In the era of "treat-to-target" (T2T) strategies for axial spondyloarthritis (axSpA), the therapeutic paradigm has shifted towards achieving both clinical remission and imaging inactivity to prevent irreversible structural damage. Under the Assessment of SpondyloArthritis international Society (ASAS) framework, bone marrow edema (BME) detected on conventional fluid-sensitive magnetic resonance imaging (MRI) sequences serves as the cornerstone for identifying active inflammation. However, a significant clinical dilemma has emerged with the widespread use of biologic disease-modifying antirheumatic drugs (bDMARDs). Observational cohorts frequently demonstrate a time-lag effect: patients rapidly achieve a clinically inactive state, yet BME signals persist on MRI. Solely relying on these residual morphological signals often leads to the misinterpretation of aseptic tissue repair or biomechanical stress as refractory inflammation, consequently prompting unwarranted escalation of immunosuppressive therapies. This prospective study introduces an integrated 18F-FDG PET/MRI bimodal system to re-evaluate BME at the molecular level. By quantifying the anomalous glycolytic activity of macrophages, we aimed to verify the existence of a "metabolically silent" BME phenotype in clinically remitted axSpA patients, thereby providing a pathophysiological rationale for optimizing imaging criteria and guiding step-down therapeutic strategies.

This prospective, single-center observational cohort study consecutively enrolled 19 adult patients fulfilling the 2009 ASAS classification criteria for axSpA. To ensure the temporal consistency of disease activity parameters, clinical assessments (including BASDAI and Patient Global Assessment) and laboratory tests (CRP, ESR) were completed within 24 hours prior to imaging. All subjects underwent whole-spine and sacroiliac joint scanning using a state-of-the-art integrated time-of-flight (TOF) 18F-FDG PET/MRI system. To mitigate the partial volume effect and noise accumulation typically seen in sub-centimeter inflammatory lesions, the Bayesian penalized likelihood (Q.Clear) reconstruction algorithm was applied with a noise suppression penalty factor of beta = 400. Concurrently, the Dixon two-point fat-water separation technique was utilized to generate T2WI pure-water maps, significantly enhancing the visual signal-to-noise ratio of BME. Image post-processing involved rigid-elastic spatial co-registration. Based on the spatial concordance between MRI structural edema and PET metabolic uptake (Maximum Standardized Uptake Value, SUVmax), patients were stratified into three distinct phenotypic subgroups. Statistical evaluations included Spearman's rank correlation for continuous variables and the Mann-Whitney U test for non-parametric comparisons. A Z-score-based unsupervised hierarchical clustering algorithm was further applied to map the multidimensional distribution of these phenotypes.

Based on bimodal imaging characteristics, the cohort was classified into Group A (double-positive active inflammation, n = 11), Group B (metabolically silent BME, n = 5), and Group C (double-negative complete remission, n = 3). Baseline characteristics, including disease duration and the proportion of bDMARDs usage, were comparable between Group A and Group B (all P = 1.000), successfully controlling for potential confounding biases related to treatment history. Bivariate analysis revealed that the SUVmax of target lesions exhibited a strong, significant positive correlation with systemic inflammatory burden and physical symptoms, specifically with the Ankylosing Spondylitis Disease Activity Score (ASDAS-CRP) (r = 0.742, P < 0.001) and BASDAI (r = 0.746, P < 0.001). Crucially, phenotypic validation demonstrated that the comprehensive disease activity of Group B was significantly lower than that of the metabolically active Group A (mean ASDAS-CRP: 0.93 vs. 2.08, P = 0.005). All subjects in Group B scored < 1.3 on the ASDAS-CRP, confirming a state of strict clinical remission. Furthermore, the unsupervised hierarchical clustering heatmap corroborated these findings. Without any prior diagnostic labels, subjects in Group B automatically congregated into an independent topological branch. This specific cluster displayed a highly uniform multidimensional profile characterized by "high edema (positive MRI BME), low metabolism (baseline PET SUVmax), and low clinical symptoms."

This study provides objective evidence for the existence of a "metabolically silent" BME subpopulation in axSpA. Despite exhibiting high-intensity fluid signals on conventional MRI mimicking active sacroiliitis, these lesions are devoid of the enhanced glycolytic activity typical of polarized M1 macrophages. The pathophysiological foundation of this phenotype likely reflects a transitional phase of aseptic tissue repair—encompassing microvascular bed remodeling, extracellular matrix deposition, and adaptive responses to biomechanical stress—rather than highly invasive immune-mediated osteitis. The integration of 18F-FDG PET/MRI effectively differentiates structural remnants from true active immunological lesions. Recognizing this metabolically inert phenotype has important clinical implications for refining current T2T imaging assessment criteria. It serves as a valuable imaging adjunct, preventing the over-interpretation of false-positive BME and safeguarding patients from the prolonged use or unnecessary dose escalation of biologic agents, thereby minimizing the risks of opportunistic infections and reducing the socioeconomic burden of axSpA management.