A regional consensus recommendation on brain atrophy as an outcome measure in multiple sclerosis

MS causes focal and diffuse damage to the brain. The focal white matter lesions are the classic hallmark of MS and appear on MRI as T₂, gadolinium enhanced T₁ lesions, or T₁-hypointense lesions (black holes) [10]. The damage also occurs in grey matter as well as diffusely in normal appearing white matter and consequently lead to BVL [11, 12]. Among MRI measures, BVL in MS is considered as one of the prognostic measures which has been directly correlated with cognitive impairment [13, 14], disability progression [15‐17], and fatigue [18]. A number of longitudinal studies showed that BVL predicts future worsening of disability and cognition. A 10-year follow-up study demonstrated that brain atrophy occurred throughout the course of the disease and was more severe in the group that showed disability progression at 5 years of follow-up. Furthermore, it was demonstrated that the overall grey matter atrophy was a better predictor of disease progression than white matter atrophy during the same follow-up period [19]. In a 13-year follow-up of 75 MS patients, Fillipi et al. showed that grey matter damage was associated with significant worsening of disability and cognitive function in 66% and 34% of patients, respectively [20]. In a study of 261 MS patients, brain atrophy and lesion load were complimentary predictors of long term disability over a 10-year follow-up period [21]. BVL has been observed at the early phases of the disease in patients with clinically isolated syndrome (CIS) [5].

With respect to age, it has been largely known that after adolescence, grey matter volume starts to decline in a linear way (−0.09%/year) while white matter decline starts after midlife, and at a higher rate (−0.2%/year) [22, 23]. Regarding cortical thickness, at midlife a global thinning was apparent, spreading over several cortical regions [24]. The decline in the thickness measured was around 0.016 mm/decade. Both volumetric and cortical thickness age-associated changes had been described, implying that age is a variable that should be considered when assessing global or regional brain volume in MS patients.

There were few reported studies exploring potential gender-associated differences and brain size. Men had larger grey and white matter volumes compared to women although after adjusting by total intracranial volume, these differences disappeared [22]. Regarding cortical thickness, men showed slightly larger cortical thickness compared to females only at midlife, and the rate of progressive thinning was similar for both groups [24]. A recent study reported that cortical thickness was associated with total intracranial volume, but not with gender [25]. Since gender and total intracranial volume are highly correlated, it is a matter of debate whether one or both variables should be considered when assessing global or regional brain volume in MS patients [26, 27].

The participants also identified several unmet needs regarding brain volume measurements and its implementation in the routine clinical practice (Table 1).

Conventional MRI techniques are considered the gold standard for diagnosing and monitoring the response to treatment in patients with MS. However, conventional MRI has limited specificity and correlation with disability measures [28, 29]. In addition, conventional MRI does not offer any information about the ongoing degenerative/reparative processes. Therefore, the panel discussed the current MRI methods and software applications that help in capturing and measuring the underlying MS pathology with high degree of specificity. Currently, segmentation- and registration-MRI based methods are used to measure brain volume. The segmentation-based method such as brain parenchymal fraction (BPF), white matter fraction (WMf), grey matter fraction (GMf), and normalized brain volume (NBV) provide assessment of global (BPF, NBV) or regional (WMf, GMf) brain volume at a single time-point in an automated way [28‐31]. However, the data extracted out of this method is usually heterogeneous and influenced by the quality of T1-weigthed images acquired, and are not recommended for longitudinal analysis. Registration-based methods measure brain volume at two time points, in order to calculate the percentage brain volume change (PBVC). These methods are robust, sensitive to changes over time, less influenced by the disruption of the quality of MR imaging acquisitions, and are most suitable for evaluating global brain volume changes but are not usually designed to analyse regional volume changes over time [32]. The sensitivity and the specificity could vary according to the cut-off value that identify the annualized PBVC. With a cut-off value of 0.4%, the sensitivity and specificity are 65% and 80%, respectively [32]. The panel listed the advantages and disadvantages for brain volume assessment in patients with MS (see Table 2.)

