western-cvd-cohorts

Orientation note — Searcher 1 (slice 1: Western prospective cohorts, hard CVD/mortality endpoints)

Harm-finding cohorts (dose-response harm at higher habitual intake)

1. S-21 - Zhong 2019 JAMA - dietary cholesterol-egg consumption and incident CVD and mortality, 6 pooled US cohorts — 6-cohort US IPD pooling (ARIC, CARDIA, CHS, Framingham Offspring, Jackson Heart, MESA); dose-response harm for CVD and all-cause mortality per +0.5 egg/day, attenuated by dietary-cholesterol adjustment.
2. S-27 - Djousse and Gaziano 2008 AJCN - egg consumption, CVD, and mortality in the Physicians Health Study — PHS-I; null for MI/stroke, clear mortality dose-response, ~2x in diabetics.
3. S-54 - Nettleton et al 2008 J Am Diet Assoc - egg, whole-grain, and dairy intake and incident heart failure in ARIC — standalone ARIC (distinct from its Zhong 2019 contribution); egg intake raises incident HF.
4. S-23 - Djousse and Gaziano 2008 Circulation - egg consumption and heart failure in the Physicians Health Study — PHS-I; HF risk rises only at ≥1 egg/day (1.28x) and ≥2/day (1.64x).
5. S-66 - Djousse et al 2020 Clinical Nutrition - egg consumption and coronary artery disease in the Million Veteran Program — US veterans (n=188,267); modest CAD harm at ≥1/day, null at low intake.
6. S-76 - Al-Ramady et al 2022 Clin Nutr ESPEN - egg consumption and acute stroke in the Million Veteran Program — same MVP cohort as #5; clean dose-response for ischemic (not hemorrhagic) stroke, but only 3.3y follow-up.

Null-to-protective Western cohorts (moderate intake, general populations)

1. S-19 - Drouin-Chartier 2020 BMJ - egg consumption and CVD in NHS, NHSII, HPFS — 3 pooled Harvard cohorts + updated 33-estimate meta-analysis; the field’s most-cited null result at ≤1 egg/day, with US/Europe null vs Asia-protective heterogeneity built in.
2. S-33 - Key et al 2019 Circulation - meat, fish, dairy, and egg consumption and ischemic heart disease in the pan-European EPIC cohort — 9-country, 409,885-participant EPIC; modest inverse egg-IHD association that disappears once early-follow-up (reverse-causation-prone) years are excluded.
3. S-45 - Larsson, Akesson and Wolk 2015 AJCN - egg consumption and heart failure, MI, and stroke in two Swedish cohorts — 70k+ Swedish men/women; null for MI/stroke/HF-in-women, one exception (HF in men at daily intake), and notably null in diabetics (contra the US pattern).
4. S-50 - Abdollahi et al 2019 AJCN - egg and cholesterol intake and incident stroke in the Kuopio Ischaemic Heart Disease Risk Factor Study — Finnish KIHD, 21-year follow-up, ApoE4-stratified; null throughout.
5. S-41 - Dawber et al 1982 AJCN - eggs, serum cholesterol, and coronary heart disease in the Framingham Study — classic small (n=912) standalone Framingham analysis; null, historically important, low power.
6. S-38 - Zazpe et al 2011 Eur J Clin Nutr - egg consumption and risk of cardiovascular disease in the SUN Project — Spanish graduate cohort; null but only 91 events, very wide CI.
7. S-58 - Qureshi et al 2007 Med Sci Monit - regular egg consumption and stroke-CVD risk in a US national cohort — NHANES I Epidemiologic Follow-up Study; null overall, but ~2x CAD risk in the diabetic subgroup (echoing Djoussé/PHS).

Weaker-fit / caveated evidence

1. S-78 - Crowe et al 2013 AJCN - ischemic heart disease risk in vegetarians vs nonvegetarians in EPIC-Oxford — the brief’s named EPIC-Oxford anchor, but its exposure is vegetarian-vs-nonvegetarian status (most EPIC-Oxford “vegetarians” still eat eggs), not an egg-intake gradient — flagged for step 2 to weight down as an indirect/confounded proxy rather than a direct dose-response test.

