A prediction model integrated genomic alterations and immune signatures of tumor immune microenvironment for early recurrence of stage I NSCLC after curative resection

Background: Surgery is the standard treatment for patients with stage I non-small cell lung cancer (NSCLC). However, postoperative recurrence leads to a poor prognosis of patients. Currently, there is no effective prognostic biomarker and perioperative treatment for patients with stage I NSCLC. Methods: One hundred thirty stage I NSCLC patients who had surgical resection were enrolled, including 69 patients who had recurrence within three years and 61 patients who had no recurrence (follow up more than five years). Whole exome sequencing was performed to evaluate gene mutation, copy number variation, and tumor mutation burden (TMB). Immunohistochemistry was carried out to assess the expression of PD-L1 and the level of CD3+ and CD8+ tumor-infiltrating lymphocytes (TILs). Tumor mutation score (TMS) was constructed with the recurrence-associated genes identified by Lasso regression. Immunoscore (IS) was built based on the location and density of CD3+ and CD8+ TILs. Logistic regression was performed to build a prediction model. 70% of patients were included in the training cohort and 30% patients in the testing cohort. P<0.05 was considered to be statistically significant. Results: Univariate analysis showed that lung adenocarcinoma (LUAD), MUC4 mutation, and high TMB were related to early recurrence (P=0.008, 0.0008, and <0.0001, respectively). CD3+ and CD8+ TILs within tumor center and invasive margin significantly negatively correlated with recurrence. EGFR mutation and PD-L1 expression had no association with recurrence. Early recurrence group had significantly higher TMS and lower IS (P<0.0001 and P=0.0003, respectively). Multivariate analysis showed that high TMS and low IS were independent predictors for early recurrence (P<0.0001 and P=0.001, respectively). Integrating TMS and IS, we built a recurrence-model with great discrimination and calibration in the training cohort (AUC =0.935; HL test, P=0.2885) and testing cohort (AUC =0.932; HL test, P=0.5515). Conclusions: High TMS and low IS were both poor prognostic factors for recurrence in stage I NSCLC. The integrated recurrence-model helps identify patients with high recurrence risk, which provides evidence for future research about perioperative treatment.

Automated summary

High TMS and low IS were identified as independent predictors for early recurrence in stage I NSCLC patients. Integration of TMS and IS in a predictive model could help identify patients at high risk of recurrence, providing evidence for perioperative treatment research in the future.