1411P - AI-powered immune phenotype predicts favorable outcomes of nivolumab (niv) plus chemotherapy (chemo) in advanced fgastric cancer (AGC): A multi-center real-world data analysis

Background

Niv+Chemo is one of the standard first-line regimens for AGC. Recent reports have consistently shown that immune phenotype (IP) assessment by an artificial intelligence (AI)-powered H&E analyzer could predict treatment response to immune checkpoint inhibitors in various cancer types. We aimed to evaluate the clinical value of AI-powered IP in AGC patients (pts) treated with Niv+Chemo as compared to Chemo.

Methods

H&E images of 585 pts (310 Niv+Chemo vs 275 Chemo) were collected from two tertiary hospitals in Korea. Inflamed IP (IIP) or Non-IIP was classified by Lunit SCOPE IO (Lunit, Seoul, Korea). Association between progression-free survival (PFS) and the IP status along with clinicopathologic factors was analyzed.

Results

During a median follow-up of 15.3 months (m), median PFS (mPFS) of Niv+Chemo group was significantly longer than that of Chemo group (8.2 vs. 5.9 m, hazard ratio [HR] 0.69, 95% confidence interval [CI] 0.57–0.84, P < 0.001). Overall, the IIP was called in 38.9% (228/585). The PFS benefit of Niv+chemo over chemo was more pronounced in the IIP: mPFS was 11.0 vs 5.8 m (HR 0.58, 95% CI 0.41–0.80, P < 0.001) in the IIP, whereas mPFS was 7.3 vs 5.9 m (HR 0.75, 95% CI 0.59–0.96, P = 0.021) in the Non-IIP. This was prominently observed in pts harboring a non-signet ring cell (SRC) histology (HR of Niv+chemo 0.47 vs. 0.66 in the IIP vs. non-IIP, respectively), as compared to pts with a SRC histology (HR of Niv+chemo 0.71 vs. 1.25 in the IIP vs. non-IIP, respectively). Among 227 pts whose PD-L1 combined positive score (CPS) by the 28-8 assay was available, the prominent PFS benefit of the IIP with Niv+chemo was consistently observed regardless of CPS: HR of Niv+chemo was 0.31 vs. 0.93 in the IIP vs. Non-IIP, respectively, for pts with CPS ≥ 5; and HR of Niv+chemo was 0.41 vs. 0.78 in the IIP and Non-IIP, respectively, for pts with CPS <5. Multivariate analysis of pts treated with Niv+Chemo revealed that the IIP was an independent factor for PFS (HR 0.64, 95% CI 0.44–0.94, P = 0.022), along with ECOG ≥ 2 (HR 4.19) and the microsatellite-high status (HR 0.39).

Conclusions

The IIP was significantly predictive for a PFS benefit of Niv+Chemo over Chemo, more prominently in pts with a non-SRC histology.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

The authors.

Funding

Lunit.

Disclosure

H. Kim: Financial Interests, Personal, Funding: AstraZeneca, Roche/Genentech; Financial Interests, Personal, Speaker, Consultant, Advisor: Bristol Myers Squibb, Ono Pharmaceuticals, Boryung Pharmaceuticals, Daiichisankyo, MSD, Boostimmune; Financial Interests, Personal, Advisory Role: Mustbio. C. Lee: Financial Interests, Institutional, Funding: Lunit. S.I. Cho, S. Song, C. Ock: Financial Interests, Personal, Full or part-time Employment: Lunit; Financial Interests, Personal, Stocks/Shares: Lunit. S. Hwang, H.Y. Kay: Financial Interests, Personal, Full or part-time Employment: Lunit. All other authors have declared no conflicts of interest.