Obtaining VoIP security through the IPsec protocol increases packet size, thus substantially increasing the network overhead. One way of mitigating such a problem consists of multiplexing calls so that the transmission of redundant data is reduced. The performance of such technique is directly linked to the number of concurrent calls; however, in more constrained scenarios, the existence of several simultaneous calls is not a premise that may be always assumed. Expanding packets' grouping time is an approach that can improve data compression in scenarios with only a few simultaneous calls; notwithstanding, such an approach reduces the quality of calls, increasing the end-to-end delay. In this work we propose an adaptive multiplexing solution based on the E-model recommendation, which objective consists of increasing the multiplexing capacity yet respecting the expected levels of quality during conversation. Results have shown that the proposed solution makes it possible to attain a higher compression rate when compared to traditional methods in favorable scenarios and better quality in unfavorable ones.