DIAGNOSIS OF PROGRESSIVE MULTIPLE SCLEROSIS FROM THE IMAGING PERSPECTIVE - A REVIEW

Abstract

Importance.

MRI demonstrates disease dissemination in space and time and may exclude multiple sclerosis (MS) mimics when applied in diagnosis of MS. Less effort has been expended in the application of MRI to progressive MS (PMS), including diagnosis of primary progressive (PP) MS and identifying patients with relapsing-remitting (RR) MS at risk to develop secondary progressive (SP) MS. We address clinical application of MRI in PMS for both diagnosis and prognosis. We also consider novel MRI indicators that reflect PMS pathophysiology.

Observations.

Although nonspecific, some spinal cord imaging features (diffuse abnormalities and lesions involving the gray matter -GM- and ≥2 white matter columns) are typical of PPMS. Both in PPMS and relapse-onset MS patients, the location of lesions in critical CNS regions (spinal cord, infratentorial regions, GM) and MRI-detected high inflammatory activity in the first years after the diagnosis predict long-term disability and future progressive disease course. These measures are currently evaluable in clinical practice.

In patients with established MS, GM involvement and neurodegeneration is associated with accelerated clinical worsening. Subpial demyelination and slowly expanding lesions are novel and promising indicators of progressive MS.

Conclusions and Relevance.

Diagnosis of PPMS is more challenging than of RRMS. No qualitative clinical, immunological, histopathological and neuroimaging features differentiate PPMS and SPMS; both are characterized by imaging findings reflecting neurodegeneration and are also influenced by aging and comorbidities.

Identification of MRI markers capable of distinguishing PPMS from RRMS and predicting evolution of RRMS to SPMS remain unmet needs. Integration of multiple parameters will likely be essential to achieve these aims.

INTRODUCTION

Magnetic resonance imaging (MRI) is central in the diagnostic work up of patients with suspected MS given its high sensitivity to demonstrate disease dissemination in space (DIS) and time and its substantial, albeit imperfect, ability to exclude other mimics of MS. From 2001 until 2017,¹ successive iterations of the McDonald criteria expended great effort to define imaging features typical for MS in patients presenting with a clinically isolated syndrome (CIS). Diagnosis of primary progressive (PP) MS remains challenging, however, and is only possible retrospectively, based on clinical assessment.

Identification of imaging features associated with PPMS, as well as features that predict evolution from relapsing-remitting (RR) to secondary progressive (SP) MS are important unmet needs. Given the advent of effective therapies for RRMS that may reduce development of SPMS and limit disability worsening in progressive MS (PMS),² the need for such imaging indicators is all the greater. Diagnosis of PMS is limited by difficulties distinguishing accumulating disability due to inflammatory disease activity from that attributable to degenerative processes associated with SPMS. Moreover, there are no accepted clinical criteria for SPMS.^3,4

This has promoted extensive research in the imaging field, facilitated by novel MRI sequences, to image pathophysiological mechanisms relevant to the pathobiology of PMS.

METHODS

A workshop was held in November 2019 in Milan, which involved neurologists and neuroradiologists. Five main clinically relevant questions regarding the role of MRI in PMS diagnosis and prognosis were identified (Box 1).

Box 1. Summary of the key questions addressed in the review and of the proposed MRI features for the identification of PMS.

Key questions and summary conclusions
(1) Is there any MRI feature that can support PPMS and SPMS diagnosis?
PPMS	Diffuse signal abnormalities in the spinal cord Spinal cord lesions involving the GM and ≥2 WM columns (axial plane) Atrophy of the lower portion of the cervical cord
SPMS	Spinal cord lesions involving the GM and ≥2 WM columns (axial plane) Atrophy of the lower portion of the cervical cord Cord GM atrophy
(2) Are there specific MRI features at disease onset able to predict disability and a progressive course?
PPMS	Gd-enhancing lesions at baseline Spinal cord lesions at baseline
SPMS	Number and volume of baseline brain T2-hyperintense lesions Increase of brain T2-hyperintense lesion volume during the first 5 years ≥2 Gd-enhancing lesions at baseline ≥1 spinal cord lesion at baseline ≥1 infratentorial lesion at baseline ≥1 cortical lesion at baseline ≥1 spinal cord lesion within 1 or 3 years ≥1 infratentorial lesion within 1 or 3 years ≥1 deep WM lesion within 1 year
(3) Are there MRI markers able to identify disability progression? In relapse-onset MS, are there MRI markers able to identify evolution to SPMS?
PPMS	New T1-hypointense lesions Cortical lesion number and volume Baseline GM damage Rate of brain atrophy
SPMS	T2 lesion volume Increase of T1-hypointense lesion volume Cortical lesion number and volume Conversion of dirty-appearing WM into focal WM lesions GM volume Rate of brain GM and deep GM atrophy
(4) Are there distinguishing MRI features between PPMS and SPMS?
None
(5) Are there candidate MRI biomarkers promising to identify MS progression?
≥4 paramagnetic rim lesions Subpial demyelination

Abbreviations: Gd=gadolinium; GM=gray matter; MRI=magnetic resonance imaging; MS=multiple sclerosis; P=progressive; PP=primary progressive; SP=secondary progressive; WM=white matter.

Experts provided a summary related to each topic (see eTable 1 for search strategy and selection criteria). A group consensus was reached during the workshop and summarized in a first draft, which was circulated among the meeting participants and additional experts in the field for critical discussion and revision. Here, we present the final conclusions from this workshop.

We also briefly discuss other promising biomarkers, including neurofilament light chain (NfL) levels, specific radiotracers used with positron emission tomography (PET), and optical coherence tomography (OCT), which have been investigated for similar purposes.

OBSERVATIONS

Diagnostic criteria for PMS

PPMS.

