Introducción

Thyroid eye disease (TED), also referred to as Graves’ orbitopathy or thyroid-associated ophthalmopathy, is the most frequent extrathyroidal manifestation of Graves’ disease (GD), affecting up to 40% of patients with GD over the course of their illness (1). TED is an autoimmune condition driven by shared antigenic targets in the thyroid and orbit, particularly the thyrotropin receptor (TSHR) and the insulin-like growth factor 1 receptor (IGF-1R), leading to orbital fibroblast activation, adipogenesis, and glycosaminoglycan deposition (2). This process results in the expansion of orbital tissues, causing a spectrum of ocular manifestations such as proptosis, eyelid retraction, diplopia, ocular surface exposure, and, in severe cases, optic neuropathy (3). Risk factors associated with TED include smoking, older age, high thyroid-stimulating immunoglobulin levels, and radioiodine therapy (1).

The clinical expression of TED varies in activity and severity, with approximately 5-10% of cases progressing to a moderate-to-severe form that significantly compromises quality of life and may threaten vision (1). Moderate-to-severe TED is typically defined by the presence of active inflammation (Clinical Activity Score ≥3) and sight-threatening features, including significant diplopia, restrictive myopathy, corneal exposure, or optic nerve compression. Management of these patients requires a tailored approach, with intravenous glucocorticoids recommended as the first-line therapy for active, moderate-to-severe disease (3). Nonetheless, a substantial proportion of patients exhibit a suboptimal response or intolerance to steroids, prompting the investigation and use of alternative treatments such as biologic therapies and targeted immunomodulators (2, 3).

In this context, decision-making in moderate-to-severe TED is complicated by variability in clinical presentation, heterogeneity in patient response, and limited comparative evidence across therapeutic options (4). In this setting, formal consensus methodologies offer a valuable approach to guide clinical practice when evidence alone is insufficient. The RAND/UCLA Appropriateness Method (RAM) is particularly suited for this purpose, as it combines a systematic literature review with expert clinical judgment to assess the appropriateness of interventions across a range of patient scenarios. Its structured, iterative process facilitates transparent recommendations while capturing the diversity of expert perspectives. Given the growing interest in optimizing individualized therapy in TED and the lack of universally accepted algorithms beyond guideline statements, RAM provides a rigorous yet flexible framework to support consensus-based clinical decision-making in this complex field (5, 6). The objective of this study was to apply the RAND/UCLA Appropriateness Method to evaluate and generate consensus-based recommendations on the medical management of patients with active, moderate-to-severe TED. While the study was conceived in the context of the Colombian health care system, with its unique challenges related to access, availability of biologics, and referral pathways, the methodological framework and findings may also be applicable and transferable to other health systems with comparable characteristics in Latin America and beyond.

Methods

Analysis

For each clinical scenario-intervention pair, individual panel ratings were aggregated to calculate the median and the interpercentile range (IPR), defined as the difference between the 70th and 30th percentiles of the distribution. To assess disagreement, the asymmetry index (AI) was calculated as the absolute difference between the midpoint of the IPR and the central point of the scale (5.0). The interpercentile range adjusted for symmetry (IPRAS) was then computed as:

IPRAS=2.35+(1.5 × AI)

Disagreement was deemed present if the IPR > IPRAS.

Appropriateness classifications followed RAND/ UCLA rules: appropriate: median 7-9, without disagreement; uncertain: median 4-6, or any median with disagreement; inappropriate: median 1-3, without disagreement.

In the third round, necessity ratings were analyzed for interventions classified as appropriate in the second round. The same analytical procedure (median, IPR, AI, IPRAS) was applied. A treatment was classified as necessary if it met all of the following criteria: median rating 7-9; no disagreement (IPR ≤ IPRAS); and fulfilled RAM necessity criteria, defined as follows: omitting the intervention would constitute improper care, a reasonable likelihood of benefit exists, and the expected benefit outweighs potential harm by a clinically significant margin.

All analyses were performed using R (version 4.3.1), with custom scripts implementing the RAND/UCLA statistical approach (5).

