Abstract A109: Urinary Metabolomic Signatures of Pancreatic Cancer Generalize Across Japanese and Thai Populations Despite Profound Baseline Divergence
Taisuke Baba, Masaki Sunagawa, Masamichi Hayashi, Kokuryo Toshio, Takashi Mizuno, Shunsuke Onoe, Nobuyuki Watanabe, Mihoko Yamada, Shoji Kawakatsu, Yosuke Ochiai, Hizuru Amano, Akinari Hinoki, Hiroo Uchida, Vasin Thanasukarn, Watcharin Loilome, Tomoki EbataAbstract
Background.
Urine is an attractive noninvasive matrix for pancreatic ductal adenocarcinoma (PDAC) biomarkers, but whether metabolomic signatures generalize across populations differing in ethnicity and diet is unknown.
Methods.
Global urinary metabolomics (Metabolon) was performed on 182 samples: Japanese discovery cohort A (60 PDAC, 44 controls), Japanese validation cohort B (21, 18), and a Thai cohort (Khon Kaen University; 19, 20). Osmolality-corrected, log2-transformed peak areas were analyzed. Fifteen bridging specimens (identical urines assayed in both batches) were included, and all disease analyses used within-cohort PDAC-versus-control contrasts, in which batch effects cancel. Cross-cohort agreement was assessed by the rank correlation of z scores, tested against a permutation null that breaks metabolite correspondence between cohorts. Generative artificial intelligence (AI) assisted in writing the analysis code; all analyses were specified and verified by the authors.
Results.
Baseline populations diverged profoundly: among healthy controls, 251 of 1,106 metabolites (23%) differed between Thai and Japanese subjects (false discovery rate (FDR) <0.05; median 2.2-fold), whereas none differed in the 15 bridging specimens. This population effect exceeded the PDAC-versus-control effect itself (1.6-fold), and controls alone were almost fully separable by nationality (leave-one-out cross-validated area under the curve (AUC), 0.997). Despite this, the PDAC signature generalized: agreement between Japanese cohort A and the Thai cohort (r = 0.499; permutation P = 0.0002) matched that between the two Japanese cohorts (r = 0.531). Generalization was bidirectional: a Thai-derived score separated PDAC from controls in Japanese cohort A (AUC, 0.827), and a Japan-derived score did so in the Thai cohort (AUC, 0.811). Twenty-nine metabolites moved concordantly in all three cohorts, all elevated in PDAC and representing multiple metabolic pathways. Discordance tracked clinical state, not ethnicity: metabolite-inferred biliary obstruction differed across cohorts (Thai, 74%; Japan A, 40%; Japan B, 5%), and 86% of Japan-B PDAC had received neoadjuvant chemotherapy. Stratifying cohort A by clinical state, correlation with the Thai cohort ranged from 0.643 (treatment-naive, obstructed) to 0.208 (postchemotherapy, nonobstructed).
Conclusions.
Urinary PDAC signatures generalize across populations whose healthy metabolic baselines are themselves almost fully discriminable. Cross-cohort discordance reflected clinical state at sampling, namely biliary drainage and neoadjuvant therapy, rather than ethnicity. Detection of pancreatic cancer from urine is robust to differences in diet and ethnicity. Generative AI (Claude, Anthropic) was used to draft and edit the text of this abstract.
Citation Format:
Taisuke Baba, Masaki Sunagawa, Masamichi Hayashi, Kokuryo Toshio, Takashi Mizuno, Shunsuke Onoe, Nobuyuki Watanabe, Mihoko Yamada, Shoji Kawakatsu, Yosuke Ochiai, Hizuru Amano, Akinari Hinoki, Hiroo Uchida, Vasin Thanasukarn, Watcharin Loilome, Tomoki Ebata. Urinary Metabolomic Signatures of Pancreatic Cancer Generalize Across Japanese and Thai Populations Despite Profound Baseline Divergence [abstract]. In: Proceedings of the AACR Conference on Pancreatic Cancer: New Frontiers in Biology and Therapeutic Development; 2026 Sep 25-28; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(18_Suppl_2):Abstract nr A109.