RESEARCH ARTICLE Optical coherence tomography in multiple sclerosis: A 3-year prospective multicenter study Friedemann Paul1 , Peter A. Calabresi2 , Frederik Barkhof3,4, Ari J. Green5, Randy Kardon6,7, Jaume Sastre-Garriga8, Sven Schippling9, Patrick Vermersch10, Shiv Saidha11, Bianca S. Gerendas12, Ursula Schmidt-Erfurth13, Catherine Agoropoulou14, Ying Zhang15, Gustavo Seifer14 & Axel Petzold16,17,18 1 NeuroCure Clinical Research Center and Experimental and Clinical Research Center, Max Delbrueck Center for Molecular Medicine and Charit e- Universit€atsmedizin Berlin, Berlin, Germany 2 Department of Neurology, Johns Hopkins Hospital, Baltimore, Maryland 3 Department of Radiology & Nuclear Medicine, Amsterdam UMC, Vrije Universiteit, Amsterdam, Netherlands 4 Institutes of Neurology & Centre for Medical Image Computing, University College London, London, UK 5 Department of Neurology, Multiple Sclerosis Center, University of California San Francisco, San Francisco, Califonia 6 Iowa City VA Center for Prevention and Treatment of Visual Loss, Department of Veterans Affairs Hospital Iowa City, University of Iowa Hospital and Clinics, Iowa City, Iowa 7 Department of Ophthalmology and Visual Sciences, University of Iowa Hospital and Clinics, Iowa City, Iowa 8 Department of Neurology/Neuroimmunology, Multiple Sclerosis Centre of Catalonia (Cemcat), Hospital Universitari Vall d’Hebron, Universitat Auto noma de Barcelona, Barcelona, Spain 9 Neuroimmunology and Multiple Sclerosis Research Section, University Hospital Zurich, Zurich, Switzerland 10 Univ. Lille, Inserm UMR1172 LilNCog, FHU Precise, Lille, France 11 Department of Neurology, Johns Hopkins University, Baltimore, Maryland 12 Department of Ophthalmology, Vienna Reading Center, Medical University of Vienna, Vienna, Austria 13 Department of Ophthalmology, Medical University of Vienna, Vienna, Austria 14 Novartis Pharma AG, Basel, Switzerland 15 Novartis Pharmaceuticals Corporation, East Hanover, New Jersey 16 Moorfields Eye Hospital, The National Hospital for Neurology and Neurosurgery, London, UK 17 Queen Square Institute of Neurology, University College London, London, UK 18 MS Center Amsterdam, Amsterdam UMC (Locatie VUmc), Amsterdam, Netherlands Correspondence Abstract Friedemann Paul, NeuroCure Clinical Research Center, Charite, Department of Objective: To evaluate changes over 3 years in the thickness of inner retinal Neurology, Experimental and Clinical layers including the peripapillary retinal nerve fiber layer (pRNFL), and com- Research Center, Lindenberger Weg 80, bined macular ganglion cell and inner plexiform layers (mGCIPL), in individu- 13125 Berlin, Germany. Tel: +49 30 450 als with relapsing-remitting multiple sclerosis (RRMS) versus healthy controls; 539705; Fax: +49 30 450 539915; E-mail: to determine whether optical coherence tomography (OCT) is sufficiently sensi- friedemann.paul@charite.de tive and reproducible to detect small degrees of neuroaxonal loss over time that correlate with changes in brain volume and disability progression as measured Funding Information by the Expanded Disability Status Scale (EDSS). Methods: Individuals with The authors wish to thank Carolyn M. Ervin RRMS from 28 centers (n = 333) were matched with 64 healthy participants. for her substantial contribution in the data OCT scans were performed on Heidelberg Spectralis machines (at baseline; analyses, as well as Mark Kirby, Aisling 1 month; 6 months; 6-monthly thereafter). Results: OCT measurements were Towell, and Marie-Catherine Mousseau highly reproducible between baseline and 1 month (intraclass correlation coeffi- (Novartis Ireland Ltd.) for their writing cient >0.98). Significant inner retinal layer thinning was observed in individuals support, funded by Novartis Pharma AG, Basel, Switzerland. FB is supported by the with multiple sclerosis (MS) compared with controls regardless of previous MS- NIHR biomedical research center at UCLH. associated optic neuritis––group differences (95% CI) over 3 years: pRNFL: 1.86 ( 2.54, 1.17) µm; mGCIPL: 2.03 ( 2.78, 1.28) µm (both Received: 21 May 2021; Revised: 1 October p < 0.0001; effect sizes 0.39 and 0.34). Greater inner retinal layer atrophy was 2021; Accepted: 6 October 2021 observed in individuals diagnosed with RRMS <3 years versus >5 years (pRNFL: p < 0.05; mGCIPL: p < 0.01). Brain volume decreased by 1.3% in Annals of Clinical and Translational individuals with MS over 3 years compared to 0.5% in control subjects (effect Neurology 2021; 8(12): 2235–2251 ª 2021 Novartis Pharma AG. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. 2235 This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. OCTiMS Study F. Paul et al. doi: 10.1002/acn3.51473 size 0.76). mGCIPL atrophy correlated with brain atrophy (p < 0.0001). There was no correlation of OCT data with disability progression. Interpretation: OCT has potential to estimate rates of neurodegeneration in the retina and brain. The effect size for OCT, smaller than for magnetic resonance imaging based on Heidelberg Spectralis data acquired in this study, was increased in early disease. changes or when applied in multicenter settings.18 Factors Introduction that influence the quality of scans and affect measure- Multiple sclerosis (MS) is a chronic inflammatory disease ments, including signal strength, location of blood vessels of the central nervous system, characterized by inflamma- in the inner retinal layers, acquisition procedures, etc., tion, demyelination, gliosis, and axonal loss.1 While the have been well described.19,20 Post-acquisition segmenta- inflammatory aspect is routinely monitored in clinical tion of healthy and thinned retinal layers also influences practice using conventional magnetic resonance imaging reproducibility, and can be affected by the robustness of (MRI), easily accessible biomarkers of neurodegeneration the software algorithm. It is therefore essential to assess are needed to monitor disease progression and assess the the test-retest variability, that is, the intersession measure- efficacy of putative neuroprotective drugs.2,3 One such ment difference in pRNFL that reflects the technical and/ potential candidate is an imaging biomarker that is found or operator-based limitations of the method but is not in the retina, namely, peripapillary retinal nerve fiber attributable to the underlying disease pathology. layer (pRNFL).4–7 Because the retina is unmyelinated, The OCTiMS (Optical Coherence Tomography in Mul- measurements of retinal layer thinning are not con- tiple Sclerosis) study is the first large prospective multi- founded by myelin and are thereby ideal for assessing center study designed to determine whether OCT imaging neuroaxonal degeneration. Both the pRNFL, containing is sufficiently sensitive to detect progression in retinal unmyelinated axons emerging from retinal ganglion cell neuroaxonal loss over 3 years in patients with relapsing- neurons, and the combined macular ganglion cell and remitting MS (RRMS) compared with healthy participants inner plexiform layers (mGCIPL), containing retinal gan- in a multicenter trial-like setting. Changes in average glion cell bodies (i.e., from which the axons forming pRNFL and mGCIPL thickness were assessed, along with pRNFL originate) and dendrites may show abnormal their potential association with brain volume loss, thinning over time and their atrophy is a sequelae of Expanded Disability Status Scale (EDSS) score, and other optic neuritis and optic neuropathy.8 Several MS-related clinical and paraclinical measures. studies have shown that thinning of pRNFL and mGCIPL is associated with increased clinical disability and brain Methods volume loss.5,9–15 Optical coherence tomography (OCT) noninvasively Participants captures cross-sectional images of the retina to generate three-dimensional high-resolution maps of the retinal The OCTiMS study (NCT02907281) was a 36-month, architecture. One significant advantage of OCT compared prospective, multicenter, noninterventional study, which with MRI is that it can be more sensitive in detecting enrolled 397 participants (333 RRMS and 64 reference small increments of neurodegeneration over time owing participants) across 28 sites in Australia, Europe and to its spatial resolution on the order of 5 microns, North America. The study was approved by respective approximately 2–3 orders of magnitude better than MRI. institutional review boards and all participants provided OCT may detect, and follow longitudinally, the degen- written informed consent before enrollment. All data were eration of retinal axons and neurons by measuring change captured and monitored by an independent clinical in pRNFL and mGCIPL thickness in patients with research organization (CRO) using electronic case record MS.16,17 OCT technology therefore has the potential to forms (eCRF). As pRNFL thinning may also occur as a provide a reliable and convenient tool to monitor disease result of MS-associated optic neuritis (MS-ON), patients progression, quantify the rate of neurodegeneration, and were categorized into those with prior history of MS-ON assess the potential neuroprotective treatment effect of and those without MS-ON (non-MS-ON) and assessed MS therapies. accordingly. For patients with a history of MS-ON in one Although it is acknowledged as an easily applicable eye, the eye with prior MS-ON was studied. For patients technique, OCT repeat measurement variability must be with a history of MS-ON in both eyes, or no optic neuri- considered when assessing within-subject longitudinal tis (ON) at all, the eye with the lowest respective pRNFL 2236 ª 2021 Novartis Pharma AG. