Protein biomarkers in pancreatic juice and serum for identification of pancreatic cancer

Open AccessPublished:May 07, 2022DOI:https://doi.org/10.1016/j.gie.2022.04.1342

      Background and Aims

      To date, surveillance of high-risk individuals for pancreatic ductal adenocarcinoma (PDAC) has not lived up to expectations, as identification of curable stages through imaging remains challenging. Biomarkers are therefore needed. Pancreatic juice (PJ) may be a promising source, because it is in direct contact with the ductal epithelial lining from which PDAC arises. We aimed to develop a panel of biomarkers from serum and PJ to detect PDAC for future surveillance purposes.

      Methods

      All patients who underwent PJ collection on secretin stimulation at the Erasmus MC were included. Both PJ and serum were evaluated. Protein levels were determined by the Lowry assay. Potential biomarkers (interleukin-8, interferon-γ, neutrophil gelatinase-associated lipocalin [NGAL], mucin 5, subtype AC [MUC5AC], mucin 2, phospholipase A2 group IB) were selected based on previously reported outcomes and assessed with enzyme-linked immunosorbent assay. Serum carbohydrate antigen 19-9 (CA19-9) values were determined by electrochemiluminescence immunoassay.

      Results

      This study included 59 cases and 126 surveilled control subjects (who underwent PJ collection), of whom 71 had a hereditary predisposition (35 genetic, 36 familial) and 55 had (suspected neoplastic) pancreatic cysts. CA19-9 values were available for 53 cases and 48 control subjects. Serum CA19-9, as well as PJ interleukin-8, NGAL and MUC5AC, were associated with PDAC independent of age, gender, and presence of diabetes mellitus. Serum CA19-9 had a significantly higher area under the curve (AUC; .86; 95% confidence interval [CI], .79-.94) than individual PJ markers (AUC, .62-.70). A combination of PJ markers and serum CA19-9 (panel 2: sensitivity 42% [95% CI, 29-57] and specificity 96% [95% CI, 86-100]) did not improve diagnostic performance compared with CA19-9 alone (sensitivity 70% [95% CI, 56-82] and specificity 85% [95% CI, 72-94]).

      Conclusions

      High levels of serum CA19-9 and PJ-derived proteins are associated with PDAC. Prospective surveillance studies including individuals at risk of developing PDAC are required to validate these findings.

      Graphical abstract

      Abbreviations:

      AUC (area under the curve), BMI (body mass index), CA19-9 (carbohydrate antigen 19.9), CBD (common bile duct), CI (confidence interval), DM (diabetes mellitus), FNB (fine-needle biopsy sampling), IFN-γ (interferon-γ), IL-8 (interleukin-8), MPD (main pancreatic duct), MUC2 (mucin 2), MUC5AC (mucin 5, subtype AC), NGAL (neutrophil gelatinase-associated lipocalin), PDAC (pancreatic ductal adenocarcinoma), PJ (pancreatic juice), PLA2G1B (phospholipase A2 group IB), PPV (positive predictive value)
      Pancreatic ductal adenocarcinoma (PDAC) is a lethal disease for which the incidence closely parallels mortality because of late presentation of symptoms and lack of effective therapy for late-stage disease.
      • Siegel R.L.
      • Miller K.D.
      • Jemal A.
      Cancer statistics, 2020.
      Surveillance programs in individuals with a hereditary predisposition have shown that early detection based on imaging is highly challenging, even when imaging modalities are combined. By the time neoplasia becomes detectable, many patients already have advanced disease.
      • Overbeek K.A.
      • Levink I.J.M.
      • Koopmann B.D.M.
      • et al.
      Long-term yield of pancreatic cancer surveillance in high-risk individuals.
      • Vasen H.
      • Ibrahim I.
      • Ponce C.G.
      • 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.
      • Canto M.I.
      • Almario J.A.
      • Schulick R.D.
      • et al.
      Risk of neoplastic progression in individuals at high risk for pancreatic cancer undergoing long-term surveillance.
      Additionally, in surveillance programs, resection of abnormalities that prove to be benign on histologic assessment are common, causing unnecessary harm.
      • Konings I.C.A.W.
      • Canto M.I.
      • Almario J.A.
      • et al.
      Surveillance for pancreatic cancer in high-risk individuals.
      Likely, the best way to improve PDAC survival is by earlier detection. Biomarkers hold promise, because structural cellular changes are expected to occur months to years before a lesion becomes apparent on imaging. Thus, they may complement imaging in ruling out malignant disease.
      • Yachida S.
      • Jones S.
      • Bozic I.
      • et al.
      Distant metastasis occurs late during the genetic evolution of pancreatic cancer.
      To date, carbohydrate antigen 19-9 (CA19-9) is the only PDAC serum marker implemented in clinical practice to assess treatment response and detect recurrence. Despite guideline recommendations,
      European Study Group on Cystic Tumours of the Pancreas. European evidence-based guidelines on pancreatic cystic neoplasms.
      use for surveillance purposes is controversial because higher values (>37 kU/L) are regularly observed in patients with no or low-grade dysplasia (limited specificity).
      • Kim J.R.
      • Jang J.Y.
      • Kang M.J.
      • et al.
      Clinical implication of serum carcinoembryonic antigen and carbohydrate antigen 19-9 for the prediction of malignancy in intraductal papillary mucinous neoplasm of pancreas.
      and low values do not rule out malignant progression (limited sensitivity). Although several other blood-based biomarkers have been proposed, thus far none has lived up to expectations or been implemented.
      • Le Calvez-Kelm F.
      • Foll M.
      • Wozniak M.B.
      • et al.
      KRAS mutations in blood circulating cell-free DNA: a pancreatic cancer case-control.
      ,
      • Meng Q.
      • Shi S.
      • Liang C.
      • et al.
      Diagnostic and prognostic value of carcinoembryonic antigen in pancreatic cancer: a systematic review and meta-analysis.
      For instance, Zhang et al
      • Zhang Y.
      • Yang J.
      • Li H.
      • et al.
      Tumor markers CA19-9, CA242 and CEA in the diagnosis of pancreatic cancer: a meta-analysis.
      reported in a meta-analysis that carcinoembryonic antigen (CEA) as compared with CA19-9 has a lower sensitivity of 40% (95% confidence interval [CI], 37-42) and similar specificity of 81% (95% CI, 79-83) and is therefore not useful in clinical practice. With a specificity of 83% (95% CI, 81-85) and a sensitivity of 68% (95% CI, 66-70), CA-242 has a similar diagnostic performance to that of CA19-9 (sensitivity 75% [95% CI, 73-77] and specificity 78% [95% CI, 75-80]) but is also not clinically implemented.
      Pancreatic juice (PJ) may serve as an alternative biomarker source. Biomarkers determined in juice are expected to be more pancreas-specific because this fluid constitutes a washout of the pancreatic ductal system and has been in close contact with the ductal cells from which PDAC originates. As compared with either PJ collection by pancreatic duct cannulation using endoscopic retrograde pancreatography or tissue sampling with FNA or fine-needle biopsy (FNB) sampling, secretin-stimulated PJ collection from the duodenal lumen is less invasive. Additionally, in contrast to needle biopsy sampling, PJ collection does not rely on a visible mass, and PJ potentially contains information on the complete range of tumor clones.
      • Choi M.H.
      • Mejlænder-Andersen E.
      • Manueldas S.
      • et al.
      Mutation analysis by deep sequencing of pancreatic juice from patients with pancreatic ductal adenocarcinoma.
      ,
      • Levink I.J.M.
      • Nesteruk K.
      • Visser D.I.
      • et al.
      Optimization of pancreatic juice collection: a first step toward biomarker discovery and early detection of pancreatic cancer.
      Based on available data, several potential biomarkers can be identified. PDAC progression is associated with a distinct state of inflammation and altered release of cytokines. Interleukin-8 (IL-8), interferon-γ (IFN-γ), and neutrophil gelatinase-associated lipocalin (NGAL) were shown to be overexpressed in dysplastic and/or PDAC tissue,
      • Kuwada Y.
      • Sasaki T.
      • Morinaka K.
      • et al.
      Potential involvement of IL-8 and its receptors in the invasiveness of pancreatic cancer cells.
      • Detjen K.M.
      • Farwig K.
      • Welzel M.
      • et al.
      Interferon gamma inhibits growth of human pancreatic carcinoma cells via caspase-1 dependent induction of apoptosis.
      • Moniaux N.
      • Chakraborty S.
      • Yalniz M.
      • et al.
      Early diagnosis of pancreatic cancer: neutrophil gelatinase-associated lipocalin as a marker of pancreatic intraepithelial neoplasia.
      and previous results in either serum or PJ demonstrated that increased concentrations of these proteins are able to distinguish PDAC cases from control subjects.
      • Moniaux N.
      • Chakraborty S.
      • Yalniz M.
      • et al.
      Early diagnosis of pancreatic cancer: neutrophil gelatinase-associated lipocalin as a marker of pancreatic intraepithelial neoplasia.
      • Noh K.W.
      • Pungpapong S.
      • Wallace M.B.
      • et al.
      Do cytokine concentrations in pancreatic juice predict the presence of pancreatic diseases?.
      • Shaw V.E.
      • Lane B.
      • Jenkinson C.
      • et al.
      Serum cytokine biomarker panels for discriminating pancreatic cancer from benign pancreatic disease.
      • Kaur S.
      • Chakraborty S.
      • Baine M.J.
      • et al.
      Potentials of plasma NGAL and MIC-1 as biomarker(s) in the diagnosis of lethal pancreatic cancer.
      • Kaur S.
      • Baine M.J.
      • Guha S.
      • et al.
      Neutrophil gelatinase-associated lipocalin, macrophage inhibitory cytokine 1, and carbohydrate antigen 19-9 in pancreatic juice: pathobiologic implications in diagnosing benign and malignant disease of the pancreas.
      Mucins are highly glycosylated proteins that are widely expressed in GI tissues, yet mucin 2 (MUC2) and mucin 5, subtype AC (MUC5AC) specifically are undetectable in normal pancreas tissue.
      • Klöppel G.
      • Kosmahl M.
      Is the intraductal papillary mucinous neoplasia of the biliary tract a counterpart of pancreatic papillary mucinous neoplasm?.
      ,
      • Balagué C.
      • Audié J.P.
      • Porchet N.
      • et al.
      In situ hybridization shows distinct patterns of mucin gene expression in normal, benign, and malignant pancreas tissues.
      Their diagnostic potential has been suggested in serum and FNA specimens but not yet for PJ.
      • Maker A.V.
      • Katabi N.
      • Gonen M.
      • et al.
      Pancreatic cyst fluid and serum mucin levels predict dysplasia in intraductal papillary mucinous neoplasms of the pancreas.
      • Kaur S.
      • Smith L.M.
      • Patel A.
      • et al.
      A combination of MUC5AC and CA19-9 improves the diagnosis of pancreatic cancer: a multicenter study.
      • Zhang J.
      • Wang Y.
      • Zhao T.
      • et al.
      Evaluation of serum MUC5AC in combination with CA19-9 for the diagnosis of pancreatic cancer.
      • Wang Y.
      • Gao J.
      • Li Z.
      • et al.
      Diagnostic value of mucins (MUC1, MUC2 and MUC5AC) expression profile in endoscopic ultrasound-guided fine-needle aspiration specimens of the pancreas.
      Because altered expression of mucins is detected in intraductal papillary mucinous neoplasms and pancreatic intraepithelial neoplasia (PanIN), increased mucin levels are expected to occur early in the development of PDAC and may therefore complement cytokines and serum CA19-9 in early detection.
      • Maker A.V.
      • Katabi N.
      • Gonen M.
      • et al.
      Pancreatic cyst fluid and serum mucin levels predict dysplasia in intraductal papillary mucinous neoplasms of the pancreas.
      • Kaur S.
      • Smith L.M.
      • Patel A.
      • et al.
      A combination of MUC5AC and CA19-9 improves the diagnosis of pancreatic cancer: a multicenter study.
      • Zhang J.
      • Wang Y.
      • Zhao T.
      • et al.
      Evaluation of serum MUC5AC in combination with CA19-9 for the diagnosis of pancreatic cancer.
      • Wang Y.
      • Gao J.
      • Li Z.
      • et al.
      Diagnostic value of mucins (MUC1, MUC2 and MUC5AC) expression profile in endoscopic ultrasound-guided fine-needle aspiration specimens of the pancreas.
      Previous studies showed that CA19-9 in PJ performed less well than the other mentioned proteins in PJ.
      • Kaur S.
      • Baine M.J.
      • Guha S.
      • et al.
      Neutrophil gelatinase-associated lipocalin, macrophage inhibitory cytokine 1, and carbohydrate antigen 19-9 in pancreatic juice: pathobiologic implications in diagnosing benign and malignant disease of the pancreas.
      To reduce harm to individuals participating in surveillance programs, a new diagnostic tool should be able to rule out malignant progression. For instance, a tool with a (close to) 100% specificity could reliably postpone surgery in individuals with indeterminate features on imaging or reduce surveillance frequency in individuals without relative or absolute indications for surgery.
      European Study Group on Cystic Tumours of the Pancreas. European evidence-based guidelines on pancreatic cystic neoplasms.
      Because multiple subtypes of PDAC have been identified,
      • Son M.
      • Kim H.
      • Han D.
      • et al.
      A clinically applicable 24-protein model for classifying risk subgroups in pancreatic ductal adenocarcinomas using multiple reaction monitoring-mass spectrometry.
      ,
      • Birnbaum D.J.
      • Begg S.K.S.
      • Finetti P.
      • et al.
      Transcriptomic analysis of laser capture microdissected tumors reveals cancer- and stromal-specific molecular subtypes of pancreatic ductal adenocarcinoma.
      each potentially characterized by distinct biomarkers, it is expected that a combination of biomarkers (rather than 1 biomarker alone) is needed to reach such high specificity.
      In this study, we aimed to develop a biomarker panel able to distinguish individuals without PDAC (high specificity; primum non nocere) to avoid unnecessary harm to participants in a surveillance population while also detecting PDAC with considerable sensitivity. To this end, we investigated the diagnostic performance of 5 biomarkers (IL-8, IFN-γ, NGAL, MUC5AC, MUC2) in both serum and PJ in addition to CA19-9 in serum.

