Layer 1 - Ranking Interventions for a Stratum

Mostly a compilation. Every row of the table below is copied from a claim page with its own source; that part combines nothing. The point of the page is the shape of what is held and what is missing — per the telos, the ceiling is itself a finding.

Superseded 2026-07-28. This header previously read “a compilation, not a synthesis… nothing here combines sources”, and sources: was empty. The final section now carries a source of its own (NNR/GBD) and makes a distinction the compiled rows do not contain — population attributable burden is not individual expected benefit. That belongs on this page because it is a claim about what layer-1 ranking is, not about any one exposure. The compilation description is retained for the table; it no longer describes the whole page.

Worked against one illustrative stratum — a man of ~60 with central adiposity and metabolic syndrome, no established CVD — because the telos names “obesity with visceral adiposity and hypertension” as a dominant exposure and a hierarchy is meaningless without a stratum to hold it fixed.

The stratum is a PARAMETER of this page, not its subject. The fabric is general; only its application is personal. Titling a page like this for a specific individual’s age and condition would invert that — the telos is explicit that the wiki “is never built around one person’s situation”, and that stratification and outcome-weighting belong to layer 3, per case, after the fact. Any actual person’s values live outside the graph, never in it.

The honest answer first

The wiki cannot yet rank these interventions against one another, and the reason is specific: it holds no baseline risk for this stratum, and no magnitude at all for the two largest levers.

Layer 1 ranks by effect size x certainty, and effect size means absolute effect at a stated baseline risk. Of eight exposures below, one has an absolute effect on a hard outcome.

What is held

Exposure	Effect	Outcome type	Certainty	Source page
Quit smoking (vs continue)	all-cause HR ~3.0; ~10 yr of life recovered if quit by 40 (removes ~90% of excess risk); still pays at any age	hard, mortality	High	Smoking and Mortality
Reduce SFA to 10%E	15 fewer CVD events per 1000 (RR 0.83, 0.70-0.98)	hard, composite	Moderate	Saturated Fat Intake and Replacement
Replace SFA with PUFA	50 fewer CVD per 1000 — at a 23.8% control event rate	hard, composite	Low	Saturated Fat Intake and Replacement
Low-carb for T2D remission	RD 0.32 at 6 mo, NNT ~3; attenuates to 0.10 by 12 mo	intermediate (HbA1c-defined)	Moderate	Carbohydrate Restriction and Type 2 Diabetes Remission
Low-carb, glycaemic control	HbA1c -0.47% at 6 mo	surrogate	High	Carbohydrate Restriction and Type 2 Diabetes Remission
Very-low-carb (NICE stratum)	weight -2.38 kg at 1 yr; HbA1c -0.38	intermediate + surrogate	Very low / Low	Diets for Weight Loss - What NICE Recommends
Low-carb vs balanced-carb, weight	~1 kg — judged not clinically important	intermediate	Moderate	Low-Carbohydrate vs Balanced-Carbohydrate Diets
Reduce sodium	-3.39 mmHg systolic; hard outcomes all very low	surrogate	High (BP only)	Sodium Intake and Blood Pressure
Reduce free sugars	isoenergetic swap for other carbohydrate: null (0.04 kg)	intermediate	Moderate	Free Sugars Intake

Read the outcome-type column before the effect column. Only the first two rows are hard outcomes. Everything else is an intermediate or a surrogate, and the telos is explicit that a surrogate is not an outcome — a marker can move the right way while patients do worse.

What is held as a RECOMMENDATION but with no magnitude

• Total diet replacement, 800-1200 kcal, maximum 12 weeks — Diets for Weight Loss - What NICE Recommends carries NICE’s recommendation, its bounds and its strength, but not its effect size. The review reports one; this wiki never extracted it. The cheapest single fix on this page.