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. F. Paul et al. OCTiMS Study thickness at baseline was included. Healthy, age-matched, bundle volume scan (see supportive information Data S1 control participants were included as the reference group for more details on techniques used). All OCT scans in all models. Participants who experienced an adverse underwent quality control (QC) assessment by the central event of MS-ON during the study were re-categorized to reading center (Vienna Reading Center [VRC]). Figure S1 the MS-ON subgroup. gives an overview of the scan patterns and the evaluation fields for each pattern used in the analysis. OCT scans were performed at baseline and at every Objectives visit (1, 6 months, and every 6 months thereafter) for The primary study objective was to evaluate change in both groups. All operators at participating sites were cer- average pRNFL thickness using OCT in patients with tified by the VRC for the use of the correct scanning pro- RRMS over a 36-month follow-up period and compare it tocol, correct pseudonymization of study-specific image with the findings from a group of healthy participants to labelling, and image transfer before patient inclusion determine whether the technology is sufficiently sensitive according to standardized certification processes to obtain to detect disease-related pRNFL changes over time. Sec- images for the above-described acquisitions. ondary objectives were as follows: to evaluate short-term Automated segmentation was performed with the man- reproducibility of pRNFL thickness measurements over a ufacturer’s software (HEYEX version 1.9.10.0, Viewing 4-week interval by test re-test estimation; to evaluate the Module version 6.0.9.0) and subsequently corrected by correlation between change in pRNFL thickness with certified and trained graders of the VRC, supervised by a change in brain volume and disability progression in retina specialist, according to a predefined reading proto- patients with RRMS. In addition, we evaluated change in col. All graders were masked to the clinical status of the mGCIPL thickness over 36 months in patients with study participants (MS or healthy control group) and any RRMS and reference participants, as well as the relation- other diagnostic information. Studies were performed and ship between change in mGCIPL thickness and change in described in accordance to the QC criteria at the VRC, brain volume and disability progression in patients with OSCAR-IB, and the Advised Protocol for OCT Study Ter- RRMS. pRNFL and mGCIPL thickness changes were also minology and Elements (APOSTEL) criteria.20,22 The compared between subgroups according to duration of detailed OCT scanning protocol is provided as supporting MS diagnosis determined at the time of enrollment. information. Finally, we explored the correlation of change in pRNFL thickness with change in other clinical and paraclinical Ophthalmology examination measures and patient-reported outcomes in patients with RRMS including visual acuity, MRI measures, cognitive An ophthalmologic examination was performed at screen- performance, and quality of life. ing and every 12 months thereafter. The examination included ophthalmic medical history, slit lamp examina- tion, dilated ophthalmoscopy, best corrected visual acuity Key inclusion/exclusion criteria testing at low 2.5% contrast and high contrast (number All participants (aged 18–65 years inclusive) were enrolled of letters read correctly on an Early Treatment Diabetic and followed up between 29 May 2012 and 24 July 2017. Retinopathy Study scale), and perimetry testing (optional Key inclusion criteria for the MS group were as follows: a at sites where required technical equipment was avail- diagnosis of RRMS as defined by the 2005 revision to the able). If, during the study, worsening of visual acuity or McDonald criteria21; MS disease duration of more than any clinically acute ON were suspected, an unscheduled 1 year (from diagnosis of MS) before study entry (screen- ophthalmology examination was performed. ing). The list of exclusion criteria is included in Table S1. Magnetic resonance imaging Optical coherence tomography All study participants underwent MRI scanning (1.5 or OCT scans in all participants were performed on Spec- 3T) of the brain at screening and at yearly intervals. In tralis OCT1 machines (Heidelberg Engineering, Heidel- the RRMS patient group, pulse-sequences included 3DT1 berg, Germany), software version 1.6.2.0 or higher, with for brain volume measurement, proton density (PD)/T2 the eye tracking function enabled for best accuracy and for new lesion count and volume, T1 for new hypointense the use of the baseline scan for alignment of subsequent lesion count and volume, and gadolinium-enhanced T1 scans on the same area of the retina. Data were collected sequences for lesion count and volume. Lesions were from a peripapillary ring, a macular volume scan, the identified by expert radiology readers and quantified by optic nerve head volume scan, and the papillomacular trained technicians using MIPAV software with local ª 2021 Novartis Pharma AG. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. 2237 OCTiMS Study F. Paul et al. thresholding. Brain volume measurements were per- both the MS-ON and non-MS-ON subgroups. The group formed using SIENA[X] (http://fsl.fmrib.ox.ac.uk). In the difference in least squares (LS) mean of change from reference participant group, 3DT1 sequences for brain baseline in average pRNFL thickness (patients with MS volume measurements (all MRI scans) and T2 lesions reference subjects), which reflects the change from base- (screening scan only) were performed. All MRIs were per- line for the MS group, after adjusting for the change in formed in a standardized manner according to the MRI the reference group (e.g., to account for age-related manual upon performance of a successful dummy-scan. change in healthy subjects), and the associated 95% CI Each MRI scan performed was previewed by the local and p-value are presented for each time point. The LS neuroradiologist for incidental findings before the scans mean of pRNFL change from baseline within each group were sent to and further processed by the MRI central is presented for Month 36. The analysis of the primary reading center at VU University Medical Center in Ams- variable was repeated for the subfield measures of pRNFL terdam. thickness: papillomacular bundle, overall temporal quad- rant, and temporal fields 6. An additional analysis was performed to evaluate the Clinical assessments change in mGCIPL thickness at Month 36 based on a EDSS mixed effect model. Change from baseline in both pRNFL and mGCIPL thickness at Month 36 was stratified by dis- Disability was assessed every 6 months by certified EDSS ease duration [defined as time from MS diagnosis to raters in clinical MS studies.23 Disability progression was study screening (≤3 years, >3 to ≤5 years, >5 years)] for defined as a one-point EDSS increase for those with a patients with MS and subgroups of ON history (MS-ON, baseline score of ≤5.0, or a 0.5-point increase for those non-MS-ON) and groups of different disease duration with a baseline score >5.0, with the EDSS progression were compared using an analysis of variance model confirmed after 6 months. (ANOVA). Regression analyses were applied to assess the relation- Symbol digit modalities test ship between the change from baseline in EDSS/percent- age change in brain volume and the change from baseline The symbol digit modalities test (SDMT) was performed in average pRNFL/mGCIPL thickness at Month 36. All every 12 months to assess attention and information pro- results are presented with nominal p-value with no cessing speed. During the test, subjects substituted sym- adjustment for multiplicity due to the observational char- bols in a row by the corresponding number and acter of the study. responded verbally.24,25 Results Patient-reported indices for multiple sclerosis