Multiple sclerosis is an immune-mediated disease of the central nervous system in which the immune system attacks myelin — the fatty insulation wrapped around nerve fibres by oligodendrocytes, which lets electrical signals travel fast and reliably [1]. Strip the insulation and conduction slows, fails, or misfires. Because the lesions can occur anywhere in the brain, spinal cord or optic nerves, the symptoms can be almost anything: visual loss in one eye, numbness, weakness, imbalance, bladder dysfunction, fatigue, cognitive slowing [2].
Two things make MS distinctive among the diseases in this collection.
The first is that its cause now appears to be largely identified, which almost nothing else in neurology can claim. A cohort of more than ten million young adults on active duty in the US military, 955 of whom developed MS, found that risk of MS increased 32-fold after infection with Epstein-Barr virus — and not after infection with other viruses including the similarly transmitted cytomegalovirus. Serum neurofilament light, a marker of neuroaxonal injury, rose only after EBV seroconversion. The authors' conclusion is unusually direct: these findings "cannot be explained by any known risk factor for MS and suggest EBV as the leading cause of MS" [3]. This did not come from nowhere — prospective serological studies had pointed the same way for two decades [4] [5], as had meta-analysis of infectious mononucleosis as a risk factor [6].
The second is that its treatment has been genuinely transformed. There are now more than a dozen disease-modifying therapies of increasing potency, and the central strategic question has shifted from whether to treat to how hard, and how soon. The centerpiece is about that question.
Start here: what the disease does
The relapsing phase. Most people present with relapsing-remitting MS: discrete attacks lasting days to weeks, followed by partial or complete recovery. Each relapse is a focal inflammatory lesion — immune cells crossing the blood-brain barrier and demyelinating a patch of white matter.
The progressive phase. Over time many people transition to secondary progressive MS, in which disability accrues steadily without discrete attacks. A minority — roughly one in ten — have primary progressive MS from the start, with steady decline and few or no relapses [7]. The two progressive forms differ in how they begin but converge in their dynamics [8] [9].
The distinction matters enormously for treatment, because the drugs work well against relapses and poorly against progression. Relapses are driven by peripheral immune cells entering the CNS, which is exactly what the drugs block. Progression appears to be driven by something more compartmentalised — meningeal collections of B cells organised into follicle-like structures have been found in secondary progressive disease and associated with more severe cortical pathology [10] [11] — and this is much harder to reach from the bloodstream.
Immunology. MS is a T-cell and B-cell disease. Large-scale genetic analysis established a primary role for cell-mediated immune mechanisms [12], with immune-related loci dominating the risk architecture [13]. Regulatory T-cell function is impaired in patients [14], and IL-17-producing cells are implicated in both the animal model and human disease [15] [16]. The animal model, experimental autoimmune encephalomyelitis, has driven much of this and has produced at least one drug directly [17] [18].
Pillar 1: measurement and diagnosis
Dissemination in space and time
The diagnostic principle is old and still governs everything: MS is demonstrated by lesions disseminated in space (in more than one part of the CNS) and disseminated in time (occurring on more than one occasion). Formalised for research protocols in 1983 [19], the criteria were reworked around MRI as the McDonald criteria in 2001 [20] and revised repeatedly since — 2005, 2010 and 2017 [21] [22] [23], with MAGNIMS supplying the MRI specifics [24].
The direction of every revision has been to allow earlier diagnosis with fewer clinical events, because effective treatment exists and waiting for a second attack costs something. The 2017 revision in particular permits CSF findings to substitute for dissemination in time [23].
MRI
MRI is the central tool. T2-weighted sequences show the accumulated lesion burden — the white-matter plaques, characteristically periventricular, juxtacortical, infratentorial and in the spinal cord — the four locations the MRI criteria count for dissemination in space [24]. Gadolinium-enhancing lesions indicate an actively inflamed lesion with a breached blood-brain barrier, and enhancement typically lasts weeks. Seeing an enhancing lesion and a non-enhancing one on a single scan demonstrates dissemination in time from one study.
MRI is also the sensitive endpoint in trials, where lesion counts move far more than clinical events do — in one head-to-head trial the mean number of new lesions per scan was 0.02 with ocrelizumab against 0.29 and 0.42 with interferon beta-1a in its two arms [25].
