J Neurol DOI 10.1007/s00415-015-7754-z
ORIGINAL COMMUNICATION
A longitudinal MRI study of cervical cord atrophy in multiple sclerosis Paola Valsasina1 • Maria A. Rocca1,2 • Mark A. Horsfield3 • Massimiliano Copetti4 Massimo Filippi1,2
•
Received: 16 March 2015 / Revised: 13 April 2015 / Accepted: 15 April 2015 Ó Springer-Verlag Berlin Heidelberg 2015
Abstract There is an urgent need of fast and accurate methods for the longitudinal quantification of cervical cord atrophy in multiple sclerosis (MS). Aim of this study was to compare a new semi-automatic method [the active surface (AS) method] with an existing cord segmentation method (the Losseff method) to measure cervical cord atrophy progression in MS patients. Using the AS and Losseff methods, normalized cervical cord cross-sectional area (CSA) was compared between 35 MS patients and 9 healthy controls (HC) at baseline and after 2.3 years of follow-up. Correlations with clinical/conventional MRI variables and a power calculation study were also performed. At follow-up, the Losseff method detected a 1 % CSA increase in HC and a 3.5 % decrease in MS patients (p = 0.01), while the AS method detected a 0.1 % decrease in HC and 3 % decrease in MS patients (p = 0.02). The AS method was more sensitive to associations with disability/conventional MRI variables and also provided Electronic supplementary material The online version of this article (doi:10.1007/s00415-015-7754-z) contains supplementary material, which is available to authorized users. & Massimo Filippi [email protected] 1
Neuroimaging Research Unit, Institute of Experimental Neurology, Division of Neuroscience, San Raffaele Scientific Institute, Vita-Salute San Raffaele University, Via Olgettina, 60, 20132 Milan, Italy
2
Department of Neurology, Institute of Experimental Neurology, Division of Neuroscience, San Raffaele Scientific Institute, Vita-Salute San Raffaele University, Milan, Italy
3
Xinapse Systems Ltd, 108 Lexden Road, West Bergholt, Essex, UK
4
Biostatistics Unit, IRCCS-Ospedale Casa Sollievo della Sofferenza, San Giovanni Rotondo, Foggia, Italy
lower numbers of subjects per arm compared to the Losseff method in a putative clinical trial scenario. Cord AS CSA measurements were more sensitive to longitudinal changes in MS patients than Losseff measurements. Cord AS CSA might be a valuable surrogate outcome measure for monitoring neuroprotection in MS. Keywords Cervical cord MRI Atrophy progression Multiple sclerosis Neuroprotection
Introduction The spinal cord is a clinically eloquent site of the CNS and is frequently affected by multiple sclerosis (MS). Focal or diffuse cord abnormalities, especially in the cervical segment, have been described in up to 90 % of MS patients [1]. Extensive axonal loss has been also well documented [2, 3] and is likely to have a major role in the clinical manifestations of the disease. As a consequence of demyelination and axonal loss, spinal cord atrophy has been frequently described in MS [4]. Spinal cord atrophy occurs early in MS patients [4, 5], is more pronounced in the progressive disease clinical phenotypes [6–10] and correlates with the severity of clinical disability [4, 6–9, 11]. In MS, the rate of progression of atrophy has been proposed as a measure of neurodegeneration [12, 13] and several clinical trials have included brain atrophy quantification to assess the effects of emerging neuroprotective therapies [14]. To date, great efforts have been made in investigating the evolution of brain atrophy in the different stages of MS [12] and the impact of current or experimental treatments on this evolution. Fast and accurate methods for the longitudinal quantification of brain atrophy
123
J Neurol
have been developed [15]. Conversely, only a few studies have examined spinal cord atrophy progression. A progressive reduction of cervical cord cross-sectional area (CSA) has been shown to occur in MS [5, 16–21], although the annual rate of cord volume loss is highly variable across studies, ranging from 1 % [20] to 5.2 % [22], with the majority of studies showing a yearly volume loss around 2–3 % [5, 16, 17, 19, 21]. Such discrepancies might be due to differences in sample sizes, follow-up durations and techniques employed to measure cord CSA. Also, the association between longitudinal modification of cord CSA and clinical worsening is discordant among studies, with some studies describing such a relationship [16, 17, 21] while others not [5, 18–20]. Probably due to the fact that the optimal method for measuring cervical cord atrophy has not yet been identified, only a small number of clinical trials [17, 23, 24] have used cord CSA as an exploratory endpoint. Among the methods currently implemented for cord CSA measurements, a new method based on active surface (AS) models is emerging [7, 8, 25, 26]. This method, which allows CSA measurements not only at specific cord levels but also along extended portions of the cord, has been successfully tested in large, cross-sectional multicenter studies [8] and seems to produce reliable values of cord area over short periods of follow-up, at least when limiting the analysis to the C2/C3 level [26]. Against this background, the main objectives of our study were: (1) to compare the sensitivity of the new AS method with the existing Losseff method [6] to measure the progression rate of cervical cord atrophy in MS patients; (2) to assess the correlation between cord atrophy development and clinical/conventional MRI variables; and (3) to perform a power calculation study using the two methods.
Materials and methods Subjects The institutional review board approved the study; all subjects gave informed consent. We studied 35 clinically definite MS patients [27] [12 relapsing remitting (RR) MS, 9 secondary progressive (SP) MS and 14 primary progressive (PP) MS] and 9 healthy controls (HC) (Table 1) with no history of neurologic dysfunction and a normal neurologic examination, at baseline and after a mean follow-up of 2.3 years (range 1.4–3.2 years). At follow-up, patients were considered clinically worsened if they had an Expanded Disability Status Scale (EDSS) score increase C1.0 when baseline EDSS was \6.0, or an EDSS score increase C0.5 when baseline EDSS was C6.0 [19]. For all subjects, exclusion criteria
123
included history of cervical trauma and evidence of cord compression and/or deformity on previous MR images. All MS patients were relapse and steroid free for at least 3 months before baseline and follow-up MRI examinations. MRI acquisition Using a 1.5 Tesla Siemens scanner (Erlangen, Germany), the following sequences were acquired from all study subjects: (a) cervical cord sagittal T1-weighted 3D magnetization-prepared rapid acquisition gradient echo (MPRAGE, scan resolution = 1.1 9 1.1 9 1.0 mm3); (b) cervical cord sagittal fast-short-tau inversion recovery (STIR, scan resolution = 1.06 9 1.06 9 3.3 mm3); (c) brain dual-echo turbo spin-echo (scan resolution = 0.97 9 0.97 9 5 mm3). Detailed acquisition parameters are reported in the Supplementary File. Conventional MRI analysis Cervical cord MS lesions at baseline and follow-up were counted on the fast-STIR images. Brain T2-hyperintense lesions were identified and T2 lesion volumes (LV) quantified, at baseline and at follow-up, using a local thresholding segmentation technique (Jim5, Xinapse System Ltd., The Northants, UK). Cord area measurements Baseline and follow-up CSA measurements were performed with the AS method and with the method proposed by Losseff et al. [6], as described in detail in the Supplementary File. For the AS method, cord CSA was estimated over a section from the top of C2 to the base of C5. An illustrative example of cord CSA measurement using the AS method on a HC and an MS patient is shown in Supplementary Figure 1. To provide a more direct comparison between the two methods for cord area measurements, cord CSA with the AS method was also calculated only in the region of C2, i.e., over the same anatomical area covered by the measurements produced by the Losseff method [7]. Normalized cord CSA was calculated for both Losseff (nLosseff) and AS (nAS) methods, adjusting the mean CSA to the intra-cranial cross-sectional area (ICCSA), as previously suggested [7, 28]. The ICCSA was measured at the level the inferior margins of the corpus callosum on an axial slice of the proton density-weighted image of each subject using a semi-automated contouring method. Such an adjustment was performed because in normal controls, the cranial size was found to be significantly correlated with cord area [5].
