What has changed in thyroid eye disease in the last five years (2020–2025)

Abstract

Background

Significant progress has been made in the management of thyroid eye disease (TED), based on the elucidation of important pathogenic mechanisms. This has led to novel therapeutics validated in randomized clinical trials. Autoreactive antigens that elicit specific orbital immune reactions have not yet been identified, although it has been shown that fibrocytes, circulating stem cells that differentiate into fibroblasts, are expressing the thyroid-stimulating hormone receptor (TSHR) and insulin-like growth factor-1 receptor (IGF-1R) and may be stimulated in the orbit by a cascade of inflammatory reactions inducing adipogenesis.

TED clinical assessment

Moderate–severe forms of TED are the target for immunosuppressive therapy. Improvement in the assessment of the active and progressive phase of disease is becoming compelling, as the outcome of a treatment depends on how early during the progressive phase the disease is treated. The clinical activity score may not always help define the right time for treating.

Therapy

In 2020, teprotumumab, an anti-IGF-1 receptor blocker, has been approved by FDA for the treatment of TED. Since then, other drugs were studied or are under investigation and will seek regulatory approval.

Microbiome and TED

A series of studies have investigated the role of microbiome in thyroid autoimmunity and TED more in detail, based on the observation that treatment with antibiotics may modify the disease phenotype in a murine model of TED.

Artificial intelligence

This approach is being studied for the assessment of TED, especially trying to standardize the use of orbital and facial images for improving the diagnosis of the disease in the early, progressive phase. In the future, these applications will allow the use of synthetic data, in addition to training on real patient images and data.

Introduction

Over the past few years, significant progress has been made in the management of thyroid eye disease (TED), based on the elucidation of important pathogenic mechanisms occurring in the setting of the disease and, even more importantly, in the advancement of therapeutics made available based on the evidence produced by relevant randomized clinical trials.

TED, also called Graves’ orbitopathy, Graves’ ophthalmopathy, and thyroid-associated ophthalmopathy, is the most frequent extra-thyroidal manifestation of Graves’ disease (GD) and is characterized by immune infiltration of the orbit, which leads to inflammation and subsequent expansion and fibrosis of orbital fat and extraocular muscles, resulting in a fairly heterogeneous clinical presentation, from mild eyelid retraction, proptosis, periorbital edema, restricted ocular motility, and significant facial disfigurement (1). In most patients, TED is well recognized as a disease with a self-limiting course, characterized by inflammation and a progressive phase with predominant inflammation of ocular soft and retro-orbital tissues followed by a burnt-out phase where irreversible scarring and eye muscle fibrosis prevail (2).

The disease may rarely progress to very severe forms characterized by corneal ulceration, or compression of optic nerve by the enlarged extraocular muscle within the orbit, which is called dysthyroid optic neuropathy (DON).

The underlying autoimmune thyroid disease is Graves’ hyperthyroidism in about 90% of patients, although TED may be associated with hypothyroidism and euthyroidism in 5–10% of cases. The disease occurs in 3 men and 16 women per 100,000 annually and in 30–50% of patients with Graves’ hyperthyroidism (3, 4). The reason why about one half of GD patients develops TED is still unknown and possibly relies on variable pathogenic mechanisms that still remain unclear (5).

It is plausible that the autoimmune process leading to TED begins with the infiltration of activated B and T cells in the orbital connective tissue and extraocular muscles, although their role remains uncertain, since autoreactive antigens that elicit specific orbital immune reactions have not yet been identified. Smith et al. have shown that fibrocytes, circulating stem cells that differentiate into fibroblasts, are expressing the thyroid-stimulating hormone receptor (TSHR) and insulin-like growth factor-1 receptor (IGF-1R) and, when migrated into the orbit, may be stimulated by a cascade of inflammatory reactions and activate adipogenesis (6, 7).

Immune cells express pro-inflammatory cytokines (e.g. IL-1β, IL-6, IL-8, IL-16, TNF-α, RANTES, and CD40 ligand) that promote activation of orbital fibroblasts. Co-activation of the TSHR and the IGF-1R in the orbit has been shown to form a signaling complex in the presence of beta-arrestin (8). This induces a cascade of inflammatory response with the release of cytokines, production of hyaluronan, and proliferation of the orbital fat tissues (9). Recently, it has also been proposed that the activation of the complex may also occur in the absence of beta-arrestin, perhaps resulting in the expression of a different clinical TED phenotype (10).

