Open questions — cataracts¶
Last curated: 2026-08-31 (independent audit pass). Every question below states why it is open with citations to the evidence that leaves it open, and what a study would have to do to close it. IDs are stable across sweeps. A question is retired only when a live search finds directly answering evidence with the relevant population, comparator and patient-important endpoint; a proxy endpoint or a single-centre feasibility result sharpens a question but does not close it.
Tier 1 — would change practice, and designable today¶
OQ-1 — Would a PROM-based surgical indication outperform an acuity threshold?¶
Why open. Visual acuity is a demonstrably poor gauge of cataract disability: the VF-14 correlated better with patients' own rating of trouble with their vision than any acuity measure and was ~3× more responsive to surgery than a generic health instrument (Steinberg 1994, PMID 8185520; Cassard 1995, PMID 7487617). Rasch-scaled instruments now exist that are brief enough for routine use, with person reliability 0.90 (Cat-PROM5) and 0.88 (Catquest-9SF) and responsiveness 1.45 and 1.47 SD (Sparrow 2018, PMID 29386619). NICE already instructs services not to restrict access on acuity (NICE NG77 recommendation 1.2.2, accessed 2026-08-31), and prototype appropriateness-and-prioritisation tools have been described (Schlenker 2023, PMID 36028007) — but no health system has published a prospective comparison of a PROM-based rule against an acuity threshold with access, outcome, equity and cost endpoints.
What would close it. A stepped-wedge or cluster-randomised implementation trial across services, randomising the referral rule, with eCSC-style outcome capture and equity analysis by sex, education and rurality.
OQ-2 — What intervention could actually deliver the 2030 effective-coverage target?¶
Why open. The 74th World Health Assembly set a 30-percentage-point eCSC increase by 2030 (Keel 2021, PMID 34237266). Modelled from 233 population-based datasets in 68 countries, global eCSC was 48.2% (39.7–57.2) in 2025 and is predicted to rise only 8.4 points (8.1–8.6) between 2020 and 2030 (McCormick 2026, PMID 41687671). Implementation research in Kenya and Nepal found outcome monitoring sporadic and rarely used in planning, with public–private partnerships weakened by fragmented financing and out-of-pocket costs (Arazi 2026, PMID 41735007). No country-level intervention study demonstrating an eCSC gain of the required magnitude was identified.
What would close it. A prospectively evaluated national or sub-national programme with pre/post eCSC measured on comparable RAAB-style surveys and an explicit theory of change.
OQ-3 — Is postoperative refraction and spectacle provision the cheapest eCSC lever available?¶
Why open. Uncorrected refractive error accounted for a median 26.4% of non-good postoperative outcomes per survey, and correcting it was estimated to yield a median 3.7-percentage-point gain in eCSC at the 6/12 threshold (PMID 41687671). No trial of routine postoperative refraction and spectacle supply with an effective-coverage endpoint was identified.
What would close it. A cluster-randomised trial of routine post-surgical refraction plus spectacle provision versus usual care, with eCSC and PROM endpoints.
OQ-4 — Which intracameral antibiotic, and at what resistance cost?¶
Why open. Only cefuroxime and moxifloxacin have randomised support (ESCRS Endophthalmitis Study Group 2007, PMID 17531690); the 6.8-million-eye network meta-analysis ranks vancomycin (OR 0.03) and cefazolin (0.09) above cefuroxime (0.18) and moxifloxacin (0.36) on observational data alone, and only the intracameral route achieved significance (OR 0.19, 99.4% CI 0.12–0.30) (Kato 2022, PMID 36258003). Vancomycin carries haemorrhagic occlusive retinal vasculitis — 36 eyes of 23 patients, all vancomycin-exposed, with usually poor visual outcomes (Witkin 2017, PMID 28110950) — and its use has fallen to 6% of surveyed US surgeons (Lieu 2024, PMID 37877364). About 2,500 patients need treatment to prevent one case (PMID 37877364).
What would close it. A registry-embedded pragmatic randomised comparison, or a federated multi-registry analysis with resistance surveillance built in.
