Drilling into formations with narrow safety pressure windows may lead to abnormal downhole conditions such as gas kick. Aiming at gas kick early detection in the managed pressure drilling process, an adaptive observer is applied to estimate downhole uncertain parameters and unknown states, and based on this, a gas kick detection method is established. A simplified gas-liquid two-phase flow model is established based on the "dual-shape" bubble representation approach, and is compared with OLGA simulation data to verify the validity of the model. A Luenberger-like adaptive observer is designed and applied to estimate the key parameters and states of downhole gas kick conditions, then a gas kick detection method is constructed by combining the Hotelling 's T² test statistics. The results indicate that the designed adaptive observer can real-timely estimate the gas volume and accurately track changes of bottom hole pressure and outlet flow rate. Compared with acoustic method, delta flow method and pit volume method, the gas kick detection method can quickly and significantly achieve early detection of different gas kick situations.