Narrowband Internet of Things (NB-IoT) is a new technology introduced by 3rd generation partnership project (3GPP) to meet low-power wide area networks (LPWAN) requirements. Accessing nodes with poor channel condition is of decisive importance in such networks. Repetition is a promising approach to enhance the network coverage in radio technologies suitable for machine-type communications including NB-IoT. This approach implies repeating the same transmission data or control signals back-to-back without getting feedback from the end point. Introducing this new feature raises the need for link adaptation in mainly two dimensions: 1) Selection of modulation and coding scheme (MCS) level and 2) Repetition number determination to guarantee the block error rate of less than 10%. In this paper, we propose a link adaptation scheme to minimize the transmission of control information for large packets that are segmented before sending. This is done by a new online SNR estimation scheme and adaptively sending downlink control information (DCI) only for necessary segments. According to the simulation results, our proposed scheme outperforms the conventional method by reducing the resource blocks consumption and active time by an average of 35% and 30%, respectively.