Multiple sclerosis is traditionally described as clinical symptoms or signs of 2 central nervous system lesions separated in time and space that are not caused by other central nervous system disease. These lesions are detected with a history and neurologic exam--the sine qua non of a diagnosis of multiple sclerosis. MRI and CSF analysis strengthen the diagnosis (Poser et al 1983; McDonald et al 2001) and are each abnormal in more than 95% of definite multiple sclerosis cases. MRI is essential to rule out mimics of multiple sclerosis. A standardized brain and cord imaging and reporting protocol from the Consortium of Multiple Sclerosis Centers (CSMC) enhances diagnosis, clinical trial assessment, and monitoring of disease activity and damage (Simon et al 2006). CSF shows ongoing inflammation and oligoclonal bands, and it can exclude alternate causes.
Criteria that define “MS” have been revised several times: 2001, 2005, and 2010 (Polman et al 2011). Criteria now allow diagnosis of multiple sclerosis when new MRI lesions define separation in time and space—ie, a new T2 MRI lesion more than 30 days or enhancement more than 3 months after the initial event (McDonald et al 2001). These criteria are helpful in diagnosing multiple sclerosis in patients after a clinically isolated demyelinating syndrome. The widened inclusion criteria for a diagnosis can lead to spurious improvements in prognosis, the “Will Rogers phenomenon” (Sormani et al 2008). Patients with milder disease have been allowed into recent studies, necessitating larger patient cohorts to see any drug benefit.
MRI in diagnosis. On monthly scans, 90% of untreated relapsing-remitting patients will have MRI activity over 9 months (Li and Paty 1999). Early active MRI lesions are typically hyperintense on T2 with a hypointense ring, possibly containing activated macrophages. Large, fluffy T2 lesions often have preserved axons and repletion of oligodendroglia. Some acute plaques enhance with gadolinium, but early on, they can be almost isointense on T2 MRI (Bruck et al 1997). Late active lesions are less hyperintense on T2 and sometimes hypointense on T1-weighted images (“black holes”), indicating axonal loss and demyelination. Inactive lesions (demyelinated or myelinating) are hypointense on T2 scans and normal or hypointense on T1 scans.
Certain MRI features are typical of multiple sclerosis, such as multiple ovoid-shaped bright lesions on T2-weighted MRI, abrupt loss of T2 signal at the gray matter (“open ring”), and periventricular lesions, often radiating up from the corpus callosum or out from the ventricle (called “Dawson fingers”), especially near the body and posterior horn of the lateral ventricle. Other typical features are a lesion greater than 5 mm; lesions in the corpus callosum, brainstem, or cortical gray matter; and lesions below the tentorium, especially in the cerebellar peduncle (Offenbacher et al 1993). Corpus callosum lesions can also be caused by vascular disease, CADASIL, Susac syndrome, tumor, Marchiafava-Bignami disease, echovirus 9, and adrenoleukodystrophy.
Twenty percent of Gd+ lesions are reoccurrences at prior sites; these are larger and have lower magnetization transfer values than new lesions. T2 volume increases by 5% to 10% per year.
New criteria allow for early diagnosis at time of first attack, with caveats when pronouncing that 1 Gd+ and 1 Gd- lesion on a single scan equals “separation in time.” There is temporal variation in the duration of enhancement, plus some regional differences that interfere with comparison of Gd+ and Gd- lesions. For instance, Gd+ MRI lesions in white matter are easier to see than lesions in the cortical gray matter. Similarly, Gd enhancement lasts 2 weeks (median) to 3 weeks (mean) (Cotton et al 2003), but there is temporal variation in the duration of enhancement. Gd+ ring-enhancing lesions with central pallor correlate with disease severity and are larger and last longer than homogenously enhancing lesions. With this temporal variation, there could be a tail of enhancing lesions several weeks after a clinically isolated syndrome, so “old” and “new” lesions could still be from the initial insult.
A leak through the tight junctions of the blood-brain barrier is usually invoked as the cause of Gd enhancement. It is also likely that T cells constantly activate endothelial cells. Activated, enlarged endothelial cells pinocytose gadolinium and possibly gadolinium-protein complexes, and could cause a capillary blush on MRI (Brown 1978; McDonald and Barnes 1989; Claudio et al 1995). Treatment with glucocorticoids blocks MRI enhancement, probably through a direct effect on endothelial cells. IFN-beta and antibodies to VLA-4 also block gadolinium enhancement, interrupting T cell-endothelial interaction.
