Outcomes in Autoimmune Retinopathy Patients Treated With Rituximab

SAMANEH DAVOUDI, NAZANIN EBRAHIMIADIB, CAGLA YASA, DAMLA D. SEVGI, RAMAK ROOHIPOOR, EVANGELIA PAPAVASILIEOU, JASON COMANDER, AND LUCIA SOBRIN

PURPOSE: To evaluate clinical and ancillary testing, UTOIMMUNE RETINOPATHY (AIR) RESULTS FROM A including adaptive optics, outcomes in autoimmune presumed immunologic attack on the by retinopathy (AIR) patients treated with rituximab. against retinal antigens. The spectrum A DESIGN: Retrospective, interventional case series. of AIR includes nonparaneoplastic AIR (npAIR), METHODS: PATIENTS: Sixteen AIR patients treated with -associated retinopathy (CAR), melanoma- rituximab. OBSERVATION PROCEDURES: All patients were associated retinopathy (MAR), and autoimmune-related treated with a loading and maintenance dose schedule of retinopathy and optic neuropathy (ARRON). There is intravenous rituximab. Visual acuity (VA), electroreti- no established treatment protocol for AIR, and the nography (ERG), and spectral-domain optical coherence evidence base for therapeutic intervention includes only tomography (SDOCT) and visual field (VF) results a few retrospective case series and case reports.1 The largest were recorded. A subset of patients was also imaged using retrospective series assessed therapeutic effects of immuno- adaptive optics scanning laser ophthalmoscopy suppressive medications including cyclosporine, predni- (AO-SLO). MAIN OUTCOME MEASURES: Rates of VA sone, infliximab, azathioprine, and mycophenolate in 30 change before vs after rituximab initiation were compared consecutive patients with AIR.2 They found potential ben- with mixed-model linear regression. efits from using immunosuppression to treat AIR, but this RESULTS: The rate of visual decline was significantly series did not include any patients treated with rituximab. less after rituximab initiation compared with the rate of Given the proposed -mediated nature of AIR, a visual decline prior to rituximab initiation (P [ .005). therapeutic agent that targets B cells and thus reduces sys- Seventy-seven percent of eyes had stable or improved temic antibody levels would seem reasonable. Rituximab is VA 6 months after rituximab initiation. Amplitudes a monoclonal antibody that targets CD20þ Bcells.This and implicit times on ERG, mean deviation on VF, central medication has Food and Drug Administration approval subfield mean thickness, and total macular volume did not for treatment of several rheumatologic and hematologic dis- decrease to a significant degree over the rituximab eases.3 There are 6 individual case reports that have treatment period. Six eyes had serial AO-SLO imaging. described the treatment of AIR with rituximab, with varying Cone densities did not change significantly over the levels of success.3–8 These 6 patients include 1 patient with treatment period. npAIR in the setting of systemic lupus erythematosus and 5 3–8 CONCLUSION: VA was stable or improved in a majority patients with CAR. The primary aim of the current study of AIR patients while they were being treated with ritux- is to report outcomes for the treatment of AIR with imab. OCT and ERG parameters, as well as AO-SLO rituximab in a larger, consecutive series of patients. cone densities, were stable during treatment. Studies A secondary aim of the study is to report on the use of with additional patients and longer follow-up periods adaptive optics scanning laser ophthalmoscopy are needed to further explore the utility of rituximab in (AO-SLO) to monitor AIR. Patients with AIR often do the management of AIR. (Am J Ophthalmol not have visible retinal pathology that can be monitored 2017;180:124–132. Ó 2017 Elsevier Inc. All rights to judge progression and response to treatment. Therefore, reserved.) ancillary testing including (ERG), visual field (VF), and spectral-domain optical coherence tomography (SDOCT) are typically used to monitor AIR patients.9 However, some patients present with advanced Supplemental Material available at AJO.com. disease and sparing of function in only a small area of See Accompanying Editorial Article on page xv. retina centrally. These patients have severely reduced Accepted for publication Apr 26, 2017. ERGs and VFs at baseline as well as OCTs with only a From the Department of Ophthalmology, Harvard Medical School, Massachusetts Eye and Ear Infirmary, Boston, Massachusetts (S.D., N.E., small amount of intact ellipsoid zone. It is difficult to judge C.Y., D.D.S., R.R., E.P., J.C., L.S.); and Department of Ophthalmology, progression in these cases. AO-SLO is a new technique Farabi Eye Hospital, Eye Research Center, Tehran University of that allows for quantitation of the central photoreceptors Medical Sciences, Tehran, Iran (N.E., R.R.). 10 Inquiries to Lucia Sobrin, Massachusetts Eye and Ear Infirmary, 243 that remain in these patients, and thus has the potential Charles St, Boston, MA 02114; e-mail: [email protected] to be a useful tool to monitor patients with AIR. There is