J Neurol Table 1 Main demographic, clinical and conventional MRI characteristics at baseline and follow-up of healthy controls (HC) and patients with multiple sclerosis (MS)
HC
MS patients
p
Mean age (years) (range)
38.9 (24–55)
43.3 (25–72)
0.26°
Men/women
2/7
11/24
0.68§
Mean disease duration at baseline (years) (range)
–
10.9 (1–37)
–
Median EDSS (range) at baseline
–
4.5 (1.0–7.0)
–
Median EDSS (range) at follow-up
–
5.0 (1.0–8.0)
–
Brain T2 LV (ml) (SD) at baseline
–
15.3 (14.8)
–
Brain T2 LV (ml) (SD) at follow-up
–
15.2 (13.9)
–
Mean number of cord lesions (range) at baseline
–
1.9 (0–5)
–
Mean number of cord lesions (range) at follow-up
–
2.2 (0–5)
–
SD standard deviation, EDSS expanded disability status scale, T2 LV T2 lesion volume ° Mann–Whitney U test;
§
Chi-square test
Statistical analysis All statistical analyses were performed using SAS release 9.1. Longitudinal changes of cord atrophy Baseline and follow-up nAS and nLosseff values were compared in a longitudinal linear modeling framework, adjusting for age and follow-up duration. Results were also reported as percentage change over time. A ‘‘time-bygroup’’ interaction analysis was performed to assess whether there were different patterns of longitudinal change of nAS and nLosseff between controls and MS patients and among clinical phenotypes. Correlation analysis Correlations between cord CSA, clinical, and conventional MRI variables were assessed using the Spearman rank correlation coefficient at the two timepoints, separately. Pairwise correlations between longitudinal changes of nAS, nLosseff, and brain T2 LV were computed using linear models, adjusted for follow-up duration. Negative binomial regression models were used to assess correlations between longitudinal changes of nAS, nLosseff and the number of new cervical cord lesions. Logistic regression models, including age and follow-up duration, were used to investigate the association between the probability of an EDSS worsening at follow-up and nAS or nLosseff changes, separately.
subjects, the distance along it was normalized by dividing it by the subject’s cervical cord length [8]. Pairwise group comparisons were made using a hierarchical linear model with repeated measurements using a spatial power covariance matrix. The hierarchical (i.e., multilevel) design was used to account for the nested design (i.e., there were many CSA measurements within each patient). Sample size calculation For each endpoint (nAS and nLosseff), we used the adjusted mean differences of CSA between baseline and 2.3 years of follow-up, along with their standard errors (SE), computed for MS patients, to perform a power calculation study. In this way, we could mimic a placebo-controlled randomized clinical trial, in which an active treatment reduces CSA loss over time. We assumed that, in the MS placebo arm, the reduction of CSA observed in this study was occurring (longitudinal linear model, age and follow-up adjusted mean = -2.42 mm2, SE = 0.38 mm2 for the AS method and = -2.36 mm2, SE = 0.57 mm2 for the Losseff method) and we considered as treatment effects 90, 75, 50, 25 and 10 % decreases of the observed full amount of CSA change. A 1:1 allocation of patients to treatment and placebo arms, 80 % power and two-sided significance level a of 0.05 (two-sample t test) were also assumed.
Results Clinical and conventional MRI measures
Analysis of normalized curves of cord CSA (AS method only) For explorative purposes, a non-parametric Nadaraya– Watson kernel regression estimator with a normal kernel was used to obtain smoothed plots of CSA measured along the cervical cord. Since cervical cord length varies between
Table 1 shows the main characteristics of HC and MS patients. During follow-up, 11 patients (31 %) had a clinical worsening and 9 MS patients formed new cervical cord T2 lesions. Brain T2 LV did not change over the follow-up period (p = 0.78), whereas the number of cord lesions was increased at follow-up vs baseline (p = 0.005).
123
J Neurol
Longitudinal changes of cord atrophy Table 2 shows the average cervical cord CSA of HC and MS patients at baseline and at follow-up, measured with the two methods. CSA values for the AS method are reported both for the portion of the cord between C2 and C5 (nAS), and for the measurements performed at C2 (nASC2). CSA measured with nAS, nAS-C2 and nLosseff did not change significantly over time in HC (Table 2), while it decreased significantly in MS patients. In MS patients vs controls, a significant ‘‘time-by-group’’ interaction of CSA change was also found for all measurements (Table 2), whereas no significant ‘‘time-by-group’’ interaction was found when considering the effect of phenotype for any method. In other words, CSA decrease over time was not significantly different among phenotypes.
regression models; for the model including the nAS measurements: p = 0.04; model including the nLosseff measurements: p = 0.05). Normalized curves of cord CSA Figure 1 shows a plot of the non-parametric kernel estimates of average CSA of controls and MS patients at baseline and follow-up. From C2 to C5, a net separation of the curves of HC from those of MS patients is identifiable. The follow-up curve for HC strictly overlapped with the baseline curve, while the follow-up curve for MS patients appears to show a CSA decrease compared with baseline, most evident in the caudal portions. The formal statistical comparison of these curves showed an effect of group (p = 0.02), but neither an effect of time (p = 0.48 for MS patients, p = 0.83 for HC) nor ‘‘time-by-group’’ interaction (p = 0.6).
Correlation analysis Sample size calculation At baseline, cord nAS was significantly associated with brain T2 LV and EDSS (Table 3). Associations between baseline nLosseff and these variables were also significant, but weaker (Table 3). At follow-up, cord nAS was significantly correlated with brain T2 LV and EDSS, while the associations between follow-up nLosseff values and these variables were not significant (Table 3). No correlations were found between the change in nAS over time and concomitant changes of brain T2 LV (p = 0.85) and cord lesion number (p = 0.83). Similarly, no correlations were found between the changes in nLosseff and conventional MRI metrics changes (p = 0.85 for T2 LV and 0.67 for cord lesions). In MS patients, the only variable associated with a clinical worsening at follow-up was age (one of the covariates of the logistic
Table 4 reports the estimated sample sizes per arm for a placebo-controlled treatment trial in which an active treatment reduces CSA loss over time. We found that the number of subjects needed to observe any treatment effect was almost halved when using the AS compared to the Losseff method (e.g., n = 72 vs n = 164 subjects to detect a 50 % treatment effect).
