Patterning Chronic Active Demyelination in Slowly Expanding/Evolving White Matter MS Lesions

Trial Design, Patients, and MR Imaging Procedures

SELs¹² and non-SELs were determined in chronic white matter lesions of the pooled population (placebo and treatment groups) of the Study to Assess the Efficacy, Safety, Tolerability, and Pharmacokinetics of BIIB033 in Participants With Relapsing Forms of Multiple Sclerosis When Used Concurrently With Avonex (SYNERGY) trial (NCT01864148), a multicenter, randomized, double-blind, placebo-controlled, dose-ranging, parallel-group, Phase 2 study. Patients were randomly allocated in a 1:2:2:2:2 ratio to 1 of 5 parallel treatment groups of opicinumab (3, 10, 30, or 100 mg/kg) or placebo, once every 4 weeks for 72 weeks. Opicinumab is a human monoclonal antibody against LINGO-1, an inhibitor of oligodendrocyte differentiation and axonal regeneration.²² All patients self-administered intramuscular interferon β-1a as a background anti-inflammatory treatment once a week, and approximately half of the population had not previously received MS disease-modifying therapies.²² SYNERGY study details have been reported previously.²² Eligible patients (18–58 years of age) had an Expanded Disability Status Scale score of 2–6 and relapsing MS, including RRMS and SPMS with relapses. Evidence of clinical or neuroimaging disease activity was required within 12 months before enrollment.²²

Axial T1-weighted (3D spoiled gradient-echo, TR = 28–30 ms, TE = 511 ms, flip angle = 27°–30°, resolution = 1 × 1 × 3 mm), axial T2-weighted (2D fast spin-echo, TR = 4500–6200 ms, TE = 66–91 ms, resolution = 1 × 1 × 3 mm), axial MTR (2 consecutive 3D spoiled gradient-echo, TR = 32–62 ms, TE = 5–11 ms, flip angle = 10°–15°, resolution = 1 × 1 × 3 mm, with and without magnetization transfer pulse), and axial DTI (2D spin-echo echo-planar imaging sequence with a diffusion gradient, TR = 9800–16,000 ms, TE = 90–132 ms, b-values = 0 and 1000, 25–32 diffusion directions, resolution = 2.5 × 2.5 × 2.5 mm) were acquired at baseline and weeks 4, 8, 12, 16, 20, 24, 48, and 72.²² Complete methods for brain MR imaging acquisitions are described in the On-line Appendix. The SEL analysis population (n = 299) represents the subset of the intention-to-treat population (n = 418) that had available T1-weighted and T2-weighted MR images at all aforementioned time points from baseline to week 72.

Identification of SELs

As previously described, SELs are detected as areas of T2 lesions, pre-existing at baseline, that show constant and concentric local expansion.¹² Before SEL detection, T2 lesions were identified in baseline scans using a semiautomated method,²³ in which a fully automated segmentation was subsequently manually reviewed and corrected by trained MR imaging readers. Identification of SELs is a 2-stage process. First, SEL candidates are identified as contiguous areas of pre-existing, nonenhancing T2 lesions that are ≥10 voxels in size and show local expansion from baseline to week 72; a minimum local expansion of 4% per year is used as a cutoff for determining SEL candidate boundaries, as in previous SEL analyses.¹² Local expansion is determined from the Jacobian determinant of the nonlinear deformation between the baseline and week 72 scans. Computation of the Jacobian is based on the pipeline proposed by Nakamura et al,²⁴ in which nonlinear registration is performed using the symmetric image normalization method,²⁵ and both T1-weighted and T2-weighted images are used for registration. The second stage of SEL detection scores each SEL candidate in turn, on the basis of the concentricity and constancy of expansion across time. Considering local expansion at all intermediate scans (weeks 4, 8, 12, 16, 20, 24, and 48) allows the identification of SEL candidates undergoing constant and gradual expansion across time; measuring concentricity allows identification of SEL candidates exhibiting inside-out radial expansion. Results pertaining to SEL analyses are presented for high-confidence SELs (with a heuristic score of ≥0).¹² Non-SELs are defined as complement regions from pre-existing, baseline, nonenhancing T2 lesions devoid of any SEL detection (irrespective of the heuristic score). SEL identification and all T1-weighted measures related to SELs and non-SELs were performed by NeuroRx Research staff, who remained blinded to all study patient-level information.

Normalization of T1-Weighted and MTR Signal Intensity

Before measuring T1-intensity change across time, T1WIs were normalized in a 2-stage process: 1) Least-trimmed squares normalized all serial T1WIs of a given patient to the baseline T1-weighted scan, and 2) T1WIs for a given subject were linearly normalized by mapping the median gray matter T1 intensity at baseline to a value of zero and mapping the median normal-appearing white matter intensity at baseline to 1. Least-trimmed squares performs linear regression between coregistered sequential scans using the 50% of voxels whose least-squares fit has the smallest sum of squared residuals.²⁶ This process normalizes intensities within a given subject on the basis of only the subset of voxels that remain relatively unchanged with time. The first stage of normalization minimizes acquisition-related intensity variation across time for a given subject, whereas the second stage provides comparable measures of T1 intensity change across different subjects.

MTR intensities were calibrated by determining the median MTR values for both gray and white matter in a healthy control subject specific to each scanner. For each new-subject scan acquired in the same scanner, the MTR value corresponding to the median healthy control gray matter was mapped to zero, while the MTR value corresponding to the median healthy control white matter was mapped to 1. In normalized MTR images, a value of zero can thus be interpreted as corresponding to healthy (ie, non-MS) gray matter, while a value of 1 can be interpreted as corresponding to healthy white matter. DTI-RD is expressed in units of 10⁻³ mm²/s and does not require normalization across scans.

Lesion-Level Visualization of Longitudinal Tissue Damage within a SEL Example

To display an example of lesion-level longitudinal tissue damage within discrete SELs, we modeled smooth voxel-based (linear fit) representations of normalized T1 (nT1), normalized MTR (nMTR), and DTI-RD intensity change across time in a high-confidence SEL. Heat map synthetic representations were produced that may represent biologic change and/or displacement. Linear models interpolating intensity change with time were used.

Statistical Analysis

The statistical analysis of SEL data was exploratory and included all patients from SYNERGY with no missing or nonevaluable T1-weighted and T2-weighted scans at any time point (baseline to week 72; SEL analysis population). No imputation of missing data was performed.

Continuous variables measuring tissue integrity (eg, MTR and DTI change from baseline in nMTR and DTI-RD) were compared between SELs and non-SELs using a Wilcoxon signed rank test, accounting for within-subject correlation of each variable. Change from baseline comparisons for continuous variables between patients with RRMS and SPMS were based on rank regression adjusted for covariates including baseline value, age, sex, and the baseline T2 volume category based on tertiles. Comparisons of baseline continuous variables between patients with RRMS and SPMS were based on rank regression, with MS type and covariates including age, sex, and baseline T2 volume category based on tertiles.

The Spearman rank correlation analysis was used to evaluate the association between change with time in continuous variables within SELs and non-SELs. Two-sided statistical tests were conducted at the 5% significance level without adjustment for multiplicity.