124 © 2017 ELSEVIER INC.ALL RIGHTS RESERVED. 0002-9394/$36.00 http://dx.doi.org/10.1016/j.ajo.2017.04.019 1 case report of a patient with MAR in whom adaptive function, hepatitis B and C, tuberculosis, and complete optics imaging was used to detect cone density reduction blood counts before starting rituximab. before any detectable structural changes on SDOCT at Two different loading regimens of rituximab were used. the time of initial diagnosis,11 but this is the first study From 2010 to 2014, all patients were treated with a loading in which AO-SLO has been used to monitor AIR patients regimen of 1000 mg every other week for 2 doses, which is longitudinally. the most common dose used for rheumatologic conditions. More recently, we used a loading dose of 375 mg/m2 (eg, 675 mg for a 1.8 m2 body surface area) weekly for 4 weeks. This dose is used more commonly for hematologic malignancies, and we switched to this dosing based on METHODS encouraging results from 2 patients treated with this regimen. Treatment was repeated every 6 months, or PARTICIPANTS: This retrospective case series was sooner if the patient had worsening of symptoms earlier. approved by the Massachusetts Eye and Ear Infirmary The rationale for regular retreatment every 6 months in (MEEI) Institutional Review Board and was compliant the absence of worsening symptoms is that the peripheral with the Health Insurance Portability and Accountability CD20þ B cells begin to recover about 6–9 months after Act. We reviewed the medical records of all AIR patients rituximab infusion.12 Retreatment consisted of 1 rituximab who were examined on the MEEI Retina/Uveitis Services 1000 mg infusion. Patients were monitored with CD20 sub- from 2010 to 2016 and were treated with rituximab. The sets, complete blood counts, liver function tests, and a com- diagnosis was based on characteristic symptoms (sudden plete metabolic profile at each infusion and every 3 months or subacute painless loss of vision, photopsias, , between infusions. Total immunoglobulin levels were photoaversion, nyctalopia) and findings, abnormal ERG monitored every 6 months. Side effects were recorded at findings including abnormal rod and/or cone responses on each visit. Some patients were treated with rituximab as full-field testing and/or any abnormality on multifocal monotherapy. Others had rituximab added to other immu- testing, presence of antiretinal antibodies, exclusion of nosuppression. Intravenous methylprednisolone sodium other diagnoses, absence of overt inflammation, and no his- succinate 100 mg and hydroxyzine 25 mg were given as tory of hereditary retinal disease.9 An evaluation for malig- part of the pretreatment regimen before each rituximab nancy was done for all patients who did not already have a infusion to decrease the chance of infusion reactions. diagnosis of cancer. All patients had blood sent for antire- tinal antibody testing by Western blot to the Ocular Immu- ASSESSMENT OF QUANTITATIVE AND QUALITATIVE nology Laboratory, Casey Eye Institute, Oregon Health and CHARACTERISTICS IN OPTICAL COHERENCE TOMOGRA- Science University. Snellen visual acuity (VA), visual PHY: SDOCT images were obtained every 3 months. symptoms, ERG, autofluorescence, VF, and Central subfield mean thickness (CSMT) and total macu- SDOCT results were recorded. OCT imaging was lar volume (TMV) were recorded from each visit from a performed with an SDOCT instrument (Spectralis, Heidel- horizontal foveal scan image. They were measured within berg, Germany). Fovea-centered images were acquired (25 the central 1-mm-diameter circle surrounding the fovea lines within a 20-degree horizontal scan and 25 lines within and the central 6-mm-diameter circle surrounding the a 20-degree vertical scan). Full-field ERGs were performed fovea, respectively. A masked ophthalmologist-grader with Burian-Allen electrodes (Hansen Labs, Coralville, ensured that the scans from which CSMT and TMV Iowa, USA) according to International Society for Clinical were recorded were of good quality and without artifact. Electrophysiology of Vision standard protocols. Multifocal electroretinography was performed with the Veris instru- FULL-FIELD AND MULTIFOCAL ELECTRORETINOG- ment (Electro-Diagnostic Imaging, Burlingame, Califor- RAPHY: Full-field and multifocal electroretinography nia, USA) and a stimulus array of 103 hexagons. A were obtained every 6 months. For the full-field electroret- subset of patients was also imaged using a prototype inography, dim scotopic, bright scotopic, 30 Hz flicker AO-SLO camera (Canon, Tokyo, Japan). amplitudes, and 30 Hz flicker implicit times were recorded. Full-field ERG results were compared with age-matched TREATMENT PROTOCOL: Patients evaluated from 2010 control normative data. Fishman and associates found to 2014 were considered for rituximab therapy only after that the threshold for a significant increase or decrease in they had failed treatment with other therapies including intervisit ERG amplitudes at a 95% confidence level for methotrexate, mycophenolate, cyclophosphamide, intra- patients with retinitis pigmentosa and visually normal sub- venous immunoglobulin, and intravitreal triamcinolone. jects was often at or above 25% and not infrequently at or From 2015 to 2016, some patients were treated with ritux- above 40% for certain stimuli and test conditions.13 There- imab as a first-line therapy because of encouraging results fore, for the present study, an increase or decrease of at least from the aforementioned case reports and with our own 40% for any given full-field ERG parameter was used to patients. Participants had testing for renal function, liver classify patients as having an improvement or worsening

