy, meta-analysis, PanIN, IPMN, mucinous cystic neoplasm, immunohistochemistry, early detection, precursor lesions, BMC Cancer