      Methods

      Study design and patient inclusion

      This case-control study included data from 3 prospective cohort studies performed at the Erasmus MC University Medical Center Rotterdam: the KRASPanc-study (MEC-2018-038), concerning patients with (suspected) sporadic PDAC undergoing diagnostic EUS or fiducial placement for stereotactic radiotherapy; the CAPS-study (MEC-2012-448), involving individuals under surveillance for a hereditary predisposition for PDAC; and the PACYFIC study (MEC-2014-021), involving individuals undergoing surveillance for suspected neoplastic pancreatic cysts. Supplementary Table 1 (available online at www.giejournal.org) shows the inclusion and exclusion criteria for these studies. We considered all patients who underwent PJ collection during EUS between August 2018 and May 2020 for inclusion. Samples were excluded if they had undergone a freeze–thaw cycle. Only 1 patient with high-grade dysplasia was identified in the study cohorts, and therefore this patient was excluded from analysis. If multiple PJ collections occurred in a single patient, the first sample was assessed for this study. In addition, serum samples collected within 3 months of PJ collection were evaluated on availability.
      The institutional medical center ethical review board approved the study, and included individuals gave written informed consent before enrollment. The study was carried out according to the ethical principles for medical research involving human subjects from the World Medical Association Declaration of Helsinki.

      Clinical data extraction

      Clinical variables were collected in the course of prospective studies. EUS-FNB pathology results and CA19-9 values were extracted from patient records if available within 3 months before or after PJ collection.

      Biomaterial collection

      PJ collection was performed with a linear echoendoscope (Pentax Medical, Tokyo, Japan) by experienced endosonographers (L.M.J.W.D., J.W.P., and M.J.B.). After insertion of the tip of the echoendoscope into the D2 segment of the duodenum, secretion of PJ was stimulated by intravenous injection of human secretin (16 μg per patient; ChiRhoClin, Burtonsville, Md, USA). Suction through the endoscopic channel was applied immediately after injection of secretin for 8 minutes by positioning of the tip close to the ampullary orifice.
      • Levink I.J.M.
      • Nesteruk K.
      • Visser D.I.
      • et al.
      Optimization of pancreatic juice collection: a first step toward biomarker discovery and early detection of pancreatic cancer.
      Within 10 minutes after collection, juice was aliquoted and snap frozen. Samples were stored at –80°C until use. Serum samples were collected by venipuncture, aliquoted, and stored at –80°C until use.

