Tissue damage detected by quantitative gradient echo MRI correlates with clinical progression in non-relapsing progressive MS

Abstract

Background:

Imaging biomarkers of progressive MS are needed. Quantitative gradient recalled echo (qGRE) MRI evaluates microstructural tissue damage in MS.

Objective:

To evaluate qGRE-derived R2t* as an imaging biomarker of MS progression compared to atrophy and lesion burden.

Methods:

Twenty-three non-relapsing progressive MS (PMS), twenty-two relapsing-remitting MS (RRMS) and eighteen healthy control participants underwent standard MS physical and cognitive neurological assessments and imaging with qGRE, FLAIR and MPRAGE at 3T. PMS subjects were tested clinically and imaged every nine months over forty-five months. Imaging measures included lesion burden, atrophy and R2t* in cortical grey matter (GM), deep GM, and normal-appearing white matter (NAWM). Longitudinal analysis of clinical performance and imaging biomarkers in PMS subjects was conducted via linear models with subject as repeated, within-subject factor. Relationship between imaging biomarkers and clinical scores was assessed by Spearman rank correlation.

Results:

R2t* reductions correlated with neurological impairment cross-sectionally and longitudinally. PMS patients with clinically defined disease progression (N=13) showed faster decrease of R2t* in NAWM and deep GM compared with the clinically stable PMS group (N=10). Importantly, tissue damage measured by R2t* outperformed lesion burden and atrophy as a biomarker of progression during the study period.

Conclusion:

qGRE-derived R2t* is a potential imaging biomarker of MS progression.

1. Introduction

Magnetic resonance imaging (MRI) plays an important role in multiple sclerosis (MS) diagnosis and monitoring ^1-3. Identification of new or gadolinium enhancing white matter (WM) lesions (WMLs) by means of T1- or T2-weighted images represents disease activity and serves as an end-point for trials of disease-modifying therapies ⁴. However, this approach to disease monitoring applies primarily to relapsing remitting MS (RRMS). Progressive MS (PMS) subtypes [non-relapsing secondary progressive (SP) MS and primary progressive (PP) MS] lack highly sensitive imaging biomarkers of progression due in part to the less inflammatory pathology of PMS ^{5, 6}. The best available biomarker of PMS currently is atrophy, especially of grey matter (GM) ⁷. However, atrophy represents the end-result of multiple pathological processes. More specific biomarkers which measure pathology prior to the manifestation of frank atrophy are needed ⁸.

Here, we assess quantitative gradient recalled echo (qGRE) imaging to measure R2t* (=1/T2t*) as a potential biomarker of progression in MS. R2t* is a component of the more commonly reported R2* (=1/T2*) measure. Previous studies showed that R2* (1/T2*) differentiates RRMS patients from healthy controls and correlates with clinical scores ^9-11. R2*, however, is also sensitive to the baseline blood oxygenation level dependent (BOLD) signal usually characterized by parameter R2’ (R2*=R2t*+R2’) that depends on physiological conditions and can fluctuate within and between scans. Therefore, we developed the qGRE approach to estimate R2t*, the sub-component of R2* most related to tissue microstructural integrity and unaffected by the BOLD signal ^12-14. R2t* correlates with neuronal density distribution across neocortex in healthy subjects ¹⁵, as well as in histopathologically verified inflammatory demyelination of WM ¹⁶. Further, decreased R2t* detects limbic system abnormalities in MS not associated with volumetric changes ¹⁷ and correlates with upper extremity dysfunction and cognitive impairment ¹⁸. Thus, qGRE R2t* is a promising biomarker of MS-related tissue damage within and outside of lesions. We hypothesized that qGRE R2t* would reflect disease progression in non-relapsing PMS patients.

This study addresses the ability of R2t* to reflect concurrent clinical progression and compares the utility of R2t* to volume loss and lesion burden imaging biomarkers. We compare these measures in age-matched PMS, RRMS and healthy subjects and longitudinally in relation to advancing impairment in PMS patients. Results from this pilot study are expected to motivate future studies using R2t*, particularly as a potential marker of progression unrelated to relapses, and to provide insight into MS-related tissue damage.

