Amiloride, fluoxetine or riluzole to reduce brain volume loss in secondary progressive multiple sclerosis: the MS-SMART four-arm RCT

Amiloride

Amiloride is a widely used diuretic and antagonises ASICs. This last mode of action is thought to be responsible for the myeloprotective and neuroprotective effects in both human and experimental models of progressive MS. Seventeen patients affected by primary progressive multiple sclerosis (PPMS) were enrolled in a Phase IIa trial, divided into a pre-treatment phase (1 year) and an amiloride treatment phase (1 year), at 5 mg bis in die (b.i.d.). During the amiloride treatment phase, patients showed a significant reduction in the whole-brain atrophy rate compared to pre treatment (p = 0.009). Also, corpus callosal radial diffusivity (RD) (myelin integrity) and thalamic mean diffusivity (MD) (structural integrity) showed a significant improvement. 30

Fluoxetine

Fluoxetine is a selective serotonin reuptake inhibitor (SSRI) that is widely used to treat depression. The main properties of this drug relevant for SPMS were the stimulation of glycogenolysis and the induction of production of brain-derived neurotrophic factor in rodent astrocyte cultures. A significantly improved cerebral white matter N-acetylaspartate (NAA)/creatine ratio was found on MRI, suggesting that the drug may improve axonal mitochondrial energy metabolism. Another possible mechanism of action of this drug is the suppression of the antigen-presenting capacity of glial cells. 31,32 Moreover, SSRIs have been demonstrated to ameliorate measures of dependence in patients with stroke. Two Phase IIa trials already tested fluoxetine in MS. In a cohort of 40 MS patients (10% SPMS) there was a significant reduction in relapse rate incidence to 0.54 (95% CI 0.29 to 0.98) with a trend towards reduction in new inflammatory lesions. 33 In the second trial, 42 patients with SPMS/PPMS (SPMS, n = 69%) were randomised in a 1 : 1 ratio to 40 mg or placebo over 2 years. 34 No statistical differences were seen in overall progression or in grey or white matter volume changes, but there were trends in favour of fluoxetine in relation to disability progression. This second trial was terminated early because of drug expiry issues and was considered to be underpowered.

Riluzole

Riluzole is a drug already licensed for motor neuron disease/amyotrophic lateral sclerosis and has two modes of action that may target neurodegeneration in SPMS: reducing glutamate release and antagonism of voltage-dependent sodium channels. 35 Sixteen patients with progressive MS were enrolled in a Phase IIa trial and studied for 1 year before treatment, followed by riluzole 50 mg b.i.d. for 1 year. 36 The primary outcome was the change in cervical spinal cord cross-sectional area, which showed a reduction from –2% (year 1) to –0.2% (year 2). Moreover, the increase in T1 hypointense lesion load was reduced from 15% in year 1 to 6% in year 2 and there was a reduction in brain parenchymal fraction of –1.0% (year 1) and –0.7% (year 2).

Magnetic resonance imaging

Magnetic resonance imaging is central to the diagnosis of MS as well as to guide and monitor treatment response. MRI-derived metrics have been extensively used as an interim outcome measure for randomised clinical trials in MS. In RRMS, quantification of new gadolinium-enhancing lesions or T2-weighted lesions from MRI is generally used as the primary outcome measure at phase 2 to test the efficacy of new DMTs in development. 37–39 In SPMS, although there may be some effect on new lesion count and lesion volume, the main MRI parameter for investigating neurodegeneration is the change (reduction) in brain volume (BV). Compared with age-matched healthy controls, there is a greater decrease in BV over time in SPMS, termed atrophy rate, which can be quantified by MRI. On average there is 0.5–1% loss of BV per year in SPMS. MRI studies have demonstrated a correlation between BV loss and disability in SPMS. A decrease in BV is seen at all stages of MS, and especially in SPMS. Neuroaxonal tissue constitutes a large proportion of BV and the increased rate of brain atrophy has been interpreted as evidence for neuroaxonal loss. Brain atrophy is correlated with disability and cognitive impairment in MS. 40–49

A large variety of advanced MRI techniques have been investigated in MS to both quantify and clarify neurodegeneration. In MS-SMART, other than BV and new and enlarging T2 lesion measures, several other MRI technologies were explored and are now described.

Proton magnetic resonance spectroscopy (MRS) was used to measure brain metabolites including glutamate, NAA and myoinositol. 50–52 High levels of glutamate, an excitatory neurotransmitter, are responsible for neuronal excitotoxicity, which was specifically targeted by riluzole as a proposed mechanism of action. NAA, a metabolite synthesised in mitochondria, is associated with axonal integrity and mitochondrial function and is used as a marker of neuroprotection and energy metabolism. Myoinositol is considered a marker of glial cell proliferation and activation (astrocytes and microglia), which are prominent features of CNS inflammation and contribute to neurodegeneration in SPMS.

Magnetisation transfer ratio (MTR) is a technique that strongly reflects the amount of myelin in the brain, but can also be influenced by inflammation and axonal density. There is an exchange of magnetisation between protons that are freely mobile and those that are bound to macromolecules. The extent of this exchange provides an indication of the amount of macromolecular structure in tissue and can be measured with the MTR. 53,54 Myelin has a major effect on MTR and in MS lower MTR is seen in demyelinated rather than remyelinated white matter lesions. The amount of signal decrease is thought to suggest damage to myelin or to the axonal membranes. The MTR measure provides an indication of the extent of axonal loss associated with new inflammatory-demyelinating white matter lesions. MTR has provided many insights regarding the evolution of demyelination and remyelination and was used in MS-SMART to measure the potential benefit on remyelination and putative neuroprotective therapy effects. 55

Diffusion tensor imaging (DTI) can quantitatively detect brain microstructural changes by calculating parameters, such as MD, which is determined by the overall water motion, and fractional anisotropy (FA), which reflects the uniformity of the direction of the diffusion of water molecules. DTI measures may be considered as markers of demyelination and axonal loss. 56 Strong diffusion-encoding gradients allow measurement of water diffusivity in the direction parallel to white matter fibres and perpendicular to this, and changes in these correlate with axonal damage and demyelination. FA derived from the diffusion tensor demonstrates the tendency of water molecules to diffuse in one direction and is a further marker of white matter integrity and myelination status. Abnormal DTI values have been demonstrated in both white matter lesions and normal-appearing white matter (NAWM) in MS, and further deterioration in the placebo arm is expected over the duration of the trial. Stabilisation of these values in treated arms of the study may indicate prevention of further tissue loss within white matter tracts.