What is EMPTY — and this is the finding

Lever	Status	Why it is empty
Physical activity	ratios only, no absolutes	The WHO guideline, its 535-page evidence annex and the 779-page PAGAC report all report hazard and risk ratios and no absolute effects for adult mortality/CVD/cancer/T2D. Staged, not ingested
Weight loss as an exposure on hard outcomes	FILLED 2026-07-28 — as a NULL/ceiling	Does Weight Loss Reduce Cardiovascular Events (Look AHEAD): lifestyle weight loss did NOT reduce hard CV events in T2D (HR 0.95). The lever is real for many outcomes; its hard-CV-event benefit is unproven via the lifestyle route
Baseline risk	FILLED as machinery (2026-07-26)	SCORE2 Baseline Risk and the ESC Treatment Thresholds holds the ESC thresholds and a read-off grid. A stratum defined by age and condition alone cannot use it — the chart also needs smoking status, a BP band, a cholesterol band and a region, and the ESC category moves between high and very high across plausible values. Those four are case inputs, supplied at layer 3, not fabric
Pharmacotherapy	FILLED 2026-07-28	Semaglutide for Cardiovascular Risk in Obesity (SELECT+STEP-1) + Statins for Primary Prevention and the Power of Zero CAC (USPSTF+Nasir). Both the realistic drug comparators the telos names. Key finding is a ceiling, not a lever: see below
Alcohol · sleep · smoking	ALL HELD 2026-07-29	Alcohol (Alcohol and Mortality and Vascular Disease) + sleep (Sleep Duration and Mortality) + smoking now ingested (Smoking and Mortality, Jha 2013) — the #1 dominant exposure, all-cause HR ~3.0 and a decade of life, now quantified and at the TOP of the ranked table above

What this page establishes

1. The one thing rankable is also the smallest lever on the list. SFA reduction has the only absolute hard-outcome effect the wiki holds — 15 fewer cardiovascular events per 1000 — and the telos’s own attention-is-an-anti-signal rule flags dietary fat composition as exactly the contested, heavily-published, small-effect question that literature volume inflates.
2. The two largest levers for this stratum are unquantified, not unfavourable. Adiposity and inactivity are empty rows, and the telos names both as dominant.
3. The rows are NOT commensurable and must not be read as a ranking. Different outcomes (CVD events vs kg vs mmHg vs HbA1c), different baseline risks (the 50-fewer-per-1000 PUFA row comes from a 23.8% event-rate population, roughly 3x the 8.5% in the row above it), different populations, different follow-up. A table that looks like a ranking and is not one is a hazard; it is laid out this way to show the gaps, not to order the interventions.
4. The binding acquisition is a baseline-risk instrument, not another exposure. Every ratio already held becomes an absolute effect the moment a stratum-specific baseline risk exists. That is one ingest — the ESC threshold table plus SCORE2 — and it converts the whole corpus.
5. For hard CV events in a low-risk person, the ceiling is now an established finding, not a gap. The pharmacotherapy and weight-loss rows filled in 2026-07-28 all land the same way — Cardiometabolic Interventions and Hard CV Outcomes in Low-Risk People: no cardiometabolic intervention the wiki holds (semaglutide, statin, or lifestyle weight loss) has a large, proven reduction in hard CV events for a low-risk primary-prevention person. So item 2 sharpens: the weight lever is no longer unquantified on hard CV outcomes — it is quantified as unproven/null via the lifestyle route, which is the ceiling itself.

What would change the answer

In priority order, all from already-staged material:

1. Ingest the ESC/SCORE2 baseline-risk layer. DONE 2026-07-26 — and it revealed the next blocker: this stratum is under-specified. To read a baseline at all it needs smoking status, a BP band, a cholesterol band and a region. Until those are stated the conversion is the author’s assumptions, not the person’s data.
2. Extract NICE’s total-diet-replacement magnitudes onto its claim page — the recommendation is held without its effect size.
3. Ingest WHO Physical Activity + PAGAC and record, as a G-gap, that the absolute layer is structurally absent from the activity literature as it reaches guidance.

Limits

• One illustrative stratum, chosen to make the gaps concrete, not induced from the corpus. The telos warns the hierarchy is stratum-dependent and that a universal ordering is not the answer. Nothing here transfers to a lean 30-year-old — and re-running it for another stratum is the intended use, not a limitation.
• This page grades coherence, never validity (method-risks R1). It says what the wiki holds and what it lacks — not what is true, and not that acting on it improves any outcome.
• No independent re-checking was done for this page. Each row is carried from its claim page and inherits that page’s audit status — nothing more.