2. Materials and methods

2.1. Study Participants

Eighteen healthy control (HC), twenty-three PMS (including four PPMS and nineteen SPMS), and twenty-two RRMS subjects were enrolled after providing informed consent. The study was approved by the Institutional Review Board. PMS patients had no relapses or gadolinium enhancing lesions within twenty-four months preceding enrollment. RRMS and HC subjects were recruited to reflect the age and sex distribution of the PMS patients to facilitate comparisons. All MS subjects in this study were recruited from John L. Trotter Multiple Sclerosis Center patients with reliable histories to confirm the clinical subtype, and MRIs within one year. RRMS subjects had not received corticosteroids for relapses within 6 months of entry. RRMS subjects had no history of disability progression unless directly related to an MS attack. The HC and RRMS subjects were scanned one time at the beginning of the study. PMS subjects were scanned every nine months for five sessions with the goal of identifying imaging biomarkers that correlated best with or predicted clinical progression in PMS patients.

One PMS patient was lost to follow up after the second session, two following the third session, and one following the fourth session. Six imaging sessions were excluded due to severe motion artifacts and six sessions due to technical difficulties. In total, ninety-two sessions (out of 115 possible) in the PMS cohort were used for data analysis.

2.2. Clinical testing

Clinical evaluation was performed on the day of MRI exam and included the Expanded Disability Status Scale (EDSS) standardized neurological examination, 25-foot timed walk (25FTW) assessment of gait, nine-hole peg test (9HPT) assessment of bilateral upper extremity function, paced auditory serial addition test (PASAT) and symbol digit modalities test (SDMT) assessments of cognitive function. Clinical assessment was performed by examiners blinded to the imaging results. For data analyses, the 25FTW and 9HPT raw scores were converted to Z-scores according to the Multiple Sclerosis Functional Composite (MSFC) guidelines ¹⁹.

2.3. Definition of clinical progression during the study interval

The PMS subjects were classified as either clinically stable or progressive across the entire study period (45 months) based on EDSS, 25FTW and 9HPT scores ²⁰. The definition of progression required that it be sustained to the end of the study period. Four subjects developed sustained progression by the second visit, five by the third, and four by the fourth. No subjects were determined to have progressed on the basis of the final study visit. EDSS score increases by one point were defined as progression if the subject’s baseline EDSS was less than 5.5; for baseline EDSS 5.5 or greater, one-half point increases defined progression. For the 25FTW score, ≥ 20% increase or converting from walking to not walking defined progression. For the 9HPT score, increasing by ≥20% or converting from able to unable to complete the test defined progression. Those PMS subjects who did not show progression in EDSS, 25FTW or 9HPT were defined as clinically stable.

2.4. Image acquisition

All subjects were imaged on a 3.0 Tesla Trio MRI scanner (Siemens, Erlangen, Germany) using a 32-channel phased-array radiofrequency head coil. qGRE data were acquired using three-dimensional multi-gradient-echo sequence with flip angle 30°, TR=50ms, voxel size 1×1×2mm³ and acquisition time 12 minutes. 10 gradient echoes, with first echo time TE1=4ms and echo spacing ΔTE=4ms were collected. In addition, fluid-attenuated inversion recovery (FLAIR) images (voxel size 1×1×3mm³) and magnetization-prepared rapid gradient-echo (MP-RAGE) (voxel size 1×1×1mm³) were acquired for purposes of atlas registration and tissue segmentation.

2.5. Image processing and segmentation

R2t* maps were generated using the qGRE approach described in a previous study ¹⁵. In brief, multi-channel data were combined using a previously published algorithm ⁹. Voxel-wise analysis was performed on combined data using the theoretical model of GRE signal relaxation ¹² and a set of post-processing algorithms that minimize adverse artifacts related to macroscopic magnetic field inhomogeneities ²¹ and physiological fluctuations ¹³.

