Aim: The aim of this study was to investigate the effectiveness of metformin in diabetes prevention in a prediabetic population across a range of fasting plasma glucose (FPG) levels at baseline. A secondary aim was to assess the effectiveness of metformin in preventing diabetes in those participants where impaired fasting glucose (IFG) was relatively more pronounced as opposed to impaired glucose tolerance (IGT). Methods: Participants randomised to metformin and placebo arms in the Diabetes Prevention Program study were stratified into cohorts according to level of FPG at baseline. Cumulative incidence of diabetes for the different cohorts was assessed. Change in FPG, insulin sensitivity, and levels of fasting insulin and proinsulin for the different cohorts were also calculated. Results: The largest reductions in incidence of diabetes and FPG occurred within prediabetic persons with a higher level of FPG at baseline. Metformin was able to stabilise insulin sensitivity in every stratified sub-cohort except one. Sub-cohorts which had higher levels of insulin sensitivity at baseline experienced the largest increases in insulin sensitivity. Metformin reduced the incidence of diabetes by 43% (RR 0.57, CI 0.4–0.9) in those prediabetic persons whose IFG was more pronounced compared to a 26% (RR 0.74 CI 0.7–0.8) when all participants in the study were included. Conclusion: The largest reductions in both incidence of diabetes and FPG occurred in prediabetic persons with a higher level of FPG at baseline. Metformin was able to stabilise insulin sensitivity and was more effective in persons with more pronounced IFG.