Irreversible Disability and Tissue Loss in Multiple Sclerosis:  A Conventional and Magnetization Transfer Magnetic Resonance Imaging Study of the Optic Nerves FREE

THE FACTORS leading to irreversible neurologic disability in multiple sclerosis (MS) are still poorly understood because of the heterogeneity of the pathologic substrates of the disease and the structural and functional complexity of the brain and the spinal cord. In this context, the optic nerve (ON) is of special interest, because it might provide a model for improving our understanding of how MS causes irreversible disability. There are several reasons for this. First, the ON is a typical site of MS lesions.¹ Second, and different from what happens for other structures of the central nervous system, the ON subserves a specific function, which can be assessed reliably by clinical and neurophysiologic methods.^2- 3 Finally, several recent studies^4- 10 have shown that magnetic resonance (MR) imaging techniques enable us not only to obtain high-quality images of the ON, but also to quantify reliably the extent of ON tissue damage and loss.

Although conventional MR imaging is very sensitive for the detection of MS lesions, it lacks specificity to the most destructive aspects of the disease (ie, severe and irreversible demyelination and axonal loss).¹¹ Also, conventional MR imaging is unable to detect the subtle changes known to occur in the normal-appearing tissue.¹¹ These limitations have been at least partially overcome by the use of modern MR imaging technology, including magnetization transfer imaging, and by measuring the volumes of different portions of the central nervous system.¹¹ Although neither of these 2 approaches provides specific information about demyelination and axonal loss, they allow estimates to be obtained of overall tissue damage in MS, which are thought to reflect more closely destructive pathologic changes of MS within and outside T2-visible lesions than conventional MR imaging does. Consistent with this, MS-related disability increases with decreasing brain and cord volumes^12- 18 and magnetization transfer ratio (MTR) values,^19- 24 and a postmortem study of patients with MS²⁵ has found a strong correlation between MTR values with the amount of residual axons in lesions and normal-appearing tissue.

In a previous study,⁹ our group showed that patients with Leber hereditary optic neuropathy (LHON), in whom the major histopathological hallmark of impaired vision is neuronal and axonal loss,²⁶ have dramatically reduced ON volumes and MTR values when compared with matched healthy volunteers. In the present study, we measured volumes and MTR of the ONs from a group of patients with MS and a single previous episode of optic neuritis and compared them with those of patients with LHON and matched healthy volunteers. To improve our understanding of the factors underlying irreversible deficits in MS, we divided affected ONs from patients with MS into 2 groups according to the degree of subsequent clinical recovery, and reasoned that if axonal loss is the major factor associated with "fixed" MS deficits, ONs with incomplete clinical recovery should have had volumes and MTR values similar to those of ONs from patients with LHON, whereas previously affected ONs with subsequent satisfactory clinical recovery should have had volumes and MTR values similar to those of ONs from healthy volunteers. Since demyelination might also contribute to reduced ON volumes and MTR values,²⁵ in patients with MS we also obtained visual evoked potentials (VEPs). Persistent delay of the P100 wave of the pattern-reversal VEP after an episode of optic neuritis is considered to be compatible with residual demyelination within the ON.^2- 3 Thus, we postulated that if irreversible demyelination is not a critical factor contributing to fixed deficits in MS, the P100 latency of ONs from patients with MS should be independent of the degree of clinical recovery.

