Glaucoma is a group of diseases that damage the optic nerve — the cable carrying vision from eye to brain. Retinal ganglion cells die, their axons are lost, and the visual field erodes from the periphery inwards. It is usually related to the pressure of fluid inside the eye, and it is the leading cause of irreversible blindness worldwide [1] [2] [3] [4].
That word does the work. Cataract, the largest single cause of blindness globally, is reversible — the lens comes out and sight returns. Vision lost to glaucoma does not come back. Treatment can stop the loss continuing; it cannot restore what has gone.
And the disease is silent while it is doing the damage. Peripheral vision is not something people monitor. The brain fills in, the other eye compensates, and the field narrows for years without being noticed. The Baltimore Eye Survey examined 5,308 people aged 40 and over and found that roughly half of all subjects with optic nerve damage from primary open-angle glaucoma, regardless of race, were unaware that they had the condition [5] [6]. Half. Found on examination, in people who had no idea.
Put those two facts together — irreversible, and silent — and you get the shape of the problem. By the time someone notices, the damage that produced the noticing is permanent. Everything in glaucoma care is organised around detecting it before that point and then holding the line for the rest of the patient's life.
Start here: pressure, and the trouble with it
Aqueous humour is made continuously inside the eye and drains out through the trabecular meshwork. If outflow is impaired, pressure rises. Raised intraocular pressure damages the optic nerve head, which behaves as a biomechanical structure — the load is borne by the lamina cribrosa, the sieve-like plate the axons pass through, and thinking of it that way reframed how the damage is understood [7] [8]. The cells that die are retinal ganglion cells, and they die by apoptosis in experimental glaucoma and after axotomy [9] [10] [11], with axonal insult occurring early in the optic nerve [12].
But pressure is a risk factor, not the definition — and the same Baltimore survey contains both halves of that statement.
On one hand, the risk of glaucomatous optic nerve damage rose steeply with screening pressure: against eyes at 15 mm Hg or below, the relative rate was 12.8 at 22–29 mm Hg and 40.1 at 30 mm Hg and above [5]. A forty-fold gradient is not a subtle association.
On the other hand, in the same 5,308 people, more than half of all glaucomatous eyes had a screening IOP below 21 mm Hg [5] — below the number conventionally called normal. That is normal-tension glaucoma, and it is not a rare curiosity; it is the majority of cases in that survey. Its progression and risk factors have been studied specifically [13] [14], with candidate contributors including vascular factors [15] [16].
So: pressure is the strongest modifiable risk factor and it is not the disease. A pressure below 21 does not exclude glaucoma, and a pressure above it does not establish glaucoma. Other risk factors matter — age, African ancestry [6], myopia [17], and a substantial genetic contribution: a multi-ethnic meta-analysis of 34,179 cases and 349,321 controls identified 127 open-angle glaucoma loci, 44 of them new, with effects broadly consistent across European, Asian and African ancestries [18] [19], on top of the single-gene forms caused by mutations in myocilin and optineurin [20].
Pillar 1: measurement and diagnosis
Glaucoma is diagnosed by putting three measurements together over time. No one of them is sufficient.
Pressure
Tonometry measures intraocular pressure. Goldmann applanation, described in 1957, remains the reference method [21], and the several available tonometers rest on different assumptions and do not always agree [22] [23]. Corneal properties confound it — a thick or biomechanically stiff cornea makes the measured pressure read higher than the true one, which is why corneal thickness is measured alongside and why in-vivo corneal biomechanics became a research topic in its own right [24].
Structure
The optic disc is examined for the cupping that reflects lost axons, and the retinal nerve fibre layer is imaged, now routinely with optical coherence tomography.
Structure changes before function does, and by years. In 1,344 eyes with elevated pressure followed with annual perimetry and nerve fibre layer photography, defects in the nerve fibre layer were identified in 88 percent of readable photographs at the moment field loss first appeared — and 60 percent of eyes already had nerve fibre layer defects six years before the field loss [25]. The same study gives the specificity context: the nerve fibre layer was read as abnormal in only 11 percent of normal eyes. Imaging the structure buys years of warning over waiting for the patient to lose field.
Function
Automated perimetry maps the visual field and quantifies the defect [26], and the trials define progression on it [27] [28]. Perimetry is the measurement that corresponds to what the patient will eventually experience, which is why it remains the reference for progression even though it lags structure.
Two mechanisms, two clinical stories
Open-angle glaucoma is the common form: the drainage angle is anatomically open, outflow resistance rises gradually, and the disease is chronic and painless [29] [30].
