Real-World and Clinical Trial Validation of a Deep Learning Radiomic Biomarker for PD-(L)1 Immune Checkpoint Inhibitor Response in Advanced Non-Small Cell Lung Cancer

Abstract

Purpose: This study developed and validated a novel deep learning radiomic biomarker to estimate response to immune checkpoint inhibitor (ICI) therapy in advanced non-small cell lung cancer (NSCLC) using real-world data (RWD) and clinical trial data.

Materials and methods: Retrospective RWD of 1,829 patients with advanced NSCLC treated with PD-(L)1 ICIs were collected from 10 academic and community institutions in the United States and Europe. The RWD included data sets for discovery (Data Set A-Discovery, n = 1,173) and independent test (Data Set B, n = 458). A radiomic pipeline, containing a deep learning feature extractor and a survival model, generated the computed tomography (CT) response score (CTRS) applied to the pretreatment routine CT/positron emission tomography (PET)-CT scan. An enhanced CTRS (eCTRS) also incorporated age, sex, treatment line, and lesion annotations. Performance was evaluated against progression-free survival (PFS) and overall survival (OS). Biomarker generalizability was further evaluated using a secondary analysis of a prospective clinical trial (ClinicalTrials.gov identifier: NCT02573259) evaluating the PD-1 inhibitor sasanlimab in second or later line of treatment (Data Set C, n = 54).

Results: In RWD Test Data Set B, the CTRS identified patients with a high probability of response to ICI with a PFS hazard ratio (HR) of 0.46 (95% CI, 0.26 to 0.82) and an OS HR of 0.50 (95% CI, 0.28 to 0.92) in the first-line ICI monotherapy cohort, after adjustment for baseline covariates including the PD-L1 tumor proportion score. In Clinical Trial Data Set C, the CTRS demonstrated an adjusted PFS HR of 1.03 (95% CI, 0.43 to 2.47) and an OS HR of 0.33 (95% CI, 0.14 to 0.91). The CTRS and eCTRS outperformed traditional imaging biomarkers of lesion size in PFS and OS for RWD Test Data Set B and in OS for the Clinical Trial Data Set.

Conclusion: The study developed and validated a deep learning radiomic biomarker using pretreatment routine CT/PET-CT scans to identify ICI benefit in advanced NSCLC.

FIG 1.

Diagram of the radiomic biomarker validation pipeline which includes image preprocessing, deep learning feature extractor, and survival model. This study evaluated one implementation of the CTRS only on the basis of imaging features and another implementation that combined the CTRS score with age, sex, treatment line, and manual lesion measurements as inputs (eCTRS). CT, computed tomography; CTRS, CT response score; eCTRS, enhanced CTRS; ICI, immune checkpoint inhibitor; OS, overall survival; PFS, progression-free survival; QC, quality check; SLD, sum of longest diameters.

FIG 2.

Kaplan-Meier survival plots for (top) PFS and (bottom) OS for patients identified to have high and low probability of ICI response by the imaging-only CTRS. Survival plots are shown for (left) 1L monotherapy cohort in the RWD Cross-Validation Data Set A (center) 1L monotherapy cohort in RWD Test Data Set B and (right) Clinical Trial Data Set C in which all patients received ICI therapy in the 2L+. 1L, first-line; 2L+, second or later line; CT, computed tomography; CTRS, CT response score; HR, unadjusted hazard ratio; ICI, immune checkpoint inhibitor; OS, overall survival; P, log-rank test P value; PFS, progression-free survival; RWD, real-world data.

FIG 3.

Forest plots demonstrating the association of the CTRS with (left) PFS and (right) OS after multivariate adjustment for sex, age, PD-L1 TPS, and histology. Multiple imputation with chained equations was used to account for variables with missing data in PD-L1 TPS and histology. Results are shown for the (top) 1L monotherapy cohort in RWD Cross Validation Data Set A, (middle) the 1L monotherapy cohort in RWD Test Data Set B, and (bottom) Clinical Trial Data Set C in which all patients received ICI therapy in the 2L+. Histology was not available for Data Set C. 1L, first-line; 2L+, second or later-line; CT, computed tomography; CTRS, CT response score; HR, Adjusted hazard ratio after multivariate adjustment with a Cox proportional hazard model and 95% CI; ICI, immune checkpoint inhibitor; OS, overall survival; P value, Wald test P value; PFS, progression-free survival; RWD, real-world data; SCC, squamous cell carcinoma; TPS, tumor proportion score.

FIG 4.

Kaplan-Meier survival plots for (top) PFS and (bottom) OS for patients identified to have high and low probability of ICI response by the eCTRS. Survival plots are shown for (left) the 1L monotherapy cohort in RWD Cross Validation Data Set A, (center) the 1L monotherapy cohort in RWD Test Data Set B, and (right) Clinical Trial Data Set C in which all patients received ICI therapy in the 2L+. 1L, first-line; 2L+, second or later line; CT, computed tomography; eCTRS, enhanced CT response score; HR, unadjusted hazard ratio; ICI, immune checkpoint inhibitor; OS, overall survival; P, log-rank test P value; PFS, progression-free survival; RWD, real-world data.

FIG 5.

The C-indices and their 95% CIs of the CTRS and eCTRS are compared with other known imaging biomarkers of RECIST SLD and total tumor volume, both measured from baseline CT images, for prediction of (top) PFS and (bottom) OS. For the RWD Cross-Validation Data Set A and RWD Test Data Set B, C-index results are presented for ICI all-comer and first-line ICI monotherapy cohorts. The results from Clinical Trial Validation Data Set C represent the 2L+ monotherapy cohort. CIs were estimated using a bootstrap method with random resampling of 1,000 times. 2L+, second or later-line; CT, computed tomography; CTRS, CT response score; eCTRS, enhanced CTRS; ICI, immune checkpoint inhibitor; OS, overall survival; PFS, progression-free survival; RWD, real-world data; SLD, sum of longest diameters.

FIG A1.

Study flow diagram. ALK, anaplastic lymphoma kinase; CT, computed tomography; EGFR, epidermal growth factor receptor; ICI, immune checkpoint inhibitor; NSCLC, non–small cell lung cancer; OS, overall survival; PFS, progression-free survival; RW, real-world.

FIG A2.

Forest plots demonstrating the association of the CTRS with (left) PFS and (right) OS after multivariate adjustment for the RWD all-comers cohort. Results are shown for the RWD Cross Validation Data Set A (upper) and the RWD Test Data Set B (lower). 2L+, second or later-line; CT, computed tomography; CTRS, CT response score; HR, hazard ratio; ICI, immune checkpoint inhibitor; OS, overall survival; PFS, progression-free survival; RWD, real-world data; SCC, squamous cell carcinoma; TPS, tumor proportion score.

FIG A3.

Kaplan-Meier survival curves stratified by the CTRS for the RWD subcohorts of patients with PD-L1 TPS ≥50% who received ICI as a monotherapy in the 1L. 1L, first-line; CT, computed tomography; CTRS, CT response score; HR, hazard ratio; ICI, immune checkpoint inhibitor; OS, overall survival; PFS, progression-free survival; RWD, real-world data; TPS, tumor proportion score.