VOL. 180 OUTCOMES IN AUTOIMMUNE RETINOPATHY PATIENTS WITH RITUXIMAB 125 of ERG, respectively. Interpretation of the multifocal elec- Japan) that automatically counts individual cones within troretinography was performed by a careful inspection of a selected area. the waves in the individual ERGs and comparing height amplitudes in relation to normative values. STATISTICAL ANALYSIS: Snellen visual acuities were converted to the logMAR scale for the statistical ana- 19 VISUAL FIELDS: VFs were obtained every 6 months with lyses. The primary outcome was the rate of VA change 30-2 Swedish interactive thresholding algorithm standard after rituximab initiation vs before rituximab initiation. testing. For the analysis, patients with coexisting glaucoma We calculated the rate of vision change by dividing the that could impact VF testing were excluded. Reliability change in VA over the corresponding number of months. parameters were checked for all VFs before interpretation. We compared the rate of VA change from the time of first Any field test with a fixation loss false-positive or false- diagnosis to the time of rituximab initiation to (1) the rate negative rate > 25% was considered unreliable and not of VA change in the first 6 months after rituximab initia- included in the analysis unless false-negative could be tion and (2) the rate of visual acuity change over the entire explained by significant VF loss.14 For patients with reliable course of rituximab treatment. Because we included both VFs, we determined whether their VFs were abnormal prior eyes of most patients, we used multilevel mixed-effects to treatment with rituximab. An abnormal VF was defined linear regression.20 By definition, these models are appro- based on the Anderson criteria: a glaucoma hemifield test priate for research designs where data for participants are outside normal limits on 2 consecutive fields or a cluster of organized at more than 1 level. 3 or more non-edge points on the pattern defect at P < Secondary outcomes were (1) changes in CSMT and 5% with 1 point at P < 1% over 2 consecutive fields or TMV on OCT, (2) dim scotopic amplitudes, bright scotopic corrected pattern standard deviation < 5% over 2 consecu- amplitudes, 30 Hz flicker amplitudes, and 30 Hz flicker im- tive fields.15 TheAndersoncriteriahavebeenusedinstudies plicit times on ERG, and (3) cone density on AO-SLO, at on a wide range of neuro-ophthalmic and retinal dis- 6 months post rituximab treatment and at the last visit. The eases.16,17 The severity of the VF loss before treatment was multilevel mixed-effects regression models included the further graded as mild, moderate, or severe. Mild loss was following covariates: age, sex, and concomitant use of other defined as mean deviation (MD) between 2 and <6dB, immunomodulatory medication with rituximab. Mean cone moderate loss was defined as MD between 6 and 12 dB, densities on AO-SLO were also compared between AIR severe loss was defined as >12 dB defect on MD, and a cases and controls and over the course of rituximab treat- value of ‘‘0’’ within the central 5% of fixation was also ment in AIR patients with multilevel mixed-effects linear considered severe. Over the course of treatment with models. All statistical analyses were performed with Stata/ rituximab, improvement or worsening on the VF were IC 12.1 (College Station, Texas, USA). defined as a 2-dB increase or decrease, respectively, in the MD if the lens, cornea, and vitreous status were stable.

ADAPTIVE OPTICS SCANNING LASER OPHTHALMOS- RESULTS COPY IMAGING ANALYSIS IN AUTOIMMUNE RETINOP- ATHY PATIENTS: The AO-SLO images were obtained THIRTY EYES IN 16 PATIENTS (37.5% MALE, MEAN AGE OF 56 with a Canon AO-SLO RD102 prototype. The technical years) were enrolled, including 6 patients with CAR, 1 details of the AO-SLO system used for this study have with MAR, and 9 with npAIR. In 2 patients, only 1 eye been published previously.18 was enrolled because the second eyes had a disease that pre- Six eyes of 4 patients with AIR were imaged serially with cluded VA and imaging assessment—1 had history of a the AO-SLO. High-quality AO-SLO images were chosen longstanding total retinal detachment and the other had from each visit. Cone densities were measured in 10 an opacified cornea from a severe corneal ulcer. The under- different 150 3 150-mm2-window measurement areas lying cancer in CAR patients was lung cancer in 2 patients, from each visit in the same locations at 1, 2, and 3 mm ovarian cancer in 2 patients, breast cancer in 1 patient, and from the fovea. The Supplemental Figure shows representa- Hodgkin lymphoma in 1 patient. Table 1 shows the tive images of how cone density was determined in 1 win- patients’ clinical characteristics. The eyes are grouped dow measurement at the same location at each visit for 1 depending on whether the Snellen VA in the eye improved patient (Supplemental Material available at AJO.com). at least 2 lines, remained stable, or decreased at least 2 lines Also, 6 healthy eyes of 4 subjects who were age- and sex- at the 6-month follow-up visit. This threshold has been matched with cases were chosen as a control group. For used in previous studies.2 the controls, cone density was measured from a single visit Fifteen patients had been treated with immunosuppres- with the same protocol used for the AIR cases. Cone den- sive therapy for their AIR prior to rituximab initiation. sity, the number of cones divided by the area of retina Of these 15 patients, 4 had only been treated with oral sampled, and distance to fovea were estimated by Photore- prednisone. One patient was treatment naı¨ve. Fourteen ceptor Analysis Software version 2.1 (Canon Inc, Tokyo, patients had treatment with agents concomitantly with

126 AMERICAN JOURNAL OF OPHTHALMOLOGY AUGUST 2017 rituximab. These agents included cyclophosphamide, treatment. Thirty percent of eyes had a posterior chamber mycophenolate, intravenous immunoglobulin (IVIg), intra-ocular lens at baseline, and none developed posterior intravitreal triamcinolone (IVT), periocular triamcinolone capsule opacity over the course of treatment. Thirty and bortezomib. The three patients who received IVIg all percent of eyes had a clear lens and forty percent had cat- received IVIg after initiation of rituximab. Two patients aracts at baseline. No cataract progression was noted over were treated with rituximab monotherapy. The mean the course of treatment and no patient underwent cataract follow-up period after rituximab initiation was 15.3 6 surgery. Five eyes (four with posterior chamber intraocular 8.7 months (range 7–34 months). Three CAR patients lens and one with early nuclear sclerosis) had intravitreal or passed away during the follow up period from their under- periocular steroid injections during the course of rituximab lying . treatment.

