This report summarizes evidence presented during the Third Annual Ignition Interlock Symposium at Vero Beach, Florida, 29 October 2002. The ignition interlock prevents a car from starting when blood alcohol concentration (BAC) is elevated. We review some of our prior work as well as introduce previously unpublished results to demonstrate the manner in which the data recorded by the alcohol ignition interlock device can serve as an advance predictor of future driving under the influence (DUI) of alcohol risks. Data used in this current report represent approximately 2,200 ignition interlock users from Alberta, Canada, and about 8,000 interlock users from Quebec, Canada; the Alberta data set contained 5.5 million breath tests and the Quebec data 18.8 million breath tests. All tests are time and date stamped and this information was used to characterize patterns of BAC and vehicle use, and the relationship between BAC elevations and DUI offenses that accumulated after the interlock was removed from the vehicles. Findings from Cox regression (Marques et al., 2003) show that BAC elevations > .02-.04% are more potent predictors of repeat DUI (p