In estimating integrated volatility using high-frequency data, it is well documented that the presence of microstructure noise presents a major challenge. Recent literature has shown that the presence of multiple observations, a common feature in datasets, brings additional difficulty. In this study, we show that the pre-averaging estimator is still consistent under multiple observations, and the related asymptotic distribution of the estimator is established. We also show that the pre-averaging estimator based on multiple observations achieves the same asymptotic efficiency as the “ideal” estimator that assumes we know the exact trading times of all transactions. Simulation studies support the theoretical results, and we also illustrate the estimator using real data analysis.