Image segmentation was achieved via a two-step process. First, MP-RAGE and FLAIR images were subjected to FreeSurfer longitudinal pipe-line analysis (version 6.0) ²² and segmentations were reviewed and errors corrected by a trained technician (MK). FreeSurfer does not accurately segment WMLs. Therefore, a recently developed intensity-based approach was used to identify WMLs ²³. For subsequent analyses, voxels were labeled as CSF or GM if both segmentation strategies agreed. Cortical and deep GM (caudate, putamen, globus pallidus and thalamic nuclei) were separated based on FreeSurfer segmentation. Normal Appearing White Matter (NAWM) was defined as voxels in which both segmentations returned a WM label. Voxels identified as lesions by both approaches or labeled as WM by FreeSurfer and lesion by the intensity-based approach were defined as lesion for the present analysis.

FreeSurfer segmentations were used to estimate the volume of brain structures for the purpose of calculating atrophy (loss of volume over time). All measures were corrected via regression for the total intracranial volume. Cortical volume (atrophy) was defined as the sum (rate of change) of the volume of all cortical labelled regions. Similarly, NAWM volume was defined as the volume of the supratentorial and cerebellar WM excluding WML volume. Finally, deep GM was the sum of the volume of bilateral caudate, putamen, globus pallidus, and thalamus). WML volume was the sum of WML labeled voxel volumes. Whole brain volume was calculated as the sum of the brain parenchyma normalized to the intracranial volume.

Mean R2t* in cortical grey matter (GM), deep GM, normal appearing white matter $(R2t_{NAWM}^{\ast })$, and individual lesions $(R2t_{Les}^{\ast })$ was computed for each subject. In MS patients, to quantify the damage in WM lesions compared to NAWM, we defined the Differential Tissue Damage Score (DTDS) in each WM lesion as

$$DTDS=\frac{R2t_{NAWM}^{\ast }-R2t_{Les}^{\ast }}{R2t_{NAWM}^{\ast }}$$

DTDS measured in this way quantifies the damage within lesions relative to the NAWM which, in MS patients, may also be abnormal ²³.

2.6. Statistical analysis

Longitudinal analysis of clinical test performance and imaging biomarkers in PMS subjects was conducted via ANOVA with subject as repeated, within-subject factor; the Dunnett test was used to correct for multiple comparison. Baseline clinical and imaging data were compared between subject groups using a one-way ANOVA with group (HC, RRMS, PMS) as factor; the Tukey test was used to correct for multiple comparison.

Spearman rank correlation tests were performed to assess the relationship between imaging biomarkers and clinical test scores. Imaging biomarkers included lesion volume, R2t* measured in cortical and deep GM, NAWM, and DTDS. Measures of clinical performance included EDSS, 25FTW, 9HPT, PASAT and SDMT. This cross-sectional analysis between HC, RRMS and PMS used only the first imaging session in the PMS cohort. Age, gender and disease duration were controlled as covariates. False discovery rate was used to correct for multiple comparisons ²⁴.

A primary interest of the present analysis was to determine which imaging biomarkers differentiated PMS patients with clinically defined progression vs. stability either at baseline or longitudinally. To assess this behavior, we fit linear mixed models with each imaging biomarker as an outcome of interest, separately. Predictor variables modeled as covariates included age, gender, and disease duration. Imaging visit was modeled as a within-subject repeated measure and subject number and group were appropriately modeled as nested random effects. The effects of interest were time (indicating longitudinal change in a variable), group (clinically defined progression vs. stability), and the interaction which encodes differential longitudinal change (i.e., slope) between the groups across time. These coefficients for each subject were then plotted for visual comparison.

Since imaging biomarkers measured in this study are characterized by high levels of shared variance, Least Absolute Shrinkage and Selection Operator (LASSO) regression was applied to identify the most salient imaging biomarkers of clinical progression ²⁵. LASSO is a modification to linear regression wherein regression coefficients (β) values are penalized such that small or redundant β values are forced to zero. Thus, LASSO models are sparse: several β values are large while many are identically 0 indicating those variables do not contribute to the model. Sparsity is controlled by a tuning variable, λ, chosen by cross-validation.