The patient-Reported Indices for Multiple Sclerosis (PRI- Baseline characteristics MuS) instrument was used to assess patients’ quality of A total of 333 patients with MS and 64 reference partici- life (QoL) and included activities specifically designed for pants without neurologic disease provided a baseline and MS. The test was performed every 12 months.26 at least one post baseline pRNFL measurement in at least one eye. Thirty-nine (11.7%) patients with MS and eight (12.5%) reference participants discontinued the study Statistical methodology (Fig. 1). Demographics were well balanced between Statistical analyses were performed using SAS (version patients with MS and reference participants, with a small 9.4). All hypothesis tests were evaluated at a 0.05 level of difference found for age (38.9 vs. 39.4, p = 0.003), which significance. All confidence intervals (CIs) were two sided was not deemed clinically relevant. Normalized brain vol- at a 95% confidence level. ume was lower in the MS groups (p < 0.0001) (Table 1). Average pRNFL thickness was summarized by group Differences between MS-ON and non-MS-ON patients in and ON history at each visit, including change from base- terms of cognitive performance, normalized brain volume line. A mixed-model repeated measures (MMRM) and disease characteristics including MS duration, EDSS, approach was applied to compare the change from base- and T2 lesion volume were all statistically non-significant line in average pRNFL thickness between groups, consid- (Table 2). Ocular characteristics, including measures of ering results from all available time points, and adjusting visual acuity and pRNFL/mGCIPL thickness, were highest for age, gender, and baseline pRNFL thickness. The refer- in eyes of heathy participants, lowest in eyes with past ence subjects were included as the reference group for MS-ON, and intermediate in eyes of patients without 2238 ª 2021 Novartis Pharma AG. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. F. Paul et al. OCTiMS Study Enrollment Eligible patients (n=422) Safety set (n=422) MS patients (n=353) Healthy controls (n=69) ♦ The safety set consisted of all participants who ♦ The safety set consisted of all participants who were enrolled into the study, i.e. those who were enrolled into the study, i.e. those who provided written informed consent to participate in provided written informed consent to participate in the study and satisfied all eligibility criteria at the study and satisfied all eligibility criteria at screening. screening. Full analysis set n=397 MS patients (n=333) Healthy controls (n=64) Excluded (n=20) Excluded (n=5) ♦ The full analysis set excluded all ♦ The full analysis set excluded all participants from the safety set who did participants from the safety set who did not provide a baseline and at least one not provide a baseline and at least one post-baseline retinal nerve fiber layer post-baseline retinal nerve fiber layer measurement in at least one eye rated as measurement in at least one eye rated as a valid assessment by the optical a valid assessment by the optical coherence tomography central reading coherence tomography central reading center. center. Completed n=350 MS patients (n=294) Healthy controls (n=56) Discontinued (n=39) Discontinued (n=8) Adverse event(s) (n=2) Subject withdrew consent (n=5) Subject withdrew consent (n=18) Lost to follow-up (n=3) Lost to follow-up (n=17) Administrative problems (n=1) Protocol deviation (n=1) Figure 1. CONSORT diagram. MS-ON (Table 3). All differences between groups (refer- included in the non-MS-ON group, but they were re- ence vs. MS and MS-ON vs. non-MS-ON) were signifi- categorized to the MS-ON group at follow-up (after the cant. Six patients developed MS-ON during the study episode). While non-severe glaucoma was not part of the period. Prior to the episode, these patients had been exclusion criteria, it should be noted that only two ª 2021 Novartis Pharma AG. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. 2239 OCTiMS Study F. Paul et al. Table 1. Baseline demographics and clinical characteristics. Group MS group Reference group difference All patients Patients without Patients with Healthy participants Characteristics with MS (N = 333) MS-ON1 (N = 235) MS-ON (N = 98) (N = 64) p-value* Age (years) Mean  SD 38.9  8.56 38.7  8.56 39.4  8.60 35.4  9.73 0.003 Median (min, 39.0 (19.0, 55.0) 39.0 (19.0, 55.0) 39.0 (21.0, 55.0) 35.0 (21.0, 53.0) max) Sex, n (%) Female 233 (70.0) 166 (70.6) 67 (68.4) 45 (70.3) 0.956 Male 100 (30.0) 69 (29.4) 31 (31.6) 19 (29.7) Race, n (%) Caucasian 318 (95.5) 223 (94.9) 95 (96.9) 61 (95.3) 0.752 Black 4 (1.2) 3 (1.3) 1 (1.0) 0 Asian 5 (1.5) 4 (1.7) 1 (1.0) 2 (3.1) Other 6 (1.8) 5 (2.1) 1 (1.0) 1 (1.6) SDMT Mean  SD 55.6  13.37 55.2  13.10 56.4  14.02 n/a n/a Median (min, 58.0 (0.0, 90.0) 56.0 (16.0, 90.0) 59.0 (0.0, 81.0) n/a max) MRI normalized brain volume (cm3) Mean  SD 1511  103 1515  99 1501  112 1589  101 <0.0001 Median (min, 1514.0 (1203.0, 1519.0 (1203.0, 1503.0 (1270.0, 1603.0 (1333.0, max) 1774.0) 1774.0) 1748.0) 1895.0) MRI, magnetic resonance imaging; MS, multiple sclerosis; MS-ON, multiple sclerosis-associated optic neuritis; n/a, not available; SD, standard devi- ation; SDMT, symbol digit modalities test. N is the total number of subjects in the analysis population (different from the number of subjects with data available). 1 Without previous MS-ON. *All p-values are two-side. For continuous variables, p-value is from two sample t-test comparing the MS and reference groups. For categorical variables, if any of the entries is <10, then p-value is from Fisher’s exact test; if all entries are >=10, then p-value is from Chi-square test. patients (0.6%) had glaucoma in the MS group and 1 for change in average pRNFL thickness at Month 36 was (1.4%) in the reference group. 1.86 ( 2.54, 1.17) lm (p < 0.0001; Table 4); this thus represents the change from baseline for the MS group after adjusting for changes (including age-related change) Changes in inner retinal layers thicknesses in healthy subjects. The difference between the two The decrease in pRNFL thickness from baseline (unad- groups for change in average pRNFL thickness were sta- justed for differences in baseline thickness at the time of tistically significant from Month 18 onward: Month 18: enrollment) over time was greater for patients with RRMS 1.13 ( 1.75, 0.52) lm (p = 0.0003); Month 24: 1.11 compared with the reference group (Fig. 2). The mean (s- ( 1.76, 0.46) lm (p = 0.0008); and Month 30: 1.16 tandard deviation [SD]) average pRNFL thickness ( 1.82, 0.49) lm (p = 0.0007). At Month 6 and Month decrease from baseline at Month 36 was 1.61 (4.60) µm 12, the group differences were 0.10 ( 0.65, 0.46) lm in patients with RRMS compared with 0.09 (2.98) µm in (p = 0.7289), and 0.55 ( 1.13, 0.03) lm (p = 0.0621), reference participants. Decreases in pRNFL thickness over respectively. The LS mean group differences in change in time were similar in the non-MS-ON and MS-ON sub- pRNFL thickness were similar regardless of MS-ON his- groups. tory (Table 4). After adjusting for potential imbalance between groups The mixed effect model analysis conducted on mGCIPL (i.e., MS vs. reference subjects) in baseline covariates measurements yielded similar results at Month 36, including age, gender, and baseline RNFL score, the least demonstrating greater thinning of mGCIPL in the MS square mean of change from baseline in RNFL at Month group compared with control participants over the 36 was 1.74 with 95% CI ( 2.04, 1.45) in the MS 3 years. The mean (SD) average mGCIPL thickness at group versus 0.12 with 95% CI ( 0.51, 0.74) in the refer- Month 36 decreased from baseline by 0.39 (3.096) lm in ence group (Table 4). The LS mean difference (95% CI) patients with MS and increased by 1.02 (5.063) lm in ref- between patients with RRMS and reference participants erence participants. The LS mean difference (95% CI) 2240 ª 2021 Novartis Pharma AG. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. F. Paul et al. OCTiMS Study Table 2. Baseline disease characteristics of the MS group. All patients with MS Patients without MS-ON1 Patients with MS-ON Characteristics (N = 333) (N = 235) (N = 98) Time since MS diagnosis Years, mean (SD) 7.28  5.75 n/a n/a 1–≤3 years, n (%) 100 (30.0) 69 (29.4) 31 (31.6) >3–≤5 years, n (%) 44 (13.2) 34 (14.5) 10 (10.2) >5 years, n (%) 189 (56.8) 132 (56.2) 57 (58.2) Previous DMT, n (%) Fingolimod 45 (13.5) n/a n/a Natalizumab 60 (18.0) n/a n/a iDMT (interferons or glatiramer acetate) 162 (48.6) n/a n/a DMF or teriflunomide 14 (4.2) n/a n/a Other 5 (1.5) n/a n/a No treatment 61 (18.3) n/a n/a EDSS score Mean  SD 1.9  1.43 1.9  1.42 2.1  1.44 Median (min, max) 2.0 (0.0, 7.0) 1.5 (0.0, 7.0) 2.0 (0.0, 6.0) MRI PD/T2 lesion volume (mm3) Mean  SD 3663  5095 3523  4659 4037  6036 Median (min, max) 1922.5 (0.0, 48514.0) 1880.0 (0.0, 34627.0) 2206.0 (66.0, 48514.0) DMF, dimethyl fumarate; iDMT, injectable disease modifying therapy; EDSS, expanded disability status scale; max, maximum; min, minimum; MRI, magnetic resonance imaging; MS, multiple sclerosis; MS-ON, multiple sclerosis-associated optic neuritis; n/a, not available; PD, proton density; SD, standard deviation. N is the total number of subjects in the analysis population (different from the number of subjects with data available). 