Results

Analysis by categories

Category 1 - Moderate TED without functional urgency or refractoriness

Across scenarios, IVGCs and teprotumumab were consistently rated necessary, reflecting strong alignment between trial-level evidence and panel consensus. Teprotumumab demonstrated superior efficacy across proptosis, CAS, and diplopia, whereas IVGCs retained a central role in inflammatory control. Tocilizumab, MMF, and periocular triamcinolone were considered appropriate in selected contexts, particularly when the inflammatory burden predominated without visual compromise. Purely cosmetic presentations lacked strong evidence for systemic escalation, though panel ratings suggested variability in expert perception of disease impact.

Category 2 - Moderate TED with refractoriness, comorbidities, or risk factors

In steroid-refractory or contraindicated patients, biologics predominated. Teprotumumab and tocilizumab were consistently rated necessary across scenarios, supported by randomized data in resistant disease. MMF and periocular triamcinolone were appropriate in selected contexts, particularly where systemic steroids posed safety concerns (e.g., uncontrolled diabetes). These findings underscore the importance of individualized therapy in moderate but complex presentations.

Category 3 - Severe TED with acute visual involvement or optic neuropathy

In vision-threatening disease, IVGCs remained foundational when not contraindicated. Teprotumumab and tocilizumab were rated necessary in multiple scenarios, particularly in refractory cases. MMF was considered appropriate but not necessary, reflecting its slower onset of action and more limited evidence in acute settings. Scenarios involving extreme emergencies (e.g., globe luxation with neuropathy) showed the highest disagreement, highlighting limited high-quality evidence in these rare but critical contexts.

Category 4 - Severe TED with major corneal complications

No systemic pharmacologic therapy was rated appropriate or necessary in corneal perforation. Management relies on urgent surgical and local interventions. Systemic immunotherapy may be considered after structural stabilization.

Consensus-based treatment algorithm

This algorithm integrates three sources: (i) formal case definitions and clinical scenarios used to elicit ratings, (ii) RAND/UCLA consensus classifications (appropriate vs. necessary), and(iii) the systematic review summarizing efficacy and safety across drug classes. Active thyroid eye disease (TED) was defined as CAS ≥3/7; moderate disease encompassed typical inflammatory signs and functional impact, whereas severe disease required the presence of optic neuropathy or major corneal complications (e.g., perforation). Refractoriness was operationalized as no or partial response after a standardized 12-week course of intravenous glucocorticoids (IVGCs), and absolute contraindications to systemic glucocorticoids were specified a priori. These criteria, together with scenario-level modifiers (e.g., smoking, uncontrolled diabetes, cosmetic burden, corneal exposure), were used to structure the decision nodes.

Among moderate cases without refractoriness or critical comorbidities, IVGCs and teprotumumab were consistently rated necessary, aligning with the evidence base supporting IVGCs as first-line therapy and the superior efficacy of teprotumumab across proptosis, diplopia, and CAS outcomes. Tocilizumab, mycophenolate mofetil (MMF), and periocular triamcinolone were considered appropriate in selected contexts, reflecting their safety-efficacy balance in non-urgent presentations (figure 1).

Figure 1. : Consensus-based treatment algorithm for moderate active thyroid eye disease. Note. This figure presents the detailed consensus-derived treatment algorithm for moderate active thyroid eye disease without optic neuropathy or corneal perforation. Therapeutic prioritization is based on RAND/UCLA appropriateness and necessity ratings across clinical scenarios, incorporating refractoriness to IVGC, contraindications to systemic glucocorticoids, inflammatory activity, and risk modifiers. Green boxes denote treatments classified as necessary (first-line). Blue boxes indicate appropriate interventions in selected cases. Dashed connectors represent escalation pathways following inadequate response or intolerance.

Source: Authors.

In moderate disease with refractoriness, contraindications, or risk-modifying comorbidities, biologics predominated. Teprotumumab and tocilizumab were rated necessary across scenarios, supported by randomized evidence in steroid-refractory contexts, while MMF and periocular triamcinolone were appropriate in selected profiles (e.g., uncontrolled diabetes favoring local steroid delivery and corneal exposure allowing broader therapeutic flexibility, including the use of botulinum toxin when deemed necessary).