In MS, both the brain and the spinal cord are affected. Indeed, up to 90% of patients with MS develop spinal cord focal lesions or diffuse abnormalities, which contribute to disability. Neuronal loss in the cord may occur for direct damage owing to white and grey matter (GM) demyelination or to secondary Wallerian degeneration. 57–61 These phenomena can be captured by quantification of spinal cord atrophy, an aspect that can be substantial in progressive MS. 62 The most common method for assessing in vivo spinal cord atrophy is to measure the cross-sectional area at specific anatomical levels, typically in the cervical region in the area from C2 to C5, using MRI. Image acquisition and segmentation of the upper cervical region are more efficient than in other parts of the cord from a technical perspective,63 and are thought to be more representative of the neurodegenerative process in the cord as lesions are more common in the cervical tract64 and this cord section may reflect destructive processes lower down in the cord via Wallerian degeneration. There is a strong clinical–radiological association between spinal cord atrophy and disability measures in MS, particularly in the progressive stages of the disease. 62,65–67 In a recent systematic review and meta-analysis (including 22 longitudinal studies and > 1000 patients), a mean upper cervical cord cross-sectional area of 73.07 mm² in all MS patients and 68.55 mm² in SPMS patients (vs. 80.87 mm² in healthy controls68) was found, with a rate of spinal cord atrophy of 1.78% per year (95% CI 1.28% to 2.27% per year), potentially supporting the use of MRI-derived cervical cord area measurement as an outcome measure of neurodegeneration in MS trials.

Optical coherence tomography

The anterior visual pathway offers the possibility to study in vivo neuroaxonal loss in MS. Since the pioneering studies of Frisén and Hoyt,69 who described qualitative changes in the retinal nerve fibre layer (RNFL) in patients with MS, and Frohman,70 who found that about 50% of ganglion cells must be lost before focal RNFL defects are detectable, many attempts have been made to improve our ability to visualise the neuronal retinal layers. Optical coherence tomography (OCT) is an increasingly recognised, non-invasive imaging tool that allows investigation of the architecture of the neural retina, which had been virtually inaccessible until the advent of high-resolution spectral domain optical coherence tomography and has fundamentally changed our ability to qualitatively and quantitatively assess the eye. 71

Optical coherence tomography relies on interferometry of near-infrared light to construct very high-resolution images of the retinal layers. The most visible layer is the RNFL, comprising unmyelinated axons in a supportive connective tissue framework. The RNFL axons originate from retinal ganglion cell bodies, and continue through the optic nerve, chiasm and tract (where they are myelinated), to synapses in the lateral geniculate bodies. In acute optic neuritis, RNFL thickness increases because of optic nerve swelling. Subsequently, its thickness reduces, indicating significant axonal loss. 72 The optic nerve lesion leads to retrograde degeneration of the RNFL, a relatively pure compartment of unmyelinated axons whose thickness can be measured sensitively and non-invasively using OCT. Therefore, quantification of peripapillary retinal nerve fibre layer (pRNFL) thickness provides a plausible biomarker of axonal loss. 73 In addition, abnormalities of the retinal layers other than the pRNFL have been observed in post-mortem specimens from patients with MS, in which 79% of eyes exhibited ganglion cell loss and 40% showed amacrine and bipolar cell loss in the inner nuclear layer. These findings have been corroborated in vivo by OCT, demonstrating thinning of the ganglion cell layer (GCL) and inner plexiform layer (IPL), and are associated with reductions in visual function and vision-specific quality of life. 74

Many studies have reported an inverse correlation between pRNFL thickness and Expanded Disability Status Scale (EDSS) score in MS, and the strongest correlation was found for patients not affected by optic neuritis. 75 A recent meta-analysis examining studies using OCT in mixed cohorts of MS patients confirmed that, when compared with healthy controls, pRNFL and the complex GCL and IPL were decreased in both MS optic neuritis and non-optic neuritis eyes. 76 One study found that patients with MS and a pRNFL thickness of up to 87 µm (as measured with a Cirrus OCT machine; Carl Zeiss, Dublin, CA, USA) or 88 µm (as measured with a Spectralis OCT machine; Heidelberg Engineering, Heidelberg, Germany) had roughly twice the risk of disability worsening during follow-up compared with patients with thicker pRNFL. 77 This risk was independent of other factors known to be associated with disability worsening, including age, disease duration, baseline level of disability (EDSS) and the use of DMTs. It was suggested that the process of neurodegeneration within the retina could reflect similar processes occurring more diffusely in the brain and spinal cord. An extensive review of the literature confirmed that using OCT in MS can provide valid, reliable and reproducible data to track the process of neurodegeneration within the retina of patients with MS with optic neuropathy. pRNFL thickness and macular volume analyses could serve as a surrogate biomarker and primary outcome measure to confirm the neuroprotective effects of new drugs. 70

Optical coherence tomography has good analytical reproducibility, is cost-effective, correlates with clinical measures (loss of visual function, cognitive decline) and is predictive of a clinical outcome (poor visual recovery). 75 Finally, atrophy of the complex GCL and IPL appears to mirror whole-brain (and particularly GM) atrophy, especially in progressive MS, supporting the use of OCT in clinical trials. 78

Despite the evidence described above, and although serial OCT-measured pRNFL thickness has been proposed as a measure of neurodegeneration for clinical trials in MS, longitudinal observations are largely confined to RRMS, underscoring the need to develop the data set in progressive MS. 76,77

Cerebrospinal fluid

Cerebrospinal fluid (CSF) measures are potential outcome measures in SPMS trials.

Neurofilaments represent a component of the mature cytoskeleton of neurons,79 and are composed of neurofilament light (NfL), neurofilament medium (NfM) and neurofilament heavy (NfH) chain subunits. All the pathological processes that cause neuroaxonal damage release neurofilament proteins into the extracellular space, CSF and, depending on the extent of damage, the peripheral blood; therefore, CSF and blood neurofilament levels are considered as markers of neuroaxonal damage.

Disease-modifying therapies for RRMS, in particular the highly active therapies, have all demonstrated reduction in CSF NfL levels (fingolimod,80 natalizumab,81 ocrelizumab82), indicating a relationship between anti-inflammatory therapies and neuronal damage resolution. In SPMS, serum NfH and CSF NfH in a cohort of patients compliant with lamotrigine (a putative neuroprotectant), demonstrated lower levels than placebo. 83 This emphasises the mechanistic utility of sampling the CSF for such markers and exploring neurofilament and other measures in an established SPMS cohort.