1 Without previous MS-ON. Table 3. Ocular baseline characteristics. MS group Reference group Group difference All patients with MS Patients without MS-ON1 Patients with MS-ON Healthy participants Number of patients N = 333 N = 235 N = 98 N = 64 p-value* Retinal thickness (lm) pRNFL 89.0  15.30 91.7  14.21 82.5  15.85 100.2  9.23 <0.0001 mGCIPL 79.3  14.22 82.2  12.91 72.5  14.94 91.9  8.42 <0.0001 Visual acuity (ETDRS letters) High contrast 56.1  9.49 57.0  8.25 53.8  11.69 59.6  5.42 <0.0001 Low contrast 27.8  11.65 29.8  10.44 22.9  12.95 34.3  7.28 <0.0001 Values are mean  SD. N is the total number of subjects in the analysis population. ETDRS, early treatment diabetic retinopathy study; mGCIPL, combined macular ganglion cell-inner plexiform layer; MS, multiple sclerosis; MS-ON; multiple sclerosis-associated optic neuritis; ON, optic neuritis; pRNFL, peripapillary retinal nerve fiber layer; SD, standard deviation. 1 Without previous MS-ON *All p-values are two-sided from two sample t-test comparing the MS and reference groups. p-values for the subgroup comparison MS-ON versus non-MS-ON were also calculated and were all <0.0001. between patients with RRMS and reference participants between groups for temporal field 6 RNFL thickness, for change in mGCIPL thickness at Month 36 (i.e., rate there was a significant decrease in the papillomacular of thinning corrected with healthy age-related attrition) bundle and temporal quadrant RNFL thickness in patients was 2.03 ( 2.78, 1.28) (p < 0.0001; Table 4). Results with MS compared with reference participants. were similar regardless of MS-ON history. Comparison of change in papillomacular bundle, tem- Effect of time from diagnosis on OCT poral quadrant pRNFL thickness, and temporal field 6 parameters pRNFL thickness between groups at Month 36 are out- lined in Table 5 (see Fig. S1 for the description of the A greater decrease in both pRNFL and mGCIPL thickness fields). While there were no significant differences was observed in patients diagnosed more recently ª 2021 Novartis Pharma AG. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. 2241 OCTiMS Study F. Paul et al. Figure 2. Mean change from baseline in pRNFL thickness over time by ON status. The change from baseline has been overlaid with the known physiological variation between measurements and is shown for months 1 to 36 1(Balk et al. Invest Ophthalmol Vis Sci 2012;53:1251–1257). SE, standard error; M, month; MS, multiple sclerosis; MS-ON; multiple sclerosis-associated optic neuritis; ON, optic neuritis; pRNFL, peripapillary retinal nerve fiber layer; n, number of eyes observed per time point. (≤3 years) than in patients with a longer time from diag- Brain volume nosis (>5 years) (p < 0.05 for pRNFL and p < 0.01 for mGCIPL, based on an ANOVA model) (Fig. 3). This The mean percentage (SD) decrease in brain volume from observation is in line with the primary analysis which baseline to Month 36 was 1.3% (1.32) for patients with demonstrated a positive correlation between baseline MS and 0.5% (0.67) for reference participants. While the pRNFL thickness and the rate of pRNFL thinning percentage change from baseline in brain volume was not (p < 0.0001). Patients with thicker pRNFL at baseline (as significantly associated with the change in average pRNFL observed in early disease) are therefore more likely to thickness, it was significantly associated with change in show faster pRNFL loss. mGCIPL thickness (p < 0.0001, Table 7). Additionally, baseline brain volume was significantly associated with baseline pRNFL thickness (p = 0.0047) but not with base- Test re-test reproducibility of pRNFL OCT line mGCIPL (p = 0.0518) Table 8. measurements The test re-test reproducibility analysis of all pRNFL mea- Expanded disability status scale surements at baseline and 1 month by OCT reported intraclass correlation coefficient (ICC) values of ≥0.98, The mean (SD) EDSS increased from baseline by 0.15 regardless of MS status or ON status (Table 6). (0.86) (median = 0.0 [0.00–0.50]) at Month 36 in the MS 2242 ª 2021 Novartis Pharma AG. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. F. Paul et al. OCTiMS Study Table 4. Change from baseline in pRNFL and mGCIPL thickness and comparison between groups at Month 36. Group difference (95% CI) in change LS mean of change from baseline (Patients with ON history from baseline MS – Reference participants, Visit group n (95% CI) adjusted for baseline) p-value 1 pRNFL thickness (µm) All Patients with MS (N = 333) 284 1.74 ( 2.04, 1.45) 1.86 ( 2.54, 1.17) <0.0001 Reference participants (N = 64) 56 0.12 ( 0.51, 0.74) Non-MS-ON Patients with MS (N = 235) 200 1.79 ( 2.12, 1.45) 1.88 ( 2.59, 1.16) <0.0001 Reference participants (N = 64) 56 0.09 ( 0.55, 0.72) MS-ON Patients with MS (N = 98) 84 1.61 ( 2.25, 0.96) 1.80 ( 2.71, 0.88) 0.0001 Reference participants (N = 64) 56 0.19 ( 0.41, 0.79) mGCIPL thickness (µm)2 All Patients with MS (N = 333) 285 0.54 ( 0.85, 0.24) 2.03 ( 2.78, 1.28) <0.0001 Reference participants (N = 64) 55 1.49 (0.80, 2.18) Non-MS-ON Patients with MS (N = 235) 201 0.47 ( 0.83, 0.12) 1.77 ( 2.51, 1.03) <0.0001 Reference participants (N = 64) 55 1.30 (0.63, 1.97) MS-ON Patients with MS (N = 98) 84 1.58 ( 2.32, 0.85) 3.89 ( 5.17, 2.61) <0.0001 Reference participants (N = 64) 55 2.31 (1.37, 3.25) N is the total number of participants in the analysis population, n is the number of participants with data available. The average pRNFL thickness was the mean of the circular scan subfields 1 to 8. Eye was nested within subject, and both were included as random effects. For participants with a history of ON, the unaffected eye was excluded. The reference participants were included as the reference group in all models. Participants who experienced an adverse event of ON during the study were re-categorized and included in the ON history subgroup. CI, confidence interval; mGCIPL, combined macular ganglion cell and inner plexiform layers; LS, least squares; MS, multiple sclerosis; MS-ON, MS- associated optic neuritis; pRNFL, peripapillary retinal nerve fiber layer. 1 Mixed model for repeated measures (MMRM) included terms for group, visit, age, gender, and the interaction of group and visit with baseline average pRNFL thickness as a covariate. 2 Mixed model includes terms for group, age, gender, and baseline mGCIPL thickness as a covariate. Table 5. Change from baseline in retinal parameters thickness, and comparison between groups at Month 36. Reference Patients with MS participants N = 333 N = 64 Papillomacular bundle thickness (µm) LS mean of change from baseline 0.55 ( 0.85, 0.24) 0.28 ( 0.36, 0.92) Group difference in change from baseline1 (95% CI) 0.82 ( 1.53, 0.12), p = 0.0215 Temporal quadrant pRNFL thickness (µm) LS mean of change from baseline 0.72 ( 1.0, 0.42) 0.13 ( 0.49, 0.75) Group difference in change from baseline1 (95% CI) 0.85 ( 1.53, 0.16), p = 0.0156 Temporal field 6 pRNFL thickness (µm) LS mean of change from baseline 0.75 ( 1.11 0.40) 0.14 ( 0.88, 0.59) Group difference in change from baseline1 (95% CI) 0.61 ( 1.42, 0.20), p = 0.1373 CI, confidence interval; LS, least squares; MS, multiple sclerosis; pRNFL, peripapillary retinal nerve fiber layer; SD, standard deviation. 1 Group difference at Month 36 was analyzed using mixed model for repeated measures (MMRM) which included terms for group, visit, age, gen- der, and the interaction of group and visit with baseline value as a covariate. N is the total number of subjects in the analysis population (different from the number of subjects with data available). ª 2021 Novartis Pharma AG. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. 