Vision
Eye exams are no longer just about glasses — the retina is the one place a doctor can directly watch the brain and blood vessels age, and AI can now read those images as a non-invasive predictor of mortality. But the moves that actually matter are older and duller than the frontier headlines: a thorough eye exam and prompt treatment of any correctable vision loss, which joined hearing on the 2024 Lancet Commission's modifiable-dementia list, albeit as one of the smaller contributors.
Vision has been reclassified along a similar arc to hearing, though on thinner evidence. The eye is no longer treated as an isolated sensory organ; it is a systemic biomarker of microvascular aging, a mitochondrial demand-test of the highest-energy tissue in the body, and a measurable input to cognitive trajectory. The 2024 Lancet Commission added untreated visual impairment to the modifiable dementia list, but as a late-life factor and with a population attributable fraction of only about 2% — against 7% each for hearing loss and high LDL ("bad") cholesterol, both of which the Commission files in midlife. The gap is about how many people are exposed rather than how strong the effect is: untreated vision loss carries about 50% higher dementia risk (relative risk 1.5; its 95% confidence interval — the range of values the data leave plausible — runs 1.4 to 1.6), slightly more than hearing loss's 40%, while the Commission's exposure prevalences are 12.7% for vision loss and 59.0% for hearing loss.[1] A 2024 study that ran a cohort analysis, a meta-analysis and Mendelian randomisation (which uses inherited variants as a natural experiment) side by side found genetic support for the hearing–dementia link across all-cause dementia and Alzheimer's disease, while on the vision side the only genetic signal was for the residual category of dementias that are neither Alzheimer's nor vascular.[2] The past few years have also produced an oculomics literature in which deep-learning models read fundus photographs to predict cardiovascular events, biological age, and all-cause mortality — a genuine research advance, though not yet standard clinical care.
Why vision is on the longevity short list
Three independent signals converge:
- Dementia (modest, unresolved). Untreated visual impairment is among the modifiable risk factors in the 2024 Lancet Commission framework, but it is one of the smaller levers — a population attributable fraction of roughly 2%, well below hearing loss or high LDL cholesterol, and causality remains unresolved (shared neuropathology, sensory deprivation, and reverse causation are all plausible).[3] AMD itself is a poor marker of dementia risk. The Commission's own meta-analysis of three studies found no significant association between age-related macular degeneration (AMD) and dementia (relative risk 1.15, 95% confidence interval 0.88–1.50 — a range wide enough to include no effect at all), and the largest published test since, a 2025 retrospective cohort across three large electronic-health-record networks, concluded that AMD "was not consistently positively associated" with Alzheimer's or vascular dementia.[4] What carries the signal is functional vision loss, not the diagnosis behind it: in a retrospective analysis of US electronic health records, blindness raised five-year Alzheimer's risk by about 29% (hazard ratio 1.29, 1.17–1.41), while non-exudative AMD in people who were not blind did not (0.95, 0.87–1.04). The authors put it no more strongly than that the link "may be partially mediated by visual impairment", and their finding that exudative AMD looked mildly protective (0.84, 0.72–0.97) is a reminder of how much residual confounding sits in a records analysis of this kind.[5] The practical implication is identical to hearing: treat the deficit early. (Evidence: moderate for the association; causality unresolved.)
- All-cause mortality. Deep-learning analysis of routine fundus photographs now generates a "Retinal Age Gap" (RAG) — the difference between predicted retinal age and chronological age — that is a genuine, if small, population-level mortality signal. UK Biobank cohort analysis (n = 35,913) found each one-year increase in the gap is associated with a 2% rise in all-cause mortality (hazard ratio, HR, 1.02), with steeper increases in non-cardiovascular and non-cancer mortality at the upper quartiles.[6]
- Cardiovascular surveillance. The retinal microvasculature is the only vascular bed visible non-invasively. Quantitative changes in vessel caliber, density, and tortuosity precede clinical cardiovascular events: in 45,161 UK Biobank participants followed a median of 12.5 years, 29 retinal vascular parameters were each associated with roughly 10–20% higher stroke risk per standard deviation, though adding them to conventional risk factors barely improved discrimination — the area under the ROC curve (AUROC), a 0-to-1 measure of how well a model separates the people who go on to have a stroke from those who do not, rose only from 0.739 to 0.752.[7] The same images also feed AI models that estimate the likelihood of coronary artery calcium rather than an Agatston score, well enough (AUROC 0.742) to reclassify borderline- and intermediate-risk adults when added to a standard risk equation — though several of that model's authors are employees or shareholders of the company that owns it.[8][9]
A retinal exam is therefore an unusually information-dense image — though the systemic-aging read-outs below remain research tools, not part of a standard eye-exam report.
Oculomics: the retina as a biological-age clock
(Weak–preliminary — research-stage, not clinical care.) The models below are genuine and rapidly improving, but none is validated or endorsed for routine screening of healthy adults, and two of the entries are still unreviewed preprints. Read this section as a preview of where the field is heading, not as advice to seek out any particular scan.
| Model / framework | Imaging modality | What it predicts | Performance |
|---|---|---|---|
| Xception network | Color fundus photography | Retinal Age Gap → all-cause mortality | Mean absolute error (MAE) ~3.55 years, against about 8.1 years for simply guessing the average age of the same UK Biobank population, or 6.7 years for guessing the average age of each sex; larger 2025–26 models reach ~2.8 years on internal validation but degrade badly out of cohort (see the third takeaway below)[10]; each 1-year RAG increase → 2% higher all-cause mortality (HR 1.02)[11] |
| RETFound | Color fundus photography | Chronological age; sex-stratified aging correlates | MAE ~2.85 years in UK Biobank (n = 71,343), 3.58 years on external validation in the Rotterdam Study; the retinal age gap tracks metabolic syndrome more strongly in men and the retinal vasculature more strongly in women[12] |
| RetiAGE | Fundus images | Probability of being over 65, used directly as a 10-year mortality score | Top vs bottom quartile: 67% higher all-cause, 142% higher cardiovascular and 60% higher cancer mortality over 10 years — off a small base, 636 cardiovascular deaths among 56,301 participants[13] |
| Cheung deep-learning model | Color fundus photography (four fields, both eyes) | Detecting existing Alzheimer's disease-dementia | AUROC 0.93 in internal validation, but 0.73–0.91 across five external test sets and 0.68–0.86 against amyloid-PET status; a case-control design that flatters apparent accuracy[14] |
| LAVA framework | Fundus images (retinal vascular morphology) | Explaining which retinal features an Alzheimer's classifier uses | An interpretability tool rather than a screening test, and it reports no AUROC at all: the underlying network reached 71–75% accuracy in five-fold cross-validation on 100 fundus images from 61 UK Biobank Alzheimer's subjects against 100 images from 80 matched controls[15] |
| FusionFM | Color fundus photography, benchmarking four eye-specific foundation models (RETFound, VisionFM, RetiZero, DINORET) | Glaucoma, diabetic retinopathy and AMD detection; diabetes and hypertension prediction | A benchmarking exercise rather than a clinical model: fusing the four gave only modest gains, and predicting systemic disease in external cohorts "remains challenging"[16] |
Three takeaways:
- The retina is a promising aging clock. The deep-learning model estimates chronological age about as accurately as a DNA-methylation clock does — a mean absolute error of 3.55 years, against the 3.3–5.2 years reported for methylation clocks — from a 30-second clinic photo rather than a blood draw. That is an accuracy comparison, not a mortality one: no study has run a retinal clock head-to-head against a methylation clock for predicting death, the retinal mortality signal itself is modest (hazard ratio 1.02 per year, with a 95% confidence interval whose lower bound touches no effect at all), and no professional body yet uses it for risk stratification.[17]
- The downstream signals differ by sex. In the same UK Biobank analysis, a high retinal age gap in men tracked metabolic syndrome most strongly, while in women both the model's attention maps and the genetic associations pointed to the retinal vasculature. Training separate male and female models did not improve accuracy, though, and the sex difference in the gap itself was very small — so this is a research finding about what retinal aging means in each sex, not yet a reason to read the same image differently in the clinic, and it is still an unreviewed preprint.[18]
- They do not travel well. Retinal clocks degrade sharply outside the population they were trained on. A 2026 model trained on 27,214 disease-free Japanese adults predicted age to within 2.78 years internally and 3.39 years in a second Japanese screening cohort, but only 8.63 years in a population-based UK hospital dataset.[19] RetiAGE fell from an internal AUROC of 0.97 in Korea to 0.76 in UK Biobank, and the Alzheimer's model above from 0.93 internally to 0.73–0.91 across five external test sets. Any number one of these tools gives you is only as good as the cohort it was calibrated in.
