Background: T + chemo is standard 1st line therapy (tx) for HER2+ GEA patients (pts). However, pts typically progress within 6-8 months, with up to 30% demonstrating loss of HER2 positivity. No HER2 targeted agents have been shown effective in the post-T setting. We report the results of M, a novel anti-HER2 monoclonal antibody + P (anti-PD-1) in HER2+ GEA pts in 2^nd line post T progression and describe a biomarker enrichment strategy that compensates for tumor heterogeneity. Methods: HER2+ (pre-T testing) PD-L1-unselected GEA pts were enrolled. Endpoints include safety, objective response rate (ORR) by RECIST, and progression free survival (PFS). Exploratory endpoints include ERBB2 amplification (amp) status pre-M+P tx circulating-tumor DNA (ctDNA) by NGS (Guardant360) and PD-L1 on archival tissue by IHC (22C3 pharmDx). Results: As of 1/10/18, 60 pts were dosed in cohort expansion; 30 in North America (NA), 30 in Asia (A). Tx was well tolerated, with 13% of pts having TRAE ≥ grade 3, 3 drug-related SAEs, and 2 cases of autoimmune hepatitis and 1 pneumonitis. ERBB2 amp was detected by ctDNA in 61% of pts tested, mainly distal gastric cancer (GC) over proximal GEJ pts, and was inversely correlated with acquired driver mutations. PD-L1 was positive in 34% of pts tested, and was independent of clinical metrics but was enriched in ctDNA ERRB2+ pts (41% vs. 33%). Of 57 evaluable pts to date in expansion (30 NA and 27 A), best ORR was 16% and disease control rate (DCR) was 54%. Both ctDNA ERBB2 amp and PD-L1 positivity predicted response (24% vs. 0% [p = .0655] and 36% vs. 5% [p = .0367], respectively). In ctDNA ERBB2+/PD-L1+ pts, which were all GC, the ORR was 57% and DCR 86%. Conclusions: M+P is a well-tolerated regimen with antitumor activity in 2^nd line HER2+ GEA. Consistent with prior tissue-based reports, many GEA pts progressing on T have lost ERBB2amp. ERBB2 status by ctDNA NGS post-T could predict response to M+P, particularly in PD-L1+ pts. Our results suggest that M+P has encouraging preliminary activity in patients with advanced GC, and that biomarker selection based on ctDNA ERBB2 and PD-L1 could enrich for the responding population. Clinical trial information: NCT02689284