Among cytokines, IL-6-has been shown to play a key role in the pathophysiology of TED, especially in the course of the inflammatory phase of the disease (11, 12, 13). Circulating IL-6 and IL-6R have been shown to be elevated in patients with Graves’ disease and active TED (14, 15, 16), independently of the effect of thyroid dysfunction (17, 18). Furthermore, the level of expression of orbital tissue IL-6 (19, 20) and orbital tissue IL-6R (21) is increased in active TED compared to inactive TED. Transcripts of IL-6 mRNA are expressed in the extraocular muscles and orbital fat of patients with TED, and the expression is related to orbital volume (22). These findings suggest that the activation of the IL-6/IL-6R pathway may be associated with TED activity and progression, which could be antagonized by IL-6/IL-6R inhibitors.

Clinical assessment of TED

TED specialists ought to match treatments with the disease clinical manifestations and the needs of individual patients, by targeting the most dominant feature in the patient’s view and taking into account efficacy, safety, and cost-effectiveness of treatment options. The prediction of an optimal treatment outcome would spare patients the frustration and the potential risks of ineffective interventions. To this end, one essential step in making a therapeutic decision is the assessment TED of activity and severity.

Clinical activity score (CAS) for the assessment of activity

A CAS has been proposed in 1989 (23), and subsequently modified in 1997 (24), as a tool to score the grade of soft tissue involvement in TED and to predict response to immunosuppressive/anti-inflammatory treatment. It mainly grades inflammatory changes with the most commonly employed 7-point score (Fig. 1). The CAS has also been endorsed by clinical researchers and guidelines as a clinical outcome measure of the efficacy of therapy in most of the RCTs conducted until 2020, especially when disease inactivation was the primary outcome of the studies. Many of the unmet needs in the management of TED demand for a more accurate diagnosis of the phase of disease that is expected to better respond to medical therapy. The limitations of the CAS in assessing the early progressive phase of TED have become more evident: it does not score disease progression in the absence of inflammation and it relies in part on the subjectivity of the clinician in assessing clinical signs, especially orbital pain as referred by the patient (25). Only the use of the extended scale of the score to 10 points, which include changes in proptosis, diplopia, and vision at a second examination, may help diagnose active TED in the progressive phase of severity. Activity can, in fact, be inferred from changes in severity across time. This approach helps detect progression in those not so rare cases where motility can worsen even in the absence of inflammation and in ‘congestive’ TED, where inflammatory features are ambiguous, due to the presence of signs of impaired venous drainage (26). Unfortunately, such an approach should rely on optimal time interval between assessments, which is not predictable given the variable time for TED progression observed from patient to patient.

Figure 1.

Clinical activity score, as proposed by Mourits et al. (23) and subsequently modified by Mourits et al. (24) with additional three points of disease progression.

For the above limitations, the Food and Drug Administration (FDA) and international regulatory advisors have lately rejected the CAS as an objective and valuable outcome for the efficacy of a therapeutic when challenged in clinical trials of TED, preferring therefore the measurement of the reduction in proptosis values (27). Although this can be more easily quantitated, proptosis measurements, when assessed with a Hertel exophthalmometer, have shown significant inter-observer variability (25). If proptosis is to be established as the most reliable outcome measure in clinical studies, MRI or CT scan orbital measurements should be standardized (28, 29).