The clinical infrastructure is still uneven, and most adults will not receive a Retinal Age Gap on a regular eye-exam printout — which is just as well, because the individual number is not yet stable enough to act on. In a reproducibility study of 67 healthy adults, photographing the same eye twice in one sitting (41 people) moved the predicted retinal age by about 2.1 years on average, and photographing it at two separate visits (26 people) by about 2.4 years — in one individual by just over nine. The population-level signal survives that noise; a single person's read-out largely does not.[20] OCT is now routine in many practices and worth having when it is offered as part of a comprehensive exam, but there is no professional-society basis for asking for OCT-angiography or an AI aging-clock read-out as a healthy-adult screening test. The high-yield ask is simply a thorough dilated exam on the cadence below.
The macula: lutein, zeaxanthin, and MPOD
The macula's structural and functional integrity depends on the bioaccumulation of two dietary xanthophyll carotenoids — lutein and zeaxanthin — that uniquely cross the blood-retina barrier and concentrate in the fovea. Their concentration is measured clinically as macular pigment optical density (MPOD).
These pigments do two physical jobs:
- Filter short-wavelength blue-violet light before it can damage photoreceptors. MPOD varies enormously between people, from essentially zero to about 1.60. At the top of that range the macular pigment absorbs roughly 97.5% of incident short-wave visible light — only 2.5% reaches the photoreceptors, against 100% in someone with no macular pigment at all.[21]
- Quench reactive oxygen species in the retina, the most metabolically demanding tissue in the body and consequently the most exposed to oxidative damage.[22]
The visual-performance consequences of higher MPOD are real and measurable in healthy adults: faster visual processing, better contrast sensitivity, and improved glare recovery. Dark adaptation is the exception that does not travel: young adults with high MPOD recover vision after a bright flash nearly two minutes faster than those with low MPOD, but in adults over 60 — the group this page is written for — dark-adaptation speed appears to be independent of MPOD.[23] Low MPOD may turn out to be a pre-disease marker for AMD, but the evidence is thin: in a single cross-sectional comparison of 30 fellow eyes of late-AMD patients against 30 age-matched healthy eyes, two-thirds of the fellow eyes (20 of 30) had critically low MPOD or shortened photoreceptor outer segments, against 5 of 30 controls (17%). No eye was followed forward, so this is a snapshot of a high-risk state, not a demonstration that low MPOD precedes disease.[24]
A newer and more speculative idea is that lutein might act systemically before it concentrates in the eye. A 2025 cross-sectional analysis of 7,554 US adults from the National Health and Nutrition Examination Survey (NHANES), matched on background characteristics, estimated biological age with the Klemera-Doubal Method — an algorithm that infers biological age from routine clinical biomarkers, applied here separately to cardiovascular, liver and kidney biomarker panels and to an overall score — and found higher combined lutein + zeaxanthin intake associated with slightly attenuated biological aging (odds ratio 0.93 — a 7% lower likelihood of accelerated aging). The same analysis reported a far larger association with death from any cause: people in the top quarter of intake died at about 60% the rate of those in the bottom quarter, roughly 40% lower mortality (hazard ratio 0.60) — an effect far too large for a single carotenoid. The authors also report a ceiling: intake beyond a threshold buys nothing further.[25] The mechanistic story is thinner than it looks. In cell and rodent work lutein lowers reactive oxygen species and lipid peroxidation, damps the cell's main inflammatory signalling switch, and delays oxidative-stress-induced senescence in retinal pigment epithelial cells and in mesenchymal stem cells.[26] The telomere and immune links are softer still: that review's only telomere evidence is a cross-sectional association between plasma lutein and white-blood-cell telomere length in 2,007 adults, and the immune-cell and telomere findings often quoted alongside it come from a gene-expression analysis inside the intake study above, not from any experiment in which lutein was given.[27] Treat this as observational and hypothesis-generating, not established: people who eat more lutein also eat more vegetables and tend toward healthier lifestyles overall, so much of the apparent benefit is likely healthy-user confounding rather than a direct effect of the carotenoid. (Evidence: weak-preliminary.)
How to actually get enough
The AREDS2 trial tested 10 mg lutein + 2 mg zeaxanthin, but it is easy to over-read the result. Adding lutein/zeaxanthin to the core antioxidant-plus-zinc formula produced no significant additional benefit in the trial's primary analysis (hazard ratio 0.90).[28] The value of lutein/zeaxanthin is that it works at least as well as beta-carotene as the carotenoid in the formula while avoiding beta-carotene's lung-cancer risk in smokers. Ten years on — five randomised years plus five further years in which everyone took the supplement — the lutein/zeaxanthin arms had about 15% less progression to late AMD than the beta-carotene arms (hazard ratio 0.85) and about 9% less than no lutein/zeaxanthin at all (hazard ratio 0.91).[29] The broader picture comes from the 2023 Cochrane review, which pooled 26 trials in 11,952 people. Adding lutein or zeaxanthin on top of the AREDS formula "may have little or no effect" on progression to late AMD — and note that almost every participant in the trial driving that comparison was already taking the AREDS formula, so this is not a test of lutein against nothing. The full antioxidant-plus-zinc formula, by contrast, "probably slows down progression to late AMD" (odds ratio 0.72 — about 28% lower odds of progression), though that pooled estimate rests on 3 of the 26 trials and 2,445 participants.[30] So lutein is best understood as one component of a formula, not a standalone eye supplement. (Evidence: moderate — Cochrane, moderate certainty for the full formula.) Most adults fall well short: typical US intake is about 1–2 mg a day, against the roughly 6–12 mg range associated with lower AMD risk at the bottom end and with cognitive benefit in supplementation trials at the top.[31]
The food-first route is straightforward:
- Cooked spinach — by far the densest practical source: about 20 mg of lutein plus zeaxanthin per cup boiled from fresh, and about 30 mg per cup from frozen.[32]
- Cooked Swiss chard — comparable to spinach, about 19 mg per cooked cup; collards are another good leafy source. Cooked kale is the surprise: about 6 mg per cooked cup on current USDA figures, roughly a third of spinach, because a cup of cooked kale weighs little.