About 10–15% of MS patients exhibit a gradual progression of disability from disease onset.⁵ According to the 2017 revision of the McDonald criteria,¹ PPMS can be diagnosed in patients with ≥12 months of disability progression (retrospectively or prospectively determined), independent of clinical relapses, who additionally satisfy two of the three following criteria: (a) one or more T2-hyperintense lesions characteristic of MS in one or more of periventricular, cortical/juxtacortical, or infratentorial brain regions; (b) ≥2 T2-hyperintense lesions in the spinal cord; (c) CSF-specific oligoclonal bands. Distinction between symptomatic and asymptomatic lesions, an element of the 2010 iteration of the McDonald criteria, was eliminated in 2017.

SPMS.

Diagnosis is based on retrospective determination of progressive disability worsening unrelated to clinical relapses over ≥6–12 months.^3,4

Clinical criteria for disability progression

Gradual worsening of disability independently from relapses over at least 6 or 12 months in relapse-onset MS and 12 months in PPMS is the gold standard to identify disability progression.^4,6

The Expanded Disability Status Scale (EDSS) score with its functional system sub-scores is the most widely used scale to quantify MS-related disability.⁷ A gradual increase of EDSS score allows to demonstrate disability progression, which should be sustained for at least 3 or 6 months (i.e., confirmed disability progression [CDP]) to exclude possible confounding effects due to recent relapses or assessment errors.^4,6

However, EDSS score has limited accuracy in identifying MS progression. Inter- and intra-rater reliability are low; using self-reported walking distance adds to problems with reliability; gait is the major indicator for worsening when scores exceed 3.5 and gait can be influenced by many non-MS issues; EDSS has poor sensitivity when scores exceed 6.0–6.5.

In a study of 16636 MS patients, the progression criteria based on 3- or 6-month CDP were confirmed only in 70% (95% CI=68,71 %) and 74% (95% CI=72,75 %) at 5 years, suggesting up to 30% overestimation of disability progression.³

A recent evaluation of 17356 patients proposed a set of combined criteria to define disability progression that had an accuracy of 87% to identify patients evolving to SPMS over 5 years. The criteria included: (a) evidence of disability progression by 1 step in patients with EDSS ≤5.5 or 0.5 steps in patients with EDSS ≥6.0, in the absence of a relapse; (b) a minimum EDSS score of 4.0 and pyramidal functional system score of 2; (c) CDP for at least 3 months, including confirmation within the leading functional system.⁴

Another study of 273 MS patients with EDSS score between 4.0 and 7.5 showed that self-reported walking distance was misclassified when compared to the actual walking distance in 145 patients (53%). Such errors were more frequent in patients using walking aids (64% vs 44%, p<.05) and in patients with PPMS (69%, p<.05).⁸

Quantitative functional composite scores have been proposed to more sensitively and objectively investigate MS disability progression. The Multiple Sclerosis Functional Composite (MSFC) score comprises quantitative measures to assess leg function/ambulation (Timed 25-Foot Walk [T25FW]), arm/hand function (9-hole Peg test [9-HPT]) and cognition (Paced Auditory Serial Addition Test).⁹

Other composite outcomes combining overall (EDSS), upper (≥20% worsening of 9-HPT) and lower (≥20% worsening of T25FW) extremity disability have been proposed to capture MS progression.^10,11 The composite scores, defined as progression on ≥1 of 3 components (EDSS, T25FW, and/or 9HPT) with a ≥20% minimum threshold change for T25FW and 9HPT, detected 6-month CDP in 59.5% of 215 SPMS patients compared to 24.7% detected with EDSS alone.¹¹

Clinical identification of MS progression is typically retrospective, challenging and delayed by months to years after onset. Moreover, clinical examination does not allow an early identification of the neurodegenerative phenomena that start from the earliest phases of the disease, several years before the occurrence of overt disability progression.

MRI features for PMS diagnosis

PPMS.

Signal characteristics on conventional brain and spinal cord MRI do not differ qualitatively between early PPMS and relapse-onset patients, although PPMS patients may have fewer brain lesions. Focal cortical lesions (CLs) have also the same MR signal features.^12,13 As a single set of criteria to demonstrate DIS in relapse-onset MS and PPMS was feasible,¹⁴ MRI criteria for diagnosis have been harmonized for both MS subgroups,¹⁵ with the exception of requiring two rather than one spinal cord lesions to define spinal cord involvement for PPMS. By reducing the number of spinal cord lesions from two to one in PPMS, the sensitivity would improve to 84% (74/88) from 77% (68/88) using the 2010 McDonald criteria.¹⁶ However, the specificity of this modification would need further investigation before implementation.

SPMS.

No reliable MRI indicators of SPMS are available. A 10-year longitudinal study of 480 relapse-onset MS patients identified a subgroup of RRMS patients with disability progression without disease activity (i.e., relapses and/or new T2 lesions) (34/480, 7.1%), who experienced higher rates of whole brain, white matter (WM), and gray matter (GM) atrophy compared to those observed in stable patients.¹⁷ The investigators referred to this subgroup as experiencing ‘silent progression’ or disease-activity-free neurodegeneration independently from focal inflammation in patients with RRMS. Atrophy evaluation may identify SPMS patients.

Other MRI features for PMS identification.

PMS patients, particularly PPMS, on average, have fewer and smaller gadolinium (Gd)-enhancing lesions compared with RRMS.^18,19 However, the number of Gd-enhancing lesions do not discriminate RRMS from PMS at an individual level; 19/45 (42%) with early PPMS had at least one Gd-enhancing lesion.²⁰

Compared with RRMS and PPMS patients, those with SPMS have a higher proportion of brain T1-hypointense lesions (i.e., ‘black holes’) (Box 2).^21,22 However, the diagnostic value of black holes in PMS is still not adequately evaluated.^21,22

Box 2. Glossary.

