1141964 research-article20222022 MSJ0010.1177/13524585221141964Multiple Sclerosis JournalA Calvi, MA Clarke MULTIPLE SCLEROSIS MSJ JOURNAL Original Research Paper Relationship between paramagnetic rim Multiple Sclerosis Journal 2023, Vol. 29(3) 352­–362 lesions and slowly expanding lesions DOI: 10.1177/ https://doi.org/10.1177/13524585221141964 13524585221141964 https://doi.org/10.1177/13524585221141964 in multiple sclerosis © The Author(s), 2022. Article reuse guidelines: sagepub.com/journals- permissions Alberto Calvi* , Margareta A Clarke* , Ferran Prados, Declan Chard , Olga Ciccarelli, Manel Alberich, Deborah Pareto, Marta Rodríguez Barranco, Jaume Sastre-Garriga , Carmen Tur , Alex Rovira and Frederik Barkhof Abstract Correspondence to: A Calvi Background: Magnetic resonance imaging (MRI) markers for chronic active lesions in MS include Queen Square MS slowly expanding lesions (SELs) and paramagnetic rim lesions (PRLs). Centre, Department of Neuroinflammation, Objectives: To identify the relationship between SELs and PRLs in MS, and their association with dis- Institute of Neurology, Faculty of Brain Sciences, ability. University College London Methods: 61 people with MS (pwMS) followed retrospectively with MRI including baseline suscep- (UCL), London, WC1B 5 EH, UK. tibility-weighted imaging, and longitudinal T1 and T2-weighted scans. SELs were computed using a.calvi@ucl.ac.uk deformation field maps; PRLs were visually identified. Mixed-effects models assessed differences in Alberto Calvi Queen Square MS Expanded Disability Status Scale (EDSS) score changes between the group defined by the presence of Centre, Department of SELs and or PRLs. Neuroinflammation, Institute of Neurology, Results: The median follow-up time was 3.2 years. At baseline, out of 1492 lesions, 616 were classi- Faculty of Brain Sciences, University College London fied as SELs, and 80 as PRLs. 92% of patients had ⩾ 1 SEL, 56% had ⩾ 1 PRL, while both were found (UCL), London, UK in 51%. SELs compared to non-SELs were more likely to also be PRLs (7% vs. 4%, p = 0.027). PRL Margareta A Clarke counts positively correlated with SEL counts (ρ= 0.28, p = 0.03). SEL + PRL + patients had greater Manel Alberich Deborah Pareto increases in EDSS over time (beta = 0.15/year, 95% confidence interval (0.04, 0.27), p = 0.009) than Alex Rovira Section of Neuroradiology, SEL+PRL-patients. Department of Radiology, Conclusion: SELs are more numerous than PRLs in pwMS. Compared with either SELs or PRLs found Hospital Universitari Vall d’Hebron, Universitat in isolation, their joint occurrence was associated with greater clinical progression. Autònoma de Barcelona, Barcelona, Spain Ferran Prados Keywords: Chronic active lesions, paramagnetic rim lesions (PRLs), slowly expanding lesions (SELs), Queen Square MS Centre, Department of volumetric MRI, susceptibility-weighted imaging (SWI), multiple sclerosis Neuroinflammation, Institute of Neurology, Faculty of Brain Sciences, Date received: 12 June 2022; revised: 9 November 2022; accepted: 11 November 2022 University College London (UCL), London UK/ Centre for Medical Image Computing (CMIC), Introduction clinical progression independent of relapses through Department of Medical Relapse-onset multiple sclerosis (MS) is character- neuro-axonal damage. Physics and Biomedical Engineering, University ised by the development of acute focal new lesions. College London, London, While acute inflammation typically lasts weeks to a On magnetic resonance imaging (MRI), the paramag- UK/e-Health Centre, Universitat Oberta de month, at pathology between 15% and 30% of lesions netic properties of iron-enriched macrophages and Catalunya, Barcelona, show evidence of chronic inflammatory activity, with microglia at the edge of lesions enable some chronic Spain Declan Chard such lesions termed chronic active or smouldering.1,2 active lesions to be identified on SWI (termed para- Olga Ciccarelli They show progressive tissue matrix damage in their magnetic rim lesions, PRLs). Hypointense rims sur- Queen Square MS Centre, Department of cores and have rims of iron-laden activated microglia/ rounding lesions have been described in both relapsing Neuroinflammation, macrophages with myelin breakdown and reactive and progressive MS using T2*-weighted or phase,5 Institute of Neurology, Faculty of Brain Sciences, astrocytes.3,4 Chronic active lesions are seen in all susceptibility-weighted imaging (SWI)6 as well as University College London phenotypes of MS and are thought to contribute to quantitative susceptibility mapping (QSM).7 It has (UCL), London, UK/ 352 journals.sagepub.com/home/msj A Calvi, MA Clarke et al. also been shown that some PRLs slowly expand over an observational study conducted at Centre Biomedical Research Centre, National Institute time more so than non-PRLs,8 despite other data d’Esclerosis Multiple de Catalunya and Section of for Health Research reporting that shrinkage can also be seen in PRLs.9 In Neuroradiology, Vall d’Hebron Barcelona Hospital (NIHR) and University College London MS, PRLs are estimated to occur in about 40% of Campus. The study received approval from the local Hospitals (UCLH), patients and on average make up ~10% of the overall Ethical Commitees of Vall d’Hebron Barcelona London, UK lesion count.10 However, limited MRI sensitivity in Hospital Campus and the Queen Square MS Centre, Marta Rodríguez Barranco detecting intracellular iron8 may suggest an underesti- University