Parkinson's disease is the second most common neurodegenerative disorder, and its global burden has been rising faster than that of any other neurological condition [1]. It is caused by the progressive loss of dopamine-producing neurons in a small midbrain structure, the substantia nigra, accompanied by aggregates of a protein called alpha-synuclein that pathologists have recognised as Lewy bodies for more than a century [2].
The disease people recognise — the tremor, the slowness, the stiffness — is the late part. By the time the first symptom appears, roughly half the relevant neurons are already gone, and the loss has been under way for about five years. That gap between when the disease starts and when it becomes visible is the most important structural fact about Parkinson's, and it is the subject of the centerpiece.
Start here: what Parkinson's disease is
The clinical core is motor parkinsonism, defined as bradykinesia — slowness and progressive decrement of movement — plus either rest tremor or rigidity [3]. The tremor is the famous feature but it is not the necessary one; some people never have it, and bradykinesia is what must be present.
Underneath, two processes run together. Dopaminergic neurons of the substantia nigra pars compacta die, removing the dopamine supply to the putamen and caudate that normally allows movement to be initiated and scaled. And alpha-synuclein misfolds and aggregates, forming the Lewy bodies and Lewy neurites that define the pathology [4]. Braak and colleagues proposed that this pathology spreads in a stereotyped sequence through the brain [5], a scheme that has been influential and also seriously criticised [6].
The disease is not only, or even mainly, a movement disorder. Non-motor features are pervasive and often precede the motor ones: constipation, loss of smell, disordered sleep, depression and anxiety, autonomic dysfunction, and later cognitive impairment and dementia [7] [8] [9] [10]. It shares a spectrum with dementia with Lewy bodies, which has the same molecular pathology in a different distribution [11].
Cause is mostly unknown. Ageing is the dominant risk factor and the reasons for that are not fully understood [12]; genome-wide association work has identified many risk loci and a substantial heritable component without producing a single dominant explanation [13]; environmental exposures contribute [14]; and the whole field owes a peculiar debt to a contaminated synthetic opioid that produced sudden parkinsonism in young people in the 1980s and gave researchers their first good animal model [15]. Broader reviews cover the epidemiology and pathogenesis [16] [17] [18].
One idea worth flagging because it keeps gaining ground: that the pathology may begin outside the brain. The dual-hit hypothesis proposed entry through the olfactory bulb and the gut [19], pathological alpha-synuclein has been shown to propagate from gut to brain in an animal model [20], and the gut microbiota can modulate motor deficits and neuroinflammation in a model of the disease [21]. This is not established in humans, but it would explain why constipation and hyposmia arrive first.
Pillar 1: measurement and diagnosis
The diagnosis is still clinical
There is no blood test that makes this diagnosis. The Movement Disorder Society criteria formalise expert clinical judgement: establish motor parkinsonism, then apply absolute exclusion criteria that rule the disease out, red flags that must be counterbalanced, and supportive criteria that argue for it — of which a clear and dramatic response to dopaminergic therapy is among the strongest [3]. Earlier criteria set the same shape [22], and neuropathological confirmation remains the benchmark against which they are judged [23].
This matters practically. A diagnosis made this way is a judgement about a pattern, it can be wrong, and it is refined over time by watching the response to treatment and the emergence of features that point elsewhere [24].
Rating severity
The MDS-UPDRS is the standard instrument: a structured scale covering non-motor and motor experiences of daily living, a motor examination, and motor complications [25]. It is what trials use as an endpoint and what makes different studies comparable. Wearable sensors are being evaluated as a more granular alternative [26].
Imaging the dopamine terminals
Dopamine transporter imaging — DaTscan — visualises the density of presynaptic dopaminergic terminals in the striatum, and it is reduced in Parkinson's disease. It has been used as a progression endpoint in trials comparing initial therapies [27]. What it does not do is distinguish Parkinson's disease from other causes of nigrostriatal degeneration; it answers "is there dopaminergic denervation?", not "is this Parkinson's disease?"
The molecular biomarker that finally arrived
For most of this disease's history there was no way to detect the causative protein in a living patient. Seed amplification assays changed that. The principle is to exploit the aggregates' own defining property: misfolded alpha-synuclein templates the misfolding of normal protein, so a tiny amount of seed in a patient sample can be amplified into a detectable signal.
A proof-of-concept study using protein misfolding cyclic amplification on cerebrospinal fluid detected as little as 0.1 picograms per millilitre of alpha-synuclein oligomers and, in blinded cohorts, identified patients with Parkinson's disease with sensitivity 88.5 percent and specificity 96.9 percent, with kinetic parameters correlating with clinical severity [28]. Related ultra-sensitive approaches quantify seeds in brain and cerebrospinal fluid [29], and large cohort work has used the assay to show substantial molecular heterogeneity among people carrying a clinical Parkinson's diagnosis [30].
