This study estimates the dose-response relationship between Body Mass Index (BMI) and crash risk in commercial motor vehicle operators. Intake data was collected on 744 new truck drivers who were training for their commercial driver's licenses at a school operated by the cooperating trucking firm during the first two-week phase of instruction. Drivers were then followed prospectively on the job using the firm's operational data for two years, or until employment separation, whichever came first. Multivariate Poisson regression and Cox proportional hazards models were used to estimate the relationship between crash risk and BMI, controlling for exposure using miles driven, trip segments, and job type. Results from the Poisson regression indicated that the risk ratio (RR) for all crashes was significantly higher for drivers in the obesity Class II and Class III categories: RR= 1.6, confidence interval 1.2-2.1 and RR= 1.49, confidence interval 1.12-1.99, respectively. Similarly, the multivariate Cox Proportional Hazard model results showed that crash risk was significantly higher for obesity class II (BMI 35 to <40; relative risk (RR) = 1.47, P = 0.02) with weaker evidence of increased crash risk for obesity class III (BMI >40; RR = 1.35, P =0.06) when compared to normal BMI (BMI 18.5 to <25). The results of this prospective study establish an association between obesity and crash risk show that driver health behavior has important spillovers for public safety.