Purpose: The challenge of optimizing patient care around radiation therapy is constantly tackled with a flow of innovations (Robot assisted surgery, Protons, Spacer, etc.). Plethoras of studies demonstrated efficacy of new techniques and solutions, but still very few discriminate patients on their likelihood to encounter side effects. Radiosensitivity tests would be key in decision making in different cases: radiation vs surgery, hypo vs normofractionnated, proton vs photontherapy, and implementation of spacers. Recent studies based on immunofluorescence experiments have stressed the role of the ATM protein in the response to radiation on skin fibroblasts, and the statistical performances of this molecular endpoint to predict individual radiosensitivity (Granzotto et al, 2016). Faster and simpler solutions based on ELISA assay on fibroblast (EF) have confirmed that pATM is a relevant biomarker (Pereira et al, IJROBP 2018). The purpose of this study was to develop a predictive assay on lymphocytes based on the quantification of the pATM proteins by ELISA method. Materials and methods. A retrospective study was performed in blind on 102 blood lymphocytes from 56 radioresistant (RR) and 46 radiosensitive (RS) patients of two cancer types, prostate and head and neck. First, all patients were divided into 2 groups, RR (CTCAE<2) and RS (CTCAE≥2). The global quantity of pATM molecules was assessed by ELISA method. Secondly a classification study was carried out to estimate the threshold value of the biomarker able to minimize the number of false positive cases (RS patients identified RR). A Support Vector Machine classification technique was used to predict the radio-sensitivity. Statistical analysis was developed on R and performed independently. Results: Following promising results obtained with IF (AUC = 0.97) and EF (AUC = 0.85), the experiments performed on lymphocytes showed that the quantities of pATM molecules in each sample were found in agreement with the CTCAE grades. Statistical results showed the following mean values concerning the classification performances for all patients: sensitivity=86 %, specificity=72 %, AUC=79%, NPV=84 % and PPV=74 %. Conclusion: This study, performed using a single blood sample, confirmed results obtained with previous other methods while proposing the base of a realistic, cost effective assay to be potentially proposed to every patient intending to be treated with RT.