Discussion Measuring spinal cord atrophy is not straightforward because of the intrinsic cord curvature, its small size and the complex anatomy of surrounding structures [29, 30]. Given
Table 2 Mean normalized cord cross-sectional areas, at baseline and at follow-up, measured with the Losseff and active surface (AS) methods in healthy controls (HC) and patients with multiple sclerosis (MS) Group
Method
Baseline
Follow-up
Cord area change (%)
Regression coefficient*
HC MS patients
2
nAS (mm ) (SD) nAS (mm2) (SD)
81.7 (3.6) 74.5 (9.2)
81.6 (3.3) 72.2 (8.6)
-0.1 -3.0
-0.12 -2.28
HC
nAS-C2 (mm2) (SD)
76.9 (5.3)
76.8 (5.3)
-0.07
MS patients
nAS-C2 (mm2) (SD)
71.5 (9.6)
69.1 (8.7)
-3.4
-2.47
HC
nLosseff (mm2) (SD)
71.7 (7.3)
72.5 (5.5)
?1.0
?0.79
-1.11
MS patients
nLosseff (mm2) (SD)
66.7 (9.9)
64.3 (9.7)
-3.4
-2.32
3.94
0.068
SE
p*
p**
0.78 0.00001
0.02
0.15
0.89
0.04
4.27
0.0002
0.30 5.12
0.31
0.01
0.0004
* Effect of time, longitudinal linear model adjusted for age and follow-up duration ** Time by group interaction, longitudinal linear model adjusted for age and follow-up duration SD standard deviation, SE standardized effect, nAS normalized cord cross-sectional area calculated with the active surface method, nAS-C2 normalized cord cross-sectional area calculated with the active surface method at C2 level, nLosseff normalized cord cross-sectional area calculated with the Losseff method
123
J Neurol Table 3 Associations between normalized cord cross-sectional areas, at baseline and at follow-up, measured with the Losseff and active surface (AS) method, and clinical/conventional MRI variables Baseline T2 LV (r, p)a
Follow-up T2 LV (r, p)a
Baseline EDSS (r, p)a
Follow-up EDSS (r, p)a
Baseline nAS
-0.49, 0.003
NA
-0.59, \0.001
NA
Baseline nLosseff
-0.32, 0.06
NA
-0.40, 0.01
NA
Follow-up nAS
NA
-0.38, 0.02
NA
-0.50, 0.002
Follow-up nLosseff
NA
-0.30, 0.07
NA
-0.31, 0.06
a
Spearman’s rank correlation coefficient
NA not applicable, LV lesion volume, EDSS expanded disability status scale, nAS normalized cord cross-sectional area calculated with the active surface method, nLosseff normalized cord cross-sectional area calculated with the Losseff method
Table 4 Estimated sample sizes per arm for a placebo-controlled treatment trial aimed at reducing cord cross-sectional area (CSA) loss over 2.3 years of follow-up (power = 80 %, significance level a = 0.05) Treatment effect (%)
Sample size per arm nAS
nLosseff
10
1751
4054
25
282
649
50
72
164
75
33
74
90
23
52
nAS normalized cord cross-sectional area calculated with the active surface method, nLosseff normalized cord cross-sectional area calculated with the Losseff method
Fig. 1 Non-parametric kernel estimate plots of cord cross-sectional area (CSA) along the cervical cord for healthy controls (red) and patients with multiple sclerosis (MS) (blue) at baseline (continuous lines) and at follow-up (dashed lines). Distance along the cervical cord was normalized by dividing it by subject’s cervical cord length: the normalized cord distance varies from 0 [most superior axial slice in which the odontoid process of the epistropheus (C2) was still visible] to 1 (inferior border of C5). See text for further details
the technical challenges of implementing accurate 3D spinal cord measurements, the majority of methods for assessing cord atrophy in MS have focussed, until recently, on analyzing small cord portions [6, 28, 29] without clear data about reproducibility over time. The AS method [7] holds great promise for overcoming some of the above-mentioned limitations. Previous reports have already shown a high intra-observer and inter-observer reproducibility of cord CSA measurements both in cross-sectional (inter-observer coefficient of variation (CoV) = 1.07 %; intra-observer CoV = 0.44 % [7]) and in short-term longitudinal designs (inter-observer CoV = 0.03–0.2 %; intra-observer CoV = 0.1–0.4 %,
over 6 months of follow-up, depending on the scanning sequence [10]). Moreover, AS cord measurements were found to be less time-consuming to perform and had the advantage of covering large portions of the spinal cord, if appropriate 3D acquisitions were available [7]. The main findings of our study were that the stability over time of cord CSA values in HC tended to be higher with AS than with Losseff measurements; and the sensitivity of the AS measurements to longitudinal modifications of cervical cord CSA in MS tended to be higher than that of Losseff measurements, even if a formal statistical comparison was not done due to the small number of subjects. In HC, significant cord CSA changes over 2 years of follow-up are not expected, especially in the age range of the controls included in our study. Previous cross-sectional studies indicated that cord volume loss occurs with aging to only a small extent in the healthy elderly [31–33], while previous longitudinal studies showed no significant changes of cord CSA values in HC over time [5, 26]. In line with these considerations, longitudinal changes of CSA were not significant in our controls. However, while cord
123
J Neurol
nAS values at follow-up were almost identical to those at baseline, a non-significant increase of cord area (?1 %) was found with nLosseff values. This unexpected result is similar to that obtained in a previous study [5], which also used the Losseff method for cord area measurements (?3 % of mean CSA over 3 years in HC). Since age-related neurodegenerative mechanisms occurring in the spinal cord are likely to result in tissue loss [34], this result is likely to be due predominantly to measurement errors. In MS patients, a significant development of cord atrophy was found with both the AS and Losseff techniques. The rate of loss of cord CSA (which was about 3–3.5 % over 2 years) was in line with previous findings [5, 17, 19, 21], but the statistical significance of the detected change was higher with the AS than with the Losseff method. This suggests that the AS method, thanks to a lower variance of measurements, might be more sensitive and, as a consequence, its application might be more effective in studies with a small sample size. In line with previous reports [4], we found that cord measurements were significantly correlated with disability. This result confirms the clinical relevance of the spinal cord, damage to which is likely to disproportionately affect the locomotor and sensory functions. Since the EDSS is strongly weighted towards locomotor disability, it has often been found to be correlated with atrophy of the cervical cord [4, 8, 9, 11]. Cord CSA was correlated with brain T2 LV, but not with the number of cervical cord lesions. The finding that cord atrophy is more strongly correlated with brain rather than cord damage is not novel [8–10] and confirms that individual cord lesions may play only a minor role in local atrophy and that Wallerian degeneration of long fiber tracts might be one of the main factors leading to tissue loss in the cord [2]. The strongest correlations between cord CSA and clinical/conventional MRI variables were found when AS measurements were used. This result strengthens the case for using the AS method for assessing cord area in cross-sectional and longitudinal studies. In our sample, the development of cord atrophy was not associated with concomitant clinical worsening. This disappointing result might be explained by the small patient sample in our study and by the relatively low proportion of MS patients (11 out of 35, about 30 %) showing a clinical worsening at follow-up. The duration of follow-up may also have played a role, since disability, as assessed with EDSS, takes years to develop. Future longitudinal studies with larger sample sizes and longer follow-up might be able to show an association between irreversible cord tissue loss and clinical changes. The analysis of cord CSA plots from C2 to C5 showed a significant effect of group, with a net separation of the curves of controls and MS patients at the two timepoints along the entire length of the cervical cord analyzed. This
123
suggests that the pathological processes we are quantifying with atrophy diffusely affect the cervical cord. While baseline and follow-up CSA plots from HC overlapped over much of their length, there was the appearance of a decrease of cord CSA in MS patients at follow-up compared to baseline. This decrease, which did not reach statistical significance, was more evident in the most caudal sections of the cord (around C5) than at the C1/C2 vertebral level. The non-uniform distribution of cord volume loss, together with the small study sample size, are likely to have caused an underpowering of the multilevel nested design of the statistical test. Future studies with larger sample size and longer follow-up may take a full advantage from the use of the AS method for the analysis of long cord segments, and might be more sensitive in detecting cord regions particularly susceptible to atrophy development over time. This study is not without limitations, including the fact that it may not be straightforward to apply the AS method, especially along large cord portions, to subjects with vertebral column disease that causes cervical cord compression. However, the technique also gave very good results (in terms of reproducibility and sensitivity to cord area changes) when applied to a limited portion of the cord around C2, as shown by our study and previous reports [10]. Moreover, the small sample size did not allow us to perform a powered comparison of cord atrophy changes among phenotypes. Finally, a high-resolution, three-dimensional acquisition sequence is needed to perform reliable cord measurements, and the application of such a sequence in a multicenter study might be problematic, especially when dealing with advanced imaging techniques, such as phase-sensitive inversion recovery sequences, as proposed by Kearney et al. [10]. Nevertheless, standard magnetization-prepared 3D T1 scans seem to be sufficient to obtain precise and reliable cord measurements. Given the need for sensitive biomarkers of neurodegeneration, we performed a power calculation study to evaluate the use of CSA as an endpoint in clinical trials. Our calculations showed that fewer subjects per arm (about the half) would be needed with the AS method, compared to the Losseff method, to detect a significant treatment effect. This result confirms previous findings [10] and suggests that: (1) the use of cord CSA as an endpoint in clinical trials would be feasible; and (2) the AS method is highly sensitive to cord volume changes, allowing small sample sizes. Together with the previously observed reliability of measurements in a multicenter setting [8], these findings suggest that cord nAS might be a valuable surrogate outcome measure for monitoring disease progression and treatment efficacy in MS. Acknowledgements This study was partially supported by a grant from Fondazione Italiana Sclerosi Multipla (FISM2014/PMS/6).
