Randomized Controlled Trial of Rituximab in Patients With Graves' Orbitopathy

Abstract

Context:

Graves' orbitopathy (GO) is a potentially sight-threatening disease for which available medical therapy is not uniformly successful. Multiple case series suggest that rituximab (RTX) may be effective therapy for GO patients.

Objective:

To determine the efficacy of RTX in GO.

Design:

It is a prospective, randomized, double-masked, placebo-controlled trial.

Setting:

The study was conducted at a large academic private practice.

Patients:

Twenty five patients with active moderate to severe GO were enrolled, and 21 completed the study to the primary endpoint.

Interventions:

Two RTX infusions (1000 mg each) or two saline infusions were given 2 weeks apart.

Main Outcome Measures:

The primary endpoint was a reduction in clinical activity score (CAS) assessed as a continuum and separately as improvement by ≥2 points at 24 weeks. Secondary endpoints included success and failure rates, proportions showing clinically significant improvement in proptosis, lid fissure width, diplopia score, lagophthalmos and disease severity, and changes in those parameters, orbital fat/ muscle volume and quality-of-life.

Results:

The treatment groups were similar in all parameters at baseline. The last observation was carried forward if the patient discontinued prematurely. No differences were found in the proportions of patients showing CAS improvement at 24 weeks (25% placebo; 31% RTX, P = .75) or in CAS decrease from baseline to 24 or 52 weeks [mean 1.5 points (1.8 SD) placebo; 1.2 (2 SD) RTX at 24 weeks, P = .73]. Similarly, there were no differences between groups in any of the secondary endpoints at either 24 or 52 weeks. There were four adverse events (AE) in 3/12 placebo patients and 11 AE in 8/13 RTX-treated patients; 5/6 moderate or severe AE occurred in the RTX group.

Conclusion:

RTX offered no additional benefit over placebo to our patients with active and moderate to severe GO and carried with it non-negligible adverse effects.

There are no uniformly effective measures that can be taken to prevent Graves' orbitopathy (GO), and the available medical therapy is often unsuccessful. Patients with mild disease may benefit from selenium treatment (1), but other systemic therapies are generally not offered due to the potential adverse effects that accompany each treatment option. Therapeutic intervention is thus reserved for patients with very active and/or severe disease (2). Intravenous corticosteroids are the mainstay of therapy and are helpful for the inflammatory symptoms and signs of the disease in 50–70% of patients (3, 4). However, this treatment is associated with significant adverse effects, requires repeated infusions over many weeks' time, and 20–30% of patients relapse following treatment. Orbital decompression is a complex surgical procedure that is offered only when the disease is particularly severe, when other treatment modalities have failed, or in the context of inactive disease requiring rehabilitative surgery (2).

Given these limitations to available therapy for GO, novel treatments that directly target pathogenic mechanisms have long been sought (5). Rituximab (RTX) is a humanized chimeric anti-CD20 monoclonal antibody that depletes both B lymphocytes in the intermediate stages of maturation and short-lived plasma cells.

While a recent literature review (6) suggested that RTX might be of benefit in GO, improvement in disease activity over time is also compatible with the natural history of the disease (7). It was in this context that we designed the current double-masked, randomized controlled trial (RCT) to study the efficacy of RTX in GO.

Materials and Methods

Study design

Eligible patients met the following criteria: age 18–80 years with moderate-to-severe and active GO (2, 8) with clinical activity score (CAS) ≥4/7, euthyroid for at least 6 weeks (defined by normal free thyroid hormone levels), had evidence of disease progression (through changes in disease activity and/or severity) during the previous 2 months or lack of improvement in the prior 6 months (assessed through patient questioning and by review of outside medical records from the referring ophthalmologist/endocrinologist and/or patient photographs). Previous steroid treatment was acceptable if discontinued ≥ 4 weeks before enrollment. Exclusion criteria were as follows: evidence of dysthyroid optic neuropathy (DON) or impending DON, contraindications to therapy with RTX (abnormal chest X-ray, HIV, hepatitis B, hepatitis C, absolute neutrophil count < 1.5 × 10⁹/L, pregnancy), prior orbital radiotherapy or decompression surgery or recent improvement in the disease (assessed as described above).