- Egg yolks — the lutein matrix is highly bioavailable due to the lipid carrier.
- Orange/yellow peppers, corn — primary zeaxanthin sources.
A reasonable midlife default: leafy greens daily, eggs regularly. Targeted AREDS2-formulation supplementation is appropriate for people with diagnosed intermediate AMD; for healthy adults, the food-first pattern is what the systemic-aging data actually rests on.
Cognition: the largest trial was null
Lutein concentrations in the macula track lutein concentrations in the brain, which makes MPOD a usable surrogate for central-nervous-system carotenoid status — a plausible reason to expect a cognitive benefit. The largest and most relevant test of that hypothesis found none. The AREDS2 cognition sub-trial randomised more than 3,000 older adults with macular degeneration and at risk of its advanced form (mean age 73) to lutein/zeaxanthin, omega-3 fatty acids, both, or placebo and followed cognitive function double-masked: lutein/zeaxanthin had no statistically significant effect on cognition, and omega-3 was likewise null. Two caveats keep it from being decisive on its own: everyone, placebo arm included, was also taking a combination of antioxidant vitamins and zinc, and the cognitive battery was administered by telephone every two years. (Evidence: strong — large RCT in the relevant age group.)[33]
The often-cited positive result comes from a much smaller study: a double-masked, placebo-controlled trial of just 51 healthy adults aged 18–30, in which a year of lutein + zeaxanthin supplementation raised MPOD and improved spatial memory, reasoning, and complex attention.[34] That is a small, preliminary trial — but not, as is usually assumed, one that can be set aside simply because it was done in the young: the same research group ran a mirror-image trial the same year, on the same computerised battery, in 51 community-dwelling older adults (mean age 74) taking 12 mg a day for a year, and reported gains in complex attention and cognitive flexibility.[35] Pooled, the evidence is split: a meta-analysis of lutein trials specifically is null,[36] while a meta-analysis of carotenoid trials in 4,402 adults found a small but statistically significant benefit — though that pool mixes carotenoids other than lutein and is itself dominated by AREDS2.[37] What separates AREDS2 from the small positive trials is not the participants' age but its size, its dose, its telephone battery against their in-person computerised one, and a population of macular-degeneration patients already taking a background supplement. The honest read: the food-first route incurs no downside, but do not expect a cognitive payoff from a lutein pill.
The vision-loss → dementia pathway
The mechanism connecting vision loss to cognitive decline mirrors the hearing story, with one important nuance:
- Sensory deprivation reduces neuroplasticity, social engagement, mobility, and environmental complexity — the same psychosocial pathway that makes hearing loss the largest single modifiable midlife dementia factor, tied with high LDL cholesterol, each accounting for an estimated 7% of global dementia cases. See Hearing and Purpose.
- Visual impairment specifically drives social withdrawal in the late midlife window because reading, driving, facial recognition, and screen use all collapse together.
- Shared pathology (amyloid-beta, oxidative stress, microvascular dysfunction) is real but may contribute less than functional impairment. The dementia signal appears to track the severity of vision loss rather than any particular eye diagnosis: a Korean health-insurance study of over six million people found dementia risk increasing with the degree of visual loss, and a 2026 records analysis found the excess Alzheimer's risk sitting with blindness rather than with AMD itself.[38] [39]
The implication: uncorrected vision is the modifiable variable. Cataract surgery, glasses, low-vision rehabilitation, and AMD treatment that preserves function are not cosmetic interventions — they preserve the sensory bandwidth that maintains cognitive engagement. The same argument is made for hearing aids, and it is worth being precise about how strong it is: the only randomised trial of hearing intervention was null on its primary cognitive endpoint,[40] and on the vision side no randomised trial has reported at all — the first, a three-year trial of spectacles for cognitive change in 760 adults over 60 in India, is still running.[41] Correcting vision is worth doing on its own merits, today; the cognitive payoff is a plausible bonus resting on observational data, not a demonstrated one.
The psychological burden of progressive vision loss is part of the clinical picture: in a 2026 meta-analysis of people with irreversible vision loss, about 27% of those with AMD screened positive for depression, against 21% across all causes of irreversible vision loss, and about 22% of the whole irreversible-vision-loss population screened positive for anxiety.[42] The excess over sighted controls is clearer for depression than for anxiety: across case-control studies, people with AMD were more likely to report depressive symptoms but not more likely to report anxiety symptoms.[43] The direction that has actually been studied runs from the vision loss to the mood disorder; there is no prospective human evidence that psychological stress accelerates AMD. Vision rehabilitation and psychological support belong in the AMD treatment plan, not after it.
The conditions you're most likely to actually get
The frontier material above is genuinely interesting, but the eye problems an aging reader will most likely develop and act on are older, better-studied, and more treatable. This is where the strongest evidence lives.
Cataract
Cataract is the leading cause of blindness worldwide — responsible for about 40% of it. In 2020, of the 43 million people who were blind, 17 million were blind from cataract, and a further 83 million had moderate-to-severe vision impairment from it.[44] The main modifiable risk factors are smoking, diabetes, systemic and inhaled corticosteroids, alcohol, and ultraviolet light — though the UV link is subtype-specific: it is most consistent for cortical cataract, the most recent systematic review of occupational solar exposure also found new support for the nuclear subtype, and it found no substantial new evidence for the posterior subcapsular one — though exposure measurement in this literature is generally poor.[45] Certified UV-blocking sunglasses are a cheap, high-value hedge.
