We conducted a retrospective, single-centre, no-profit, clinical study aimed at assessing the long-term effects of sirolimus, used as second-line treatment, on the outcome of moderate-to-severe, active GO compared to ivGCs. The research design entailed the inclusion of all consecutive patients treated with either sirolimus or methylprednisolone over 34 consecutive months.

The study was performed in a Tertiary Referral Center, namely the University Hospital of Pisa. The study was approved by the local Ethic Committee (Comitato Etico Area Vasta Nord-Ovest, approval no. 21672_MARINO) and performed in accordance with the International Conference on Harmonisation Good Clinical Practice guidelines and the principles of the Declaration of Helsinki.

Inclusion criteria were: (1) men and women aged between 18 and 75 years; (2) moderate-to-severe, active GO, defined by the presence of a Clinical Activity Score (CAS) ≥ 3 on a 7-point scale associated with at least one of the following criteria in the most affected eye: (i) exophthalmos ≥ 2 mm compared with normal for gender and race; (ii) inconstant or constant diplopia; and (iii) lid retraction ≥ 2 mm; (3) written, signed informed consent, including compliance with requirements and restrictions listed in the consent form; and, only for the sirolimus group: (4) previous treatment with ivGCs, administered more than 24 weeks prior to starting sirolimus.

Exclusion criteria were: (1) optic neuropathy, as defined by the 2021 EUGOGO guidelines [4]; (2) treatment with GCs or other immunosuppressive medications, and/or selenium within the 24 weeks prior to the current treatment; (3) previous orbital radiotherapy; (4) previous orbital decompressive surgery; (5) mental illness preventing comprehensive informed consent.

Signed informed consent was obtained from all patients.

The primary objective of the study was the overall outcome of GO at 48 weeks, based on a composite evaluation, as recommended by the 2021 EUGOGO guidelines [4, 28]. Overall response was defined as the presence of at least two of the following criteria in the most affected eye, without worsening in any of the same measures in the other eye: (1) reduction ≥ 1 point in 5-scale CAS (spontaneous and gaze-evoked pain excluded); (2) reduction ≥ 2 mm in exophthalmos; (3) reduction ≥ 2 mm in eyelid aperture; (4) increase ≥ 8° in the sum of eye muscle ductions.

Secondary objectives were: (1) overall GO outcome at 24 weeks; at 24 and 48 weeks: (2) outcome of single eye features, with the addition of diplopia (disappearance or improvement, namely change from constant to inconstant, intermittent, or absent, or from inconstant to intermittent or absent); (3) outcome of quality of life (QoL) [increase ≥ 6% of GO-specific QoL questionnaire (GO-QoL) score]; (4) TSH-receptor antibodies (TRAbs); and (5) GO relapse at 48 weeks (prevalence of patients with a worsening of at least two of any eye features from 24 to 48 weeks).

Safety monitoring was performed. Adverse events were documented and graded according to National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE), version 5.0.

Patients in the sirolimus group were treated with a loading dose of sirolimus of 2 mg orally on the first day, given on fasting, approximately at 10 am, followed by 0.5 mg per day for 12 weeks. Patients in the methylprednisolone group received intravenous methylprednisolone according to the following, previously described [4], protocol: 500 mg weekly for 6 weeks, 250 mg weekly for the subsequent 6 weeks (cumulative dose: 4.5 g). Within the methylprednisolone group, all patients were given omeprazole 20 mg/daily across the treatment period to prevent gastrointestinal adverse events, whereas post-menopausal women were treated with bisphosphonates (i.e. alendronate 70 mg/week) and vitamin D (i.e. cholecalciferol 25.000 IU/month) to prevent bone loss.

All patients underwent an ophthalmological assessment at baseline, 24 and 48 weeks, which included the following procedures: (1) exophthalmometry (Hertel exophthalmometer); (2) measurement of eyelid aperture; (3) assessment of diplopia (Gorman score) [4]; (4) ocular ductions; (5) assessment of the corneal status; (6) examination of the fundi; (7) visual acuity (Snellen chart); and (8) evaluation of CAS [29]. Patients were evaluated by two ophthalmologists (M.N.M, C.P) at the same time at all visits, in order to minimize inter- and intra-observer variations.

The following blood tests were performed at baseline and every four weeks up to 24 weeks: free thyroxine (FT4), by chemiluminescence immunoassays (Vitros Immunodiagnostics, Raritan, NJ); thyroid stimulating hormone (TSH), by immunochemiluminometric assay (Immulite 2000, Siemens Healthcare, Gwynedd, UK). TRAbs were measured at baseline, 24 and 48 weeks by enzyme-linked immunoassay (ElisaRSR™ TRAb 3rd Generation, Cardiff, UK). The following blood tests were performed at baseline and every two weeks up to 24 weeks: blood count, sodium, potassium, phosphate, creatinine, aspartate aminotransferase, alanine aminotransferase, creatine phosphokinase, alkaline phosphatase, fasting blood glucose, total cholesterol, low-density lipoprotein-cholesterol, high-density lipoprotein-cholesterol, triglycerides, all by enzymatic-colorimetric assays (Roche, Mannheim, Germany), urinalysis by automated urine chemistry analyzer (UNIMAX, Menarini, Firenze, Italy) and automated urine sediment analyzer (sediMAX conTRUST PRO, Menarini, Firenze, Italy), and urine culture by calibrated loop method (Sidecar, Alifax, Polverara, Italy). Rapamycin was measured in the sirolimus group at 12 weeks, 2–3 days prior to the end of treatment, by chemiluminescent microparticle immunoassay (ARCHITECT Sirolimus, Abbott, Chicago, Illinois, USA).

A specific questionnaire for GO was used to assess QoL [30]. Questionnaire comprises two subscales: (1) visual functioning (eight questions concerning limitations attributable to decreased visual acuity, diplopia, or both), and (2) appearance (eight questions referring to limitations in psychosocial functioning attributable to changes in appearance). Questions are scored as severely limited (one point), a little limited (two points), or not limited at all (three points). The total score as well as the two subscales were converted into percentages according to the following formula: (total points*100)/(number of questions answered*3). A higher percentage means a better QoL. Patients filled the questionnaire at baseline, 24 and 48 weeks. Patients were defined responders when an increase by at least 6% from baseline occurred.

Data were collected and recorded in a database. The following database validation procedures were employed: allowed character checks, batch totals, missing records check, cardinality check, digits check, consistency check, control totals, cross-system consistency check, data type check, hash totals, limit check, logic check, presence check, range check, spelling and grammar check, and uniqueness check.

In a previous study, conducted in 15 patients per group, we observed 86.6% of responders in patients given sirolimus vs 26.6% in patients given methylprednisolone at 24 weeks [27]. Assuming that similar results would have been observed at 48 weeks, we estimated that a total of 18 patients would have been sufficient for statistical significance with P < 0.05 by Fisher’s exact test for primary outcome in the final analysis, with a statistical power of 0.8. Thus, the number of patients studied, namely 40, largely exceeds that estimate.

Continuous variables are presented as mean (SD) or median (IQR), as appropriate, and differences were assessed by ANOVA with Bonferroni’s correction or Mann–Whitney. Categorical data were compared by two-tailed Fisher’s exact test. The impact of the baseline features of the two groups on the study outcomes was assessed by binary logistic regression. Analyses were performed using SPSS version 21.0 (IBM, New York, NY).