Each eligible patient was given the choice of trial participation, treatment with IV glucocorticoids, orbital decompression surgery (if appropriate) or close observation and the potential for disease progression was discussed. In order to identify progression at an early stage, patients were re-evaluated every 2 months for the first 6 months and patients who showed deterioration were removed from the trial and the appropriate therapeutic options were offered.

RTX (1000 mg) was administered IV to the intervention group at baseline and 2 weeks later, preceded by acetaminophen and 100 mg of methylprednisolone 1 h before the infusion (standard premedication with RTX for prevention of infusion reactions). The placebo group received 1 L of normal saline, acetaminophen, and 100 cm³ of normal saline on the same schedule.

The primary endpoint was CAS reduction assessed as a continuum and separately as an improvement by ≥ 2 points. Secondary endpoints included a dichotomous success/failure rate (decrease in CAS ≥ 2 points and no need for additional therapy for the eye disease vs either CAS decrease of < 2 points or need for additional therapy for proven or suspected sight-threatening disease), and change between baseline and 24 or 52 weeks in lid fissure width, lagophthalmos, proptosis, orbital fat and/or muscle volume, subjective diplopia score using the Gorman scale (9), and quality of life (QOL) measured using the Physical and Mental Component Summary scores of the Medical Outcomes Study 12-Item Short Form Health Survey (SF-12) (10). This instrument was selected rather than GO-QOL (11) because there has been no validation study for the Dutch-generated instrument in a North American population. In addition, we determined the proportion of patients in each group showing clinically significant improvement in various predetermined disease parameters. Results of the last evaluation for patients who discontinued the trial prior to week 24 or 52 were carried forward to either 24 or 52 weeks as the final evaluation for that patient and were used for the outcome analysis.

The study was approved by the Institutional Review Board (IRB) of Mayo Clinic Rochester. Informed consent was obtained from each patient using a form approved by the IRB after full discussion of the inclusion-exclusion criteria, protocol, and potential adverse events. The study was funded by the National Institutes of Health (NIH).

Clinical assessment and follow-up

Baseline assessment included medical history, physical examination (endocrinologist), ocular evaluation (ophthalmologist), orbital CT, TSH, free T4, thyrotropin receptor antibody (TRAb; radioreceptor assay) levels and CD 19+ B cell count. The duration of the disease was defined after discussion with the patient regarding the onset of signs or symptoms that later led to the diagnosis. If the patient did not recall the onset, the date of the first visual abnormality recorded in the medical records was used. QOL was assessed using the SF-12 questionnaire, a shorter version of the SF-36v2 Health Survey using just 12 questions to measure functional health and well-being from the patient's point of view (12). Patients were re-evaluated at 8, 16, 24, and 52 weeks by an endocrinologist and ophthalmologist, both masked to the patient's study group. The same ophthalmologist and endocrinologist followed each patient throughout the study with the exception of 9/112 individual patient visits were cross-coverage was necessary due to scheduling conflicts. Two endocrinologists (R.S.B. and M.N.S.) were involved in the study, each following approximately half of the patients. While three ophthalmologists participated in the study, one of them (J.A.G.) followed 22/25 patients. Adverse events were recorded and classified as moderate if they required and resolved with therapy, and severe if they required therapy and the condition interfered with normal functioning. Ophthalmologic visits included measurement of proptosis [(using Krahn exophthalmometer (13)], lagophthalmos, lid fissure width, determination of the Gorman diplopia score, and NOSPECS classification. The biochemical tests, CD19+ B cell count, ophthalmologic parameters and QOL data were collected at each visit. Orbital CT was repeated at 52 weeks. ELISA testing of interleukin-6 (IL-6; reference range < 3 pg/mL), soluble IL-6 receptor (sIL-6R; < 60 ng/mL), and chemokines C-C motif ligand 5 (CCL5/RANTES; < 40 ng/mL) and C-X-C motif ligand 10 (CXCL10; < 150 pg/mL) were performed by ELISA (R&D Systems, catalog #D6050, DR600, DRN00B, and DIP100, respectively). TRAb measurements were performed using the TSH/TRAb modular E170 assay (Roche) with reference range < 1.75 IU/L.