Activity enters the hierarchy with a magnitude [2026-07-27]

Layer 1 requires magnitudes comparable enough to make the ordering visible — “an effect stated only as ‘associated with’ cannot be ranked against anything.” Physical activity now has one, and it is large: HR 0.34 (0.27-0.43) at the maximum for device-measured total activity, HIGH certainty -> Physical Activity Dose and Mortality.

Two things this changes about the ranking, both stratum-dependent.

1. The big-rock framing holds for the inactive. Near-total inactivity is already named as a dominant exposure; the magnitude now supports that placement rather than merely asserting it.
2. For someone already active, the ceiling arrives early — and that is itself the finding. Maximal risk reduction is reached at ~24 min/day of MVPA. Above that, the curve flattens, so additional activity is a small lever competing with other small levers. Per the ceiling rule, “your remaining levers here are small” is a result that licenses someone to stop optimizing, not a failure to find something.

The caution that keeps this honest: the ranking is by expected effect x certainty, and this estimate is observational and between-quartile, so reverse causation is not excluded by design. A large HR from cohorts does not outrank a smaller one from trials automatically.

A published ranking arrives — on the wrong axis [2026-07-28]

NNR - Nordic Nutrition Recommendations 2023 is the first source the wiki holds that publishes an explicit ordering of dietary risk factors for a defined population, via its collaboration with the Global Burden of Disease project:

«As described in the collaboration between the Global Burden of Disease project and the NNR2023 project, a diet low in whole grains is the highest- ranked dietary risk factor in the Nordic and Baltic countries. Across all countries, low whole grains diets are responsible for one fifth of the total burden of disease attributed to dietary factors and it is the greatest overall contributor to ischemic heart disease and colon and rectum cancer (Knudsen et al, 2025).» [EXTRACTED (NNR - Nordic Nutrition Recommendations 2023) chunk 06]

«Despite their conservative methodology, the collaboration between GBD and the NNR2023 project observed that a diet high in red meat is the fourth-highest dietary risk factor for Disability Adjusted Life Years (DALYs) in the Nordic and Baltic countries. It is ranked second highest in Denmark and Iceland» [EXTRACTED (NNR - Nordic Nutrition Recommendations 2023) chunk 06]

This looks like exactly what this page says it lacks. It is not, and the difference is the point.

	This page’s table	The GBD/NNR ranking
Quantity ranked	absolute effect on one person in a stratum, per unit exposure change	attributable burden across a population (DALYs)
What drives the number	effect size x baseline risk	effect size x baseline risk x prevalence of the exposure
Answers	what do I gain by changing this?	what is costing this population the most?
Same quantity?	NO	NO

Population attributable burden is not individual expected benefit, and the wedge between them is exposure prevalence. A risk factor can top a population ranking because it is common while offering a given individual a small gain — and a rare exposure with a large individual effect can sit far down the same list while being the dominant lever for the person who has it. The ranking answers a public-health resource-allocation question; layer 1 asks a personal one. Reading the first as the second is a unit-of-analysis error, and it is a seductive one because both are honestly called “ranking dietary risk factors”. [INFERRED (NNR - Nordic Nutrition Recommendations 2023) — the two rankings are NNR's; the prevalence-wedge distinction is this page's]

What it does license, which is not nothing:

• A prevalence-weighted sanity check on where the wiki has spent its attention. Low whole-grain intake ranks first and red meat fourth in this population, while the wiki’s only rankable absolute effect is saturated-fat reduction. That is a coverage signal about the corpus, not a claim about any person.
• Stratum-dependence, demonstrated rather than asserted. The red-meat rank moves from fourth overall to second in Denmark and Iceland — the same exposure, the same evidence, a different ordering because the population differs. The telos’s “the hierarchy is stratum-dependent, not a fixed list” now has a worked instance instead of a statement.
• A named instrument to acquire. Knudsen et al. 2025 is the GBD/NNR collaboration paper and is not held; NNR reports its conclusions, not its method or its attributable fractions. AWAITS a source carrying population attributable fractions with their exposure-prevalence inputs — that, not another guideline, is what would let a population ranking be converted toward an individual one.

