This brief deals with the problem of guaranteed cost control for a class of uncertain networked control systems with time-varying delay. An improved predictive controller design strategy is proposed to compensate for the delay and data dropout in both the forward and backward channels to achieve the desired control performance. The varying controller gains which are designed to vary with delays can lead to less conservative results. Meanwhile, an algorithm involving a convex optimization problem is presented to achieve a suboptimal guaranteed cost. Furthermore, a numerical simulation and a practical experiment are given to illustrate the effectiveness of the proposed method.