2243 OCTiMS Study F. Paul et al. ALL Non-MS-ON MS-ON 0 Mean change from baseline in pRNFL thickness (µm) (SE) -0.5 -1 -1.5 n=326 n=225 n=101 -2 -2.5 n=73 n=55 n=18 -3 n=161 n=113 -3.5 n=48 -4 ALL Non-MS-ON MS-ON 0 Mean change from baseline in mGCIPL thickness(µm) (SE) n=325 n=225 n=100 -0.5 -1 n=115 n=76 n=58 n=164 -1.5 n=18 -2 n=49 -2.5 ≤3 years, >3-≤5 years >5 years Figure 3. Change from baseline in pRNFL and mGCIPL thickness at Month 36, stratified by time from MS diagnosis. Pairwise comparisons between the three MS duration subgroups were conducted based on an analysis of variance (ANOVA) using Tukey’s HSD test (unadjusted for differences in baseline thickness). Only the differences between <=3 and >5 groups were found significant (p < 0.05 and p < 0.01 for pRNFL and mGCIPL, respectively). Note that no patient with disease <1 year was included, as per the inclusion criteria. n is the number of eyes with data available. mGCIPL, combined macular ganglion cell and inner plexiform layers; MS, multiple sclerosis; MS-ON; multiple sclerosis-associated optic neuritis; pRNFL, peripapillary retinal nerve fiber layer; SE, standard error. Table 6. Test re-test reproducibility of pRNFL OCT measurements. MS group Reference group All patients with MS Patients without Patients with Healthy participants N = 656 MS-ON1 N = 537 MS-ON N = 119 N = 122 n 642 524 118 121 Baseline, (µm) 88.84  15.35 91.52  13.89 76.86  15.87 100.45  9.15 Month 1, (µm) 88.96  15.41 91.64  14.01 77.03  15.78 100.13  9.34 Difference, (µm) 0.12  1.82 0.14  1.86 0.01  1.60 0.35  1.73 95% CI for mean ( 0.02, 0.26) ( 0.02, 0.30) ( 0.28, 0.30) ( 0.66, 0.04) p-value 0.1036 0.0840 0.9429 0.0271 ICC (95% CI) 0.993 (0.992, 0.994) 0.991 (0.989, 0.992) 0.995 (0.993, 0.996) 0.982 (0.974, 0.987) Values are mean  SD unless otherwise indicated CI, confidence interval; ICC, intraclass correlation coefficient; MS, multiple sclerosis; MS-ON; multiple sclerosis-associated optic neuritis; N is the total number of eyes in the analysis population, n is the number of eyes with data available. 1 Without previous MS-ON 2244 ª 2021 Novartis Pharma AG. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. F. Paul et al. OCTiMS Study Table 7. Regression analysis: relationship between percentage brain percentage of patients with T1 gadolinium enhancing volume change from baseline to Month 36 and change in overall lesions and new PD/T2 lesions are presented in (Table 9). average mGCIPL thickness (µm) from baseline to Month 36. There was no linear relationship (as assessed by a regres- Regression parameter (A) Parameter estimate (SE) p-value sion analysis) observed between the change from baseline in T1 gadolinium enhancing lesion count or volume and mGCIPL thickness (µm) 0.022 (0.005) <0.0001 the change from baseline in average pRNFL thickness at Age (years) 0.002 (0.008) 0.7986 Duration of disease (years) 0.014 (0.013) 0.2734 different time points. However, there was a statistically MS-ON history (yes/no) 0.081 (0.176) 0.6468 significant inverse relationship between the number of Goodness of fit: adjusted R-squared = 0.061 new PD/T2 lesions from baseline and change from base- line in pRNFL thickness at Month 36, with a parameter The most affected eye was used for analyses. Full analysis set (MS estimate (SD) for pRNFL thickness of 0.69 (0.24) and reference individuals, N = 397). (p = 0.0048), suggesting that patients with a greater mGCIPL, macular ganglion cell and inner plexiform layers; MS-ON, multiple sclerosis-associated optic neuritis; SE, standard error. increase in the number of T2 lesions had more pRNFL thinning (Table 10). There was no linear relationship between the change Table 8. Regression analysis: relationship between normalized brain from baseline in PD/T2 lesion volume and change from volume at baseline and (A) pRNFL thickness at baseline; (B) mGCIPL baseline in pRNFL thickness observed at any time thickness at baseline. point. Regression parameter (A) Parameter estimate (SE) p-value pRNFL thickness (µm) 1.0 (0.35) 0.0047 Age (years) 3.3 (0.60) <0.0001 Table 9. Change in MRI lesions in patients with MS over the study Duration of disease (years) 2.8 (0.96) 0.0043 period. MS-ON history (yes/no) 3.9 (12.57) 0.7552 Goodness of fit: adjusted R-square = 0.166 Patients with MS, N = 333 Regression parameter (B) Parameter estimate (SE) p-value Presence of T1 Gd-enhancing lesions, n (%) Baseline 50 (15.2) mGCIPL thickness (µm) 0.8 (0.39) 0.0518 Month 12 67 (22.1) Age (years) 3.2 (0.63) <0.0001 Month 24 75 (26.8) Duration of disease (years) 2.5 (1.01) 0.0124 Month 36 77 (27.5) MS-ON history (yes/no) 3.2 (13.60) 0.8168 T1 Gd-enhancing lesion count, mean  SD Goodness of fit: adjusted R-squared = 0.138 Baseline 0.3  1.07 Month 12 0.5  1.48 The most affected eye was used for analyses. Full analysis set (MS Month 24 0.8  2.00 and reference individuals, N = 397). Month 36 0.9  2.77 mGCIPL, macular ganglion cell and inner plexiform layers; MS-ON, T1 Gd-enhancing lesion volume (mm3), mean  SD multiple sclerosis-associated optic neuritis; ON, optic neuritis; pRNFL, Baseline 30.2  119.19 peripapillary retinal nerve fiber layer; SE, standard error. Month 12 54.9  167.89 Month 24 83.8  267.85 Month 36 100.3  323.88 patient group, with a mean (SD) time to onset of EDSS Presence of new PD/T2 lesion, n (%) progression of 18.6 (8.5) months. Twenty-three (7%) Change from baseline at Month 12 132 (43.1) patients had a 6 month confirmed disability progression by Change from baseline at Month 24 171 (59.8) Change from baseline at Month 36 180 (64.3) end of study. PD/T2 lesion count, mean  SD No linear relationship was observed between the Change from baseline at Month 12 +2.4  6.34 change from baseline in EDSS score and change from Change from baseline at Month 24 +4.7  10.09 baseline in average pRNFL thickness, regardless of dis- Change from baseline at Month 36 +5.9  12.81 ease progression. The change from baseline in mGCIPL PD/T2 lesion volume (mm3), mean  SD thickness was not correlated with the change in EDSS Change from baseline at Month 12 + 92.9  1045.11 score either. Change from baseline at Month 24 +153.2  1156.19 Change from baseline at Month 36 +247.5  1311.52 Gd-enhancing, gadolinium-enhancing; MRI, magnetic resonance imag- MRI lesions ing; MS, multiple sclerosis; PD, proton density. N is the total number Changes from baseline in T1 gadolinium enhancing and of subjects in the analysis population (different from the number of PD/T2 lesion count and volume as well as in the subjects with data available). ª 2021 Novartis Pharma AG. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. 2245 OCTiMS Study F. Paul et al. Table 10. Regression analysis: relationship between the number of new PD/T2 lesions from baseline and the change in pRNFL thickness Safety from baseline. None of the serious adverse events reported were related Parameter estimate to study procedures. There were no new adverse safety Regression parameter (SE) p-value signals identified from vital sign measurements during the study. pRNFL thickness (µm) 0.69 (0.24) 0.0048 Baseline PD/T2 lesion volume (mm3) 0.00 (0.00) 0.0699 Age (years) 0.43 (0.09) <0.0001 Discussion Duration of disease (years) 0.05 (0.14) 0.7117 MS-ON history (yes/no) 37.19 (8.42) <0.0001 The OCTiMS study was designed to evaluate change in Goodness of fit: adjusted R-squared = 0.161 pRNFL thickness in patients with RRMS followed for up to 3 years, compared with a group of reference partici- The most affected eye was used for analyses. pRNFL, peripapillary retinal nerve fiber layer; MS-ON, multiple pants (without neurologic or ophthalmic disease), to sclerosis-associated optic neuritis; SE, standard error. determine whether retinal OCT measurements are suffi- ciently accurate, reproducible, and sensitive to measure Visual function disease progression over time. In this study, the biggest multicenter OCT MS study performed to date, we Changes from baseline in mean visual acuity were detected a greater decrease in pRNFL thickness in patients observed in both groups (MS and reference), and for high with MS compared with reference participants over a 36- contrast and low contrast visual acuity at all time-points. month period. The accuracy of repeated measurements These changes were small, ranging from 1.6 to 1.1 let- was excellent, indicated by an ICC greater than 0.98. ters in the RRMS group and from 1.5 to 0.0 letters in The mean (SD) average pRNFL thickness decreased the reference group. from baseline by 1.61 (4.600) µm at Month 36 in patients A significant linear correlation (Pearson’s correlation) with MS compared with only 0.09 (2.982) µm in refer- was observed between change from baseline in low con- ence control participants, representing an effect size of trast visual acuity (total number of letters recognized at 0.39 (Cohen’s d). The mean change from baseline for the low contrast level) and change from baseline in GCIPL MS group after adjusting for age-related changes in thickness at month 36 (p < 0.01). No such correlation healthy subjects were 1.86 (95% CI 2.54, 1.17) lm was observed for high contrast visual acuity (total num- over the 3-year period. The difference in change from ber of letters recognized at high contrast level). There baseline between the two groups for average pRNFL was no linear relationship observed between the change thickness was statistically significant from Month 18 from baseline in average pRNFL thickness and the change