Black holes

Black holes are defined as areas with unequivocal hypointensity compared with normal-appearing WM and GM matter on T1-weighted images obtained with spin-echo sequences at 1.5 T and have a corresponding T2-hyperintense WM lesion. Black holes are classified as acute or chronic (persistent or permanent). Whereas acute black holes reflect transient edema and inflammation, chronic black hole are those T1-hypointesities persisting ≥6 months after their first appearance on MRI and are characterized, pathologically, by severe demyelination and axonal loss.126,127

Diffuse spinal cord abnormalities

Abnormal areas of subtle increases of signal intensity on PD-weighted or STIR images, between that of focal lesions and normal-appearing spinal cord, lacking well-demarcated borders from adjacent normal-appearing cord.128

Slowly expanding lesions

Pathologically, slowly expanding (also known as chronic active or smoldering) lesions represent up to 57% of chronic lesions (although credible estimates are in the range of 25%). They are often characterized by a ‘rim’ of iron-laden activated microglia/macrophages, although not in all lesions,98 and signs of peripheral slow but ongoing demyelination and axonal loss around an inactive core without substantial BBB damage, thus reflecting a compartmentalized pathological process.79,81,94 Slowly expanding lesions have been investigated in vivo in lesions that progressively increase in size and show a paramagnetic rim on susceptibility-based MRI,85,94–96 corresponding, pathologically, to peripheral iron-laden microglia, or by evaluating the gradual expansion on conventional T1- and T2-weighted sequences of lesions showing a progressive decline of T1-hypointensity.97,100

Dirty-appearing white matter

DAWM (or diffusely abnormal WM) is defined as area with ill-defined borders and with signal intensity on T2- and/or PD-weighted MRI between that of focal WM lesions and the surrounding normal-appearing WM and isointense to the signal of the nearby cortical GM.62–64 Typically, DAWM is noted in the periventricular regions especially parieto-occipital or in the centrum semiovale. Pathologically, DAWM is characterized by inflammatory infiltrates, BBB disruption, demyelination, gliosis and axonal loss, which are less severe compared to focal WM lesions.129

Abbreviations: BBB=blood-brain barrier; DAWM=dirty-appearing white matter; GM=gray matter; MRI=magnetic resonance imaging; PD=proton density; STIR=short tau inversion recovery.

Spinal cord diffuse T2 signal abnormality²¹ (Box 2, Figure 1) is commoner in PPMS (19/31, 61%) and SPMS (10/32, 31%) than in RRMS (6/28, 21%) (p<.01).²¹ It is associated with spinal cord atrophy, sensorimotor and bowel/bladder manifestations and disability.²¹ Diffuse abnormality has not been incorporated into MRI diagnostic criteria^1,15,23 because clinical scans are insufficiently reliable or specific to identify it, and it is dependent on the type of T2-weighted sequence used. This abnormality has not been described in CIS patients.^24,25 However, it was the only spinal cord alteration in 3/10 (30%) PPMS patients and in 8/147 (5%) RRMS patients without focal lesions.²⁵

Figure 1. Summary of the MRI features supporting PPMS and SPMS diagnosis.

Features supporting diagnosis for both PPMS and SPMS include: A) PSIR axial sequence of the spinal cord (SC) (top and middle rows), lesions involving the GM and ≥2 WM columns on axial plane. No lesions detected in a RRMS female patient with 4 years of disease duration (left), whereas a lesion affecting the right anterior and posterior column, the left posterior column and the central GM is visible at C3-C4 level in a SPMS female patient with a disease duration of 22 years (right). A schematic representation of spinal cord subdivisions of GM and WM columns is shown in the bottom row. B) atrophy of the lower portion of the cervical SC. On the right, a sagittal 3D T1-weighted sequence reveals atrophy of the lower portion of the cervical SC (C7) (cord surface in blue) in a SPMS male patient with disease duration of 26 years. On the left, a sagittal 3D T1-weighted sequence (cord surface in green) of a RRMS male patient with disease duration of 6 years is shown. SC atrophy was quantified using the cord finder toolbox (Jim 7.0). C) Diffuse signal abnormalities in the SC (orange arrowheads) on a T2-weighted sequence in a patient with PPMS. D) GM atrophy of the SC in SPMS.. Compared to a RRMS female patient with a disease duration of 7 years (left), greater SC GM atrophy at C3/C4 level is visible on PSIR sequence in a SPMS patient with a disease duration of 19 years (right) using a local thresholding segmentation technique (Jim 7.0).

Abbreviations: ant=anterior column; GM=gray matter; L=left; mm=millimeter; MS=multiple sclerosis; post=posterior column; PP=primary progressive; PSIR=phase-sensitive inversion recovery; R=right; RR=relapsing-remitting; SC=spinal cord; SP=secondary progressive; WM=white matter.

Confluent cord lesions may resemble longitudinally extensive myelitis and diffuse cord changes seen in other inflammatory disorders or vascular disorders. Axial T2 sequences typically show that diffuse lesions in MS represent confluence of discrete lesions, many peripherally located, rather than a single, homogeneous longitudinally extensive lesion. Diffuse cord changes may occur in inflammatory disorders such as HTLV-1 or HIV-associated myelitis and systemic vasculitis, or in vascular disorders such as dural arteriovenous fistula, in which abnormal flow voids indicate dilated veins and suggest the diagnosis.