VISUAL ACUITY OUTCOMES: Table 2 shows the mean OPTICAL COHERENCE TOMOGRAPHY, ELECTRORETINO- rate of vision change before and after treatment from a GRAM, AND VISUAL FIELD OUTCOMES: In the entire multilevel mixed-effects linear model adjusting for age, group of patients, TMV and CMST did not change signif- sex, and use of immunomodulatory medications concomi- icantlybeforevsaftertreatment(at6monthsoratthe tantly with rituximab. In Table 2 the results are listed for last visit) (Table 2). We did observe an overall decrease the entire cohort of eyes as well as divided depending on in mean TMV over the period of rituximab treatment whether the eye showed VA improvement, stability, or (8.08 mm2 before treatment to 7.97 mm2 at last follow- worsening, as described previously for Table 1. In the joint up after rituximab initiation). This difference was greater analysis of all eyes, logMAR visual acuity was declining at a in the 7 eyes that had significant decreases in VA rate of 0.032 units/month prior to rituximab initiation. For (8.07 mm2 before treatment to 7.68 mm2 at last follow- the first 6 months after rituximab initiation, logMAR visual up after rituximab initiation). However, the declines acuity improved at a rate of 0.005 units/month (P ¼ .04), were not statistically significant at 6 months post treat- and from the time of rituximab initiation to the last ment or at the last follow-up when compared with follow-up, visual acuity declined at a slower rate of pretreatment TMV in the overall group or in any of the 0.0005 units/month (P ¼ .005). These rates of visual subgroups. Of note, none of the patients had cystoid change also showed statistically significant improvement macular edema on OCT. in the subanalyses of eyes with stable and improving VA Supplemental Table 2 shows results of ERG testing prior group at 6 months and final follow-up. In the subanalysis to treatment with rituximab and also the percent change in of patients with worsening visual acuity, the rate of visual amplitudes and implicit times 6 months post treatment decline was not different in the first 6 months after rituxi- (Supplemental Material available at AJO.com). None of mab initiation or over the entire rituximab treatment the patients showed significant changes (more than 40%) period (P values ¼ .66 and .69, respectively) when in any ERG amplitudes or implicit times over the course compared with the rate prior to rituximab initiation. The of rituximab treatment. One-third of eyes had extinguished Figure shows the logMAR visual acuity as a function of ERGs at the initial visit and did not have serial ERG testing. time. There is a decline in visual acuity in the 6 months For the VF analysis, 2 patients were excluded because they prior to rituximab initiation that is slowed over the first did not have reliable VFs (Patients 9 and 16). One patient year after rituximab initiation. was excluded because she had coexisting normal tension To evaluate the effect of concomitant immunosuppres- glaucoma (Patient 14). Among patients with reliable VFs, sion on VA outcomes, we examined the logMAR VA in 22 eyes (92%) had an abnormal VF based on the Anderson patients according to the other agent they received along criteria before initiation of rituximab. Two eyes (8%) had with rituximab (Supplemental Table 1; Supplemental normal VFs before rituximab and they remained normal dur- Material available at AJO.com). For patients who received ing course of treatment (Patient 3). Eight eyes had mild VF more than one agent alongside rituximab, we categorized loss at baseline (Patients 2, 8, 10, and 13). Of these 8 eyes, 1 them according the most potent immunosuppressive agent eye progressed to moderate VF loss (left eye of Patient 13) they received using the following potency hierarchy: cyclo- over the course of rituximab treatment. Two eyes had phosphamide > antimetabolites > IVIg > IVT. The moderate VF loss at baseline and both of these eyes remained patients who received rituximab alone and rituximab stable over the course of rituximab therapy. Twelve eyes had with mycophenolate showed improvement in mean vision severe VF loss before initiation of treatment (Patients 1, 4, 5, after rituximab initiation. Patients who received cyclo- 7, 11, 12, and 15); 4 of these worsened during course of treat- phosphamide, IVIg or IVT showed stability or decline in ment (Patients 7 and 15). No eyes showed an improvement VA after rituximab initiation. in VF over the course of treatment. VA decline was attributed to ongoing photoreceptor death. We tried to ensure there were no other causes that ADAPTIVE OPTICS SCANNING LASER OPHTHALMOS- could explain the visual decline in these patients. Lens sta- COPY OUTCOMES: Six eyes of 4 patients had serial AO- tus was recorded in the medical chart over the course of SLO imaging of sufficiently high quality to permit image

VOL. 180 OUTCOMES IN AUTOIMMUNE RETINOPATHY PATIENTS WITH RITUXIMAB 127 128

TABLE 1. Clinical Characteristics Regarding Rituximab Therapy and Antiretinal Antibody Results for Each Patient

Interval Between VA at VA 6 Months Months From First Length of Molecular Diagnosis and Rituximab After Rituximab RituximabInfusion Follow-up Weights VA at 6 Number, Diagnosis, Sex, Age (y)/ Rituximab Initiation Initiation to Most After Rituximab Side Effects During Rituximab of ARA on a a months Accompanying Disease Initiation (mo) (Snellen Chart) (Snellen Chart) Medication Before Rituximab Medication With Rituximab Recent Infusion Initiation (mo) Therapy Western Blot

Improved 1, npAIR, F, 21 7 20/200 20/30 Prednisone, mycophenolate Mycophenolate 6 8 Leukopenia 34, 46, 60, 62 20/125 20/25 2, npAIR, F, 69 4 20/200 20/25 Prednisone - 6 8 - 45, 72 20/200 20/40 Stable 3, npAIR, F, 36 6 20/50 20/60 Prednisone, mycophenolate, Cyclophosphamide 8 8 - 23, 28, 29, 30, 20/60 20/60 cyclosporine 41, 43, 46, 62 b 4, npAIR, M, 75 16 20/100 20/100 Prednisone, methotrexate, IVT, cyclophosphamide 24 25 - 36, 40, 50, 70, cyclophosphamide, IVT 22, 32, 36, 42, 44, 46, 70