Perivenular spaces are prominent due to inflammation (lymphocyte cuffing) (Ge et al 2005). A “sand-like appearance” of the high convexity white matter is from dilated Virchow-Robin spaces (Achiron and Faibel 2002). These dilated spaces correlate with Gd+ lesions (Wuerfel et al 2008), suggesting there is diffuse activation of multiple areas of the brain.
Arcuate U fiber lesions below gyri are bright on T2. These arcs form an “open ring” and are a strong indicator of demyelinating disease. Antibody and complement in plaques (Lucchinetti type II) causes a T1-enhancing ring and hypointense T2 ring (Konig et al 2008). Restricted apparent diffusion coefficient (ADC) on MRI is more common in demyelinating disease than in tumors or abscesses.
T2 lesion volume has a weak correlation with disability (r = 0.2 to 0.3). T1 holes correlate more strongly, especially in secondary progressive multiple sclerosis (approximately 0.8). Without treatment, 56% of acute black holes will remain as permanent black holes. Callosal atrophy is a poor prognostic sign.
Spinal cord lesions are usually only 1 to 2 segments long, often multiple, and predominantly cervical (66%) (Bot et al 2004). They are diffuse in only 13% and are most likely in chronic multiple sclerosis. Almost all are peripherally located, not central, perhaps related to inflammation near pial veins. In recently diagnosed multiple sclerosis, cord lesions help define dissemination in space (85% are positive with brain plus cord MRI versus 66% using brain MRI alone). Over 90% of patients with primary progressive disease have brain lesions.
Normal-appearing white matter evinces slight abnormalities on MRI and MR spectroscopy and also more cerebral perfusion several months before overt MRI lesions appear (Goodkin et al 1998; Filippi et al 1999; Wuerfel et al 2004). MR spectroscopy shows biochemically abnormal lipid peaks in gray matter. Scans with 8 Tesla magnets reveal multiple lesions in gray matter and many in subpial and perivenous locations (Rammohan 2003). PET scans show decreased cerebral glucose utilization in frontal and parietal cortex, and this correlates with low NAA/Cr ratios on magnetic resonance spectroscopy (MRS) and low oxygen use on fMRI. These measures decline over time in multiple sclerosis and suggest hypometabolism and mitochondrial respiratory dysfunction.
Magnetization transfer (MTR) values in normal-appearing white matter are lowest in chronic progressive multiple sclerosis and continue to decline. MTR values are slightly low in benign multiple sclerosis, but do not decrease over time (Filippi et al 1999). Gray matter MTR loss correlates with progression on the EDSS and long-term disability. Of patients with clinically or lab-supported multiple sclerosis, 1.5% have negative conventional MRI scans yet low magnetization transfer values in pons, cerebellum, and periventricular areas (Filippi et al 1999). Changes are fastest in the corpus callosum. The mechanism for abnormal MTR may involve different immune or neurotrophic responses than in MRI-positive multiple sclerosis. T2+ plus T1+ MTR lesions have more demyelination and more axonal swelling and are often chronic inactive compared to T2+ only lesions (Fisher et al 2007). Gd+ lesions show some recovery in MTR. Diffusion tensor white matter tractography shows damage to the cingulum (executive function, attention, and speed) and U-fibers (attention, verbal and visuospatial speed, and memory) that correlate with cognitive impairment (Mesaros et al 2012).
Brain atrophy is present in early multiple sclerosis and averages 0.9% per year, versus 0.1% to 0.2% in healthy controls. It evolves 5 times more rapidly than in normal brains, and up to 14 times faster during active secondary progressive multiple sclerosis. Atrophy is largely from axonal loss, but also from demyelination and contraction of the neuropil. Axonal loss and T2 lesion volume are only partially correlated. Gray matter volume loss appears at first presentation in the thalamus and putamen, caudate, cerebellum, and cortex. White and gray matter atrophy correlate with cognitive decline and can be measured with third ventricular volume, a reflection of thalamic loss. Cerebellar gray atrophy correlates with loss of locomotion. Brain atrophy correlates better with changes in disability than T2 MRI lesions do, and is most predictive in the cervical cord (Zivadinov and Bakshi 2004). Cervical cord atrophy is pronounced in primary and secondary progressive multiple sclerosis. Atrophy is predicted by the presence of early T2 hypointensity and T1 Gd+ lesions (Simon et al 2000), T1 “black holes,” and low brain volume in some studies. T2 hypointensity is likely from iron deposition (perhaps neurotoxic) and correlates with later brain atrophy. CNS atrophy can also be measured with transcranial sonography of the third ventricle (width vs. disability is inversely correlated at r = -0.6). IFN-beta and glatiramer therapy slow atrophy. Atrophy is slower with weekly IFN-beta than with high-dose, high-frequency interferon, possibly because the latter reduces inflammation (“pseudoatrophy”) or because there are differences in neurotrophin induction.
