Graves’ orbitopathy (GO), an autoimmune condition affecting approximately half of patients with Graves’ disease (GD), significantly impairs quality of life and psychosocial well-being. In 3%–5% of cases, the condition progresses to severe, sight-threatening optic neuropathy (1). Pathological studies have demonstrated high expression of thyroid-stimulating hormone receptors (TSHR) in orbital tissues and fibroblasts (2, 3). Anti−TSH receptor autoantibodies (TRAb), which serve as a critical biomarker for GO, bind to TSHR and activate downstream inflammatory pathways, including the PI3K-mTOR and PKA-cAMP signaling cascades. This molecular interaction promotes the accumulation of hyaluronic acid, adipogenesis, and lacrimal gland enlargement, ultimately to orbital tissue remodeling. The resulting changes give rise to various clinical manifestations of GO, including eyelid retraction, periorbital edema, proptosis, diplopia, lacrimation, and orbital pain (4).

Significant advances have been made in the detection of TRAb. The third-generation TRAb assay, which utilizes TSHR-coated ELISA plates, is the most widely employed method, achieving specificity and sensitivity exceeding 97% (5, 6). However, this assay measures total thyrotropin-binding inhibitory immunoglobulin (TBII) and does not distinguish between functionally stimulating antibodies (thyroid-stimulating immunoglobulin, TSI) and blocking antibodies (TBAb). Although TSI represents the truly pathogenic antibody subset, its measurement has traditionally been limited by procedures that are complex, time-consuming, and costly. Recently, a novel bridge-based TSI binding assay has been developed, which directly quantifies TSI levels by exploiting the formation of a “bridge” between two distinct TSHR molecules. This assay is highly specific for the active receptor conformation, although it may also detect a subset of blocking antibodies in addition to TSI (7). It has demonstrated superior concordance with clinical diagnoses of GD compared with conventional TBII assays and shows a positive correlation with the TSI bioassay (8–10).

Quantification of TSI specifically assesses the stimulatory component of TBII and may provide enhanced diagnostic sensitivity for GD and related autoimmune thyroid disorders. However, whether TSI outperforms TBII in evaluating GO activity, severity, or treatment response remains a matter of debate (11–14). Clinical manifestations of GO also exhibit marked age-related heterogeneity, with older patients frequently presenting with more severe disease. While specific biomarkers, such as CD3+ CD4+ T cells, have shown potential in age-stratified analyses (15, 16), it remains unclear whether TBII or TSI levels exhibit similar age-dependent variations in their associations with GO features.

In addition, both TBII and TSI hold promise as biomarkers of treatment response. Previous studies, including recent comparisons of TBII, bridge-based TSI, and bioassay-based TSI, have demonstrated their utility in predicting outcomes following IVMP monotherapy (17–20). Furthermore, the lipid profile, particularly low-density lipoprotein (LDL) levels, has been implicated in immune dysregulation and may serve as a predictor for the response to IVMP in GO patients (21). However, despite the recommendation of IVMP and mycophenolate mofetil (MMF) combination therapy as first-line treatment for moderate-to-severe active GO in the 2021 EUGOGO guidelines (22), whether these biomarkers can predict treatment outcomes remains unexplored.

The clinical activity score (CAS) is a standard tool for assessing GO activity; however, studies suggest that it may not fully capture disease activity, particularly in Asian populations, where low CAS scores can coincide with progressive clinical manifestations (23, 24). Quantitative orbital MRI provides a valuable objective complement to clinical assessment. During active inflammation, increased signal intensity within the extraocular muscles on fat-suppressed sequences (such as T2-Dixon)—correlates with the severity of inflammation. This can be quantified using signal intensity ratios relative to reference tissues such as cerebral white matter (25, 26) with the maximum value recorded as the maximum signal intensity ratio (SIR max). Additionally, increased extraocular muscle volume (EMV) has been associated with greater disease severity in GO (27, 28). Lacrimal gland involvement is also common, and the maximum cross-sectional area of the lacrimal gland (LG area) is recognized as a highly sensitive imaging marker of disease activity (29).

Therefore, this retrospective study has two objectives: (1) to comprehensively evaluate the correlations of TBII and TSI levels with clinical manifestations of GO and quantitative orbital MRI parameters (SIR max, EMV, LG area), and to investigate whether these correlations are age dependent; and (2) to compare the prognostic value of baseline TBII and TSI levels in predicting treatment outcomes following IVMP-MMF combination therapy.