For severe disease with acute optic neuropathy, IVGCs (when not contraindicated), teprotumumab, and tocilizumab were consistently rated necessary, in line with the standard-of-care role of IVGCs and trial-level data supporting biologics in refractory cases. MMF was deemed appropriate but not necessary, reflecting weaker evidence and a slower onset. Extreme emergency scenarios, such as globe luxation with acute neuropathy, showed the most extraordinary polarization, with uncertain-by-disagreement ratings recorded for MMF and teprotumumab in the global uncertainty set, reflecting the paucity of robust data in extreme emergencies (figure 2).

Figure 2. : Consensus-based treatment algorithm for severe active thyroid eye disease Note. Thisfigure outlines the consensus-based management algorithm for severe active thyroid eye disease, defined by the presence of optic neuropathy or major corneal complications. Immediate interventions for vision-threatening disease are highlighted, followed by escalation strategies for refractory or contraindicated cases. Red boxes denote emergency or immediate management steps. Green boxes indicate therapies rated as necessary. Blue boxes represent appropriate treatments in selected scenarios. Gray boxes identify interventions not considered primary therapy in the acute setting.

Finally, in cases of severe corneal perforation, no systemic pharmacologic intervention was rated appropriate or necessary, indicating surgical or local management as the preferred approach. After the perforation is controlled, prompt systemic therapy should be initiated.

To facilitate bedside clinical application of these consensus recommendations, a simplified clinical decision pathway is presented in figure 3. This practical algorithm integrates disease severity, refractoriness, and urgency into a streamlined framework intended for rapid clinical use, while detailed scenario-based consensus algorithms are provided in figures 1 and 2.

Figure 3.: Practical clinical decision pathway for active moderate-to-severe thyroid eye disease. Note. This simplified clinical algorithm summarizes consensus-based therapeutic priorities for active thyroid eye disease (TED) (Clinical Activity Score ≥3). The pathway integrates disease severity (moderate vs. severe), refractoriness to intravenous glucocorticoids (IVGC), and urgency of visual compromise into a streamlined framework intended for rapid bedside application. Green boxes indicate therapies rated as necessary or first-line according to the RAND/UCLA consensus process. Blue boxes represent appropriate treatments in selected clinical contexts. Red boxes denote emergency or immediate interventions. Gray boxes indicate therapies that are not considered first-line treatments in the specified scenario. For detailedscenario-level consensus algorithms and full therapeutic stratification, see figures 1 and 2.

Source: Authors

Discussion

Our consensus study using the RAND/UCLA Appropriateness Method provides a structured appraisal of medical therapies for active, moderate-to-severe TED in the context of the Colombian health care system. The systematic review confirmed intravenous glucocorticoids (IVGCs) as the cornerstone of treatment while also highlighting the selective role of biologics such as teprotumumab and tocilizumab, and the consistent benefit of mycophenolate as an adjunct or alternative immunosuppressant. Panel deliberations further refined these findings, distinguishing between scenarios where therapies were deemed appropriate, uncertain, or inappropriate, thereby offering practical guidance for clinical decision-making across heterogeneous clinical presentations. Notably, ratings were based exclusively on clinical efficacy and safety, as panelists were explicitly instructed not to consider costs or availability.

When compared with both the EUGOGO 2021 guidelines (4) and the ATA/ETA 2022 consensus statement (8), our findings align in reaffirming intravenous glucocorticoids as the cornerstone of therapy for active moderate-to-severe TED and in supporting the selective use of biologics such as teprotumumab and tocilizumab. A key point of divergence relates to mycophenolate mofetil (MMF). EUGOGO recommends MMF, in combination with IVGCs, as a preferred first-line regimen, while ATA/ ETA considers it an acceptable therapy, particularly in combination with glucocorticoids. In contrast, our panel did not classify MMF as necessary. This more cautious stance reflected two main concerns: the retraction of pivotal publications, which undermined confidence in the robustness of the evidence (8), and the perception among several panelists that the therapeutic effect of MMF appeared less consistent when compared with the more consistent benefits demonstrated with teprotumumab or IVGC monotherapy. This divergence illustrates how evolving evidence and expert interpretation can shift the weighting of the same therapeutic options, even within consensus frameworks.