J Neurol Conflicts of interest
The authors declare no conflict of interest.
Ethical standard Approval was obtained from the local ethical standards committee on human experimentation and written informed consent from all subjects before enrolment. The study has been performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments.
References 1. Nijeholt GJ, van Walderveen MA, Castelijns JA et al (1998) Brain and spinal cord abnormalities in multiple sclerosis. Correlation between MRI parameters, clinical subtypes and symptoms. Brain 121(Pt 4):687–697 2. Evangelou N, DeLuca GC, Owens T, Esiri MM (2005) Pathological study of spinal cord atrophy in multiple sclerosis suggests limited role of local lesions. Brain 128(Pt 1):29–34 3. DeLuca GC, Ebers GC, Esiri MM (2004) Axonal loss in multiple sclerosis: a pathological survey of the corticospinal and sensory tracts. Brain 127(Pt 5):1009–1018 4. Filippi M, Rocca MA (2013) Multiple sclerosis: Linking disability and spinal cord imaging outcomes in MS. Nat Rev Neurol 9(4):189–190 5. Rashid W, Davies GR, Chard DT et al (2006) Increasing cord atrophy in early relapsing-remitting multiple sclerosis: a 3 year study. J Neurol Neurosurg Psychiatry 77(1):51–55 6. Losseff NA, Webb SL, O’Riordan JI et al (1996) Spinal cord atrophy and disability in multiple sclerosis. A new reproducible and sensitive MRI method with potential to monitor disease progression. Brain 119(Pt 3):701–708 7. Horsfield MA, Sala S, Neema M et al (2010) Rapid semi-automatic segmentation of the spinal cord from magnetic resonance images: application in multiple sclerosis. Neuroimage 50(2):446–455 8. Rocca MA, Horsfield MA, Sala S et al (2011) A multicenter assessment of cervical cord atrophy among MS clinical phenotypes. Neurology 76(24):2096–2102 9. Lukas C, Sombekke MH, Bellenberg B et al (2013) Relevance of spinal cord abnormalities to clinical disability in multiple sclerosis: MR imaging findings in a large cohort of patients. Radiology 269(2):542–552 10. Kearney H, Rocca MA, Valsasina P et al (2014) Magnetic resonance imaging correlates of physical disability in relapse onset multiple sclerosis of long disease duration. Mult Scler 20(1):72–80 11. Daams M, Weiler F, Steenwijk MD et al (2014) Mean upper cervical cord area (MUCCA) measurement in long-standing multiple sclerosis: Relation to brain findings and clinical disability. Mult Scler 20(14):1860–1865 12. Barkhof F, Calabresi PA, Miller DH, Reingold SC (2009) Imaging outcomes for neuroprotection and repair in multiple sclerosis trials. Nat Rev Neurol 5(5):256–266 13. Zivadinov R, Reder AT, Filippi M et al (2008) Mechanisms of action of disease-modifying agents and brain volume changes in multiple sclerosis. Neurology 71(2):136–144 14. Filippi M, Preziosa P, Rocca MA (2014) Magnetic resonance outcome measures in multiple sclerosis trials: time to rethink? Curr Opin Neurol 27(3):290–299 15. Vrenken H, Jenkinson M, Horsfield MA et al (2013) Recommendations to improve imaging and analysis of brain lesion load and atrophy in longitudinal studies of multiple sclerosis. J Neurol 260(10):2458–2471 16. Ingle GT, Stevenson VL, Miller DH, Thompson AJ (2003) Primary progressive multiple sclerosis: a 5-year clinical and MR study. Brain 126(Pt 11):2528–2536
17. Lin X, Tench CR, Turner B, Blumhardt LD, Constantinescu CS (2003) Spinal cord atrophy and disability in multiple sclerosis over four years: application of a reproducible automated technique in monitoring disease progression in a cohort of the interferon beta-1a (Rebif) treatment trial. J Neurol Neurosurg Psychiatry 74(8):1090–1094 18. Stevenson VL, Ingle GT, Miller DH, Thompson AJ (2004) Magnetic resonance imaging predictors of disability in primary progressive multiple sclerosis: a 5-year study. Mult Scler 10(4):398–401 19. Agosta F, Absinta M, Sormani MP et al (2007) In vivo assessment of cervical cord damage in MS patients: a longitudinal diffusion tensor MRI study. Brain 130(Pt 8):2211–2219 20. Furby J, Hayton T, Altmann D et al (2010) A longitudinal study of MRI-detected atrophy in secondary progressive multiple sclerosis. J Neurol 257(9):1508–1516 21. Lukas C, Knol DL, Sombekke MH et al (2015) Cervical spinal cord volume loss is related to clinical disability progression in multiple sclerosis. J Neurol Neurosurg Psychiatry 86(4):410–418 22. Stevenson VL, Leary SM, Losseff NA et al (1998) Spinal cord atrophy and disability in MS: a longitudinal study. Neurology. 51(1):234–238 23. Leary SM, Miller DH, Stevenson VL, Brex PA, Chard DT, Thompson AJ (2003) Interferon beta-1a in primary progressive MS: an exploratory, randomized, controlled trial. Neurology 60(1):44–51 24. Montalban X, Sastre-Garriga J, Tintore M et al (2009) A singlecenter, randomized, double-blind, placebo-controlled study of interferon beta-1b on primary progressive and transitional multiple sclerosis. Mult Scler 15(10):1195–1205 25. Klein JP, Arora A, Neema M et al (2011) A 3T MR imaging investigation of the topography of whole spinal cord atrophy in multiple sclerosis. AJNR Am J Neuroradiol 32(6):1138–1142 26. Kearney H, Yiannakas MC, Abdel-Aziz K et al (2014) Improved MRI quantification of spinal cord atrophy in multiple sclerosis. J Magn Reson Imaging 39(3):617–623 27. Polman CH, Reingold SC, Banwell B et al (2010) Diagnostic criteria for multiple sclerosis: 2010 revisions to the McDonald criteria. Ann Neurol 69(2):292–302 28. Lin X, Blumhardt LD, Constantinescu CS (2003) The relationship of brain and cervical cord volume to disability in clinical subtypes of multiple sclerosis: a three-dimensional MRI study. Acta Neurol Scand 108(6):401–406 29. Tench CR, Morgan PS, Constantinescu CS (2005) Measurement of cervical spinal cord cross-sectional area by MRI using edge detection and partial volume correction. J Magn Reson Imaging 21(3):197–203 30. Agosta F, Filippi M (2007) MRI of spinal cord in multiple sclerosis. J Neuroimaging 17(Suppl 1):46S–49S 31. Agosta F, Lagana M, Valsasina P et al (2007) Evidence for cervical cord tissue disorganisation with aging by diffusion tensor MRI. Neuroimage 36(3):728–735 32. Valsasina P, Horsfield MA, Rocca MA, Absinta M, Comi G, Filippi M (2012) Spatial normalization and regional assessment of cord atrophy: voxel-based analysis of cervical cord 3D T1weighted images. AJNR Am J Neuroradiol 33(11):2195–2200 33. Yanase M, Matsuyama Y, Hirose K et al (2006) Measurement of the cervical spinal cord volume on MRI. J Spinal Disord Tech 19(2):125–129 34. Cruz-Sanchez FF, Moral A, Tolosa E, de Belleroche J, Rossi ML (1998) Evaluation of neuronal loss, astrocytosis and abnormalities of cytoskeletal components of large motor neurons in the human anterior horn in aging. J Neural Transm 105(6–7):689–701
123
ORIGINAL COMMUNICATION
A longitudinal MRI study of cervical cord atrophy in multiple sclerosis Paola Valsasina1 • Maria A. Rocca1,2 • Mark A. Horsfield3 • Massimiliano Copetti4 Massimo Filippi1,2
•
Received: 16 March 2015 / Revised: 13 April 2015 / Accepted: 15 April 2015 Ó Springer-Verlag Berlin Heidelberg 2015
Abstract There is an urgent need of fast and accurate methods for the longitudinal quantification of cervical cord atrophy in multiple sclerosis (MS). Aim of this study was to compare a new semi-automatic method [the active surface (AS) method] with an existing cord segmentation method (the Losseff method) to measure cervical cord atrophy progression in MS patients. Using the AS and Losseff methods, normalized cervical cord cross-sectional area (CSA) was compared between 35 MS patients and 9 healthy controls (HC) at baseline and after 2.3 years of follow-up. Correlations with clinical/conventional MRI variables and a power calculation study were also performed. At follow-up, the Losseff method detected a 1 % CSA increase in HC and a 3.5 % decrease in MS patients (p = 0.01), while the AS method detected a 0.1 % decrease in HC and 3 % decrease in MS patients (p = 0.02). The AS method was more sensitive to associations with disability/conventional MRI variables and also provided Electronic supplementary material The online version of this article (doi:10.1007/s00415-015-7754-z) contains supplementary material, which is available to authorized users. & Massimo Filippi [email protected] 1