College London. The inclusion criteria Jaume Sastre-Garriga mation of the true number of such lesions. PRLs are were: (1) a confirmed diagnosis of relapsing-remit- Neurology- Neuroimmunology specific for MS11 and having four PRLs is associated ting (RRMS) or clinically isolated syndrome (CIS) Department, Multiple with greater motor and cognitive disability at an ear- according to the revised 2017 McDonald criteria;23 Sclerosis Centre of Catalonia (CEMCAT), lier age,12 but it remains uncertain if they preferen- (2) availability of baseline SWI and fluid-attenuated Vall d’Hebron Barcelona tially accumulate in progressive MS.13,14 inversion recovery (FLAIR) or T2-weighted scans; Hospital Campus, Barcelona, Spain (3) at least three consecutive 3D magnetisation-pre- Carmen Tur Chronic lesion activity can also be detected with pared rapid gradient-echo (MPRAGE) scans of ade- Queen Square MS Centre, Department of slowly expanding lesions (SELs) using deformation- quate image quality. Neuroinflammation, based volumetric MRI. As with PRLs, SELs are seen Institute of Neurology, Faculty of Brain in all MS phenotypes,15,16 appear more numerous with The scans were performed on a 3T system (Tim Trio; Sciences, University progressive (primary and secondary) progressive Siemens, Erlangen, Germany) using a 12-channel College London (UCL), London, UK/Neurology- compared with relapsing-remitting MS,15 and have phased-array head coil, with the following acquisition Neuroimmunology been associated with neurological and cognitive dis- parameters as previously described:6 (1) transverse Department, Multiple Sclerosis Centre of ability.17,18 SELs also represent a high fraction of the fast FLAIR (TR = 9000 ms, TE = 87 ms, TI = 2500 ms, Catalonia (CEMCAT), total lesion burden.18 SELs have a greater T1 hypoin- flip angle = 120°, voxel size = 0.49×0.49×3.0 mm3), Vall d’Hebron Barcelona Hospital Campus, tense volume than non-SELs,17,19 lower magnetisa- (2) sagittal T1-weighted 3D MPRAGE (TR = 2300 ms, Barcelona, Spain tion transfer ratio (MTR),20 more abnormal diffusivity TE = 2.98 ms, TI = 900 ms, voxel size = 1.0×1.0× Frederik Barkhof measures,21 and are more likely to become persistent 1.2 mm3) and (3) transverse SWI (TR = 33 ms, Queen Square MS Centre, Department of black holes (PBH),22 suggesting greater neuro-axonal TE1 = 6.08 ms, TE2 = 24.6 ms, flip angle = 15°, voxel Neuroinflammation, damage. While both represent chronic active lesions, size = 0.65×0.65×3 mm). Institute of Neurology, Faculty of Brain Sciences, it is unclear how SELs and PRLs are related, or have University College different associations with clinical outcomes. At each MRI session, Expanded Disability Status London (UCL), London, UK/Centre for Medical Scale (EDSS)24 scores were assessed by experienced Image Computing In this work, we hypothesised that (1) lesions with par- MS neurologists. Changes in EDSS scores were (CMIC), Department of Medical Physics and amagnetic rims would be more likely to expand and so assessed between baseline and the last MRI session. Biomedical Engineering, would be classified as SELs; (2) both lesion types Confirmed disability progression (CDP) was based on University College London, London, UK would be associated with a higher lesion volume and an EDSS change ⩾ 1.5 or ⩾ 1.0 when baseline EDSS Biomedical Research more abnormal brain-derived measures; and (3) that score was 0 or > 0, respectively, as previously Centre, National Institute for Health Research patients who had both PRLs and SELs would have a described,25 confirmed 6 months after the last MRI (NIHR) and University poorer clinical prognosis than those who did not. session. College London Hospitals (UCLH), London, UK/Radiology We aimed to: (1) investigate the association between & Nuclear medicine, VU University Medical PRLs, identified on the baseline SWI, and SELs, Lesion and brain segmentation, SEL detection Centre, Amsterdam, The computed using volumetric MRI, in relapse-onset Lesions were identified using the automated lesion Netherlands MS; (2) evaluate their impact on other MS-specific prediction algorithm from the Lesion Segmentation *counts positively correlated with SEL markers (total lesion volume and brain-derived met- Tool (LST),26 part of Statistical Parametric Mapping rics); (3) evaluate groups defined by the presence or (SPM) software package, based on the analysis of absence of PRLs and SELs; (4) assess the independ- baseline FLAIR images. The output of the LST seg- ent and combined contribution of PRLs and SELs to mentation was manually quality-checked and, if the evolution of disability. needed, corrected using Jim v7.0 (Xinapse Systems, Aldwincle, UK) by experienced raters (AC & MC). Materials and methods The 2D FLAIR images were resampled to 1 mm iso- tropic space. The 3D-MPRAGE images were co-reg- Participants, MRI acquisitions and clinical istered to the lesion masks. assessments Participants, who had all given informed consent for SEL-derived metrics were obtained with the in-house use of their data, were retrospectively selected from algorithm:18 lesions seen on FLAIR over 10 mm3 journals.sagepub.com/home/msj 353 Multiple Sclerosis Journal 29(3) were classified as SELs or non-SELs based on the subtype at the patient level, reporting median and presence or absence of detectable expansion (as meas- range, or mean and standard deviation (SD). Lesion ured by the Jacobian determinant) on T1-weighted volumes were log-transformed to produce