This is the most consequential change in Parkinson's diagnostics in decades, because it opens the possibility of defining the disease biologically rather than by its motor signs — which, as the centerpiece shows, appear only after most of the damage is done.
Prodromal markers
Three features, individually unremarkable, together identify people at substantially elevated risk years before diagnosis, and the MDS has published research criteria that combine them into a probability [31].
REM sleep behaviour disorder — acting out dreams because the normal paralysis of REM sleep fails — is the strongest single marker; its pathophysiology involves precisely the brainstem structures affected early in synucleinopathy [32] [33]. Hyposmia is common, measurable, and has been shown prospectively to predict future Parkinson's disease [34] [35] [36] [37]. Constipation completes the trio.
None of these is remotely specific on its own. Their value is in combination, and in a future where there is something to offer people identified early.
Centerpiece: a simple simulatable model of the presymptomatic window
Why does a disease that has been destroying neurons for years announce itself all at once?
The answer is that the nigrostriatal system has reserve. Surviving dopaminergic terminals increase their output, and the striatum compensates, until the losses outrun what compensation can hide. The model is arithmetic, and both of its parameters are measured.
An [18F]dopa PET study scanned 32 patients twice, a mean of 18 months apart, and measured two things in the same population: the rate at which putamen dopaminergic function declines — 4.7 percent of normal per year, with putamen the most sensitive measure — and the level at which motor symptoms begin, 75 percent of normal putamen uptake (the caudate, less affected, was at 91 percent) [38].
Taking the decline as linear, the presymptomatic period follows by division rather than assumption: (100 − 75) / 4.7 = 5.3 years.
That figure is checked twice against numbers the arithmetic was never given. The same PET study concluded independently that the preclinical period is "unlikely to be longer than seven years" [38] — and 5.3 satisfies it. More strikingly, a completely different method reached almost the same answer: counting pigmented neurons in post-mortem substantia nigra across 20 Parkinson's cases and 36 controls, Fearnley and Lees estimated the presymptomatic phase at about five years [39]. In-vivo imaging and counting cells in dead brains, landing within a year of each other.
The compensation gap is the mechanism, and it is computable from those two papers together. At the moment symptoms begin, the pathology series finds 48 percent of caudal nigral neurons already lost — so 52 percent remain — while the imaging series finds putamen dopaminergic function still at 75 percent of normal [39] [38]. Roughly half the cells are dead and three quarters of the function is intact. The surviving terminals are working harder, and that 23-point gap is the silent years, quantified.
Two consequences follow, and they are the reason this section exists.
First, anyone diagnosed with Parkinson's disease has had it for years. The diagnosis is not the beginning of the disease; it is the moment compensation failed.
Second, any therapy that protects neurons is being tested far too late. A drug given at diagnosis is being asked to help a system that has already lost half its cells. This is the strongest argument for the biomarker work above: a molecular assay that detects misfolded alpha-synuclein before motor onset [28] [30], combined with prodromal criteria [31], is what would make it possible to test neuroprotection during the window when there is something left to protect.
Three honest limits. A constant linear rate of decline is an approximation — the post-mortem series found neuron loss to be exponential and faster early [39] — so the straight line is a decade-scale average, not a claim about any given year. The right panel shows how sensitive the answer is: the same arithmetic with a 3 percent annual decline gives more than eight years, and with 7 percent gives under four. And the two thresholds being compared are different quantities measured in different cohorts by different methods; that they agree is meaningful, but it is agreement between estimates, not a single measurement.
Pillar 2: treatment
The honest framing for this section: the symptomatic treatment of Parkinson's disease is genuinely excellent, and none of it is known to slow the disease.
Levodopa
Levodopa remains the most effective drug, more than fifty years after its introduction. It is a dopamine precursor that crosses into the brain and is converted to dopamine; it is given with carbidopa, a peripheral decarboxylase inhibitor that stops that conversion happening outside the brain, which both increases delivery and reduces nausea.
A long-standing anxiety that levodopa might accelerate the disease was tested directly. The ELLDOPA trial found that levodopa-treated patients had better motor scores at the end of the study even after washout, though the imaging substudy pointed the other way — a discordance the authors did not resolve [40]. The practical conclusion the field drew is that withholding levodopa from someone who needs it is not justified.
The other classes, and what each is for
Dopamine agonists act directly on dopamine receptors. Starting with an agonist rather than levodopa produces fewer motor complications but more somnolence, oedema and hallucinations [41] [27] — and, importantly, impulse control disorders: pathological gambling, hypersexuality, compulsive shopping and eating, which are strongly associated with dopamine agonist use and are frequently missed unless specifically asked about [42].
