A MACHINE learning model, SCORPIO, has shown promise in predicting the success of immune checkpoint inhibitors (ICIs) for patients with cancer. By leveraging routine blood tests and clinical data, this model offers a simpler and cost-effective alternative to advanced genomic or immunologic assays. SCORPIO was trained using data from 9,745 patients treated across 21 cancer types and has proven to be highly accurate in forecasting patient outcomes.
In a comprehensive study, SCORPIO outperformed traditional biomarkers such as tumor mutational burden (TMB) in predicting overall survival. The model achieved a median area under the curve (AUC) value of 0.763 for survival at 6, 12, 18, 24, and 30 months, while TMB had much lower predictive performance with AUC values ranging from 0.503 to 0.543. Additionally, SCORPIO demonstrated superior accuracy in predicting clinical benefit, including tumor response or prolonged stability, achieving AUC values of 0.714 and 0.641, compared to TMB’s performance of 0.546 and 0.573.
External validation using global Phase III clinical trials and real-world data from Mount Sinai Health System confirmed SCORPIO’s robust performance across diverse cancer types and healthcare settings. The model was trained on clinical and laboratory variables, such as complete blood count and metabolic profile, providing a reliable, non-invasive way to assess the likelihood of clinical benefit from ICIs.
This innovative tool could significantly improve decision-making in oncology by offering a reliable, accessible means of predicting which patients will benefit from immunotherapy. With its proven ability to adapt across cancer types, SCORPIO has the potential to affect personalized cancer treatment, enhancing outcomes and optimizing resource use in clinical settings.
Reference: Yoo SK et al. Prediction of checkpoint inhibitor immunotherapy efficacy for cancer using routine blood tests and clinical data. Nat Med. 2025. doi: 10.1038/s41591-024-03398-5.
Anaya Malik | AMJ