Statistical considerations

The study was planned, monitored, and analyzed with the direct involvement of a specialist statistician (P.T.). Sample sizes were computed based on the expected drop of 2.9 and 1.5 points in the CAS score in RTX (14) and placebo group, respectively (15). A sample size of 15 per group will have 80% power to detect a mean difference in CAS of 1.4, assuming the common standard deviation is 1.27 using a two group t test with a 0.050 two-sided significance level, and 98% power to detect a difference in CAS mean values of 2.0. SAS procedure PROC PLAN was used to design a completely randomized trial. Each patient was assigned to one of two groups (Rituximab or Placebo) based on the output generated by this procedure, using blocks of six. Patients were stratified on smoking status (smoker defined as having smoked within the last year) and substratified according to CAS (4–5 or 6–7). The randomization plan was available only to the pharmacist attached to this protocol.

Data analysis

Continuous variables were compared between the two groups using t test or the Wilcoxon rank sum test, while categorical variables were analyzed using the χ²/Fisher exact test. Nonparametric methods were used throughout where normality and χ² assumptions did not hold. General linear models were used to access treatment effect on change in CAS between baseline and 24 weeks. A P-value < .05 was considered significant. All statistical analyses were performed using SAS (version 9.2) and JMP (version 10.1) software.

Results

Enrollment, retention, and trial completion

Over the duration of the study, a total of 636 patients (18–80 years of age) with GO were evaluated by endocrinologists and/or ophthalmologists at the Mayo Clinic. The vast majority were excluded, because they had mild and/or inactive disease. Fifty nine patients met the inclusion criteria, and 25 opted to participate in the trial; 12 were enrolled in the placebo arm, 13 in the RTX arm (Figure 1). Of these, 13/25 were referred by external physicians, 8 were self-referred, and 4 were evaluated only at Mayo Clinic. The primary end point at 24 weeks was met by 21/25 patients and 19/21 continued in the trial to 52 weeks (Figure 1). Our goal was to enroll 30 patients in total. However, enrollment was slower than expected, owing to a great extent to concerns expressed by patients regarding the potential side-effect profile of RTX. In addition, other patients opted out of the trial because they could readily obtain RTX (without chance of randomization) outside of the trial from practicing clinicians. Thus, study costs increased significantly which necessitated closing the trial before full enrollment was obtained. Given that the final sample size was smaller than anticipated, post hoc power calculations were performed and revealed 75% power for detecting a CAS difference of 1.4 and 96% for detecting a CAS difference of 2 points. Each of the enrolled patients received both infusions, except for one in the RTX group who did not receive the second infusion owing to a severe vasculitic reaction following the first infusion.

Figure 1.

Flow chart of patient participation in the trial from enrollment to primary endpoint (per protocol) at 24 weeks and trial completion at 52 weeks.

Demographics and clinical characteristics at baseline

The placebo and RTX groups were similar at baseline with respect to age, gender, and smoking status (Table 1). All smokers were current smokers while 10 of the nonsmokers were prior smokers (quit range 4–35 years prior to study enrollment). There were no differences between the placebo and RTX groups in prior corticosteroid treatment, TRAb level, CAS scores or any of the quantitative ocular parameters. Disease duration was also similar (estimated median 299 days, IQR: 253–595 for placebo; 373 days, IQR 240–1080 for RTX, P = .79) with two patients in the placebo and four in the RTX group with a duration of > 2 years. In 24/25 patients the disease was progressive at the start of the trial; one patient in the RTX group entered the trial with stable, but clearly active and moderately severe disease. All 10 patients (4 in RTX group and 6 in the placebo, P = .33) treated with corticosteroids had completed therapy at least 1.5 months prior to trial enrollment and all patients reported worsening of the disease after corticosteroids were discontinued. No patient had an elevated TSH level at baseline and no patient received RAI during the study. Euthyroidism was actively monitored at all study visits and maintained through medication adjustment as needed (only one patient in each group had a transiently elevated TSH before the primary outcome assessment). Individual patient values for the baseline clinical characteristics and all study parameters at baseline, 24 and 52 weeks are listed in Supplemental Table 1.

Table 1.