This study evaluated TBII levels using third-generation assays and quantified TSI using the IMMULITE 2000 system in newly diagnosed GO patients and those with a history of IVMP treatment. The results demonstrated that both TBII and TSI levels were positively correlated with orbital tissue involvement in the newly diagnosed group. Moreover, both assays exhibited comparable efficacy in differentiating active from inactive disease within this cohort, aligning with previous findings (12, 13, 20). This concordance is likely attributable to the predominance of TSI over TBAb in GO, as the latter is detectable in only a minority of cases (14, 37). Additionally, our analysis revealed that TSI showed significantly higher detection rates than TBII in both active and inactive subgroups within the newly diagnosed population. This enhanced sensitivity supports the superior utility of TSI for disease detection across the clinical spectrum of thyroid eye disease, particularly in inactive cases, which often present with subtle features such as mild extraocular muscle enlargement without overt clinical or inflammatory imaging findings. This observation further suggests that TSI may offer additional diagnostic value in differentiating GO from other orbital conditions with similar presentations, such as idiopathic orbital inflammation, allergic conjunctivitis, and IgG4-related disease, which may manifest with only mild eyelid swelling and LG enlargement on orbital imaging (38, 39).

In contrast, within the IVMP history group, neither TBII nor TSI showed significant correlations with disease activity parameters (SIR max, EMV, CAS), nor were they able to discriminate between active GO patients. Previous evidence indicates that IVMP significantly reduces TBII levels, with the most pronounced effects occurring within the first 3 months; however, this influence diminishes beyond 6 months and may even be followed by a rebound (40–42). In our cohort, the median interval since the last IVMP treatment was 4 (3, 7.5) months, placing patients in a transitional phase during which the uniformity of the initial treatment effect wanes, leading to considerable heterogeneity in individual antibody decay rates. It is this dissolution of a consistent treatment effect and the resulting interindividual variability—rather than a uniformly suppressed antibody state—that likely disrupts the stable correlation between antibody titers and the orbital disease burden. This phenomenon may account for the diminished predictive utility of both assays in this specific patient group.

A key finding of this study was the critical influence of age on the ability of TBII and TSI levels to reflect orbital involvement and disease activity. Previous studies have reported inconsistent correlations between TBII levels and CAS in younger cohorts (median/mean age <45 years) (13, 43–45), whereas significant associations have been consistently observed in older populations (median/mean age > 45 years) (14, 20, 46–48). Our data are consistent with this pattern, demonstrating improved biomarker accuracy in reflecting orbital involvement among older patients with GO. The mechanisms underlying this age-dependent effect warrant further consideration. Initial hypotheses proposed age-related differences in TSHR expression within orbital tissues; however, recent evidence has shown no significant correlation between orbital TSHR abundance and age (49). Instead, age-related functional differences in orbital fibroblasts (OFs)—the principal effector cells in GO—may contribute to this phenomenon. Notably, older patients with muscle-predominant GO exhibit distinct circular RNA expression profiles and differential pathway activation in OFs, suggesting that cellular responses to autoantibodies may be modulated by age-dependent molecular programs (50). Cytokine-stimulated OFs from older patients exhibit proliferative responses that correlate with clinical activity markers. In contrast, OFs from younger patients are predisposed to adipogenic differentiation, which is more closely associated with the development of proptosis (51). Furthermore, downstream signaling following TSHR activation involves cross talk with the insulin-like growth factor 1 receptor (IGF-1R), a pathway critically modulated by β-arrestin-1 and β-arrestin-2. While β-arrestin-2 primarily mediates TSHR desensitization, β-arrestin-1 enhances TSHR signaling activation (52). Notably, altered β-arrestin expression—particularly reduced β-arrestin-2 levels—has been implicated in aging-related conditions such as neurodegeneration and type 2 diabetes (53), suggesting that diminished β-arrestin-2 in older individuals may enhance the efficiency of TSHR-mediated inflammatory signaling.