      Protein analysis

      Concentrations of candidate proteins (IL-8, IFN-γ, NGAL, MUC5AC, MUC2) in serum and PJ were assessed by enzyme-linked immunosorbent assay according to the protocol of the manufacturer. Phospholipase A2 group IB (PLA2G1B; pancreas standard marker) and total protein concentration were assessed to control for collection- or patient-related differences between biomaterials.
      For IL-8 and IFN-γ (No. 88-8086, No. 88-7316; Thermo Fisher Scientific, Waltham, Mass, USA), immunosorb plates were coated with protein-specific capture antibody overnight at 4°C. Plates were blocked with enzyme-linked immunosorbent assay diluent for 1 hour at room temperature and incubated with PJ (100 μL per well) at 4°C overnight. Then, biotin-conjugated detection antibody was added at room temperature for 1 hour, followed by avidin-horseradish peroxidase (HRP) for 30 minutes at room temperature. Antibody binding was visualized by addition of tetramethylbenzidine (TMB) substrate, reactions were terminated by adding sulfuric acid, and absorbance was read at 450 nm. Measurements of MUC2, MUC5AC, and NGAL (OKIA00076, OKEH02839, OKEH02840; Aviva Systems Biology, San Diego, Calif, USA) and PLA2G1B (No. MBS703283, PJ only; My Biosource, San Diego, Calif, USA) concentrations were performed similarly using precoated and preblocked plates. Before NGAL analysis, PJ was diluted 1:30. For total protein concentration, a Lowry protein assay (Bio-Rad, Hercules, Calif, USA) was executed. PJ or serum samples of cases and (different types of) control subjects were equally distributed among enzyme-linked immunosorbent assay plates.

      Statistical analysis

      Depending on distributional properties of measures, descriptive data are expressed as means with standard deviation, medians with range, or percentages. Statistical significance was assessed using the Mann-Whitney U test for continuous data and either the Fisher exact test or χ2 test for categorical data. Individuals were considered to have high-risk features when presenting with 1 absolute indication for surgery (enhancing mural nodule/solid lesion ≥5 mm, caliber change of the main pancreatic duct [MPD] suggesting MPD obstruction, and MPD dilation ≥10 mm) or with 2 or more relative indications for surgery (MPD dilation 5-9.9 mm, cyst diameter ≥40 mm, new-onset diabetes mellitus [DM], acute pancreatitis, and enhancing mural nodule <5 mm).
      European Study Group on Cystic Tumours of the Pancreas. European evidence-based guidelines on pancreatic cystic neoplasms.
      For primary outcome analysis, PDAC cases and control subjects were compared with the Mann-Whitney U test. Additionally, control subjects from the PACYFIC cohort with a cystic lesion with absolute or relative indications for surgery
      European Study Group on Cystic Tumours of the Pancreas. European evidence-based guidelines on pancreatic cystic neoplasms.
      were compared with control subjects without these indications to evaluate the influence of heterogeneity in the control group (Mann-Whitney U test). To evaluate whether marker concentrations were associated with PJ quality, they were correlated with total protein (ruling out collection-related differences between groups) and PLA2G1B (representing true pancreas-derived material) concentrations.
      Multiple logistic regression models were created to test whether the biomarkers were associated with PDAC, independent of age, gender, and presence of DM. Age and body mass index (BMI; per group) were correlated with biomarker concentrations using “Spearman’s correlation.” To evaluate the markers, association with gender, chemotherapy (treatment-naive vs post-treatment), resectability (resectable vs locally advanced pancreatic cancer [combination of non- and borderline resectable disease]), and caliber change (as a measure of MPD obstruction), concentrations were compared within groups (cases and control subjects) by the Mann-Whitney U test. To rule out the influence of common bile duct stenting, a subanalysis was executed comparing cases and control subjects without a common bile duct stent (Mann-Whitney U test).
      Diagnostic performance was evaluated for those biomarkers that significantly differed between cases and control subjects (Mann-Whitney U test) and were associated with PDAC independent of age, gender, DM, and BMI (multiple logistic regression described earlier). For these markers, receiver-operating characteristic curves were created, resulting in an area under the curve (AUC) with 95% CIs, a paired-sample analysis was performed to compare AUC values between panels and markers. Two panels were created: a PJ-only panel (IL-8, NGAL, and MUC5AC) and a combined panel (serum CA19-9 as well as PJ IL-8, NGAL, and MUC5AC). Given diagnostic performances of panels are based on the number of biomarkers with a positive test.
      Because implementation of a biomarker should not result in additional stress or anxiety, unnecessary diagnostic procedures, harm (because of unnecessary surgery or biopsy sampling), and costs, we aimed for high specificity (>85% for individual markers) and positive predictive value (PPV). In this study, selection of cutoff values was based on the specificity. A potential biomarker panel should eventually have a specificity >95% with fair sensitivity (>50%). For CA19-9, a cutoff of ≥37 kU/L was used, as is customary in clinical practice.
      European Study Group on Cystic Tumours of the Pancreas. European evidence-based guidelines on pancreatic cystic neoplasms.
      CIs of sensitivity, specificity, PPV, and accuracy are represented by “exact” Clopper-Pearson CIs. These analyses were performed in SPSS (version 27; SPSS Inc, Chicago, Ill, USA), and figures were created using GraphPad (Prism version 9; GraphPad Software, La Jolla, Calif, USA). Survival curves were created, and Cox regressions were performed using R (“ggplot2,” “survival,” “survminer”; Rstudio, PBC, Boston, Mass, USA).

      Results

      Patient cohort

      Patient characteristics are reported in Table 1. Of 185 consecutive individuals who met the inclusion criteria, 59 individuals had histologically confirmed PDAC. Indications for EUS in this case group were tissue acquisition for suspected PDAC (n = 38 [63.3%]; these cases were treatment-naïve) and fiducial placement to enable stereotactic radiotherapy (n = 21 [35.0%]; these patients had undergone chemotherapy before PJ collection).
      Table 1Clinical characteristics of study participants
      Cases (n = 59)Control subjects (n = 126)P value
      Median age, y (interquartile range)68.0 (11.0)62.0 (16.0)<.001
      Male gender36 (61.0)42 (33.3)<.001
      Median body mass index, kg/m2 (interquartile range)23.2 (4.4)25.6 (5.1)<.001
      Familial/genetic predisposition0 (.0)71 (56.3)<.001
       Member of familial pancreatic cancer family36 (28.6)
      CDKN2A p1625 (19.8)
      BRCA2 + 2 blood relatives with PDAC5 (4.0)
      BRCA1 + 2 blood relatives with PDAC1 (.8)
      PALB2 + 2 blood relatives with PDAC1 (.8)
      BRCA2 + CDKN2A p161 (.8)
      STK11/LKB12 (1.6)
      Diabetes mellitus24 (40.7)16 (12.7)<.001
      Indication EUS<.001
       Suspected PDAC38 (64.4)4 (3.2)
       Fiducial placement21 (35.6)
       Surveillance0 (.0)122 (96.8)
      CBD stent in situ<.001
       CBD stent in situ9 (15.3)0 (.0)
       No CBD stent but CBD dilation14 (23.7)3 (2.4)
       No CBD stent and no CBD dilation36 (61.0)123 (97.6)
      Relative/absolute indications for surgery
      European Study Group on Cystic Tumours of the Pancreas. European evidence-based guidelines on pancreatic cystic neoplasms.
      59 (100.0)26 (20.6)<.001
       Enhancing mural nodule or hypodense lesion59 (100.0)4 (3.2)
       Caliber change to main pancreatic duct44 (74.6)0 (.0)
       Diffuse pancreatic duct dilation >5 mm0 (.0)14 (11.1)
       Carbohydrate antigen 19.9 ≥37 kU/L37 (62.7)7 (5.5)
       Cyst size >40 mm0 (.0)2 (1.6)
       New-onset diabetes
      Development of diabetes mellitus in the last 2 years.
      9 (15.2)2 (1.6)
       Recent acute pancreatitis
      Present in the last 2 years.
      2 (3.4)6 (4.8)
       Lymphadenopathy26 (44.1)0 (4.0)
      Working diagnosis<.001
       No abnormalities43 (34.1)
       Unspecified cyst12 (9.5)
       Side-branch IPMN53 (42.1)
       Main-duct IPMN14 (8.8)
       Mucinous cystic neoplasm1 (.8)
       Neuroendocrine tumor1 (.8)
       Indeterminate, not suspect for malignancy2 (1.6)
       Resectable PDAC
      Resectable PDAC (as based on Dutch pancreatic cancer audit guideline): superior mesenteric artery, no contact; celiac axis, no contact; common hepatic artery, no contact; superior mesenteric vein and portal vein, ≤90-degree contact.
      11 (18.6)
       Locally advanced PDAC48 (81.4)
      Distant metastases (on imaging)10 (16.9)0 (.0)<.001
      Values are n (%) unless otherwise defined.
      CBD, Common bile duct; IPMN, intraductal papillary neoplasm; PDAC, pancreatic ductal adenocarcinoma.
      Development of diabetes mellitus in the last 2 years.
      Present in the last 2 years.
      Resectable PDAC (as based on Dutch pancreatic cancer audit guideline): superior mesenteric artery, no contact; celiac axis, no contact; common hepatic artery, no contact; superior mesenteric vein and portal vein, ≤90-degree contact.
      The control group consisted of 126 individuals: 122 undergoing EUS in the course of surveillance for a genetic or familial predisposition (n = 71 [56.3%]) or for suspected neoplastic pancreatic cysts (n = 51 [40.4%]) and 4 (3.2%) for an unsubstantiated PDAC suspicion.
      Control subjects had a median follow-up duration of 31 months (interquartile range, 6) during which none of these individuals developed PDAC. Fourteen control subjects had a surveillance duration shorter than 12 months: 10 resigned surveillance (and did not present with PDAC later in time), 3 patients died, and 1 was released from surveillance after a mucinous cystic neoplasm was found during surgery. These control subjects were not excluded from analysis.
      Four control subjects underwent pancreatic surgery, of which 3 had intraductal papillary mucinous neoplasms with low-grade dysplasia (2 gastric, 3 intestinal-type) and 1 had mucinous cystic neoplasm with low-grade dysplasia. Of these, 1 control subject died because of abdominal sepsis and multiorgan failure 8 days after resection. During the follow-up period, 2 control subjects developed other malignancies: 1 individual with a hereditary increased risk of PDAC was diagnosed with stage IV mamma carcinoma 13 months after collection, and 1 had a melanoma at the time of collection and died 7 months after collection.

