Background: Systemic inflammatory scores have been developed as tools to aid clinicians in prognostication and patient (pt) selection for clinical trials. We compared the accuracy of five prognostic scores to predict overall survival (OS) in pts with advanced pancreatic adenocarcinoma (PDAC). Methods: Pts with advanced PDAC enrolled on the COMPASS trial (NCT02750657) from 2015 to 2020 were included. All pts had biopsies for whole genome and RNA sequencing prior to standard first-line chemotherapy in the advanced setting. Prognostic risk was calculated using: neutrophil-to-lymphocyte ratio (NLR; >5 = high), platelet-to-lymphocyte ratio (PLR; > 150 = high), Prognostic Nutritional Index (PNI = albumin + 5 x lymphocytes. PNI < 45 = high risk), Gustave Roussy Immune Score (GRIm-S; NLR>6 = 1 point, albumin <35 = 1 point, LDH > upper limit of normal [ULN] = 1 point. GRIm-S ≥2 = high risk), and Memorial Sloan Kettering Prognostic Score (MPS; NLR >4 and albumin < 40 = high risk). OS was estimated using the Kaplan-Meier method and compared between risk groups (high vs. not-high) for each scoring system using the log-rank test. Cox proportional hazards models were used to analyze the association between each prognostic score and OS, adjusting for baseline clinical and genomic factors. Results: In total, 263 pts with advanced PDAC cancer were included, with median follow up of 32.9 (95% CI 15.9-64.2) months. Median OS in the intention to treat population was 9.3 months (95% CI 8-10.2). PLR and PNI were not prognostic. High risk NLR (N=85, 32%), GRIm-S (N=47, 18%) and MPS (N=46, 17%) identified pts with poor prognosis. The GRIm-S and MPS were most significant: median OS in high vs low risk pts 6.4 vs. 10 months p<0.001 (GRIm-S) and 6.3 vs. 10 months p=0.002 (MPS). On multivariable analyses, high risk NLR, GRIm-Score and MPS were each associated with poor OS after adjusting for baseline clinical and genomic factors (Table). For all models, ECOG ≥1 (N=165, 63%); the basal-like Moffitt RNA subtype (N=49, 20% vs 80% classical) and low HRDetect scores (N=31, 13%) were significantly associated with poor OS. However these scores did not associate with RNA based classifiers or HRD scores and can therefore provide additional prognostic information. Conclusions: Both the GRIm-S and MPS are highly prognostic in PDAC and are scores easily used in the clinical setting and may help in clinical trial selection. Genotypic correlates are being explored.

¹Multivariable model for each prognostic score adjusted for age, sex, ECOG, Moffitt subtype and HRDetect score.