Rare patients can develop nephrogenic systemic fibrosis (NSF) from certain preparations of gadolinium. NSF is more common with low glomerular filtration rate and diabetes, and it can be diagnosed with a skin biopsy.
CSF in diagnosis. The CSF is the best non-MRI marker of multiple sclerosis. CSF reflects CNS inflammation and the absence of a blood-brain barrier leak (Freedman et al 2005). The CSF, in order of increasing frequency and importance for a diagnosis of multiple sclerosis, shows elevated protein, a moderate increase in white blood cells, increased IgG, IgG/albumin index, IgG synthesis rate (Tourtellotte or Link formulas), and oligoclonal bands. Bands are present in 95% of multiple sclerosis patients, but only 80% of primary progressive patients (Siritho and Freedman 2009; Freedman personal communication 2011). The index, synthesis rate, and oligoclonal bands are increased in black compared to white patients by 30% to 40%, suggesting more active inflammation (Rinker et al 2007). The IgG indices correlate with CSF IL-10 levels. A positive MRI and CSF oligoclonal bands in clinically isolated syndromes predict conversion to definite multiple sclerosis in 90% of patients, at 9 times the rate with negative tests.
When suspected cases of multiple sclerosis have no CSF IgG bands, a repeat study will show new bands in half of them (Thompson and Freedman 2006). With only one CSF band, 1 in 3 later develops more oligoclonal bands and 1 in 27 has cerebral lymphoma (Davies et al 2003). Patients already diagnosed with multiple sclerosis who have negative oligoclonal bands, however, are less likely to develop more bands over time (Siritho and Freedman 2009). Serum bands are increased in 44% of multiple sclerosis patients and are 10 times more common in women than in men (Thompson and Freedman 2006). Inflammatory conditions such as SSPE, Lyme disease, syphilis, Behçet disease, lupus, and adrenoleukodystrophy often have unique CSF bands.
Intrathecal synthesis of antibodies to measles was described in 1962. Many other viruses, such as HHV-6, are also targets of a polyspecific B cell response. An index of CSF antibodies to measles, rubella, and herpes zoster (the MRZ reaction) improves sensitivity (Felgenhauer 1992), but is low in neuromyelitis optica. Antibodies to galactocerebroside, triose-phosphate isomerase, and glyceraldehyde-3-phosphate are elevated. These antibodies inhibit glycolytic enzyme activity and could disrupt neuroaxonal function. Reports of CSF antibodies to MOG and MBP have not been replicated, perhaps because of conformational changes in the protein.
The CSF cell count is often slightly elevated at 5 to 10 per cubic um. Eighty percent are T cells in stable multiple sclerosis and in healthy controls; T cells rise to 90% in active multiple sclerosis (Reder and Arnason 1985). CSF white blood cells are 90% CD3+ T cells (70% CD4, 20% CD8), 3% natural killer, 4% macrophages, and 5% B cells (Cepok et al 2001).The CD4/CD8 ratio reflects the blood (2/1) in relapsing-remitting multiple sclerosis, but CSF CD8 cells fall in progressive disease. Many CSF T and B lymphocytes are often activated blasts (Noronha et al 1980). B cells are at lower levels than in blood and are often blasts. A high CSF B cell to monocyte ratio in CSF correlates with IgG levels and with rapid disease progression in relapsing and progressive multiple sclerosis (Cepok et al 2001).