We studied 30 patients with relapsing-remitting and secondary progressive MS.²⁷ All patients had to have a single previous episode of acute optic neuritis confirmed by a complete and standardized neuro-ophthalmologic assessment (including color vision and visual field testing) and to have been clinically stable for at least 6 months. In addition, they had had neither relapses nor corticosteroid treatment during the preceding 3 months. Optic nerves from patients with MS were classified as follows: (1) clinically affected with incomplete or no recovery when best-corrected visual acuity (VA), assessed with wall charts, was less than 20/25 at least 6 months after the attack; (2) clinically affected with recovery when best-corrected VA was equal to or greater than 20/25 at least 6 months after the attack; and (3) clinically unaffected. Two control groups were identified. The first consisted of 18 sex- and age-matched healthy volunteers, with no history of neurologic diseases and with normal results of neurologic and ophthalmologic examinations. The second control group consisted of 10 age-matched patients with LHON documented by 1 of the 3 primary mitochondrial DNA mutations (7 patients had the 11778, 2 the 3460, and 1 the 14484 mutation). A complete ophthalmologic evaluation was performed by one physician (S.B.), who was unaware of the MR imaging results, in all subjects. In patients with MS, disability was measured with the Expanded Disability Status Scale.²⁸ Additional demographic and clinical characteristics of the subjects studied are reported in Table 1. Local ethical committee approval and written informed consent from all the subjects were obtained before study initiation.

The ONs of all subjects were imaged in a single session, by means of a standard head coil and a scanner operating at 1.5 T. The following sequences were obtained: (1) T2-weighted turbo spin echo (repetition time [TR], 4230 milliseconds; echo time [TE], 119 milliseconds; echo train length, 15; 15 coronal slices with 3-mm thickness; interslice gap, 0.3 mm; matrix size, 180 × 512; and field of view [FOV], 156 × 250 mm); (2) T1-weighted spin echo (TR, 500 milliseconds; TE, 14 milliseconds; 15 coronal slices with 3-mm thickness; interslice gap, 0.3 mm; matrix size, 224 × 512; and FOV, 156 × 250 mm); and (3) 2-dimensional gradient echo (TR, 640 milliseconds; TE, 12 milliseconds; flip angle, 20°; 15 contiguous coronal slices with 5-mm thickness; matrix size, 256 × 256; and FOV, 250 × 250 mm), with and without an off-resonance radiofrequency saturation pulse (offset frequency, 1.5 kHz; gaussian envelope duration, 7.68 milliseconds; flip angle, 500°). During ON imaging, all subjects were asked to close their eyes and possibly avoid eye movements. We also obtained dual-echo turbo spin-echo (TR, 3300 milliseconds; TE, 16-98 milliseconds; echo train length, 5; 24 contiguous axial slices with 5-mm thickness; matrix size, 192 × 256; and FOV, 188 × 250 mm) images of the brain from all subjects.

The volumes of the ON were measured on T1-weighted images. A single observer (M.I.), unaware of subjects' identity, calculated the areas of the ON sections from 11 consecutive 3-mm-thick slices, starting from the first slice showing the chiasm and moving backward (ie, toward the eye) and using a local thresholding techniquefor tissue segmentation.²⁹ The ON volumes were then calculated by multiplying the sum of these areas for the slice thickness. From the corresponding 2 gradient echo images and after image coregistration with the use of an algorithm based on mutual information,³⁰ we derived quantitative MTR maps of the ON, as extensively described elsewhere.²¹ After ON and chiasm segmentation from the MTR maps, the average MTR value was calculated for the same tract considered for ON volume measurements. Because of the different slice thickness, 7 consecutive 5-mm-thick MTR slices, starting from the first slice showing the chiasm and going backward, were used. The MTR value for each slice was derived from a small region of interest (4 pixels) placed in the center of the ON section. The pixels adjacent to the nerve borders were excluded from the analysis, to minimize the influence of partial volume effects from the cerebrospinal fluid. An ON was considered to be atrophic or to have abnormal MTR values when its volume or MTR values were at least 2 SDs below the mean values obtained for the ONs from the healthy volunteers. Dual-echo images of the brain were reviewed by 2 observers (M.I. and S.G.), unaware of subjects' identity, to identify hyperintense lesions.

A black-and-white checkerboard pattern on a square television screen, reversing every 0.7 milliseconds, was used to obtain pattern-reversal VEPs. The entire stimulating field was 11°; 30' check size was used. The contrast was kept constant at 70%. The cortical responses were recorded from Oz referred to Cz. The bandwidth was 1 to 250 Hz. An average of 100 responses were used to obtain the VEP. The P100 latency was considered abnormal if it was more than 2.5 SDs above the mean of the healthy volunteers.