      Relative and absolute indication for surgery

      With regard to morphologic changes, 43 control subjects (34.1%) had no abnormalities on imaging, and 80 (63.4%) had a cystic lesion. According to European guidelines,
      European Study Group on Cystic Tumours of the Pancreas. European evidence-based guidelines on pancreatic cystic neoplasms.
      26 control subjects (20.6%) had a relative or absolute indication for surgery at the time of PJ collection (Table 1). Of these, 10 (7.9% of control subjects) had 1 absolute indication or 2 or more relative indications for surgery and could have undergone surgery according to these guidelines.

      Serum: cases versus control subjects

      Serum samples were available for 116 of 185 patients (48 cases, 68 control subjects), and CA19-9 was determined in 101 of 185 patients (53 cases, 48 control subjects). Time between PJ and (biobank) serum sampling was ≤3 weeks for all samples; the median time between PJ and CA19-9 measurement was 9 days (interquartile range, 14). Total protein concentration did not differ between investigated groups (P = .21) (Fig. 1A), and none of the investigated serum biomarkers was correlated with the total protein concentration in serum (Supplementary Fig. 1. available online at www.giejournal.org).
      Figure thumbnail gr1
      Figure 1Serum (A-G) and pancreatic juice (PJ; H-N) protein levels in cases as compared with control subjects. A, Total protein concentration in serum was not different between cases and control subjects. B and C, CA19-9 (B) and IL-8 (C) concentration was higher in serum from cases than from control subjects. D-G, Concentration of IFN-γ (D), NGAL (E), MUC5AC (F), and MUC2 (G) in serum did not differ between cases and control subjects. H and I, Based on concentration of total protein (H) and PLA2G1B (I), the quality of pancreatic juice was not different for cases and control subjects. J and K, IL-8 concentration was higher in PJ from cases than from control subjects, whereas IFN-γ concentration did not differ. L-N, Concentrations of NGAL (L), MUC5AC (M), and MUC2 (N) were higher in PJ from cases than from control subjects. Groups were compared with Mann-Whitney U test. The dark green line indicates the median concentration per group. CA19-9, Carbohydrate antigen 19.9; IFN-γ, interferon-γ; IL-8, interleukin-8; MUC2, mucin 2; MUC5AC, mucin 5, subtype AC; NGAL, neutrophil gelatinase-associated lipocalin.
      Concentrations of CA19-9 (P < .0001) and IL-8 (P = .006) were higher in cases than in control subjects. None of the other serum proteins differed between groups (Fig. 1B-G). MUC2 was detectable in 15 individuals (7 cases [14%], 8 control subjects [12%]) and MUC5AC in 5 individuals (2 cases [4%], 3 control subjects [4%]). Additionally, the presence of serum MUC2 and MUC5AC was not related to the development of other malignancies within the 12 months before or after serum collection. No differences in serum protein levels were found between control subjects with and without absolute or relative indications for surgery (P > .05 for all markers; not shown).
      European Study Group on Cystic Tumours of the Pancreas. European evidence-based guidelines on pancreatic cystic neoplasms.

      PJ: cases versus control subjects

      Total protein (P = .35) and PLA2G1B (P = .24) concentrations (as a measure of PJ quality) did not differ between cases and control subjects (Fig. 1H and I). PLA2G1B levels correlated with the concentration of total protein (r = .38, P < .001), IL-8 (r = .13, P = .05), IFN-γ (r = .27, P < .001), NGAL (r = .23, P < .001), and MUC2 (r = .14, P = .05) but not with MUC5AC (r = .01; P = .85) (Supplementary Fig. 1).
      When comparing cases with control subjects, IL-8 (P = .0001) and NGAL (P < .0001) concentrations were higher for cases, whereas IFN-γ concentration did not differ from control subjects (P = .13) (Fig. 1J and L). Concentrations of the investigated mucins, MUC5AC (P = .01) and MUC2 (P = .04), were higher in cases than control subjects (Fig. 1M and N). For the control group, none of these markers differed between patients with and without absolute or relative indications for surgery (P > .05; not shown).
      European Study Group on Cystic Tumours of the Pancreas. European evidence-based guidelines on pancreatic cystic neoplasms.

      Relation of markers to clinical characteristics

      Because statistical differences in baseline characteristics were present between cases and control subjects (Table 1), we investigated their influence on protein concentrations. IL-8 concentration in PJ (r = .27, P = .04) (Fig. 2A) and serum (r = .55, P < .0001) (Fig. 2B) was correlated with age for cases but not for control subjects. Biomarker levels were similar for male and female patients and those with or without DM and did not correlate to BMI (not shown). Multivariable logistic regression showed that IL-8 (P = .003), NGAL (P = .003), and MUC5AC (P = .03) concentrations in PJ and CA19-9 concentration in serum (P = .05) were associated with PDAC, independent of age, gender, and DM (Supplementary Tables 2 and 3, available online at www.giejournal.org).
      Figure thumbnail gr2
      Figure 2Association of biomarkers and clinical parameters. A and B, Interleukin (IL)-8 concentration in cases was correlated with age for pancreatic juice (PJ; A) and serum (B). This correlation was not found in control subjects or for any other investigated proteins (not shown). C and D, Caliber change (as a measure of pancreatic duct obstruction) did not influence the pancreatic content in PJ (C), whereas PJ IL-8 concentration was higher in patients with caliber change than in those without (D). This association was not found for the other investigated proteins (not shown). E-I, PJ quality and biomarker concentration did not differ between cases with previous chemotherapy and treatment-naïve cases. J-N, In treatment-naïve cases (n = 38), marker concentrations were not different in patients with early-stage pancreatic ductal adenocarcinoma (resectable) compared with those with (locally) advanced disease. Significance was tested with Spearman correlation (A and B) and Mann-Whitney U test (C and D). The dark green line indicates the median per group (C and D). CA19-9, Carbohydrate antigen 19.9; MUC2, mucin 2; MUC5AC, mucin 5, subtype AC; NGAL, neutrophil gelatinase-associated lipocalin; LAPC, locally advanced pancreatic cancer; PLA2G1B, phospholipase A2 group IB.
      For cases, a clinical parameter that might affect PJ composition is MPD obstruction (suspected by a caliber change on imaging), because this may decrease pancreatic content in PJ. However, we did not observe lower concentrations of PLA2G1B in such cases (P = .63) (Fig. 2C). Although IL-8 concentration was higher in individuals with a caliber change in the MPD (P = .05) (Fig. 2D,), none of the other investigated PJ markers was influenced by caliber change (not shown). Furthermore, common bile duct stenting may cause local inflammation and increase biomarker concentrations. Therefore, we performed a subanalysis after excluding the 9 cases with a common bile duct stent. Concentrations of IL-8 (P = .003), NGAL (P < .0001), and MUC5AC (P = .04) in PJ and IL-8 (P = .03) and CA19-9 (P < .0001) in serum remained higher in cases than in control subjects. Differences in MUC2 in PJ no longer reached significance (P = .06). Because cases who are treatment-naïve may show different results (eg, other inflammatory state, potential tumor reduction) than those who had already undergone chemotherapy, we compared these groups. However, both the PJ quality (as based on the PLA2G1B concentration) (Fig. 2E) and the marker concentrations (Fig. 2E-I) were not different between groups. These were also not different for patients with resectable disease (early cancer) or locally advanced disease (ie, locally advanced pancreatic cancer) (Fig. 2J-N).