Another relevant divergence concerns the use of rituximab. Both the EUGOGO 2021 guidelines and the ATA/ETA 2022 consensus statement acknowledge rituximab as an acceptable option in selected cases of active, moderate-to-severe TED, particularly in patients resistant to glucocorticoids. Our panel, however, expressed greater reservations. While recognizing that some randomized trials and observational studies suggest benefit, the group discussed at length the potential risk of dysthyroid optic neuropathy (DON) associated with rituximab therapy (9, 10), a complication reported in clinical practice and highlighted by panelists as a serious safety concern. This perception, coupled with the inconsistency of trial results, led to a more conservative classification of rituximab in our consensus, underscoring the importance of weighing not only efficacy signals but also the potential for rare but sight-threatening adverse events in therapeutic decision-making.

The emergence of biologic therapies is reshaping the therapeutic landscape of TED, with particular attention to tocilizumab and teprotumumab. Tocilizumab has shown consistent benefits in patients with glucocorticoid-resistant disease, including a randomized trial demonstrating significant reductions in CAS (11). Yet, it remains without formal regulatory approval for TED in Colombia. Teprotumumab, on the other hand, has been approved in several countries, including Colombia, as the first targeted therapy specifically indicated for TED and has demonstrated robust efficacy across all clinical domains (12-16). Although not widely available, its broader incorporation into clinical practice would represent a paradigm shift, especially for patients with rapidly progressive disease or an inadequate response to glucocorticoids. Our consensus classified both agents as appropriate in defined scenarios, underscoring their potential role in advancing individualized care and highlighting the need for regulatory pathways and access strategies to ensure their eventual integration into clinical practice.

Beyond systemic interventions, our consensus also highlighted the value of local therapies, which remain underrepresented in international guidelines. Both periocular triamcinolone and botulinum toxin were classified as appropriate in several scenarios and even necessary in some contexts, underscoring their perceived utility as adjuncts. Evidence from the systematic review further supports these findings. A recent meta-analysis of 30 studies (n=299 patients treated for eyelid retraction; n=205 for TED-associated strabismus) reported an 84% pooled success rate of BTX (Botulinum toxin type A) for eyelid retraction, with a mean MRD (margin reflex distance) reduction of 2.4 mm and mainly transient adverse events such as ptosis (13%) and diplopia (2%)(17). In contrast, outcomes in strabismus were less consistent, with only 24% achieving resolution of diplopia and the majority eventually requiring surgery.

For periocular triamcinolone, the prospective study by Bordaberry et al. (18) evaluated repeated peribulbar injections in patients with active, moderate-to-severe TED, including those with optic neuropathy. Significant reductions in CAS and improvement in visual function were observed, with two-thirds of patients experiencing complete resolution of optic neuropathy and minimal adverse events. Nevertheless, the non-randomized design, lack of controls, and potential for selection and detection bias limit confidence in these results. Despite these limitations, both the systematic review and the consensus panel converged on the view that periocular triamcinolone represents a valuable local option, particularly when systemic steroids are contraindicated or insufficient.

Collectively, these findings position local therapies as pragmatic, well-tolerated adjuncts for selected patients with TED. Their inclusion among interventions rated appropriate or necessary by our panel emphasizes the importance of broadening treatment algorithms beyond systemic agents and highlights the need for controlled studies to define their optimal role.

A key strength of this study is the use of the RAND/UCLA Appropriateness Method, which integrates the best available evidence with structured expert judgment (6, 19). This approach goes beyond narrative expert consensus by applying iterative rating rounds, moderated discussion, and quantitative measures of agreement and disagreement (IPR, IPRAS, AI). As such, it provides a transparent and reproducible framework that captures both convergence and divergence of expert opinion, allowing for nuanced recommendations that reflect the complexity of TED management.

Nonetheless, certain limitations should be acknowledged. The panel size was relatively small, and although composed of experts in endocrinology and ophthalmology, broader representation might have yielded additional perspectives. The study was conceived in the Colombian health care setting, which may affect generalizability; however, the methodology ensures relevance to similar contexts. Notably, costs and availability were intentionally excluded from consideration, consistent with RAND/UCLA principles, which may limit direct applicability to policy-making. Finally, for several therapies, especially local interventions and some immunomodulators, the evidence base remains limited and heterogeneous, thereby constraining the strength of the recommendations (20).