Cataract surgery is one of the highest-value operations in medicine, and its benefits reach beyond acuity. In a randomised trial of 306 women over 70, expedited first-eye surgery — about four weeks' wait against a routine twelve-month wait — cut the overall rate of falling by about a third (rate ratio 0.66) and cut fractures from 8% to 3%. The proportion of women who fell at least once did not change; the gain is in repeat falls. (Evidence: strong — RCT.)[46] A nuance worth knowing: the companion trial of second-eye surgery found a very similar reduction in the falling rate (rate ratio 0.68, against 0.66 in the first-eye trial), but it was smaller — 239 women against 306 — and the result did not reach statistical significance. The range its data are compatible with runs from a 61% reduction to a 19% increase, so the trial simply cannot tell a large benefit from none, and its authors concluded that the effect "remains uncertain" rather than absent. Read that as an underpowered trial, not as a reason to decline the second operation — which did improve stereopsis, confidence and visual disability.[47] Observational data also link cataract extraction to lower dementia risk, though less cleanly than the headlines suggest. The best-known cohort — 3,038 older adults followed for a mean of roughly eight years — found 29% lower dementia risk after surgery (hazard ratio 0.71, 95% confidence interval 0.62 to 0.83), with glaucoma surgery, which does not restore vision, showing no such association; that internal control rests on only 105 operations, and is too imprecise to exclude a benefit as large as the cataract one. Residual confounding remains the live worry.[48] Pooling since then has gone both ways, though on different outcomes: a 2024 meta-analysis of 24 studies and 558,276 people found a 25% lower risk of long-term cognitive decline but graded the certainty "very low to low" and called for randomised trials,[49] while another pooling 720,075 people found no significant effect on all-cause dementia at all.[50] (Evidence: weak — inconsistent observational pooling, no randomised data.)
Glaucoma
Glaucoma is asymptomatic until it is advanced, and intraocular pressure is the only major modifiable risk factor. Higher-risk groups include older adults, people with a family history, those of Black or Hispanic/Latino ancestry, high myopes, and people with diabetes. The practical message is simply to be examined on schedule, because early disease is silent and the vision it takes does not come back.
Treatment has a clear evidence-based front line. The LiGHT trial randomised 718 people with newly diagnosed glaucoma or ocular hypertension to selective laser trabeculoplasty (a quick outpatient laser) versus daily eye drops; at three years, 74% of laser-first patients still needed no drops at all, and no one in the laser arm had needed glaucoma surgery, against 11 patients in the drops arm.[51] The six-year extension, which 524 of the original participants completed, found about 70% of laser-treated eyes still at target pressure with no drops and no surgery, less disease progression (19.6% versus 26.8% of eyes) and fewer trabeculectomies (13 versus 32 eyes) — though the extension's own primary outcome, quality of life, showed no difference between the arms.[52] That establishes laser as a reasonable first-line option, not just a fallback. (Evidence: strong — RCT.) One calibration on screening: the US Preventive Services Task Force concluded in 2022 that evidence is insufficient to recommend population glaucoma screening in asymptomatic adults aged 40 and older — a grade I statement, meaning neither for nor against, and still the operative position — so the case-finding here happens within a comprehensive eye exam rather than as a standalone screening program.[53] Exercise acutely lowers intraocular pressure — roughly twice as much in sedentary people as in the already fit — and is broadly good for the eye, but evidence that it prevents glaucoma is weak.[54] The one caveat for lifters: in a study of 20 healthy young adults, holding the breath through a heavy repetition raised pressure substantially more than breathing normally, and the effect was largest on big compound lifts. Whether that matters clinically is untested, but if you have glaucoma or ocular hypertension it is worth raising your lifting technique with your ophthalmologist rather than assuming it is neutral.[55]
Dry eye disease
Dry eye becomes common with age and is more frequent in women: US survey data put diagnosed prevalence at about 7% of adults overall, rising from under 3% in young adults to nearly 19% in those 75 and older.[56] A widely believed remedy does not hold up: the large, double-masked DREAM trial gave symptomatic patients 3,000 mg/day of fish oil and found it no better than placebo on dry-eye signs or symptoms — a null the trial's own withdrawal extension did nothing to overturn, though that extension was tiny — 43 patients randomised to continue or stop — and is better read as uninformative than as confirmatory.[57] (Evidence: strong for the null — large RCT.)[58] Prolonged screen work worsens evaporative symptoms mainly because people blink less often and less completely while staring at a display — partial blinks fail to resurface the tear film.[59] That is what the 20-20-20 rule (every 20 minutes, look 20 feet / 6 metres away for 20 seconds) is meant to counter — though the numbers are a clinical convention rather than a trial result, and the one controlled test of the break schedule, in 30 adults over a single 40-minute session, found no effect on symptoms, reading speed or accuracy. Treat it as a free, sensible prompt to blink and refocus, not as a proven therapy.[60]
Myopia and refractive error
For children, the strongest lever against myopia is outdoor time. A cluster-randomised trial added 40 minutes of outdoor school time per day and cut 3-year myopia onset from about 40% to 30% — roughly a quarter fewer new cases.[61] The 2024 Cochrane review of randomised outdoor-time trials puts that in context: pooling three trials at two years, outdoor time cut myopia onset by about 16% (relative risk 0.84, 95% confidence interval 0.72 to 0.98 — a range that stays below 1, so a real effect), while the three-year estimate rests on a single 394-child study and is too imprecise to call either way (relative risk 0.77, 0.59 to 1.01). (Evidence: moderate — randomised, moderate-certainty Cochrane.)[62] The effect is mainly on onset: across those trials refraction differed by a small but real 0.13 dioptres at two years, and by three years the estimate was too imprecise to separate from no difference at all; axial elongation differed only by hundredths of a millimetre. For a child who is already myopic there is a separate and now-mainstream toolkit — low-dose atropine drops, myopia-control spectacle lenses, orthokeratology and multifocal soft contact lenses all slowed progression in a 2025 Cochrane network meta-analysis of 104 trials in 17,509 children — though most of those estimates are low or very low certainty, most trials ran only one to two years, and what happens after treatment stops is unresolved. That is a paediatric ophthalmology conversation, not a self-serve one. (Evidence: moderate — Cochrane network meta-analysis, mostly low-certainty.)[63] The adult relevance is that high myopia raises lifetime risk of retinal detachment, myopic maculopathy, and glaucoma — a reason for high myopes to keep to a more frequent exam schedule.
Diabetic retinopathy
Blood-sugar control is the dominant modifiable lever here. In type 1 diabetes the effect is large: intensive control cut the onset of retinopathy by 76% and slowed its progression by 54% over a mean 6.5 years, at the cost of a two-to-threefold increase in severe hypoglycemia. (Evidence: strong — RCT.)[64] In type 2 diabetes it is real but far more modest: retinopathy progressed in 7.3% of intensively treated participants against 10.4% on standard therapy over four years in ACCORD-Eye — and that has to be read alongside the fact that the trial's intensive-glycemia arm was stopped early for excess deaths among people at high cardiovascular risk, so "tighter is better" is not unconditional advice for an older adult with heart disease. (Evidence: moderate — RCT, with a safety caveat.)[65] The eye benefit did persist for years after that trial ended, even once glycated haemoglobin (HbA1c) levels had converged: 5.8% versus 12.7% progression at follow-up.[66] One caution that matters if you are starting insulin, a glucagon-like peptide-1 (GLP-1) receptor agonist, or having bariatric surgery: large, rapid drops in blood glucose can transiently worsen retinopathy before the long-term benefit appears, which is why a retinal exam belongs at the start of any aggressive tightening rather than after it.[67] See GLP-1 receptor agonists. This is really a metabolic problem expressed in the eye — see Metabolic flexibility and Glycemic index for the systemic picture.