Volumes and MTR values among the 5 groups of ONs (MS affected, with no or incomplete recovery; MS affected, with recovery; MS unaffected; normal; and LHON) were compared by means of a random effect model (random intercept), which allows correlations existing for any of the variables tested between the 2 ONs from the same individual to be accounted for. Four contrasts were decided a priori on the basis of the available degrees of freedom and biological considerations. The following were the a priori contrasts we tested: (1) normal ON vs MS clinically unaffected ON; (2) MS clinically unaffected ON vs MS affected ON with recovery; (3) MS affected ON with recovery vs MS affected ON with incomplete or no recovery; and (4) MS affected ON with incomplete or no recovery vs LHON ON. Univariate correlations were assessed with the Spearman rank correlation coefficient, with P values corrected to account for the correlations existing between the 2 ONs from the same individual. A multivariable linear regression model was used to generate a composite MR score to be correlated with the VA in patients with MS. This was computed by means of a linear combination of ON volumes (reflecting the extent of tissue loss) and MTR values (reflecting the intrinsic abnormality of the residual tissue). The weight of each MR parameter resulted from the coefficients estimated by the regression model. The magnitude and the significance of the correlation was evaluated by a nonparametric Spearman correlation analysis. Pairwise comparisons were performed with a Student t test for unpaired data, again correcting for the correlation existing between the 2 ONs of the same individual.

In the MS population, there were 18 clinically unaffected ONs, 20 affected ONs with recovery (VA was 20/25 in 5 patients and 20/20 in 15 patients), and 22 affected ONs with incomplete or no recovery (VA was light perception in 1 case, counting fingers in 2, 20/100 in 5, 20/80 in 2, 20/70 in 1, 20/40 in 2, and 20/30 in 9). In patients with LHON, VA was 20/400 bilaterally in 3 patients, counting fingers OD and 20/100 OS in 2 patients, light perception OD and hand movements OS in 1 patient, counting fingers bilaterally in 2 patients, and hand motions bilaterally in 1 patient.

No brain or ON lesions were detected on any of the MR images obtained from healthy volunteers and patients with LHON. All patients with MS had abnormal brain MR images and met the criteria of Fazekas et al.³¹ The MR images of 2 ONs from patients with MS (1 with no or incomplete recovery and 1 with recovery) were of suboptimal quality, and, as a consequence, they were not considered in the following analyses. Five MS ONs with incomplete or no recovery (24%) and 3 of those with recovery (16%) had a hyperintense lesion on the dual-echo images. The lengths of these lesions ranged from 3 to 9 mm. Seven of them were located in the intracanalicular portion, while 1 was located in the intraorbital portion of the ON.

Table 2 gives the mean volumes and MTR values from the ONs of all the subjects. Volumes and MTR values of the affected MS ONs with incomplete or no recovery were both significantly lower (P = .002 and P<.001, respectively) than the corresponding quantities of the MS ONs with recovery, whereas both of these quantities did not differ from those of the ONs from patients with LHON. Also, volumes and MTR values of the affected MS ONs with recovery did not differ from the corresponding quantities of clinically unaffected ONs, which in turn had volumes and MTR values similar to those of the ONs from healthy volunteers. The same results were obtained when a VA cutoff of 20/50 was used for incomplete recovery (data not shown).

Visual evoked potentials were obtained from 25 patients with MS (14 clinically unaffected ONs, 18 affected ONs with recovery, and 18 affected ONs with incomplete or no recovery). The VEPs could not be obtained from 9 ONs with incomplete or no recovery. The mean P100 latencies were 99.6 milliseconds (SD, 4.5 milliseconds) for healthy control subjects, 114.1 milliseconds (SD, 19.6 milliseconds) for clinically unaffected ONs (P = .008 vs control subjects), 126.7 milliseconds (SD, 22.6 milliseconds) for affected ONs with recovery (P<.001 vs control subjects), and 154.5 milliseconds (SD, 20.3 milliseconds) for affected ONs with incomplete or no recovery whose P100 could be elicited (P<.001 vs control subjects). In Table 3, the numbers and percentages of ON with atrophy, abnormal MTR values, and abnormal P100 latency are reported for each of the categories studied. The percentage of ONs with atrophy and abnormal MTR values were similar in MS ONs with incomplete or no recovery and ONs from patients with LHON.

