Performance is a key issue in the implementation of tools for policy-based management of large and complex networked systems. When a system is characterized by millions of policies, the policy decision point is usually a performance bottleneck for the whole system. Although a few researchers have proposed cache-based methods to improve the efficiency of the policy decision point, the mutable variables, e.g., time, location, and temperature, are usually hard to be dealt with. The reason is that when applicable policies contain mutable variables, the policy decision point has to re-evaluate conditions in policies for events with the same targets. This paper thus proposes a novel Mutable Variable Aware Cache mechanism, by which the entries in the cache are aware of the mutable variables. The conditions or sub-conditions that do not contain mutable variables in the applicable policies will be evaluated once during the life cycle of a policy decision point for specific events, and the evaluation results will be appended to the cache. This optimization can greatly reduce the cost of fetching the values of mutable variables and evaluating the conditions or sub-conditions. It can, therefore, improve the performance of the policy decision point. We classify the possible situations and propose key algorithms. We also conduct a performance evaluation, which shows that the Mutable Variable Aware Cache mechanism can significantly improve the efficiency of a policy decision point.