Baseline Data

Variable	Placebo (n = 12)	Rituximab (n = 13)	P
Age - mean (sd)	61.8 (11.0)	57.6 (12.7)	.31
Female gender - no. (%)	8 (66.7)	9 (69.2)	.89
Caucasian race - no. (%)	12 (100)	13 (100)
CAS - mean (sd)	5.3 (1.0)	4.9 (1.0)	.36
Smoking - no. (%)	2 (16.7)	2 (15.4)	.93
GO duration, days - median (IQR)	299 (253–595)	373 (240–1080)	.79
TRAb, IU/L - median (IQR)	19.5 (2.2–28.8)	20 (9–60)	.44
Dermopathy/Acropachy - no. (%)	1 (8.3%)	2 (15.4)	.59
Prior steroid therapy - no. (%)	6 (50)	4 (31)	.33
Progressive GO - no. (%)	12 (100)	12 (92.3%)	.33
Proptosis left (Krahn) mm - mean (sd)	23.0 (2.4)	24.2 (3.3)	.36
Proptosis right (Krahn) mm - mean (sd)	23.3 (3.8)	24.6 (3.0)	.32
Proptosis left (CT) mm - mean (sd)	17.3 (2.6)	18.2 (2.7)	.34
Proptosis right (CT) mm - mean (sd)	17.2 (3.3)	19.0 (2.6)	.24
Lagophthalmos left, mm - median (IQR)	0 (0;1)	1 (0;2.5)	.27
Lagophthalmos right, mm - median (IQR)	0 (0;0)	0 (0;1)	.12
Lid fissures left (mm) - mean (sd)	9.8 (2.0)	11.1 (2.8)	.30
Lid fissures right (mm) - mean (sd)	9.0 (2.7)	10.9 (1.5)	.06
Diplopia score - median (IQR)	2 (1–3.75)	2 (1–2.5)	.74
Total orbital volume right (mm3) - median (IQR)	22.5 (20–26)	22.8 (20.5–25.9)	.76
Total orbital volume left (mm3) - median (IQR)	21.9 (19.9–24.7)	23.2 (20.2–25.6)	.64
QOL SF-12 physical -median (IQR)	39.9 (33.5;48.8)	53.2 (42.1;56.7)	.14
QOL SF-12 mental-median (IQR)	46.1 (36;51.3)	44.3 (32.8;50.7)	.83
IL-6 (pg/ml) - median (IQR)	3.0 (0.9–9.7)	2.2 (1.9–3.4)	.60
IL-6R (ng/ml) - median (IQR)	23.9 (21.8–25.4)	24.7 (22.9–25.3)	.77
CCL-5 (ng/ml) - median (IQR)	3.5 (3.1–3.6)	3.5 (3.3–3.6)	.77
CXCL10 (pg/ml) - median (IQR)	340.6 (213–511.5)	124.6 (106.8–483.2)	.24

Change in CAS

CAS at 24 and 52 weeks had improved from baseline in most patients in both groups (Figure 2, A and B). At the primary end point (24 weeks) both placebo and RTX groups had a lower CAS score than baseline (Figure 2C) and the drop in CAS between baseline and 24 weeks was also similar (Table 2). By 24 weeks CAS had improved by ≥ 2 points in 3/12 placebo patients and 4/13 RTX-treated patients (P = .75; Figure 3A). By 52 weeks CAS had improved from baseline by ≥ 2 points in 6/10 placebo patients and 7/10 RTX-treated patients (P = .64; Figure 3C). We analyzed the CAS data to determine whether RTX-treatment might accelerate disease inactivation (CAS < 3). We found that at 24 weeks, 4/13 patients in the RTX group and 2/12 in the placebo had inactive disease (P = .41) and by 52 weeks inactive disease was present in 6/10 RTX-treated patients and 4/10 placebo patients (P = .37).

Figure 2.

Change in CAS and TRAb levels over the course of the study showing individual patient data for (A) CAS - placebo group; (B) CAS - rituximab-treated group; (C) mean CAS ± SD for each group; and (D) median TRAb levels for each group. Patients who discontinued the trial prior to week 52 were evaluated before discontinuation and those data were carried forward to either 24 weeks (for the 5 patients who discontinued prior to or at week 24) or 52 weeks (for the single patient discontinued from the trial after 24 weeks) as the final evaluation for that patient. There were no significant differences in mean CAS or TRAb levels between the study groups at any time point.