Assessment of severity

TED severity has been classified by the European Group on Graves’ Orbitopathy (EUGOGO) as mild, moderate–severe, or sight-threatening in a consensus paper in 2008 (30): this clinical grading has been widely accepted and, despite the quite heterogeneous manifestations comprised in the moderate–severe group of patients, it has provided an effective way to standardize the approach to treatment. The clinical characteristics that confer significant severity in TED patients are the degree of proptosis, which best correlates with the disfiguring features of the disease, and the altered eye muscle motility, which usually impacts the ability of patients to perform routine daily activities. These clinical signs worsen during the progressive phase of TED but rarely improve with disease inactivation. This is why treatment, to be effective, is expected to significantly impact these features and not just inflammation. In addition, proptosis can be measured and quantified and this has led to the FDA approval of study designs for TED RCTs with improvement of proptosis as the primary efficacy end point. In an effort to standardize and measure also the improvement of motility in TED, the diplopia score, proposed by Gorman (31), has been used in RCTs over the past few years. Unfortunately, this score has evident limitations, as it relies on the patient’s subjective report of diplopia, which may not be perceived, for instance, in very severely limited motility, when the eye movements are blocked due to marked inflammatory enlargement and subsequent scarring. A more objective method of assessing changes in eye motility may be the measurement of ductions in various directions of gaze, which can be quantified in degrees (32, 33). A composite outcome score, which includes improvement of the CAS of 2 points, proptosis of 2 mm, palpebral aperture of 2 mm, diplopia of at least 1 class, and the quality of life of at least 6 points, has been proposed by EUGOGO in 2012 (34). A few studies have included this parameter as an outcome (35, 36, 37, 38), but this has not been adopted in most of the ongoing RCTs. One reason for this is that since the approval of teprotumumab, regulatory agencies consider proptosis as a primary efficacy outcome (27) with diplopia and quality of life as secondary end points. A major limitation of the composite score is that it does not allow us to score the efficacy that drugs may have on different clinical features of TED based on the variable phenotypes of presentations (39). Improvement of proptosis or diplopia, but not inflammation, may be the therapeutic outcome expected by the patients, whose main concern has to be valued in a shared clinician/patient decision for treatment. The recent ETA/ATA consensus statement on TED management (40) has been developed along these lines of decision.

Targeted therapy

Progress in the knowledge of the efficacy of novel drugs in TED has changed the treatment paradigm. While steroids have been shown to mainly target inflammation, IGF-1R blockade with teprotumumab (TEP) is effective in reducing proptosis and improving diplopia (41, 42), IL-6 with tocilizumab (TCZ) targets improving both inflammation and proptosis (43, 44), and CD20 targeting with rituximab (RTX) has been associated with a reduction in disease relapse, in addition to inactivation early in the course of disease (35, 45, 46) (Fig. 2).

Figure 2.

Treatment plan as suggested by the ETA/ATA consensus statement on management of TED.

Targeted therapy offers now to the clinician the option of treating more specifically the dominant feature of the disease, also taking into account the clinical manifestation of TED that is most relevant for the patient’s disability or dysfunction. Patients express their disabling features when asked (47), and the use of the Graves’ orbitopathy quality of life (GO-QoL) questionnaire as an outcome of clinical studies has allowed us to match the patients’ expectation from treatment and the predicted effectiveness of the therapy that the clinician chooses (48).

In this context, the task force of the ETA/ATA consensus document on the management of TED has suggested that the choice of treatment, when possible, should not be made based on a first line–second line approach (39) but on the evidence of the efficacy of a specific drug on the dominant clinical feature of the disease in a determined patient (40). This approach to therapy will help in the near future, when considering the use of drugs that are now being studied in RCTs and in view of potential combinatory treatment for TED. Until now, no studies have investigated the therapeutic potential of combining drugs that, by targeting different disease mechanisms, may synergistically affect the heterogeneous features of TED.

Approval of first drug for TED

The FDA approval of the IGF-1R blocker teprotumumab as the first pharmacological treatment for TED in 2020 (26) has led to its use as first-line therapy in the USA. TEP has been shown in two randomized controlled trials (41, 42) to be effective, when compared to placebo, in improving inflammation of active TED of less than nine months in 59% of patients compared to 21% and, for the first time, in significantly reducing (>2 mm) proptosis and improving the grade of diplopia assessed with the Gorman score, in 83 and 70% of patients, respectively, with an overall response (CAS reduction and proptosis response) in 78% of patients. Compared to currently available therapies, this treatment has shown a consistent and significant effect on proptosis, rarely observed with other drugs or only in small subgroups of patients. TEP has also been approved in other countries, including Brazil, Saudi Arabia, and the UK. As a result of a recent RCT (OPTIC-J) (49) showing also improvement of proptosis in an Asian cohort of patients, TEP is now approved in Japan. The European Medicinal Agency has recommended granting a marketing authorization in the European Union for teprotumumab (Tepezza), for the treatment of adults with moderate to severe TED (50). To date, the use in Europe and the UK has been unfunded by the National Health Services’ pending negotiations of the cost and specific indications. In the absence of cost-effectiveness studies, the cost of a cycle of therapy with teprotumumab is extremely elevated, when compared to available, off-label drugs (40).