Cerebrospinal fluid and other tests
Oligoclonal bands — immunoglobulin bands present in CSF but not in serum — indicate intrathecal antibody production and are found in the large majority of people with MS. They are supportive rather than specific, and under the 2017 criteria can substitute for dissemination in time [23].
Neurofilament light is a marker of neuroaxonal damage measurable in CSF and now in blood [26], and its behaviour in the EBV cohort — rising only after seroconversion — was part of what made that study persuasive [3].
Disability is scored on the Expanded Disability Status Scale, which is the outcome measure in essentially every trial cited here, and which is heavily weighted toward walking ability at its upper range — a real limitation when much of MS disability is cognitive and fatigue-related.
Centerpiece: a simple simulatable model of the cost of waiting
For years the standard approach was escalation: start with a safer, moderately effective drug, and move to something stronger only if the patient kept relapsing. The alternative — start with a highly effective drug immediately — has steadily gained ground. The arithmetic behind that shift is simple enough to state exactly.
Two placebo-controlled phase 3 trials report annualized relapse rates directly. In FREEDOMS, the annualized relapse rate was 0.18 with fingolimod against 0.40 with placebo [27]. In DEFINE, it was 0.17 with dimethyl fumarate against 0.36 with placebo [28]. Recomputing each trial's relative reduction from its own two rates gives 55.0 percent and 52.8 percent — two independent trials, different drugs, different placebo arms, agreeing within 2.2 percentage points on a quantity neither was asked to produce jointly.
Higher-efficacy agents do better. AFFIRM reported that natalizumab reduced the rate of clinical relapse at one year by 68 percent, alongside a 42 percent reduction in the risk of sustained disability progression and 92 percent fewer gadolinium-enhancing lesions [29] [30].
Now model a patient who spends d years on ineffective therapy before switching. With relapse rates constant within each state, cumulative relapses by time T are the untreated rate times the delay, plus the treated rate times the remainder.
The structural result is the point, and it is not about the size of any number. After the switch, the delayed and immediate curves run exactly parallel. The relapses accumulated during the delay are not caught up later — they persist as a fixed offset for the rest of the patient's life. A five-year delay costs about 1.1 relapses permanently on moderate-efficacy therapy; delay does not slow recovery, it creates a deficit that is never made up.
The right panel adds the counterintuitive corollary. The permanent excess is the delay multiplied by the difference in relapse rates — so the more effective the therapy being withheld, the more each year of waiting costs. A year of delay costs 0.22 relapses if the alternative is moderate-efficacy therapy and 0.27 if it is high-efficacy therapy. The better the drugs get, the worse escalation looks.
That is the teaching point: treating hard and early is not about impatience, it is about the fact that inflammatory damage is cumulative and irreversible.
Three honest limits, and the third is the most important. A constant relapse rate is an approximation — relapse frequency declines with disease duration and age — so the ten-year projections illustrate the shape of the penalty rather than predicting counts for an individual. The high-efficacy rate is derived by applying one trial's relative reduction to another trial's placebo arm, which assumes comparable populations. And relapses are an imperfect proxy for what patients actually care about: much disability accrues independently of relapses, which is precisely the problem the progressive forms present and which no relapse-based model can capture.
Pillar 2: treatment
Acute relapses
High-dose corticosteroids shorten a relapse. They do not change its eventual outcome or the disease's course, and they are not maintenance therapy.
The disease-modifying ladder
Moderate efficacy. Interferon beta and glatiramer acetate were the first agents to work, and glatiramer's original trial showed reduced relapse rate and improved disability [31]. Interferon beta given after a first demyelinating event delays conversion to clinically definite MS [32]. Oral dimethyl fumarate joined this tier with the relapse-rate reduction quoted above [28]. These drugs are safe, well characterised, and roughly halve relapses.
Higher efficacy splits by mechanism.
Blocking entry to the CNS. Natalizumab is a monoclonal antibody against α4-integrin that prevents lymphocytes crossing the blood-brain barrier — a mechanism first demonstrated in the animal model [18] — and it is among the most effective agents available [29].