OQ-5 — Is immediate sequential bilateral surgery safe with respect to bilateral rare harm?¶
Why open. BICAT-NL established refractive non-inferiority (97% vs 98% of second eyes within ±1.0 D; difference −1%, 90% CI −3 to 1) with €403 lower societal cost and zero endophthalmitis in 865 randomised patients (Spekreijse 2023, PMID 37201546). Zero events cannot bound bilateral risk when the unilateral rate is 0.066% (PMID 36258003); the update literature explicitly calls for randomised registry studies powered for bilateral endophthalmitis (Spekreijse 2023, PMID 36206058).
What would close it. A federated analysis across national registries with pre-specified bilateral-event definitions, or a registry-randomised design.
OQ-6 — Does steroid rather than NSAID prophylaxis reduce posterior capsule opacification?¶
Why open. A 13,368-patient registry analysis found lower Nd:YAG capsulotomy rates with steroid than NSAID monotherapy after adjustment (HR 0.70, 95% CI 0.52–0.88, P = .001), with no added benefit from combination therapy (HR 1.11, 0.68–1.80) (Hecht 2020, PMID 32061757). The Cochrane synthesis of 66 studies found no difference between anti-inflammatory regimens except for an experimental immunotoxin (Findl 2010, PMID 20166069). The conflict is unresolved and no randomised trial has used capsulotomy as a prespecified endpoint.
What would close it. A randomised comparison of postoperative steroid versus NSAID with Nd:YAG capsulotomy at 3 years as the primary endpoint.
OQ-7 — Which anti-inflammatory regimen prevents clinically important macular oedema rather than OCT-only thickening?¶
Why open. Clinical pseudophakic cystoid macular oedema occurs in 1–2% of eyes 2–10 weeks after surgery, while angiographic and OCT changes without functional consequence are far commoner, and the natural course is often benign with spontaneous recovery over months (Radeck 2020, PMID 32468102); reported clinical incidence spans 0.1–2.35% (Zur 2017, PMID 28351047). Agent choice is unsettled: a network meta-analysis of 11 RCTs and 2,175 subjects found bromfenac significantly better than nepafenac 0.1% for one-month visual acuity (p < 0.001) but no significant differences in foveal thickness or IOP (Almasri 2024, PMID 39095467). Large systematic reviews reach contradictory conclusions specifically because PCMO is defined and resolved differently across studies (Han 2019, PMID 30953417), and high-quality trials are lacking for periocular and intravitreal steroid, bevacizumab and vitrectomy.
What would close it. A trial using a core outcome set with a prespecified, functionally anchored definition of clinically important macular oedema — not foveal thickness alone — as the primary endpoint.
OQ-8 — Which infants, if any, should receive a primary intraocular lens?¶
Why open. IATS found no visual difference at 10.5 years (median 0.89 vs 0.86 logMAR, P = 0.82) with far more surgery in the IOL arm (adverse events 81% vs 56%; additional intraocular operations 72% vs 16%) (Lambert 2020, PMID 32077909; Plager 2014, PMID 25077835). The trial was not powered to identify subgroups, and 43% of children randomised to aphakia had received a secondary IOL by 10.5 years anyway (Drews-Botsch 2026, PMID 41962549).
What would close it. A prospective study testing a prespecified selection rule (corneal diameter, anterior chamber depth, family capacity for contact-lens care) against default aphakia.
OQ-9 — Can aphakic glaucoma after infant cataract surgery be prevented, not merely predicted?¶
Why open. Risk rises with follow-up — glaucoma 9% at 1 year, 17% at 5 and 22% at 10 years; glaucoma-or-suspect 12%, 31% and 40% — and is unrelated to IOL status (Freedman 2021, PMID 33331850). Shallow preoperative anterior chamber depth predicts it (2.76 ± 0.48 vs 3.08 ± 0.38 mm; OR 5.8, 95% CI 1.8–18.9) (Wong 2026, PMID 41419074). Earlier surgery protects against amblyopia but is itself the strongest risk factor for aphakic glaucoma (Kuhli-Hattenbach 2020, PMID 32076840). No intervention has been tested for reducing incidence.