Efficacy has subsequently been proven in cohorts of patients with steroid resistance (51) and also in patients with DON due to compression of the optic nerve (52, 53). When employed in patients with inactive, long-standing TED, a lesser degree of improvement of proptosis has been observed, also with a limited effect on diplopia (54, 55).

An important drawback in the treatment of active TED with TEP is the observed elevated disease relapse rate (up to 40%) occurring after the completion of one cycle of therapy (usually lasting 24 weeks) (54, 56). Other data from retrospective and uncontrolled clinical studies have reported a rate of proptosis regression in up to 60% of patients (57).

Reasons for this are unclear. A possible explanation is that the 24-week cycle of treatment may not cover the entire duration of the course of active TED in some of the patients (Fig. 3). Alternatively, while there is evidence that blockade of IGF-1R affects adipogenesis downstream, it may have only a partial effect on the cytokine secretion of cells infiltrating the orbit, which occurs earlier in the active inflammatory phase of TED (1, 6, 7). Disease relapse has been shown in a good number of patients to respond to a second cycle of TEP (54), if available and tolerated. Treatment of relapsing patients with alternative drugs is a plausible option but has not been tested in clinical trials as of yet.

Figure 3.

This diagram illustrates how disease duration may affect the therapeutic outcome. If the course of therapy (vertical red arrow and dotted red line) does not extend throughout the entire phase of TED activity (red line), at the end of treatment, the disease may continue to progress (horizontal red arrow) and it may require additional therapy (vertical gray arrow and dotted gray line) until it evolves to the burnt-out phase (shaded area).

The safety profile of TEP is generally favorable, although several adverse events deserve attention, especially the increased rate of hyperglycemia occurring in about 30% of patients, especially those with pre-existing IGT (58). Worsening of inflammatory bowel disease and hearing impairment has also been reported in 10% of patients, in about half of them irreversible or only partially improving after therapy withdrawal (59). Assessment of auditory function before administering TEP is highly suggested, since a significant proportion of patients with Graves’ disease have been reported to suffer from asymptomatic hearing dysfunction (60).

Other IGF-1 receptor-blocking drugs

Based on the success of therapy with TEP in TED, several other anti-IGF-1R antibodies have been developed and are being studied in randomized trials (Table 1). Recently, a recombinant anti-IGF-1R antibody (Sycume), administered intravenously, has been studied in a RCT (RESTORE 1) (61) and approved (March 2025) in China as a first-line therapy for TED patients with active TED. Veligrotug (VDRN-001), a humanized anti-IGF-1 monoclonal antibody with a different half-life and administered i.v., has completed a phase 3 study: the results have been communicated by the study sponsor (Viridian) but not yet been published. Subcutaneous VDRN-002 and VDRN-003 are also being studied in phase 2 and 3 trials, and lonigutamab, an anti-IG-1R antibody binding to a different domain from the one of TEP, also administered subcutaneously is being developed in phase 1–2 studies. Linsitinib, a small-molecule IGF-1R antagonist administered orally, twice daily, is being studied in a phase 2 study, after showing efficacy also in a mouse model of TED, with anticipated efficacy on preclinical studies (62).

Table 1.

Targeted and immune suppressive drugs studied or being studied in randomized controlled trials in TED.

Therapeutic class/drug	Route of administration	RCT vs placebo	TED clinical phase	Results	Label
Anti-IGF-1 receptor
Teprotumumab	IV	One phase 2 and three phase 3 OPTIC and OPTIC X	Active (3 studies); inactive (1 study)	Published (38, 39, 46, 50, 51)	FDA-approved; Brazil, Emirates, Japan; and EMA (UK pending)
Sycume	IV	Phase 3	Active	Published (57)	China
Veligrotug	IV	Phase 3	Active (thrive study); inactive (thrive 2 study)	Not published	Not approved
VDRN 003	SC	Phase 3	Active (2 ongoing studies, reveal 1 and 2)	=	=
Linsitinib	Oral	Phase 2b	Active	Not published	=
Lonigutamab	SC	Phase 1–2	Active (ongoing)	=	=
Anti-IL-6 receptor
Tocilizumab	IV	Phase 2b	Active (1 study)	Published (41)	Not approved
Tocilizumab	IV	Phase 2 vs IVGC	Active (1 study)	Not published	=
Satralizumab	SC	Phase 3	Active (2 studies)	Not published	Not approved
FcRn inhibition
Batoclimab	SC	Phase 3	Active (2 studies, +1 extension study)	Proof of concept (73)	Not approved
Efgartigimod	SC	Phase 3	Stopped for futility (2 studies)	Not published	Not approved
Anti-IL-17A
Secukinumab	SC	Phase 2	Active	Published (79)	Not approved

