Background: In patients with mismatch repair deficient (dMMR) colorectal cancers (CRCs), we previously reported that loss of expression of MSH2 and MSH6 (MutS co-loss) was associated with better response to ICIs and longer median overall survival (mOS) compared to loss of expression of MLH1 and PMS2 (MutL co-loss). Here, we expanded our analysis and included gastrointestinal (GI) non-CRCs and explored the impact of dual vs. monotherapy ICIs on mOS in GI (CRC and non-CRC) cancers. Methods: Specimens were profiled by next-generation sequencing (592, NextSeq; WES, WTS NovaSeq) (Caris Life Sciences, Phoenix, AZ). MMR/microsatellite instability (MSI) status was determined by immunohistochemistry (IHC) of MMR protein. Real world OS was extracted from insurance claims and calculated using Kaplan-Meier estimates for molecularly defined cohorts from first treatment with ICIs (Nivolumab, Nivo; Ipilumumab, ipi; or Pembrolizumab, pembro) to last contact. Statistical significance was determined using chi-square and Mann-Whitney U test with p-values adjusted for multiple comparisons (q < 0.05). Results: The GI non-CRC cohort (N= 19,767) included cancers of the esophagus, stomach, gastroesophageal junction, pancreas, bile duct and small bowel with 97 (0.49%) patients having MutS co-loss, and 494 (4.03%) patients having MutL co-loss. MutS co-loss was associated with increased KRAS (45.4% vs 25.2%, q<0.01), CDKN2A (29.2% vs 9.0%), GNAS (27.8% vs 7.8%) and SMAD4 (17.5% vs 5.9%) mutations compared to MutL co-loss. Independent of treatment, MutS co-loss (N=74) had improved mOS compared to MutL co-loss (N=332) (40.4 m vs. 26.2 m, HR = 0.66; (95% CI: 0.46-0.95), P=0.024). In patients treated with ICIs, the mOS in MutS co-loss (N=21) was better compared to MutL co-loss (N=76) (not reached (NR) vs. 25.4 m, HR= 0.23 (95% CI: 0.07-0.75), p=0.008). At 3 years, more than 80% of the patients with MutS co-loss were alive. Of particular importance, when looking at all GI (CRC and non-CRC) patients, the mOS of MutL co-loss treated with ipi/nivo (N=18) trended for better mOS compared to MutL co-loss treated with pembro (N=215) (NR vs. 28.2m (HR=0.39; (95% CI:0.14-1.07), p=0.057), while the mOS of MutS co-loss treated with ipi/nivo (N=6) was not different compared to MutS co-loss treated with pembro (N=44) (NR vs. NR, HR=0.75 (95% CI: 0.094-5.92), p=0.78). Conclusions: In ICI-treated GI non-CRCs, the mOS was longer in MutS co-loss compared to MutL co-loss. In ICI-treated GI (CRC and non-CRC) patients with MutL co-loss, there was a trend for better survival with ipi/nivo compared to pembro. Our data suggest that the MutS vs. MutL status may guide the choice of ICIs regimen (Dual vs. Monotherapy) but more data are needed.