What would close it. A trial of a modifiable factor — surgical timing window, technique, or an IOP-surveillance and early-treatment protocol — with 10-year glaucoma incidence as the endpoint.
OQ-10 — Should intraocular lens formula recommendations in national guidance be updated?¶
Why open. NICE prescribes Haigis or Hoffer Q below 22.00 mm, Barrett Universal II (or SRK/T) between 22.00 and 26.00 mm, and Haigis or SRK/T above 26.00 mm (NICE NG77 recommendation 1.3.5, accessed 2026-08-31), and the May 2025 surveillance decision was explicitly not to update this section. Contemporary network meta-analyses in the same axial-length bands rank Kane, Olsen, EVO and Hill-RBF above several of these — though usually without statistical significance (Zhang 2026, PMID 40971912; Ma 2024, PMID 37726043; Li 2026, PMID 41651446).
What would close it. A large multicentre registry study with standardised constants and devices, as the authors of both major network meta-analyses request.
OQ-11 — Who benefits enough from presbyopia-correcting optics to accept the photic trade-off?¶
Why open. Trifocals give better uncorrected near acuity than monofocals (MD −0.32 logMAR, 95% CrI −0.46 to −0.19) and EDOF better intermediate acuity, with distance comparable (Cho 2022, PMID 36136323; Li 2024, PMID 38627651); the same network meta-analysis found no statistically significant differences in contrast, glare or halos, while the review literature consistently reports them (Sieburth 2019, PMID 30993062) and dysphotopsia affects up to 67% early and 2.2% persistently at one year (Pusnik 2022, PMID 36676002). Objective screening cut-offs (age ≤62, ocular scatter index ≤1.25, dysfunctional lens index ≥7.67) exist only retrospectively (Fernández 2023, PMID 36871115). NICE meanwhile instructs the NHS not to offer multifocal IOLs at all (recommendation 1.4.2).
What would close it. A randomised trial with a validated dysphotopsia and contrast instrument as a co-primary endpoint alongside spectacle independence, and prespecified screening criteria.
Tier 2 — needs enabling measurement or mechanism work first¶
OQ-12 — Can a drug be shown to reach the human lens nucleus?¶
Aggregation reversal is established in vitro and in animal lenses (Zhao 2015, PMID 26200341; Xu 2020, PMID 33313297), and delivery systems are in preclinical development (Zhang 2026, PMID 41323208; Jiang 2025, PMID 40336002). As of 2026-08-31 no published human pharmacokinetic study demonstrating lens-nucleus penetration at an active concentration was identified. The single registered interventional cataract-progression study is a 30-participant phase 1/2 implant trial in post-vitrectomy eyes (NCT05026632).
OQ-13 — Is there a non-invasive human marker of lens protein state?¶
The central quantitative claim about nuclear cataract is that oxidation begins once nuclear glutathione falls below roughly 2 mM (Truscott 2005, PMID 15862178), yet oxidative markers are measured in tissue and animal models, not in living human eyes (Lee 2024, PMID 37882550). Without such a marker no preventive trial can demonstrate target engagement, and phase 2 must use multi-year opacity grading — which is what made AREDS large and slow (AREDS Research Group 2001, PMID 11594943).
OQ-14 — Is the cortex–nucleus transport barrier modifiable?¶
The barrier that develops in the fourth decade is the proposed proximate cause of nuclear cataract and simultaneously obstructs antioxidant delivery to the compartment that needs it (Truscott 2000, PMID 10971179; Braakhuis 2019, PMID 31137834). No study has attempted to measure or shift it in human lenses.
OQ-15 — Can grading systems be cross-calibrated?¶
LOCS III, the decimalised Oxford system, the WHO simplified system, Scheimpflug and AS-OCT metrics and deep-learning graders are all in use (Chylack 1993, PMID 8512486; Sparrow 2000, PMID 10652171; Thylefors 2002, PMID 11821974; Mackenbrock 2024, PMID 38242135). No bidirectional conversion functions with prediction intervals have been published, so burden and progression estimates pooled across systems carry an unquantified error. Nor has a minimum clinically important difference in LOCS III units ever been established, despite ≥0.5-unit change occurring in 54% of one cohort within a year (Srinivasan 1997, PMID 9486033).