These findings have important implications for both clinical practice and research. They provide structured, scenario-based recommendations that can inform local and regional guidelines in Latin America and other comparable health systems. They also emphasize the role of emerging biologics, the more cautious appraisal of older immunosuppressants, and the potential utility of local therapies such as periocular triamcinolone and botulinum toxin.

Future research should prioritize randomized controlled trials of these interventions, as well as long-term comparative effectiveness studies and consensus updates that incorporate evolving therapies and regulatory approvals, ensuring that therapeutic algorithms for TED remain both evidence-based and adaptable.

Conclusions

This RAND/UCLA consensus study provides a structured evaluation of medical therapies for active, moderate-to-severe thyroid eye disease. The panel confirmed intravenous glucocorticoids as the necessary first-line therapy while highlighting the emerging role of biologics such as teprotumumab and tocilizumab and the selective use of local therapies including periocular triamcinolone and botulinum toxin. Mycophenolate was judged appropriate but not necessary, reflecting concerns about the reliability of its evidence base. Rituximab was appraised conservatively due to inconsistent trial results and safety concerns, particularly the risk of dysthyroid optic neuropathy.

By integrating systematic evidence with structured expert judgment, this study offers scenario-based recommendations that complement and refine international guidelines. These findings may serve as a reference for clinicians in Colombia and other comparable health systems, while also emphasizing the need for ongoing research to clarify the role of immunomodulators, biologics, and local therapies in individualized treatment algorithms.

Authors’ contributions

Carlos E. Builes-Montaño: Conceptualization, methodology, project administration, and supervision; María G. Mejía-López: Methodology, project administration; Alejandro Román-González: Methodology, project administration, final approval; María Del S. Cabarcas-Solano: Writing, review, and editing, final approval; Marta L. Muñoz-Cardona: Writing, review, and editing, final approval; Natalia Aristizábal-Henao: Writing, review, and editing, final approval; Katherine Restrepo-Erazo: Writing, review, and editing, final approval; Jennifer Camargo González: Writing, review, and editing, final approval; Hernando Vargas-Uricoechea: Writing, review, and editing, final approval; Henry M. Arenas-Quintero: Writing, review, and editing, final approval; Alejandro A. Castellanos-Pinedo: Writing, review, and editing, final approval.

Ethical implications

This work was a systematic review and consensus statement; therefore, no specific ethical considerations were identified. The study did not directly involve patients, human participants, or animals and was based exclusively on a review of the scientific literature and contributions from multidisciplinary experts.

Funding

Festina Lente support was funded by the Asociación Colombiana de Endocrinología, Diabetes y Metabolismo. This work was funded by the Asociación Colombiana de Endocrinología, Diabetes y Metabolismo, through an unrestricted educational grant from Amgen.

Conflicts of interest

The authors declare the following conflicts of interest: Natalia Aristizábal-Henao has received academic consulting fees from Amgen. Katherine Restrepo-Erazo has participated in conferences and advisory boards for AstraZeneca, Boehringer Ingelheim, Novo Nordisk, Merck, Sanofi, Recordati, Eli Lilly, Bayer, and Abbott. Hernando Vargas-Uricoechea has received speaking honoraria from Sanofi, Abbott, and Boehringer Ingelheim. Carlos E. Builes-Montaño has received consulting and speaking fees from Sanofi, Novo Nordisk, Novartis, Recordati Rare Diseases, Valentech, Janssen, Abbott, and Boehringer Ingelheim, and is a shareholder of Festina Lente. Marta Lucía Muñoz-Cardona has received travel support from Amgen. María del Socorro Cabarcas-Solano, Maria G. Mejía-López, Alejandro Román-González, Jennifer Camargo, and Alejandro Castellanos-Pinedo report no conflicts of interest.

Data statement

No data was collected in the development of this manuscript.

Use of artificial intelligence (AI)

The authors acknowledge the use of Grammarly, an artificial intelligence-based tool, exclusively for language editing purposes. This tool was not used for generating ideas, analyzing data, or creating scientific content. All intellectual contributions, including study design, data interpretation, conclusions, and final approval of the manuscript, remain the sole responsibility of the authors.

Acknowledgements

Festina Lente provided methodological support and was responsible for medical writing and editorial assistance. The authors assume full responsibility for the content and conclusions expressed in this manuscript.