Mitochondria, supplements, and saffron
Photoreceptors — cones especially — carry the densest mitochondrial load in the retina, and the retina ages faster than most organs, losing roughly 70% of its ATP output over a lifetime.[68] The leading hypothesis for age-related visual decline in structurally intact eyes is therefore a mitochondrial energy problem — declining ATP synthesis in the retinal pigment epithelium and photoreceptor layer, with rising reactive oxygen species and a cascade that may culminate in drusen, RPE apoptosis, and AMD. In aged mice, a month of daily 670 nm exposure improved retinal electroretinogram amplitudes by about 25%, without restoring them to young levels.[69]
The supplement landscape splits into three tiers:
Tier 1 — established AMD-modifying:
- AREDS-type formulation (10 mg lutein, 2 mg zeaxanthin, vitamins C and E, zinc, copper) for people with intermediate AMD or advanced AMD in one eye. The original AREDS trial cut progression to advanced AMD by about a quarter over an average 6.3 years — a 20% five-year probability of progressing on the formula against 28% on placebo — but only in people who already had intermediate AMD or advanced disease in one eye.[70] [71] In people with early AMD the five-year risk of progressing was just 1.3% — far too few events for any benefit to be demonstrable — and eyes with no AMD were never enrolled at all, so the formula has never been tested in healthy eyes. That is the single most important thing to know before buying it. (Evidence: strong — large RCT.) AREDS2 then swapped beta-carotene for lutein/zeaxanthin after beta-carotene was linked to lung cancer in smokers, and randomised a 25 mg zinc dose against the original 80 mg: the two performed identically at five years and again at ten, with no difference in adverse effects between them, so there is no efficacy reason to seek out the higher dose that some products still use.[72] [73] This is a treatment for diagnosed intermediate disease, not a healthy-adult preventive.
Tier 2 — preliminary, surrogate-endpoint only:
- Saffron (Crocus sativus L.) — 20–30 mg/day. The evidence here is thin and easy to over-sell. There is no Cochrane review and no meta-analysis of saffron for AMD — the trials are too few, too small, and too heterogeneous in design and outcome measure to pool, which is why the most recent independent systematic review, published in 2026, gathered ten clinical trials of saffron in eye disease and declined to pool any of them.[74] The largest randomised trial (100 participants, a placebo-controlled crossover with three-month arms) found a clinically trivial gain of well under one line of visual acuity — 0.69 letters.[75] The 12-month open-label extension found visual acuity actually drifted slightly worse (1.6 letters, which the authors suggest may reflect cataract progression) even as an electroretinogram surrogate marker improved.[76] Most saffron trials are small, several are open-label, and the literature clusters within a few author groups and saffron-producing regions. The 2026 narrative review that makes the case for saffron is written by the group that ran nearly all of the trials it reviews, with a first author affiliated with a saffron company and two authors named as co-inventors on the patent covering the branded ophthalmic saffron the review recommends — grounds for caution about publication and conflict-of-interest bias.[77] A separate randomised pilot in primary open-angle glaucoma — 34 patients already on two topical drops, given 30 mg/day of oral saffron extract — reported an intraocular pressure about 3 mmHg lower than placebo at four weeks, a benefit that disappeared once saffron was stopped.[78] Bottom line: interesting mechanism, but the outcome data are preliminary, surrogate-endpoint, and not a reason to add saffron to established care. (Evidence: weak-preliminary.)
Tier 3 — systemic mitochondrial support, essentially no retinal-outcome data:
CoQ10, NAD+ precursors (NMN, NR), and resveratrol all have plausible mitochondrial mechanisms and are sometimes marketed for eye health, but none has meaningful clinical-outcome data in the retina. They are general mitochondrial/longevity supplements, not eye supplements, and are better evaluated against the core supplement evidence on their own merits.
For healthy midlife adults, the honest read: AREDS-type formulas belong to people with diagnosed intermediate AMD; saffron is preliminary and surrogate-endpoint at best; and the mitochondrial stack has no retinal outcome data. None of these is a reason to skip the interventions with real evidence — cataract and glaucoma care, correcting refractive error, not smoking, and a good diet.
Photobiomodulation: the 670 nm finding
One of the more surprising recent findings in ophthalmic gerontology is that brief morning exposure to deep-red light (670 nm) may measurably improve age-related visual decline — though the supporting trials are tiny and the endpoints are surrogates, so treat this as an intriguing early signal rather than a proven intervention.
The mechanism is biophysical: long-wavelength deep-red light penetrates tissue and is absorbed by cytochrome c oxidase in the mitochondrial electron transport chain, which in aged mice improves retinal function on electroretinography by roughly a quarter.[79] A proposed second step — that long wavelengths reduce the viscosity of the nanoscopic water layer around the rotating ATP synthase pumps and so raise their mechanical efficiency — is the trial authors' own hypothesis rather than a separately demonstrated finding.[80]
In a UCL Institute of Ophthalmology study of 20 adults aged 38–70, a single three-minute exposure to 670 nm light improved cone-mediated color contrast thresholds by about 17% on the blue-yellow axis and 12% on the red-green axis; the ten participants retested a week later still showed a smaller gain. The authors note the sample was too small to break the result down by age, sex or color axis.[81]
Timing may matter more than dose. Six of the twenty participants repeated the protocol months later in the early afternoon and showed no improvement at all — consistent with a circadian gating of mitochondrial ATP production, which in fruit flies peaks in the morning hours when 670 nm light raises ATP. But six people, all of them also in the morning arm and with no independent control group, is a hypothesis rather than a finding, and no other group has reproduced it.[82] [83]
Not every study agrees, though the counter-evidence is no stronger than the positive evidence. An uncontrolled 12-month pilot treated 31 people with intermediate AMD and 11 healthy older adults with daily 670 nm light at five times the irradiance used in the UCL work, and found no benefit in the intermediate-AMD group on the structural endpoints that matter — no change in drusen volume and no improvement in outer nuclear (photoreceptor) layer thickness — and no functional gain either. The one significant result went the other way: scotopic thresholds improved by 1.77 decibels in the healthy-ageing group. Different endpoints, a five-fold higher dose, no sham arm and no randomisation — it neither confirms nor refutes the color-contrast finding.[84]
The honest read depends on which device you mean. For consumer 670 nm devices, the positive data are still a handful of small, short pilots from one research group using a color-contrast surrogate endpoint, the one 12-month pilot in intermediate AMD was uncontrolled and null, irradiance and spectral purity vary widely between products, and long-term clinical outcomes are unknown. Plausible and low-risk, but the confidence should be low. (Evidence: weak-preliminary.) For clinic-delivered multiwavelength photobiomodulation — a different intervention, given in a retina clinic at three wavelengths (590, 660 and 850 nm) in short courses repeated every four months — the picture changed between 2024 and 2026. LIGHTSITE III, a double-masked randomised sham-controlled trial of 148 eyes in 100 people with dry AMD, met its prespecified visual-acuity endpoint at 21 months with a gain of about six letters, and by 24 months 6.8% of treated eyes had progressed to geographic atrophy against 24.0% of sham eyes.[85] A second 12-month multicentre sham-controlled trial, by a different group, in 138 eyes with early or intermediate dry AMD, reported drusen volume falling on treatment while it rose on sham, with a between-group advantage of just under four letters of acuity.[86] The US Food and Drug Administration granted that device de novo marketing authorization in November 2024. Both trials have device-industry ties, neither has independent long-term replication, and a 2026 review of the wider field still finds the evidence too heterogeneous in devices, wavelengths and dosing to call photobiomodulation established therapy for most eye conditions.[87] So: if you have intermediate dry AMD this is now a real question for a retina specialist rather than a frontier curiosity — and it is still no reason to buy a home red-light panel. (Evidence: moderate for clinic-delivered multiwavelength photobiomodulation in dry AMD; weak-preliminary for consumer 670 nm devices in healthy eyes.)