OQ-16 — What risk-adjustment model should cataract registries use?¶
Predictors of poor acuity outcome are known — age, short axial length, ocular comorbidity, AMD, diabetic retinopathy, amblyopia, corneal pathology, previous vitrectomy, and intraoperative PCR (OR 5.74) (Sparrow 2012, PMID 22441022) — and surgeon-level PCR variation is greatest where case numbers are smallest, with the authors stating that acceptable benchmarks cannot be set without case-mix adjustment (Johnston 2010, PMID 19680280). No validated adjustment model is in shared use across EUREQUO or national audits (Lundström 2012, PMID 22541829).
OQ-17 — Is there a cataract PROM with demonstrated cross-cultural measurement invariance?¶
Catquest-9SF is unidimensional with acceptable precision across 13 studies and 15,289 participants, but 9 of 13 showed significant mistargeting and the authors recommend revalidation in every new population (Kabanovski 2020, PMID 31862206). The minimum important difference differs by baseline function (−0.5 vs −1.80 logits) (Grimfors 2022, PMID 36494767), so a single service-wide responder threshold is not currently defensible.
OQ-18 — How can rare and bilateral harms be estimated without misleading zero-event conclusions?¶
Endophthalmitis at 0.066% (PMID 36258003), dropped nucleus at 0.071% (Lundström 2020, PMID 32126043) and late IOL dislocation at 1.7% cumulative at 25 years (Pueringer 2011, PMID 21683329) are all below the resolution of conventional trials. Registry federation is the proposed answer but no methodological standard for it exists in this literature.
OQ-19 — What decomposes the female excess in cataract blindness?¶
Women bear 60% of cataract blindness and 59% of MSVI (Vision Loss Expert Group 2024, PMID 38461217), have higher age-adjusted incidence of nuclear, cortical and any cataract (Kanthan 2008, PMID 17900695), and have lower effective surgical coverage (RR 1.20, 95% CI 1.15–1.25) (McCormick 2022, PMID 36240806). No analysis separates longevity, incidence and access contributions.
OQ-20 — Does cataract surgery cause diabetic retinopathy progression, or reveal it?¶
Meta-analysis of 15 studies (7,287 patients) found progression RR 1.46 (95% CI 1.28–1.66) but no significant effect in paired-eye studies (RR 0.85, 0.39–1.83) (Lee 2025, PMID 39179126); propensity-matched US data show 1-year NPDR→PDR hazard ratios of 1.45–1.58 (Loya 2025, PMID 39956206). The paired design is the one that controls systemic confounding and its confidence interval is too wide to settle the question.
OQ-21 — What is the contemporary incidence of posterior capsule opacification?¶
The pooled 1-, 3- and 5-year estimates of 11.8%, 20.7% and 28.4% date from 1998 and predate universal sharp-edge optic design (Schaumberg 1998, PMID 9663224). Modern figures span 10% to 30% depending entirely on endpoint definition (Konopińska 2021, PMID 34199147; Milazzo 2014, PMID 25455552; Gu 2022, PMID 34727350). No updated meta-analysis restricted to modern sharp-edged hydrophobic IOLs was identified.
OQ-22 — What is the true clinical burden of Nd:YAG-induced IOL damage?¶
A systematic review of 53 publications found morphological damage from micro-pits to crater formation and delamination, measurable optical degradation after central pit formation, and symptomatic damage requiring explantation in up to 9.3% of retrospectively analysed cases — while stating explicitly that robust clinical data linking laboratory findings to patient outcomes are scarce and calling for large prospective studies, especially in premium IOLs (Borkenstein 2025, PMID 40862048).
OQ-23 — Would occupational or recreational UV protection reduce cataract incidence?¶
The exposure–response is quantified (doubling of cumulative UV-B: OR 1.60, 95% CI 1.01–2.64 for cortical cataract) (Taylor 1988, PMID 3185661), the causal relationship is supported by animal and in vitro work (Löfgren 2017, PMID 27260484), the intervention is cheap and specific (wraparound sunglasses blocking below 400 nm) (Roberts 2011, PMID 21617534), and a WHO/ILO burden estimate is in development (Tenkate 2019, PMID 30737039). No randomised or quasi-experimental evaluation with a cataract endpoint was identified.
