The emergent claim. Assemble the modifiable-risk-factor lists the wiki holds for four leading age-related outcomes — dementia, cardiovascular disease, cancer, Parkinson’s — and a small set of exposures recurs across most of them in the same direction: adiposity, physical (in)activity, diabetes/glycaemia, blood pressure, LDL, smoking, excessive alcohol. These are the same big rocks the cardiometabolic cluster already ranks. No single source says this — each disease body is disease-specific by construction — so the synthesis is the overlap itself.
What it changes about what to do. A person does not need a separate anti-dementia diet, an anti-cancer diet, and an anti-heart-disease diet. Because the levers are shared, pulling the big rocks once buys risk reduction on several patient-important outcomes simultaneously — which strengthens their Layer-1 ranking rather than adding competing tasks: a lever that moves three outcomes outranks one that moves one at equal per-outcome effect -> Layer 1 - Ranking Interventions for a Stratum. The dementia body reasons the same way locally — pulling the shared levers buys a second patient-important outcome (cognition) on top of CV events. (inferred from Livingston et al., 2024) This page lifts that from a footnote on one disease page to a cross-disease structural claim.
The frame compounds with time, not only across diseases. The shared core is midlife-onset with long latency — in the dementia body’s life-course table, LDL, blood pressure, adiposity, smoking and diabetes are midlife factors driving late-life incidence -> Dementia Prevention and Modifiable Risk Factors — so a 10-year, single-disease risk calculator understates these levers on both axes: neither the multi-outcome breadth above nor the decades of compounding exposure. The decision counterpart is the lifetime-benefit frame — act early despite a low 10-year number -> Lifetime Benefit - The Frame for Younger Adults. Its logic transports beyond CVD (the qualitative case for early action); its CVD-specific LIFE-CVD model and CVD-free-years currency do not, and the PAFs still do not sum (caveat below). The ESC below-40 “lifestyle over drugs” bound transports too — but whether it strengthens off-CVD is open, since the shared-substrate hypothesis below implies the CVD preventive drugs (statins, antihypertensives, glycaemic control) may themselves be dementia levers, not a no-drug case.
The parameter table — matched exposure, per-disease direction
The matched parameter is the exposure and the direction of its association, NOT the effect
magnitude: each column’s RR is for a different outcome (dementia vs PD vs cancer incidence vs CV
events), estimated in different populations, so the numbers are co-membership evidence, not a
commensurable magnitude — do not read across a row as a same-size effect. Blank = not a credibly
evidenced lever for that disease in the sources held (a G-gap, not a null).
| Exposure | Dementia (Livingston, RR/HR) | Parkinson’s (Chen, RR) | Cancer (WCRF) | CVD (vault) | Direction shared? |
|---|---|---|---|---|---|
| Physical inactivity | activity 0.80 (0.77-0.84) protective | activity 0.79 (0.68-0.91) protective | activity a protective cause | established lever | yes — all four |
| Adiposity / obesity | midlife 1.31 (1.02-1.68) | — (not credible) | convincing/probable cause, 12 of 17 sites | established lever | yes (3 of 4) |
| Diabetes / glycaemia | HR 1.24 per 5-yr earlier onset | 1.31 (1.10-1.57) | (via adiposity) | established lever | yes |
| High blood pressure | untreated 1.42 (1.15-1.76) | — | — | established lever | yes (dementia + CVD) |
| High LDL | >3 mmol/L HR 1.33 | — | — | established lever | yes (dementia + CVD) |
| Smoking | midlife 1.30 (1.18-1.45) harm | 0.64 (0.60-0.69) protective | a cause (not WCRF-graded — see note) | established harm | NO — flips in PD |
| Excess alcohol | >21 u/wk 1.22 (1.01-1.48) | — | a cause | J-curve, arm artifact-suspect | yes (dementia + cancer) |
| Coffee / caffeine | — | 0.55-0.67 protective | — | — | PD-only signal |
(Livingston et al., 2024) (Chen et al., 2021) (World Cancer Research Fund & American Institute for Cancer Research, 2018)
Note on the smoking/cancer cell. Smoking is a cause of cancer, but the WCRF Continuous Update Project cited here does not grade tobacco — its scope is diet, nutrition, activity and body fatness, so it only acknowledges smoking as background, never as a graded exposure. The cell records the well-established external fact, not a WCRF finding; the WCRF-graded cancer levers in this table are body fatness, alcohol and physical activity.