Neuroimaging Research Unit, Institute of Experimental Neurology, Division of Neuroscience, San Raffaele Scientific Institute, Vita-Salute San Raffaele University, Via Olgettina, 60, 20132 Milan, Italy
2
Department of Neurology, Institute of Experimental Neurology, Division of Neuroscience, San Raffaele Scientific Institute, Vita-Salute San Raffaele University, Milan, Italy
3
Xinapse Systems Ltd, 108 Lexden Road, West Bergholt, Essex, UK
4
Biostatistics Unit, IRCCS-Ospedale Casa Sollievo della Sofferenza, San Giovanni Rotondo, Foggia, Italy
lower numbers of subjects per arm compared to the Losseff method in a putative clinical trial scenario. Cord AS CSA measurements were more sensitive to longitudinal changes in MS patients than Losseff measurements. Cord AS CSA might be a valuable surrogate outcome measure for monitoring neuroprotection in MS. Keywords Cervical cord MRI Atrophy progression Multiple sclerosis Neuroprotection
Introduction The spinal cord is a clinically eloquent site of the CNS and is frequently affected by multiple sclerosis (MS). Focal or diffuse cord abnormalities, especially in the cervical segment, have been described in up to 90 % of MS patients [1]. Extensive axonal loss has been also well documented [2, 3] and is likely to have a major role in the clinical manifestations of the disease. As a consequence of demyelination and axonal loss, spinal cord atrophy has been frequently described in MS [4]. Spinal cord atrophy occurs early in MS patients [4, 5], is more pronounced in the progressive disease clinical phenotypes [6–10] and correlates with the severity of clinical disability [4, 6–9, 11]. In MS, the rate of progression of atrophy has been proposed as a measure of neurodegeneration [12, 13] and several clinical trials have included brain atrophy quantification to assess the effects of emerging neuroprotective therapies [14]. To date, great efforts have been made in investigating the evolution of brain atrophy in the different stages of MS [12] and the impact of current or experimental treatments on this evolution. Fast and accurate methods for the longitudinal quantification of brain atrophy
123
J Neurol
have been developed [15]. Conversely, only a few studies have examined spinal cord atrophy progression. A progressive reduction of cervical cord cross-sectional area (CSA) has been shown to occur in MS [5, 16–21], although the annual rate of cord volume loss is highly variable across studies, ranging from 1 % [20] to 5.2 % [22], with the majority of studies showing a yearly volume loss around 2–3 % [5, 16, 17, 19, 21]. Such discrepancies might be due to differences in sample sizes, follow-up durations and techniques employed to measure cord CSA. Also, the association between longitudinal modification of cord CSA and clinical worsening is discordant among studies, with some studies describing such a relationship [16, 17, 21] while others not [5, 18–20]. Probably due to the fact that the optimal method for measuring cervical cord atrophy has not yet been identified, only a small number of clinical trials [17, 23, 24] have used cord CSA as an exploratory endpoint. Among the methods currently implemented for cord CSA measurements, a new method based on active surface (AS) models is emerging [7, 8, 25, 26]. This method, which allows CSA measurements not only at specific cord levels but also along extended portions of the cord, has been successfully tested in large, cross-sectional multicenter studies [8] and seems to produce reliable values of cord area over short periods of follow-up, at least when limiting the analysis to the C2/C3 level [26]. Against this background, the main objectives of our study were: (1) to compare the sensitivity of the new AS method with the existing Losseff method [6] to measure the progression rate of cervical cord atrophy in MS patients; (2) to assess the correlation between cord atrophy development and clinical/conventional MRI variables; and (3) to perform a power calculation study using the two methods.
Materials and methods Subjects The institutional review board approved the study; all subjects gave informed consent. We studied 35 clinically definite MS patients [27] [12 relapsing remitting (RR) MS, 9 secondary progressive (SP) MS and 14 primary progressive (PP) MS] and 9 healthy controls (HC) (Table 1) with no history of neurologic dysfunction and a normal neurologic examination, at baseline and after a mean follow-up of 2.3 years (range 1.4–3.2 years). At follow-up, patients were considered clinically worsened if they had an Expanded Disability Status Scale (EDSS) score increase C1.0 when baseline EDSS was \6.0, or an EDSS score increase C0.5 when baseline EDSS was C6.0 [19]. For all subjects, exclusion criteria
123
included history of cervical trauma and evidence of cord compression and/or deformity on previous MR images. All MS patients were relapse and steroid free for at least 3 months before baseline and follow-up MRI examinations. MRI acquisition Using a 1.5 Tesla Siemens scanner (Erlangen, Germany), the following sequences were acquired from all study subjects: (a) cervical cord sagittal T1-weighted 3D magnetization-prepared rapid acquisition gradient echo (MPRAGE, scan resolution = 1.1 9 1.1 9 1.0 mm3); (b) cervical cord sagittal fast-short-tau inversion recovery (STIR, scan resolution = 1.06 9 1.06 9 3.3 mm3); (c) brain dual-echo turbo spin-echo (scan resolution = 0.97 9 0.97 9 5 mm3). Detailed acquisition parameters are reported in the Supplementary File. Conventional MRI analysis Cervical cord MS lesions at baseline and follow-up were counted on the fast-STIR images. Brain T2-hyperintense lesions were identified and T2 lesion volumes (LV) quantified, at baseline and at follow-up, using a local thresholding segmentation technique (Jim5, Xinapse System Ltd., The Northants, UK). Cord area measurements Baseline and follow-up CSA measurements were performed with the AS method and with the method proposed by Losseff et al. [6], as described in detail in the Supplementary File. For the AS method, cord CSA was estimated over a section from the top of C2 to the base of C5. An illustrative example of cord CSA measurement using the AS method on a HC and an MS patient is shown in Supplementary Figure 1. To provide a more direct comparison between the two methods for cord area measurements, cord CSA with the AS method was also calculated only in the region of C2, i.e., over the same anatomical area covered by the measurements produced by the Losseff method [7]. Normalized cord CSA was calculated for both Losseff (nLosseff) and AS (nAS) methods, adjusting the mean CSA to the intra-cranial cross-sectional area (ICCSA), as previously suggested [7, 28]. The ICCSA was measured at the level the inferior margins of the corpus callosum on an axial slice of the proton density-weighted image of each subject using a semi-automated contouring method. Such an adjustment was performed because in normal controls, the cranial size was found to be significantly correlated with cord area [5].