Primary objective

The primary objective of the MS-SMART study was to establish whether or not any of the three selected drugs (i.e. amiloride, fluoxetine and riluzole) were able to decrease the progression of BV loss in people with SPMS over 96 weeks, as assessed by MRI-derived percentage brain volume change (PBVC).

Secondary objectives

The secondary objectives were to:

1. establish that a multiarm trial strategy was an efficient way of screening drugs in SPMS and can become the template for future work

2. explore any anti-inflammatory drug activity (measured by counting the new and enlarging T2-weighted white matter lesions)

3. examine for evidence of pseudo-atrophy by MRI

4. examine the clinical effect of neuroprotection as measured by clinician- and patient-reported outcome measures (PROMs)

5. collect basic health-related quality-of-life data.

Exploratory objectives

The exploratory objectives were to:

1. assess neuroprotection in new lesions by estimating persistent new T1-weighted hypointense lesion (or ‘black holes’) count

2. assess cortical neuroprotection by evaluation of GM volume change

3. evaluate myelination using MTR

4. assess spinal cord neuroprotection using MRI

5. quantify neuronal mitochondrial dysfunction, prevention of glial cell inflammation and prevention of excitotoxicity as measured by MRS metabolites

6. quantify neuroprotection using DTI as an index of white matter integrity

7. evaluate neuroprotection using OCT measures

8. quantify neuroprotection using CSF neurofilaments

9. investigate treatment effect and predictive models of brain atrophy (by MRI) and clinical disability (by EDSS) at 96 weeks using baseline MRI and disability scores.

Inclusion criteria

• Confirmed diagnosis of SPMS. Steady progression rather than relapse must be the major cause of increasing disability in the preceding 2 years. Progression can be evident from either an increase of at least 1 point in EDSS score or clinical documentation of increasing disability in the patient’s notes.

• EDSS score of 4.0–6.5.

• Aged 25–65 years, inclusive.

• Women and men with partners of childbearing potential had to use an appropriate method of contraception.

• Women must have a negative pregnancy test within 7 days prior to the baseline.

• Willing and able to comply with the trial protocol (e.g. can tolerate MRI and fulfils the requirements for MRI).

• Give written informed consent.

Exclusion criteria

• Pregnant or breastfeeding females.

• Baseline MRI scan not of adequate quality for analysis.

• Significant organ comorbidity.

• Relapse within 3 months of baseline visit.

• Intravenous or oral steroid treatment for a MS relapse/progression within 3 months of baseline visit.

• Use of simvastatin at 80-mg dose within 3 months of baseline visit.

• Commencement of fampridine within 6 months of baseline visit.

• Use of immunosuppressants (e.g. azathioprine, methotrexate, ciclosporine) or first-generation disease-modifying therapies (β-interferons, glatiramer acetate) within 6 months of baseline visit.

• Use of mitoxantrone, natalizumab, alemtuzumab, daclizumab, fingolimod, fumarate, teriflunomide, laquinomod or other experimental disease-modifying therapy within 12 months of baseline visit.

• PPMS.

• RRMS.

• Known hypersensitivity to the active substances and their excipients to any of the active drugs for this trial.

• Use of a SSRI within 6 months of the baseline visit.

• Current use of tamoxifen.

• Current use of herbal treatments containing St. John’s wort.

• History of bleeding disorders or current use of anticoagulants.

• Use of monoamine oxidase inhibitors, phenytoin, L-tryptophan and/or neuroleptic drugs, lithium, chlorpropamide, triamterene or spironolactone within 6 months of the baseline visit.

• Current use of potassium supplements.

• Significant signs of depression and a Beck Depression Inventory, version 2 (BDI-II), score of ≥ 19.

• Bipolar disorder.

• Epilepsy/seizures.

• Receiving or previously received electroconvulsive therapy treatment.

• Glaucoma.

• Routine screening blood values: liver function tests > 3 × upper limit of normal of site reference ranges (aspartate aminotransferase, alanine aminotransferase, bilirubin, gamma-glutamyl transferase); potassium levels of < 2.8 mmol/l or > 5.5 mmol/l; sodium levels of < 125 mmol/l, creatinine levels of > 13 mmol/l; a white blood cell count of < 3 × 10⁹/l; lymphocytes count of < 0.8 × 10⁹/l; a neutrophil count of < 1.0 × 10⁹/l; a platelet count of < 90 × 10⁹/l; and haemoglobin levels of < 80 g/l.

Recruitment of participants

The identification of the potential MS-SMART participants was conducted through several routes:

• in clinics run by principal investigators/neurologists, at participating hospital and clinic sites

• using existing MS research and other neurological databases, such as the Scottish Health Research Register (SHARE) (URL: www.registerforshare.org), containing contact details of people who previously consented to be contacted about research

• by primary care referrals

• by participants’ self-referral through a dedicated MS-SMART website.

Pre-screening routes to check patients’ eligibility were allowed:

• in-clinic briefing of potential participants about the study directly by a member of the clinical team

• initial telephone contact from a member of the research team to explain the trial.

All the contacted potential participants were provided with a study patient information leaflet (PIL) and then at an interval no less than 24 hours after receiving the PIL they were re-contacted to determine if they wanted to proceed to face-to-face screening. If required, with permission, the participant’s general practitioner (GP)/neurologist was also contacted to provide written evidence that the patient met all the relevant eligibility criteria, before the screening visit in order to avoid unnecessary visits.

Following the pre-screening review at the London and Edinburgh sites, the potential participants were also provided with PILs for the optional relevant substudies.

The anonymity of all ineligible patients and all eligible patients who declined participation was maintained in every recruiting site. Anonymised information was collected including age, sex, date of screening, reason not eligible to participate (if applicable), reason for declining participation despite eligibility (if applicable) and any other reason for non-participation (if applicable).

Investigational medicinal products

The intervention was a random allocation to amiloride (5 mg twice daily), or fluoxetine (20 mg twice daily), or riluzole (50 mg twice daily), or placebo (one capsule twice daily). For the first 4 weeks, participants were asked to take only one capsule daily (Figure 1). Details on dose modification and stopping rules are reported elsewhere. 29

FIGURE 1.

Drug titration scheme. AE, adverse event. Adapted from Connick et al. 29 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Adapted from Connick et al. 29 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

The MS-SMART placebo comprised a size 00 capsule identical to the overencapsulated fluoxetine/amiloride/riluzole.