3. Results

Demographic information and clinical test performance were summarized and compared between MS subtypes (Table 1). Age and disease duration were each well matched between MS cohorts. The RRMS patients were chosen to match the PMS patients, and thus had an average age of 54.9 years with a range spanning 5 decades. PMS subjects had significantly higher baseline EDSS, 25FTW and 9HPT (dominant and non-dominant) scores compared to RRMS subjects (p<0.001, p<0.01, p<0.001, p<0.001, respectively). The RRMS cohort performed significantly better on SDMT than the PMS cohort (p<0.001). No differences were found in PASAT performance.

Table 1.

Demographic and clinical information of healthy controls and multiple sclerosis subjects.

		PMS	RRMS	Healthy Control
Number of subjects		23	22	18
Number of subjects that progressed over time		13	N/A	N/A
Disease duration		16.6±9.5	15.9±8.4	N/A
Number of imaging sessions		92	22	18
Age mean ± SD (years) (range)		56.8±8.6 (31-75)	54.9±10.1 (22-72)	48.8±14.5 (23-76)
Female/Male		15/8	18/4	13/5
EDSS mean ± SD (range)		6.0±1.5 (2.5-8.5)	2.4±1.1 (1-5)	N/A
25FTW mean ± SD (second) (range)		46.1±63.7 (4.2-165.8)	4.5±1.1 (2.7-7.7)	N/A
9HPT mean ± SD (second) (range)	Dominant	116.1±231.3 (18.7-777)	22.8±6.2 (15.2-45.5)	N/A
	Non-dominant	127.4±251.1 (21.4-777)	22.0±3.6 (16.3-31.8)	N/A
SDMT mean ± SD (range)		41.5±14.4 (10-63)	54.7±8.2 (37-80)	N/A
3 sec PASAT mean ± SD (range)		43.5±13.4 (14-60)	44.9±10.9 (27-58)	N/A

Of the twenty-three PMS subjects, nineteen were SPMS and four were PPMS. The clinical tests scores of PMS subjects were from session one. (Tests scores of subsequent sessions are shown in figure 1.) Disease duration was calculated by subtracting the date of diagnosis from the date of baseline scan. The disease durations of thirteen PMS subjects with clinically defined progression and ten stable PMS subjects were 16.7±9.8 years and 16.3±9.6 years, respectively. PPMS and SPMS patients did not significantly vary on any demographic parameter.

The PMS group was assessed longitudinally for changes in clinical test scores (Figure 1). Within the PMS group, Visit five showed significantly higher (more impaired) scores in EDSS, 25FTW and 9HPT than baseline (p<0.05, p<0.05, p<0.001, respectively). In contrast, cognitive function test scores (PASAT and SDMT) remained stable over the five sessions (p>0.05). Pairwise statistical comparisons for each test and visit were determined (supplemental Table S1). Clinical test performances were compared between the RRMS and PMS group (Figure 1). RRMS compared to the baseline visit of the PMS cohort showed significantly better performances in EDSS, 25-foot walk, 9-hole peg test (dominant and non-dominant hand) and SDMT (p<0.001, p<0.05, p<0.001, p<0.001, p<0.01, respectively). Baseline clinical scores did not predict future MS clinical progression (Figure S1). Clinical scores of PPMS and SPMS cohorts were differentially graphed over time for visual comparison (Figure S2).

Figure 1. Clinical assessments of the PMS cohort over four years.