J Neurol Table 1 Main demographic, clinical and conventional MRI characteristics at baseline and follow-up of healthy controls (HC) and patients with multiple sclerosis (MS)
HC
MS patients
p
Mean age (years) (range)
38.9 (24–55)
43.3 (25–72)
0.26°
Men/women
2/7
11/24
0.68§
Mean disease duration at baseline (years) (range)
–
10.9 (1–37)
–
Median EDSS (range) at baseline
–
4.5 (1.0–7.0)
–
Median EDSS (range) at follow-up
–
5.0 (1.0–8.0)
–
Brain T2 LV (ml) (SD) at baseline
–
15.3 (14.8)
–
Brain T2 LV (ml) (SD) at follow-up
–
15.2 (13.9)
–
Mean number of cord lesions (range) at baseline
–
1.9 (0–5)
–
Mean number of cord lesions (range) at follow-up
–
2.2 (0–5)
–
SD standard deviation, EDSS expanded disability status scale, T2 LV T2 lesion volume ° Mann–Whitney U test;
§
Chi-square test
Statistical analysis All statistical analyses were performed using SAS release 9.1. Longitudinal changes of cord atrophy Baseline and follow-up nAS and nLosseff values were compared in a longitudinal linear modeling framework, adjusting for age and follow-up duration. Results were also reported as percentage change over time. A ‘‘time-bygroup’’ interaction analysis was performed to assess whether there were different patterns of longitudinal change of nAS and nLosseff between controls and MS patients and among clinical phenotypes. Correlation analysis Correlations between cord CSA, clinical, and conventional MRI variables were assessed using the Spearman rank correlation coefficient at the two timepoints, separately. Pairwise correlations between longitudinal changes of nAS, nLosseff, and brain T2 LV were computed using linear models, adjusted for follow-up duration. Negative binomial regression models were used to assess correlations between longitudinal changes of nAS, nLosseff and the number of new cervical cord lesions. Logistic regression models, including age and follow-up duration, were used to investigate the association between the probability of an EDSS worsening at follow-up and nAS or nLosseff changes, separately.
subjects, the distance along it was normalized by dividing it by the subject’s cervical cord length [8]. Pairwise group comparisons were made using a hierarchical linear model with repeated measurements using a spatial power covariance matrix. The hierarchical (i.e., multilevel) design was used to account for the nested design (i.e., there were many CSA measurements within each patient). Sample size calculation For each endpoint (nAS and nLosseff), we used the adjusted mean differences of CSA between baseline and 2.3 years of follow-up, along with their standard errors (SE), computed for MS patients, to perform a power calculation study. In this way, we could mimic a placebo-controlled randomized clinical trial, in which an active treatment reduces CSA loss over time. We assumed that, in the MS placebo arm, the reduction of CSA observed in this study was occurring (longitudinal linear model, age and follow-up adjusted mean = -2.42 mm2, SE = 0.38 mm2 for the AS method and = -2.36 mm2, SE = 0.57 mm2 for the Losseff method) and we considered as treatment effects 90, 75, 50, 25 and 10 % decreases of the observed full amount of CSA change. A 1:1 allocation of patients to treatment and placebo arms, 80 % power and two-sided significance level a of 0.05 (two-sample t test) were also assumed.
Results Clinical and conventional MRI measures
Analysis of normalized curves of cord CSA (AS method only) For explorative purposes, a non-parametric Nadaraya– Watson kernel regression estimator with a normal kernel was used to obtain smoothed plots of CSA measured along the cervical cord. Since cervical cord length varies between
Table 1 shows the main characteristics of HC and MS patients. During follow-up, 11 patients (31 %) had a clinical worsening and 9 MS patients formed new cervical cord T2 lesions. Brain T2 LV did not change over the follow-up period (p = 0.78), whereas the number of cord lesions was increased at follow-up vs baseline (p = 0.005).
123
J Neurol
Longitudinal changes of cord atrophy Table 2 shows the average cervical cord CSA of HC and MS patients at baseline and at follow-up, measured with the two methods. CSA values for the AS method are reported both for the portion of the cord between C2 and C5 (nAS), and for the measurements performed at C2 (nASC2). CSA measured with nAS, nAS-C2 and nLosseff did not change significantly over time in HC (Table 2), while it decreased significantly in MS patients. In MS patients vs controls, a significant ‘‘time-by-group’’ interaction of CSA change was also found for all measurements (Table 2), whereas no significant ‘‘time-by-group’’ interaction was found when considering the effect of phenotype for any method. In other words, CSA decrease over time was not significantly different among phenotypes.
regression models; for the model including the nAS measurements: p = 0.04; model including the nLosseff measurements: p = 0.05). Normalized curves of cord CSA Figure 1 shows a plot of the non-parametric kernel estimates of average CSA of controls and MS patients at baseline and follow-up. From C2 to C5, a net separation of the curves of HC from those of MS patients is identifiable. The follow-up curve for HC strictly overlapped with the baseline curve, while the follow-up curve for MS patients appears to show a CSA decrease compared with baseline, most evident in the caudal portions. The formal statistical comparison of these curves showed an effect of group (p = 0.02), but neither an effect of time (p = 0.48 for MS patients, p = 0.83 for HC) nor ‘‘time-by-group’’ interaction (p = 0.6).
Correlation analysis Sample size calculation At baseline, cord nAS was significantly associated with brain T2 LV and EDSS (Table 3). Associations between baseline nLosseff and these variables were also significant, but weaker (Table 3). At follow-up, cord nAS was significantly correlated with brain T2 LV and EDSS, while the associations between follow-up nLosseff values and these variables were not significant (Table 3). No correlations were found between the change in nAS over time and concomitant changes of brain T2 LV (p = 0.85) and cord lesion number (p = 0.83). Similarly, no correlations were found between the changes in nLosseff and conventional MRI metrics changes (p = 0.85 for T2 LV and 0.67 for cord lesions). In MS patients, the only variable associated with a clinical worsening at follow-up was age (one of the covariates of the logistic
Table 4 reports the estimated sample sizes per arm for a placebo-controlled treatment trial in which an active treatment reduces CSA loss over time. We found that the number of subjects needed to observe any treatment effect was almost halved when using the AS compared to the Losseff method (e.g., n = 72 vs n = 164 subjects to detect a 50 % treatment effect).