Physical activity now reaches a fifth disease family — age-related eye disease — but disease-specifically. PA → cataract is graded WCRF probable (RR 0.90, 0.86–0.94) — another protective cell for the workhorse lever, on the same WCRF matrix as the cancer column (Cataract and Modifiable Exposures). But its stablemate PA → AMD did not reach significance (RR 0.92, CI 0.84–1.01 crosses 1; graded limited, no conclusion — Age-Related Macular Degeneration and Modifiable Exposures). So even within one organ the shared lever does not transfer uniformly — a within-family instance of the same “the logic breaks disease-specifically” caution the Parkinson’s exception makes below. Both estimates are observational (Aune/Jayedi lab, F-not-E vs the other Aune MAs), discounted for healthy-user confounding (cataract E-value 1.46). (Aune et al., 2026)
The lever-specificity cuts the other way too. Where PA fails on AMD, a different lever — the Mediterranean-diet pattern — reaches significance for AMD progression (prospective-cohort HR 0.77, 0.67–0.88; case–control OR 0.66, 0.54–0.81; cross-sectional null), though on observational, gold-by-design-only evidence discounted for healthy-user confounding (Age-Related Macular Degeneration and Modifiable Exposures). These are a different exposure than PA, so this is a distinction, not a tension — but it sharpens the caution: “shared levers” is doubly disease-and-lever-specific — not every big-rock lever reaches every disease, and the lever that reaches one disease need not be the one that reaches its organ-neighbour. (Marques-Couto et al., 2025)
And the classic vascular big rocks do reach AMD — in the expected direction. Babaker 2025 (18 observational studies) finds AMD occurrence significantly associated with smoking (OR 1.86), hypertension (1.24), CVD (1.44), and diabetes (1.44), while BMI, stroke and blood lipids are NS (Age-Related Macular Degeneration and Modifiable Exposures). So the AMD column is not uniformly lever-poor: it is PA specifically that fails there, whereas the vasculature-injuring big rocks associate as they do across the dementia and cardiometabolic families — consistent with the shared vascular substrate injured by the same exposures reading below. Two discounts keep this an F/refinement and not a strong claim: these are route-(a) prognostic associations (who develops AMD), not demonstrated levers, and they come from a crude-and-adjusted-mixed observational pool with no absolute risk — so they reinforce the ranking already set by each rock’s own outcomes rather than opening an AMD-specific lever. (Babaker et al., 2025)
Where the logic breaks — Parkinson’s is the exception, and it is instructive
Parkinson’s does not obey the shared-lever pattern, and reading it as if it did would misdirect action:
- Only two shared levers hold — physical activity (RR 0.79, PD’s strongest and most design-robust signal) and, modestly, diabetes (1.31). Adiposity, blood pressure and LDL were assessed and found non-significant for PD (not merely unstudied), and alcohol is not a credible PD lever — so pulling the big rocks does far less for PD than for dementia/CVD/cancer, and does it through a narrower set.
- The direction flips on smoking and coffee, and the flip is an artifact, not a lever. PD’s strongest protective associations — smoking 0.64, caffeine 0.55 — run opposite to every other disease, and the source attributes them to reverse causation over PD’s long prodrome (patients quit more easily pre-diagnosis), not neuroprotection: «its protective effect for PD development does not impact public health strategies for reducing smoking». (Chen et al., 2021) So the one place the shared-lever logic would license a harmful action (smoke to protect the brain) is exactly where the signal is an artifact -> The U-Shaped Association Artifact. The lesson generalizes: a disease-specific protective exposure that contradicts a well-established harm is a prompt to run the artifact check, never to act.
The disease-specific levers — what does NOT transfer
Sharing a core does not make the diseases interchangeable. The non-shared levers are where disease-specific effort still pays:
- Dementia-specific: hearing loss, uncorrected vision, education / cognitive reserve, depression, traumatic-brain-injury protection, social contact, air pollution -> Dementia Prevention and Modifiable Risk Factors.
- Cancer-specific (diet-adjacent): red and processed meat, dietary fibre (protective), plus the non-diet bulk (tobacco, infections, occupational carcinogens) -> Diet Physical Activity and Cancer Prevention, Red and Processed Meat and Cancer.
