With the rapid explosion of data volume from mobile networks, edge caching has received significant attentions as an efficient approach to boost content delivery efficiency by bringing contents near users. In this article, cache-enabled heterogeneous networks (HetNets) considering the limited backhaul is analyzed with the aid of the stochastic geometry approach. A hybrid caching policy, in which the most popular contents are cached in the macro BSs tier with the deterministic caching strategy and the less popular contents are cached in the helpers tier with the probabilistic caching strategy, is proposed. Correspondingly, the content-centric association strategy is designed based on the comprehensive state of the access link, the cache and the backhaul link. Under the hybrid caching policy, new analytical results for successful content delivery probability, average successful delivery rate and energy efficiency are derived in the general scenario, the interference-limited scenario and the mean load scenario. The simulation results show that the proposed caching policy outperforms the most popular caching policy in HetNets with the limited backhaul. The performance gain is dramatically improved when the content popularity is less skewed, the cache capacity is sufficient and the helper density is relatively large. Furthermore, it is confirmed that there exists an optimal helper density to maximize the energy efficiency of the cache-enabled HetNets.