Compared with the PMS cohort at baseline, the RRMS cohort (red dots) showed significantly better clinical test scores. EDSS, 25-foot walk and 9-hole peg test showed an increasing trend of disability and motor dysfunction for the longitudinal PMS cohort over five imaging sessions, suggesting MS progression within that group. For EDSS, 25FTW and 9HPT (non-dominant hand), there were significant differences between scores in visit five and visit one (p<0.05, p<0.05, p<0.001, respectively). 25FTW, 9HPT, PASAT and SDMT were converted to Z scores. Each dot represents one subject. The gray area represents the standard error. One-way ANOVA with repeated measures was used to compare the difference between visit one and each subsequent visit of the PMS group. Dunnett test was used to correct for multiple comparisons. Two-sample t-test was used to compare the RRMS cohort to the first visit of PMS group. *** p < 0.001, ** p <0.01, * p<0.05, n.s. p>0.05.

We next investigated the relation between the R2t* of different brain tissues and clinical test performance. In the cross-sectional analysis of forty-five MS subjects (22 RRMS and 23 PMS), R2t* measurements in NAWM and cortical GM, but not deep GM, correlated significantly with the EDSS, and motor and cognitive test scores (Table 2, Table S2, Table S3). The association between most clinical scores and R2t* was stronger in cortical GM compared to NAWM. R2t* in WMLs, quantified relative to NAWM within each subject as the DTDS, showed significant correlations with the EDSS and motor-related clinical scores (25FTW, 9HPT), but not cognitive test scores (PASAT, SDMT). Importantly, lesion volume was not significantly correlated with any clinical scores.

Table 2. R2t* in the cortical GM, NAWM and lesions correlated with clinical assessments in forty-five MS subjects.

R2t* measurements in both cortical GM and NAWM correlated with motor and cognitive tests. In addition, R2t* in cortical GM showed stronger correlation with clinical tests than NAWM. Mean lesion differential tissue damage score (DTDS) computed using R2t* measures within lesions showed significant correlations with the EDSS and motor assessments. No significant correlation between lesion volume and clinical scores was found. Spearman rho and p values were computed in R, with age, disease duration and gender as covariates. All listed p values are after multiple comparison correction using false discovery rate. Statistically significant correlations are highlighted in color.

	R2t* (Cortical GM)/s−1		R2t* (Deep GM)/s−1		R2t* (NAWMJ/s−1		Mean DTDS		Lesion Volume/mm3
	rho	p	rho	p	rho	p	rho	p	rho	p
EDSS	−0.72	p<0.001	−0.14	0.147	−0.49	p<0.001	0.46	p<0.001	0.18	0.058
25FTW	−0.74	p<0.001	−0.24	0.009	−0.53	p<0.001	0.52	p<0.001	0.12	0.235
9HPT (Dominant)	0.45	p<0.001	−0.06	0.585	0.30	p<0.001	−0.34	p<0.001	−0.01	0.910
9HPT (NonDominant)	0.56	p<0.001	−0.04	0.665	0.41	p<0.001	−0.38	p<0.001	−0.09	0.402
PASAT	0.29	0.002	−0.14	0.155	0.31	p<0.001	0.05	0.628	−0.16	0.096
SDMT	0.45	p<0.001	0.03	0.792	0.36	p<0.001	−0.28	0.002	−0.07	0.483

We next examined R2t* in the RRMS and PMS groups cross-sectionally and in the PMS group longitudinally (Figure 2). At baseline, cortical GM R2t* was lower in the PMS cohort compared to RRMS and control group (p<0.05). Lesion DTDS showed larger (worse) values in PMS compared to RRMS. (Table S4). In PMS, both cortical and deep GM R2t* declined over the five imaging sessions (Figure 2, Table S5). R2t* in NAWM did not change significantly across visits, however, the data exhibited increasing variability at later visits. DTDS declined over the last four visits, likely due to the decrease of R2t* in NAWM at later imaging sessions, making the differences between R2t* in NAWM and lesion smaller (Figure 2). Longitudinal R2t*, lesion volume and tissue damage score measurements for PMS subjects, with SPMS and PPMS subjects differentiated, have been graphed for visual comparisons in Figure S3. R2t* within WM lesions did not show significant change over five imaging sessions (Figure S4).

Figure 2. Longitudinal R2t*, lesion volume and tissue damage score measurements for non-relapsing PMS subjects compared with healthy control and RRMS measurements.

