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

Earlier diagnosis of pancreatic ductal adenocarcinoma is key to improving overall survival in patients with this hard-to-treat cancer. We independently validated two previously identified plasma-based metabolic signatures for exclusion of pancreatic ductal adenocarcinoma in cohorts with an increased annual risk. The METAPAC study was a prospective, multicentre, investigator-masked, enrichment design, phase 4 trial done in 23 centres in Germany. Patients with pancreatic lesions identified by diagnostic imaging that required further diagnostic assessment were recruited and followed up for 24 months. Targeted quantitative plasma metabolite analysis was done on a liquid chromatography-tandem mass spectrometry platform. The improved metabolic (i-Metabolic) signature consisted of 12 analytes plus carbohydrate antigen (CA) 19-9, and the minimalistic metabolic (m-Metabolic) signature consisted of four analytes plus CA 19-9. The primary endpoint of the study was the exclusion of pancreatic ductal adenocarcinoma with an 85% specificity and the highest possible diagnostic accuracy. All statistical analyses were done per protocol. This study is registered with the German Clinical Trials Register (DRKS00010866). Between Sept 9, 2016, and April 8, 2022, 1370 patients with CT-identified pancreatic lesions necessitating further diagnostic assessment were screened, of whom 1129 patients (489 with pancreatic ductal adenocarcinoma, 640 controls) were included in the primary analysis (median age 67 years [IQR 58-75]; 556 [49%] female, 572 [51%] male). The control group consisted of high-risk individuals with acute pancreatitis (11 [1%] of 1129 participants), chronic pancreatitis (113 [10%]), intraductal papillary mucinous neoplasms (232 [21%]), cystic lesions other than intraductal papillary mucinous neoplasms (271 [24%]), and metastases of extrapancreatic origin (13 [1%]). The i-Metabolic signature detected pancreatic ductal adenocarcinoma with an area under the curve (AUC) of 0·846 (95% CI 0·842-0·849), specificity of 90·4% (89·8-91·1), sensitivity of 67·5% (66·9-68·0), and balanced accuracy of 80·5% (80·2-80·8), compared with CA 19-9 alone (AUC 0·799 [0·797-0·802], p<0·0001; specificity 79·1% [78·7-79·4]; sensitivity 81·8% [81·5-82·0]; balanced accuracy 80·6% [80·4-80·9]). The m-Metabolic signature detected pancreatic ductal adenocarcinoma with an AUC of 0·846 (95% CI 0·842-0·849; p<0·0001 vs CA 19-9 alone), specificity of 93·6% (93·1-94·0), sensitivity of 59·9% (59·3-60·4), and accuracy of 79·0% (78·8-79·2). In a population of 242 individuals with new-onset diabetes (three cases of incident pancreatic ductal adenocarcinoma), the m-Metabolic signature (without CA 19-9) significantly discriminated patients with pancreatic ductal adenocarcinoma from those without (p=0·038). AUC, specificity, and sensitivity remained constant after random bootstrapping for a prevalence of pancreatic ductal adenocarcinoma between 1% and 20%. Two plasma-based metabolic signatures showed significant improvement in performance compared with CA 19-9 alone in excluding pancreatic ductal adenocarcinoma in a prospective real-world cohort. These findings could offer a surveillance tool in patients with an annual risk of pancreatic ductal adenocarcinoma of 1% to reduce unnecessary invasive procedures and facilitate earlier detection of resectable disease. Federal Ministry of Education and Research (BMBF, Germany).