      Diagnostic performance

      We generated receiver-operating characteristic curves to assess the diagnostic performance of markers that were independently associated with the presence of cancer (serum CA19-9 and PJ IL-8, NGAL, and MUC5AC) (Fig. 3). As previously described, we aimed at high specificity. Cutoff values and corresponding sensitivities, specificities, accuracy, and PPV of individual markers are presented in Table 2. Our results show that CA19-9 (cutoff ≥37 kU/L) was able to differentiate between cases (n = 53) and control subjects (n = 48) with a sensitivity of 69.8% (95% CI, 55.7-81.7) and specificity of 85.4 (95% CI, 72.2-93.9).
      Figure thumbnail gr3
      Figure 3Diagnostic performance of both individual biomarkers associated with pancreatic ductal adenocarcinoma independent of age, gender, and presence of diabetes mellitus and the created panels (panel 1: interleukin [IL]-8, neutrophil gelatinase-associated lipocalin [NGAL], and mucin 5, subtype AC [MUC5AC] in pancreatic juice [PJ]; panel 2: serum CA19-9 and IL-8, NGAL, and MUC5AC in PJ). A, Receiver-operating characteristic curves of the different individual markers and panels. For NGAL, 2 cutoffs are indicated: Cutoff A could be used for NGAL as individual marker (high specificity) and cutoff B for the panel (high total diagnostic performance). Ranges are 95% confidence intervals. The numbers are equal to the number of possible values related to the diagnostic performance on the curve. B, Differences (Δ) between AUC values of the different markers and panels; P values are generated by a paired-sample analysis. AUC, Area under the curve; CA19-9, carbohydrate antigen 19.9.
      Table 2Diagnostic performance of individual markers and generated panels
      ProteinNo. of cases/control subjectsArea under the curveCutoff concentrationSensitivity

      %
      Specificity

      %
      Accuracy

      %
      Positive predictive value

      %
      Individual markers
      CA19-953/48
      Patient numbers were variable, as 1 concentration of IL-8 in PJ (case) was missing and 85 concentrations of serum CA19-9 were missing (7 cases, 78 control subjects).
      .86 (.79-.94)37 kU/L69.8 (55.7-81.7)85.4 (72.2-93.9)77.2 (67.8-85.0)84.1 (72.3-91.5)
      IL-8 in PJ58/126
      Patient numbers were variable, as 1 concentration of IL-8 in PJ (case) was missing and 85 concentrations of serum CA19-9 were missing (7 cases, 78 control subjects).
      .68 (.58-.77)15 pg/mL44.8 (31.7-58.5)92.1 (85.9-96.1)77.2 (70.4-83.0)72.2 (57.3-83.4)
      NGAL in PJ59/126
      Patient numbers were variable, as 1 concentration of IL-8 in PJ (case) was missing and 85 concentrations of serum CA19-9 were missing (7 cases, 78 control subjects).
      .70 (.62-.77)993 ng/mL28.8 (17.8-42.1)87.3 (80.2-92.6)68.7 (61.4-75.3)51.5 (36.6-66.1)
      321 ng/mL76.27 (63.4-86.4)59.5 (50.4-68.2)64.9 (57.5-71.7)46.9 (40.6-53.2)
      MUC5AC in PJ59/126
      Patient numbers were variable, as 1 concentration of IL-8 in PJ (case) was missing and 85 concentrations of serum CA19-9 were missing (7 cases, 78 control subjects).
      0.62 (.53-.71)3.7 pg/mL39.0 (26.6-52.6)81.0 (73.0-87.4)67.6 (60.3-74.3)48.9 (37.2-60.8)
      Panels
      Panel 1: PJ (≥2 positive)
      The number of markers that showed a concentration above the defined cutoffs (CA19.9, 37 kU/L; IL-8, 15 pg/mL; NGAL, 321 ng/mL; MUC5AC, 3.7 pg/mL).
      58/126
      Patient numbers were variable, as 1 concentration of IL-8 in PJ (case) was missing and 85 concentrations of serum CA19-9 were missing (7 cases, 78 control subjects).
      .78 (.71-.85)As indicated
      The number of markers that showed a concentration above the defined cutoffs (CA19.9, 37 kU/L; IL-8, 15 pg/mL; NGAL, 321 ng/mL; MUC5AC, 3.7 pg/mL).
      51.7 (38.2-65.1)85.7 (78.4-91.3)75.0 (68.1-81.1)62.5 (50.4-73.2)
      Panel 1: PJ (3 positive)
      The number of markers that showed a concentration above the defined cutoffs (CA19.9, 37 kU/L; IL-8, 15 pg/mL; NGAL, 321 ng/mL; MUC5AC, 3.7 pg/mL).
      58/126
      Patient numbers were variable, as 1 concentration of IL-8 in PJ (case) was missing and 85 concentrations of serum CA19-9 were missing (7 cases, 78 control subjects).
      15.5 (7.4-27.4)99.2 (95.7-100)72.8 (65.8-79.1)90.0 (53.9-98.6)
      Panel 2: PJ + serum (≥3 positive)
      The number of markers that showed a concentration above the defined cutoffs (CA19.9, 37 kU/L; IL-8, 15 pg/mL; NGAL, 321 ng/mL; MUC5AC, 3.7 pg/mL).
      52/48
      Patient numbers were variable, as 1 concentration of IL-8 in PJ (case) was missing and 85 concentrations of serum CA19-9 were missing (7 cases, 78 control subjects).
      .84 (.77-.92)As indicated42.3 (28.7-56.8)95.8 (85.8-99.5)68.0 (57.9-77.0)91.7 (73.2-97.8)
      Values in parentheses are 95% confidence intervals. For the individual markers that were independently associated with pancreatic ductal adenocarcinoma (CA19-9, PJ IL-8, and PJ NGAL), cutoffs were selected aiming for specificity (sensitivity >25%). Two biomarker panels were generated: a PJ-only panel (panel 1: IL-8 and NGAL) and a combined panel (panel 2: serum CA19-9, PJ IL-8, and PJ NGAL).
      PJ, Pancreatic juice; CA19-9, carbohydrate antigen 19.9; IL-8, interleukin-8; MUC5AC, mucin 5, subtype AC; NGAL, neutrophil gelatinase-associated lipocalin.
      Patient numbers were variable, as 1 concentration of IL-8 in PJ (case) was missing and 85 concentrations of serum CA19-9 were missing (7 cases, 78 control subjects).
      The number of markers that showed a concentration above the defined cutoffs (CA19.9, 37 kU/L; IL-8, 15 pg/mL; NGAL, 321 ng/mL; MUC5AC, 3.7 pg/mL).
      For the biomarkers in PJ, an IL-8 concentration of 15 pg/mL differentiated between cases (n = 58) and control subjects (n = 126) with high specificity (92.1%; 95% CI, 85.9%-96.1%), a sensitivity of 44.8% (95% CI, 31.7-58.5), and a moderate PPV (72.2%; 95% CI, 57.3-83.4). NGAL (993 ng/mL) had a specificity of 87.3% (95% CI, 80.2-92.6), sensitivity of 28.8% (95% CI, 17.8-42.1), and PPV of 51.5% (95% CI, 36.6-66.1) for differentiating between cases (n = 59) and control subjects (n = 126) (Table 2, Figs. 3 and 4). Serum CA19-9 had a significantly higher AUC than the individual PJ markers (P = .001) (Fig. 3B). Only 1 case (and no control subjects) had a positive test for all 4 markers (serum CA19-9 as well as PJ IL-8, NGAL, and MUC5AC).
      Figure thumbnail gr4
      Figure 4The presence of a positive or negative test of tested individual biomarkers and panels for cases and control subjects in relation to gender, DM, indication of procedure, and working diagnosis based on imaging. For the individual biomarkers and the 2 panels, the cutoff as shown in was used to differentiate between a positive and negative test (for NGAL was 993 ng/mL as we aimed for high specificity). Patients with suspected PDAC were treatment-naïve, whereas patients with an indication for fiducial placement had undergone chemotherapeutic treatment. The presence of a relative or absolute indication (RI/AI) was based on the European evidence-based pancreatic cyst guideline.
      European Study Group on Cystic Tumours of the Pancreas. European evidence-based guidelines on pancreatic cystic neoplasms.
      CA19-9, Carbohydrate antigen 19.9; DM, diabetes mellitus; FNB, fine-needle biopsy sampling; IL-8, interleukin-8; MUC5AC, mucin 5, subtype AC; NGAL, neutrophil gelatinase-associated lipocalin; PDAC, pancreatic ductal adenocarcinoma; PJ, pancreatic juice; IPMN, intraductal papillary mucinous neoplasm; NET, neuroendocrine tumor.
      Next, 2 biomarker panels were tested: 1 combining PJ IL-8, NGAL, and MUC5AC (panel 1) and another that included serum CA19-9 (panel 2). For PJ markers in these panels, cutoffs with the highest sum (sensitivity + specificity) were selected to increase the overall performance (and not just the specificity). For PJ IL-8 and MUC5AC, these cutoffs had high specificity and were used to calculate diagnostic performance of the individual marker. For NGAL, we selected a different cutoff (321 ng/mL). For panel 1 (tested in 58 cases and 126 control subjects), having 2 or more markers with a positive test was related to a decent diagnostic performance (sensitivity, 51.7% [95% CI, 38.2-65.1]; specificity, 85.7% [95% CI, 78.4-91.3]; accuracy, 75.0% [95% CI, 68.1-81.1]; PPV, 62.5% [95% CI, 50.4-73.2]). In case of 3 biomarkers with a positive test, the specificity became close to 100% yet the sensitivity was 15.5% (95% CI, 7.4-27.4) (Table 2, Fig. 3A). With regard to panel 2 (tested in 52 cases and 48 control subjects), having 3 or more biomarkers with a positive test was able to differentiate between cases and control subjects with a sensitivity of 42.3 (95% CI, 28.7-56.8), specificity of 95.8% (95% CI, 85.8-99.5), accuracy of 68.0% (95% CI, 57.9-77.0), and PPV of 91.7% (95% CI, 73.2-97.8).
      Based on the AUC, both panels performed better than individual PJ markers but were not superior to CA19-9 alone. Panel 2 performed better than panel 1 (P < .001) (Fig. 3B). Of note, patient numbers for serum CA19-9 and panel 2 were lower (52 cases, 48 control subjects) than those for PJ markers and panel 1 because of missing CA19-9 values.
      As a comparison, sensitivity of EUS-FNB (at first try) in our cohort was 73%. The specificity could not be calculated, because none of the control subjects underwent EUS-FNB at the same procedure. Figure 4 shows an overview of the investigated biomarkers and panels in relation with (other) patient characteristics.

