125P Assessing the accuracy of multiple prognostic scores for immune checkpoint inhibitors (ICI) in patients with advanced solid tumors

Many patients fail to achieve a clinical benefit from ICI. Several scores have been developed improve ICI candidates selection but it is uncertain which one better predicts patients’ outcome. Here, we performed direct comparison of the most successful scores. This sub-analysis immunoblood prospective observational study that enrolled diagnosed with advanced solid tumors treated Main clinicopathological data were retrieved medical records and responses assessed according RECIST 1.1 criteria. LIPI, RMH, PMH, dNLR, NLR, PIPO GRIm calculated. Receiving operator characteristics (ROC) curves their area under curve (AUC) used predict PFS durable (DCB; stable disease≥6 months or better). Associations PFS, OS DCB, where Cox logistic regressions. Scores’ correlation was Spearman rho. Significance set at p<0.05. We recruited 155 (65% male, mean age 63). NSCLC (28%), colorectal (20%) breast (9%) H&N (6%) cancer melanoma frequent tumor types. Frequency high risk/bad outcome group each score were: LIPI 13%, RMH 36%, PMH 54%, 14%, 6%, NLR 32% dNRL 27%. Fair accuracy in identifying higher risk progression mild predicting DCB observed for (AUC PFS: 0.7, 95%CI: 0.6-0.8; AUC DCB: 0.6, 0.5-0.8) AUC: 0.5-0.7) All other provided poor/no accuracy. No significant difference between (both p>0.05). Additionally, only associated (p=0.001; p<0.001), (p<0.001; p=0.001) (p=0.034; p=0.010) univariate analyses. At multivariate analyses remained significantly (p=0.030; p=0.021) (p=0.012; p<0.001). A strong both (rho=0.72, p<0.001) observed. sufficiently reliable assessing prognosis They superior analyzed our population highly correlated.