To evaluate the adherence to study drug, each participant was requested to bring back the unused study drug at each study visit and was asked about adherence. Participants had to record the number of capsules taken by using a diary card and to indicate any reason for non-adherence.

Screening visit

After the pre-screening phase, and at least 24 hours after receiving the PIL, patients underwent a formal face-to-face screening visit that involved documentation of written informed consent, review of medical history, concomitant medications, assessment of EDSS and BDI-II score, full physical clinical examination, recording of vital signs and collection of blood tests. After the screening phase, patients had a baseline brain MRI. In the majority of the recruitment sites, it was possible to perform the baseline MRI scan on the same day as the screening visit, decreasing travel costs and time commitment. A standard mandatory MRI protocol (defined as core MRI) was performed for all participants recruited in all sites. The London and Edinburgh sites performed optional MRI scan protocols (defined as advanced MRI), which included brain MTR, spectroscopy and DTI scans at the Edinburgh site, and brain MTR, spectroscopy and upper cervical cord at the UCL site. Each site subsequently transferred all scans to the Queen Square MS Centre Trial Office for central quality control and approval. Upon acceptance of the core MRI scans, the participants were then invited to take part to the baseline visit. At the London site, participants were also invited to take part in the optional OCT and CSF substudies. At the Edinburgh site, participants were invited to take part in the optional OCT substudy. Figure 2 shows the substudies carried out at the London and Edinburgh sites.

FIGURE 2.

Diagrammatic representation of the substudies.

Baseline visit

The activities at the baseline visit included detailed neurological evaluation, randomisation and collection of participant self-reported questionnaires. Rejection of the advanced MRI protocol only did not impede the enrolment of patients and, whenever possible, the rejected advanced MRI scans were repeated at the baseline visit. After completion of the baseline visit and randomisation, the study drugs, emergency card and patient diary were given to the participant.

For patients who consented to the substudies, OCT and lumbar puncture were also performed.

Follow-up visits

After the baseline visit, participants were reviewed at weeks 4, 8, 12, 24, 36, 48, 72 and 96 (Figure 3). At the week 48 and week 96 visits, a full neurological assessment was repeated and patients completed self-reported questionnaires. The MRI scan, instead, was repeated at week 24 and week 96. Standard clinical laboratory blood tests (chemistry and haematology) were performed for safety monitoring at each visit. Finally, at week 100, a safety telephone call was performed.

FIGURE 3.

Patient visits. MSFC, multiple sclerosis functional composite; SDMT, Symbol Digit Modalities Test; SLCVA, Sloan Low Contrast Visual Acuity.

For patients taking part in the substudies, lumbar puncture was repeated at weeks 48 and 96, and OCT was repeated at week 96.

At each study site, there were treating physicians [responsible for dose adjustment and adverse event (AE) monitoring], research nurses and trained NeuroStatus-certified assessing physicians masked to the patients’ history, who performed baseline and yearly clinical outcome assessments. Follow-up assessments were undertaken preferably by the same assessing physician to improve consistency.

Primary end point

Percentage brain volume change measured at 96 weeks using the Structural Image Evaluation using Normalization of Atrophy (SIENA) technique was chosen as the primary outcome measure in this trial. SIENA is an automated method that registers the follow-up scan to the baseline scan and produces an integral of the edge motion existing in each voxel in both scans. It then directly calculates the PBVC from those values. 47,85 PBVC is considered as a marker of neurodegeneration. 48

Secondary end points

Secondary MRI end points were count of new and enlarging T2 lesions at 96 weeks and PBVC at 24 weeks (to estimate pseudoatrophy). Although new and enlarging T2 lesions appear to be less relevant than brain atrophy as a measure of neuroprotection in SPMS, they were included as a core outcome measure in order to detect an unanticipated immunomodulatory effect.

Clinical secondary end points were EDSS, Timed-25-Foot Walk (T25FW), 9-Hole Peg Test (9HPT), Paced Auditory Serial Addition Test (PASAT), Multiple Sclerosis Functional Composite (MSFC), Symbol Digit Modalities Test (SDMT), high-contrast visual acuity (HCVA) (100%), and Sloan Low Contrast Visual Acuity (SLCVA) (5%, 2.5%, 1.25%) obtained at baseline, 48 and 96 weeks. Time to first relapse was recorded. Patient-reported outcomes of the following were also measured at baseline and at 48 and 96 weeks: Multiple Sclerosis Impact Scale 29 items, version 2 (MSIS29v2), Multiple Sclerosis Walking Scale, version 2 (MSWSv2), Neurological Fatigue Index (NFI), Numeric Pain Rating Scale (NPRS), Brief Pain Inventory (BPI), Neuropathic Pain Scale (NPS) and health-related quality-of-life [EuroQol-5 Dimensions, five-level version (EQ-5D-5L)]. The MSFC z-score was normalised using participants’ baseline scores.

Exploratory end points

Exploratory MRI end points were as follows:

• Proportion of new and enlarging T2 lesions at 24 weeks being persistently T1 hypointense at 96 weeks. Persistently T1 hypointense lesions exhibit greater axonal loss and could be a measure of neuroprotection. This end point was collected by all trial sites as part of the core MRI protocol.

• Percentage GM volume change (to assess cortical neuroprotection) at 96 weeks. GM atrophy is abundant in SPMS and shows robust correlations with disability. GM atrophy was investigated as an additional measure of neuroprotection. This end point was collected by all trial sites as part of the core MRI protocol.

• MR spectroscopy metabolite concentration including measures of NAA (reversal of neuroaxonal mitochondrial dysfunction), myoinositol (prevention of glial cell inflammation) and glutamate (prevention of excitotoxicity) at baseline and at 96 weeks. MR spectroscopy was part of the advanced MRI substudy and was performed at the London and Edinburgh sites.

• MTR in the GM and in the new lesions to determine demyelination and remyelination and MTR of NAWM to investigate tissue integrity. It was measured at baseline and at 96 weeks. MTR was part of the optional advanced MRI substudy and was performed at the London and Edinburgh sites.

• Upper cervical cord cross-sectional area (to investigate cord atrophy) measured on cervical cord three-dimensional (3D) phase-sensitive inversion recovery (PSIR) MRI scans acquired at baseline and 96 weeks. Spinal cord MRI was part of the optional advanced MRI substudy and was performed at the London site.