A study that separately visualized spinal cord GM and WM reported that more patients with focal lesions extending to GM and involving at least two WM columns had RRMS vs. CIS (odds ratio [OR]=8.3, 95% confidence interval [CI]=2.5–27.6; p=.001) and SPMS vs. RRMS (OR=10.1, 95% CI=1.2–85.3; p=.03) (Figure 1);²⁴ there was no difference between SPMS and PPMS (OR=3.1, 95% CI=0.3–31.8; p=.34).²⁴ In a multicenter study, the central cord, which is primarily GM, was more often affected in PPMS than RRMS patients (peak t value=4.5 at C3; p<.05).²⁶

Spinal cord atrophy may be related to PMS. In the cervical cord, C7 area was smaller in patients with SPMS within 1 year of transition from RRMS (n=25, 56.0 ± 8.8 mm²) than other groups (RIS=63.7 ± 9.8 mm, n=34; RRMS=64.8 ± 10 mm² n=31; p≤.003) and was associated with SPMS (β=7.6, p=.004).²⁷ The lower cervical cord is selectively more vulnerable in SPMS for uncertain reasons; nonetheless, a craniocaudal pattern of spinal cord atrophy may herald SPMS (Figure 1). A recent multicenter study found that atrophy between C1/C2 and C5 was found in RRMS, whereas extension to C6-C7 was detected in PMS; C5/C6 involvement was more typical of SPMS than PPMS.²⁸ Measurement of cervical and thoracic cord GM area (Figure 1) together with brain GM volume may be superior to measurement of whole brain or cord atrophy in distinguishing PMS from RRMS (area under the curve=0.90).^29,30

Early predictors of progression

In relapse-onset MS, the number and volume of brain T2-hyperintense WM lesions at disease onset predicts long-term disability (Figure 2).^31–33 T2-hyperintense lesion numbers and volumes (LVs) on initial MRI (median [range] T2 LV=2.5 [0.0,55.0] cm³ vs. 0.7 [0.0,13.7] cm³) and increase in brain T2-hyperintense LV within the first five years after disease onset were greater in CIS patients who developed SPMS after 20 years compared to those who remained RRMS (T2 LV growth rate per year [bootstrap 95% CI]=2.89 [1.78,4.01] cm³ vs. 0.80 [0.63,0.99] cm³; p<.001) (Figure 2).³¹

Figure 2. Summary of MRI predictors of subsequent disability progression and evolution to SPMS at disease onset.

In PPMS early predictors are A) gadolinium (Gd)-enhancing (orange arrowheads) on post-contrast T1-weighted sequences³⁵ and B) spinal cord (SC) lesions at baseline (orange arrowheads),⁴⁰ STIR sequence. In SPMS early predictors are C) the number and volume of baseline brain lesions (T2-FLAIR sequences);³¹ D) increase of brain lesion volume (LV) during the first 5 years (orange arrowheads) (T2-FLAIR sequences);³¹ E) ≥2 Gd-enhancing lesions at baseline (orange arrowheads) (post-contrast T1-weighted sequence);³⁴ F) ≥1 SC lesion at baseline or ≥1 SC lesion within 1–3 years (orange arrowheads) (STIR sequence);³⁴ G)≥1 infratentorial lesion at baseline or ≥1 infratentorial lesion within 1–3 years (orange arrowheads) (T2-FLAIR sequence);³⁴ H) ≥1 cortical lesion at baseline (orange arrowheads) (DIR sequence);⁴⁴ I) ≥1 new deep WM lesion within 1 year (orange arrowheads) (T2-FLAIR sequence).³⁴ Relevant OR, HRs or global T2 LV that are associated with disability progression and evolution to SPMS at disease onset and that were derived from the literature and discussed in the text are also reported.

Abbreviations: cm=centimeter; DIR=double inversion recovery; FLAIR=fluid-attenuated inversion recovery; Gd=gadolinium; HR=hazard ratio; LV=lesion volume; MRI=magnetic resonance imaging; MS=multiple sclerosis; OR=odds ratio; PP=primary progressive; SC=spinal cord; SP=secondary progressive; STIR=short tau inversion recovery; WM=white matter.

Having at least two Gd-enhancing lesions (OR [95% CI]=3.16 [1.08,9.23]; p=.035) and one spinal cord lesion (OR [95% CI]=4.71 [1.72,12.92], p=.003) at CIS onset, and the occurrence of one spinal cord lesion (at 1 year, OR [95% CI]=5.72 [1.67,19.56], p=.005) or one infratentorial lesion within one or 3 years (at 1 year, OR [95% CI]=7.02 [2.06,23.94], p=.002) after disease onset predicted conversion to SPMS after 15 years (C-statistic=0.86 and accuracy=91% at 1 year; 0.89 and 88% at 3 years) (Figure 2).³⁴ The number of Gd-enhancing lesions on baseline scans was associated with disability progression over 5 years in patients with PPMS (Figure 2) (OR [95% CI]=1.28 [1.04,1.58]; p=.02).³⁵

Lesion topography at disease onset, particularly location in brainstem^33,36,37 and spinal cord^34,38,39 (Figure 2), is associated with greater disability worsening. A 30-year longitudinal study confirmed the association of baseline infratentorial lesions (≥1 vs. 0) (OR [95% CI]=20.3 [5.4,75.6]; accuracy=78%; p<.001), new infratentorial (≥1 vs. 0) (OR [95% CI]=19.3 [5.7,65.6]; accuracy=65%; p<.001) and new deep WM lesions (≥1 vs. 0) (OR [95% CI]=14.9 [3.3,68.1]; accuracy=65%; p<.001) one year after MS onset with long-term disability progression and SPMS evolution (Figure 2).³³