Angle-closure glaucoma is different. The iris blocks the drainage angle, and when it closes acutely the pressure rises fast — a painful red eye with blurred vision, halos, headache and vomiting, and a genuine emergency, because pressure of that magnitude damages the eye within hours and can damage the corneal endothelium as well [31]. Laser peripheral iridotomy is the standard intervention across the angle-closure spectrum [32], and because the lens contributes to the crowding, early lens extraction has been tested as a treatment for primary angle-closure glaucoma [33] [34]. Angle closure has its own genetic risk loci [35] and its own long-term outcome data after an acute attack [36].
Centerpiece: what a millimetre of mercury is worth
Lowering intraocular pressure is the only intervention shown to change the course of this disease, and the Early Manifest Glaucoma Trial priced it.
The trial randomised 255 patients aged 50–80 with newly detected early open-angle glaucoma — median visual-field mean deviation −4 dB, median IOP 20 mm Hg, mostly found by population screening — to immediate laser trabeculoplasty plus betaxolol, or to no initial treatment, and followed them for a median of six years. Treatment reduced pressure by 5.1 mm Hg, or 25 percent, maintained throughout. Progression occurred in 58 of 129 treated patients (45 percent) against 78 of 126 controls (62 percent), P = 0.007, and occurred significantly later in those treated [37].
The check the arithmetic was never given. The trial states its treatment effect twice in a single sentence — "5.1 mm Hg or 25%" — and reports the cohort's median pressure separately, as 20 mm Hg. Those two statements of the same effect imply a baseline: 5.1 divided by 0.25 is 20.4 mm Hg. The paper never performs that division; doing it recovers its own cohort's pressure to within 0.4 mm Hg, or 2 percent. Two independently reported quantities, internally consistent.
What the pressure bought. The absolute risk reduction is 16.9 percentage points, a number needed to treat of 5.9 over six years — treat six people with early glaucoma and prevent one from progressing. Spread across the 5.1 mm Hg achieved, that is 3.3 percentage points of absolute risk per millimetre of mercury. On the relative scale, interpolating between the trial's two points, each millimetre multiplies progression risk by 0.939 — a 6.1 percent relative reduction per mm Hg.
And an honest constraint on that number, which the figure states. This is a two-point interpolation on the risk scale, not a Cox model. With a control event rate of 62 percent, a risk ratio is strongly attenuated toward 1 relative to the corresponding hazard ratio, so a hazard-based per-millimetre figure is necessarily larger than this one. They are different quantities and should not be quoted interchangeably — which matters, because the per-mm Hg figure circulates widely without anyone saying which scale it is on.
The gradient is corroborated from a completely different design. The Baltimore Eye Survey's cross-sectional relative rates — 12.8 at 22–29 mm Hg and 40.1 above 30, against 15 mm Hg or below [5] — are a risk gradient in untreated prevalent disease, measured by prevalence survey rather than by randomisation. Different method, same direction, much larger range because it spans a wider pressure span. AGIS examined the relationship between achieved pressure control and visual field deterioration in treated eyes [38], and UKGTS was the first placebo-controlled trial to test whether a pressure-lowering drop preserves visual function, randomising newly diagnosed patients to latanoprost or placebo with time to visual field deterioration within 24 months as the primary outcome [39].
The teaching point. Pressure is the one lever we can pull. It is not the whole disease — more than half of glaucomatous eyes in Baltimore had pressures below 21 [5] — but it is the only variable that treatment can move and that moving demonstrably changes outcomes. That is why treatment targets pressure even when it already looks normal, and why the normal-tension trials asked exactly that question [40] [41].
Three honest limits. EMGT enrolled early glaucoma with a median pressure of 20 and excluded pressures above 30, so the per-millimetre figure derived here is anchored in that range and should not be extrapolated to advanced or high-pressure disease — the figure draws everything beyond 5.1 mm Hg as a dashed line for that reason. The trial's treated arm received laser plus a beta-blocker, so this prices pressure lowering, not any particular drug. And progression here is a composite of field and disc criteria over six years, not blindness.
Pillar 2: treatment
Every proven treatment does one thing: lower pressure. There is nothing else established, and that includes in normal-tension glaucoma, where the question of whether lowering an already-normal pressure helps was put directly and is the reason those patients are treated at all [40] [41].
Drops
Prostaglandin analogues are first-line — latanoprost, travoprost, bimatoprost — given once daily, and latanoprost is the drug UKGTS tested against placebo [39] [42]. Beta-blockers such as timolol, alpha-agonists such as brimonidine, carbonic anhydrase inhibitors such as dorzolamide, and the newer rho-kinase inhibitors are added or substituted.