onwards. The changes observed in pRNFL thickness from baseline in high or low contrast visual acuity. between patients with MS and healthy controls are consis- tent with numerous clinical studies and meta- Symbol digit modalities test analyses.16,27–29 Separate analyses performed on the tem- poral quadrant and the papillomacular bundle also The mean (SD) SDMT total score increased in patients demonstrated significant differences between groups at with MS from 55.7 (13.2) at baseline to 58.2 (14.22) at Month 36, consistent with reports that the temporal Month 36. There appears to be no visible linear correla- quadrant is the most involved pRNFL region in MS.30,31 tion as observed graphically between the change from However, the statistical significance of the differences baseline in SDMT total score and average pRNFL thick- were less marked than for the pRNFL. ness (Pearson correlation coefficient = 0.097634). To accurately assess the change in pRNFL thickness during the OCTiMS study, it was important to control Patient-reported indices for multiple for the presence of MS-ON, a key confounding factor in sclerosis absolute pRNFL thickness values.28,29 Indeed, the average degree of pRNFL atrophy following MS-ON is There was no change from baseline in PRIMuS QoL total 20.38 µm, representing about one fifth of normal score observed in patients with MS at Month 24 and 36. pRNFL thickness. In eyes without clinically manifested There was no visible linear correlation as observed graphi- MS-ON, the level of atrophy is about one third of this cally between the change from baseline in PRIMuS QoL (~7 µm).29 In our study, pRNFL thickness decreased sim- total score or PRIMuS activities total score and change in ilarly in patients with MS regardless of MS-ON history, average pRNFL thickness (Pearson correlation coefficients indicating that the rate of further pRNFL thinning after 0.01668 and 0.0074, respectively). the initial optic neuritis-related thinning has occurred is 2246 ª 2021 Novartis Pharma AG. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. F. Paul et al. OCTiMS Study no different than eyes of patients with RRMS without a only reflect axonal degeneration, but also neuronal degen- history of optic neuritis. However, we cannot exclude a eration (as measured by mGCIPL change), and both history of subclinical MS-ON in the absence of confound- appear to decrease over time. Similarly, another recent ing ophthalmological pathology in apparent non-MS-ON study reported mGCIPL thinning at a very early stage, patients. Moreover, in light of recent work on inter-eye followed by a plateau.40 The data suggest that the difference thresholds for detecting unilateral MS-ON, a mGCIPL complex may be capable of showing thickness different assignment strategy of patients to MS-ON and changes at earlier time points than the pRNFL layer con- non-MS-ON subgroups might have changed diagnostic taining primarily axons. groupings.32,33 In this study we adhered to the 2005 revi- A secondary endpoint of OCTiMS was to evaluate the sion of the MS diagnostic criteria.21 Interestingly, the short-term reproducibility of the pRNFL thickness mea- pRNFL changes we observed are similar to those mea- sure at study start by test/re-test estimation after a 4-week sured by the older time domain OCT technology used in interval in patients with MS and control participants. The a longitudinal multicenter US study.34 As in OCTiMS, purpose of this analysis was to determine if adjustments this work showed similar patterns of longitudinal pRNFL in the assumptions and sample size were required if the thinning in non-MS-ON and MS-ON eyes. However, it measurement variability was larger than expected. Repro- should be noted that in OCTiMS, due to differences in ducibility for the average pRNFL thickness was excellent standard deviations the effect size in the subset of patients at a group level, with ICC values greater than 0.98. The without MS-ON (0.5 [Cohen’s d]) is numerically higher minute average change in pRNFL from baseline to Month than that of MS-ON patients (0.26 [Cohen’s d]). 1 was 0.12 µm for MS patient eyes and 0.35 µm for the Abnormal thinning of the mGCIPL layer has also been reference eyes. Overall, the difference was not statistically associated with MS.28 The mGCIPL analysis presented significant and remains well within the recognized degree here supports the primary endpoint. Results show a of physiological variation.36 Our results are in line with decrease in mean (SD) mGCIPL thickness of 0.39 (3.096) prior evidence that demonstrate the high reliability of µm in patients with MS over 36 months compared with repeated scans utilizing eye tracking, Automatic Real- an increase of 1.02 (5.063) µm in reference participants, Time (ART) averaging, auto-registration to a baseline with a rate of thinning corrected with healthy age-related scan during subsequent scan acquisitions, and automated attrition of 2.03 (95% CI 2.78, 1.28) µm over the 3- segmentation with manual correction of macular OCT year period. The increase in mGCIPL observed in the ref- scans, thereby supporting their use in multicenter set- erence group reflects the degree of physiological variation tings.41 also observed with pRNFL (Fig. 2) and consistent with Other secondary and exploratory objectives were to earlier observations.35,36 An additional factor may be that identify a potential relationship between the extent of the dendritic layer of GCIPL may be dynamic and chang- pRNFL thinning and different measures of neuronal dam- ing more over time in normal eyes compared to eyes age and disease severity. No clear associations were where damage is present. observed between change in pRNFL thickness and change Interestingly, we observed greater thinning of both in visual function, EDSS, brain volume, cognitive perfor- pRNFL and mGCIPL in patients with a shorter disease mance, or measures of QoL. By contrast, change in duration, which aligns with previous smaller studies.37–39 mGCIPL did significantly correlate (p < 0.01) with low This observation is in line with the positive correlation contrast visual acuity, a sensitive measure of macular between baseline pRNFL thickness and the rate of pRNFL function. In addition, patients with more new T2/PD thinning. It may be due to a larger dynamic range from lesions showed greater RNFL thinning at Month 36 which the retinal layer can thin, or to faster pRNFL loss (p < 0.01), aligning with previous work that suggest an in earlier disease, perhaps as a consequence of axonal loss interrelation of inflammatory activity and retinal tissue driven by more active inflammation. In the subgroup of loss.42 There was no observed correlation between EDSS patients with earlier disease, the effect size for pRNFL is worsening and change in average RNFL thickness, con- numerically higher compared with the overall MS popula- trary to the findings of previous studies.13,39 Although tion (0.43 vs. 0.39). The largest effect size is observed in recent evidence suggests that EDSS may correlate better non-MS-ON patients with ≤3 years of MS duration with mGCIPL thickness than with pRNFL thickness,40,43 (0.75). This finding suggests that OCT measures may be this was not the case in our study either. However, it is more useful for documenting neuroaxonal loss in early important to note that the majority of the RRMS popula- disease, especially as MRI measures of brain atrophy are tion in this study had well-controlled disease, with only less sensitive at this stage. 7% of patients progressing during the 3 years, and a In the present study, we confirm that neurodegenera- mean (SD) EDSS worsening at Month 36 of 0.15 (0.86), tion occurring in early stage of disease onset may not which may account for the failure to detect any ª 2021 Novartis Pharma AG. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. 