In patients with MS, ON volumes and MTR were moderately correlated (r = 0.46, P = .001). The VA was correlated with ON volume (r = 0.39, P = .01), MTR values (r = 0.49, P = .001), and P100 latency (r = −0.57, P<.001). No statistically significant correlation was found between P100 latency and MTR values (r = −0.10), whereas P100 latency and ON volumes were moderately correlated (r = −0.31, P = .05). A moderate correlation was also found between the VA and the MR composite score (r = 0.53, P<.001).

There is increasing appreciation that neurodegeneration is the major factor underlying the accumulation of disability in MS and that inflammatory demyelination alone is not sufficient to explain the neurologic dysfunction in this disease.³² Although neurodegeneration has been known to occur in MS from the early descriptions of Charcot, the major pieces of evidence suggesting the importance of this aspect of the disease for disability in MS come from MR spectroscopy studies,³² showing that the level of brain N-acetyl-aspartate, a marker of axonal integrity,³³ is strictly correlated with the degree of disability in patients with MS.^34- 36 Accumulation of disability in MS can be secondary to a progressive disease course or to incomplete recovery after acute relapses.²⁷ Previous studies have been mainly focused on understanding the progressive accumulation of disability,³² whereas the factors related to incomplete recovery after an acute injury are still to be elucidated. This lack of knowledge is secondary to the fact that, because of the anatomic and functional complexity of the brain and the spinal cord and the multifocality of the disease, it is virtually impossible to define the exact role of regional tissue damage in determining fixed neurologic deficits in MS. In this respect, the assessment of MS-related tissue damage in the ON can be viewed as an excellent model to investigate the effect of the various components of MS abnormality on the clinical manifestations of the disease. Until recently, however, MR imaging of the ON was limited by its small size, motion artifacts, and the effect of surrounding cerebrospinal fluid, lipid, and bony structures.³⁷ Although MR imaging of the ON remains technically challenging, several studies have shown that with careful optimization, it is possible to obtain high-quality T1- and T2-weighted images, as well as reliable quantitative measurements of MR quantities, including the MTR.^4- 10

Against this background, the present study was performed to increase our understanding of the factors associated with irreversible disability after MS relapses. This study has demonstrated a reduction in the volume and MTR values of the ON from patients with MS, which, after an episode of optic neuritis, showed no or only an incomplete functional recovery. The volume and MTR reductions found in these ONs were not different from those found in patients with LHON, a condition in which there is loss of neurons and axons.²⁶ In contrast, affected ONs that recovered after the disease injury had volumes and MTR values higher than those of ONs with incomplete or no recovery and similar to those of clinically unaffected ONs and of ONs from healthy volunteers. We also showed a graded relationship between these 2 MR quantities and VA in patients with MS. Although neither ON volume reductions nor MTR decreases are specific markers of axonal loss, we interpret these findings as an indication that loss of axonal integrity is associated with persistent functional deficit in MS due to incomplete recovery from relapses.

Several previous studies have shown that brain and spinal cord atrophy occurs in MS and correlates with the degree of disability.^12- 18 Although the pathologic basis of atrophy in MS is not fully elucidated, axonal loss is likely to be an important contributing factor, since axons represent the largest proportion of the white matter volume,³⁸ and there is evidence of considerable axonal damage in MS lesions and normal-appearing tissue.^39- 41 All of this is in keeping with the findings of this study showing that a similar average degree of ON volume and a similar frequency of ON atrophy are detectable in the ON with incomplete or no recovery from patients with MS and in those from patients with LHON.

