This correspondence paper analyzes the performance of a multiuser underlay uplink non-orthogonal multiple access network. A subset of users are selected for concurrent uplink transmission in any coherence interval while ensuring that the interference temperature limit (ITL) of the primary receiver is satisfied by the overall interference from all these users. Analytical expressions are derived for non-outage probability and throughput of the secondary network. Channel knowledge is utilized for intelligent apportioning of the ITL between all transmitting nodes in a novel channel-aware ITL apportioning(CA-ITLA) scheme. The performance of the scheme is compared to benchmark schemes like static ITL apportioning (S-ITLA) and orthogonal multiple access (OMA). It is shown that exploiting channel state information (CSI) allows the CA-ITLA scheme to achieve large throughput gains. It is shown that indirectly exploiting primary channel knowledge through a CSI dependent ITL results in even larger throughputs. Useful upper bounds onperformance are also derived with static and CSI-dependent ITL. Monte Carlo simulation results validate the presented analysis