2247 OCTiMS Study F. Paul et al. correlation between EDSS progression and pRNFL/ of incorporating OCT measures in RRMS clinical trials in mGCIPL thinning. the future. A decrease of 1.3% and 0.5% from baseline in brain volume was observed at Month 36 in patients with MS Conclusion and control subjects, respectively, representing an effect size of 0.76 (Cohen’s d). While baseline brain volume The OCTiMS study prospectively assessed pRNFL thick- was associated with baseline pRNFL thickness, there was ness change using OCT in patients with MS in a multi- no clear association between the extent of brain volume center setting. The OCT measurements were shown to be loss and the change in pRNFL thickness from baseline. highly reproducible, and a significant difference in pRNFL However, we did find that decreases in average mGCIPL thinning was observed between patients with MS from thickness were correlated with more brain atrophy over 18 months onwards compared with healthy controls. This the 3-year study. This finding is in line with results was supported by similar results obtained for mGCIPL from a 4-year study by Saidha et al.,12 suggesting that thickness, of which changes were found to correlate with mGCIPL atrophy may mirror the severity of disease brain volume loss. Based on Heidelberg Spectralis data activity and whole-brain atrophy. In their study, the cor- acquired in this study, the effect size for the OCT mea- relation between mGCIPL and brain atrophy in patients sures (RNFL, GCIPL) was between small to medium in with RRMS was stronger when MS-ON effects were this overall stable RRMS population, and smaller than excluded. that found for MRI brain volume. Of note, the effect size It should be noted that Caucasians made up 94% of for OCT measures was larger in patients in the early the study population, which is a limitation given that stages of the disease course and without any history of RNFL thickness and rates of thinning are known to differ MS-ON. Considering the rapid advances in OCT imaging between ethnic groups.44,45 Another limitation is that only toward higher resolution, OCT measures are getting more the affected eye of patients with unilateral MS-ON was and more reliable, increasing their potential to become a studied, which, in this subgroup, may increase the risk of convenient tool to monitor disease progression, and a floor-effect which could lead to underestimating the quantify the level of neurodegeneration in MS popula- level of neurodegeneration. Furthermore, in spite of con- tions, particularly in the early disease course. siderable collaborative network efforts of the neuro- ophthalmological community, a multi-rater, validated Acknowledgments quality control metric for OCT has not yet been estab- lished and implemented in multicenter studies.19,22 Like- The authors wish to thank Carolyn M. Ervin for her substan- wise, reproducibility of inner nuclear layer thickness of tial contribution in the data analyses, as well as Mark Kirby, bipolar cells, a promising biomarker for inflammation in Aisling Towell, and Marie-Catherine Mousseau (Novartis MS, has not yet been optimized and applied.46,47 Ireland Ltd.) for their writing support, funded by Novartis A strength of this study is that there were substantial Pharma AG, Basel, Switzerland. Frederik Barkhof is sup- efforts made in to ensure that identical machines and ported by the NIHR biomedical research center at UCLH. software are used across sites for optimal comparability of Sincere gratitude also to Alina Czechner, the senior reader at OCT measures, which may not be possible in all multi- the Vienna Reading Center, who has supervised most of the center studies. However, mGCIPL change measured in manual OCT analysis, to Gordon Francis and G€ oril Karlsson OCTiMS may not be very accurate, as shown in patients (Novartis Pharma AG) who contributed to the design and with MS > 5 years who had no mGCIPL change (Fig. 3), initiation of the study and to Annik Laflamme (Novartis likely due to the limitations of the OCT imaging tech- Pharma AG, Basel at time of study) who assisted in the ini- nique used and less frequent sampling. It should be noted tial development of this publication. The authors also thank that since the completion of this work the performance patients, investigators, and site staff who participated in and sensitivity of OCT imaging have increased substan- these studies, including all OCTiMS Principal Investigators tially. At the time of the protocol initiation, the standard (Table S2). In particular, the authors would like to extend recommended macular acquisition on the Heidelberg their special thanks to Dr Pablo Villoslada for his ongoing Spectralis was 25 B scans. The current Heidelberg seg- commitment and fundamental contribution towards the mentation software requires at least 61 B scan density, completion of this work. highlighting the technical advances that are rapidly occur- ring in OCT acquisition and post-acquisition analyses. Conflicts of Interest This will likely improve the accuracy of these measure- ments, as has happened in the MRI arena over the last Friedemann Paul is a member of the Novartis OCTIMS three decades, and further supports the potential benefits study steering committee and the MedImmune/Viela Bio 2248 ª 2021 Novartis Pharma AG. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. F. Paul et al. OCTiMS Study steering committee; has received research support and per- neurology and has served on scientific advisory boards sonal compensation for activities with Alexion, Chugai, for Biogen, Genzyme, Genentech Corporation, EMD Ser- Biogen, Bayer, Merck Serono, Teva, Genzyme, Novartis, ono & Celgene. He is the PI of investigator-initiated Shire, Roche, Actelion, Celgene, and MedImmune. studies funded by Genentech Corporation and Biogen Peter A. Calabresi has received research funding from Idec, and received support from the Race to Erase MS Biogen, Genentech and Annexon; consulting honoraria foundation. He has received equity compensation for from Biogen, Disarm Therapeutics, and Nervgen. consulting from JuneBrain LLC, a retinal imaging device Frederik Barkhof has received grants from National developer. Institute for Health, Research University College London Bianca Gerendas reports consulting fees from Roche Hospitals, Biomedical Research Centre. He reports other and Novartis Ophthalmology (not relevant to the work from Neurology, other from Brain, other from Radiology, presented herein) and receives funding from DXS. other from MSJ, other from Neuroradiology, personal fees Ursula Schmidt-Erfurth has received personal fees from from Springer, personal fees and other from Bayer, personal Bayer, Novartis, Genentech, and Roche. fees from Biogen, personal fees from Roche, grants from Axel Petzold is part of the steering committee of the TEVA, grants from Merck, grants from Biogen, personal ANGI network which is sponsored by ZEISS, steering fees from IXICO Ltd, personal fees from Novartis, personal committee of the OCTiMS study which is sponsored by fees and other from GeNeuro, grants from IMI-EU, grants Novartis and reports speaker fees from Heidelberg- from GE Healthcare, grants from UK MS Society, grants Engineering. from Dutch Foundation MS Research, grants from NWO, Catherine Agoropoulou, Ying Zhang, and Gustavo Sei- grants from NIHR, outside the submitted work. fer are employees of Novartis. Ari J. Green reports grants and other support from Inception Biosciences; grants from the National Multiple References Sclerosis Society and from the US National Institutes of 1. Reich DS, Lucchinetti CF, Calabresi PA. Multiple sclerosis. Health; other support from MedImmune, Mylan, Sandoz, N Engl J Med. 2018;378(2):169-180. Dr Reddy, Amneal, Momenta, Synthon, and JAMA Neu- 2. Tur C, Moccia M, Barkhof F, et al. Assessing treatment rology, outside the submitted work; and has received outcomes in multiple sclerosis trials and in the clinical grant support from Novartis for participating in the setting. Nat Rev Neurol. 2018;14(2):75-93. OCTIMS study. 3. Sastre-Garriga J, Pareto D, Battaglini M, et al. MAGNIMS Randy Kardon served on the OCTIMS Steering Com- consensus recommendations on the use of brain and mittee and receives honoraria from Novartis for this spinal cord atrophy measures in clinical practice. Nat Rev activity. RK reports other support from MedFace LLC Neurol. 2020;16(3):171-182. and FaceX LLC and has a patent to use pupil and eye 4. Parisi V, Manni G, Spadaro M, et al. Correlation between movement recordings to diagnose eye and CNS disorders, morphological and functional retinal impairment in such as traumatic brain injury. RK has received an unre- multiple sclerosis patients. Invest Ophthalmol Vis Sci. stricted grant from Heidelberg Engineering in support of 1999;40(11):2520-2527. development of a Neuro-Toolbox analysis suite for optic 5. Gordon-Lipkin E, Chodkowski B, Reich DS, et al. Retinal nerve disorders using OCT. nerve fiber layer is associated with brain atrophy in Jaume Sastre-Garriga has received speaking or consult- multiple sclerosis. Neurology. 2007;69(16):1603-1609. ing honoraria from Merck, Biogen, Genzyme, Celgene, 6. Frau J, Fenu G, Signori A, et al. A cross-sectional and and Novartis. longitudinal study evaluating brain volumes, RNFL, and Sven Schippling received fees for speaking or consulting cognitive functions in MS patients and healthy controls. from Bayer Healthcare, Biogen, Genzyme/Sanofi, Merck, BMC Neurol. 2018;18(1):67. Novartis, Roche and Teva; research grants from Gen- 7. Kuchling J, Paul F. Visualizing the central nervous system: zyme/Sanofi, Novartis, and Swiss Multiple Sclerosis Soci- imaging tools for multiple sclerosis and neuromyelitis ety; and he is now an employee of Roche, which was not optica spectrum disorders. Frontiers in Neurology. involved in the study. 