Complex event processing (CEP) systems are currently widely used in large-scale enterprises for the processing of high and dynamically changing rates of input events using large number of complex rules. Given the hardware limitations of vertically scaled CEP solutions, horizontal scalability has become an essential requirement for modern CEP systems. In this paper, we propose an adaptive load-balancing technique via rule distribution (called ARD) for a cluster of CEP engines that provides horizontal scalability for CEP systems. Our experiments show our proposed technique provides higher scalability and yields higher throughput in comparison with two previously proposed non-adaptive load-balancing techniques, namely VISIRI and SCTXPF, when the system faces with variable workload. In addition, ARD keeps the system balanced more often.