Dots not yet connected¶
Cross-domain junctions where two bodies of evidence in this knowledge base both exist, but no study has joined them.
| # | Dot A | Dot B | The missing junction | Powers |
|---|---|---|---|---|
| 1 | Uncorrected refractive error explains a median 26.4% of non-good outcomes (PMID 41687671) | Cataract surgery is $9–$1,600/QALY in developing countries (PMID 17383730) | Cost-effectiveness of bundling postoperative refraction and spectacles into the surgical episode | Programme design and financing |
| 2 | Formula-specific prediction error in dioptric bands (PMID 37726043) | Catquest-9SF minimum important difference by baseline function (PMID 36494767) | Mapping dioptres of prediction error onto PROM change patients notice | IOL counselling and formula choice |
| 3 | Female excess in cataract blindness (PMID 38461217) | eCSC sex gap RR 1.20 (PMID 36240806) and girls under-operated for bilateral childhood cataract (PMID 26992842) | Decomposition of the female burden into incidence, longevity and access components | Equity policy and outreach design |
| 4 | Lens epithelial cell senescence drives delayed PCO, reversible by dasatinib+quercetin in models (PMID 40660668) | Epithelial injury opens the lens to oxygen and drives cataractogenesis (PMID 35508906) | Whether one senolytic strategy could address both cataract formation and PCO | Drug development strategy |
| 5 | Sharp optic edge cuts 3-year Nd:YAG rate by RR 0.21 (PMID 34398965) | Nd:YAG damages IOLs, with explantation in up to 9.3% of analysed damaged cases (PMID 40862048) | Lifetime system cost and harm model of lens choice including capsulotomy sequelae | Procurement and lens selection |
| 6 | Cataract surgery associated with DR progression RR 1.46 (PMID 39179126) | Multifocal optics discouraged in diabetic macular disease (PMID 41343850) | Prospective lens-selection strategy in diabetic eyes with retinopathy-progression and PROM endpoints | Lens selection in diabetes |
| 7 | Zero endophthalmitis in 865 randomised ISBCS patients (PMID 37201546) | Endophthalmitis base rate 0.066% (PMID 36258003) | Federated registry methodology for bounding bilateral rare-event risk | Sequencing policy |
| 8 | Waiting deteriorates acuity and quality of life (PMID 11778807; PMID 17322459) | Falls fall from 1.17 to 0.81 per year after first-eye surgery (PMID 35702892) | Prioritisation by preventable functional loss rather than acuity | Waiting-list design |
| 9 | Preoperative anterior chamber depth predicts 10-year aphakic glaucoma (OR 5.8) (PMID 41419074) | Myopic shift after infant IOL is 0.35 D/month to age 1.5 years (PMID 28215452) | An infant biometric model integrating glaucoma risk and refractive trajectory | Paediatric surgical strategy |
| 10 | Resident complication rate halves after ~40 cases (PMID 21107259) | eCSC must rise 30 points by 2030 (PMID 34237266; PMID 41687671) | Scale-up designs that protect quality during rapid volume growth | Workforce planning |
| 11 | Registry anaesthesia associations: topical carries higher PCR and endophthalmitis risk (PMID 36449673) | Randomised evidence shows only small pain differences between topical and sub-Tenon (PMID 26308931) | A trial powered for complications rather than pain, or a case-mix-adjusted registry analysis | Anaesthetic policy |
| 12 | Marketed cataract-dissolving drops have no reliable evidence (PMID 28245346) | Variable quality of cataract information on short-video platforms (PMID 40755956) | Measurement of delayed presentation and worse outcomes attributable to unvalidated products | Consumer safety and public information |
Maintenance rule¶
At each sweep: re-run the search behind every asserted absence on this page (stale absences are the commonest real error), re-check any registry status quoted, and either retire a question with the citation that answers it or record the date on which the absence was reconfirmed.