Table 2.

Changes in Disease Parameters

Variable	Placebo	Rituximab	P
CAS 24 weeks - mean (sd)	3.8 (1.4)	3.7 (1.9)	.93
CAS 24 weeks change - mean (sd)	1.5 (1.8)	1.2 (2)	.73
CAS 52 weeks - mean (sd)	2.4 (2.0)	2.0 (1.7)	.64
CAS 52 weeks change - mean (sd)	2.9 (2.3)	3.0 (1.6)	.91
Lid fissures R change 0–24 weeks (mm) - median (IQR)	−0.5 (-1;1.75)	0 (-1;1)	.98
Lid fissures L change 0–24 weeks (mm) - median (IQR)	0.5 (-1;1.75)	0 (-1.5;1)	.49
Lid fissures R change 0–52 weeks (mm) - median (IQR)	0 (-0.75;1.5)	0 (0;1)	.68
Lid fissures L change 0–52 weeks (mm) - median (IQR)	0 (-1;1.25)	−1 (-2;0.75)	.51
	−3.2 (8.9)	−5.3 (6.9)	.50
TRAb change 24 weeks (median, IQR)	−0.25 (-9.8;0)	−4.2 (-8.8;0)	.73
CT proptosis R change 0–52 weeks (mm) - mean (sd)	0.0 (1.8)	0.82 (1.4)	.97
CT proptosis L change 0–52 weeks (mm) - mean (sd)	0.0 (1.9)	0.1 (1.2)	.87
Lagophthalmos R change 0–24 weeks (mm) - median (IQR)	0 (-0.5 to 0)	0 (0 to 0)	.21
Lagophthalmos L change 0–24 weeks (mm) - median (IQR)	0 (0 to 0.5)	0 (0 to 1)	.92
Lagophthalmos R change 0–52 weeks (mm) - median (IQR)	0 (0 to 0)	0 (-0.3 to 0.6)	.52
Lagophthalmos L change 0–52 weeks (mm) - median (IQR)	0 (0 to 0.5)	0 (0 to 2.1)	.51
TOV R at 52 weeks (mm3) - mean (sd)	23.0 (4.9)	22.9 (2.9)	.99
TOV R change 0–52 weeks (mm3) - mean (sd)	0.0 (-0.6)	−0.1 (0.8)	.78
TOV L at 52 weeks (mm3) - mean (sd)	23.0 (4.8)	22.8 (3.0)	.90
TOV L change 0–52 weeks (mm3) - mean (sd)	−0.3 (1.2)	−0.4 (0.9)	.84
TMV R at 52 weeks (mm3) - mean (sd)	7.3 (0.8)	8.6 (2.0)	.07
TMV R change 0–52 weeks (mm3) - mean (sd)	0.4 (0.8)	0.9 (1.0)	.20
TMV L at 52 weeks (mm3) - mean (sd)	7.7 (0.8)	8.3 (2.4)	.46
TMV L change 0–52 weeks (mm3) - mean (sd)	0.2 (1.7)	0.4 (1.2)	.69
TFV R at 52 weeks (mm3) - mean (sd)	15.5 (4.8)	13.6 (3.0)	.30
TFV R change 0–52 weeks (mm3) - mean (sd)	−0.5 (0.9)	−0.5 (2.0)	.95
TFV L at 52 weeks (mm3) - mean (sd)	15.1 (4.8)	14.3 (2.4)	.62
TFV L change 0–52 weeks (mm3) - mean (sd)	−0.52 (1.0)	−0.7 (0.8)	.50
Diplopia score at 24 weeks - median (IQR)	2.5 (0;4)	3 (2;3.5)	.36
Diplopia score at 52 weeks - median (IQR)	1.5 (0.75;3)	2 (0.75;3.25)	.64
Change in diplopia score (0–24 weeks) - median (IQR)	0 (-0.75;0)	0 (-1.5;0)	.21
Change in diplopia score (0–52 weeks) - median (IQR)	−1 (-1;3)	−0.5 (-1;0.25)	.69
QOL SF-12 physical at 24 weeks - median (IQR)	40.3 (38.5;52.1)	45.9 (43.6;50.8)	.33
QOL SF-12 mental at 24 weeks - median (IQR)	46.1 (35.4;57.4)	52.8 (36.1;56.7)	.91
QOL SF-12 physical at 52 weeks - median (IQR)	46.7 (41.1;53.6)	51.8 (49.8;55.5)	.29
QOL SF-12 mental at 52 weeks - median (IQR)	49.4 (35.4;56.1)	55.1 (42.8;57.8)	.18

Abbreviations: TOV, total orbital volume; TMV, total muscle volume; TFV, total fat volume.