• Four measures were derived from DTI acquisition at baseline and at 96 weeks: FA, axial diffusivity (AD), RD and MD. The peak width of skeletonised diffusivity (PWSD) was used as the outcome measure for each metric. This quantifies dispersion of values across all tracts within an individual subject and, therefore, offers improved responsiveness over measures of central tendency in a disease characterised by multifocal pathology. DTI was part of the optional advanced MRI substudy and was performed at the Edinburgh site.

Additional exploratory end points were:

• Measurement of CSF neurofilament levels as part of the CSF substudy at the London site. The primary outcome of the CSF substudy was CSF NfL levels at 48 weeks. A secondary outcome was CSF NfL levels at 96 weeks and the change from 48 to 96 weeks. Additional CSF substudy outcomes included a panel of biomarkers: neural cell adhesion molecule (NCAM), NfH, glial fibrillary acidic protein (GFAP), ferritin, soluble cluster of differentiation 14 (CD14), matrix metalloproteinase 9 (MMP9) and neopterin.

• Measurement of the peripapillary RNFL and retinal nerve GCL and IPL (GCL + IPL complex) with OCT to determine the extent of retinal layer thinning were obtained at baseline and at 96 weeks. OCT scans were acquired as part of the OCT substudy at the London and Edinburgh sites.

Core magnetic resonance imaging

The following MRI sequences were obtained at all three MRI assessment visits for all participants:

• sagittal localiser to identify the subcallosal line

• axial dual echo fast/turbo spin echo proton density (PD)/T2 weighted from foramen magnum to vertex, in plane resolution 1 mm², contiguous 3-mm slices

• axial fluid attenuated inversion recovery from foramen magnum to vertex, in plane resolution 1 mm², contiguous 3-mm slices

• axial T1 from foramen magnum to vertex, in plane resolution 1 mm², contiguous 3-mm slices

• sagittal 3D T1 gradient echo with voxel resolution of 1 mm³.

Before the beginning of the trial, each site provided a ‘dummy scan’ (carried out on a healthy volunteer or a person with MS), which was sent the central MRI facility (Queen Square MS Centre Trial Office, London) for review and agreement on MRI parameters. Detailed MRI parameters according to scanner make and model are reported elsewhere. 29 Quality control feedback was generated by review at the central MRI facility soon after scan acquisition and was provided to the site.

The percentage brain volume change was calculated using the SIENA method. After the receipt of digital imaging and communications in medicine images to the central MRI facility and after quality control, the T2 lesions were outlined on PD scans by trained personnel blinded to clinical data using a semi-automatic method (Jim software, version 7.0, Xinapse Systems, Essex, UK). On further MS expert review, T2 lesion masks were used to lesion fill the 3D T1-weighed images using a patch-based method86 and, from these, brains were extracted and segmented into GM [i.e. cortical grey matter (CGM) plus deep grey matter (DGM)] and white matter volumes. 87 Finally, the SIENA method was applied to the 3D T1 images and the segmentations to calculate a percentage change in BV between baseline and week 96 scans.

Percentage change in GM volume was calculated by taking the difference in the GM volumes between 96 weeks and baseline divided by the baseline GM volume.

Advanced magnetic resonance imaging

An advanced MRI protocol was used at the London and Edinburgh sites.

At both the London and Edinburgh sites, the following scans were acquired as part of this advanced protocol:

• brain MTR

• brain MRS.

At the London site only, cervical cord 3D-PSIR scans were performed.

At the Edinburgh site only, brain DTI was performed.

Sequence generation

Randomisation was by a central, internet-based, secure password-protected randomisation database. The random allocation sequence was generated by the programmers at a UK Clinical Research Collaboration-registered trials unit [Edinburgh Clinical Trials Unit (ECTU)].

Type of randomisation

Patients were randomised to amiloride, fluoxetine, riluzole or placebo in a 1 : 1 : 1 : 1 ratio using a minimisation algorithm balanced according to sex, age (< 45 years or ≥ 45 years), baseline EDSS score (4.0–5.5; 6.0–6.5) and centre, with a random component to maintain unpredictability of treatment allocation.

Implementation

Eligible patients were randomised by study site personnel via a secure web-based randomisation service.

Blinding

Participants and all other personnel directly involved in the study were masked to treatment allocation. Amiloride, fluoxetine, riluzole and placebo capsules were overencapsulated to obtain an identical appearance to ensure that treatment allocation remained concealed to both staff and participants.

The MRI data were processed independently at a central reading site (Queen Square MS Centre Trial Office, UCL, London, UK) by staff unaware of trial group assignments.

Primary outcome

The primary end point, the PBVC between baseline and 96 weeks, was calculated using a normal linear model to compare the three active treatment group arms with placebo adjusting for baseline normalised BV and the minimisation variables: age, sex, treatment centre (as a fixed effect) and baseline EDSS score. Baseline normalised BV was entered into the model as a continuous variable, as were the minimisation variables of age and EDSS score. Treatment centre was included as an explanatory factor variable with the UCL centre as the reference category. Brighton, Truro and Plymouth each had small numbers of patients (< 11 patients) and so these sites were combined in a ‘South Coast Other’ category. Similarly, the sites Liverpool, Stoke and Newcastle also had small numbers of patients (< 11 patients) and so were combined into a ‘Northern Other’ category. The efficacy measure for each active treatment was the mean difference in PBVC change versus placebo. All patients for whom baseline and 96-week BV data were available were included in the analysis according to the treatment group to which they were randomised, irrespective of which treatment(s) they might have received. Dunnett-adjusted 95% simultaneous confidence intervals (CIs) and p-values were calculated using a single-step Dunnett procedure in order to adjust for the multiple pairwise comparisons and maintain the overall family-wise error rate of a false-significant result < 5% for the primary outcome analysis. No formal comparison of the active treatments was undertaken.

The primary analysis was a complete-case analysis based on the intention-to-treat (ITT) population. This meant that all patients for whom baseline BV and PBVC at 96-week values were available were included in the analysis according to the treatment group to which they were randomised, regardless of the treatment they actually received. Patients who withdrew from the trial or who were non-compliant with medication were also included, provided baseline BV and PBVC data were recorded for them. In a secondary analysis, the same analysis was based on the per protocol population. The per protocol population included all randomised patients who were adherent to the protocol and compliant with the allocated treatment throughout the duration of follow-up. For this trial, patients were considered compliant with medication if they reported taking, on average, ≥ 90% of their prescribed medication (taking account of planned down-titrations and deferred up-titrations) in the 30 days preceding each clinic visit. Any patients who withdrew from the trial or who were not compliant with medication were excluded from the per protocol population. The per protocol population was formally agreed, prior to database lock, by the chief investigator who was blinded to treatment allocation.

