Abstract
Tumor burden may have opposing influences on immunotherapy outcomes by enhancing neoantigen load or imposing barriers to immune responses. Whether tumor burden is a specific determinant of immunotherapy benefit or a general prognostic factor irrespective of drug mechanism remains ambiguous. We performed a post hoc individual patient-level data analysis of eight prospective trials, including patients with non-small cell lung cancer, hepatocellular carcinoma (HCC), bladder cancer, and renal cell carcinoma to determine the association between tumor burden and immunotherapy and conventional therapy efficacy. Objective response rates were higher among patients with low tumor burden treated with either atezolizumab or conventional therapies. Low tumor burden was also associated with improved progression-free survival in most cancer types and overall survival in all cancer types irrespective of treatment class. Tumor burden effects were dose-dependent across cancer types. An exception to this was in HCC, in which sorafenib treatment was uniquely associated with improved antitumor effects in high tumor burden cancers. Comparisons of outcomes within tumor burden strata showed that atezolizumab is superior to conventional therapy in improving overall survival but not progression-free survival in both high and low tumor burden patients compared with conventional therapies in non-small cell lung cancer, HCC, and bladder cancer. These findings demonstrate that tumor burden is a histology-agnostic and dose-related prognostic factor rather than an immunotherapy-specific predictive biomarker. Cross-treatment analyses suggest that surrogate endpoints may inadequately account for the detrimental impact of tumor burden, with implications for nonrandomized studies and early-phase trials, particularly in which variations in tumor burden exist in the underlying population. SIGNIFICANCE: Tumor burden is not a specific predictive marker of immunotherapy benefit but has substantial prognostic effects across cancer types and treatment classes. These results highlight the importance and broad applicability of tumor burden as stratification or selection markers in trial design, especially because surrogate endpoints do not wholly capture the detrimental effects of tumor burden on overall survival.