Figure 3.

Proportion of patients in each study group showing clinically significant improvement from baseline to 24 weeks (A and B) or baseline to 52 weeks (C and D). Improvement was defined as decrease in any of the following: (CAS) ≥ 2 points; proptosis ≥ 2 mm; diplopia score from 3 or 4 to 0, 1 or 2; lid fissure width ≥ 3 mm; and NOSPECS classification by 1 or 2 classes (A and C). Success and failure rates (B and D) were defined as a composite of decrease in CAS ≥ 2 points and no need for additional therapy (success) or either CAS decrease of < 2 points or need for additional therapy (failure). There were no significant differences between the study groups in any parameters at any time point.

Volumetric CT analyses

CT-measured muscle volume, fat volume and total orbital volume were not significantly changed from baseline in either group (Table 2 and Supplemental Table 1). Proptosis decreased minimally by a mean of < 1 mm in both groups over 52 weeks (Table 2) and had improved by ≥ 2 mm in 3/10 placebo and 2/10 RTX patients (P = .61; Figure 3C).

Quantitative clinical parameters and NOSPECS

Mean change in exophthalmometer-measured proptosis between baseline and 24 or 52 weeks was similar between groups (Table 2). Improvements by ≥ 2 mm were observed in 4/12 placebo and 2/13 RTX-treated patients at 24 weeks (P = .29; Figure 3A) and in 3/10 patients in each group at 52 weeks (Figure 3C). While there was no change in median diplopia score from baseline to 24 weeks in either group there was a slight worsening in that score from baseline to 52 weeks in both groups (Table 2). There were no differences between groups at 24 or 52 weeks in the proportion of patients with significant improvement in diplopia, lid fissure width, or lagophthalmos (Figure 3, A and C). At 24 weeks, 8/12 of placebo and 6/13 of RTX patients improved by ≥ 1 class in NOSPECS score while two patients in each group improved by ≥ 2 classes (Figure 3A). By 52 weeks, 7/10 in both groups improved by 1 class and 4/10 in both groups improved by 2 classes (Figure 3C). Deterioration by NOSPECS criteria was seen in 2 RTX patients by week 24 and in 3 placebo patients by week 52 (Supplemental Table 2).

Success/failure rates and QOL assessment

We found that 9/13 patients in the RTX group (69%) and 9/12 patients in the placebo group (75%) failed at 24 weeks with no difference between the groups (P = .75; Figure 3B). At 52 weeks, 48% of patients overall failed with equal distribution between groups (P = .85; Figure 3D). In a subgroup analysis excluding the two RTX patients who developed DON, we again found no difference in CAS change or failure rate between groups. We found no significant differences between groups in the QOL scores or their changes (Table 2). In both groups, the scores improved during the study.

Impact of prior corticosteroid treatment on disease course

We subdivided each of the treatment groups based on prior therapy with glucocorticoids and evaluated the CAS changes between baseline and 24 weeks. In the placebo group the CAS change was 1.7 ± 2.3 points in the 6/12 corticosteroid-treated patient and 1.3 ± 1.4 in the patients without prior corticosteroids (P = .76), while in the RTX group the CAS change was 1 ± 1.4 points in the 4/13 corticosteroid-treated group and 1.3 ± 2.3 in those without corticosteroid exposure (P = .76).

Immune parameters

TRAb levels decreased during the trial in both patient groups without any difference between the groups (Figure 2D). There were no differences between the study groups at any time point examined in IL6, sIL6R, CXCL10 or CCL5 levels. The CD19 count was significantly lower by week 8 in RTX-treated patients (14 cells/μL) compared with placebo (283 cells/μL, P < .001; normal range 31–409 cells/μL). At the nadir for each patient, as measured at the 8, 16 or 24 week visit, 12/13 RTX patients had < 5 cells/μL while CD19 B cells were undetectable in 8/13. The single patient who received only one infusion of RTX because of a vasculitic reaction, showed no change in CD19 count at 8 weeks compared with baseline (157 cells/μL each).

