Methods. Data came from the U.S.–Mexico Border Diabetes Prevention and Control Pro- ject survey (2001–2002), a stratified, multistage sample representative of the border popula- tion. The analytic sample included adults from all racial and ethnic backgrounds with diabetes (n = 665), including 333 Mexicans and 268 U.S. Hispanics. Smoking behaviors were based on self-reports. Age- and gender-specific prevalence of smoking behavior was estimated and lo- gistic regression was used for mediation analysis of group differences.