The longitudinal PMS cohort showed significant decreasing of R2t* (greater microstructural tissue damage) in cortical GM and deep GM over five imaging sessions, performed nine months apart. R2t* of cortical GM and deep GM in visit five were significantly different from R2t* in visit one (p<0.001 and p<0.01, respectively). At baseline, GM R2t* of PMS subjects was significantly lower than healthy control and RRMS cohorts (p<0.05). Lesion volume of PMS subjects increased over four years. Specifically, the lesion volume in visit five was significantly higher than the lesion volume of visit one (p<0.01). The line represents the mean value and the gray area represents the standard error. One-way ANOVA with repetitive measurements was used to compare the difference between visit one and each subsequent visit of the PMS group. Dunnett test was used to correct for multiple comparisons. One-way ANOVA was used to compare the healthy control, RRMS and baseline of the PMS group. Tukey test was used to correct for multiple comparisons. *** p < 0.001, ** p <0.01, * p<0.05, n.s p>0.05.

To determine if imaging biomarkers distinguished stable (n=10) vs. clinically progressing (n=13) non-relapsing PMS patients, we calculated the intercept and rate of change in R2t*, lesion volume, and atrophy using mixed effect modeling (Table 3, Figure 3). PMS subjects with clinically defined progression showed faster decline of R2t* in NAWM and deep grey matter (trend level) as evidenced by the small p-values in the Visit by Group interaction (Table 3). No measure showed a main effect of group. These statistical tests were significant considered in isolation but not significant when adjusted for multiple comparisons. Significant differences in deep GM R2t* and NAWM R2t* were found between the stable versus progressing PMS groups at the final time point (Figure S5).

Table 3. Inference tests resulting from linear mixed effects model.

t and p values corresponding to effects of interest are shown. These statistics were extracted from the linear mixed effects model adjusted for age, disease duration, and gender. Effects of Visit indicated changes over time that were present in both groups; effects of group represent mean differences across the groups that were present across visits; the interaction of Visit: Group was the primary interest in the present analysis, indicating changes over time that were different between groups.

	Visit		Group		Visit:Group
	t	p	t	p	t	p
CGM R2t*	−3.00	0.0069	−3.89	0.70	−0.70	0.49
DGM R2t*	−1.01	0.33	0.86	0.40	−2.07	0.053
NAWM R2t*	−0.71	0.49	0.21	0.84	−2.39	0.027
DTDS R2t*	−2.02	0.056	0.26	0.79	−0.16	0.88
CGM volume	0.36	0.72	−0.54	0.60	0.11	0.92
DGM volume	−0.55	0.59	−0.22	0.83	−1.55	0.14
NAWM volume	0.08	0.94	−0.64	0.53	−1.01	0.33
Lesion volume	1.24	0.23	−0.17	0.87	1.00	0.33

Figure 3. Rate of R2t* decline in NAWM and deep grey matter was faster in PMS subjects with clinically defined progression than in stable PMS subjects.

Thirteen PMS subjects showed clinically defined progression, whereas ten subjects were stable on clinical tests. Each dot represents one subject. Each row corresponds to an imaging biomarker. The first and third columns of graphs show rate of change and the second and fourth columns show the baseline values. Rate of change of R2t* in NAWM was significantly faster in PMS with clinical progression. Rate of decrease of R2t* in deep GM was nominally faster in the PMS group with clinical progression than the stable group, but did not reach statistical significance. All other p values were not statistically significant.

Logistic LASSO was used to identify the unique contribution of imaging biomarkers to the prediction of clinical progression. Predictors included in the model were initial and rate of change in R2t*, lesion volume, and atrophy (Table 4). Imaging data were standardized as Z-scores, resulting in regression coefficients (β values) that could be compared across variables. Baseline imaging values were not strongly associated with later progression. However, changes in deep GM and NAWM R2t* were associated with clinical progression. The largest β values were observed in the rate of change in R2t* of deep GM, followed by NAWM.

