Abstract 6820: Artificial intelligence (AI) assisted analysis of plasma cytokine levels in a phase II trial using metronomic chemotherapy on gastrointestinal (GI) cancer patients

Abstract We previously reported a metronomic chemotherapy regimen (UFT, a 5-fluorouracil prodrug; 100 mg/twice per day p.o.; plus cyclophosphamide as an i.v. bolus on day 1 and then 50 mg/day p.o.; plus celecoxib 200 mg/twice a day p.o.); evaluated in the COMET trial (ID NCT02926911; Allegrini et al. Angiogenesis 2012: 15(2):275-86) of GI cancer patients. At AACR 2024 we reported analysis of 75 cytokines in the plasma of these patients. Here, we report levels of 22 not previously evaluated cytokines, analyzed for clinical correlations in this trial, in 34 out of the 38 patients. We used AI to perform part of the statistical analysis. AI assisted in the writing of the R or Python programs implemented. We report the identification of baseline plasma levels of soluble Tie-2 (sTie2) and osteopontin as potential predictive biomarkers for the overall survival (OS) of participants to this trial. sTie 2 values showed a positive correlation with OS (r = 0.43; p = 0.019), while osteopontin was inversely related (r = -0.41; p = 0.024). Having defined cut-off levels (389 pg/ml for sTie 2 and 412 pg/ml for osteopontin), we observed a stronger association with OS in subgroups of patients above or below the thresholds, as detected by Kaplan-Meier analysis and log-rank tests (p < 0.01, in both tests). The chosen cut-offs also were predictive for patients with better Progression Free Survival (PFS). Patients with osteopontin levels below the cut-off had a PFS of 5.5 months, compared to 2 months in patients above the cut-off (p=0.002). For sTie2 patients with above cut-off levels had a PFS of 5.9 months versus 2.2 months in patients below the cut-off (p=0.002). We also used cytokine levels to generate predictive patient scores. A score of 1 was assigned for each of sTie2 ≤ 388.8 pg/mL, Osteopontin > 412.5 pg/mL, and sHer3 > 185.4 pg/mL baseline levels. We then arranged patients with relation to Stable and Progressive Disease as follows: A) Low Risk (Score 0): Stable Disease at Recist criteria: 62.5% Progressive Disease: 37.5%. B) Intermediate Risk (Score 1): Stable Disease: 45.8% Progressive Disease: 54.2%. C) High Risk (Score 2-3): Stable Disease: 23.1% Progressive Disease: 76.9%. Our results demonstrate an approach using AI to assist in biomarker analysis of clinical trials, and to stratify patients to identify those likely to benefit from metronomic chemotherapy. Citation Format: Ryley Stewart, Juan C. Silva Espinoza, Briana Elise Pinales, Cuipeng Qiu, Jiangying Zhang, Paola Orlandi, Vladimir Georgiev, Alessandro Michelangeli, Kazumasa Fujiwara, Giacomo Allegrini, Giulio Francia, Guido Bocci, Maria L. Manca. Artificial intelligence (AI) assisted analysis of plasma cytokine levels in a phase II trial using metronomic chemotherapy on gastrointestinal (GI) cancer patients [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 6820.