2011P - Machine-derived features of tumor infiltrating lymphocytes (TILs) as biomarkers of clinical outcomes to platinum chemotherapy in small cell lung cancer (SCLC)

Background

Platinum chemotherapy (PC) is the mainstay of treatment in SCLC. Most patients relapse with poor overall survival (OS) and develop resistance to further therapy. Presently, there are no biomarkers to predict response/resistance to initial treatment. TILs have shown to serve as a prognostic factor in several solid cancers but remain largely unexplored in SCLC. In this study, we explore the association of density and spatial arrangement of TILs with clinically relevant outcomes.

Methods

TIL features were extracted using computational algorithms from H&E whole slide images of pre-treatment SCLC biopsy across two cohorts, D1 (University Hospitals) and D2 (Roswell Park) (n=213). D1 (n=101) was used to identify the most prognostic features and train two classifiers to predict objective response to PC and OS. The classifier's performance was evaluated in an independent cohort D2 (n=112). D1 was split into Training (St)/Test (Sv) set (40/60) based on overall response to PC per RECIST; responders=complete/partial response, non-responders=stable/progressive disease and OS status. Boruta method of feature selection was used to select the most discriminating features. Linear classifier with qr factorization, linear least square approach was used to build a model for predicting response and Kaplan-Meier with log-rank test for estimating survival curve and risk stratification; Cox’s proportional hazards with elastic-net penalization for risk score. The model was evaluated using Area Under the Curve (AUC) and Precision Recall Curves (PRC) for OR and in terms of Hazard Ratio (HR), C-index (CI) and p-values for OS prediction.

Results

Features related to density variation of clusters of TILs and non-lymphocytes within the tumor area were associated with response to PC and OS. For St, AUC=0.96, PRC=0.98 (HR=2.84; 1.34-6.02, p=0.005, CI=0.75 0.60-0.90), AUC=0.66, PRC=0.76 (HR=2.06; 1.15-0.84, p=0.013, CI=0.7; 0.56-0.84) for Sv. For D2, validation AUC=0.67 and PRC=0.79 (HR=1.99; 1.08-3.7, p=0.026, CI=0.69; 0.54-0.83).

Conclusions

We demonstrate that TIL density and architecture features are associated with OR and OS. Additionally, independent multi-site validation is warranted.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

The authors.

Funding

Roswell Park Cancer Center's Alliance Foundation.

Disclosure

P. Jain: Financial Interests, Institutional, Other, Institutional Principal investigator: Harpoon Therapeutics; Financial Interests, Institutional, Other, Principal Investigator: Debiopharm; Financial Interests, Institutional, Other, Site Principal Investigator: Iovance Biotherapeutics, AADi Bioscience, SOPHiA Genetics, Sanofi; Financial Interests, Institutional, Local PI: Harpoon Therapeutics, Iovance Biotherapeutics, AADi Bioscience, SOPHiA Genetics, Sanofi; Financial Interests, Institutional, Coordinating PI: Debiopharm; Non-Financial Interests, Advisory Role: G1 Therapeutics; Non-Financial Interests, Other, Invited speaker: American Lung Association. A. Madabhushi: Financial Interests, Personal, Advisory Board, Serve on SAB and consult.: SimbioSys; Financial Interests, Personal, Advisory Board: Aiforia, Picture Health; Financial Interests, Personal, Full or part-time Employment: Picture Health; Financial Interests, Personal, Ownership Interest: Picture Health, Elucid Bioimaging, Inspirata Inc; Financial Interests, Personal, Royalties: Picture Health, Elucid Bioimaging; Financial Interests, Institutional, Funding: AstraZeneca, BMS, Boehringer-Ingelheim, Eli Lilly. All other authors have declared no conflicts of interest.