      Predictive value of biomarkers

      During a median follow-up of 30 months (range, 19-38), 43 patients (73%) died. Survival analysis was only performed for patients who underwent PJ collection (for suspected PDAC) at the time of diagnosis (aka treatment-naïve, n = 38). The other subgroup (who underwent chemotherapy) underwent PJ collection 7 months after diagnosis. Cox regression analysis showed that resectability (detected in earlier stage), serum CA19-9, and PJ IL-8, NGAL, and MUC5AC (Fig. 5) were not predictive for survival in this patient cohort.
      Figure thumbnail gr5
      Figure 5Survival of patients based on resectability (A) and with a positive or negative test for serum CA19-9 (B), PJ IL-8 (C), PJ NGAL (D), and PJ MUC5AC (E). HR, Hazard ratio (calculated by Cox regression and adjusted for patient age); CA19-9, carbohydrate antigen 19.9; LAPC, locally advance pancreatic cancer; CI, confidence interval; IL-8, interleukin-8; MUC5AC, mucin 5, subtype AC; NGAL, neutrophil gelatinase-associated lipocalin.

      Discussion

      Here we assessed the diagnostic performance of promising protein biomarkers in serum and PJ for PDAC (and high-grade dysplasia) detection. We showed that concentrations of CA19-9 and IL-8 (but not NGAL, MUC5AC, MUC2, and IFN-γ) in serum and IL-8, NGAL, MUC5AC, and MUC2 (but not IFN-γ) in PJ are significantly higher in cases than in control subjects. CA19-9 in serum and IL-8 and NGAL in PJ were associated with PDAC, independent of age, gender, BMI, and presence of DM. A panel of these 3 markers was able to differentiate between cases and control subjects with a specificity and PPV of 100%, which is far higher than that achieved with serum CA19-9 alone.
      Imaging-based surveillance programs have not yet convincingly shown improved survival in individuals undergoing pancreatic surveillance for hereditary risk. If based only on imaging and clinical features (as is advocated in clinical guidelines), 10 control subjects (7.9%) with indications for surgery would have undergone surgery, which may have been unnecessary because they did not develop PDAC during the median follow-up period of 16.5 months. Thus, tools to improve the early diagnosis of pancreatic cancer are urgently needed, in particular biomarkers, which are relatively easy and cheap to determine in laboratories worldwide. Implementation of a biomarker in a surveillance program may serve 2 goals: selecting individuals at increased risk who are eligible for increased surveillance, in which case a lower specificity is accepted, and to support decision-making regarding additional diagnostic procedures and even treatment including surgery. In the latter case, the marker should have a high specificity to avoid unnecessary harm. The current criterion standard (EUS-FNB) is not able to address this because of adverse events related to EUS-FNB (eg, pancreatitis).
      PJ is an attractive biomarker source and a good alternative for serum. Measurement of IL-8, NGAL, MUC5AC, and MUC2 in PJ outperformed measurement of these markers in serum. This may be explained by the close contact of PJ with the pancreatic ductal epithelial cell lining, which contains the cells of origin of PDAC. Although serum collection carries a negligible risk, PJ collection, if collected from the duodenal lumen (after secretin stimulation) during EUS (without cannulation or occlusion of the ampullary orifice), also carries a low and acceptable procedural risk. Although contamination of PJ with blood, bile, and duodenal fluid may lower detectable concentrations of pancreas-derived biomarkers,
      • Levink I.J.M.
      • Nesteruk K.
      • Visser D.I.
      • et al.
      Optimization of pancreatic juice collection: a first step toward biomarker discovery and early detection of pancreatic cancer.
      ,
      • Doyle C.J.
      • Yancey K.
      • Pitt H.A.
      • et al.
      The proteome of normal pancreatic juice.
      ,
      • Sadakari Y.
      • Kanda M.
      • Maitani K.
      • et al.
      Mutant KRAS and GNAS DNA concentrations in secretin-stimulated pancreatic fluid collected from the pancreatic duct and the duodenal lumen.
      serum may contain biomolecules derived from other tissues as well. In our study, total protein and PLA2G1B in PJ,
      • Meng Q.
      • Shi S.
      • Liang C.
      • et al.
      Diagnostic and prognostic value of carcinoembryonic antigen in pancreatic cancer: a systematic review and meta-analysis.
      as measures for PJ dilution, were comparable between our cases and control subjects.
      Serum CA19-9 is currently the only biomarker used in clinical practice for PDAC but only for treatment response prediction and detection of disease recurrence. It is less effective for early-stage PDAC detection.
      • Kim J.R.
      • Jang J.Y.
      • Kang M.J.
      • et al.
      Clinical implication of serum carcinoembryonic antigen and carbohydrate antigen 19-9 for the prediction of malignancy in intraductal papillary mucinous neoplasm of pancreas.
      ,
      • Ciprani D.
      • Morales-Oyarvide V.
      • Qadan M.
      • et al.
      An elevated CA 19-9 is associated with invasive cancer and worse survival in IPMN.
      ,
      • Murakami M.
      • Nagai Y.
      • Tenjin A.
      • et al.
      Proposed cut-off value of CA19-9 for detecting pancreatic cancer in patients with diabetes: a case-control study.
      Based on the European evidence-based guideline for cystic lesions,
      European Study Group on Cystic Tumours of the Pancreas. European evidence-based guidelines on pancreatic cystic neoplasms.
      which advocates CA19-9 >37 kU/L as a relative indication for surgery, 7 control subjects in our cohort (none of which developed PDAC within a 16.5-month follow-up) would have wrongfully undergone surgery. Conversely, in favor of CA19-9 (despite the very low numbers), 4 of 5 individuals in our surveillance cohort who did undergo surgery had low-grade dysplasia coinciding with serum CA19-9 <37 kU/L, whereas the 1 individual with high-grade dysplasia had an increased CA19-9 level (60 kU/L).
      In the current study, PJ NGAL differentiated cases from control subjects with a sensitivity of 30.5% and a specificity of 87.3 at a cutoff of 993 ng/mL. Kaur et al
      • Kaur S.
      • Baine M.J.
      • Guha S.
      • et al.
      Neutrophil gelatinase-associated lipocalin, macrophage inhibitory cytokine 1, and carbohydrate antigen 19-9 in pancreatic juice: pathobiologic implications in diagnosing benign and malignant disease of the pancreas.
      discriminated PDAC (n = 58) from control subjects (n = 47) without pancreatic disease (Non-pancreatic non-healthy [NPNH]) with a higher sensitivity (79%; 95% CI, 67-89) and similar specificity (83%; 95% CI, 61-95) at a cutoff of 138 ng/mL. This difference may be because Kaur et al included control subjects without visible abnormalities and we included a more heterogeneous control group (including individuals at increased risk for PDAC). Hogendorf et al
      • Hogendorf P.
      • Durczyński A.
      • Skulimowski A.
      • et al.
      Neutrophil gelatinase-associated lipocalin (NGAL) concentration in urine is superior to CA19-9 and Ca 125 in differentiation of pancreatic mass: preliminary report.
      compared PDAC (n = 21) with chronic pancreatitis (n = 15) and found a sensitivity of 71% and specificity of 73% at a cutoff of 27 ng/mL. This low cutoff value may be related to an alternative dilution of fluid related to their method of collection (ie, intraoperatively from the common bile duct). Budzynska et al
      • Budzynska A.
      • Nowakowska-Dulawa E.
      • Marek T.
      • et al.
      Differentiation of pancreatobiliary cancer from benign biliary strictures using neutrophil gelatinase-associated lipocalin.
      compared malignant biliary structures (n = 22) because of cholangiocarcinoma or PDAC with benign biliary strictures (n = 18) and found that serum NGAL levels had no discriminative value. However, bile NGAL levels differentiated between the 2 groups with a sensitivity of 77% and specificity of 72% with a cutoff of 459 ng/mL, and, similar to our study, a combination of bile NGAL and serum CA19-9 improved the sensitivity to 91% and specificity to 67%. For IL-8 in PJ, our results show a sensitivity of 44% and specificity of 92% at a cutoff of 15 pg/mL. Our results are comparable with those of Noh et al,
      • Noh K.W.
      • Pungpapong S.
      • Wallace M.B.
      • et al.
      Do cytokine concentrations in pancreatic juice predict the presence of pancreatic diseases?.
      where IL-8 in (secretin-stimulated) PJ was able to differentiate between PDAC (n = 38) and control subjects without abnormalities on EUS (n = 41) with a sensitivity of 74% and specificity of 95% at a cutoff of 23 pg/mL.
      The variety of biomarker cutoff values in medical literature is a commonly known limitation of biomarker studies and complicates quick study comparisons and meta-analyses. A selected cutoff depends on the goal of the test (high sensitivity or high specificity). For instance, for CA19.9 a cutoff of 37 kU/L is generally used. In our study, this cutoff led to a sensitivity of 69.8% and specificity of 85.4%, whereas for a surveillance population a cutoff of 83 kU/L may be preferable (related to a sensitivity of 62.3% and specificity of 100% in our study). It is not feasible to have distinct cutoff values per race or indication, yet it is important to critically appraise cutoffs before implementation in guidelines and clinical practice.
      This study involved 3 prospective cohorts (2 surveillance cohorts and 1 cohort with sporadic PDAC cases). Ideally, to answer if early detection of PDAC is possible using the selected biomarkers, only PDAC cases detected in the surveillance cohort would have to be included. However, these cases are rare, and it would take years to reach the appropriate sample size. To overcome this problem, we included patients with sporadic PDAC. Some of these PDAC cases were detected early (18% resectable), which is in accordance with the literature on sporadic PDAC. Extrapolation of the current data to a surveillance population should be done with caution, because high-risk individuals with a hereditary predisposition or pancreatic cystic neoplasms may have a distinct natural disease course (with distinct molecular changes) compared with patients with sporadic PDAC. Our control group was heterogeneous (comprising individuals with either a genetic or familial predisposition or suspected neoplastic cystic neoplasms), which may further complicate extrapolation to other groups. Additionally, because we did not have histologic confirmation of the control group, we cannot formally exclude the possibility that some of these patients may have had undetectable lesions at the time of sampling. However, the fact that none developed PDAC during 16.5 months of follow-up argues against this. In the current study, survival comparison between patients with a negative and positive test was not the primary objective. Potentially because of an insufficient sample size for this analysis, we did not find a statistical difference. However, the survival curve of serum CA19.9 and PJ IL-8 showed a trend and may become significant in studies with a sufficient sample size for this analysis.
      In conclusion, levels of MUC5AC, IL-8, and NGAL in PJ together with serum CA19-9 provide a biomarker panel for PDAC detection that is highly specific, fairly sensitive, and safe to collect and could be readily implemented in EUS-based surveillance programs. Future prospective studies are needed to validate this panel in a longitudinal surveillance cohort and investigate if combining these markers with imaging results in increased sensitivity and may allow an earlier identification of PDAC patients.