Discussion Measuring spinal cord atrophy is not straightforward because of the intrinsic cord curvature, its small size and the complex anatomy of surrounding structures [29, 30]. Given
Table 2 Mean normalized cord cross-sectional areas, at baseline and at follow-up, measured with the Losseff and active surface (AS) methods in healthy controls (HC) and patients with multiple sclerosis (MS) Group
Method
Baseline
Follow-up
Cord area change (%)
Regression coefficient*
HC MS patients
2
nAS (mm ) (SD) nAS (mm2) (SD)
81.7 (3.6) 74.5 (9.2)
81.6 (3.3) 72.2 (8.6)
-0.1 -3.0
-0.12 -2.28
HC
nAS-C2 (mm2) (SD)
76.9 (5.3)
76.8 (5.3)
-0.07
MS patients
nAS-C2 (mm2) (SD)
71.5 (9.6)
69.1 (8.7)
-3.4
-2.47
HC
nLosseff (mm2) (SD)
71.7 (7.3)
72.5 (5.5)
?1.0
?0.79
-1.11
MS patients
nLosseff (mm2) (SD)
66.7 (9.9)
64.3 (9.7)
-3.4
-2.32
3.94
0.068
SE
p*
p**
0.78 0.00001
0.02
0.15
0.89
0.04
4.27
0.0002
0.30 5.12
0.31
0.01
0.0004
* Effect of time, longitudinal linear model adjusted for age and follow-up duration ** Time by group interaction, longitudinal linear model adjusted for age and follow-up duration SD standard deviation, SE standardized effect, nAS normalized cord cross-sectional area calculated with the active surface method, nAS-C2 normalized cord cross-sectional area calculated with the active surface method at C2 level, nLosseff normalized cord cross-sectional area calculated with the Losseff method
123
J Neurol Table 3 Associations between normalized cord cross-sectional areas, at baseline and at follow-up, measured with the Losseff and active surface (AS) method, and clinical/conventional MRI variables Baseline T2 LV (r, p)a
Follow-up T2 LV (r, p)a
Baseline EDSS (r, p)a
Follow-up EDSS (r, p)a
Baseline nAS
-0.49, 0.003
NA
-0.59, \0.001
NA
Baseline nLosseff
-0.32, 0.06
NA
-0.40, 0.01
NA
Follow-up nAS
NA
-0.38, 0.02
NA
-0.50, 0.002
Follow-up nLosseff
NA
-0.30, 0.07
NA
-0.31, 0.06
a
Spearman’s rank correlation coefficient
NA not applicable, LV lesion volume, EDSS expanded disability status scale, nAS normalized cord cross-sectional area calculated with the active surface method, nLosseff normalized cord cross-sectional area calculated with the Losseff method
Table 4 Estimated sample sizes per arm for a placebo-controlled treatment trial aimed at reducing cord cross-sectional area (CSA) loss over 2.3 years of follow-up (power = 80 %, significance level a = 0.05) Treatment effect (%)
Sample size per arm nAS
nLosseff
10
1751
4054
25
282
649
50
72
164
75
33
74
90
23
52
nAS normalized cord cross-sectional area calculated with the active surface method, nLosseff normalized cord cross-sectional area calculated with the Losseff method
Fig. 1 Non-parametric kernel estimate plots of cord cross-sectional area (CSA) along the cervical cord for healthy controls (red) and patients with multiple sclerosis (MS) (blue) at baseline (continuous lines) and at follow-up (dashed lines). Distance along the cervical cord was normalized by dividing it by subject’s cervical cord length: the normalized cord distance varies from 0 [most superior axial slice in which the odontoid process of the epistropheus (C2) was still visible] to 1 (inferior border of C5). See text for further details
the technical challenges of implementing accurate 3D spinal cord measurements, the majority of methods for assessing cord atrophy in MS have focussed, until recently, on analyzing small cord portions [6, 28, 29] without clear data about reproducibility over time. The AS method [7] holds great promise for overcoming some of the above-mentioned limitations. Previous reports have already shown a high intra-observer and inter-observer reproducibility of cord CSA measurements both in cross-sectional (inter-observer coefficient of variation (CoV) = 1.07 %; intra-observer CoV = 0.44 % [7]) and in short-term longitudinal designs (inter-observer CoV = 0.03–0.2 %; intra-observer CoV = 0.1–0.4 %,
over 6 months of follow-up, depending on the scanning sequence [10]). Moreover, AS cord measurements were found to be less time-consuming to perform and had the advantage of covering large portions of the spinal cord, if appropriate 3D acquisitions were available [7]. The main findings of our study were that the stability over time of cord CSA values in HC tended to be higher with AS than with Losseff measurements; and the sensitivity of the AS measurements to longitudinal modifications of cervical cord CSA in MS tended to be higher than that of Losseff measurements, even if a formal statistical comparison was not done due to the small number of subjects. In HC, significant cord CSA changes over 2 years of follow-up are not expected, especially in the age range of the controls included in our study. Previous cross-sectional studies indicated that cord volume loss occurs with aging to only a small extent in the healthy elderly [31–33], while previous longitudinal studies showed no significant changes of cord CSA values in HC over time [5, 26]. In line with these considerations, longitudinal changes of CSA were not significant in our controls. However, while cord
123
J Neurol
nAS values at follow-up were almost identical to those at baseline, a non-significant increase of cord area (?1 %) was found with nLosseff values. This unexpected result is similar to that obtained in a previous study [5], which also used the Losseff method for cord area measurements (?3 % of mean CSA over 3 years in HC). Since age-related neurodegenerative mechanisms occurring in the spinal cord are likely to result in tissue loss [34], this result is likely to be due predominantly to measurement errors. In MS patients, a significant development of cord atrophy was found with both the AS and Losseff techniques. The rate of loss of cord CSA (which was about 3–3.5 % over 2 years) was in line with previous findings [5, 17, 19, 21], but the statistical significance of the detected change was higher with the AS than with the Losseff method. This suggests that the AS method, thanks to a lower variance of measurements, might be more sensitive and, as a consequence, its application might be more effective in studies with a small sample size. In line with previous reports [4], we found that cord measurements were significantly correlated with disability. This result confirms the clinical relevance of the spinal cord, damage to which is likely to disproportionately affect the locomotor and sensory functions. Since the EDSS is strongly weighted towards locomotor disability, it has often been found to be correlated with atrophy of the cervical cord [4, 8, 9, 11]. Cord CSA was correlated with brain T2 LV, but not with the number of cervical cord lesions. The finding that cord atrophy is more strongly correlated with brain rather than cord damage is not novel [8–10] and confirms that individual cord lesions may play only a minor role in local atrophy and that Wallerian degeneration of long fiber tracts might be one of the main factors leading to tissue loss in the cord [2]. The strongest correlations between cord CSA and clinical/conventional MRI variables were found when AS measurements were used. This result strengthens the case for using the AS method for assessing cord area in cross-sectional and longitudinal studies. In our sample, the development of cord atrophy was not associated with concomitant clinical worsening. This disappointing result might be explained by the small patient sample in our study and by the relatively low proportion of MS patients (11 out of 35, about 30 %) showing a clinical worsening at follow-up. The duration of follow-up may also have played a role, since disability, as assessed with EDSS, takes years to develop. Future longitudinal studies with larger sample sizes and longer follow-up might be able to show an association between irreversible cord tissue loss and clinical changes. The analysis of cord CSA plots from C2 to C5 showed a significant effect of group, with a net separation of the curves of controls and MS patients at the two timepoints along the entire length of the cervical cord analyzed. This
123
suggests that the pathological processes we are quantifying with atrophy diffusely affect the cervical cord. While baseline and follow-up CSA plots from HC overlapped over much of their length, there was the appearance of a decrease of cord CSA in MS patients at follow-up compared to baseline. This decrease, which did not reach statistical significance, was more evident in the most caudal sections of the cord (around C5) than at the C1/C2 vertebral level. The non-uniform distribution of cord volume loss, together with the small study sample size, are likely to have caused an underpowering of the multilevel nested design of the statistical test. Future studies with larger sample size and longer follow-up may take a full advantage from the use of the AS method for the analysis of long cord segments, and might be more sensitive in detecting cord regions particularly susceptible to atrophy development over time. This study is not without limitations, including the fact that it may not be straightforward to apply the AS method, especially along large cord portions, to subjects with vertebral column disease that causes cervical cord compression. However, the technique also gave very good results (in terms of reproducibility and sensitivity to cord area changes) when applied to a limited portion of the cord around C2, as shown by our study and previous reports [10]. Moreover, the small sample size did not allow us to perform a powered comparison of cord atrophy changes among phenotypes. Finally, a high-resolution, three-dimensional acquisition sequence is needed to perform reliable cord measurements, and the application of such a sequence in a multicenter study might be problematic, especially when dealing with advanced imaging techniques, such as phase-sensitive inversion recovery sequences, as proposed by Kearney et al. [10]. Nevertheless, standard magnetization-prepared 3D T1 scans seem to be sufficient to obtain precise and reliable cord measurements. Given the need for sensitive biomarkers of neurodegeneration, we performed a power calculation study to evaluate the use of CSA as an endpoint in clinical trials. Our calculations showed that fewer subjects per arm (about the half) would be needed with the AS method, compared to the Losseff method, to detect a significant treatment effect. This result confirms previous findings [10] and suggests that: (1) the use of cord CSA as an endpoint in clinical trials would be feasible; and (2) the AS method is highly sensitive to cord volume changes, allowing small sample sizes. Together with the previously observed reliability of measurements in a multicenter setting [8], these findings suggest that cord nAS might be a valuable surrogate outcome measure for monitoring disease progression and treatment efficacy in MS. Acknowledgements This study was partially supported by a grant from Fondazione Italiana Sclerosi Multipla (FISM2014/PMS/6).