Age-related changes in immune cell composition may also contribute to the observed effects. Age-associated B cells (ABCs), a unique subset that accumulates with aging, are characterized by a CD11c⁺T-bet⁺ phenotype and rely on TLR7/TLR9 signaling and Th1 cytokines, including IFN-γ and IL-21, for their differentiation (54, 55). These cells are enriched in autoimmune diseases, such as IgG4-related disease, where they promote inflammation through the production and secretion of autoantibodies and pro-inflammatory cytokines, including TNF-α and IFN-γ. In GO, ABCs may amplify TRAb-associated inflammation by promoting IgG4 synthesis—consistent with the elevated IgG4 levels observed in orbital tissues from patients with severe disease—and by secreting cytokines that activate OFs (56, 57). Additionally, ABCs can influence T-cell polarization toward a Th17 phenotype, thereby enhancing IL-17–mediated pro-inflammatory signaling and glycosaminoglycan production by fibroblasts (55, 58).

Regulatory T cells (Tregs), a key immunoregulatory CD4⁺ T-cell subset, also exhibit age-related functional decline, marked by reduced STAT3 activation and impaired suppression of IL-17–producing T cells. This diminished Treg-mediated control in older individuals permits the expansion of pathogenic T cells and exacerbates pro-inflammatory responses (59, 60). In GO, these dual alterations likely impair the regulation of TRAb-induced orbital inflammation, thereby amplifying tissue injury driven by TRAb signaling. Future studies should aim to elucidate the mechanisms underlying the cross talk between TRAbs and age-associated immune cell populations.

In addition, we evaluated the prognostic value of TBII and TSI for treatment outcomes following first-line therapy with IVMP combined with MMF in patients with moderate to severe active GO. Elevated TSI levels, increased SIR max on MRI, and reduced HDL concentrations emerged as significant independent predictors of poor treatment response, whereas TBII levels showed no significant prognostic utility. This finding diverges from prior studies assessing IVMP monotherapy, which reported predictive value for both TBII and TSI (20, 61). Several factors may account for this discrepancy. Most notably, our study evaluated a combination regimen of IVMP plus MMF, in contrast to IVMP monotherapy in earlier research. Additionally, our response assessment incorporated MRI parameters alongside CAS to adjudicate ambiguous cases, whereas some previous studies relied solely on CAS. Other methodological differences include a shorter median disease duration in our cohort (3 months), inclusion of a broader set of prognostic variables—such as imaging metrics and lipid profiles—in our multivariate model, and evaluation of sustained treatment response at 3 months post-IVMP (following MMF completion), rather than immediate post-treatment outcomes reported elsewhere.

Interestingly, recent studies have shown that HDL exerts immunomodulatory effects by suppressing the activation, proliferation, and cytokine production of CD4+ T lymphocytes (62), which constitute the most abundant lymphocyte population infiltrating orbital tissues in GO, exhibiting both chemotactic and pro-inflammatory characteristics (63). These findings highlight the importance of identifying biomarkers that reliably reflect disease severity, thereby facilitating more accurate outcome prediction.

Both TBII and TSI demonstrated only weak correlations with the total CAS, a result consistent with previous reports (46, 61). Several factors may account for these modest associations. First, GO disease activity is likely governed by a multifactorial process extending beyond TRAb titers, encompassing the local cytokine milieu, fibroblast responsiveness, and the well-established cross talk between the TSHR and IGF-1R signaling pathways (4, 64). Second, antibody concentrations measured in the peripheral circulation may not accurately reflect their bioactive levels within orbital tissues. Finally, in nearly half of all patients, hyperthyroidism precedes the onset of ocular symptoms (65), suggesting that peak antibody titers may occur before the onset of maximal clinical activity. Given the cross-sectional design of our study, these dynamic temporal dissociations may not have been adequately captured.

This study has several limitations. First, it employed a retrospective design, and TSI levels were not measured using bioassay methods, which may limit direct comparability with previous studies. Second, the disease duration was generally short among patients included in the analysis; as a result, its potential impact was not considered in the regression model. Finally, the sample size for the treatment response analysis was relatively small. Future multicenter, prospective studies with larger patient cohorts are warranted to improve the statistical power of longitudinal prognostic modeling.

In summary, in newly diagnosed GO patients, both TBII and TSI levels were associated with disease activity, with their diagnostic utility notably enhanced in older individuals. The bridge-based TSI assay demonstrated a higher positivity rate compared with TBII and emerged as an independent predictor of treatment outcome in patients receiving IVMP combined with MMF. However, no significant difference was observed between TSI and TBII in their ability to distinguish between active and inactive diseases.