Blue light: deconstructed
The opposite end of the visible spectrum has been the subject of a decade of consumer-product marketing that the clinical evidence does not support.
A 2025 systematic review and meta-analysis of randomized crossover trials of blue-light-blocking glasses (BBGs) pooled just three double-masked trials and 49 participants, and found no statistically significant effect on any objective sleep outcome — sleep onset latency fell by about five minutes, an estimate so imprecise that the data are equally consistent with the glasses making sleep onset ten minutes slower, and wake-after-sleep-onset barely moved. That is an underpowered null rather than a demonstration that the glasses do nothing.[88] BBGs also show minimal effect on contrast sensitivity, color discrimination, or digital eye strain.[89]
The strongest single source is the 2023 Cochrane review of blue-light-filtering lenses (17 randomised trials, five to 156 participants each, with follow-up from less than a day to five weeks): it found the lenses probably have little or no effect on best-corrected visual acuity and may make no difference to visual fatigue, and — importantly — could not determine whether the lenses affect sleep at all, because the six trials in 148 people that measured it split three-to-three and the certainty was rated very low. There was no randomised evidence at all on macular health. The reviewers pooled nothing and no outcome reached high certainty. (Evidence: moderate — a careful review of low-certainty trials.)[90] For scale: the blue light a phone or tablet emits at full brightness has been measured at roughly 0.1–0.4% of the international exposure limit for blue light, against about 10% for looking at a blue sky on a sunny day — so a screen delivers something on the order of one percent of the blue light of open sky, and typical blue-blocking glasses filter only a fraction of that.[91] The American Academy of Ophthalmology's public guidance says the same in words rather than numbers: there is "no scientific evidence that blue light from digital devices causes damage to your eye", and "the Academy does not recommend any special blue light-blocking eyewear for computer use" — the 20-20-20 rule and sensible screen ergonomics are what's evidence-based, though the Academy also notes that blue light's effects are still being researched.[92]
Two practical caveats:
- Most of what is sold as "blue-blocking" is far too weak to change circadian signaling — but the strong versions are not nothing. A characterisation of 26 commercial models introduced a melanopic daylight filtering density (mDFD) metric and set the bar for a true blue blocker at 1.0. The products measured spanned 0.01 to 3.43, and the clear and near-clear lenses sold in most optical shops for screen use fell below the bar; under office LED lighting only about 31% of the filters tested cut melanopic input enough to reach the recommended evening level. In the 16 field studies of glasses that did clear the threshold, worn in the hours before bed, at least one sleep or circadian outcome improved every time — and these authors argue that the pooled null results partly reflect lumping clear lenses in with real filters. Two of the paper's five authors work for a blue-light-filtering software company.[93]
- Daytime full-spectrum natural light exposure is what matters. The biologically necessary long-wavelength input for retinal mitochondria comes from outdoor light, and morning sun anchors the circadian clock through melanopsin-expressing retinal ganglion cells. See Sun exposure and Circadian rhythms.
The pragmatic takeaway: skip the clear "computer glasses" — they do nothing for eye strain or acuity, and there is no good evidence they help sleep or macular health — get outdoors in the morning, dim screens at night, and treat insomnia with sleep hygiene (or evaluation for sleep-disordered breathing) rather than with optical filters. If you want an optical intervention for sleep specifically, only a genuinely deep orange lens worn in the last hour or two before bed has any physiological basis, and even then it is a weaker lever than turning the lights down.
Lifestyle inputs to retinal aging
The same systemic interventions that move cardiovascular and dementia risk also move retinal-aging biomarkers — which is the point of treating the retina as a biosensor in the first place.
Circadian alignment
The retina has its own peripheral circadian clock, and the clearest demonstration of what it does comes from flies: disrupting the Clock:Cycle complex in adult Drosophila photoreceptors changes the expression of 15–20% of their genes, including the core phototransduction machinery, disrupts global chromatin accessibility, and causes light-dependent retinal degeneration alongside light-independent oxidative stress.[94] The mammalian counterpart of that heterodimer is CLOCK:BMAL1, and the human evidence is so far associative rather than mechanistic — but the principle transfers: photoreceptor proteins degrade on light exposure and must be re-synthesized on a daily schedule; circadian disruption mis-times that synthesis and accelerates photoreceptor decline.[95] — an expert-consensus editorial rather than primary data. People with AMD report delayed sleep-wake timing and worse morning function than healthy controls.[96] See Circadian rhythms.
Aerobic exercise
A systematic review of 21 studies found that aerobic exercise interventions in adults, with or without cardiovascular risk factors, produced wider central retinal arteriolar equivalents and narrower venular equivalents — the favorable microvascular signature — and that higher habitual activity tracked with narrower venules in both children and adults. In the same review, twelve weeks of high-intensity interval training improved retinal flicker-light dilatation, a marker of endothelial reactivity, in older patients with cardiovascular risk factors.[97] A separate cross-sectional analysis of nearly 43,000 people with type 2 diabetes found higher physical activity associated with lower odds of moderate and severe non-proliferative retinopathy, and its longitudinal arm found about 19% lower risk of developing retinopathy — though in that cohort both arterioles and venules were wider, a reminder that the vessel-calibre signature is not identical across populations.[98] Exercise also lowers intraocular pressure: by about 2.7 mmHg immediately after a resistance-training session, and — as a lasting change in resting pressure — by about 4.6 mmHg in people with glaucoma after three months of aerobic training. It also raises circulating brain-derived neurotrophic factor (BDNF), with retinal BDNF about 20% higher in trained mice.[99] See Zone 2 and VO₂ max.