In addition, based on the ITT population defined above, the effect of missing data on the primary outcome analysis was investigated by performing three separate sensitivity analyses.

The sensitivity analyses involved:

1. Missing 96-week PBVC observations were imputed using a pattern-mixture approach to multiple imputation under a missing-not-at-random (MNAR) assumption for missing values. 102–104 A regression approach was implemented utilising fully conditional specification and the imputation model contained the same variables as for the final analysis. Thirty imputation data sets were created. The imputed values for PBVC at 96 weeks were then adjusted to reflect the MNAR assumption by adding a constant value equal to the observed standard deviation in the primary outcome at 96 weeks. In the clinical context of MS-SMART, the MNAR concern was that at least some of the missing data at 96 weeks would be due to deteriorating patient outcomes and disability. We report how large an amount can be added to the imputed values without changing the clinical interpretation of the trial, which was suggested as a possible sensitivity analysis for clinical trials by White et al. 105 If there were any missing baseline BV values then these will be imputed using the same multiple imputation method but without the MNAR adjustment. In this case we made the missing-at-random (MAR) assumption,102 which means that given the other data observed in the trial and the statistical analysis model being used, the missingness of baseline BV data are assumed to be independent of baseline BV.

2. A standard multiple imputation analysis was performed assuming that any missing outcome values were MAR. We used the same multiple imputation method as in (1) except no adjustment of imputations with respect to a MNAR assumption was performed in this case.

3. A complete-case approach was used as for the primary analysis except that any outliers more than 4 standard deviations away from the mean were excluded for the PBVC outcome.

Secondary magnetic resonance imaging outcome

Secondary clinical outcome

No adjustment for multiplicity was made when analysing the secondary and exploratory end points. The interpretation of secondary and exploratory outcome analyses will be suitably cautious to reflect the high number of outcomes considered.

When the change over time in discrete or continuous outcomes (i.e. EDSS, 9HPT, PASAT, MSFC, SDMT, SLCVA, MSIS29v2, MSWSv2, NFI, NPRS, BPI, NPS and EQ-5D-5L) were found to be reasonably normally distributed, these were compared between the active treatment and placebo groups using normal linear models. As for the primary outcome analysis, the linear models included trial arm as an explanatory factor variable (with placebo as the reference category), the baseline variable (corresponding to the outcome variable) and the minimisation variables: age, gender, treatment centre and baseline EDSS score. Treatment centre was included as an explanatory factor variable with the UCL centre as the reference category. Brighton, Truro and Plymouth each had small numbers of patients (< 11 patients) and so these sites were combined in a ‘South Coast Other’ category. Similarly, the sites of Liverpool, Stoke and Newcastle also had small numbers of patients (< 11 patients) and so were combined into a ‘Northern Other’ category. Normality could not be assumed for MSFC and so this outcome was transformed using the signed square root transformation prior to analysis, and results were checked using Mann–Whitney–Wilcoxon non-parametric comparison tests. For the EDSS outcome only, the 95% CIs were computed using a bootstrap method because of the ordinal nature of the outcome variable. Cox proportional hazard models (adjusting for the minimisation variables) were used for time to first relapse and T25FW, with the difference between each active treatment and placebo being expressed in terms of a hazard ratio. The variable ‘time to progress to a given EDSS score’ was not analysed using a Cox proportional hazard model as described in the protocol because it was only recorded at a maximum of three time points. Instead, the proportion of trial participants with an increase in EDSS score of at least 1.0 point at 96 weeks relative to baseline was analysed using a multiple logistic regression model adjusting for the minimisation variables.

For the NPS, the same linear regression analysis method was conducted as above, except that the analysis was applied to the individual questionnaire items rather than to any total score because the items are individually meaningful. For question 8 of the NPS, because this question has three categorical options and is in a different format to the other items, frequencies and percentages were calculated for each option at follow-up. Logistic regression was used to determine if there was a significant difference between the active treatment arms and placebo for each of the three binary options separately. These logistic regression models all adjusted for the minimisation variables to enable adjusted odds ratios to be computed.

Corresponding analyses were performed for clinical secondary outcomes measured at 48 weeks.

Exploratory outcomes

Statistical modelling assessed whether baseline MRI or CSF neurofilament levels or disability measures could be used to predict PBVC or EDSS score outcome at 96 weeks. Separate linear regression models were fitted for each potential baseline predictor of PBVC or EDSS score at 96 weeks. Trial arm was included as an explanatory factor variable in each model.

To investigate if treatment effect depends on baseline variables, the same analysis was performed as above, but with an interaction term included in the models representing the interaction between trial arm and the baseline variable. Specifically, we tested for interactions of treatment effect with the following baseline variables: Multiple Sclerosis Walking Scale (MSWSv2), MSIS total (MSIS29v2), MSIS physical, MSIS psychological, cross-sectional area of the upper cervical spinal cord, MSFC, EQ-5D-5L index, T25FW, 9HPT, SDMT score, PASAT, EDSS score at baseline, EDSS score at randomisation, HCVA, SLCVA 5%, SLCVA 2.5%, SLCVA 1.25%, CSF: NfH chain, CSF: NfL chain, T2LV, NAA and N-acetylaspartylglutamate concentration, deep grey matter volume (DGMV), total BV, cortical grey matter volume (CGMV), EQ-5D visual analogue scale (VAS), mean myoinositol concentration, mean Glx concentration, mean T2 hyperintense lesions MTR, mean whole-brain MTR, mean GM MTR, and mean NAWM MTR. Interactions found to be statistically significant suggest that the magnitude of the treatment effect differs according to the value of the baseline variable.

Regarding CSF neurofilament levels, we assessed the unique contribution of neurofilament levels (either NfL or NfH or both together) in predicting PBVC treatment response at 96 weeks over and above other baseline variables that were found to predict treatment response in the above analyses. This consisted of fitting a multiple linear regression model to PBVC at 96 weeks and comparing the model results with and without the neurofilament variables. The R² coefficient and residual standard deviation were reported as indicators of model fit.

