Document resource
Background Despite widespread use of programmed death-ligand 1 (PD-L1) immunohistochemistry in gastrointestinal cancers, most immune checkpoint inhibitor trials fail to stratify meaningful benefit. In 2024, an FDA advisory panel rejected expanding immunotherapy for PD-L1-negative esophageal cancer, citing poor predictive response. This highlights the need for a more effective biomarker strategy. We hypothesized that PD-L1 adds limited predictive value once microsatellite instability (MSI) is known, and a tiered strategy incorporating MSI, tumor mutational burden (TMB), and POLE mutations could better guide treatment while reducing unnecessary testing.Methods We conducted a systematic review of phase II/III trials in advanced colorectal and gastric cancers reporting objective response rates stratified by MSI and PD-L1 status. Trials were included if they defined assay methodology and histology. Pooled response rates were calculated using a random-effects model; incomplete stratifications were qualitatively summarized. Clinical and genomic data from The Cancer Genome Atlas (TCGA) for colorectal and gastric tumors (n ≈ 1,025) were analyzed to evaluate overall survival based on MSI, TMB, and POLE mutation status using Kaplan-Meier and Cox regression. The cost of PD-L1 testing was estimated using 2023 Medicare reimbursement data.Results Seven trials were included: KEYNOTE-059, KEYNOTE-177, KEYNOTE-061/062, CheckMate-142, ATTRACTION-2, JAVELIN Gastric 100, and LEAP-005. Tumors with high MSI had a pooled objective response rate of 51% (95% CI, 45-57%), versus 9% in microsatellite-stable tumors. Within MSI-high tumors, PD-L1 status did not distinguish responders; in CheckMate-142, response rates were comparable, i.e 46% for PD-L1-positive and 41% for PD-L1-negative cases. In microsatellite-stable tumors, PD-L1-positive cases still had low response rate of 13%. The Cancer Genome Atlas analysis showed that only MSI-high and TMB-high tumors had superior overall survival (hazard ratio 0.67, p = 0.003). POLE mutations were seen in 2% colorectal and 3.3% of GI cancers, but with a high response rate of 48%. Economic data confirmed that eliminating routine PD-L1 testing could save $6 million annually in the United States without reducing the accuracy of immunotherapy selection.Conclusions PD-L1 status adds minimal predictive value once MSI is known and should be considered for de-implementation in gastrointestinal cancers outside harmonized research protocols. In contrast, MSI, TMB, and POLE mutations are robust predictors of immunotherapy benefit and should guide biomarker-based treatment selection. A tiered approach—(1) universal MSI testing, (2) reflex TMB and POLE testing in microsatellite-stable tumors, and (3) IFN-γ gene profiling for ambiguous cases—can streamline immunotherapy selection, enhance precision, and eliminate unnecessary spending on low-utility PD-L1 testing.