Sequestering lymphocytes in lymph nodes. Fingolimod, a sphingosine-1-phosphate receptor modulator, traps lymphocytes in lymph nodes so they cannot reach the CNS [33]. It beat placebo [27] and beat intramuscular interferon head-to-head [34].
Depleting B cells. That B-cell depletion works at all was a surprise, since MS was long framed as a T-cell disease — but rituximab reduced inflammatory lesions and relapses in relapsing-remitting MS [35], and ocrelizumab, which selectively depletes CD20+ B cells, outperformed interferon beta-1a in two identical phase 3 trials [25]. Anti-CD20 therapy is now a mainstay, and its success is a substantial piece of evidence about the underlying immunology.
Risk, and why it is manageable
Natalizumab carries a serious hazard: progressive multifocal leukoencephalopathy, a brain infection caused by JC virus reactivation, first reported in treated patients [36] [37]. Quantification made it usable rather than prohibitive. Among 99,571 natalizumab-treated patients there were 212 confirmed cases — 2.1 per 1,000 — and every one of the 54 patients with pre-diagnosis samples was anti-JC virus antibody positive. In antibody-negative patients the estimated incidence was 0.09 cases or fewer per 1,000 [38]. Risk stratified by antibody status, prior immunosuppression and treatment duration turned an unacceptable risk into a manageable one for most patients — a good example of a biomarker rescuing a drug.
Progressive disease, which remains the hard problem
Every agent above works chiefly on relapses. In primary progressive MS, ocrelizumab was the first therapy to show benefit against placebo in a phase 3 trial — with T2 lesion volume falling 3.4 percent against a 7.4 percent rise on placebo and brain-volume loss of 0.90 against 1.09 percent — but with no significant difference in the physical component of quality of life, and more neoplasms (2.3 versus 0.8 percent) [39]. The benefit is real and modest, and that gap between the relapsing and progressive forms is the central unmet need in MS.
Everything else
Symptomatic management — spasticity, bladder dysfunction, fatigue, pain, mobility — and rehabilitation are a large part of care and are not optional extras. Depression and anxiety are substantially more common in MS than in the general population [40].
Pillar 3: what is unresolved
What to do with the EBV finding. If EBV infection is effectively necessary for MS [3], the implications range from an EBV vaccine as primary prevention to EBV-directed therapy in established disease. Neither exists yet, and the mechanism connecting a near-universal infection to a relatively rare disease is not established — most people infected with EBV never develop MS, so EBV is necessary but nowhere near sufficient.
Progression independent of relapses. Disability accrues in many patients without attacks, apparently driven by compartmentalised CNS inflammation and neurodegeneration that current drugs do not reach [10] [11]. Agents that penetrate the CNS are the obvious direction.
Remyelination. No approved therapy repairs myelin. That repair is biologically possible has been shown by transplanted oligodendrocyte progenitors remyelinating in animal models [41]; translating that into a drug is unsolved.
Biomarkers for individual decisions. Neurofilament light is the most promising monitoring marker [26], and the natalizumab experience shows what a good stratifying biomarker is worth [38]. What is missing is anything predicting which patient needs high-efficacy therapy up front.
When to stop. With effective long-term suppression and an ageing treated population, de-escalation and discontinuation are increasingly pressing and largely unstudied.
Dig deeper in lmmol
MS belongs with the other immune-mediated diseases in this collection, and the comparison is unusually pointed. Rheumatoid arthritis and inflammatory bowel disease share the treat-early logic, the B-cell-depletion and integrin-blocking mechanisms, and the same problem of choosing among them — natalizumab was developed for both MS and Crohn's disease, and the first PML cases appeared in both [37] [38]. The instructive divergence is TNF: inhibiting it transformed RA and IBD, but anti-TNF therapy has been reported to cause demyelination [42], which is a reminder that immune-mediated diseases are not interchangeable. Among the neurological diseases, Parkinson's disease and Alzheimer's disease show the contrast between degeneration that is primarily neuronal and damage that is primarily immune — though MS's progressive phase blurs that line. Depression is markedly more common here than in the general population [40] — see depression. For another disease where a common virus is necessary but not sufficient for a much rarer outcome, see HIV as the contrasting case of a virus that is both necessary and sufficient. The full collection is at health.