J Neurol Conflicts of interest
The authors declare no conflict of interest.
Ethical standard Approval was obtained from the local ethical standards committee on human experimentation and written informed consent from all subjects before enrolment. The study has been performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments.
References 1. Nijeholt GJ, van Walderveen MA, Castelijns JA et al (1998) Brain and spinal cord abnormalities in multiple sclerosis. Correlation between MRI parameters, clinical subtypes and symptoms. Brain 121(Pt 4):687–697 2. Evangelou N, DeLuca GC, Owens T, Esiri MM (2005) Pathological study of spinal cord atrophy in multiple sclerosis suggests limited role of local lesions. Brain 128(Pt 1):29–34 3. DeLuca GC, Ebers GC, Esiri MM (2004) Axonal loss in multiple sclerosis: a pathological survey of the corticospinal and sensory tracts. Brain 127(Pt 5):1009–1018 4. Filippi M, Rocca MA (2013) Multiple sclerosis: Linking disability and spinal cord imaging outcomes in MS. Nat Rev Neurol 9(4):189–190 5. Rashid W, Davies GR, Chard DT et al (2006) Increasing cord atrophy in early relapsing-remitting multiple sclerosis: a 3 year study. J Neurol Neurosurg Psychiatry 77(1):51–55 6. Losseff NA, Webb SL, O’Riordan JI et al (1996) Spinal cord atrophy and disability in multiple sclerosis. A new reproducible and sensitive MRI method with potential to monitor disease progression. Brain 119(Pt 3):701–708 7. Horsfield MA, Sala S, Neema M et al (2010) Rapid semi-automatic segmentation of the spinal cord from magnetic resonance images: application in multiple sclerosis. Neuroimage 50(2):446–455 8. Rocca MA, Horsfield MA, Sala S et al (2011) A multicenter assessment of cervical cord atrophy among MS clinical phenotypes. Neurology 76(24):2096–2102 9. Lukas C, Sombekke MH, Bellenberg B et al (2013) Relevance of spinal cord abnormalities to clinical disability in multiple sclerosis: MR imaging findings in a large cohort of patients. Radiology 269(2):542–552 10. Kearney H, Rocca MA, Valsasina P et al (2014) Magnetic resonance imaging correlates of physical disability in relapse onset multiple sclerosis of long disease duration. Mult Scler 20(1):72–80 11. Daams M, Weiler F, Steenwijk MD et al (2014) Mean upper cervical cord area (MUCCA) measurement in long-standing multiple sclerosis: Relation to brain findings and clinical disability. Mult Scler 20(14):1860–1865 12. Barkhof F, Calabresi PA, Miller DH, Reingold SC (2009) Imaging outcomes for neuroprotection and repair in multiple sclerosis trials. Nat Rev Neurol 5(5):256–266 13. Zivadinov R, Reder AT, Filippi M et al (2008) Mechanisms of action of disease-modifying agents and brain volume changes in multiple sclerosis. Neurology 71(2):136–144 14. Filippi M, Preziosa P, Rocca MA (2014) Magnetic resonance outcome measures in multiple sclerosis trials: time to rethink? Curr Opin Neurol 27(3):290–299 15. Vrenken H, Jenkinson M, Horsfield MA et al (2013) Recommendations to improve imaging and analysis of brain lesion load and atrophy in longitudinal studies of multiple sclerosis. J Neurol 260(10):2458–2471 16. Ingle GT, Stevenson VL, Miller DH, Thompson AJ (2003) Primary progressive multiple sclerosis: a 5-year clinical and MR study. Brain 126(Pt 11):2528–2536
17. Lin X, Tench CR, Turner B, Blumhardt LD, Constantinescu CS (2003) Spinal cord atrophy and disability in multiple sclerosis over four years: application of a reproducible automated technique in monitoring disease progression in a cohort of the interferon beta-1a (Rebif) treatment trial. J Neurol Neurosurg Psychiatry 74(8):1090–1094 18. Stevenson VL, Ingle GT, Miller DH, Thompson AJ (2004) Magnetic resonance imaging predictors of disability in primary progressive multiple sclerosis: a 5-year study. Mult Scler 10(4):398–401 19. Agosta F, Absinta M, Sormani MP et al (2007) In vivo assessment of cervical cord damage in MS patients: a longitudinal diffusion tensor MRI study. Brain 130(Pt 8):2211–2219 20. Furby J, Hayton T, Altmann D et al (2010) A longitudinal study of MRI-detected atrophy in secondary progressive multiple sclerosis. J Neurol 257(9):1508–1516 21. Lukas C, Knol DL, Sombekke MH et al (2015) Cervical spinal cord volume loss is related to clinical disability progression in multiple sclerosis. J Neurol Neurosurg Psychiatry 86(4):410–418 22. Stevenson VL, Leary SM, Losseff NA et al (1998) Spinal cord atrophy and disability in MS: a longitudinal study. Neurology. 51(1):234–238 23. Leary SM, Miller DH, Stevenson VL, Brex PA, Chard DT, Thompson AJ (2003) Interferon beta-1a in primary progressive MS: an exploratory, randomized, controlled trial. Neurology 60(1):44–51 24. Montalban X, Sastre-Garriga J, Tintore M et al (2009) A singlecenter, randomized, double-blind, placebo-controlled study of interferon beta-1b on primary progressive and transitional multiple sclerosis. Mult Scler 15(10):1195–1205 25. Klein JP, Arora A, Neema M et al (2011) A 3T MR imaging investigation of the topography of whole spinal cord atrophy in multiple sclerosis. AJNR Am J Neuroradiol 32(6):1138–1142 26. Kearney H, Yiannakas MC, Abdel-Aziz K et al (2014) Improved MRI quantification of spinal cord atrophy in multiple sclerosis. J Magn Reson Imaging 39(3):617–623 27. Polman CH, Reingold SC, Banwell B et al (2010) Diagnostic criteria for multiple sclerosis: 2010 revisions to the McDonald criteria. Ann Neurol 69(2):292–302 28. Lin X, Blumhardt LD, Constantinescu CS (2003) The relationship of brain and cervical cord volume to disability in clinical subtypes of multiple sclerosis: a three-dimensional MRI study. Acta Neurol Scand 108(6):401–406 29. Tench CR, Morgan PS, Constantinescu CS (2005) Measurement of cervical spinal cord cross-sectional area by MRI using edge detection and partial volume correction. J Magn Reson Imaging 21(3):197–203 30. Agosta F, Filippi M (2007) MRI of spinal cord in multiple sclerosis. J Neuroimaging 17(Suppl 1):46S–49S 31. Agosta F, Lagana M, Valsasina P et al (2007) Evidence for cervical cord tissue disorganisation with aging by diffusion tensor MRI. Neuroimage 36(3):728–735 32. Valsasina P, Horsfield MA, Rocca MA, Absinta M, Comi G, Filippi M (2012) Spatial normalization and regional assessment of cord atrophy: voxel-based analysis of cervical cord 3D T1weighted images. AJNR Am J Neuroradiol 33(11):2195–2200 33. Yanase M, Matsuyama Y, Hirose K et al (2006) Measurement of the cervical spinal cord volume on MRI. J Spinal Disord Tech 19(2):125–129 34. Cruz-Sanchez FF, Moral A, Tolosa E, de Belleroche J, Rossi ML (1998) Evaluation of neuronal loss, astrocytosis and abnormalities of cytoskeletal components of large motor neurons in the human anterior horn in aging. J Neural Transm 105(6–7):689–701
123
