ADVERTISEMENT
Artificial Intelligence-Powered Spatial Analysis of TILs Predicts Treatment Outcomes With ICIs in NSCLC
To analyze the spatial distribution of tumor-infiltrating lymphocytes (TILs), which helps to predict the effectiveness of immune checkpoint inhibitors (ICIs), researchers from the Republic of Korea developed a whole-slide image analyzer powered by artificial intelligence (AI). They then showed that this analyzer could detect 3 immune phenotypes that correlated with tumor response to ICIs and survival in 2 independent cohorts of patients with advanced non–small-cell lung cancer (NSCLC) treated with ICIs (J Clin Oncol. 2022. Published online March 10. doi:10.1200/JCO.21.02010).
Although ICIs are a standard treatment for advanced NSCLC that expresses programmed death ligand-1 (PDL-1), outcomes depend on the tumor microenvironment, for which no standard biomarker exists. Measuring the spatial distribution of TILs may serve as a biomarker, but accomplishing this manually limits this method’s utility and objectivity as well as the reproducibility of its results.
“In theory, TILs are the main activator of antitumor immunity and could be a promising biomarker if TILs can be objectively assessed throughout the whole tumor microenvironment,” wrote Se-Hoon Lee, MD, PhD, Division of Hematology-Oncology, Department of Medicine, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Republic of Korea, and colleagues.
Their proof-of-concept study aimed to turn theory into practice by showing that the AI-powered Lunit SCOPE IO spatial analyzer could segment and measure multiple histologic components, including TILs, on whole-slide images. This study also showed that the 3 immune phenotypes identified by the analyzer correlated with the outcomes of ICI treatment, which may make the analyzer a useful tool for optimizing treatment selection.
The 3 immune phenotypes detectable by the analyzer are inflamed, immune-excluded, and immune-desert. When compared with the latter 2 immune phenotypes, the inflamed immune phenotype correlated with greater local immune cytolytic activity, a higher rate of response to ICIs, and prolonged progression-free survival (PFS) in patients who had this phenotype.
According to the researchers, 44% of tumors they analyzed were inflamed, 37% were immune-excluded, and 19% were immune-desert. In those with an inflamed immune phenotype, median PFS was 4.1 months, and OS was 24.8 months. In contrast, in those with an immune-excluded phenotype, median PFS was 2.2 months and OS was 14 months, and in those with an immune-desert phenotype, median PFS was 2.4 months, and OS was 10.6 months.
Furthermore, patients with a PDL-1 tumor proportion score (TPS) <1% had a 32% incidence of the inflamed immune phenotype, those with a PDL-1 TPS of 1% to 49% had a 43% incidence of this phenotype, and those with a PDL-1 TPS ≥50% had a 57% incidence of this phenotype.
These findings are particularly useful in light of the superiority of pembrolizumab to standard chemotherapy in tumor response rate and survival in patients with a PDL-1 TPS ≥50% vs. the similar outcomes found for both treatments in patients with a PDL-1 TPS between 1% and 49%. “Therefore, development of a novel biomarker to predict ICI response in the clinical setting in patients with metastatic NSCLC with PD-L1 TPS 1%-49% is highly warranted,” wrote Dr. Lee and team.
In addition, when the researchers used their AI model to analyze TILs in whole-slide images, they found a positive correlation between the TPS generated by the model and the TPS determined by pathologists (P<.001).
“This is potentially a supplementary biomarker to TPS a determined by a pathologist,” the researchers concluded regarding their analyzer’s result.
