We explored this issue considering bivariate spline models, where smoking intensity and duration were treated as interacting continuous exposures.