search_scope

Started from the planner’s named target cohort families and discovery hubs (Rong 2013 BMJ dose-response meta-analysis; the meta-analysis portion of Drouin-Chartier 2020 BMJ; Shin 2013 AJCN; Krittanawong 2021 and Godos 2021 dose-response meta-analyses; Alexander 2016 noted as egg-industry-funded, not itself searched further since it added no new primaries beyond what the other hubs surfaced). For each named family (Harvard NHS/NHSII/HPFS, US pooled/Zhong, PHS/WHS, EPIC pan-European, EPIC-Oxford, Rotterdam, ARIC standalone, Framingham standalone, SUN, UK Biobank, Scandinavian/Finnish) ran targeted WebSearch queries by cohort + author name, then verified every candidate’s exact citation, abstract, and effect sizes via NCBI E-utilities (esearch/esummary/efetch against PubMed) and, where PubMed/journal pages were paywalled, WebFetch against open PMC full text. Citation counts pulled from the Semantic Scholar Graph API by DOI/PMID (one straggler, Qureshi 2007, timed out repeatedly against S2’s rate limit and against a Crossref bibliographic-query fallback — left unknown). WebSearch quota (~200 calls, shared across the session) was exhausted about two-thirds through the search; the remainder (Million Veteran Program follow-up details, Framingham/PHS abstract precision, funding/COI text) was completed via WebFetch and direct NCBI/Crossref/Semantic Scholar API calls over curl instead. One incidental find (Million Veteran Program) surfaced via a search for “a distinct large 21st-century US cohort beyond NHS/HPFS/ARIC/PHS/MESA” rather than being planner-named; NHANES I Epidemiologic Follow-up Study (Qureshi 2007) likewise surfaced incidentally while chasing an “ARIC standalone” lead that turned out to be NHANES, not ARIC.

exclusions

• Hu FB et al. 1999 JAMA (original Harvard NHS/HPFS egg-CVD analysis) — superseded by Drouin-Chartier 2020’s larger, updated analysis of the same 3 cohorts; not minted to avoid double-weighting the Harvard family.
• Djoussé L et al. 2021 Clinical Nutrition, “Egg consumption, overall diet quality, and risk of type 2 diabetes and coronary heart disease: a pooling project of US prospective cohorts” (9 US cohorts, n=103,811) — boundary case with slice 3: T2D is a co-headline outcome here, not a secondary subgroup, and its constituent cohorts substantially overlap with PHS/WHS/ARIC/CHS/MESA already represented in this pool. Recorded, not minted; left for slice 3 or the step-2 consolidator to judge.
• “Egg consumption and risk of coronary artery disease, potential amplification by high genetic susceptibility” (AJCN, China-PAR/ChinaMUCA/InterASIA cohorts) — Chinese cohort, belongs to slice 2 (Asian/multinational). Not minted.
• Rotterdam Study — searched specifically (Kromhout, Voortman as likely authors) but found no dedicated egg-intake-vs-hard-endpoint primary; only Rotterdam papers on dairy specifically or a composite “nutrient-rich foods” score turned up, neither isolating egg intake. Gap, not an exclusion.
• UK Biobank — searched specifically but found no standalone Western UK Biobank primary isolating egg intake against a hard CVD/mortality endpoint (only the Chinese-cohort paper above, plus general diet-quality/polygenic-risk UK Biobank papers that don’t isolate eggs, surfaced). Gap.
• ARIC’s contribution inside Zhong 2019’s 6-cohort pooling is not itself separately mintable as “ARIC standalone” — Nettleton 2008 (a genuinely separate, differently-windowed ARIC analysis, HF-specific) was minted in its place to satisfy that named family.

Slice paragraph

14 sources across 11 distinct Western cohort families (Harvard NHS/NHSII/HPFS; the ARIC+CARDIA+CHS+FOS+JHS+MESA “Lifetime Risk Pooling Project”; Physicians’ Health Study — 2 papers, HF and CVD/mortality; pan-European EPIC; EPIC-Oxford; SUN/Spain; Framingham standalone; Cohort of Swedish Men + Swedish Mammography Cohort; Kuopio Ischaemic Heart Disease Risk Factor Study/Finland; ARIC standalone; NHANES I Epidemiologic Follow-up Study; Million Veteran Program — 2 papers, CAD and stroke), safely clearing the ≥6-family independence bar. Both sides the brief asked for are present: harm-finding (Zhong; both Djoussé/PHS papers; Nettleton/ARIC; both MVP papers) and null-to-protective (Drouin-Chartier/Harvard; EPIC pan-European; Larsson/Sweden; KIHD/Finland; Dawber/Framingham; Zazpe/SUN; Qureshi/NHANES). A cross-cutting heart-failure-specific harm signal replicates independently in 3 unrelated cohorts (PHS, ARIC, and Swedish men at daily intake only), and the diabetic-subgroup interaction (egg looks worse for people who already have diabetes) replicates in PHS and NHANES but is notably absent in the Swedish cohort — a genuine unresolved heterogeneity worth flagging downstream. Two real gaps: no dedicated Rotterdam or UK Biobank egg-vs-hard-endpoint primary could be found despite targeted searching (both may simply not exist as standalone published analyses, as opposed to being missed). One structural weakness worth flagging to step 2: 4 of the 14 sources (both PHS papers, both MVP papers) share Luc Djoussé as senior/last author across 2 of the 11 cohort families — a single research group’s output, not 4 independent analytic teams, even though the underlying cohort data are genuinely distinct.