It is well known that a single NURBS patch is inadequate to describe design domains with complex geometry, due to its tensor product nature. In practical engineering applications, complex geometry is often represented by several non-uniform rational B-splines (NURBS) patches with multiple interfaces in computer-aided design (CAD). To solve minimum mean compliance topology optimization problems in this situation, a novel density-based isogeometric topology optimization approach combined with Nitsche’s method is proposed. In this method, adjacent NURBS patches are connected using Nitsche’s method to enforce kinematic and mechanical equilibrium on interfaces. Moreover, the coercivity of the Nitsche-type bilinear form is regulated by the stabilization parameters, which are independently estimated using a robust weighted form for each element. The performance of the proposed method is demonstrated through a series of 2D and 3D numerical examples in complex design domains.