Previous literature indicates that pre-diagnostic diabetes and blood glucose levels are inversely related to glioma risk. To replicate these findings and determine whether they could be attributed to excess glucose consumption by the preclinical tumour, we used data from the Apolipoprotein MOrtality RISk (AMORIS) (n = 528,580) and the Metabolic syndrome and Cancer project (Me-Can) cohorts (n = 269,365). We identified individuals who were followed for a maximum of 15 years after their first blood glucose test until glioma diagnosis, death, emigration or the end of follow-up. Hazard ratios (HRs), 95% confidence intervals (CIs) and their interactions with time were estimated using Cox time-dependent regression. As expected, pre-diagnostic blood glucose levels were inversely related to glioma risk (AMORIS, P trend = 0.002; Me-Can, P trend = 0.04) and pre-diagnostic diabetes (AMORIS, HR = 0.30, 95% CI 0.17 to 0.53). During the year before diagnosis, blood glucose was inversely associated with glioma in the AMORIS (HR = 0.78, 95% CI 0.66 to 0.93) but not the Me-Can cohort (HR = 0.99, 95% CI 0.63 to 1.56). This AMORIS result is consistent with our hypothesis that excess glucose consumption by the preclinical tumour accounts for the inverse association between blood glucose and glioma. We discuss additional hypothetical mechanisms that may explain our paradoxical findings.