Immunotherapies provide significant survival benefits in many cancers, but the association of tumor microenvironment (TME) with clinical outcomes of immunotherapy remains to be explored. We aim to utilize comprehensive profile of TME-related components and molecular mechanisms to accurately predict the cancer immunotherapy benefit, and further evaluate the predictive utility of the combination of TME and radiomics landscape.
This study analyzed RNA sequencing or whole-exome data of 3,663 patients treated with immunotherapy from the phase II IMvigor210 trial, Gene Expression Omnibus and cBioPortal database, and of 4,547 patients from The Cancer Genome Atlas. We built a novel predictive TME score integrating immune checkpoints, human leukocyte antigens, immune cells, long non-coding RNA, and tumor mutation burden using computational algorithms, and further combined it with radiomics landscape.
The analysis of IMvigor210 trial identified that patients with high TME scores had strikingly greater overall survival (OS) benefits from immunotherapy compared with patients with low TME scores (hazard ratio 0.41, 95% CI, 0.30 to 0.57; P < 0.0001). The TME score showed encouraging immunotherapeutic predictive ability for OS (20-month, AUC = 0.81) and response (AUC = 0.77). These findings were successfully validated in other cohorts across multiple cancer types. Moreover, the combination of TME score and radiomics landscape was superior to the variable alone in predicting the patient prognosis for OS (AUC = 0.91, 0.87, 0.75 for the combination, TME score and radiomics landscape, respectively).
This study developed a TME score that achieved accurate prediction of immunotherapy benefit in patients with cancer. The study is also the first to highlight that the combination of TME signature and radiomics landscape has predictive implications for immunotherapy and merits prospective validation in future clinical trials.
Legal entity responsible for the study
National Science and Technology Major Project (2020ZX09201021), Medical Artificial Intelligence Project of Sun Yat-Sen Memorial Hospital (YXRGZN201902), National Natural Science Foundation of China (81572596, 81972471, U1601223), Natural Science Foundation of Guangdong Province (2017A030313828), Guangzhou Science and Technology Major Program (201704020131), Guangdong Science and Technology Department (2017B030314026), Special Funds for the Cultivation of Guangdong College Students’ Scientific and Technological Innovation (PDJH2019A0212), National Students’ Innovation and Entrepreneurship Training Program (201910571001), Guangdong Medical University College Students’ Innovation Experiment Project (ZZZF001).
All authors have declared no conflicts of interest.