Time-restricted eating
In mice given a chemical retinal insult, intermittent fasting started before the insult preserved the retinal pigment epithelium and photoreceptors, suppressed microglial and Müller-cell overactivation, and reduced retinal reactive oxygen species. Starting the protocol later in life still helped, but progressively less: near-complete protection in young mice became only partial at 12 months and weaker again at 16 months.[100] That is a male-mouse toxin model rather than AMD, and it tests prevention rather than treatment, so read it as mechanism — the same mechanism (AMPK / SIRT1 activation, mTOR suppression, mitochondrial biogenesis) that time-restricted eating is thought to work through systemically.[101] In humans, a 2026 systematic review of intermittent fasting in adults aged 60 and over ranked 16:8 time-restricted eating among the most effective protocols for weight and body-mass-index reduction without loss of lean mass — though that ranking came from a network meta-analysis of seven trials measuring body composition only, while the review's cardiometabolic and cognitive findings rest on a narrative synthesis of observational studies. Those observational data carry a caution that runs against the assumption that a narrower window is always better: eating windows of 10 hours or less were linked to worse outcomes including lower cognitive scores — the underlying arterial-stiffness study used a cut-off of 11 hours — and overnight fasts longer than about 12.4 hours to 58% higher cardiovascular mortality in this age group, in a U-shaped pattern where both extremes fared worse than the middle. Note that a 16:8 schedule has an 8-hour window and so falls inside that band — the same review ranks it well on body composition and flags it on vascular measures, from two different tiers of evidence. The review's own authors call the observational half hypothesis-generating rather than causal.[102] Window length is worked through properly in Fasting.
UV protection
Chronic UV exposure is an established risk factor for cataract and contributes to AMD. The intervention is unglamorous and cheap: real UV-blocking sunglasses on bright days. See Sun exposure.
A practical vision protocol for healthy midlife adults
- Get a comprehensive dilated eye exam by age 40, then on a risk-adjusted cadence. For asymptomatic adults without risk factors, the American Academy of Ophthalmology's schedule is every 2–4 years from 40 to 54, every 1–3 years from 55 to 64, and every 1–2 years from 65 — with each band tightened by one step (1–3 years at 40–54, 1–2 years at 55–64) if you have a family history of glaucoma, Black or Hispanic/Latino ancestry, or high myopia.[103] With diabetes the Academy is more specific: an exam at diagnosis for type 2 diabetes, five years after onset for type 1, and at least yearly thereafter in both.[104] This is a full dilated exam that checks for cataract, glaucoma, diabetic retinopathy, and AMD — not a population "glaucoma screening," which the evidence does not yet support. OCT is worth having when it's offered as part of that exam; there is no need to seek out OCT-angiography or an AI "aging clock" as a healthy-adult test.
- Treat any correctable visual loss promptly. Glasses, cataract surgery, AMD management — the dementia and mortality literature points the same way as the hearing literature, though the case for vision is thinner than the case for hearing rather than the same as it: the genetic evidence supports the hearing–dementia link for all-cause dementia and Alzheimer's disease, while on the vision side it reaches only the residual non-Alzheimer, non-vascular dementia category.[105] A condition you've been told about but not addressed is the modifiable factor.
- Eat the macula. Leafy greens daily (cooked spinach is the densest practical source), eggs regularly, orange/yellow peppers and corn for zeaxanthin. Aim for the dietary pattern, not the supplement, unless you have intermediate AMD — in which case AREDS2 is established and your ophthalmologist will guide you.
- Wear real UV-blocking sunglasses on bright days. Broad-spectrum coverage; the cheap pair from the gas station is fine if it's certified.
- Skip the clear "computer glasses". The clear or near-clear anti-blue lenses sold for screen use do not move eye strain or visual acuity, and there is no good evidence they help sleep or macular health — though the pooled sleep nulls are underpowered rather than decisive. Only a genuinely deep orange lens worn in the last hour or two before bed has any physiological case, and dimming the room is the stronger lever. If you struggle with screens, treat that with the 20-20-20 rule — a free, sensible prompt to blink and refocus rather than a proven therapy — plus brightness and font adjustments and a real ophthalmologic workup.
- Get morning sunlight. It anchors the circadian clock and provides full-spectrum input to the retina. See Circadian rhythms and Sun exposure.
- Train both ways and don't smoke. Exercise improves retinal microvasculature on objective imaging; smoking is a large independent AMD risk factor. The ear and the eye respond to the same systemic levers as the heart and brain.
- Don't count on saffron, home red-light devices, or "eye supplements." The saffron and consumer 670 nm data are small, short, and surrogate-endpoint; the one longer 670 nm study was an uncontrolled pilot and was null on structure; and NMN/NR/resveratrol/CoQ10 have no retinal-outcome evidence. If you have diagnosed intermediate AMD, the established move is an AREDS-type formula under your ophthalmologist's guidance — and clinic-delivered multiwavelength photobiomodulation, a different intervention from a home panel, is now a question worth raising with a retina specialist.
- Treat hearing loss in parallel. The two sensory pathways feed into the same dementia pathway, and the risks stack: people with both impairments carry roughly 50% higher dementia risk than people with neither — a risk the authors judge to exceed that of either impairment alone and of the two combined.[106] Treating both is worth doing on its own merits; whether treating both helps cognition more than treating one has never been tested. See Hearing.
- Sleep close to 8 hours and aim for circadian regularity. Photoreceptor protein turnover is clock-gated; AMD patients show measurable circadian disruption. Sleep is not unrelated to the eye.
What's overhyped or wrong
- "Blue light from screens damages your eyes." It does not, in adults under normal exposure conditions. The American Academy of Ophthalmology is explicit about this; the multi-billion-dollar BBG market is consumer marketing ahead of clinical evidence.[107]
- "Lutein supplements are all you need for eye health." Lutein is one input. Microvascular health (blood pressure, glycemic control, smoking status, exercise) drives more of the variance than carotenoid intake does. Take both seriously.
- "AMD is just an eye disease." It is the visible expression of a systemic mitochondrial-and-microvascular aging process. The retina is the diagnostic; the disease is closer to the rest of the body's aging biology than to a localized pathology.
- "Red light therapy reverses aging vision." The 670 nm mechanism is real, but the human evidence for consumer devices is a handful of tiny pilots on a color-contrast surrogate, and a 12-month uncontrolled pilot found no structural benefit. Clinic-delivered multiwavelength photobiomodulation for dry AMD is a separate intervention with sham-controlled randomised trials behind it; the home 670 nm panel is plausible and low-risk, not proven.
- "Omega-3 supplements treat dry eye." The large DREAM trial found fish oil no better than placebo. Dry eye is better managed with lid hygiene, artificial tears, screen breaks, and evaluation for the underlying cause.
- "Lutein supplements alone slow AMD." Cochrane finds lutein or zeaxanthin added on top of the AREDS formula has little or no effect; the benefit belongs to the full antioxidant-plus-zinc formula, and overwhelmingly to people who already have intermediate AMD — about 78 fewer progressions to late AMD per 1,000 people with intermediate AMD, against about 4 per 1,000 in early AMD.
- "Cataract surgery is purely cosmetic / can wait indefinitely." It cuts falls in a randomised trial and is associated with lower dementia risk. Untreated cataract in the late midlife window is a quietly expensive choice.
