Clinical Significance of Tumour Mutation Burden in Immunotherapy Across Multiple Cancer Types: an Individual Meta-Analysis

Zhenyu Yang,Shiyou Wei,Yulan Deng,Zihuai Wang,Lunxu Liu
DOI: https://doi.org/10.1093/jjco/hyaa076
IF: 2.925
2020-01-01
Japanese Journal of Clinical Oncology
Abstract:Background: Biomarkers for stratifying patients that could benefit from immune checkpoint inhibitors are necessary. Tumour mutation burden has recently become a promising biomarker in cancer, but the associations between tumour mutation burden and outcomes of immune checkpoint inhibitors treatment were not well-documented in present studies. Methods: We searched PubMed, Web of Science and EMBASE databases up to 1 October 2019. Studies evaluated the association between tumour mutation burden and clinical outcomes were included. Hazard ratios and odds ratios were applied to estimate the association of tumour mutation burden score with overall survival, progression-free survival and response rate, respectively. The best cut-off value was chosen by best discriminated overall survival using Contal and O'Quigley method. Results: Twenty-two studies involving 6171 patients in diverse cancers were included. The individual participant data meta-analysis demonstrated that high tumour mutation burden was associated with better overall survival (HR = 0.57, 95% CI = 0.50-0.64) and progression-free survival (HR = 0.50, 95% CI = 0.40-0.63) and higher response rate. The best cut-off values in each cancer type were 17.7/MB in non-small cell lung cancer, 7.9/MB in bladder cancer, 6.1/MB in melanoma, 12.3/MB in colorectal cancer, 6.9/MB in esophagogastric cancer, 10.5/MB in head and neck cancer. The pooled meta-analysis showed the prognosis value was robust and the sensitivity, specificity and area under the receiver operating characteristic curves in predicting response rates were 0.63, 0.71 and 0.73, respectively. Conclusions: The present meta-analysis indicates tumour mutation burden is a promising predictor of immune checkpoint inhibitors therapy but the cut-off value differs in different cancers.
What problem does this paper attempt to address?