Multivariate analysis

Multivariate analysis of CAS change at 24 weeks, after adjusting for age, gender, smoking status, prior corticosteroid use, TRAb level, and duration of GO showed that women, irrespective of group assignment, had a better response than men, consistent with the known more severe and therapy-resistant course of GO in men (16). No other risk factor was found to be an independent predictor of outcome (Supplemental Table 3).

Dysthyroid optic neuropathy

Two patients, both in the RTX group, developed DON during the study. One patient developed DON 3 weeks into the trial and the other at 24 weeks. Both patients were therefore removed from the trial, underwent orbital decompression and both responded well with recovery of visual acuity and visual fields. One placebo-treated patient experienced severe pain, constant diplopia, increased periocular swelling and inability to close the right eye 4 months into the trial with visual changes suggestive of DON. She showed significant improvement in pain, swelling and color perception following IV glucocorticoids, subsequently undergoing bilateral orbital decompression for correction of excess proptosis. A fourth patient, who developed a vasculitis following her infusion of RTX (6 days after drug administration), experienced worsening ocular pain and diplopia over the subsequent 6 weeks and consequently underwent orbital decompression.

Adverse effects

Study patients experienced a total of 15 AEs (4 placebo and 11 RTX) occurring in 11 patients (3 placebo and 8 RTX; Table 3). Of those, six events were considered to be moderate or severe (5 in the RTX group).

Table 3.

Adverse Effects

Side-effect type	Placebo (N)	Rituximab (N)	Overall (N)
Myalgias/arthralgias	2	2	4
Skin (rash, itching)	0	2	2
Infectious (bronchitis, conjunctivitis)	1	1	2
Vasculitis	0	1	1
Optic neuropathy	0	2	2
Severe lacrimation	0	1	1
Gastrointestinal (tongue pain, abdominal pain, diarrhea)	1	1+1	2+1
Moderate/severea	1	5	6
Total events	4	11	15
Total patients	3	8	11

^aBold numbers indicate moderate/severe adverse effects.

Discussion

We initiated our RCT of RTX for the treatment of patients with GO based on several encouraging reports suggesting benefit from this agent (14, 17, 18). Our trial results, however, demonstrated no therapeutic benefit over the placebo in our population of patients with active and moderate to severe GO. In addition, there were no differences in the secondary assessments that included multiple quantitative measures of disease severity, the NOSPECS classification of severity, a composite assessment of treatment failure, and a QOL questionnaire. We found instead that CAS in most patients in both study groups improved over the course of the trial, while the individual markers of disease severity remained essentially unchanged except for diplopia that worsened in both groups. The NOSPECS classification, our composite success rate, and QOL improved over the course of the trial, likely reflecting the improvement in CAS. The decline in CAS in both groups at 24 weeks was similar to the improvement seen in disease activity (range 0.8–2.1 CAS points) over a similar time period in the placebo arms of RCTs of octreotide in patients with active GO (15, 19–21). Evaluation of our data at 52 weeks revealed further decline in CAS from baseline (2.9 points in placebo and 3.0 points in RTX). Overall improvement was present in roughly 50% of the cohort at 52 weeks. These data are consistent with the natural history of the disease (7, 22, 23).

The discrepancy between our clearly negative results and the promising findings in the reported case series (14, 17, 18) may be explained in part by sampling, measurement, and/or reporting bias in the latter. Case series are especially susceptible to these forms of bias (24), particularly when an instrument utilizing subjective features, such as the CAS, is used in open label studies. Further, while CAS improvement was reported in most of these studies, the overall 1 year failure rate is not negligible, as 5/41 (12%) of the similarly RTX-treated patients required orbital decompression within 1 year of therapy (14, 18, 25–30). Finally, simultaneous treatment with glucocorticoids, reported in some (28) or in all RTX-treated patients (26, 27) represents an important confounding factor.

