The rapid expansion of the Internet of Things (IoT) has intensified demands for scalable traffic management, real-time visibility, and resilient control in large, heterogeneous networks. Conventional Software-Defined Networking (SDN) architectures, typically based on centralized control and static forwarding, struggle to address the dynamic, data-intensive behavior of Software-Defined IoT (SD-IoT) systems. To overcome these limitations, this paper proposes MC-LBTO, a modular multi-controller framework that integrates programmable data plane intelligence with adaptive, secure coordination among distributed controllers to optimize load balancing and network efficiency. MC-LBTO comprises three cooperative modules: PDSM (P4-enabled Dynamic State Monitoring) for in-switch traffic observation and analytics; PALB (P4-based Adaptive Load Balancer) for latency-aware and fair traffic distribution; and STAM (Secure Trusted Adaptive Multi-Control) for consistent, fault-tolerant inter-controller operation. Experimental evaluations demonstrate that PDSM reduces controller CPU utilization by 35.7%, enhances flow-state detection accuracy to 98.3%, and lowers monitoring latency by 22.6% compared with OpenFlow-based monitoring. PALB achieves a 36% reduction in request latency, a 25% throughput increase, and a load distribution variance of only 5.5%, outperforming both traditional and modern probabilistic baselines. STAM enhances control-plane robustness with a Mean Time to Recovery (MTTR) of 75 ms, a 70% reduction in packet loss, and state consistency and security indices (SCI/SLI ≥ 9) under failure conditions. Collectively, these results confirm that MC-LBTO enables a scalable, secure, and self-adaptive SD-IoT architecture that maintains low overhead, balanced resource utilization, and fast recovery, offering a technically grounded framework for dependable and high-performance IoT networking.