Spinal cord lesions predict worse disease evolution when present early in the disease course, independent of classification as RIS^40,41 or CIS^34,39 (Figure 2). Twenty-five of 71 (35%) subjects with RIS had asymptomatic spinal cord lesions.⁴⁰ Asymptomatic spinal cord lesions increased the risk of clinical conversion up to 75.3 fold (95% CI=16.1,350.0; p<.001); 21 out of 25 (84%) with RIS and asymptomatic cord lesions progressed to CIS (n=19) or PPMS (n=2) over 1.6 years.⁴⁰ In another study, spinal cord lesions were more common in RIS cases who developed PPMS (15/15, 100%) compared to those who developed relapse-onset MS (72/113, 64%) (p=.005) or remained without clinical evidence of MS (i.e., RIS) (74/324, 23%) (p<.001) over 5 years.⁴¹ CIS patients with spinal cord lesions had a 5.6-fold higher risk (95% CI=0.9–35.4, p=.065) of reaching an Expanded Disability Status scale (EDSS) score ≥3.0; this risk was greatest in those whose initial attack was non-spinal (hazard ratio [HR] [95% CI]=29.8 [1.1,786.5]; p=.042).³⁹ Baseline spinal cord lesion number (β [95% CI]=0.40 [0.26,0.54]; p<.001), increase in spinal cord lesion number (β [95% CI]=0.19 [0.11,0.28]; p<.001) and higher rate of spinal cord atrophy (β [95% CI]=−0.15 [−0.21,−0.09]; p<.001) are associated independently with higher EDSS (R²=0.53) 5 years after disease onset in non-spinal CIS patients.⁴² In RRMS within 2 years from disease onset, spinal cord lesions were an independent predictor of EDSS score ≥4.0 after 7 years (β [95% CI]=4.4 [2.1,9.0; p<.001]).⁴³

In 219 relapse-onset MS patients, the number of CLs at disease onset predicted conversion to SPMS after 7 years (HR=2.16, 4.79, and 12.3 for 2, 5, and 7 CLs, respectively; p<.001), and time to progression (up to 4 years earlier on average) (Figure 2).⁴⁴ No patient without CLs at baseline entered the SP phase and few (1.8%) reached an EDSS score ≥4.0 at last follow-up.⁴⁴

Markers of disease worsening

WM lesions.

WM LV of untreated MS patients increases by 5–10% per year,⁴⁵ higher in those with PMS than with RRMS.¹⁸ In patients with longstanding, severe MS, multifocal T2 lesions expand, and new ones appear, ultimately leading to confluent T2 lesions in the WM. WM LV and confluent lesions have been proposed as predictors of SPMS onset and disease progression (Figure 3). However, the correlations between T2 LV and disability are only moderate at best. Progressive accumulation of WM lesions in patients with severe MS has been reported in some studies^46,47 but not in others.⁴⁸ Differences in the degree of concomitant accumulation of new lesions, enlargement of pre-existing lesions and shrinkage of others in the cohorts analyzed could explain the discrepant conclusions of these studies.

Figure 3. Summary of the MRI markers to identify disability progression and SPMS evolution during the disease course.

In both PPMS and SPMS MRI markers are A) increase of number or volume of brain T1-hypointense lesions.⁵⁴ Baseline and 1-year follow-up T1-weighted sequences of a PPMS patient. At follow-up, three new T1-hypointense lesions are visible (orange arrowheads); B) cortical lesion number and volume.⁵⁵ A DIR sequence showing three cortical lesions (orange arrowheads) from a PPMS patient is shown; C) baseline thalamic (blue) and cortical (green) damage and atrophy.⁵¹ Quantification on MT imaging (left) and 3D T1-weighted sequences (right); D) rate of whole brain, brain GM (green) and deep GM (blue) atrophy.⁶⁹ Quantification on a 3D T1-weighted sequence using SIENA and FIRST software. In SPMS also E) brain T2-hyperintense lesion volume (LV) (T2-FLAIR sequence)⁵¹ and F) conversion of DAWM (orange arrowheads and dotted orange areas) into focal WM lesions (T2-FLAIR sequence).⁶⁵ Further promising MRI markers are G) the presence of ≥4 rim positive lesions on susceptibility-based MRI (orange arrowheads)⁸⁵ and H) presence of subpial demyelination (orange arrowheads) on T2* sequence at ultra-high field (i.e., 7 Tesla).¹⁰¹ Figure 3H is adapted from Kilsdonk et al.¹⁰¹ with permission. Relevant OR, HRs or global T2 LV that were suggested to identify disability progression and SPMS evolution during the disease course and that were derived from the literature and discussed in the text are also reported.

Abbreviations: cm=centimeter; DAWM=dirty-appearing white matter; DIR=double inversion recovery; FLAIR=fluid-attenuated inversion recovery; GM=gray matter; LV=lesion volume; MRI=magnetic resonance imaging; MS=multiple sclerosis; MT=magnetization transfer; MTR=magnetization transfer ratio; OR=odds ratio; PP=primary progressive; SP=secondary progressive; y=year; WM=white matter.

Black holes.

The relationship between black holes and disability has been assessed with conflicting results.^{21,22,45,49–54} In cross-sectional studies, black hole LV showed mild-to-moderate correlation with EDSS, especially when only the most hypointense voxels were evaluated.^{21,22,45,49,50}

Longitudinally, in relapse-onset MS, EDSS worsening over 10 years was most strongly associated with the combination of higher baseline black hole numbers (r=0.42; p<.001) and increasing black hole LV (r=0.53; p<.001).⁵² The change in black hole LV over 1-year was the only predictor of MS severity score (MSSS) after 12 years, explaining 20% of the variance in MSSS; baseline and 1-year changes of T2 and Gd-enhancing LV, and of brain and ventricular volume fractions were not retained in the model.⁵³ Conversely, baseline black hole volume did not predict worsening of disability assessed using MSSS or EDSS over up to 13 years.⁵¹ In PPMS, new T1-hypointense lesions after 15 months were associated with EDSS worsening after 15 years (β=0.28; p=.003) (Figure 3).⁵⁴

GM lesions.

CLs have been detected in all MS clinical phenotypes, but their burden and accumulation over time are higher in PMS and are associated with clinical worsening (Figure 3).^12,13,55,56 In a 5-year longitudinal study, CL volume (OR [95% CI]=1.7 [1.4,2.3]; p<.001), age (OR [95% CI]=1.2 [1.1,1.3]; p=.001) and cerebellar cortical volume (OR [95% CI]=0.2 [0.1,0.4]; p<.001) predicted evolution from RRMS to SPMS.⁵⁵ Focal lesions frequently affect the thalamus.^57,58 PMS patients have a higher thalamic lesion burden, especially lesions with a discrete/ovoid shape,⁵⁷ and are preferentially located in regions close to the CSF.⁵⁸ In PPMS and SPMS, thalamic lesion burden correlates with clinical disability.^57,58

Spinal cord lesions.