The same considerations apply to the interpretation of the MTR changes seen in this study. Low MTR values correspond to areas where the relative proportion of water bound to macromolecules is reduced, indicating a loss of microstructural tissue integrity. Consistent with this, several studies have shown that brain and cervical cord MTR changes are correlated with disability in patients with MS.^19- 24 Axonal loss is likely to be an important contributor to MTR decreases in MS for several reasons. First, a postmortem study of patients with MS has shown that a strict correlation exists between MTR values and percentage amounts of residual axons.²⁵ Second, MTR reduction has been found to correlate well with N-acetyl-aspartate–creatine ratio measured in MS lesions.⁴² Third, low MTR values have been found in animal models of wallerian degeneration⁴³ and diffuse axonal injury.⁴⁴

Given that ON volume and MTR reductions can also be related to persistent demyelination, we obtained VEPs from patients with MS. Because an increased P100 latency weeks or months after an acute episode of optic neuritis is considered to be a marker of residual demyelination,^2- 3 we focused our analysis on this component of the VEP. Although the ONs with incomplete or no recovery were those in which the P100 was more frequently abnormal, we also found that a large proportion of clinically unaffected ONs (64%) and affected ONs with subsequent recovery (89%) had increased P100 latencies. The mean P100 latencies of clinically unaffected ONs and affected ONs with subsequent recovery were also significantly increased compared with those from healthy subjects. These findings, which agree with those of previous studies,^2,5 support the concept that demyelination per se is not sufficient to explain fixed neurologic symptoms after acute damage.

In this study, we also showed significant correlations between VA and ON volumes and MTR values. The composite MR score, based on a measure reflecting the amount of tissue loss (ON volume) and on a measure reflecting the integrity of the residual tissue (ON MTR), was also correlated with the VA. The magnitude of these correlations was, however, moderate and perhaps lower than expected. There are several reasons that might explain this finding. First, VA is not only related to the amount of tissue damage occurring in the ON, but also to that occurring along the entire visual pathways. Given the high interpatient variability in MS lesion location, the amount of abnormality occurring beyond the ON might well be a factor reducing the strength of the correlation. Second, functional MR imaging studies have shown a marked intersubject variability of cortical adaptive reorganization after MS injury,^45- 46 including optic neuritis.^47- 48 Again, this might yet be an additional confounding factor. Third, atrophy and abnormal MTR values are found in 20% to 30% of clinically unaffected ONs and in about 20% to 40% of the affected ONs with subsequent satisfactory clinical recovery. This indicates that the system is redundant enough to limit, at least to a certain extent, the consequences of tissue loss and damage. Finally, as mentioned earlier, ON volumes and MTR values do not simply reflect tissue loss and damage, and, as a consequence, other factors might lead to underestimation (eg, reactive gliosis in the case of ON volume measurements) or overestimation (eg, demyelination in the case of MTR measurements) of the extent of "disabling" disease, thus reducing the magnitude of the correlation with the clinical outcome.

Accepted for publication September 24, 2001.

Author contributions: Study concept and design (Drs Inglese, Bianchi, Comi, and Filippi); acquisition of data (Drs Inglese, Ghezzi, Bianchi, Gerevini, and Martinelli); analysis and interpretation of data (Drs Inglese, Gerevini, Sormani, and Filippi); drafting of the manuscript (Drs Inglese and Filippi); critical revision of the manuscript for important intellectual content (Drs Inglese, Ghezzi, Bianchi, Gerevini, Sormani, Martinelli, Comi, and Filippi); statistical expertise (Dr Sormani); administrative, technical, or material support (Drs Inglese, Ghezzi, Bianchi, Gerevini, Martinelli, and Comi); supervision (Dr Filippi).

Corresponding author and reprints: Massimo Filippi, MD, Neuroimaging Research Unit, Department of Neuroscience, Scientific Institute and University Ospedale San Raffaele, Via Olgettina 60, 20132 Milan, Italy (e-mail: filippi.massimo@hsr.it).