more images. To account for differences in the follow-up normally distributed values. Associations between time intervals, the SEL pipeline incorporated a nor- measures were initially assessed with Spearman malisation step that computed the z-score of Jacobian. (‘ρ’) correlation coefficients. Then, we performed In previous work, we subdivided SELs into possible Poisson and linear regressions, for the counts and and definite categories, but here in our descriptive the volumes, and checked the normality of the resid- analyses we combined both into a single SEL cate- uals. A Chi-square test was used to assess the differ- gory, also referred to as SEL candidates.19 However, ences in the proportion of SEL and non-SEL lesions the models including the clinical variables were eval- that were PRLs. Patients were stratified according uated within all SEL sub-types. to the following criteria: ⩾ 1 SEL and ⩾ 1 PRL (SEL + PRL +); ⩾ 1 SEL and 0 PRLs (SEL + PRL−); For brain extraction, tissue segmentation and parcel- 0 SELs (SEL−). Since the number of SEL− patients lation, Geodesic Information Flows 27 was applied to was small, the subsets with and without PRLs were T1-weighted volumetric scans, as previously not considered separately. Differences in those sub- described.28 Using these segmentations, we obtained groups were assessed using linear regressions. The the normalised brain volume (NBV), normalised cor- analyses were repeated separating the cohort into tical grey matter (CGM) and normalised deep grey those with shorter (< 2 years) and longer (⩾ 2 years) matter (DGM) volumes. follow-ups. Mixed-effects regression models adjusted for age, PRL detection on SWI gender, relapse rate, baseline total lesion volume, and FLAIR and SWI images were rigidly aligned using treatment status (at the end of the study) were used to NiftyReg, and each lesion within the lesion mask was assess the relationship between the change in EDSS identified with a unique identification number that (dependent variable) and the MRI measures as inde- was kept after transforming the masks to SWI space pendent variables, i.e. counts and log-volumes of using nearest neighbour interpolation.28 The regis- SELs (possible and definite) as observed over the tered lesions were then assessed by two independent study period and presence of PRLs at baseline or their raters (AC and MC), using 3D Slicer, to identify PRLs co-occurrence. The interaction term between each (on SWI) and confirm the hyperintensity (on FLAIR). metric and the time at follow-up was used, and the The criteria for the definition of a PRL were: (1) a random effects included the subject identification partial or complete rim of hypointense signal relative number. Multiple logistic regressions adjusted for the to the lesion core and surrounding white matter; (2) same covariates were applied to investigate the risk of correspondence to the lesion’s edge on FLAIR; (3) a CDP explained by within-patient counts or log vol- rim visible on at least two consecutive slices. We umes of SELs (possible or definite). excluded PRLs that did not correspond to lesion edges or to hypointense areas on the FLAIR, and care was also taken not to mistake veins and signals from the Results white/grey matter border for PRLs. The raters inde- pendently marked 20 scans before the start of the Cohort demographics and clinical features study (Cohen’s Kappa = 0.87) and discussed why dif- The demographical and clinical characteristics of the ferences had arisen. They then independently cohort are presented in Table 1. The median age at reviewed all the remaining cases, and in the case of baseline was 34.4 years (range 14.1–64.9); the median any disagreements, it was adjudicated by two experi- disease duration was 0.4 years (range 0.1–16.6) and enced radiologists (AR and FB). the majority of patients were female (69%). The median time to the last follow-up MRI scan was 3.2 years (range 0.7–8.3 years). At baseline, 6 patients Statistical analysis were classified as having had a CIS and 55 were Analysis was performed with R and statistical signifi- relapsing-remitting (RRMS); at the last follow-up 6 cance was reported at p < 0.05. Lesions were individ- still had a CIS, 53 were RRMS and 2 converted to ually classified as a SEL or a non-SEL, and whether secondary-progressive multiple sclerosis (SPMS). they were also a PRL, allowing the coincidence of The median EDSS at baseline was 1.5 (range 0–4.5), SEL and PRL to be assessed at the lesion level. Lesion and the mean EDSS change was 0.16 (SD = 1.33). 14 counts and volumes were also analysed for each (23%) patients developed CDP by the end of the 354 journals.sagepub.com/home/msj A Calvi, MA Clarke et al. Table 1. Clinical-demographic characteristics of the patients enrolled in the study. Demographics and clinical features Number of patients 61 Female, n (% over total) 42 (69 %) Age at baseline, median (range), years 34.4 (14.1–64.9) Disease duration, median (range), years 0.4 (0.1–16.6) Time to intermediate scan, median (range), years 0.8 (0.4–6.4) Time to final scan, median (range), in years 3.2 (0.7–8.3) EDSS at baseline, median (range) 1.5 (0–4.5) EDSS at final scan, median (range) 1.5 (0–5.5) EDSS changea, mean (SD) 0.16 (1.33) MS phenotype at baseline CIS = 6; RRMS = 55 MS phenotype at last follow-up CIS = 6; RRMS = 53; SPMS = 2 Number of relapses between baseline and final scan, median (range) 0 (0–7) Time elapsed since the last relapse and the baseline