IVGC, intravenous glucocorticoids; TED, thyroid eye disease.

IL-6 blockade

The activation of IL-6/IL-6R pathway that has been shown to be associated with TED progression can be antagonized by IL-6/IL-6R inhibitors, whose safety and efficacy have been shown in several autoimmune and inflammatory conditions (63).

Tocilizumab (TCZ)

TCZ is a humanized anti-IL-6R monoclonal antibody that binds specifically to both membrane-bound and soluble IL-6 receptor (sIL-6Rα), thereby antagonizing all IL-6-mediated signaling pathways (64, 65). It has been recommended as a second-line treatment option and for the inactivation of moderate-to-severe TED (39, 40), based on a phase 3, double-blind, multicenter study that included 32 patients resistant to a previous course of glucocorticoids conducted in Spain (44). TCZ, administered i.v. in 4 doses (8 mg/kg) every 4 weeks, induced inactivation of TED, measured as a reduction in baseline CAS ≥ 2 points at week 16 (the primary end point), in a greater proportion of patients when compared to placebo (93.3 vs 58.8%; P = 0.04). Proptosis reduction at week 16 was also significantly greater as compared with placebo, but less than 2 mm (median (interquartile range): −1.5 (−2.0 to 0.5) vs 0.0 (−1.0 to 0.5) mm; P = 0.01).

Unfortunately, the significant treatment benefits achieved at week 16 were not maintained through week 40, which may be explained, at least in part, by the variable times of TED duration from diagnosis. The dosing schedule used in this study was based on studies in rheumatoid arthritis and not TED. In non-randomized studies, TCZ therapy showed an inconsistent improvement in both CAS and proptosis (66, 67, 68). Of note, treatment with tocilizumab has also been associated with reductions in titers of TSH receptor antibodies (69). More recent studies have also shown similar efficacy when TCZ has been administered subcutaneously (70, 71).

TCZ is generally well tolerated in patients with TED. No severe adverse events are generally reported. The most frequently observed side effects are increases in hepatic enzymes, hypercholesterolemia, and transient neutropenia (68). While the Spanish randomized trial did not report on relapse rate after TCZ, all other studies have reported relapse rates of TED less than 10% (67, 68, 69). Recently, a phase 2b, single-blind, multicenter tocilizumab in active moderate–severe Graves’ orbitopathy (TOGO) trial (NCT04876534) has been completed in Italy. This study compared the safety and efficacy of intravenous TCZ with the standard of care (intravenous methylprednisolone) in 64 individuals with TED. Preliminary data, showing a significantly greater proportion of patients improving the CAS and proptosis with TCZ than steroids, have been reported at the International Thyroid Congress (June 2025), and the manuscript on the final results of the trial is in preparation.

Based on the data from the Spanish study, the European Group on Graves’ Orbitopathy (EUGOGO) 2021 guidelines (36) and the consensus document from the American and European Thyroid Associations (37) have recommended TCZ as a second-line treatment option to improve inflammation of active, moderate-to-severe TED.

Satralizumab (STZ)

STZ is a monoclonal antibody that antagonizes binding of IL-6 to both mIL-6Rα and sIL-6Rα, similarly to TCZ. It is approved for the treatment of neuromyelitis optica (72, 73).

Two phase 3, multicenter randomized controlled trials called SatraGO-1 and SatraGO-2 (NCT05987423 and NCT06106828, respectively) have been designed to test the safety and efficacy of STZ in patents with TED who are either treatment-naïve or previously treated with glucocorticosteroids. These trials include approximately 240 participants with active, moderate-to-severe TED or chronic, inactive TED who are receiving treatment with subcutaneous STZ or placebo every 2 weeks for the first three doses, followed by every 4 weeks for a total of 7 doses. The primary end point of both studies is the proportion of patients with a ≥2 mm reduction in proptosis of the study eye at week 24. Responders will then be re-randomized to either continue treatment with STZ or placebo until week 48 and followed up until week 72. Results are expected by 2026.