      Acknowledgment

      We acknowledge K. A. Overbeek, L. G. B. Wermelink, D. M. De Jong, A. S. Audhoe, D. I. Visser, Y. M. M. Mulder, A. J. Berkhout. S. C. Jaarsma, and S. van de vondervoort (all from the Department of Gastroenterology & Hepatology, Erasmus MC, University Medical Center, Rotterdam, The Netherlands). They did not fulfill the requirements for authorships but substantially contributed to the project.

      Appendix

      Figure thumbnail fx2
      Supplementary Figure 1Correlations between concentrations of the tested markers. Numbers indicate the correlation coefficient (Spearman r); bold numbers with asterisk (∗) are significant (P < .05). IFN-γ, Interferon-γ; IL-8, interleukin-8; MUC2, mucin 2; MUC5AC, mucin 5, subtype AC; NGAL, neutrophil gelatinase-associated lipocalin; PJ, pancreatic juice; PLA2G1B, phospholipase A2 group IB.
      Supplementary Table 1Overview of inclusion criteria of the 3 prospective cohort studies
      Prospective cohortInclusion criteriaExclusion criteria
      KRASPancPatients who undergo EUS for (suspected) PDAC either as part of a diagnostic process or fiducial placement before radiotherapeutic treatment.Age <18 y
      CAPSIndividuals who, after evaluation by a clinical geneticist, have an estimated 10-fold increased risk of developing PDAC. This includes the following: (1) Carriers of a gene mutation in CDKN2A or STK11, regardless of the family history of pancreatic cancer

      (2) Carriers of a gene mutation in BRCA1, BRCA2, p53, or mismatch repair gene with a family history of PDAC in ≥2 family members

      (3) Familial PDAC (familial pancreatic cancer) kindreds, defined as individuals with at least 2 first-degree relatives with PDAC; 3 relatives with pancreatic cancer, either first- or second-degree relatives; or 2 second-degree relatives with pancreatic cancer of which ≥1 was aged <50 y at the time of diagnosis
      Age <18 y, personal history of pancreatic cancer, individuals unable to provide informed consent, severe medical illness, PRSS1 gene carrier, contraindication for EUS because of anatomic abnormalities/surgery
      PACYFICIndividuals with a suspected neoplastic pancreatic cyst (either newly or previously diagnosed or previously operated on) for which cyst surveillance is warranted, according to the treating physicianAge <18 y, history of chronic pancreatitis, suspected pseudocyst (simple, thin-walled cyst that developed in the course of acute pancreatitis, as documented by sequential imaging studies), suspected serous cystadenoma (typical microcystic lesion with lobulated outlines, a calcified central scar and/or cyst fluid carcinoembryonic antigen levels <5 ng/mL), von Hippel-Lindau disease, and limited life expectancy (<2 y)
      PDAC, Pancreatic ductal adenocarcinoma.
      Supplementary Table 2Multivariable analysis of markers in pancreatic juice
      Pancreatic juice markerInterleukin-8 (pg/mL)Neutrophil gelatinase-associated lipocalin (μg/mL)Mucin 5, subtype AC (pg/mL)Mucin 2 (pg/mL)
      Odds ratio (95% CI)P valueOdds ratio (95% CI)P valueOdds ratio (95% CI)P valueOdds ratio (95% CI)P value
      Marker1.03 (1.01-1.05).0032.53 (1.38-4.64).0031.16 (1.01-1.32).031.00 (1.00-1.00).52
      Age (y)1.04 (1.00-1.08).081.05 (1.00-1.09).031.06 (1.02-1.11).0071.06 (1.01-1.10).01
      Female gender (y/n).53 (.25-1.09).08.41 (.20-.83).01.44 (.22-.89).02.41 (.21-.82).01
      Diabetes mellitus (y/n)4.14 (1.82-9.42)<.0014.37 (1.95-9.77)<.0014.14 (1.86-9.21)<.0013.66 (1.67-8.01).001
      Constant.05.05.04.03.17.01.04.03
      CI, Confidence interval.
      Supplementary Table 3Multivariable analysis of markers in serum
      Serum markerInterleukin-8 (pg/mL)Carbohydrate antigen 19.9 (kU/L)
      Odds ratio (95% CI)P valueOdds ratio (95% CI)P value
      Marker1.08 (1.00-1.17).071.04 (1.02-1.06)<.001
      Age (y)1.06 (1.01-1.11).021.03 (.97-1.11).33
      Female gender (y/n).29 (.12-.70).006.43 (.14-1.29).13
      Diabetes mellitus (y/n)3.50 (1.26-9.68).02.57 (.17-1.96).37
      Constant.06.12.08.34
      CI, Confidence interval.