A b

MERICAN 5, CAR, F, 49 /Hodgkin 24 20/60 20/60 Prednisone - 6 7 - 65 lymphoma 6, npAIR, F, 37 28 20/200 20/200 Prednisone Cyclophosphamide 16 16 Infusion reaction with 23 20/200 20/200 ventricular tachycardia J UNLOF OURNAL 7, npAIR, M, 49 15 20/200 20/125 Prednisone, IVIg, IVIg, bortezomib, 26 31 - 90–92 20/60 20/60 bortezomib, mycophenolate, mycophenolate cyclophosphamide 8, npAIR, F, 64 48 20/30 20/25 Prednisone, periocular Periocular triamcinolone, 10 16 Sinusitis 29, 92 20/25 20/30 triamcinolone, cyclophosphamide O

PHTHALMOLOGY methotrexate c 9, CAR, M, 75 /Lung cancer 24 20/60 20/60 IVT IVT 6 7 - 45, 46, 94 20/50 20/40 10, npAIR, M, 58 12 20/125 20/125 - Cyclophosphamide 6 10 - 45, 59, 82 20/125 20/125 c 11, CAR, F, 86 /Ovarian 24 20/60 20/60 IVT IVT 6 7 Infusion reaction 35, 46 cancer 20/80 20/80 12, npAIR, F, 66 48 20/20 20/20 Prednisone, IVT, IVIg IVIg, cyclophosphamide 1 34 Infusion reaction 40, 42, 46 20/20 20/20 13, CAR, M, 61/Lung cancer 12 20/40 20/30 Prednisone Cyclophosphamide 12 18 - 40, 44, 62 Worsened 20/200 CF 14, CAR, F, 71/Ovarian 6 20/200 CF Prednisone, IVT IVIg 16 21 - 40, 102, 62 cancer 20/80 20/200 15, MAR, M, 65/Melanoma 5 20/40 20/60 Prednisone, IVT IVT, IVIg 6 16 - 30 20/40 20/400 c 16, CAR, F, 80 /Breast 13 20/40 20/80 Prednisone, IVT, Periocular triamcinolone, 6 12 Infusion reaction 70–72 cancer 20/40 20/80 periocular cyclophosphamide triamcinolone A

GS 2017 UGUST ARA ¼ antiretinal antibodies; CAR ¼ cancer-associated retinopathy; CF ¼ count fingers; IVIg ¼ intravenous immunoglobulin; IVT ¼ intravitreal triamcinolone; MAR ¼ melanoma-associated retinopathy; npAIR ¼ non-paraneoplastic autoimmune retinopathy. a For the visual acuity results for each patient, the right eye values are listed above the left eye values, except for patients who only had 1 eye that could be evaluated. b The left eye of Patient 4 and Patient 5 could not be evaluated owing to corneal scar and chronic total retinal detachment, respectively. c These patients died because of their underlying cancers. TABLE 2. Results of Visual Acuity and Spectral-Domain Optical Coherence Tomography Quantitative Measurements Before and After Treatment With Rituximab

Groups Based On Visual Mean Over 6 P Value, Comparing Mean From Rituximab P Value, Comparing Acuity Results Mean Before Months After 6 Months Post-Rituximab Initiation to Last Last Follow-up to With Treatment Rituximab Rituximab to Pre-Rituximaba Follow-up Pretreatmenta

Rate of logMAR Total (28 eyes) 0.032 0.005 .04 0.0005 .005 vision change Improved (4 eyes) 0.088 0.111 <.0001 0.803 <.0001 (units per month) Stable (18 eyes) 0.222 0.0006 <.0001 0.004 <.0001 Worsened (6 eyes) 0.266 0.405 .66 0.372 .69 TMV Total (30 eyes) 8.08 8.00 .37 7.97 .30 Improved (4 eyes) 7.85 7.90 .07 7.93 .07 Stable (19 eyes) 8.33 8.33 .70 8.33 .70 Worsened (7 eyes) 8.07 7.78 .20 7.68 .15 CSMT Total (30 eyes) 276 275 .78 275 .83 Improved (4 eyes) 276 274 .53 274 .53 Stable (19 eyes) 286 286 .86 287 .87 Worsened (7 eyes) 268 266 .91 264 .68