FcRn inhibition

The neonatal fragment crystallizable receptor (FcRn) is a unique cellular receptor with affinity for both IgG and albumin, even though completely unrelated proteins. FcRn is broadly expressed on both hematopoietic and non-hematopoietic cells and functions as a recycling mechanism to prevent degradation and extend the half-life of circulating IgG and albumin (74, 75). This molecule facilitates the destruction of IgG, reducing pathogenic IgG and IgG immune complexes, without expected effects on other Ig isotypes, and complement plasma cells, B cells, or other cells of the immune system. Several FcRn inhibitors selectively targeting IgG recycling are now being tested in studies, for their potential use in antibody-mediated autoimmune diseases (76).

In 2023, data from a preliminary study with batoclimab (IMVT-1401) in TED have shown reductions in IgG levels, with consequent reductions of circulating levels of pathogenic IgG anti-TSH-R-Ab (77). This was a multicenter, randomized, double-blind, placebo-controlled, phase 2b trial conducted at 19 sites in Canada, Europe, and the United States (NCT03938545) of three batoclimab dosing regimens (680, 340, and 255 mg) vs placebo administered weekly for 12 weeks. The primary end point was the proptosis response (decrease of at least 2 mm) at 12 weeks in addition to safety and tolerability in patients with moderate-to-severe, active TED. IMVT-1401 induced an early (2 weeks) and significant (30–60%) dose-related reduction in total IgG and both binding and stimulating antibodies to the TSH-R in all three groups compared to placebo, but no difference in proptosis responder rates with IMVT-1401 or placebo was observed. While no change was reported in the diplopia score, an improvement in the EUGOGO GO-QoL score was seen with the two higher-dose groups of batoclimab at 12 and 19 weeks.

There was a reduction in serum albumin in patients treated with the higher doses of batoclimab, but this was completely reversible at the end of treatment and no adverse events were reported in association with the hypoalbuminemic state. Post hoc analysis revealed that even the increase in low-density lipoprotein cholesterol (LDL-C) was reversible after drug discontinuation. Efgartigimod (EFG), another FcRn inhibitor, approved for the treatment of myasthenia gravis. The peculiarity of EFG is that while blocking FcRn for immunoglobulins does not affect the binding of albumin, a decrease in serum albumin is not expected. EFG has been the object of two RCTs in patients with TED (NCT06307626), which have recently been stopped due to a lack of drug effectiveness as per the interim analysis of the data.

Based on the currently available data, FcRn inhibitors do not seem to provide an effective therapeutic alternative for active TED. Since preliminary data with IMVT-1401 have shown that inhibition of the FcRn receptor brings about a decrease in total circulating IgG and in TSH-R antibodies, a recent proof-of-concept study in patients with Graves’ hyperthyroidism has been designed to assess the potential increase in disease remission rate after batoclimab discontinuation. A RCT (NCT05907668) is currently underway (78).

Very recently, data from the ORBIT study (NCT04737330), aimed to evaluate the efficacy and safety of secukinumab, an IL-17A inhibitor, in subjects with active, moderate-to-severe TED were published (79). The primary end point of the study was an overall response of a reduced CAS of ≥2 points and a reduction of ≥2 mm in proptosis from baseline at week 16. Secukinumab did not show clinical efficacy versus placebo when treating patients with active, moderate-to-severe TED.

The role of gut microbiome in TED

The gut, being the widest area of bacterial access in the body, constitutes the main reservoir of T cells in humans. The gut microbiota is characterized by a complexity of residing bacteria, viruses, protozoa, and fungi, and the microbiome is the genomes of all the microbiota. The neonates inherit the microbiome from maternal microbiota, and a stable gut microbiota takes years to establish in the adult, depending also on feeding habits, hygiene, stress, smoking, and the use of antibiotics. The gut microbiota is composed by different bacteria, grouped in Firmicutes and Bacteroidetes, and may be affected by the geographical position and the age of individuals, as well as their being healthy or sick. The microbiota composition may protect against pathogens, affect the metabolism of nutrients, break down carcinogens, and modulate the immune system. Theoretically, a healthy microbiome keeps the balance between Treg and TH17 cells, a dysregulation of which is implied by any kind of dysbiosis and may be an initiating mechanism for autoimmunity (80, 81).

