Pancreatic ductal adenocarcinoma (PDAC) has a dismal prognosis and is highly chemoresistant. Early detection is the only means to impact long-term survival, but screening methods are lacking. Given the complex and heterogeneous nature of pancreatic cancer, unbiased analytical methods such as metabolomics by nuclear magnetic resonance (NMR) spectroscopy show promise to identify disease-specific molecular fingerprints. NMR profiles constitute a fingerprint of the biofluid, reporting quantitatively on all detectable small biomolecules. NMR spectroscopy was applied to investigate the urine metabolome of PDAC patients (n = 33) and to detect altered metabolic profiles in comparison with healthy matched controls (n = 54). The spectral data were analyzed using multivariate statistical techniques. Statistically significant differences were found between urine metabolomic profiles of PDAC and control individuals (p < 10(-5)). Group discrimination was possible due to average concentration differences of several metabolite signals, pointing to a multimolecular signature of the disease. The robustness of the determined statistical model is confirmed by its predictive performance (sensitivity = 75.8%, specificity = 90.7%). Additionally, the method allowed for a neat separation between spectral profiles of individuals with intermediate and advanced pathologic staging, as well as for the discrimination of samples based on tumor localization. NMR spectroscopy analysis of urinary metabolic profiles proved successful in identifying a complex molecular signature of PDAC. Furthermore, results of a descriptive-level analysis show the possibility to follow disease evolution and to carry out tumor site mapping. Given the high reproducibility and the noninvasive nature of the analytical procedure, the described method bears potential to impact large-scale screening programs.