- Parkinson’s-specific: the coffee/tea/urate signals (weak, mostly artifact-suspect) and the case-control-only pesticide/occupational signal -> Parkinsons Disease Modifiable Risk Factors.
The unifying mechanism — a candidate, marked as inference
Why do the same exposures recur? The most parsimonious reading is that the shared core is the set that damages a common cardiometabolic-vascular substrate: hypertension, diabetes, adiposity, LDL and smoking all injure the vasculature and drive systemic metabolic/inflammatory dysfunction, and vascular contribution to dementia is established while adiposity’s cancer link runs through metabolic/inflammatory pathways. On this reading the overlap is not coincidence — several age-related diseases share an upstream substrate, so the exposures that degrade it are levers for all of them. This is a mechanism hypothesis, not an outcome finding: it is corroborated by the co-membership above but not independently tested here, and it does not privilege any single exposure as the cause (an effect arrives only when its whole condition set is present) -> Layer 1 - Ranking Interventions for a Stratum.
How to read this — the caveats that bind
- The magnitudes are population-attributable, not per-person. The dementia PAF (45.3%) and WCRF’s preventable-fraction estimates are modelled population quantities; removing a factor does not remove that share of one person’s risk, and the factors overlap so the PAFs do not sum -> Dementia Prevention and Modifiable Risk Factors.
- The relative risks are borrowed from observational data and assumed causal for most of these factors; the shared-lever claim inherits that uncertainty. Physical activity is the most robust leg (consistent direction across all four, dose-response, some trial support); smoking-in-PD is the least (a named artifact).
- The loop is open. No operation here has graded the shared-lever recommendation against a realized multi-disease outcome. The synthesis says a well-informed advisor would prioritize the shared core for its multi-outcome coverage; whether doing so lowered any individual’s combined risk is a validity fact the wiki cannot see.
Extension — the same two workhorse levers preserve FUNCTION, not just prevent disease incidence
The table above is built from disease-incidence bodies. The wiki now also holds musculoskeletal / quality-of-life outcomes — Frailty, Knee Osteoarthritis and Modifiable Levers, Chronic Pain and Physical Activity — and the workhorse levers recur on this axis too, in the same direction. Physical activity is the genuinely cross-cutting one: it improves frailty status, mobility and ADLs and is the safe function lever across arthritis and chronic pain — it recurs on all three function pages and all four incidence rows. Adiposity / weight loss is narrower on the function axis: it is an evidenced lever for knee OA specifically (reduces joint load, inflammation, pain), not a general function lever — and it is NOT a frailty lever (weight loss is itself a Fried frailty criterion and can harm the frail), so its breadth here is knee-OA-via-overweight, whereas on the incidence axis it is broad (dementia/cancer/CVD). Knee OA now sits on BOTH axes, and the two carry different evidence weight — the distinction is the decision-relevant part: obesity is the dominant knee-OA incidence lever too (it carries the large majority — ~77% — of the radiographic-KOA attributable fraction -> Knee Osteoarthritis Incidence and Risk Factors) and sits on the same adiposity machinery as the cardiometabolic levers (so this is NOT a distinct pathway — it is consistent with the shared-substrate reading below, not a counter-case). But that incidence figure is prognostic / route-(a): no held evidence shows that losing weight in a normal-BMI-trajectory person prevents incident OA, whereas the function/symptom benefit has a causal anchor (IDEA, in the already-symptomatic). So for knee OA the proven weight-loss lever is on the FUNCTION axis; the incidence axis is prognostically-implied only. This is co-membership evidence again, NOT a commensurable magnitude — the outcomes here are frailty status, WOMAC pain and physical function, not disease incidence, so do not read a shared lever as an equal-size effect. The decision consequence compounds the Layer-1 point: physical activity especially earns its top ranking not only because it lowers dementia/CVD/cancer/PD incidence but because the same act preserves function, cuts pain, and can improve frailty status (on a thin evidence base — see Frailty) — one regimen, many patient-important outcomes across both the disease-incidence and the function/QoL axes. The break-points differ by outcome (weight loss presupposes overweight and is contraindicated in the frail; the falls benefit weakens once frailty is established), so the caveats stay outcome-specific -> Layer 1 - Ranking Interventions for a Stratum.