The domain opener for age-related eye disease (cataract · AMD). Cataract is the leading global cause of blindness — «Cataract is the leading cause of blindness and impaired vision globally and accounts for 50% of all global blind ness» — but unlike AMD it is typically reversible by surgery, so the patient-important outcome is vision/function conditional on surgical access, not irreversible sight loss. (Aune et al., 2026)
The physical-activity lever — WCRF probable
Aune 2026 (SR-MA of prospective cohorts, gold-tier) is the only source held. It is the stronger arm of the paper: physical activity → lower cataract risk graded probably causal on WCRF criteria (the same graded-causality matrix as Diet Physical Activity and Cancer Prevention). (Aune et al., 2026)
- High vs low PA: RR 0.90 (95% CI 0.86–0.94), I2 = 74%, 10 cohorts / 8 publications, 163,065 cases, 1,914,137 participants — a 10% relative risk reduction. Excluding two outlying studies cut heterogeneity to 19% with the estimate essentially unchanged (RR 0.89, 0.87–0.92). No publication bias (Egger p = 0.39). (Aune et al., 2026)
- Dose-response: RR 0.91 per 20 MET-hours/week (0.84–0.99, I2 = 66%, n = 3), monotone over the studied range — no nonlinearity (pnonlinearity = 0.32); the authors report a 25% risk reduction at 50 MET-hours/week. The shape claim is weak (only 3 cohorts had usable dose data; the studied range is not stated in the extracted text — read the 25%-at-50 figure as a within-range point, not an optimum). (Aune et al., 2026)
- By PA type: running RR 0.70 (0.60–0.82, n = 2) > total PA 0.89 (0.83–0.95, n = 2) > leisure-time 0.95 (0.93–0.97, n = 2). The running estimate rests on 2 self-selected runner cohorts (National Runners’/Walkers’ Health Study) — a healthier-than-average, more-intense stratum; do not read the 0.70 as the general dose. (Aune et al., 2026)
Studied populations: cataract follow-up 7–12.1 y; mostly Europe + US; modified Newcastle-Ottawa quality mean 5.7 (median 5.9) / 8 (moderately high). The main quality gaps were unvalidated PA assessment and unclear exclusion of prevalent cases at baseline. (Aune et al., 2026)
Why this is an association discounted for healthy-user confounding, not a demonstrated lever
This is observational cohort evidence. The binding threat is confounding by the healthy-active-person profile — physical activity travels with not-smoking, better diet, and higher SES, each plausibly protective for the lens. Three facts bound how much of the 10% is causal:
- The E-value is modest. «The E-value for the association between physical activity and cataract was 1.46 (lower CI: 1.31), suggesting an unadjusted risk fac tor would need to have a moderate association with both physical activity and cataract risk to fully explain away the observed association.» A moderate unmeasured confounder suffices — so the association is not robust to strong confounding, consistent with a small true effect partly inflated by healthy-user selection. (Aune et al., 2026)
- Confounder adjustment was real but incomplete. Most cohorts adjusted BMI, smoking, alcohol, and hormone therapy; the overall association survived subgroup analysis stratified by adjustment. Only one study adjusted for sun exposure (a lens-specific risk factor) and still found a reduced risk. (Aune et al., 2026)
- Measurement error runs toward the null here, not away from it — «Measurement errors in both the exposure and outcome may have affected the results, however, given the cohort design of the included studies, any bias would most likely have attenuated the associations toward the null.» So the self-reported-PA measurement error (the domain’s binding constraint, Measurement Error in Dietary Assessment) argues the true effect is at least as large as measured, not smaller — a point pulling the other way from the confounding discount. (Aune et al., 2026)
The U/J-artifact lens (The U-Shaped Association Artifact) does not apply — the contrast is high-vs-low with a monotone dose-response and all component RRs below 1.00; there is no protective-arm reversal to interrogate.
Mechanism — a mediator chain, only partly cashed
Aune’s stated pathway is metabolic-mediation: PA improves weight control, lowers type-2-diabetes risk (diabetes carries a ~3-fold cataract risk), and lowers blood pressure — and adiposity, diabetes, and hypertension are each cataract risk factors; PA may also reduce ocular oxidative stress, inflammation, and advanced-glycation-end-product formation. (Aune et al., 2026)
- The adiposity channel is treated as a mediator, and the mediated fraction is not quantified. «It is possible that certain fac tors that were adjusted for in some studies, including adi posity, hypertension and diabetes, may be on the causal pathway between physical activity and cataract, and thus represent overadjustments. Further studies should investigate whether these factors may mediate part of the observed associations.» So the source flags that studies adjusting for adiposity/HTN/ diabetes may be over-adjusting away part of a real PA effect — but it does not put a number on the mediated share, and calls for further work. The oxidative-stress channel is-grade mechanism, not an outcome finding. (Aune et al., 2026)
Established (non-PA) risk factors — context for stratification
Aune lists the established cataract risk factors as diabetes, higher BMI, hypertension, smoking, ultraviolet light, and steroid use. Several are themselves modifiable big rocks already held elsewhere in the fabric (glycaemic control, weight, blood pressure, smoking cessation), which is the route-(a) point: absolute cataract benefit from any lever scales with baseline risk, so it is largest in the diabetic / high-BMI / smoking stratum even though the relative PA effect is modest. (Aune et al., 2026)
Decision relevance (Layer 1)
Vision loss and its prevention are patient-important function/QoL outcomes on the one health axis (independence, quality of life). But physical activity is already a big rock by construction — non-substitutable for its pleiotropy (Physical Activity Dose and Mortality) — so a vision-specific protective arm adds to PA’s case rather than creating a new lever. For a reasonably-healthy person the decision be physically active is already settled on mortality and cardiometabolic grounds; the cataract arm is confirmatory and marginal, not decision-turning. It bites hardest, via route (a), where cataract baseline risk is high (diabetes, high BMI) and where a person is weighing activity specifically for eye health. Honest summary: modest (RR 0.90), observational, WCRF-probable — a real but small confirmatory arm on an already-pulled lever.
Held threads / gaps
- Single-source, observational. Aune 2026 is the sole anchor; no RCT, no Mendelian-randomization, and no objective (accelerometer) PA measure is held for this outcome — all are named gaps.
- A diet arm for cataract is not held. Aune notes fruit-and-vegetable adjustment materially moved the PA estimate (between-subgroup pheterogeneity = 0.02), hinting a diet lever exists — a directional gap, no SR held.