CSMT ¼ central subfield macular thickness (mm); TMV ¼ total macular volume (mm3). aMixed-effects linear regression test adjusted for age, sex, and concomitant use of immunomodulatory medications with rituximab. analysis during their treatment course. An example of serial ANTIRETINAL ANTIBODY TESTING: Only Patients 2 imaging for 1 patient is shown in the Supplemental Figure. and 4 had serum antiretinal antibodies repeated after ritux- Mean cone density in AIR patients at baseline (14 366 6 imab treatment. Eight months after rituximab therapy, 3365) was less than that of normal controls (16 349 6 Patient 2’s 45-kDa antibody became undetectable but there 726), but this did not reach statistical significance in the was no change in the results for the 72-kDa antibody. Nine linear multilevel mixed model (P ¼ .10). Supplemental months after rituximab treatment, Patient 4’s antiretinal Table 3 shows the multilevel mixed-model results for antibody levels decreased 50% by Western blot. cone density over the rituximab treatment period (Supple- mental Material available at AJO.com). Of these 6 eyes with AIR, 5 eyes had stable VA and 1 eye (Patient 16’s right eye) had decreasing VA during rituximab treatment. DISCUSSION This eye with decreasing VA had a lower cone density at all visits compared to the other 5 eyes. Mean cone densities IN THIS COHORT OF 16 PATIENTS WITH AIR, PATIENTS WERE did not change to a statistically significant degree over experiencing visual decline prior to rituximab initiation the course of therapy (P ¼ .30). and this decline was slowed after rituximab initiation. This difference in the rate of visual acuity change before SIDE EFFECTS AND RITUXIMAB DISCONTINUATION: vs after rituximab was statistically significant. OCT, Six patients developed side effects. One patient who was ERG, and AO-SLO outcomes remained stable over the taking cyclophosphamide concomitantly with rituximab period of rituximab treatment. developed sinusitis. One patient who was also taking For the main outcome of VA, over the course of rituxi- mycophenolate developed leukopenia requiring adjust- mab treatment 4 eyes of 2 patients (14%) had an improve- ment of the mycophenolate dose. Four patients had infu- ment in VA, 19 eyes of 11 patients (63%) were stable, and sion reactions. Three were minor reactions that were 7 eyes of 4 patients (23%) showed worsening. There are treated with an adjustment of the infusion rate. The fourth several reasons why some patients responded significantly reaction was ventricular tachycardia, which required a to rituximab while others worsened. Part of the difference prolonged desensitization protocol with each subsequent in response may be owing to the point in the disease course infusion. Patient 12 chose to discontinue rituximab infu- at which treatment is initiated. The 2 patients who had vi- sions because the palpitations and headaches she experi- sual improvement with rituximab were treated within 4 enced as part of her infusion reaction persisted after her and 7 months, respectively, of their symptom onset and first set of infusions for several days. Patient 14 stopped before there was severe loss of function on ERG. Also, no her infusions because of progressive visual loss despite pathology was found on OCT, color fundus photography, rituximab therapy. None of the patients diagnosed with and autofluoresence in these 2 patients. Two of the 4 pa- npAIR have been diagnosed with a malignancy during or tients who had a decline in vision had extinguished after treatment with rituximab. ERGs prior to rituximab initiation. They had severe outer

VOL. 180 OUTCOMES IN AUTOIMMUNE RETINOPATHY PATIENTS WITH RITUXIMAB 129 .15 many patients in this series had advanced disease before rituximab was initiated, making it difficult for them to achieve an improvement in vision.

.22 Quantitative measures of TMV and CSMT were stable over the treatment course. We did observe an overall decrease in TMV over the period of rituximab treatment,

.30 but this was not statistically significant at 6 months post

VA (logMAR) treatment or at the last visit when compared with pretreat- Initiation of rituximab ment TMV. Loss or disruptions of the ellipsoid zone with

.40 decreased CSMT at the time of diagnosis have been observed in some studies of AIR patients,21,22 but there are no large longitudinal studies of CSMT in AIR. We .52 -6 0 6 12 did not find a significant decrease in CSMT in our Time (months) patients over the course of treatment. Ellipsoid zone and photoreceptor outer segment volume were not available FIGURE. Mean visual acuity as a function of time before and after rituximab initiation. for this study, but may be more sensitive measures than TMV and CSMT in AIR. An analysis with these measures may be able to demonstrate a significant difference with treatment. retinal loss on OCT, suggesting that photoreceptor death We did not find significant changes in the ERG parame- was the primary cause of their poor VA. Their worsening ters in any patients in our study. Some patients did show may be a reflection of the late stage of disease at rituximab improvement in certain parameters, but it is not possible initiation, when widespread death of retinal cells has to say with certainty whether these changes were true already occurred, and/or the refractory nature of their improvement or within the normal variability of the inflammation given prior failure with other agents. testing.13 Fox and associates reported the first case of npAIR There was variability in response to treatment based on that showed improvement in ERG in addition to improve- adjunctive therapy used with rituximab (Supplemental ment in symptoms and visual function.3 One month after Table 1). The patients who gained vision were those who receiving rituximab, the dark-adapted rod-mediated ERG received rituximab alone or rituximab with mycopheno- amplitudes more than doubled and dark-adapted maximal late. Patients who received cyclophosphamide, IVIg, or response A-waves also showed clinically significant in- IVT had, on average, stable or worsening vision after ritux- creases in amplitude. Other case reports have not found a imab. The difference may be owing to selection bias, with significant improvement in ERG parameters, similar to patients with less severe or earlier disease being treated less what we found in this study.4,5,7 Prompt treatment to halt aggressively with rituximab monotherapy or with an retinal cell death is likely key to achieving improvement adjunctive antimetabolite, and patients with more severe in ERG parameters in AIR patients. or rapidly advancing disease being treated with more We usually check antiretinal antibodies on our patients potent (cyclophosphamide) or intensive (IVIg) adjunctive only once at initial presentation to help establish the diag- therapies. nosis. There is a significant cost for antiretinal antibody Three of the case reports on AIR treatment with rituxi- testing, which some patients have to pay for out of pocket, mab documented improvement in VA with treatment.3,5,7 and this limits the ability to obtain the test repeatedly. In A fourth described a patient with CAR in whom visual addition, there is not yet any strong evidence that serial deterioration was successfully halted and vision was antiretinal antibody testing can be used as a reliable indi- stable for 9 months after rituximab initiation until the cator of disease progression and/or treatment patient passed away because of the underlying cancer.4 response.9,23,24 Indeed, in the consensus panel paper on Two case reports, both describing CAR patients with nonparaneoplastic AIR, agreement could not be reached Waldenstrom macroglobulinemia, documented a lack of for the utility of repeating antiretinal antibody testing to response to rituximab.6,8 Thus, of these previously monitor AIR patients.9 We also emphasize that ARA published case reports, half describe improvement with testing can be positive in normal controls and patients rituximab, which is higher than the rate seen in the with other eye diseases.25 Therefore a diagnosis of AIR present study. This disparity is likely owing to publication should not be given based on positive ARA testing alone. bias; single case reports showing positive results are more Instead the patient’s clinical symptoms, examination, and likely to be published than those that do not show ancillary testing should be paramount in establishing the improvement. In this series of patients consecutively diagnosis. Establishing the diagnosis accurately is treated with rituximab at 1 institution, only a minority of especially critical when considering exposing a patient patients showed a significant improvement in vision, to a drug with potentially severe side effects, such as though most achieved stabilization. One caveat is that rituximab.