An important limitation of our study is the small number of participants, additionally impacted by early withdrawals. As a result, it may have lacked the power to detect benefit. To assess the extent to which this might have impacted our results, we conducted a sensitivity analysis in which we added five theoretical patients to our study data; two added to the RTX group demonstrating the best results achieved in the RTX group (an improvement in CAS of 5 points/no failure) and three added to the control group showing the worst response achieved in that group (deterioration in CAS of 1 point/failure). The repeat analysis of these 30 patients failed to identify a benefit for RTX (improvement in CAS 1 ± 1.9 points in the placebo group; 1.7 ± 2.3 in RTX; P = .35). Despite this, it is clear that a larger study is necessary to clearly define the potential role of this therapy in GO. We also acknowledge the measure of disease activity that we used, CAS, may not fully describe the active phase of the disease in some patients. However it remains a parameter employed by most recent clinical trials in the field to assess clinical activity (3, 4, 31). In addition, owing to our practice patterns, we generally enrolled patients into the trial during their first visit to our clinic. As such, we assessed recent disease progression through careful questioning of the patient and by review of outside medical records from the referring ophthalmologist/endocrinologist. We acknowledge this as a less reliable gauge of progression than allowing a several week window between the initial visit and enrollment.

We encountered a relatively high rate of moderate or severe adverse events in our study (one in the control group and five in RTX-treated patients, two of which were DON), which is higher than reported in previous case series (14, 27, 28). However, serious adverse events associated with RTX therapy (ie, sudden death and DON) have been reported by others (26, 29).

It does not appear that the duration of disease differed between our patients (about 1 year) and those reported in most of the case series showing benefit. However, because we had only three patients in our study (two in the RTX group) having a disease duration of 6 months or less, our ability to assess this subgroup was limited.

In our study, prior corticosteroid treatment was used in both groups. It might be argued that the improvement experienced in the control group was influenced by continued benefit from prior corticosteroid therapy. In a recent study of IV corticosteroids in active moderate to severe GO, the beneficial response was noted by 6–12 weeks of therapy (4). Furthermore, no patient in that trial, or in an earlier trial by Kahaly (3), was described to have suffered deterioration after corticosteroid therapy that was subsequently followed by sustained improvement. In our study, no patient in either group received therapeutic corticosteroids within the 6 weeks prior to entering the study. In some cases outside medical records were not available to us to determine whether a patient received prophylactic glucocorticoids in combination with radioactive iodine (RAI) therapy. However we know that the therapy was reasonably remote as the time from RAI administration to trial enrollment was a median of 42 months (IQR 17–90) for placebo and 41 months (IQR 10–192) for RTX, with only one patient having received RAI < 6 months prior to enrollment (5 months). These data taken together argue against an impact of previous corticosteroid therapy on the observed improvement in CAS seen in both of our study groups.

The strengths of our RCT include that all the participating investigators were masked as to the patients' study group. In addition, both subjective (eg, QOL) and multiple objective response variables were measured and patients were followed for a sufficiently long period of time to allow observation of late responses. That we found no significant differences between study groups in any of these parameters as assessed by multiple masked observers, including patients, ophthalmologists, endocrinologists, and radiologists, supports the internal consistency of our data. Despite Mayo Clinic being a tertiary referral center, our GO patient population is similar to many other referral centers in that our patients are both self- and physician-referred. We followed our usual clinical practice of offering to each patient the appropriate treatment options or, if they met enrollment criteria, the additional option of participation in the trial. Finally, we followed a therapeutic protocol known to induce B cell depletion and shown to be beneficial for other autoimmune conditions (32, 33). This same protocol was also commonly used in the reported case series of RTX-treated GO patients.

In conclusion, we found that for our patients RTX treatment offered no additional benefit over placebo, while being associated with a non-negligible side effect profile. Given the limitation that our study included a relatively small number of patients and recognizing the known heterogeneity of GO, it is possible that subsets of GO patients, such as those with active disease of shorter duration or steroid-naïve patients, might benefit from this treatment. However, our data suggest that the use of RTX may not be justified in the population of GO patients having active and moderate to severe disease of approximately or greater than 1 year's duration. The potential role of RTX in the treatment of patients with GO can be ultimately defined only through larger multicenter RCTs that might include only patients with short disease duration or be stratified based on this variable.