Greater spinal cord lesion burden is consistently associated with more severe disability, a progressive phenotype and faster disability progression.^26,28,59

The relevance of spinal cord lesions for clinical disability is supported by the recent recognition of PPMS phenotype in the setting of an isolated CNS demyelinating lesion, termed progressive solitary sclerosis.⁶⁰ Patients typically have a single critical lesion in the CNS, located in the spinal cord typically at the cervico-medullary junction, oligoclonal bands and experience progressive motor impairment.⁶⁰

Dirty-appearing WM.

Some investigators have reported that dirty-appearing (DA) WM (Box 2, Figure 3) is associated with higher EDSS score^61,62 while others have not.^63–65

A recent study found DAWM on proton density (PD)- and T2-weighted images in 88/348 (25.3%) RRMS patients.⁶⁴ After 8 years, DAWM burden was unchanged in the majority of RRMS patients (61/88, 69.3%), decreased in 25/88 (28.4%) and increased in 2/88 (2.3%). DAWM was associated with more severe brain atrophy (−4.8% vs. −4.2%; p=.038), but not with baseline and longitudinal changes of WM lesion burden and EDSS score.⁶⁴ The lack of association with disability could be explained by a gradual evolution of DAWM into WM lesions. A recent study of 589 SPMS patients showed greater transformation of DAWM into focal WM lesions after 3 years in those with disability progression than those without (2.70 cm³ vs. 1.76 cm³; p<.001).⁶⁵

DAWM can be found in up to 6% of healthy controls (HC)⁶⁴ and also occurs in other conditions such as cerebrovascular disease,⁶⁶ and therefore may not be MS-specific⁶⁶ and could be influenced by altered cerebrospinal fluid pulsatility linked to hypertension.⁶⁷ Furthermore, the prevalence of DAWM is associated with MR field strength.⁶⁴

Atrophy.

Brain^17,68–71 and spinal cord^{27,28,72–74} atrophy is associated with clinical disability and are useful prognostic markers (Figure 3). This is particularly relevant to PMS, a disease phenotype characterized by prominent brain and spinal cord atrophy.

Cerebral GM atrophy is an important indicator of clinical disability (Figure 3).^68–71 Compared with HC (mean [standard deviation, SD] annualized GM fraction loss=−0.028 [0.24] %), the 4-year GM fraction change was 8.1 times greater in RRMS (−0.23 [0.34] %), 12.4 times greater in RRMS converting to SPMS (−0.35 [0.37] %), and 14 times greater in established SPMS (−0.39 [0.50] %).⁷⁰ A large multicenter 2.4-year longitudinal study showed the annualized rate of deep GM and cortical atrophy was faster in SPMS (−1.45% and −1.11%, respectively) and PPMS (−1.66% and −0.79%, respectively) compared to HC (−0.94% and −0.34%, respectively) and RRMS (−1.34% and −0.67%, respectively). Deep GM atrophy occurred most rapidly and was associated with disability progression (β [95% CI]=−0.04 [−0.02,−0.06]; p=.006).^69,71 The pattern of GM atrophy progression was similar in relapse-onset MS and PPMS, first manifest in deep GM and subsequently spreading to involve an increasing number of cortical regions in PMS.⁶⁸ The cerebellum and the striatum atrophied earlier in relapse-onset MS than in PPMS.⁶⁸

GM damage also predicts long-term prognosis. In relapse-onset MS, a combined model including baseline GM volume (OR [95% CI] =0.71 [0.51,1.00]; p=.04) and T2 LV (OR [95% CI]=1.13 [1.04,1.24]; p=.005) predicted evolution to SPMS after 13 years (C-index=0.84).⁵¹ In PPMS, a model including GM mean diffusivity (a measure of microstructural abnormalities) (β=3.86; p=.03), GM atrophy (β=−0.24; p=.05) and new T1-hypointense lesions during the first 15 months of follow-up (β=0.28; p=.003) predicted 15-year disability (Figure 3) (R²=0.61).⁵⁴

Spinal cord atrophy progresses at a faster rate in PMS patients and in those experiencing disability worsening than those who do not.^27,28,72,74

PPMS and SPMS: similar or not?

Despite their historical separation, whether PPMS and SPMS are different in terms of pathophysiology is uncertain. The ratio of women to men in relapse-onset MS is greater than in PPMS; an excess of women does not occur in PPMS.^75–77 PPMS patients are older and more likely to have comorbid conditions and age may contribute to disability and to MRI changes in PPMS.^75–78 By contrast, genetic, epidemiological, immunological, histopathological and MRI findings are similar arguing against a fundamental distinction between PPMS and SPMS.^{75,76,79–82} Members of multiplex pedigrees may present with different clinical phenotypes.⁸⁰ RIS cases can evolve to either RRMS or PPMS⁴¹ and 28% of PPMS patients experience relapses.⁸³

Disability progression is a function of age and not pre-progressive disease course.^75,76,82,83 The rate of disease progression is similar in PPMS and SPMS patients after reaching EDSS≥4.0.^75,76

The burden of T2-hyperintense, T1-hypointense and Gd-enhancing lesions has been reported to be lower in PPMS vs. SPMS^18,19but this is now disputed.^35,84 Similar to what occurs in relapse-onset MS, although with a lower prevalence, Gd-enhancing lesions are more frequent in the earliest phases and decline over time.³⁵