Two problems with drops deserve stating plainly. The first is that they have real adverse effects, both local and systemic, and topical antiglaucoma medication has been reviewed specifically on that account [43] [44]; chronic exposure, particularly to preservatives, damages the ocular surface [45] [46].
The second is adherence, and it is the largest gap between what treatment can do and what it does. Persistence with topical glaucoma medication has been measured and is poor [47]. A silent disease treated with a drop that stings, in an eye that feels fine, for the rest of a person's life, to prevent something they cannot perceive — every element of that works against taking it.
Laser
Laser trabeculoplasty improves outflow through the trabecular meshwork; EMGT's treated arm received it as part of the initial intervention [37]. Selective laser trabeculoplasty is the modern form, its mechanisms have been reviewed [48], and repeated low-energy protocols are under investigation [49].
An honest gap: the LiGHT trial, which is the evidence behind offering selective laser trabeculoplasty as first-line therapy ahead of drops, is not in this substrate. The crawl returned mechanism reviews but not the trial, so this review states no efficacy or cost-effectiveness figures for SLT-first treatment.
Surgery
Trabeculectomy creates a guarded channel for aqueous to drain under the conjunctiva, forming a bleb. It remains the standard filtering operation, was compared against conventional management in early open-angle glaucoma [50], has been reviewed as practised in the modern era [51] [52], and its blebs are graded systematically because their appearance predicts function [53]. Tube shunts are the alternative when trabeculectomy is unsuitable or has failed.
Minimally invasive glaucoma surgery trades some pressure-lowering for a much better safety profile, and is usually done alongside cataract surgery. A systematic review found that adding an iStent to cataract surgery improved the likelihood of drop-free glaucoma control (relative risk 1.6, 95% CI 1.4–1.8 at medium-term and 1.6, 1.4–1.9 at long-term follow-up) and conferred 2.0 mm Hg greater pressure reduction (95% CI 1.3–2.7) than cataract surgery alone [54]. Two millimetres is modest — put it through the figure above and it is worth roughly 12 percent relative risk reduction — but it comes with drop-free control, which given the adherence problem is worth more than the pressure number alone suggests.
What treatment cannot do
It cannot restore lost vision. Retinal ganglion cells that have died do not come back, and the field defect is permanent. This is worth being explicit about with patients, because "we're treating your glaucoma" is easily heard as "this will get better."
Pillar 3: what is unresolved
Neuroprotection. If ganglion cell death is the final common path, protecting those cells independently of pressure would be a second lever. The idea has been pursued for decades through ischaemia and excitotoxicity models [55] [56], with candidates including minocycline [57], cell-transplantation approaches [58] and, more recently, nicotinamide, on the basis that it protects against mitochondrial dysfunction [59]. The route to clinical trials and precision medicine has been mapped [60]. Nothing has yet reached practice.
Detecting it earlier and monitoring it better. Given that half of prevalent cases are unaware [5] and that structural damage precedes field loss by years [25], the case for better detection is strong. Deep learning applied to fundus images and OCT is the most active area — validated systems exist for diabetic retinopathy and related eye diseases [61] [62] [63], with public benchmarks for glaucoma specifically [64] and challenge datasets for optic-disc segmentation and glaucoma classification. Home monitoring of the visual field has been tested for uptake and performance [65].
Why some eyes progress at normal pressure. The vascular and biomechanical hypotheses remain live [15] [7], as do oxidative-stress and inflammatory contributions [66], and normal-tension progression has its own risk-factor literature [13].
Turning genetics into something useful. The 127 loci are a starting point, and the authors note that several compounds targeting POAG risk genes are potential therapeutic candidates [18]. Whether polygenic risk can identify who to screen is the more immediate question.
Dig deeper in lmmol
Glaucoma opens the vision domain for this collection, and its nearest neighbour is type 2 diabetes and glycemic control — diabetic retinopathy is the other great chronic eye disease, the two share the imaging and deep-learning infrastructure that now screens for both [61] [62], and diabetes is itself managed by holding a number that the patient cannot feel, exactly as glaucoma is. Hypertension is the closest structural analogue in the whole collection: a silent, symptomless pressure that damages an end organ over decades, where the treatment target is a number rather than a symptom and adherence is the binding constraint. Osteoarthritis offers the opposite lesson on imaging — there, the X-ray and the symptom come apart and imaging over-reads disease, whereas in glaucoma the imaging under-reads nothing and is years ahead of the symptom. And Alzheimer's disease and Parkinson's disease share the neurodegeneration problem: retinal ganglion cells, like neurons elsewhere, die by apoptosis and cannot be replaced [9], which is why neuroprotection has been pursued in all three and delivered in none. The full collection is at health.