Background: Trastuzumab deruxtecan (T-DXd) is an antibody-drug conjugate (ADC) approved for treatment of HER2-positive GC/gastroesophageal junction adenocarcinoma (GEJA) after chemotherapy (Japan) and for advanced/metastatic GC/GEJA after a trastuzumab-containing regimen (US/Israel/Singapore). Conventional HER2 immunohistochemistry (IHC) scoring identifies potential responders to trastuzumab but may not be suitable for next-generation ADCs capable of bystander killing, such as T-DXd [Modi NEJM 2022]. Use of computational pathology–based tools enables precise and accurate quantification of HER2 expression and spatial analysis, which enables design of better scores to reflect the T-DXd mechanism of action. Methods: Quantitative continuous scoring (QCS) was performed to quantify HER2 expression and distribution per cell in digitized images of HER2-stained (IHC, Ventana 4B5) GC. A QCS-derived signature was identified, called the continuous spatial proximity score (cSPS), which considers spatial location of tumor cells in addition to HER2 expression level and can be used as a surrogate for potential bystander activity. We retrospectively stratified patients in DESTINY-Gastric01 (DG-01; NCT03329690; T-DXd arm: n=123^a; standard of care [SOC] arm: n=59^a) using cSPS to give a binary QCS biomarker positive (BM+) or negative (BM−) status. QCS cSPS signature was independently validated on the DESTINY-Gastric02 cohort (DG-02; NCT04014075; n=71^a). Results: In the DG-01 cohort, enrolled per HER2-positive status by IHC using ASCO/CAP GC scoring guidelines, median progression-free survival (mPFS) was 5.5^a mo with T-DXd and 2.8^a mo with SOC. QCS-derived cSPS scores showed significant survival benefit for BM+ vs BM− in the DG-01 T-DXd arm (P<0.0001; HR=0.37 [0.24-0.58]) and longer mPFS for BM+ (8.3 mo) vs BM− (3.9 mo). Within the SOC arm, this signature showed the inverse prognosis (P=0.034; HR=2.17 [1.06-4.44]; mPFS, 2.8 mo for BM+ vs 4.9 mo for BM−), strongly suggesting the signature is predictive of T-DXd response. This signature was validated in the DG-02 cohort (mPFS, 5.5 mo on enrollment by IHC status, 10.1 mo for BM+ vs 3.7 mo for BM−; HR=0.24 [0.13-0.45]; P<0.0001). Conclusions: We established QCS, a novel scoring method for tissue biomarkers. Our validated data and results support possible future use of this quantitative approach against manual scoring for identification of patients with GC who may benefit from T-DXd therapy. Clinical trial information: NCT03329690, NCT04014075.

NA, not applicable. ^a Results reported for a subset with available digital images.