Time-Dependent Exposures and the Fixed-Cohort Bias