And a caution NNR supplies against its own ranking. The red-meat sentence opens «Despite their conservative methodology», and the preceding sentence records that the GBD review «concluded that there is a weak association between unprocessed red meat consumption and colorectal cancer, breast cancer, ischemic heart disease and type 2 diabetes». So a fourth-place burden rank coexists with a weak underlying association — which is the prevalence wedge visible inside NNR’s own text, and the cleanest available demonstration that a high rank is not a large individual effect. [EXTRACTED (NNR - Nordic Nutrition Recommendations 2023) chunk 06]

Self-critique of the NNR/GBD addition [run 2026-07-28, before commit]

• The seductive claim was NOT written. NNR hands this page a ranked list of dietary risk factors, and this page opens by saying it cannot rank anything. Presenting the GBD ranking as the fix would have closed the page’s central gap with the wrong quantity — and it would have read as the single most valuable find of the revisit. The parameter comparison was built first and returned NO.
• Unit-of-analysis check: this is the failure mode the table catches. Population attributable burden and individual absolute benefit are both honestly describable as “how much this exposure matters”, differ by a prevalence factor, and are reported in the same document.
• Prior-page reconciliation: done in the open. The header’s “nothing here combines sources” and the empty sources: were both falsified by this addition; the supersession is noted in place rather than silently rewritten.
• Attribution split. The two rankings and the «Despite their conservative methodology» caveat are NNR’s, quoted. The prevalence-wedge argument is the wiki’s and is tagged [INFERRED].
• Absence claim, scoped. “the first source the wiki holds that publishes an explicit ordering of dietary risk factors” is a claim about the wiki’s holdings, verifiable from the registry, not a claim about the literature.
• Residual: the underlying GBD paper (Knudsen et al. 2025) is not held. Everything here rests on NNR’s two-sentence report of it — no attributable fractions, no method, no confidence intervals. Recorded as an AWAITS rather than treated as a held magnitude.

The largest relative effects here are the ones a person cannot change [2026-07-28, ESC chunk 03]

This page’s table holds dietary and activity exposures. ESC’s psychosocial chapter supplies effect sizes that dwarf most of them — for exposures that are not individually modifiable.

«The strongest association has been found between low income and CVD mortality, with a RR of 1.76 [95% confidence interval (CI) 1.45-2.14].» [EXTRACTED (ESC - CVD Prevention Guidelines 2021) chunk 03]

And on psychosocial stress: it «is associated, in a dose-response pattern, with the development and progression of ASCVD, independently of conven- tional risk factors and sex». [EXTRACTED (ESC - CVD Prevention Guidelines 2021) chunk 03]

RR 1.76 is larger than any dietary relative effect this wiki holds. It is also a social position, not a behaviour — and that is the finding, not an aside.

What this does to the hierarchy, stated carefully.

• It is a ceiling statement, and the telos says ceilings are results. If the largest available relative effects attach to income and social position, then the achievable gain from optimising the dietary levers on this page is bounded well below the total variation in outcomes — which licenses someone to stop optimising, rather than implying they should try harder.
• It does NOT belong in the ranked table, and adding it would break the table’s own rule. Layer 1 ranks remediable gaps; an exposure a person cannot change has no intervention to rank, however large its association. Recording it here, outside the table, is the honest placement.
• The confounding structure is stated by ESC itself, and it names the culprits — psychosocial stress «has direct biological effects, but is also highly correlated with socioeconomic and behavioural risk factors (e.g. smoking, poor adherence)». Smoking and adherence are exactly the behavioural exposures the telos names as dominant, so part of what the 1.76 measures is already accounted for elsewhere in any ranking, and treating it as additive would double-count. [INFERRED (ESC - CVD Prevention Guidelines 2021) — the figures and the correlation warning are ESC's; the double-counting caution and the placement decision are this page's]

What would make this actionable rather than merely sobering: a source on interventions that change the stressor rather than the social position — job control, social connection, stress-management programmes — with effects on hard outcomes. The corpus holds none. ESC recommends screening for depression, anxiety and insomnia, which is a detection step, not an intervention with a measured outcome effect. AWAITS such a source; this is a G gap with a named shape, not a vague one.