CSF reflects damage to brain cells. Protein markers include axon cytoskeleton markers (neurofilament light chains, present in first multiple sclerosis attacks; neurofilament heavy chains, highest in secondary progressive multiple sclerosis, but present even in clinically isolated syndromes; actin; NAA; tau; and tubulin), membrane markers (24S-hydroycholesterol, apoE4), glial markers (GFAP, S-100b), amyloid precursor protein (Teunissen et al 2009), and the endothelial marker, endothelin. N-acetyl aspartate (NAA) is abundant enough in neurons (10 mM) to be detectable with magnetic resonance spectroscopy, but decreases in CSF during secondary progressive multiple sclerosis as neurons die or lose function. Tau protein, a marker of axonal damage, increases 2-fold in progressive multiple sclerosis, but also increases in other inflammatory diseases, and levels correlate with the IgG index. 14-3-3, a neuronal, axonal, and glial protein, is present in 10% of patients with transverse myelitis and multiple sclerosis (de Seze et al 2002). In clinically isolated syndromes, 14-3-3 predicts an earlier conversion to multiple sclerosis. It is elevated in various dementias after extensive damage of the brain, especially in Creutzfeldt-Jacob disease where levels are high. Reduced ATP metabolites, perhaps reflecting high energy demand, correlate with more severe progression. Cystatin C may be uniquely cleaved by endogenous CSF proteases. Urinary myelin basic protein-like material (not MBP) also increases in progressive multiple sclerosis and correlates with the number of MRI T1 black holes. A high level of myelin basic protein (MBP) in CSF and MBP-like material in the urine reflects damage to myelin and oligodendroglia in progressive multiple sclerosis (Whitaker et al 1995). S-100b protein increases during flares. None of these is diagnostic in itself, but multiplex analysis coupled with reliable assays may be used in the future.
CSF neurotrophic factors also rise at some times. During recovery, neural cell adhesion molecule (N-CAM) and ciliary neurotrophic factor (CNTF) increase. During exacerbations, nerve growth factor sometimes increases, although it falls as the disease becomes advanced. However, there are low levels of CSF growth hormone, which is neuroprotective and induces insulin-like growth factor-1 and remyelination. DJ-1 (PARK7) in astrocytes and neurons stabilizes neuroprotectant Nrf2. DJ-1 is elevated 6-fold in relapsing-remitting Japanese multiple sclerosis versus non-inflammatory disease and correlates well with the multiple sclerosis severity scale (MSSS; r = 0.509) (Hirotani et al 2008).
Evoked potentials in diagnosis. Evoked potentials are occasionally helpful for confirming multiple sclerosis (eg, when MRI and CSF are normal), but they should not be used for the routine diagnosis of multiple sclerosis. The frequency of abnormal evoked potentials in definite multiple sclerosis is visual=90% (VEP), auditory=80% (BAEP), and somatosensory=70% (SSEP). VEP become less variable 4 weeks after the onset of optic neuritis, and are now quite standardized compared to the pre-MRI era. VEP latency reflects demyelination and has less spontaneous recovery than other visual measures such as fields and acuity. Auditory evoked potentials are seldom helpful in making the diagnosis. Neurophysiological studies such as vestibular evoked myogenic potentials, multifocal visual evoked potentials, motor (magnetic) evoked potentials, and the P300 event-related potential could also provide information about central nervous system function and prognosis (Leocani and Comi 2000).
Serum tests in diagnosis. Sedimentation rate or C-reactive protein, antinuclear and anticardiolipin antibodies (Cuadrado et al 2000), Sjögren syndrome A and B antibodies, angiotensin converting enzyme, vitamin B12 and methylmalonate, and possibly vitamin D levels should be ordered when appropriate. C-reactive protein, a marker for inflammation from many etiologies, has a modest correlation with relapses, progression, and MRI activity.
DNA transcription from many genes is controlled by methylation. Circulating methylated DNA profiles are highly abnormal in multiple sclerosis plasma.
Ophthalmologic diagnosis. The optic neuritis workup is funduscopy, visual acuity, perhaps optical coherence tomography (below), a neurologic exam, MRI, and possibly lumbar puncture. Perimetry shows a scotoma that is typically central or diffuse but sometimes is peripheral. Low-contrast Sloan letter charts are more sensitive than the standard Snellen measure of visual acuity. A loss of one line of low-contrast acuity correlates with a 3 mm2 increase in MRI T2 lesion volume; one line of high-contrast loss correlates with a 6 mm2 increase in MRI T2 lesions (Wu et al 2007). Loss of acuity can reflect post-geniculate white matter damage.
Visual evoked potentials are often abnormal in the affected eye, but return to normal within 2 years in one third of eyes. Low contrast stimuli and multifocal visual evoked potentials are more sensitive than conventional tests. Visual evoked potentials are more sensitive than optical coherence tomography (OCT) (Naismith et al 2009). In patients with multiple sclerosis and normal visual acuity, but no history of optic neuritis, subclinical optic tract lesions are detected with visual evoked potentials (82%), contrast sensitivity (73%), OCT (60%), pupillary light reflex (52%), flight of colors (36%), and color vision (Ishihara plates) (32%) (van Diemen et al 1992; Naismith et al 2009).
