Tumour response heterogeneity as a powerful independent predictor of treatment outcome in advanced lung adenocarcinoma: a retrospective analysis

Abstract

Tumour response heterogeneity to treatment is common across different foci in the same patient with cancer. However, the effect of heterogeneity on clinical outcome remains unclear. Here, we developed a quantitative assessment of tumour response heterogeneity and explored the correlation of tumour response heterogeneity with the clinical outcome. Patients eligible for this retrospective study had advanced lung adenocarcinoma, with 3-10 measured foci, an Eastern Cooperative Oncology Group status of 0-1, and received first-line chemotherapy or targeted therapy. Percent change of the longest diameter in each measurable lesion were assessed as per Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1. The tumour response heterogeneity score was the absolute value of the dispersion coefficient of the percentage value of each target lesion. Patients were randomly divided (1:1) into a learning and a validation set. The optimal cutoff value of tumour response heterogeneity score was identified according to progression-free survival information in the discovery cohort. We then confirmed and analysed the correlation of tumour response heterogeneity scores with clinicopathological features in the validation cohort. Next-generation sequencing was used to investigate a potential mechanism linking tumour response heterogeneity scores to clinical outcomes. Data from 174 patients were collected between Jan 1, 2016, and Dec 30, 2020. 101 (58%) patients had been treated with platinum-based doublet chemotherapy and 73 (42%) with targeted therapy. Median follow-up was 19·8 months (IQR 3·71-13·75). In the discovery cohort (n=85), the optimal cutoff point of tumour response heterogeneity score was defined as 0·46. Patients with high tumour response heterogeneity score had worse progression-free survival than patients with low tumour response heterogeneity (median 4·5 months [95% CI 3·6-8·6] vs 15·8 months (12·8-not reached), hazard ratio [HR] 4·43 [95% CI 2·14-9·16]; p<0·0001). In the validation cohort (n=89), high tumour response heterogeneity score was shown to be associated with significantly shorter progression-free survival (median 5·1 months (95% CI 3·5-7·8) vs 12·9 months (10·4-28·7), HR 2·69 [95% CI 1·59-4·55]; p<0·0001). The median value of tumour response heterogeneity score followed the order of progressive disease, stable disease, and partial response (p<0·0001). Patients with high tumour response heterogeneity score had worse overall response rate than patients with low tumour response heterogeneity score (30·0% vs 68·0%; Fisher's exact test p<0·0001). In the overall population (n=174), high tumour response heterogeneity score was further shown to be an independent biomarker for progression-free survival by univariate and multivariate analysis. Moreover, tumour response heterogeneity score was able to further distinguish the outcomes in partial response and stable disease subgroup. In the partial response group, high tumour response heterogeneity score was associated with significantly shorter progression-free survival in the partial response subgroup (HR 2·79 [95% CI 1·24-6·29; p=0·0006) and the stable disease subgroup (HR 2·55 [95% CI 1·44-4·49]; p=0·0015). Compared with low tumour response heterogeneity score, high tumour response heterogeneity score was associated with higher tumour mutation burden and high frequency variation of cell cycle signal pathway. The tumour response heterogeneity score, a novel metric for tumour response heterogeneity, was a powerful independent predictor of outcome in advanced lung adenocarcinoma. The tumour response heterogeneity score further stratified patients with the same RECIST assessment results and could be a useful addition to the RECIST assessment system. Further prospective studies are warranted. National Natural Science Foundation of China (grant 82072565) and Bejing Xisike Clinical Oncology Research Foundation (grant Y-2019AZZD-0386).