MAO-B inhibitors block the enzyme that degrades dopamine in the brain, giving a modest symptomatic benefit and a useful adjunct effect. COMT inhibitors block peripheral levodopa breakdown, extending each dose. The evidence for all classes has been assembled in evidence-based medicine reviews [43] [44] [45].
Motor complications, which are the real long-term problem
After some years on levodopa, the smooth response gives way to motor fluctuations — "wearing off" before the next dose — and dyskinesia, involuntary writhing movements at peak effect. The mechanism is thought to relate to pulsatile stimulation of denervated striatal receptors: as terminals are lost, the brain can no longer buffer the peaks and troughs of an oral drug taken three or four times a day.
The therapeutic logic that follows is continuous delivery. Subcutaneous apomorphine infusion reduced off-time significantly against placebo in a randomised trial in patients with persistent fluctuations [46], and intestinal levodopa gel infusion serves the same purpose. Some symptoms respond poorly to any of it — freezing of gait in particular is only partly levodopa-responsive [47].
Deep brain stimulation
For selected patients, electrodes implanted in the subthalamic nucleus or globus pallidus interna, connected to a pulse generator, substantially improve motor function and fluctuations. The evidence is unusually good for a surgical therapy: early reports of subthalamic stimulation in advanced disease [48] were followed by five-year follow-up [49], a randomised comparison against best medical therapy [50], a head-to-head trial of pallidal versus subthalamic targets [51], and EARLYSTIM, which showed benefit in patients with early motor complications rather than only end-stage disease [52] [53].
Deep brain stimulation treats the symptoms that dopamine treats. It does not help the features that are not dopaminergic — cognitive decline, autonomic failure, falls — and patient selection is most of the art.
Everything else
Physiotherapy, occupational therapy, speech therapy and exercise have real evidence behind them; cueing training improved gait-related mobility in a randomised trial [54]. Non-motor symptoms need their own management and have their own evidence base [55] [56]. Multidisciplinary care is not an optional extra in a disease with this many domains.
Pillar 3: what is unresolved
Disease modification, which is the whole game. No treatment has been shown to slow progression. The most direct strategy targets alpha-synuclein itself — immunotherapy blocked uptake and templated propagation of misfolded protein in animal models [57] [58] [59] — and the translation of that into human benefit has not yet succeeded. Trophic-factor approaches have also been tested: intraputamenal GDNF delivered by convection-enhanced infusion did not meet its primary endpoint in a randomised trial [60].
Biological staging. With a seed amplification assay that detects the causative protein [28] [30], it becomes possible to define Parkinson's disease by its biology rather than its motor signs — the same shift that reorganised Alzheimer's research. That is a change in what the word "Parkinson's" refers to, and it is happening now.
Repurposed agents. The most discussed is the GLP-1 receptor agonist exenatide, supported by preclinical neuroprotection in dopaminergic models [61] [62] [63] and by a randomised, double-blind, placebo-controlled trial reporting a motor advantage that persisted after washout [64] [65], with an insulin-resistance rationale behind it [66]. Whether this reflects disease modification or a durable symptomatic effect is exactly the question a positive early trial cannot settle.
Gene and cell therapy. AAV-delivered GAD, intended to normalise subthalamic output, progressed from an open-label safety study [67] to a sham-surgery-controlled randomised trial with modest benefit [68]. Cell replacement has a long and difficult history, and patient-derived induced pluripotent stem cells have made autologous approaches conceivable [69].
And the mechanism itself. Oxidative stress [70], mitochondrial dysfunction, impaired protein clearance, alpha-synuclein secretion and cell-to-cell spread [71] and neuroinflammation are all implicated, and which is cause and which is consequence is not settled [72].
Dig deeper in lmmol
Parkinson's completes lmmol's set of three protein-aggregation neurodegenerative diseases, and the comparison is the most instructive thing about it. Alzheimer's disease is the one furthest along the path Parkinson's is now taking — from a clinical syndrome to a biologically defined disease with fluid biomarkers — and the seed amplification assays described above are Parkinson's version of that shift [28]. Huntington's disease is the opposite pole: a single fully penetrant genetic cause with a presymptomatic period you can measure from a blood test at birth, where Parkinson's has many risk loci and no such certainty [13] — and the two diseases were quantitatively linked in the founding neurochemical study of brain dopamine in both [73]. Obstructive sleep apnea is worth reading alongside the prodromal section, since both are diseases where a sleep symptom is the visible tip of something systemic [32]. Depression is both a prodromal feature and a common comorbidity here [56], and stroke provides the contrast of sudden versus slow neuronal loss. The full collection is at health.