Retinal tomographs (optical coherence tomography, OCT) reproducibly measure retinal nerve fiber layer thickness (RNFL), retinal ganglion cells, and macular volume. One third of macular volume is from neurons, and macular edema correlates with visual function. OCT can define acute optic neuritis, demonstrate a second lesion in clinically isolated syndromes, and monitor atrophy and progression as a surrogate marker to complement the neurologic exam. OCT usually shows RNFL thinning 3 to 6 months after optic neuritis. OCT is abnormal in half of the “normal” fellow eyes in multiple sclerosis patients after an episode of optic neuritis. Thinning of RNFL is faster in optic neuritis eye (69 um) than in “normal fellow” eye (partially affected, at 95 um) and healthy control eyes (103 um) (Trip et al 2006). RNFL loss occurs over time in some patients, even without symptoms of optic neuritis. Optic nerve cross-sectional area on MRI correlates with RNFL thickness (r = 0.66). OCT measures correlate with visual evoked potential amplitude, but not with latency. Atrophy is associated with motor disability (r = 0.2-0.4) and cognitive problems (r = 0.5) (Toledo et al 2008). RNFL loss is greater in progressive multiple sclerosis than in relapsing-remitting multiple sclerosis; loss is even worse in neuromyelitis optica, especially in the superior and inferior quadrants. With current technology, OCT is not a replacement for evoked potential and MRI, especially in clinically isolated syndromes, because it is less sensitive.
MRI studies show multiple cerebral white matter lesions in one fourth to three fourths of optic neuritis patients, depending on the series; most of the MRI lesions involve the visual radiations.
The spinal fluid in optic neuritis contains elevated protein, mild lymphocytosis, elevated IgG index, and oligoclonal bands (50% to 70% vs. 95% in multiple sclerosis).
Quality of life scales in diagnosis. Quality of life scales are inexpensive, simple to administer, and measure a wide range of the problems seen in multiple sclerosis (Cella et al 1996; Cella et al 2012). They predict changes in disability and are objective measures of therapeutic outcomes. However, present scales are insensitive and have not contributed to trial monitoring.
Confusion in diagnosis and misdiagnosis. In some patients with clinically definite multiple sclerosis but negative MRI, other techniques such as evoked potentials or magnetic transfer imaging will show damage. When CSF has no oligoclonal bands, as in approximately 3% of patients, prognosis is better and brain MRI lesions are fewer. Four years after a diagnosis of multiple sclerosis, only half of these oligoclonal band-negative patients become positive (Zeman et al 1996). Similarly, one third of patients with one oligoclonal band on follow-up will develop multiple oligoclonal bands (Davies et al 2003).
Incidental, unexpected multiple sclerosis-like lesions on an MRI scan, without symptoms or signs of multiple sclerosis, are often referred to neurologists (“radiologically isolated syndrome”). Autopsy studies mentioned above show a significant reservoir of undetected multiple sclerosis. However, multiple sclerosis-like MRI lesions could be from vascular disease, tumor, and possibly migraine headache. Yet, in 30 patients with incidental brain MRI lesions, 23 (77%) developed new MRI lesions by 6 months, and 11 (37%) had clinical conversion to multiple sclerosis (Lebrun et al 2008). In 41 patients (7 treated) followed for 2.7 years, 59% had new MRI lesions; 30% had clinical attacks at a median time of 5.4 years (Okuda et al 2009). A quarter of patients have cognitive impairment and 40% have frontotemporal cortical MRI lesions. Asymptomatic cord lesions, present in one third of patients with these brain lesions, predict a future attack or progression in 80%. Thus, early treatment is reasonable for some cases with MRI lesions only, but commitment to a therapy after a misdiagnosis is a danger.
MRI lesions can lead to MRI-supported misdiagnosis of multiple sclerosis. Patients may be treated with expensive and dangerous drugs, “therapeutic mislabeling.” Reluctance to reverse the misdiagnosis by neurologists raises ethical issues because continued therapy does not provide full disclosure of a patient’s condition. Secondly, in patients with definite multiple sclerosis, MRI-based decisions to change treatment to another partially effective therapy should be made with caution (eg, one new lesion after years of clinically stable disease). Premature termination of a quite effective therapy in this highly variable disease can potentially harm patients. Imaging is expensive and can lead to decisions that are not evidence-based. Personalized medicine versus evidence-based decisions are discussed in (Caplan 2011).
A diagnosis of multiple sclerosis should be judiciously questioned when there are “red flags” such as:
In the absence of objective evidence for multiple sclerosis or other disease, follow-up investigations should be kept to a minimum.