- "Eye exams just check whether you need glasses." A modern dilated exam with OCT screens for diabetic retinopathy, AMD, glaucoma, and increasingly the systemic biomarkers above. The signal density per visit is high.
Further reading
- Livingston G et al. Dementia prevention, intervention, and care: 2024 report of the Lancet Standing Commission. Lancet 2024.[108]
- Zhu Z et al. Retinal age gap as a predictive biomarker for mortality risk. British Journal of Ophthalmology 2023.[109]
- Grimbly MJ et al. Estimating biological age from retinal imaging: a scoping review. BMJ Open Ophthalmol 2024.[110]
- Nusinovici S et al. Retinal photograph-based deep learning predicts biological age, and stratifies morbidity and mortality risk. Age and Ageing 2022.[111]
- Trofimova O et al. Deep learning aging marker from retinal images unveils sex-specific clinical and genetic signatures. medRxiv 2025 (preprint, not peer-reviewed).[112]
- Weinreb RN et al. A framework for healthcare from the eye: oculomics as a powerful window to systemic health. Ophthalmology 2026.[113]
- Rim TH et al. Deep-learning-based cardiovascular risk stratification using coronary artery calcium scores predicted from retinal photographs. Lancet Digital Health 2021.[114]
- Cheung CY et al. A deep learning model for detection of Alzheimer's disease based on retinal photographs: a retrospective, multicentre case-control study. Lancet Digital Health 2022.[115]
- Yousefzadeh N et al. Neuron-level explainable AI for Alzheimer's disease assessment from fundus images. Scientific Reports 2024.[116]
- Stringham JM et al. Lutein across the Lifespan: From Childhood Cognitive Performance to the Aging Eye and Brain. Curr Dev Nutr 2019.[117]
- Renzi-Hammond LM et al. Effects of a lutein and zeaxanthin intervention on cognitive function — RCT in healthy young adults. Nutrients 2017.[118]
- Tao M et al. The association between lutein and zeaxanthin intake and multi-level biological aging. Frontiers in Nutrition 2025.[119]
- Ye J et al. Effects and mechanisms of lutein on aging and age-related diseases. Antioxidants 2024.[120]
- Macular pigment optical density and photoreceptor outer segments as predisease AMD biomarkers. PMC 2020.[121]
- Carotenoids in age-related macular degeneration — systematic review. Antioxidants 2021.[122]
- Saffron therapy for ongoing AMD treatment — 12-month extension. PMC 2024.[123]
- Maggi MA et al. Saffron as a retinal neuroprotectant: a narrative review of preclinical studies and clinical results. Antioxidants (Basel) 2026. Written by the trial group itself, with a first author affiliated to the saffron company whose patented preparation the review promotes.[124]
- Gu X et al. Age-related macular degeneration and dementia: association through pathogenesis or visual impairment? J Alzheimers Dis 2026.[125]
- Sivapathasuntharam C et al. Aging retinal function is improved by near infrared light (670 nm) that is associated with corrected mitochondrial decline. Neurobiology of Aging 2017.[126]
- Morning exposure to deep red light improves declining eyesight. UCL News 2021.[127]
- Shinhmar H et al. Weeklong improved colour contrasts sensitivity after single 670 nm exposures associated with enhanced mitochondrial function. Scientific Reports 2021.[128]
- Khorrami-Nejad M et al. Blue-light-filtering spectacle lenses in managing vision-related symptoms — an updated review. PMC 2026.[129]
- Luna-Rangel FA et al. Efficacy of blue-light blocking glasses on actigraphic sleep outcomes: a systematic review and meta-analysis of randomized controlled crossover trials. Frontiers in Neurology 2025.[130]
- Glickman GL et al. Optimizing the potential utility of blue-blocking glasses for sleep and circadian health. Translational Vision Science & Technology 2025.[131]
- Circadian disruption & retinal physiopathology — expert consensus. PMC 2022.[132]
- AMD and circadian preference. PMC 2025.[133]
- Streese L et al. Physical activity and exercise improve retinal microvascular health. Atherosclerosis 2020.[134]
- Intermittent fasting attenuates glial hyperactivation and photoreceptor degeneration in NaIO3 AMD model. PMC 2025.[135]
- Couto-Alfonso S et al. Intermittent fasting and healthy aging in older adults: a systematic review of cardiometabolic, mental health and cognitive outcomes with a network meta-analysis of anthropometric measures. Nutrients 2026.[136]
- Antioxidants in age-related macular degeneration — lights and shadows. PMC 2025.[137]
- Chew EY et al. Effect of omega-3 fatty acids, lutein/zeaxanthin, or other nutrient supplementation on cognitive function: the AREDS2 randomized clinical trial. JAMA 2015.[138]
- Evans JR, Lawrenson JG. Antioxidant vitamin and mineral supplements for slowing the progression of age-related macular degeneration. Cochrane Database of Systematic Reviews 2023.[139]
- Chew EY et al. Long-term outcomes of adding lutein/zeaxanthin and omega-3 fatty acids to the AREDS supplements — AREDS2 Report 28. JAMA Ophthalmology 2022.[140]
- Singh S et al. Blue-light filtering spectacle lenses for visual performance, sleep, and macular health in adults. Cochrane Database of Systematic Reviews 2023.[141]
- Global estimates on the number of people blind or visually impaired by cataract — meta-analysis 2000–2020. Eye 2024.[142]
- Harwood RH, Foss AJE et al. Falls and health status in elderly women following first eye cataract surgery: a randomised controlled trial. British Journal of Ophthalmology 2005.[143]
- Foss AJE et al. Falls and health status in elderly women following second eye cataract surgery: a randomised controlled trial. Age and Ageing 2006.[144]
- Lee CS et al. Association between cataract extraction and development of dementia. JAMA Internal Medicine 2022.[145]
- Gazzard G et al. Selective laser trabeculoplasty versus eye drops for first-line treatment of ocular hypertension and glaucoma (LiGHT): a multicentre randomised controlled trial. Lancet 2019.[146]
- Gazzard G et al. Laser in Glaucoma and Ocular Hypertension (LiGHT) Trial: six-year results of primary selective laser trabeculoplasty versus eye drops. Ophthalmology 2023.[147]
- US Preventive Services Task Force. Screening for primary open-angle glaucoma: recommendation statement. JAMA 2022.[148]
- Farrand KF et al. Prevalence of diagnosed dry eye disease in the United States among adults aged 18 years and older. American Journal of Ophthalmology 2017.[149]
- Asbell PA et al. n-3 fatty acid supplementation for the treatment of dry eye disease (DREAM). New England Journal of Medicine 2018.[150]
- He M et al. Effect of time spent outdoors at school on the development of myopia among children in China: a randomized clinical trial. JAMA 2015.[151]
- Grewal MK et al. A pilot study evaluating the effects of 670 nm photobiomodulation in healthy ageing and age-related macular degeneration. Journal of Clinical Medicine 2020.[152]