In addition, a meta-analytic approach was used to assess the value of using MRI or disability or neurofilament variables recorded at 24 or 48 weeks as potential surrogate end points;107 the rationale being that we aimed to predict the effect of treatment on PBVC/EDSS score at 96 weeks based on the observed treatment effect on a surrogate end point. Bivariate mixed-effects models were used to assess surrogacy separately for each investigational treatment arm versus placebo comparison, based on the estimated treatment effect measured across different centres. Each potential surrogate end point was analysed separately. Coefficients of determination at the centre level were reported for each treatment and each surrogate end point. The analysis was conducted using the package ‘Surrogate’ within R software [The R Foundation for Statistical Computing, Vienna, Austria; URL: www.r-project.org (accessed 3 February 2020)].

Proportion of new and enlarging T2 lesions at 24 weeks being persistently T1 hypointense at 96 weeks

This outcome was analysed using multiple linear regression adjusting for the minimisation variables.

Percentage grey matter volume change

The percentage of GM volume change was statistically compared between trial arms using a multiple linear regression method adjusting for baseline and the minimisation variables.

Magnetic resonance spectroscopy

Changes in the NAA, myoinositol, and Glx metabolite concentrations were statistically compared between trial arms using a multiple linear regression method adjusting for baseline and the minimisation variables.

Magnetic transfer ratio

Lesion, GM and NAWM MTR measures were statistically compared between trial arms using a multiple linear regression method adjusting for baseline and the minimisation variables.

Diffusion tensor imaging

Peak width of skeletonised diffusivity-FA, AD, RD and MD were statistically compared between trial arms using a multiple linear regression method adjusting for baseline and the minimisation variables.

Cervical cord magnetic resonance imaging

The MUCCA at week 96 was statistically compared between trial arms using a multiple linear regression method adjusting for baseline and the minimisation variables.

Adherence to trial protocol

Only one protocol violation was recorded: eight patients at the Leeds site were randomised prior to baseline data collection. Eligibility was rechecked and all patients were confirmed to be eligible.

Protocol deviations (see Report Supplementary Material 1, Table 1) associated with randomised patients were recorded on 317 occasions, with similar frequencies across treatment groups. Most deviations (239; 75%) related to assessments occurring outside protocol-specified time windows. A further six deviations not linked to randomised patients were also recorded.

Adherence to trial medication

Overall, 337 participants were adherent to allocated trial medication. Adherence was similar across treatment groups: amiloride, 83 of 111 (75%); fluoxetine, 87 of 111 (78%); riluzole, 84 of 111 (76%); and placebo, 83 of 112 (74%). Eighty-five participants permanently discontinued their assigned treatment after randomisation [amiloride, 20 (18%); fluoxetine, 24 (22%); riluzole, 22 (20%); and placebo, 19 (17%)].

Adherence to remaining in the trial

Nineteen patients (4%) withdrew from the trial post randomisation (three deceased, one on instruction from their treating clinician and 15 at the request of the participant). Two withdrawals (both at the request of the participant) occurred after the 96-week primary end point. A further 13 patients (3%) could not be contacted and were recorded as lost to follow-up. Fifty-two patients (12%) did not attend the 96-week MRI follow-up: amiloride 12 (11%), riluzole 12 (11%), fluoxetine 15 (14%) and placebo 13 (12%).

Adherence to study blinding

The treatment allocation was unblinded six times during the course of the trial and the reasons were as follows: two deaths (one in the riluzole arm and one in the fluoxetine arm), three SAEs with hospitalisation (two patients were on riluzole, one was on placebo) and one for evidence of clinical worsening, which was suspected to be due to study drugs (the patient was on fluoxetine).

The success or otherwise of the study blinding was assessed via questionnaire for the participant or participant’s doctor. A total of 117 out of 386 participant respondents and 202 out of 389 clinician respondents did not venture a guess and selected the ‘Don’t know’ option. Out of those participants making a guess, we found that only 138 out of 269 participants (51%) correctly guessed their treatment allocation. The corresponding kappa statistic was calculated to be 0.039 (95% CI –0.060 to 0.139), which suggests very poor agreement between the participant’s guess and the true result. For clinicians, 111 out of 187 (59%) correctly guessed the participant’s treatment allocation, corresponding to a kappa statistic of 0.127 (95% CI –0.005 to 0.259). This again suggests poor agreement, albeit the observed agreement was slightly higher for the clinician than for the participant.

Primary outcome

There was no significant difference between any of the active arms and the placebo arm in the PBVC at week 96 (Table 6 and Figure 7). This finding was confirmed in sensitivity analysis (see Table 6).

FIGURE 7.

Dot plot for PBVC at 96 weeks. Patients are stratified by trial arm. The vertical lines show the mean ± 1 standard deviation.

The percentage brain volume change at 96 weeks was –1.35% in the overall cohort (see Table 7), corresponding to a PBVC of around –0.7% per year. In the placebo arm, the PBVC was similar to the overall cohort (see Table 7).

Secondary outcomes

New and enlarging T2 lesions

The number of new and enlarging T2 lesions after 24 and 96 weeks is reported in Table 9.

The adjusted rate ratio for the number of new and enlarging T2 lesions detected at the 96-week MRI scan is reported in Table 10. An adjusted rate ratio of < 1 indicates a lower rate of new and enlarging T2 lesions in the active treatment arm. The model was based on n = 400 patients. The scale parameter of the over-dispersed Poisson model was estimated to be 2.542.

There was no significant difference in the number of new and enlarging T2 lesions detected at the 96-week MRI scan for amiloride, or for riluzole versus placebo (Table 10).

Clinical outcomes

Detailed summary statistics for the clinical secondary outcomes at 48 and 96 weeks, split by trial arm, are reported in Report Supplementary Material 1 (Table 2). Multiple regression analyses for the clinical secondary outcomes at 48 weeks are also reported in Report Supplementary Material 1 (Table 3).

Patient-reported outcomes

Adjusted mean differences for MSIS29v2, MSWSv2, EQ-5D-5L and NFI (95% CIs) and p-values for each of the treatment arms versus placebo are reported in Table 13. AMDs for pain questionnaires (i.e. BPI, NPRS, NPS) are reported in Report Supplementary Material 1 (Table 4).

The multiple regression model for each outcome included trial arm as an explanatory factor variable (with placebo as the reference category), the baseline measurement and the minimisation variables: age, gender, treatment centre and EDSS score at randomisation. Negative AMDs are indicative of observed treatment benefit for all variables except EQ-5D.