scan, median (range), in days 114 (0–2783) Number (%) of patients with CDP* 14 (23%) Number (%) of patients treated at baseline 16 (26%) Number (%) of patients treated at last follow-up 51 (84%) Number (%) treated with first line DMT 33 (54%) Number (%) treated with first and second line DMTs 18 (30%) EDSS: Expanded Disability Status Scale; SD: standard deviation; MS: multiple sclerosis; CIS: clinically isolated syndrome; RRMS: relapsing-remitting multiple sclerosis; SPMS: secondary-progressive multiple sclerosis; CDP: confirmed disability progression; DMT: disease-modifying treatment. aEDSS change calculated as the difference between EDSS at final scan and EDSS at baseline. *CDP was defined by an EDSS change ⩾ 1.5 if baseline EDSS = 0 or EDSS change ⩾ 1.0 when baseline EDSS > 0, respectively, and confirmed at least > 6 months after the last scan. Table 2. Patient subgroups according to SEL and PRLs. Presence of SELs SEL+ (⩾1 SEL) SEL− n = 56 n=5 Presence of PRLs PRL + 31 3 (⩾ 1 PRL) n = 34 PRL− 25 2 n = 27 SEL: slowly expanding lesion; PRL: paramagnetic rim lesion. study. At baseline, 16% of patients were receiving a without SELs (SEL−, n = 5, 8%). None of the SELs, disease-modifying treatment (DMT), while 84% had PRLs or brain volume measures was different been treated with a DMT at the final follow-up. between the group followed up for < 2 years (n = 15) compared to those followed up for ⩾ 2 years (n = 46) (Supplementary Table 1). The prevalence of SELs and PRLs at a patient level Nearly all patients had at least one SEL (92%, n = 56), SELs and PRLs, and their co-localisation at a while slightly more than half had ⩾ 1 PRL (56%, lesion level n = 34). Table 2 shows that the majority of patients Overall 1492 lesions were identified. Among those, were SEL + PRL + (n = 31, 51%), followed by 616 were SELs (41%) and 876 were non-SEL (59%). SEL + PRL− (n = 25, 41%) and finally, patients On the baseline SWI, 80 PRLs were identified, journals.sagepub.com/home/msj 355 Multiple Sclerosis Journal 29(3) Table 3. Lesion-level analysis of frequency of SEL-derived lesions by PRL. Lesion type PRL No PRLs Proportion PRL over lesions of the corresponding type Total lesions (n = 1492) 80 1412 0.05 SEL (n = 616) 43 573 0.07 Non-SEL (n = 876) 37 839 0.04 PRL: paramagnetic rim lesion; SEL: slowly expanding lesion. Figure 1. Stacked bar plot showing the total number of lesions classified as SEL or non-SEL, and the correspondent number of PRL within each subtype. The figure, drawn with R, shows the counts of lesions divided between the SEL and the non-SEL categories, and the corresponding PRLs in each subtype. The proportion of PRLs was higher among the SELs than among the non-SELs (7% vs. 4%, p = 0.027). corresponding to 5% of the total lesions (Table 3). 43 association between baseline BPF and SEL volumes lesions were both SELs and PRLs, representing 54% (ρ = −0.48, p < 0.001), which was accounted for by (43/80) of all the PRLs and 7% (43/616) of all the an association with total lesion volumes in the partial SELs (Figure 1). The proportion of PRLs was higher correlations. No significant associations were found among the SELs than among the non-SELs (7% vs. between SEL or PRL measures with the regional 4%, p = 0.027). An example of a lesion that corre- baseline brain volumes (CGM, DGM). sponds to both a SEL and a PRL is shown in Figure 2. MRI measures in patient subgroups based on the Associations between SELs, PRLs and brain presence of SEL and PRLs volume measures Table 5 presents a descriptive analysis of MRI and The MRI characteristics of the cohort at the patient clinical measures among patients classified as level are presented in Table 4. A positive correlation SEL + PRL +, SEL + PRL− and SEL−. Within the between SEL counts and SEL volumes, with baseline subgroups, SEL + PRL + patients had the highest PRL counts was seen (ρ= 0.28, p = 0.03 and ρ= 0.29, total lesion counts (median 28, p = 0.001) and the p = 0.02, respectively). However, when we performed highest total lesion volume (median 3.7 ml, a partial correlation between these measures, taking p = 0.005). Consistent with this, in a multiple linear into account the total lesion burden, no significant regression model, the SEL category was associated associations were found. There was a negative with total lesion volume (beta = 5.7 ml, 95% 356 journals.sagepub.com/home/msj A Calvi, MA Clarke et al. Figure 2. Example of a PRL at baseline and the correspondence to a SEL. The images show from left to right: SWI at baseline (the top left rectangle shows a magnification of the paramagnetic rim lesion), FLAIR with lesion mask countered in red, Jacobian map superimposed indicating that the PRL corresponds to a SEL. The images are from a 45-year-old person with a CIS and the MRI was performed 4 months after onset. At baseline the EDSS was 0 and last follow- up (~4 years) was 1.5. CIS: clinically isolated syndrome; EDSS: Expanded Disability Status Scale; PRL: paramagnetic rim lesion; FLAIR: fluid-attenuated inversion recovery; PRL: paramagnetic rim lesion; SWI: susceptibility-weighted imaging; SEL: slowly expanding lesion. Table 4. Radiological measures: counts and volumes and brain-derived measures. Radiological measures Total lesion count at baseline (n), median [range] 20 [1–80] SEL count (n), median [range] 5 [0–41] non-SEL count (n), median [range] 10 [0–55] PRL count at baseline (n), median [range] 1 [0–8] PRL/total lesion count ratio, mean (SD) 0.05 (0.10) SEL/total lesion count ratio, mean (SD) 0.41 (0.23) Total lesion volume at baselinea, median [range], ml 2.2 [0.1–67.0] Total SEL volume, median [range], ml 0.6 [0–65.3] Non-SEL volume, median [range], ml 0.9 [0–27.1] Total SEL/total lesion volume ratio, mean (SD) 0.39 (0.28) PRL volume at baseline, median [range], ml 0.03 [0–1.0] PRL/total lesion volume ratio, mean (SD) 0.05 (0.09) Gadolinium-enhancing lesions at baseline, count (n), median [range] 0 [0–40] New lesions at final session, count (n), median [range] 1 [0–24] NBV volume at baseline, mean (SD), ml 1505.7 (75.1) CGM volume at baseline, mean (SD), ml 839.9 (47.9) DGM volume at baseline, mean (SD), ml 49.4 (3.7) BPF at baseline, mean (SD) 0.77 (0.02) SEL: slowly expanding lesion; PRL: paramagnetic rim lesion; SD: standard deviation; NBV: normalised brain volume; CGM: cortical grey matter; DGM: deep grey matter; BPF: brain parenchymal fraction. All the measures are referred to the patient level. aTotal lesion volume, and as consequence all the SEL-derived volumes (as subsets of the total volume) were based on the segmentation on the FLAIR image. journals.sagepub.com/home/msj 357 Multiple Sclerosis Journal 29(3) Table 5. Clinical-demographic and radiological characteristics in groups according to presence of SEL and PRLs. SEL + PRL + SEL + PRL− SEL− p-value* (n = 31) (n = 25) (n = 5) EDSS at baseline, median [range] 1.5 [0; 4.5] 1 [0; 3] 1 [0; 2] 0.187a EDSS at final follow-up, median [range] 1.5 [1; 5.5] 1 [0; 5.5] 1 [1; 2] 0.106a EDSS change, mean (SD) 0.27 (1.12) 0.06 (1.62) 0.00 (1.00) 0.806a Age at baseline, median [range] 37.2 [17.7; 67.6] 34.2 [17.5; 47.4] 35.7 [23.5; 41.9] 0.325a Disease duration, median [range] 0.2 [0.1–16.7] 0.4 [0.1–13.5] 0.3 [0.1; 0.3] 0.518a Total lesion count, median [range] 28 [4–80] 9 [1–77] 5 [1–24] 0.006a Non-SEL count, median [range] 15 [1–55] 5 [0–44] 5 [1–24] 0.029a SEL count, median [range] 9 [1–41] 4 [1–39] 0 [0] 0.008a Total lesion volume, median [range] 3.7 [0.5–66.9] 0.7 [0.02–6.0] 0.5 [0.1–27.1] 0.005a Non-SEL volume, median [range] 2.1 [0.1–18.3] 0.4 [0–2.5] 0.5 [0.1–27.1] 0.020a SEL volume, median [range] 1.2 [0.1–65.3] 0.2 [0.01–4.2] 0 [0] 0.007a NBV, mean (SD) 1497.2 (83.4) 1512.4 (64.8) 1525.1 (77.9) 0.636a CGM volume, mean (SD) 834.2 (55.8) 846.5 (38.6) 841.6 (39.6) 0.638a DGM volume, mean (SD) 49.0 (4.2) 49.8 (3.0) 50.5 (3.8) 0.552a BPF, mean (SD) 0.76 (0.02) 0.77 (0.01) 0.77 (0.02) 0.130a SEL: slowly expanding lesion; PRL: paramagnetic rim lesion; EDSS: expanded disability status scale; SD: standard deviation; NBV: normalised brain volume; CGM: cortical grey matter; DGM: deep grey matter; BPF: brain parenchymal fraction. aUnivariate linear regression. *Bold indicates significant results (p-value <0.05). confidence interval (CI) (0.8, 10.6), p = 0.025), but SEL and PRL associations with EDSS progression not baseline NBV or regional brain volumes. The In mixed-effects regression models (see Supplementary percentages of pwMS who changed from untreated Table 5), a greater EDSS increase over time was asso- to treated from baseline to the end of the study were ciated with higher definite SEL counts (beta = 0.01/ not different between the SEL + PRL + and the year, 95% CI (0.001, 0.03), p = 0.045) and definite SEL + PRL− groups (Supplementary Table 2). SEL volumes (beta = 0.01/year, 95% CI (0.001, 0.01), PwMS who were free from relapses within the p = 0.044). When adjusted for age at baseline, sex and 3 months before study entry (n = 41) did not reveal disease duration, treatment status, and baseline lesion significant differences in the MRI measures (e.g. counts and volumes, the total lesion counts and vol- number of SELs and PRLs) compared to those with umes were also associated with EDSS progression. relapses in the previous 3 months (Supplementary The presence of ⩾ 1 PRL was associated with a Table 3). In addition, pwMS with ⩾ 5 gadolinium- greater increase in EDSS over time (beta = 0.15/year, enhancing lesions had similar MRI characteristics to 95% CI (0.01, 0.30), p = 0.044) relative to pwMS those with < 5 gadolinium-enhancing lesions without PRLs. The SEL + PRL + subgroup had (Supplementary Table 4). greater increases in EDSS scores over time than SEL+ PRL− subgroup (SEL + PRL + beta = 0.15/ year, 95% CI (0.04, 0.27), p = 0.009 vs. SEL+ PRL− The associations of SELs and PRLs with beta = −0.02/year 95% CI (−0.13, 0.08), p = 0.678). demographics No differences were found between patients in the SEL counts were positively associated with the SEL + PRL + and SEL− subgroups (beta = −0.11/ patients’ age at baseline (ρ= 0.37, p = 0.004) as well as year, 95% CI (−0.44, 0.22), p = 0.515) or between the SEL volume (ρ= 0.36, p = 0.004). SEL lesion vol- SEL + PRL− and SEL− (beta =-0.09/year, 95% CI umes were higher in pwMS with longer disease dura- (−0.42, 0.23) p = 0.575). When the model was restricted tion (beta = 0.10, 95% CI (0.07, 0.14), p < 0.001) and to the subgroup of subjects with < 5 gadolinium- who were older at baseline (beta = 3.0 95% CI (1.7, enhancing lesions, the results did not change. In addi- 4.3), p < 0.001). PRL lesion count and volume were tion, the exclusion of pwMS who experienced a not associated with any of the demographic factors relapse in the previous 3 months showed similar trends explored. No associations were found between gender of EDSS increase for the SEL + PRL + groups in com- and either SEL-derived or PRL measures. parison to the SEL + PRL− (adjusted