130 AMERICAN JOURNAL OF OPHTHALMOLOGY AUGUST 2017 Recently, Dabir and associates reported on an MAR case variable. Although we had at least 6 months of follow-up where structural changes were not detectable on SDOCT, on all patients, this is still a relatively short follow-up but were detected on the adaptive optics imaging where the period. Patients were treated with rituximab at different cone density was reduced.11 We obtained serial AO-SLO stages of disease. Most patients were taking a concomitant imaging in a subset of patients. Mean cone densities did immunosuppressive medication with rituximab, so it is not decrease significantly during course of treatment, difficult to isolate the effect of rituximab in these patients. which may indicate halting of cell death. As might be ex- However, most patients were on other immunosuppressive pected, the mean cone densities in AIR cases were less than agents prior to starting rituximab with progressive deterio- those in healthy controls. For some patients, cone densities ration and showed stabilization or improvement that was were significantly diminished. For example, Patient 16 had temporally associated with rituximab initiation. Similar a mean cone density (12 799 cones per mm2) that was much to the previous case series on npAIR,2 we did not have a lower than that of the age- and sex-matched control (16 control group of untreated patients to compare to. We 450 cones per mm2). The difference in cone densities did, however, compare the rate of VA change prior to ritux- between AIR patients and controls was not statistically sig- imab initiation to the rate of VA change after rituximab nificant, however. The failure to achieve statistical signifi- initiation and found that the rate of visual decline was cance may be owing to the small number of patients who less likely after rituximab initiation. We did not routinely had AO-SLO images of sufficient quality to be included. check antiretinal antibodies after treatment, so we are AO-SLO image quality is very sensitive to conditions not able to comment on their value for monitoring disease such as dry eyes and cataracts. The number of AO-SLO and predicting response to treatment; this is a subject that images we included in these analyses is small because we should be pursued with future studies. only included reliable, high-quality images. The AO-SLO prototype is limited to imaging 3 mm The side effects attributable to rituximab in this series around the fovea and it can only detect cone cells because were able to be treated and/or reversed. The risk of infec- of optical resolution. Because AIR can affect the mid- tions, particularly with prolonged exposure to rituximab, periphery or far periphery in certain patients, AO-SLO is an important concern. Patients should be appropriately may have missed some progressive retinal cell loss. Newer educated about these risks. Close monitoring during infu- versions of AO-SLO machines that capture ganglion cells sions is also paramount, as rituximab infusion reactions and rod cells in a wider field of the retina are being devel- are relatively common and in some cases will require oped, and they may be helpful for following AIR progres- modified infusion protocols, including some with desensi- sion and monitoring effects of the treatment. One tization. There is a possible risk of unmasking a subclini- additional limitation is that we did not have serial VF cal, underlying malignancy that is being suppressed by testing on all of our patients. the of npAIR patients with initiation of In summary, the majority of AIR patients treated with any immunosuppressive therapy. We always discuss this rituximab showed stabilization of vision over the treatment risk with our npAIR patients. All of our npAIR patients course. Most of these patients had worsening VA before have an evaluation for possible underlying malignancies rituximab while they were taking other immunosuppressive directed by their primary care physician and/or an oncol- agents. The 2 patients who showed significant VA gains ogist before initiating therapy. For our paraneoplastic AIR were treated early on in their disease course with rituximab. patients, we always have a discussion with their oncologist The majority of patients whose vision worsened despite prior to initiating any immunosuppressive therapy. We treatment were initiated on rituximab at an advanced stage did not observe any new malignancies in any patients of their disease. These results suggest that earlier treatment over the rituximab treatment course; however, the initiation with rituximab may be advantageous. However, follow-up period was short for detection of this potential studies with additional patients and longer follow-up pe- long-term complication. riods are needed to evaluate the effects of rituximab pro- This study has limitations. It is a retrospective series and spectively and further explore the utility of SDOCT and the length of posttreatment follow-up with rituximab was AO-SLO in the monitoring of AIR patients.

FUNDING/SUPPORT: NO FUNDING OR GRANT SUPPORT. FINANCIAL DISCLOSURES: LUCIA SOBRIN IS A CONSULTANT FOR Santen (Osaka, Japan). The following authors have no financial disclosures: Samaneh Davoudi, Nazanin Ebrahimiadib, Cagla Yasa, Damla D. Sevgi, Ramak Roohipoor, Evangelia Papavasilieou, and Jason Comander. All authors attest that they meet the current ICMJE criteria for authorship. The authors thank Norihiko Utsunomiya and Koji Nazato from Canon (Tokyo, Japan) for their assistance with the AO-SLO instrument installation and maintenance for this project.