It is important to target the neurodegenerative process in MS patients, in addition to the inflammation component. In general, the results of brain volume data vary from one study to another, which could be attributed to the different mechanism of action of DMTs, the ability of these drugs to cross the blood brain barrier and the heterogeneity of the patient population. In phase III clinical trials, several DMTs demonstrated reduction in the rate of brain volume loss in relapsing-remitting multiple sclerosis (RRMS) patients at variable levels with different response time (Table 3.) [6‐8, 33‐49]. However, one should be careful in interpreting data across trials with different patient population, using different methods in measuring brain volume. In both placebo-controlled as well as active comparator trials, most of the DMTs provided a delayed effect around the second year of treatment with the exception of daclizumab and fingolimod [6‐8, 33‐54]. In the FREEDOMS trials, fingolimod demonstrated an effect on brain volume within 6 months of treatment [6, 8]. The first reported brain atrophy study with intramuscular (IM) interferon (IFN)-β-1a demonstrated lower rate of BVL than placebo in the second year of treatment of RRMS patients (−0.23% IFN-β-1a compared to −0.51% in the placebo group; p = 0.03) [33]. However, the subcutaneous (sc) IFN-β-1a produced conflicting results in both CIS and RRMS patients [33, 34, 50, 51]. Data from placebo controlled trials on sc IFN-β-1b and BVL in RRMS patients are not published to date [52, 53]. Glatiramer significantly reduced BVL by 40% at 9–18 months and 25% at 0–18 months compared to placebo [35]. Three trials compared glatiramer and IFN-β and showed a marginal reduction in BVL in one study [36‐38]. In two trials, natalizumab increased the rate of BVL in the first year and then significantly reduced it versus placebo in the second year [40, 41]. Teriflunomide failed to demonstrate reduction in the rate of BVL compared to placebo [42]. Dimethyl fumarate produced a 21% reduction in BVL compared to placebo in the DEFINE study and failed to show any statistically significant reduction in BVL in the CONFIRM study [43, 44]. In the CARE-MS I and II, alemtuzumab reduced the rate of BVL to 24–42% compared to sc IFN-β-1a [45, 46]. Both the ALLEGRO and the BRAVO studies reported reduction of brain atrophy with laquinimod [47, 48]. Daclizumab had a marginal, but a significant favorable effect on brain atrophy compared to IM IFN-β-1a [49]. In the three trials (FREEDOMS I & II and TRANSFORMS), fingolimod also significantly reduced BVL to 28–45% and the reduction of brain atrophy was evident early during the course of the treatment [6‐8].

The panel discussed incorporating BVL as a key measure and part of the treatment strategy, which focuses on achieving no evidence of disease activity (NEDA). The measures under NEDA-3 were focused on the composite measures of absence of relapses, disability progression (based on expanded disability status scale (EDSS) scores), and MRI activity (new or enlarging of T₂ lesion) [30, 31]. As more evidence is accumulating with respect to the importance of brain atrophy during different stages of MS, absence of BVL may be incorporated to the NEDA measures to be a total of four measures (NEDA-4). Incorporating BVL in NEDA-4, allows a more comprehensive and balanced assessment, capturing both focal and diffuse disease activity [55, 56]. In the pooled analysis of the two FREEDOMS studies with fingolimod, patients on fingolimod were 4 times were more likely to achieve NEDA-4 than those who were on placebo at two years [57]. The advisors discussed the following case as an example of a patient who achieved a NEDA-4 during a period of 3 years. A 35-year old female was diagnosed with MS and had a highly active disease (more than 20 active lesions) at baseline. A disease modifying therapy was initiated in 2012. Over the 3-year longitudinal follow-up, there was no evidence of disease activity (absence of relapses, disease progression and MRI new/enlarging lesions). The change in the annualized brain volume change during the observational period was −0.089, which was below the threshold considered in NEDA 4 of −0.4% (Fig. 1).

The panel acknowledged that cognitive ability is not readily observable in routine neurological examinations, and self-reported cognitive complaints are confounded by mood and other subjective symptoms. One of the core cognitive deficits in MS patients is slowing of information processing speed, which can be subtle and difficult to assess without formal neuropsychological testing. In addition, there is a problem with the testing-retesting reliability as well as confounders at the time of testing. The advisors listed the most common tools that are utilized for cognitive assessment in their clinical practice such as the BICAMS (Brief International Assessment for Cognition for MS), CDT (The Clock Drawing Test), the MSNQ (MS Neuropsychological Screening Questionnaire), the PASAT (The Paced Auditory Serial Addition Task), the SDMT (Symbol Digit Modalities Test), and finally the WLG (Word List Generation) (Table 4.).

The panel utilized a standard format of questions for assessing cognition, including; 1) the types of tools they would use to test cognitive assessment, 2) the types of MS patients in whom they would apply these tests 3) the timing of the assessment 4) where they would apply these tests 5) and finally who would apply these tools. They used a voting system to rate and highlight the importance and the clinical relevance of each cognition assessment test the in the real world setting. Each voting participant could vote multiple times for different categories.

Of the 12 expert voting panel, 75% voted for the Symbol Digit Modalities Test (SDMT) as the most common test they would utilize for cognition. All voted to use the SDMT in RRMS and equal distribution of votes for the rest of the MS spectrum. All of the advisors recommended to have the assessment test at 12 months during the office visit and 83% of the experts recommended the tools to be conducted by the nurse, since it is time consuming for most neurologists (Table 4).

The advisors utilized the same methodology for assessing brain volume, including; 1) the types of methods they would use to assess BVL, 2) the types of MS patients they would target 3) the timing of the assessment during disease activity 4) and who would conduct these studies. They also used a voting system to rate importance and the clinical applicability of these methods in the routine clinical practice and each voting member was able to vote multiple times for different categories.

As for the brain volume assessment (Table 5.), all (100%) of the advisors recommended the Structural Image Evaluation, using Normalization, of Atrophy (SIENA), a registration-based method, to measure BVL mainly in patients with radiologically isolated syndrome (RIS), CIS, & RRMS at the 6–12 month time frame for most participants. It was recommended to have this measurement taken by the MRI technician.