difference 358 journals.sagepub.com/home/msj A Calvi, MA Clarke et al. beta = 0.21, 95% CI (0.01, 0.43), p = 0.059). In the PRLs expand by > 10% of their initial volume.9 logistic model, predicting CDP, SEL and PRL counts Despite the PRLs and SELs being associated in our and volumes were not statistically significant. work, their correlation appears to be mainly driven by the total lesion load, consistent with previous observations.9,12 Discussion In this work, we assessed the associations between This study found associations of lower BPF with MRI markers of the chronic active lesions in pwMS. higher SEL volumes, suggesting a potential link We found a moderate association between slowly between chronic inflammation and global neuro- expanding lesions and SWI rim lesions, and that SELs axonal damage. However, no clear differences in outnumber PRLs by a factor of 10. In addition, this brain volume measures were found between the work suggests that the co-occurrence of both MRI SEL+ PRL+ and other subgroups, which may be due markers is an adverse prognostic factor. to the short disease duration of the present cohort. In our MS cohort, we found that 92% developed at From a clinical perspective, our results suggest that least one SEL over a heterogeneous follow-up time there is an association between SELs, higher age at (median 3, range 0.7–8.3 years), which was a higher baseline and longer disease duration, which are well- percentage compared to previous studies.29,16 known predictors for a worse clinical outcome in However, the proportion of patients with at least one MS.31,32 The presence of SEL and PRLs in combina- SEL was similarly high in patients with acute lesion tion predicted clinical disability, as measured by the activity (89% in Beynon et al.19). In our study, the EDSS, suggesting that as additive factors they might SEL detection rate did not seem to be affected by the be involved in the perpetuation of inflammation and presence of patients with high inflammatory activity reduced remyelination. The early deposition of para- (n = 7 had ⩾ 5 gadolinium-enhancing lesions). magnetic materials within PRLs could indicate a spe- Alternative explanations include differences in the cific adverse pathogenetic process, such as a more SEL detection technique or in the clinical characteris- aggressive macrophages-microglia profile,33 favour- tics of the population. On the other hand, 56% had at ing expansion and conversion to SEL. least one PRLs at SWI, in line with previous reports.6,12 Despite being early in their disease course, patients on This work has several limitations. First of all, the average had 5 SELs (40% of the overall lesions), MRI scans were retrospectively collected, thus comparable to more heterogeneous MS phenotypes (4 increasing the heterogeneity of the cohort. In addi- to 19.5 SELs per patient).15,18,30 By contrast, only a tion, the acquisitions of the SWI images were in 2D, low number of baseline lesions were PRLs (80/1492, which could impact the visualisation of PRLs. 5%), suggesting that close to disease onset only a However, the use of 3D T1-weighted images for the small subset are characterised by ring-like signal. A SEL computation ensured a superior spatial resolu- recent study evaluating up to 7 years after baseline has tion. Second, the short period of observation for this shown that PRLs can persist in progressive disease, cohort could have impacted the accuracy of SEL and a half (50%) of them show a slow volume growth detection. Third, the criteria for SEL definition and and exhibit more destructive MRI features, as assessed pipeline used differ between centres, thus making by prolonged T1 relaxation times.9 However, it was comparisons across studies difficult. Due to the high also suggested that, after a period of growth, PRLs prevalence of SELs within the cohort, an assessment stabilise with reduced susceptibility and rim attenua- of patients based on the presence or absence of SELs tion thereafter, and eventually some lose their rims. alone could be of limited clinical value. From the distribution of the SEL counts observed, a cut-off of This work indicates a moderate association between four SELs may be worth considering in future stud- PRLs and SELs in recently diagnosed relapse-onset ies. This study was also limited by having a baseline MS. Despite this, the proportion of PRLs that were SWI scan only and the temporal evolution of PRLs also classified as SELs exceeded the proportion of could not be assessed or compared with SELs. Other PRLs classified as non-SELs. Moreover, the group of studies, however, have not found a large change in patients having both SELs and PRLs (SEL+ PRL+) the proportion of PRL over time.9 The relationship had the highest lesion counts and volumes. On 7T between contrast enhancement, SELs and PRLs, was MRI, an association between lesion expansion and not assessed. However, when we excluded the group PRLs was previously found.8 Newly developed PRLs of pwMS with an outlier gadolinium lesion count or have been shown to expand by ~30% in volume over relapse activity in the last 3 months, the results of the 3 years,8 while after 7–10 years on average 50% of models assessed remained materially unchanged. As journals.sagepub.com/home/msj 359 Multiple Sclerosis Journal 29(3) a recently diagnosed MS cohort, they had a limited J. Sastre-Garriga serves as co-Editor for Europe on disability progression, and only a small minority the editorial