INTRODUCTION: Life expectancy evaluation is crucial when selecting patients who may benefit from phase I studies. The Royal Marsden Hospital (RMH) prognostic score, based on three objective variables (number of metastatic sites, lactate dehydrogenase (LDH) and serum albumin) was validated in patients treated with cytotoxics and targeted therapies. We aimed to determine if those factors were applicable to immune-checkpoint therapies (ICTs) in phase I trials and to evaluate new variables that may preclude a better prognosis in patients receiving ICT. PATIENTS AND METHODS: We conducted a retrospective analysis of survival risk factors in a discovery cohort of 155 patients enrolled into ICT phase I trials at our institution. We computed univariate analysis and multivariate analysis (MVA) of demographics, clinical and biological data to assess their prognostic value for overall survival (OS). MVA results were used to build a prognostic score for OS. A validation cohort of 113 patients enrolled in phase I ICT trials was used to prospectively validate this score. RESULTS: A total of 155 patients (M/F: 83/72; median age 59) receiving an experimental ICT between March 2012 and January 2016 were included in the discovery cohort. An MVA assessing the RMH score variables showed that low albumin (hazard ratio [HR] 1.73, 95% confidence interval [CI] 1.05-2.86) and LDH > upper limit normal (ULN) (HR 1.88, 95% CI 1.12-3.15) were independent negative prognostic factors for OS. Interestingly, neutrophil-to-lymphocyte ratio (NLR) > 6 (HR 1.75, 95% CI 1.04-2.95) was associated with a decrease in OS. The number of metastases was not associated with a poorer outcome for this ICT cohort (HR 0.83, 95% CI 0.51-1.35). A risk score based on the results of the MVA (NLR > 6 = 1; LDH > ULN = 1; albumin < 35 g/l = 1) showed that patients presenting a high score (>1) had a significantly shorter OS (20.4 weeks; 95% CI 5.7-35.2) compared to those with a low score (0 or 1) (68.9 weeks; 95% CI 50-83.7) (HR 2.9, 95% CI 1.87-4.64). In the validation cohort of 113 patients, again the patients presenting a high score showed an inferior OS (HR 6.3, 95% CI 2.7-14.8). CONCLUSION: In ICT phase I trials, traditional prognostic variables included in the RMH score may be suboptimal to determine patient's prognosis. In this context, the NLR is a significant prognostic variable. The Gustave Roussy Immune Score, based on albumin, LDH and NLR, allows a better selection of patients for ICT phase I trials.