Background: MMRd and MSI are prognostic for survival in many cancers and predict resistance to fluoropyrimidines in colon cancer. We examined the relationship between MMRd, MSI and survival in patients (pts) with resectable gastroesophageal cancer randomised to surgery alone or perioperative ECF chemotherapy in the MRC MAGIC trial. Methods: For MMR, tumour sections were assessed for expression of MMR proteins MLH1, MSH2, MSH6 and PSM2. MSI status was determined using the Promega MSI Analysis System (NR-21, BAT-26, BAT-25, NR-24, MONO-27). MMR status was correlated with MSI and the relationship between MMR status and survival was assessed. Results: MMRd results were available for 302 pts (66% resected pts). Sixteen (5%), 3 (1%), 2 (1%) and 18 (6%) pts were deficient in MLH1, MSH2, MSH6 and PMS2 respectively. Fourteen of 16 MLH1-ve tumours (88%) with available MSI results had MSI compared to 2/269 (1%) MLH1+ve tumours. Pts with MLH1-ve tumours treated with chemotherapy plus surgery had inferior median survival compared to MLH1+ve pts (7.2m vs 22.3m) whereas for pts treated with surgery alone the effect was in the opposite direction; median survival for MLH-ve pts was not reached, that for MLH1+ve pts was 20.3m. The interaction test between MHL1 and treatment was significant p=0.01. MSH2, MSH6 and PMS2 did not demonstrate similar relationships. Conclusions: In MAGIC, 88% of MLH1-ve pts demonstrated MSI and MLH1 deficiency was associated with a positive prognostic effect in pts treated with surgery alone and a differentially negative prognostic effect in pts treated with chemotherapy. These results are consistent with those reported for MSI status in this dataset. If independently validated, either MSI or MMRd (specifically MLH1) status on pre-operative biopsies could be used to select patients for peri-operative chemotherapy.

*only patients with date of surgery included in survival analysis