The 3-Axis Integrative Framework (3-AIF) for Difficult-to-Treat Rheumatic Disease: An ODE-Based Systems Model with Transcriptomic Support and with Clinical Plausibility Assessment

Difficult-to-treat (D2T) rheumatoid arthritis (RA), defined by EULAR in 2021, affects about 12% of RA patients and is often refractory to sequential single-pathway biological therapy. A mechanistic framework that explains D2T states as system-level phenomena has been lacking. We developed the 3-Axis Integrative Framework (3-AIF), a computational model based on three biological axes: Axis 1 (mucosal tolerance via AID-IgA-Microbiota), Axis 2 (energy-gated danger sensing via the Nerve-Adipose-Immune unit), and Axis 3 (cellular stress execution via ISR/ISRmt, gated by mTORC1-AMPK). The framework was formulated as a six-variable ordinary differential equation (ODE) system with Hill-function metabolic gating. Numerical simulation produced three trajectories: healthy resolution, chronic attractor trapping corresponding to D2T states, and therapeutic recovery with a lag phase. Phase portrait analysis showed bistable dynamics with two stable equilibria separated by an unstable saddle point. Bifurcation analysis showed fold catastrophe with hysteresis. Sensitivity analysis identified three dominant parameters: NAM activation coefficient (α2), energy drain rate (δ1), and recovery capacity (β3). We describe the Integrative Neuromuscular Acupoint System (INMAS), a wet-needling protocol combining peripheral nerve stimulation with metabolic co-factor micro-injection. To illustrate potential clinical relevance, we present an illustrative retrospective 15-patient case series covering seven rheumatic conditions: SLE (n=3, including lupus nephritis, post-splenectomy intractable SLE, and low-activity SLE with ISR-dominant fatigue), axial spondyloarthritis (n=1), RA (n=2, including seropositive RA with ILD and long-standing RA with JAK inhibitor-controlled pain but persistent ESR elevation), dermatomyositis (n=1), Sjogren syndrome (n=3), Behcet disease with fibromyalgia (n=1), and non-inflammatory chronic pain (n=4). The dominant axis varied across patients: Axis 1 in 3, Axis 2 in 7, Axis 3 in 3, and multi-axis in 2. Twelve of 14 INMAS-treated patients were observed to show trajectory patterns directionally consistent with model predictions during descriptive follow-up, while two patients with irreversible regulatory circuit disruption (post-splenectomy, advanced neurodegeneration) showed limited sustained response. These findings suggest that D2T disease may represent entrapment within a multi-axis attractor and may not be fully addressed by escalation of single-pathway immunosuppression alone.