The prevalence of paramagnetic rim lesions (PRLs) on susceptibility-based MRI⁸⁵ and of CLs^12,13 do not differ between PPMS and SPMS. Microstructural tissue abnormalities and irreversible tissue loss consistently occur in PPMS and SPMS and predate clinical progression.^86,87

Ageing can contribute similarly to disability progression in PPMS and SPMS. Trophic CNS support, compensatory mechanisms, and remyelinating capacity decline with age, limiting functional recovery and brain plasticity, and contributing to neurodegenerative processes.^88,89 Mitochondrial abnormalities and higher oxidative stress have greater impact on PMS than RRMS.⁹⁰

Comorbidities unrelated to MS also have detrimental effects, especially in older PMS patients, possibly accelerating disease progression.^78,91,92 Cerebrovascular diseases accelerate brain ageing, promote hypoperfusion with consequent hypoxic WM lesion accumulation, and contribute to neurodegeneration.⁹³ Cerebrovascular risk factors, including higher body mass index and dyslipidemia, are independently associated with higher EDSS and disability progression.^78,91,92

Collectively, these observations suggest that disease progression is similar in PPMS and SPMS, and that age and comorbid diseases may confound any reported differences. A recent revision of MS phenotypic categories combined PPMS and SPMS and classified PMS by activity (presence or absence of relapses or MRI activity),⁵ reflecting the shift in current thought on this issue.

New MRI markers of progression

Slowly expanding lesions.

Slowly expanding lesions (i.e., with a slow but consistent expansion over time)⁹⁴ have been proposed as a pathological substrate of disability progression in MS (Box 2).^79,81,94 These lesions can show a paramagnetic rim on susceptibility-based sequences (Figure 3),^85,94–96 limited repair, lower T1 signal intensity compared to quiescent lesions, and a progressive decline of T1-hypointensity indicating ongoing tissue destruction.⁹⁷ Patients with ≥4 PRLs have more severe brain atrophy, earlier motor disability and cognitive impairment and a 3.2-fold higher prevalence of PMS than those with ≤3 PRLs (8/29 [28%] vs. 4/45 [9%]; p<.05).⁸⁵

However, not all chronic active lesions show iron-laden rims in pathological studies.⁹⁸ Additionally, they are not specific to PMS; rim lesions occur in all MS phenotypes, at widely different reported rates: CIS/RRMS from 6% to 53%, PMS from 7% to 62%. Some studies reported a higher frequency in PMS compared to RRMS⁹⁹ while others did not.⁸⁴ Linear expansion over time on T1- and T2-weighted sequences occurs in a similar proportion of patients with RRMS and PPMS (68% vs. 72%, respectively).¹⁰⁰

Subpial demyelination.

Pathology studies suggest that demyelination extending inward from the brain pial surface may be MS-specific and more widespread in PMS than in RRMS. However, subpial demyelination is very poorly detected at standard field strengths.¹⁰¹ Conventional and quantitative MRI techniques at ultrahigh field strengths could improve the detection (Figure 3). SPMS patients have greater decrease of magnetization transfer ratio values in the outer cortical surface compared to RRMS,^102,103 but increased T2*, probably reflecting reduced myelin and iron content, is more widespread in PMS than RRMS.¹⁰⁴

Leptomeningeal enhancement.

Focal/linear leptomeningeal enhancement on delayed post-contrast T2-fluid-attenuated inversion recovery (T2-FLAIR) images, presumably reflecting meningeal inflammatory aggregates, has been described in 25% of MS patients using 3T MRI¹⁰⁵ and in 90% using 7T MRI.¹⁰⁶ Clinical and pathological data suggest an association of these MR findings with cortical atrophy, prior or current inflammation and cortical (subpial) demyelination.^105–109 However, focal leptomeningeal enhancement is not MS-specific, and is detected in other inflammatory CNS diseases¹¹⁰ and with ageing.¹¹¹ It is detectable more frequently in PMS (33%) compared to RRMS (19%)¹⁰⁵ but its discriminative power is low; its prognostic value is unknown.

Other candidate biomarkers of progression

Serum neurofilament light chain.

Neurofilaments are structural scaffolding proteins exclusively expressed in neurons and represent promising biomarkers for neuro-axonal injury. They are released into serum upon CNS damage and can be reliably quantified with SIMOA technology.

In a study of 246 MS patients,¹¹² sNfL levels were positively associated with EDSS score (β=1.141 [95% CI=1.106,1.178]; p<.001) and recent EDSS worsening (β=1.294 [95% CI=1.090,1.536]; p=.003). PMS had higher sNfL concentrations compared to CIS/RRMS (41.4 [IQR=32.1,57.2] pg/ml vs. 27.2 [IQR=19.2,57.2] pg/ml; β=1.205 [95% CI=1.106,1.418]; p=.029). This was confirmed in a longitudinal study of 259 MS patients with a median follow-up of 6.5 years (PMS=41.9 [31.9,55.7]) pg/ml vs. RRMS=29.7 [21.2,42.2] pg/ml; β [95% CI]=1.154 [1.059,1.258]; p=.001).¹¹³

Higher baseline sNfL levels have been also associated with a greater risk of EDSS worsening over 1 year,^112,113 EDSS score after 5 years (r=0.26; q=.012),¹¹⁴ 3-month CDP (>60 pg/mL vs. <30 pg/mL: hazard ratio [95% CI]=1.94 [0.97,3.87]; p=.06),¹¹⁵ and SPMS conversion over 5 years (converters vs. non-converters: median sNfL=31.7 vs. 16.9 pg/mL; q=.001).¹¹⁴

Longitudinal sNfL level changes also correlated with EDSS changes. In a study of 42 RRMS patients, EDSS score increased by 0.53 (95% CI=0.14,0.91, p=.009) points with a 10-fold increase in NfL over 2 years.¹¹⁶ In a 12-year longitudinal study of 607 patients, EDSS worsening and change in sNFL levels was correlated (β [95% CI]=1.015 [1.007,1.023]; p< .001).¹¹⁷

sNfL is still limited as a biomarker that identifies progression. Age, disease activity and treatments significantly influence sNfL. Moreover, sNfL changes are not specific for any neurological disease. Short-term variations and the influence of comorbidities still limit its application.