Table 4: Imaging variables that correlated with clinical progression.

LASSO logistic regression was fit with the 10 imaging parameters as predictors and the group (clinically stable PMS vs. clinical progression PMS) as outcome. Input variables were normalized to Z scores such that values would exist on the same scale. λ=0.143, the tuning parameter, was determined and optimized by cross validation. CGM = cortical grey matter; DGM = deep grey matter; NAWM = normal appearing white matter; DTDS= differential tissue damage score, WB = whole brain.

Predictor	β value
Initial Value
R2t* - CGM	0
R2t* - DGM	0
R2t* - NAWM	0
R2t* - Lesion DTDS	0
Lesion Volume	0
Volume – CGM	0
Volume – DGM	0
Volume - NAWM	0
Volume – WB	0
Rate of Change
R2t* - CGM	0
R2t* - DGM	−0.423
R2t* - NAWM	−0.015
R2t* - Lesion DTDS	0
Lesion Volume	0
Volume – CGM	0
Volume – DGM	0
Volume - NAWM	0
Volume – WM	0

R2t* and R2* are related quantities. Compared with R2*, R2t* was calculated by introducing new model parameters to remove the signal contribution caused by BOLD effect. To investigate the practical implications of this approach, we calculated mean R2* in the CGM, DGM, and NAWM in each group and found the topography to be similar in R2t* and R2* (Figure S6). The interaction of Visit by Group in NAWM R2t* (p = 0.027) was similar to the R2* data (p = 0.024). This represents only a cursory evaluation of the differential explanatory power of R2t* vs. R2*. A more comprehensive evaluation and comparison of R2t* and R2* is reserved for future work.

4. Discussion

MRI has significantly contributed to our understanding of MS pathology. WMLs, detected on gadolinium enhanced T1- and T2-weighted images, reliably correlate with clinical relapses in RRMS ⁴. Based on this, new or enhancing lesions are included as outcome measures in ongoing clinical trials and represent disease activity potentially due to suboptimal treatment of RRMS ²⁶. However, no currently available imaging biomarker, apart from atrophy, similarly correlates with progression in PMS. Because atrophy occurs subsequent to damage, identification of progression by atrophy is possible only in retrospect. This has, at least partially, slowed development of treatments of PMS. The current results demonstrate preliminary evidence that R2t* correlates with clinical impairment more closely than does atrophy, most likely because R2t* identifies tissue degeneration that is additional to the atrophy. These findings motivate the use of this technique in larger future studies to assess its utility as a biomarker of MS progression.

qGRE quantitatively measures R2t* (=1/T2t*) ¹², which is a sub-component of total R2* GRE signal decay (R2*=1/T2*) ¹⁵. Previous studies showed that quantitative determination of R2* differentiates RRMS patients from healthy controls and correlates with neurological impairment ^9-11. R2t* is more specific to tissue microstructure than total R2* because it measures a part of the GRE signal decay (transverse relaxation rate) that depends strongly on the local cellular-matrix environment of water molecules (a major source of MRI signal) and is not affected by the BOLD contributions to the R2* signal^{12, 13, 21}. Thus, R2t* is more closely related to intrinsic tissue integrity than R2* and is a more biologically relevant biomarker. It is also important to note that the data analysis in this study included the post-processing algorithms that minimize adverse artifacts related to macroscopic magnetic field inhomogeneities ²¹ and physiological fluctuations ¹³. Prior results also showed significant associations between CNS R2t* reduction and MS clinical severity ^{27, 28}. Importantly, the present results expand those findings to non-relapsing clinical progression examined longitudinally and demonstrate that R2t*-based measures outperform atrophy- and lesion-based measures as clinical biomarkers.

