<p>Patch antennas are widely used in wireless applications, where the radiation patch configurations should be designed carefully to adjust the antenna radiation characteristics. To meet the demand for improving radiation performance, the antenna configuration tends to be more complex. Due to the full-field coupling effect between the radiation patch (denoted as design domain) and other components of the patch antenna (i.e., the multi-layer substrates and complex supporting structures, denoted as non-designable domain), the entire antenna must be totally included in the full-wave analysis model for the topology optimization of the radiation patch, which significantly increases the computational cost of the iterative optimization. To address this issue, this paper proposes an efficient topology optimization method for patch antenna design using the condensation technique. The patch antenna performance is computed using the volume-surface integral equation-based method of moments, and the radiation patch configurations are described by an additional impedance layer on the metallic surface. Based on this, the impedance submatrices for designable and non-designable domains are constructed independently. Then, the governing equation and the adjoint equations for sensitivity analysis are condensed using the condensation technique, significantly reducing the computation time and the memory usage. The main contribution of the proposed method is its ability to effectively design the radiation patch in complex antenna configurations that satisfy the design requirements, including the radiation polarization and the impedance matching characteristics. Typical patch antennas are designed to verify the effectiveness of the proposed method. Compared with directly solving the uncondensed governing equations, the proposed method saves significant computational time in each iteration of the optimization design.</p>