Journal articles

Chataway J, De Angelis F, Connick P, Parker RA, Plantone D, Doshi A, John N, et al. Efficacy of three neuroprotective drugs in secondary progressive multiple sclerosis (MS-SMART): a phase 2b, multiarm, double-blind, randomised placebo-controlled trial [published online ahead of print January 22 2020]. Lancet Neurol 2020.

Connick P, De Angelis F, Parker RA, Plantone D, Doshi A, John N, et al. Multiple Sclerosis-Secondary Progressive Multi-Arm Randomisation Trial (MS-SMART): a multiarm phase Iib randomised, double-blind, placebo-controlled clinical trial comparing the efficacy of three neuroprotective drugs in secondary progressive multiple sclerosis. BMJ Open 2018;8:e021944.

Marshall I, Thrippleton MJ, Bastin ME, Mollison D, Dickie DA, Chappell FM, et al. Characterisation of tissue-type metabolic content in secondary progressive multiple sclerosis: a magnetic resonance spectroscopic imaging study. Neurol 2018;265:1795–1802.

Conference abstracts

Chataway J, Chandran S, Miller D, Connick P, Giovannoni G, Pavitt S, et al. (2016). Recruitment in Multiple Sclerosis Trials: The MS-SMART Experience. Presented at the Annual Meeting of the Association of British Neurologists (ABN), Brighton, UK, 17–19 May, 2016.

Connick P, Miller D, Pavitt S, Giovannoni G, Wheeler-Kingshott C, Weir C, et al. MS-SMART Trial Design and Recruitment Status: A Multi-arm Phase IIB Randomised Double Blind Placebo-controlled Clinical Trial Comparing the Efficacy of Three Repositioned Neuroprotective Drugs in Secondary Progressive Multiple Sclerosis. Presented at the European Committee for Treatment and Research in Multiple Sclerosis (ECTRIMS), London, 14–17 September 2016.

De Angelis F, Cameron J, Connick P, Miller DH, Pavitt S, Giovannoni G, et al. Optical Coherence Tomography in Secondary Progressive Multiple Sclerosis: A Baseline Data Report from the MS-SMART Trial. Presented at the European Committee for Treatment and Research in Multiple Sclerosis (ECTRIMS), London, 14–17 September 2016.

Doshi A, Plantone D, De Angelis F, Wheeler-Kingshott C, Ciccarelli O, Chataway J. Exploring Cognition In Secondary Progressive Multiple Sclerosis (SPMS). Presented at the Annual Meeting of the Association of British Neurologists (ABN), Brighton, UK, 17–19 May 2016.

Plantone D, Solanky B, Pavitt S, Giovannoni G, Wheeler-Kingshott C, De Angelis F, et al. Initial Cross-sectional MR Spectroscopy Analysis of a Cohort of Secondary Progressive MS Patients Enrolled in the MS-SMART Trial. Presented at the European Committee for Treatment and Research in Multiple Sclerosis (ECTRIMS), London, 14–17 September 2016.

De Angelis F, Gomez AG, Parker R, Plantone D, Doshi A, Barkhof F, et al. Examining Cross-sectional Relationships of Optical Coherence Tomography, Cervical Cord MRI and Disability in Secondary Progressive MS. Presented at the 7th Joint European Committee for Treatment and Research in Multiple Sclerosis (ECTRIMS)/Americas Committee for Treatment and Research in Multiple Sclerosis (ACTRIMS), Paris, 25–28 October 2017.

Calvi A, Prados F, Tur C, De Angelis F, John N, Doshi A, et al. Characterizing the Slowly Evolving Lesions (SELs) in a Cohort of Secondary Progressive Multiple Sclerosis Patients. Presented at the 4th Congress of the European Academy of Neurology (EAN), Lisbon, 16–19 June 2018.

Chataway J, De Angelis F, Connick P, Parker R, Plantone D, Doshi A, et al. MS-SMART Trial: A Multi-arm Phase 2b Randomised, Double Blind, Parallel Group, Placebo-controlled Clinical Trial Comparing the Efficacy of Three Neuroprotective Drugs in Secondary Progressive Multiple Sclerosis [NCT01910259]. Presented at the 34th Congress of the European Committee for Treatment and Research in Multiple Sclerosis (ECTRIMS), Berlin, 10–12 October 2018.

De Angelis F, Garcia A, Stutters J, Eshaghi A, Parker R, Plantone D, et al. Retinal Nerve Fibre Layer Thickness in the Temporal Quadrant Predicts Clinical Disability and Cervical Cord Cross-sectional Area in Secondary Progressive Multiple Sclerosis. Presented at the 34th Congress of the European Committee for Treatment and Research in Multiple Sclerosis (ECTRIMS), Berlin, 10–12 October 2018.

Doshi A, De Angelis F, Muhlert N, Prados F, Stutters J, Macmanus D, et al. Cognition in Secondary Progressive Multiple Sclerosis (SPMS). Presented at the 34th Congress of the European Committee for Treatment and Research in Multiple Sclerosis (ECTRIMS), Berlin, 10–12 October 2018.

Doshi A, De Angelis F, Muhlert N, Stutters J, Eshaghi A, Prados F, et al. Multiple Sclerosis Impact Scale and Brain Volume are Independent Predictors of Cognitive Impairment in Secondary Progressive Multiple Sclerosis. Presented at the 4th Congress of the European Academy of Neurology (EAN), Lisbon, 16–19 June 2018.

Monteverdi A, Solanky B, De Angelis F, Plantone D, Stutters J, John N, et al. Exploring Metabolite Profiling of Patients with Secondary Progressive Multiple Sclerosis. Presented at International Society for Magnetic Resonance in Medicine (ISMRM), Paris, 16–21 June 2018.

Ammoscato F, De Angelis F, Plantone D, Parker RA, Wheeler-Kingshott CAMG, Pavitt S, et al. CSF Neurofilament Heavy Chain Release (NfH) is Important in Secondary Progressive MS (SPMS) – Results from the MS-SMART Trial. Presented at the 35th Congress of the European Committee for Treatment and Research in Multiple Sclerosis (ECTRIMS)/24th Annual Conference of Rehabilitation in MS (RIMS), Stockholm, 11–13 September 2019.

John NA, Solanky B, De Angelis F, Stutters J, Prados F, Plantone D, et al. A Neurometabolic Profile of Secondary Progressive Multiple Sclerosis: The Relationship between Brain Metabolites and Clinical Disability. Presented at the Americas Committee for Treatment and Research in Multiple Sclerosis (ACTRIMS) Dallas, TX, 28 February–2 March 2019.