The Nth intervention does not deliver its trial effect [2026-07-28, ESC chunk 06]

This page ranks interventions as if each carried its measured effect independently. ESC states the constraint that breaks that assumption:

«The incremental benefit of medication when added to an already complex regimen is often uncertain. Moreover, care for multimorbid CVD patients is often fragmented and given by multiple providers, complicating decision- making and adherence to recommended treatment.» [EXTRACTED (ESC - CVD Prevention Guidelines 2021) chunk 06]

Why this belongs on a ranking page rather than a clinical one. A ranked list invites the reading do the top item, then the next, then the next. If each addition erodes adherence to the ones already in place, the list is not additive and its tail is worth less than its stated effects imply — possibly less than nothing, if a marginal addition displaces a larger established one.

This is a telos provision with a source attached. Adherence is part of the effect; cognitive cost, decision fatigue and opportunity cost are real constraints, not excuses. ESC supplies a guidance body saying so about its own recommendations, which is stronger than the wiki asserting it.

Three bounds, because the claim is easy to over-extend.

• ESC’s sentence is about medication in multimorbid patients. Whether it transfers to stacking dietary and activity changes in a healthier person is not established here — plausible by the same mechanism, unevidenced in this source.
• «often uncertain» is not «often absent». ESC claims the incremental benefit is not reliably known, not that it is zero. Reading it as the latter would be the stronger, unsupported claim.
• It does not reorder anything on this page. It bears on how far down the list is worth going, which is a different question from which item is largest — and it reinforces the ceiling finding above rather than competing with it. [INFERRED (ESC - CVD Prevention Guidelines 2021) — the quoted constraint is ESC's; applying it to the additivity assumption behind a ranked list is this page's]

The absolute layer arrives for one exposure — and it shrinks the lever [2026-07-28, WHO SFA Annex 6]

This page’s binding complaint is that it holds almost no absolute effects. WHO’s Annex 6 has now been read, and it supplies the full per-outcome absolute profile for saturated-fat reduction. The result does not enlarge the top row; it bounds it.

Outcome	Absolute per 1000	Certainty
Cardiovascular events	15 fewer (25 fewer to 2 fewer)	Moderate
All-cause mortality	2 fewer (6 fewer to 2 more)	Moderate
CVD mortality	1 fewer (4 fewer to 2 more)	Low
CHD mortality	1 fewer (3 fewer to 3 more)	Low
Stroke	2 fewer (7 fewer to 6 more)	Very low
Type 2 diabetes	1 fewer (4 fewer to 3 more)	Low

Full profile with relative effects and study counts: Saturated Fat Intake and Replacement.

Three consequences for the hierarchy.

• The one rankable exposure is rankable on ONE outcome. Cardiovascular events clears the null; every other row crosses it, including all-cause mortality at Moderate certainty. So the table’s top row is not “SFA reduction prevents 15 events per 1000” as a general benefit — it is that, on composite cardiovascular events, and approximately nothing measurable elsewhere.
• It sharpens the existing finding rather than overturning it. This page already said the one rankable thing is also the smallest lever on the list, and invoked the telos’s attention-is-an-anti-signal rule. The annex confirms it from the inside: the most-studied dietary exposure in the corpus yields one significant absolute effect on one composite outcome.
• The ceiling statement gets firmer. Combined with the ESC finding above — that the largest relative effects attach to social position — the picture is a small, well-measured dietary lever sitting under much larger unmodifiable ones. That is a result, and the telos says to report it as one.

What is still empty is unchanged. Physical activity still has ratios without absolutes; pharmacotherapy, weight loss as an exposure, alcohol, sleep and smoking remain empty rows. Annex 6 filled one cell of one row — the SFA line — and demonstrated how much work an absolute layer takes per exposure.