VOL. 180 OUTCOMES IN AUTOIMMUNE RETINOPATHY PATIENTS WITH RITUXIMAB 131 REFERENCES tudes in normal subjects and patients with retinitis pigmen- tosa. Retina 2005;25(8):1014–1021. 1. Grewal DS, Fishman GA, Jampol LM. Autoimmune retinop- 14. Bourne RR, Jahanbakhsh K, Boden C, et al. Reproducibility athy and antiretinal antibodies: a review. Retina 2014;34(5): of visual field end point criteria for standard automated peri- 827–845. metry, full-threshold, and Swedish interactive thresholding 2. Ferreyra HA, Jayasundera T, Khan NW, He S, Lu Y, algorithm strategies: diagnostic innovations in glaucoma Heckenlively JR. Management of autoimmune retinopathies study. Am J Ophthalmol 2007;144(6):908–913. with immunosuppression. Arch Ophthalmol 2009;127(4): 15. Budenz DL, Rhee P, Feuer WJ, McSoley J, Johnson CA, 390–397. Anderson DR. Comparison of glaucomatous visual field 3. Fox A, Jeffrey B, Hasni S, Nussenblatt R, Sen HN. Rituximab defects using standard full threshold and Swedish interactive treatment for nonparaneoplastic autoimmune retinopathy. threshold algorithms. Arch Ophthalmol 2002;120(9): Can J Ophthalmol 2015;50(6):e101–104. 1136–1141. 4. Dy I, Chintapatla R, Preeshagul I, Becker D. Treatment of 16. Okuma S, Mizoue S, Ohashi Y. Visual field defects and cancer-associated retinopathy with rituximab. J Natl Compr changes in macular retinal ganglion cell complex thickness Canc Netw 2013;11(11):1320–1324. in eyes with intrachoroidal cavitation are similar to those 5. Mahdi N, Faia LJ, Goodwin J, Nussenblatt RB, Sen HN. A in early glaucoma. Clin Ophthalmol 2016;10:1217–1222. case of autoimmune retinopathy associated with thyroid car- 17. Thomas R, George R. Interpreting automated perimetry. cinoma. Ocul Immunol Inflamm 2010;18(4):322–323. Indian J Ophthalmol 2001;49(2):125–140. 6. Sen HN, Chan CC, Caruso RC, Fariss RN, Nussenblatt RB, 18. Park SP, Chung JK, Greenstein V, Tsang SH, Chang S. A Buggage RR. Waldenstrom’s macroglobulinemia-associated study of factors affecting the human cone photoreceptor den- retinopathy. Ophthalmology 2004;111(3):535–539. sity measured by adaptive optics scanning laser ophthalmo- 7. Or C, Collins DR, Merkur AB, Wang Y, Chan CC, scope. Exp Eye Res 2013;108:1–9. Forooghian F. Intravenous rituximab for the treatment of 19. Lange C, Feltgen N, Junker B, Schulze-Bonsel K, Bach M. cancer-associated retinopathy. Can J Ophthalmol 2013; Resolving the clinical acuity categories ‘‘hand motion’’ and 48(2):e35–e38. ‘‘counting fingers’’ using the Freiburg Visual Acuity Test 8. Liu Y, Magro C, Loewenstein JI, et al. A man with paraneo- (FrACT). Graefes Arch Clin Exp Ophthalmol 2009;247(1): plastic retinopathy plus small fiber polyneuropathy associated 137–142. with Waldenstrom macroglobulinemia (lymphoplasmacytic 20. Burton P, Gurrin L, Sly P. Extending the simple linear regres- lymphoma): insights into mechanisms. Ocul Immunol Inflamm sion model to account for correlated responses: an introduc- 2015;23(5):405–409. tion to generalized estimating equations and multi-level 9. Fox AR, Gordon LK, Heckenlively JR, et al. Consensus on mixed modelling. Stat Med 1998;17(11):1261–1291. the diagnosis and management of nonparaneoplastic autoim- 21. Abazari A, Allam SS, Adamus G, Ghazi NG. Optical coher- mune retinopathy using a modified Delphi approach. Am J ence tomography findings in autoimmune retinopathy. Am J Ophthalmol 2016;168:183–190. Ophthalmol 2012;153(4):750–756. 756.e751. 10. Roorda A, Duncan JL. Adaptive optics ophthalmoscopy. 22. Lima LH, Greenberg JP, Greenstein VC, et al. Hyperauto- Annu Rev Vis Sci 2015;1:19–50. fluorescent ring in autoimmune retinopathy. Retina 2012; 11. Dabir S, Mangalesh S, Govindraj I, Mallipatna A, Battu R, 32(7):1385–1394. Shetty R. Melanoma associated retinopathy: a new dimen- 23. Forooghian F. The uncertainty regarding antiretinal anti- sion using adaptive optics. Oman J Ophthalmol 2016;8: bodies. JAMA Ophthalmol 2015;133(7):744–745. 125–127. 24. Forooghian F, Cao S, Cui J, Matsubara JA. The enigma of 12. Edwards JC, Szczepanski L, Szechinski J, et al. Efficacy of autoimmune retinopathy. Int Ophthalmol Clin 2015;55(2): B-cell-targeted therapy with rituximab in patients with rheu- 81–91. matoid arthritis. N Engl J Med 2004;350(25):2572–2581. 25. Ten Berge JC, Schreurs MW, Vermeer J, Meester-Smoor MA, 13. Fishman GA, Chappelow AV, Anderson RJ, Rotenstreich Y, Rothova A. Prevalence and clinical impact of antiretinal an- Derlacki DJ. Short-term inter-visit variability of ERG ampli- tibodies in uveitis. Acta Ophthalmol 2016;94(3):282–288.

132 AMERICAN JOURNAL OF OPHTHALMOLOGY AUGUST 2017