OBJECTIVE A genome-wide association study in the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial identified two markers (rs57922 and rs9299870) that were significantly associated with cardiovascular mortality during intensive glycemic control and could potentially be used, when combined into a genetic risk score (GRS), to identify patients with diabetes likely to derive benefit from intensive control rather than harm. The aim of this study was to gain insights into the pathways involved in the modulatory effect of these variants. RESEARCH DESIGN AND METHODS Fasting levels of 65 biomarkers were measured at baseline and at 12 months of follow-up in the ACCORD-Memory in Diabetes (ACCORD-MIND) MRI substudy (n = 562). Using linear regression models, we tested the association of the GRS with baseline and 12-month biomarker levels, and with their difference (Δ), among white subjects, with genotype data (n = 351) stratified by intervention arm. RESULTS A significant association was observed between GRS and ΔGLP-1 (glucagon-like peptide 1, active) in the intensive arm (P = 3 × 10−4). This effect was driven by rs57922 (P = 5 × 10−4). C/C homozygotes, who had been found to derive cardiovascular benefits from intensive treatment, showed a 22% increase in GLP-1 levels during follow-up. By contrast, T/T homozygotes, who had been found to experience increased cardiac mortality with intensive treatment, showed a 28% reduction in GLP-1 levels. No association between ΔGLP-1 and GRS or rs57922 was observed in the standard treatment arm. CONCLUSIONS Differences in GLP-1 axis activation may mediate the modulatory effect of variant rs57922 on the cardiovascular response to intensive glycemic control. These findings highlight the importance of GLP-1 as a cardioprotective factor.