2020;11:450. Patrick Vermersch received honoraria and consulting 8. Frohman EM, Balcer LJ, Calabresi PA. Multiple sclerosis: fees from Biogen, Sanofi-Genzyme, Novartis, Teva, Merck, can retinal imaging accurately detect optic neuritis? Nat Roche, Imcyse, AB Science, and Celgene. He also received Rev Neurol. 2010;6(3):125-126. research supports from Novartis, Sanofi-Genzyme, and 9. Siger M, Dziez gielewski K, Jasek L, et al. Optical coherence Roche. tomography in multiple sclerosis: thickness of the retinal Shiv Saidha has received consulting fees from Medical nerve fiber layer as a potential measure of axonal loss and Logix for the development of CME programs in brain atrophy. J Neurol. 2008;255(10):1555-1560. ª 2021 Novartis Pharma AG. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. 2249 OCTiMS Study F. Paul et al. 10. D€orr J, Wernecke KD, Bock M, et al. Association of Assessment for Multiple Sclerosis (BICAMS). Mult Scler. retinal and macular damage with brain atrophy in 2012;18(6):891-898. multiple sclerosis. PLoS One. 2011;6(4):e18132. 25. Benedict RH, DeLuca J, Phillips G, LaRocca N, Hudson 11. Saidha S, Sotirchos ES, Oh J, et al. Relationships between LD, Rudick R. Validity of the Symbol Digit Modalities retinal axonal and neuronal measures and global central Test as a cognition performance outcome measure for nervous system pathology in multiple sclerosis. JAMA multiple sclerosis. Mult Scler. 2017;23(5):721-733. Neurol. 2013;70(1):34-43. 26. Doward LC, McKenna SP, Meads DM, Twiss J, Eckert BJ. 12. Saidha S, Al-Louzi O, Ratchford JN, et al. Optical The development of patient-reported outcome indices for coherence tomography reflects brain atrophy in multiple multiple sclerosis (PRIMUS). Mult Scler. 2009;15(9):1092- sclerosis: a four-year study. Ann Neurol. 2015;78(5):801- 1102. 813. 27. Khanifar AA, Parlitsis GJ, Ehrlich JR, et al. Retinal nerve 13. Martinez-Lapiscina EH, Arnow S, Wilson JA, et al. Retinal fiber layer evaluation in multiple sclerosis with spectral thickness measured with optical coherence tomography domain optical coherence tomography. Clin Ophthalmol. and risk of disability worsening in multiple sclerosis: a 2010;20(4):1007-1013. cohort study. Lancet Neurol. 2016;15(6):574-584. 28. Petzold A, de Boer JF, Schippling S, et al. Optical 14. Shi CE, Jiang H, Gameiro GR, et al. Visual function and coherence tomography in multiple sclerosis: a systematic disability are associated with focal thickness reduction of review and meta-analysis. Lancet Neurol. 2010;9(9):921- the ganglion cell-inner plexiform layer in patients with 932. multiple sclerosis. Invest Ophthalmol Vis Sci. 2019;60 29. Petzold A, Balcer LJ, Calabresi PA, et al. Retinal (4):1213-1223. layer segmentation in multiple sclerosis: a systematic 15. Vidal-Jordana A, Pareto D, Cabello S, et al. Optical review and meta-analysis. Lancet Neurol. 2017;16 coherence tomography measures correlate with brain and (10):797-812. spinal cord atrophy and multiple sclerosis disease-related 30. Bock M, Brandt AU, D€ orr J, et al. Patterns of retinal nerve disability. Eur J Neurol. 2020;27(11):2225-2232. fiber layer loss in multiple sclerosis patients with or 16. Oertel FC, Zimmermann HG, Brandt AU, Paul F. Novel without optic neuritis and glaucoma patients. Clin Neurol uses of retinal imaging with optical coherence tomography Neurosurg. 2010;112(8):647-652. in multiple sclerosis. Expert Rev Neurother. 2019;19(1):31- 31. Lamirel C, Newman NJ, Biousse V. Optical coherence 43. tomography (OCT) in optic neuritis and multiple 17. You Y, Graham EC, Shen T, et al. Progressive inner sclerosis. Rev Neurol. 2010;166(12):978-986. nuclear layer dysfunction in non-optic neuritis eyes in MS. 32. Nolan-Kenney RC, Liu M, Akhand O, et al. Optimal Neurol Neuroimmunol Neuroinflamm. 2018;5(1):e427. intereye difference thresholds by optical coherence 18. Lange AP, Sadjadi R, Costello F, Guber I, Traboulsee AL. tomography in multiple sclerosis: an international study. Reproducibility of retinal nerve fiber layer measurements Ann Neurol. 2019;85(5):618-629. with manual and automated centration in healthy subjects 33. Xu SC, Kardon RH, Leavitt JA, Flanagan EP, Pittock SJ, using spectralis spectral-domain optical coherence Chen JJ. Optical coherence tomography is highly sensitive tomography. ISRN Ophthalmol. 2012;2012:860819. in detecting prior optic neuritis. Neurology. 2019;92(6): 19. Tewarie P, Balk L, Costello F, et al. The OSCAR-IB e527-e535. consensus criteria for retinal OCT quality assessment. 34. Talman LS, Bisker ER, Sackel DJ, et al. Longitudinal study PLoS One. 2012;7(4):e34823. of vision and retinal nerve fiber layer thickness in multiple 20. Schippling S, Balk LJ, Costello F, et al. Quality control for sclerosis. Ann Neurol. 2010;67(6):749-760. retinal OCT in multiple sclerosis: validation of the 35. Balk LJ, Oberwahrenbrock T, Uitdehaag BM, Petzold A. OSCAR-IB criteria. Mult Scler. 2015;21(2):163-170. Physiological variation of retinal layer thickness is not 21. Polman CH, Reingold SC, Edan G, et al. Diagnostic caused by hydration: a randomised trial. J Neurol Sci. criteria for multiple sclerosis: 2005 revisions to the 2014;344(1-2):88-93. "McDonald Criteria". Ann Neurol. 2005;58(6):840-846. 36. Balk LJ, Sonder JM, Strijbis EM, et al. The physiological 22. Cruz-Herranz A, Balk LJ, Oberwahrenbrock T, et al. The variation of the retinal nerve fiber layer thickn1ess and APOSTEL recommendations for reporting quantitative macular volume in humans as assessed by spectral optical coherence tomography studies. Neurology. 2016;86 domain-optical coherence tomography. Invest Ophthalmol (24):2303-2309. Vis Sci. 2012;53(3):1251-1257. 23. Kurtzke JF. Rating neurologic impairment in multiple 37. Balk LJ, Cruz-Herranz A, Albrecht P, et al. Timing of sclerosis: an expanded disability status scale (EDSS). retinal neuronal and axonal loss in MS: a longitudinal Neurology. 1983;33(11):1444-1452. OCT study. J Neurol. 2016;263(7):1323-1331. 24. Langdon DW, Amato MP, Boringa J, et al. 38. Kuhlmann T, Lingfeld G, Bitsch A, Schuchardt J, Br€ uck Recommendations for a Brief International Cognitive W. Acute axonal damage in multiple sclerosis is most 2250 ª 2021 Novartis Pharma AG. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. F. Paul et al. OCTiMS Study extensive in early disease stages and decreases over time. 45. Kimbrough DJ, Sotirchos ES, Wilson JA, et al. Brain. 2002;125(Pt 10):2202-2212. Retinal damage and vision loss in African American 39. Cilingir V, Batur M. Axonal degeneration independent of multiple sclerosis patients. Ann Neurol. 2015;77 inflammatory activity: is it more intense in the early stages (2):228-236. of relapsing-remitting multiple sclerosis disease? Eur 46. Saidha S, Sotirchos ES, Ibrahim MA, et al. Microcystic Neurol. 2020;83(5):508-516. macular oedema, thickness of the inner nuclear layer of 40. Pietroboni AM, Carandini T, Dell’Arti L, et al. Evidence of the retina, and disease characteristics in multiple sclerosis: retinal anterograde neurodegeneration in the very early a retrospective study. Lancet Neurol. 2012;11(11):963-972. stages of multiple sclerosis: a longitudinal OCT study. 47. Gelfand JM, Cree BA, Nolan R, Arnow S, Green AJ. Neurol Sci. 2020;41(11):3175-3183. Microcystic inner nuclear layer abnormalities and 41. Oberwahrenbrock T, Traber GL, Lukas S, et al. neuromyelitis optica. JAMA Neurol. 2013;70(5):629-633. Multicenter reliability of semiautomatic retinal layer segmentation using OCT. Neurol Neuroimmunol Neuroinflamm. 2018;5(3):e449. Supporting Information 42. Ratchford JN, Saidha S, Sotirchos ES, et al. Active MS is Additional supporting information may be found online associated with accelerated retinal ganglion cell/inner in the Supporting Information section at the end of the plexiform layer thinning. Neurology. 2013;80(1):47-54. article. 43. Saidha S, Syc SB, Durbin MK, et al. Visual dysfunction in multiple sclerosis correlates better with optical coherence Table S1. List of exclusion criteria for participants in the tomography derived estimates of macular ganglion cell MS group and reference group (when applicable). layer thickness than peripapillary retinal nerve fiber layer Data S1. Optical coherence tomography (OCT). thickness. Mult Scler. 2011;17(12):1449-1463. Figure S1. Different scan patterns used in the OCTiMS 44. Samarawickrama C, Wang JJ, Huynh SC, et al. Ethnic study for OCT imaging and respective evaluation areas differences in optic nerve head and retinal nerve fibre for each scan. layer thickness parameters in children. Br J Ophthalmol. Table S2. OCTiMS Principal Investigators (ordered alpha- 2010;94(7):871-876. betically by countries). ª 2021 Novartis Pharma AG. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. 2251