board of Multiple Sclerosis Journal and converted to a secondary-progressive form. As a as Editor-in-Chief in Revista de Neurología, receives result, the models assessing disability had small research support from Fondo de Investigaciones effect sizes or were borderline significant. Finally, Sanitarias (19/950) and has served as a consultant/ assessing the effects of treatment exposure was not speaker for Biogen, Celgene/Bristol Meyers Squibb, an objective of the study, due to the relatively small Genzyme, Novartis and Merck. sample size. However, the percentages of patients C. Tur has received 2021 Merck’s Award for the who changed treatment over the course were similar Investigation in Multiple Sclerosis, Junior Leader La and the treatment status at the end of the study was Caixa Fellowship in 2020, ECTRIMS Post-doctoral not significantly associated with the EDSS, which Research Fellowship in 2015; honoraria and support suggests that for our sample treatment was unlikely for travelling from Merck Serono, Sanofi, Roche, to have had a substantial effect on the group differ- TEVA Pharmaceuticals, Novartis, Biogen, Bayer, ences observed. Ismar Healthcare. A. Rovira serves on scientific advisory boards for In conclusion, this study suggests that SELs are a Novartis, Sanofi-Genzyme, Synthetic MR, common finding and a subset of them can coincide TensorMedical, Roche, Biogen, and OLEA Medical, with PRLs. Despite a modest association of PRLs and and has received speaker honoraria from Bayer, SELs, their combination, in association with an over- Sanofi-Genzyme, Merck-Serono, Teva Pharmaceutical all increase in lesion burden, could be of prognostic Industries Ltd, Novartis, Roche, Bristol-Myers and relevance as a predictor of MS severity early in the Biogen. disease course. F. Barkhof is supported by the NIHR Biomedical Research Centre initiative at UCLH, and he serves on Disclosures the editorial boards of Brain, European Radiology, A. Calvi is supported by the ECTRIMS post-doc fel- Journal of Neurology, Neurosurgery & Psychiatry, lowship (2022), previously received a UK MS Society Neurology, Multiple Sclerosis, and Neuroradiology, PhD studentship (2020), a Guarantors of Brain ‘Entry’ in the steering committee and iDMC member for clinical fellowship (2019), and an ECTRIMS- Biogen, Merck, Roche, EISAI. Consultant for Roche, MAGNIMS fellowship (2018). Biogen, Merck, IXICO, Jansen, Combinostics. M. A. Clarke was previously a National Multiple Research agreements with Novartis, Merck, Biogen, Sclerosis Society (RG-1901-33190) post-doctoral fel- GE, Roche. Co-founder and shareholder of Queen low (2019–2022) and an ECTRIMS-MAGNIMS fel- Square Analytics LTD. low (2018). F. Prados received a Guarantors of Brain fellowship Declaration of Conflicting Interests 2017–2020 and is supported by National Institute for The author(s) declared no potential conflicts of inter- Health Research (NIHR), Biomedical Research est with respect to the research, authorship, and/or Centre initiative at University College London publication of this article. Hospitals (UCLH). D. Chard is a consultant Hoffmann-La Roche. In the Funding last 3 years, he has been a consultant for Biogen, The author(s) received no financial support for the has received research funding from Hoffmann-La research, authorship, and/or publication of this Roche, the International Progressive MS Alliance, article. the MS Society, and the NIHR UCLH Biomedical Research Centre, and a speaker’s honorarium from Data Availability Novartis. Fully anonymised data and analyses are available O. Ciccarelli received research funding from NIHR after review by University College London and Biomedical Research Centre initiative at UCLH, UK Hospital Universitari Vall d’Hebron, after completing and National MS Societies, Rosetrees trust; she is an an application form detailing specific requirements, Associate Editor for Neurology. rationale, and proposed use followed by a data-shar- D. Pareto DP has received a research support from ing agreement. Instituto Carlos III (PI18/00823) and a research con- tract from Biogen Idec. ORCID iDs M. Alberich has nothing to disclose in relation to this Alberto Calvi https://orcid.org/0000-0002-1953- work. 2803 360 journals.sagepub.com/home/msj A Calvi, MA Clarke et al. Margareta A Clarke https://orcid.org/0000-0001- 9. Dal-Bianco A, Grabner G, Kronnerwetter C, et al. 5531-796X Long-term evolution of multiple sclerosis iron rim Declan Chard https://orcid.org/0000-0003-3076- lesions in 7 T MRI. Brain 2021; 144(3): 833–847. 2682 10. Ng Kee Kwong KC, Mollison D, Meijboom R, Jaume Sastre-Garriga https://orcid.org/0000-0002- et al. The prevalence of paramagnetic rim lesions in 1589-2254 multiple sclerosis: A systematic review and meta- Carmen Tur https://orcid.org/0000-0003-1849- analysis. PLoS ONE 2021; 16: e0256845. 3184 11. Meaton I, Altokhis A, Allen CM, et al. Alex Rovira https://orcid.org/0000-0002-2132- Paramagnetic rims are a promising diagnostic 6750 imaging biomarker in multiple sclerosis. Mult Scler J. Epub ahead of print 26 August 2022. DOI: Supplemental Material 10.1177/13524585221118677. Supplemental material for this article is available 12. Absinta M, Sati P, Masuzzo F, et al. Association online. of chronic active multiple sclerosis lesions with disability in vivo. JAMA Neurol 2019; 76: 1474– 1483. 