Assessing the sensitivity of treatment effect estimates to differential follow-up rates: implications for translational research