e21609 Background: We prospectively recorded common clinical and laboratory parameters of pts with metastatic NSCLC treated with 2nd line ICIs to evaluate their potential value in a clinical outcome prediction model. Methods: Data on patient (age, PS, BMI) and disease characteristics (histology, sites/number of metastases), smoking status, use/duration of co-medications [proton pump inhibitors (PPIs) inhalational/p.os steroids, antibiotics (ATB)], laboratory values [neutrophil/lymphocyte ratio (NLR), LDH] and response to previous therapy were prospectively collected. Pts were categorized as having steroids if they had received steroids > 10mg for ≥10d (starting from 15d before or within the first 3 months of treatment). Prolonged ATB administration was defined as ATBs ≥14d (30d before or within the first 3 months). JAD Bio (www.jadbio.com), an Automated Machine Learning service that analyzes biological data with emphasis on feature selection was used to create predictive models. Results: 66 pts were evaluated; median follow up time was 6.37 months. 15.2% of patients had PR, 34.8% SD and 50% PD. Median PFS and OS were 3.5 and 6.77 months, respectively. Steroids had negative impact on disease stabilization (PR+SD) rates (p = 0.042) and PFS (p = 0.013) but not OS (p = 0.051). ATBs negatively affected RR (p = 0.046) only, whereas, prolonged ATB exposure was associated with lower RR (p = 0.007), PFS (p = 0.0001) and OS (p = 0.001). In multivariate analysis, steroids [HR = 2.54 (CI:1.23-5.29, p = 0.012)], prolonged ATBs [HR = 3.46 (CI:1.72-6.95, p = 0.0001)], liver HR = 2.92 (CI:1.45-5.78, p = 0.003) and bone HR = 2.06 (CI:1.041-4.10, p = 0.038) metastases independently predicted for shorter PFS. Only prolonged ATBs [HR = 2.52 (CI: 1.39-4.54, p = 0.002)] and bone metastases [HR = 2.26 (CI: 1.23-4.17, p = 0.009)] independently predicted for shorter OS. Analysis of the investigated parameters by JAD Bio predicted disease stabilization with an accuracy of 71%. Importantly, PDL-1 status was not included due to high rates of missing data. Conclusions: Our results corroborate previous evidence on the detrimental role of prolonged ATB administration on ICIs efficacy, possibly related to perturbation of the gut microbiota. Modeling, including other significant parameters from larger patient cohorts, could result in a robust estimation of outcome.