Histopathological studies have demonstrated that MS tissue damage is complex, involving not only demyelination but also axonal injury and transections, and gliosis in WMLs, GM and NAWM ²⁹. Progressive MS is thought to be due to continuing tissue injury mediated by multiple mechanisms beyond inflammatory activity ³⁰. We assayed central nervous system tissue damage using R2t* as a marker of pathology in patients with RRMS and PMS. PMS patients had greater tissue injury (reduced R2t*) compared to similarly aged RRMS patients and this injury correlated closely with clinical test performance. Consistent with ongoing damage, we found longitudinal reductions in R2t* across much of the brain in PMS patients. Importantly, this accumulating damage occurred largely outside of WMLs. Damage specific to lesions relative to NAWM (quantified with the DTDS) decreased over time because of the relatively larger increase in pathology in NAWM compared to discrete lesions.

PMS patients were further stratified into progressing or stable groups over the study period. Almost all PMS patients demonstrated at least mildly worsened (but not always statistically significant) clinical test performance across the study period. Critically, patients with confirmed progression, indicating significantly worsened impairment, had the most dramatic declines in tissue integrity in deep GM and NAWM. The association of deep GM atrophy with MS-related disability, specifically in progressive disease, is well documented ^{31, 32} but tissue atrophy is likely the end result of multiple pathologies. We speculate that the association between deep GM abnormalities and clinically defined progression is mediated by the central role of these structures in motor function, which is in turn key for defining MS disability. Overall, the present results suggest that tissue integrity measured by R2t* reflects neuropathology that has not yet become manifest as significant atrophy.

Atrophy is a commonly used imaging biomarker in progressive MS. In this study, we provided evidence that R2t* outperformed atrophy as a real-time correlate of accumulating clinical disability. Previous work from our group showed that R2t* positively correlated with neuronal density ¹⁵. In addition, R2t* normalization paralleled clinical improvement over 14 months in a patient with biopsy confirmed demyelinating white matter pathology ²⁷. These relationships demonstrate the biological relevance of R2t* as an imaging biomarker, although the histological correlates of the reduced R2t* measurements are yet to be fully defined. Evidence of better performance of R2t* than atrophy include increased effect size as revealed by linear mixed effects modeling (Table 3) and LASSO logistic regression (Table 4) which showed that R2t*-based metrics outperformed atrophy-based metrics for identifying accumulating disability in MS. This increased power could draw from two processes. First, R2t* might be a more powerful measure for identifying MS-related pathology. Alternatively, or in conjunction, R2t* could be measured more accurately than atrophy (volumetric measurements); a measure like R2t*, which is not exquisitely dependent on tissue segmentation, may be more robust in clinical applications.

This study has some limitations. The cohort was relatively small and only the PMS group (23 PMS patients) was followed longitudinally. Small studies are prone to sampling error. This is evident in our data in the form of unexpected variations in tissue volume and in clinical scores (e.g., SDMT). Nevertheless, we detected differential changes in those who progressed versus those who did not, suggesting that R2t* might serve as a surrogate for progression in early phase clinical trials in PMS. The progressive patients were followed every nine months for five visits, a short time compared to the time-course of PMS. Some patients might have clinically progressed but remained below the detection threshold. In addition, learning effects might have masked cognitive decrements on repeated testing. We were able to match RRMS and PMS cohorts by age but matching the two subtypes by both age and impairment was not possible due to the near-definitional requirement for PMS to have more significant clinical impairment. R2t* in NAWM was used as reference to calculate DTDS. Since NAWM in MS subjects is known to experience diffuse damage ³³, DTDS tends to underestimate the degree of tissue damage. The histological basis of R2t* signal changes in the deep GM is uncertain. Measured R2t* reflects the combined effects of microstructural tissue integrity and non-heme iron deposition. This question will be a subject of our future studies employing a recently developed method for separating cellular and iron contributions to GRE signal ³⁴.

5. Conclusion

We measured qGRE R2t* as a biomarker for tissue microstructural damage in brains of RRMS and PMS patients. We examined longitudinal changes in R2t* in non-relapsing PMS patients with and without clinically confirmed progression over the study period. We found that decreased R2t* in deep GM and NAWM correlated with clinical progression. This finding supports the potential of using R2t* as a quantitative imaging biomarker of ongoing MS progression.