13. Mehta V, Pei W, Yang G, et al. Iron is a sensitive References biomarker for inflammation in multiple sclerosis 1. Luchetti S, Fransen NL, van Eden CG, et al. lesions. PLoS ONE 2013; 8(3): e57573. Progressive multiple sclerosis patients show 14. Harrison DM, Li X, Liu H, et al. Lesion heterogeneity substantial lesion activity that correlates with clinical on high-field susceptibility MRI is associated with disease severity and sex: A retrospective autopsy multiple sclerosis severity. AJNR Am J Neuroradiol cohort analysis. Acta Neuropathol 2018; 135(4): 2016; 37(8): 1447–1453. 511–528. 15. Elliott C, Wolinsky JS, Hauser SL, et al. Slowly 2. Frischer JM, Weigand SD, Guo Y, et al. Clinical and expanding/evolving lesions as a magnetic resonance pathological insights into the dynamic nature of the imaging marker of chronic active multiple sclerosis white matter multiple sclerosis plaque. Ann Neurol lesions. Mult Scler 2019; 25(14): 1915–1925. 2015; 78(5): 710–721. 16. Preziosa P, Pagani E, Meani A, et al. Slowly 3. Bagnato F, Hametner S, Yao B, et al. Tracking expanding lesions predict 9-year multiple sclerosis iron in multiple sclerosis: A combined imaging and disease progression. Neurol Neuroimmunol histopathological study at 7 tesla. Brain 2011; 134(Pt. Neuroinflamm 2022; 9(2): e1139. 12): 3602–3615. 17. Elliott C, Belachew S, Wolinsky JS, et al. Chronic 4. Popescu BF, Frischer JM, Webb SM, et al. white matter lesion activity predicts clinical Pathogenic implications of distinct patterns of iron progression in primary progressive multiple sclerosis. and zinc in chronic MS lesions. Acta Neuropathol Brain 2019; 142(9): 2787–2799. 2017; 134(1): 45–64. 18. Calvi A, Carrasco FP, Tur C, et al. Association of 5. Yao B, Ikonomidou VN, Cantor FK, et al. slowly expanding lesions on MRI with disability in Heterogeneity of multiple sclerosis white matter people with secondary progressive multiple sclerosis. lesions detected with T2*-weighted imaging at 7.0 Neurology 2022; 98(17): e1783–e1793. tesla. J Neuroimaging 2015; 25(5): 799–806. 19. Beynon V, George IC, Elliott C, et al. Chronic lesion 6. Clarke MA, Pareto D, Pessini-Ferreira L, et al. activity and disability progression in secondary Value of 3T susceptibility-weighted imaging in progressive multiple sclerosis. BMJ Neurol Open the diagnosis of multiple sclerosis. AJNR Am J 2022; 4(1): e000240. Neuroradiol 2020; 41(6): 1001–1008. 20. Elliott C, Belachew S, Fisher E, et al. MRI 7. Zhang Y, Gauthier SA, Gupta A, et al. Quantitative characteristics of chronic MS lesions by phase rim susceptibility mapping and R2* measured changes detection and/or slowly expanding properties (4101). during white matter lesion development in multiple Neurology 2021; 96(15 Suppl.): 4101. sclerosis: Myelin breakdown, myelin debris degradation and removal, and iron accumulation. 21. Klistorner S, Barnett MH, Yiannikas C, et al. AJNR Am J Neuroradiol 2016; 37(9): 1629–1635. Expansion of chronic lesions is linked to disease progression in relapsing–remitting multiple sclerosis 8. Dal-Bianco A, Grabner G, Kronnerwetter C, et al. patients. Mult Scler 2021; 27(10): 1533–1542. Slow expansion of multiple sclerosis iron rim lesions: Pathology and 7 T magnetic resonance imaging. Acta 22. Calvi A, Tur C, Chard D, et al. Slowly expanding Neuropathol 2017; 133(1): 25–42. lesions relate to persisting black-holes and clinical journals.sagepub.com/home/msj 361 Multiple Sclerosis Journal 29(3) outcomes in relapse-onset multiple sclerosis. lesion filling in multiple sclerosis. Neuroimage 2016; Neuroimage Clin 2022; 35: 103048. 139: 376–384. 23. Thompson AJ, Banwell BL, Barkhof F, et al. 29. Elliott C, Arnold DL, Chen H, et al. Patterning Diagnosis of multiple sclerosis: 2017 revisions of chronic active demyelination in slowly expanding/ the McDonald criteria. Lancet Neurol 2018; 17(2): evolving white matter MS lesions. AJNR Am J 162–173. Neuroradiol 2020; 41(9): 1584–1591. 24. Kurtzke JF. Rating neurologic impairment in multiple 30. Elliott C, Arnold DL, Chen H, et al. Patterning sclerosis: An expanded disability status scale (EDSS). chronic active demyelination in slowly expanding/ Neurology 1983; 33(11): 1444–1452. evolving white matter MS lesions. AJNR Am J 25. Coles AJ, Twyman CL, Arnold DL, et al. Neuroradiol 2020; 41(9): 1584–1591. Alemtuzumab for patients with relapsing multiple 31. University of California San Francisco MS-EPIC sclerosis after disease-modifying therapy: A Team: Cree BA, Gourraud PA, Oksenberg JR, et al. randomised controlled phase 3 trial. Lancet 2012; Long-term evolution of multiple sclerosis disability 380(9856): 1829–1839. in the treatment era. Ann Neurol 2016; 80(4): 26. Roura E, Oliver A, Cabezas M, et al. A toolbox 499–510. for multiple sclerosis lesion segmentation. 32. Dekker I, Eijlers AJC, Popescu V, et al. Predicting Neuroradiology 2015; 57(10): 1031–1043. clinical progression in multiple sclerosis after 6 and 27. Cardoso MJ, Modat M, Wolz R, et al. Geodesic 12 years. Eur J Neurol 2019; 26(6): 893–902. information flows: Spatially-variant graphs and their 33. Zarruk JG, Berard JL, Passos dos Santos R, et al. application to segmentation and fusion. IEEE Trans Expression of iron homeostasis proteins in the spinal Visit SAGE journals online Med Imaging 2015; 34(9): 1976–1988. journals.sagepub.com/ cord in experimental autoimmune encephalomyelitis home/msj 28. Prados F, Cardoso MJ, Kanber B, et al. A multi- and their implications for iron accumulation. SAGE journals time-point modality-agnostic patch-based method for Neurobiol Dis 2015; 81: 93–107. 362 journals.sagepub.com/home/msj