      References

        • Siegel R.L.
        • Miller K.D.
        • Jemal A.
        Cancer statistics, 2020.
        CA Cancer J Clin. 2020; 70: 7-30
        • Overbeek K.A.
        • Levink I.J.M.
        • Koopmann B.D.M.
        • et al.
        Long-term yield of pancreatic cancer surveillance in high-risk individuals.
        Gut. 2022; 71: 1152-1160
        • Vasen H.
        • Ibrahim I.
        • Ponce C.G.
        • 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; 34: 2010-2019
        • Canto M.I.
        • Almario J.A.
        • Schulick R.D.
        • et al.
        Risk of neoplastic progression in individuals at high risk for pancreatic cancer undergoing long-term surveillance.
        Gastroenterology. 2018; 155: 740-751
        • Konings I.C.A.W.
        • Canto M.I.
        • Almario J.A.
        • et al.
        Surveillance for pancreatic cancer in high-risk individuals.
        BJS Open. 2019; 3: 656-665
        • Yachida S.
        • Jones S.
        • Bozic I.
        • et al.
        Distant metastasis occurs late during the genetic evolution of pancreatic cancer.
        Nature. 2010; 467: 1114-1117
      1. European Study Group on Cystic Tumours of the Pancreas. European evidence-based guidelines on pancreatic cystic neoplasms.
        Gut. 2018; 67: 789-804
        • Kim J.R.
        • Jang J.Y.
        • Kang M.J.
        • et al.
        Clinical implication of serum carcinoembryonic antigen and carbohydrate antigen 19-9 for the prediction of malignancy in intraductal papillary mucinous neoplasm of pancreas.
        J Hepatobil Pancreat Sci. 2015; 22: 699-707
        • Le Calvez-Kelm F.
        • Foll M.
        • Wozniak M.B.
        • et al.
        KRAS mutations in blood circulating cell-free DNA: a pancreatic cancer case-control.
        Oncotarget. 2016; 7: 78827-78840
        • Meng Q.
        • Shi S.
        • Liang C.
        • et al.
        Diagnostic and prognostic value of carcinoembryonic antigen in pancreatic cancer: a systematic review and meta-analysis.
        OncoTarget Ther. 2017; 10: 4591-4598
        • Zhang Y.
        • Yang J.
        • Li H.
        • et al.
        Tumor markers CA19-9, CA242 and CEA in the diagnosis of pancreatic cancer: a meta-analysis.
        Int J Clin Exp Med. 2015; 8: 11683-11691
        • Choi M.H.
        • Mejlænder-Andersen E.
        • Manueldas S.
        • et al.
        Mutation analysis by deep sequencing of pancreatic juice from patients with pancreatic ductal adenocarcinoma.
        BMC Cancer. 2019; 19: 11
        • Levink I.J.M.
        • Nesteruk K.
        • Visser D.I.
        • et al.
        Optimization of pancreatic juice collection: a first step toward biomarker discovery and early detection of pancreatic cancer.
        Am J Gastroenterol. 2020; 115: 2103-2108
        • Kuwada Y.
        • Sasaki T.
        • Morinaka K.
        • et al.
        Potential involvement of IL-8 and its receptors in the invasiveness of pancreatic cancer cells.
        Int J Oncol. 2003; 22: 765-771
        • Detjen K.M.
        • Farwig K.
        • Welzel M.
        • et al.
        Interferon gamma inhibits growth of human pancreatic carcinoma cells via caspase-1 dependent induction of apoptosis.
        Gut. 2001; 49: 251-262
        • Moniaux N.
        • Chakraborty S.
        • Yalniz M.
        • et al.
        Early diagnosis of pancreatic cancer: neutrophil gelatinase-associated lipocalin as a marker of pancreatic intraepithelial neoplasia.
        Br J Cancer. 2008; 98: 1540-1547
        • Noh K.W.
        • Pungpapong S.
        • Wallace M.B.
        • et al.
        Do cytokine concentrations in pancreatic juice predict the presence of pancreatic diseases?.
        Clin Gastroenterol Hepatol. 2006; 4: 782-789
        • Shaw V.E.
        • Lane B.
        • Jenkinson C.
        • et al.
        Serum cytokine biomarker panels for discriminating pancreatic cancer from benign pancreatic disease.
        Mol Cancer. 2014; 13: 114
        • Kaur S.
        • Chakraborty S.
        • Baine M.J.
        • et al.
        Potentials of plasma NGAL and MIC-1 as biomarker(s) in the diagnosis of lethal pancreatic cancer.
        PLoS One. 2013; 8: e55171
        • Kaur S.
        • Baine M.J.
        • Guha S.
        • et al.
        Neutrophil gelatinase-associated lipocalin, macrophage inhibitory cytokine 1, and carbohydrate antigen 19-9 in pancreatic juice: pathobiologic implications in diagnosing benign and malignant disease of the pancreas.
        Pancreas. 2013; 42: 494-501
        • Klöppel G.
        • Kosmahl M.
        Is the intraductal papillary mucinous neoplasia of the biliary tract a counterpart of pancreatic papillary mucinous neoplasm?.
        J Hepatol. 2006; 44: 249-250
        • Balagué C.
        • Audié J.P.
        • Porchet N.
        • et al.
        In situ hybridization shows distinct patterns of mucin gene expression in normal, benign, and malignant pancreas tissues.
        Gastroenterology. 1995; 109: 953-964
        • Maker A.V.
        • Katabi N.
        • Gonen M.
        • et al.
        Pancreatic cyst fluid and serum mucin levels predict dysplasia in intraductal papillary mucinous neoplasms of the pancreas.
        Ann Surg Oncol. 2011; 18: 199-206
        • Kaur S.
        • Smith L.M.
        • Patel A.
        • et al.
        A combination of MUC5AC and CA19-9 improves the diagnosis of pancreatic cancer: a multicenter study.
        Am J Gastroenterol. 2017; 112: 172-183
        • Zhang J.
        • Wang Y.
        • Zhao T.
        • et al.
        Evaluation of serum MUC5AC in combination with CA19-9 for the diagnosis of pancreatic cancer.
        World J Surg Oncol. 2020; 18: 31
        • Wang Y.
        • Gao J.
        • Li Z.
        • et al.
        Diagnostic value of mucins (MUC1, MUC2 and MUC5AC) expression profile in endoscopic ultrasound-guided fine-needle aspiration specimens of the pancreas.
        Int J Cancer. 2007; 121: 2716-2722
        • Son M.
        • Kim H.
        • Han D.
        • et al.
        A clinically applicable 24-protein model for classifying risk subgroups in pancreatic ductal adenocarcinomas using multiple reaction monitoring-mass spectrometry.
        Clin Cancer Res. 2021; 27: 3370
        • Birnbaum D.J.
        • Begg S.K.S.
        • Finetti P.
        • et al.
        Transcriptomic analysis of laser capture microdissected tumors reveals cancer- and stromal-specific molecular subtypes of pancreatic ductal adenocarcinoma.
        Clin Cancer Res. 2021; 27: 2314
        • Doyle C.J.
        • Yancey K.
        • Pitt H.A.
        • et al.
        The proteome of normal pancreatic juice.
        Pancreas. 2012; 41: 186-194
        • Sadakari Y.
        • Kanda M.
        • Maitani K.
        • et al.
        Mutant KRAS and GNAS DNA concentrations in secretin-stimulated pancreatic fluid collected from the pancreatic duct and the duodenal lumen.
        Clin Translat Gastroenterol. 2014; 5: e62
        • Ciprani D.
        • Morales-Oyarvide V.
        • Qadan M.
        • et al.
        An elevated CA 19-9 is associated with invasive cancer and worse survival in IPMN.
        Pancreatology. 2020; 20: 729-735
        • Murakami M.
        • Nagai Y.
        • Tenjin A.
        • et al.
        Proposed cut-off value of CA19-9 for detecting pancreatic cancer in patients with diabetes: a case-control study.
        Endocrinol J. 2018; 65: 639-643
        • Hogendorf P.
        • Durczyński A.
        • Skulimowski A.
        • et al.
        Neutrophil gelatinase-associated lipocalin (NGAL) concentration in urine is superior to CA19-9 and Ca 125 in differentiation of pancreatic mass: preliminary report.
        Cancer Biomark. 2016; 16: 537-543
        • Budzynska A.
        • Nowakowska-Dulawa E.
        • Marek T.
        • et al.
        Differentiation of pancreatobiliary cancer from benign biliary strictures using neutrophil gelatinase-associated lipocalin.
        J Physiol Pharmacol. 2013; 64: 109-114