Optical coherence tomography.

OCT yields high-resolution images of the retina and has been proposed to identify PMS and predict disability progression.

A study of 541 MS patients demonstrated that the retinal nerve finer layer (RNFL) was thinner in patients with SPMS (mean [SD]=85.5 [14.3] μm) and PPMS (mean [SD]=80.5 [15.4] μm) compared to CIS (mean [SD]=98.2 [8.4] μm) and RRMS (mean [SD]=92.9 [13.0] μm).¹¹⁸

Similarly, in another cross-sectional study of 113 MS patients and 38 healthy controls, the RNFL the was thinner in those with PMS compared to RRMS (mean [SD]=38.63 [6.55] vs. 42.05 [7.26]; p=.023), as was the ganglion-cell/inner plexiform layer (GCIPL) (mean [SD]=60.01 [8.55] vs. 65.31 [8.87]; p=.006) and outer plexiform layer (OPL) (mean [SD]=17.96 [1.70] vs. 18.51 [1.30]; p=.033).¹¹⁹

Retinal atrophy is accelerated in patients with PMS. A recent study showed that PMS was associated with faster peripapillary RNFL (pRNFL) (β [95% CI]=−0.34% [−0.52,−0.16%] annualized percent rate of change; p<.001), GCIPL (β [95% CI]=−0.27% [−0.41,−0.12%] annualized percent rate of change; p<.001), inner nuclear layer (INL) (β [95% CI]=−0.10% [−0.18,−0.02%] annualized percent rate of change; p=.01),and outer nuclear layer (ONL) (β [95% CI]=−0.13% [−0.24,−0.03%] annualized percent rate of change; p=.01) atrophy, compared to RRMS.¹²⁰

In a large multicenter trial of 879 MS patients, those with the lowest pRNFL thickness category at baseline (≤87 μm) had increased risk of disability worsening over 3 years (HR [95% CI]=1.75 [1.19,2.59]; p=.005) compared with those in the highest tertile (>98 μm).¹²¹

Events of optic neuritis and focal WM lesions in the retro-geniculate portion of the visual pathway can influence OCT measures. Moreover, standardized data acquisitions and rigorous quality control are necessary and further rigorous and prospective studies are required before accepting a role of OCT to identify PMS.

Positron emission tomography. ¹¹C-PK11195 is a first generation tracer for translocator protein (TSPO). Greater ¹¹C-PK11195 uptake, believed to mirror accumulation of activated microglia, was found in SPMS compared to RRMS in the NAWM (mean [SD] 1.38 [0.09] vs. 1.26 [0.05]; p=.018) and in the rim of chronic active lesions (mean [SD] 1.33 [0.08] vs. 1.20 [0.12]; p=.043) but not in GM structures.¹²² Another study showed a more widespread increased cortical ¹¹C-PK11195 uptake in SPMS compared to RRMS, and an association of this finding with EDSS score (r=0.52; p=.026), especially in SPMS (r=0.87; p=.009).¹²³

Uptake of a second generation tracer for TSPO (¹¹C-PBR28) was greater in SPMS compared to RRMS patients in WM lesions, NAWM, deep GM and cortex but not in cortical lesions, although no formal statistical comparisons were performed.¹²⁴ Similarly, ¹⁸F-PBR06 uptake in deep GM was higher in SPMS compared to RRMS (+10.8%; p=.002).¹²⁵

PET tracers have been proposed also to investigate demyelination and remyelination processes and neuronal damage; however no study has evaluated specific features of PMS. PET is an encouraging technique for monitoring pathophysiological substrates contributing to MS progression; however, several aspects limit its application to identify PMS. PET has a lower spatial resolution than MRI and uses ionising radiation, thus raising safety concerns regarding serial radionuclide administration. Moreover, specificity of the radiotracers is limited, since some expression has been detected in astrocytes and endothelial cells; PET abnormalities are often not disease-specific. Standardized protocols for acquisition and analysis have not yet been defined. Preliminary results should be confirmed by larger prospective studies with stratification by MS phenotypes.

CONCLUSIONS

No definitive qualitative clinical, immunological, histopathological and neuroimaging features differentiate PPMS and SPMS, whereas both are characterized by neurodegenerative phenomena and a gradual and irreversible accumulation of clinical disability, which is also influenced by ageing and comorbidities.

A confident diagnosis of PPMS is more difficult than RRMS. PPMS is partly a diagnosis of exclusion, since it can be mimicked by other conditions clinically and radiologically. Diffuse signal abnormalities and lesions involving the GM and at least 2 WM columns on spinal cord MRI may be relatively specific to this stage of the disease, but confirmation is required.

Both in PPMS and relapse-onset MS patients, MRI features at disease onset predict long-term disability and a progressive disease course, including lesion location in critical CNS regions (i.e., spinal cord, infratentorial regions, and GM) and a high inflammatory activity in the first years after disease onset (Box 1).

Clinical worsening is associated with imaging measures of GM involvement and neurodegeneration; however, detection validation and standardization need to be implemented at the individual-patient level.

Discovery of MR markers capable of detecting evolution from RRMS to SPMS remains an unmet need. Multiparametric MRI studies are needed, as it is unlikely that a single MR method will be an optimal discriminator.

Novel candidate imaging biomarkers such as subpial demyelination and slowly expanding lesions/PRLs may identify PMS, but should be further investigated.

The contribution of these promising MRI measures, combined with other biomarkers, such as NfL level quantification¹¹⁷ or OCT assessment,¹²⁰ to improve PMS identification should be explored.