A TED murine model immunized with TSH-R was developed, and the gut microbiota was analyzed in the INDIGO multicenter European study in 2018 (82). The characteristics of this model were hyperthyroidism, orbital changes, adipogenesis, and some signs of GO, and gut microbiota had an increased operational taxonomic unit of the phylum Firmicutes, but decreased Bacteroidetes when compared with controls. Bacteroidetes and Firmicutes were the most represented of the seven phyla identified. When mice were treated with a probiotic (lab 4), they showed an increased Firmicutes population and an increase in TSAbs, whereas when mice were treated with the antibiotic vancomycin, the result was an increase in Bacteroidetes with a decrease in TSAbs (83).

These data showed that the gut microbiome may have a role in the induction of GD in mice, suggesting also that treatments with appropriate antibiotics may improve outcomes for patients with autoimmune thyroid disease, including TED (84, 85).

In 2023, the INDIGO Consortium studied the microbiota in patients with GD and TED, compared to healthy controls (86). A gut microbiota perturbation similar to that observed in the murine model was observed, characterized by an increase in Actinobacteria and a decrease in Bacteroidetes, resulting in a Firmicutes:Bacteroidetes ratio higher in GD/TED patients than in controls. This might have also been due to hyperthyroidism itself, as serum FT4 concentrations were more elevated in GD than in TED patients. Bacteroides spp. were also found to be associated with smoking in TED patients. These data show that the gut microbiota may have a role in the increased incidence of AITD, which could be reduced by antibiotic treatment.

Studies in humans with AITD have shown substantial geographic variation: 95% of these studies are carried out in China, where the population has a different microbiota compared to that of Europeans. It is not surprising that they have reported totally different data, although all have shown reduced gut microbiota diversity (85).

Artificial intelligence in TED

Artificial intelligence (AI) refers to computer systems that can perform tasks that typically require human intelligence. AI works as a tool that may help us analyze patient data and recognize clinical patterns to support decisions on disease management, potentially improving patient care. Recently, large language models, or LLMs, have been preferred as AI systems because they have been trained on an enormous amount of data collected from the medical literature (87). They can process complex medical information and provide responses that are coherent and understandable. Some of these models, such as GPT-4, Claude, or Med-PaLM, are already being studied in analyzing laboratory data.

In a recent publication on the applicable methods to assess TED, the authors reviewed the different AI models employed over the last couple of decades that have tried to approach both diagnosis and management of TED (88). Deep learning models have been used by our group in 2000. The objective of the work was to seek predictors for a correct diagnosis of TED activity and progression using a multilayer perceptron model (backpropagation neural network) that was trained on a large series of more than 350 patients. The model prediction ability was concordant with the clinical assessment in 69–70% of patients (89).

Convolutional neural networks have recently been employed to improve and standardize reading of CT and MRI imaging of TED patients (90), and this has led to accuracy up to 87–89% when compared to expert professionals (radiologists). Similar models have been used to assess CAS and active TED analyzing facial images, with an accuracy rate between 78 and 98% when compared to diagnosis made by expert clinicians (91, 92, 93).

Very recently, a smartphone-based, machine learning-assisted system has been devised to assess the CAS from digital images of TED, examined by oculoplastic surgeons, reaching a prediction accuracy of active disease >90% and >82% when compared to in-person assessment by the same specialists (94).

Visual assessment is essential in diagnosing and monitoring TED. Large language model applications allow the use of synthetic data, in addition to training on real patient images and data. This might envisage future application of AI algorithms also in processing real images and clinical pictures.

Declaration of interest

The author is on the advisory board of ArgenX, Immunovant, Lycia, Amgen, Viridian, Merck, and Roche, receives speaker fee from Amgen, and receives research funding (to the institution) for clinical trials from Roche, Viridian, Sling Therapeutics, Immunovant, Amgen, ArgenX, and RSR. Dr Salvi is currently the President of the European Group on Graves’ orbitopathy (EUGOGO).

Funding

This work was funded by Fondazione IRCCS Cà Granda, Milan, Ricerca Corrente.