The Effects of Missing Data Due to Study Dropout on Longitudinal Analysis Inference Using Outcome-Dependent Sampling