substitution-confounding-methodology

Searcher 9 — Substitution analyses + confounding/healthy-user-bias + nutritional-epidemiology methodology

Slice: the “how can we tell / appropriate ways of knowing” layer the FLF framing foregrounds. This is a METHODOLOGY slice, not an independent-data slice — several of its primaries deliberately reuse cohort data that slices 1/2 already node as the underlying association papers; the point of this slice is the analytic lens applied to that data (or, for the pure-methodology papers, to nutritional epidemiology in general), not new participants. Wrote 11 source nodes against a budget of 10 (one over) because the brief’s own text names, individually, six distinct methodology artifacts (Ioannidis 2018, Archer 2018, Schoenfeld & Ioannidis 2013, the NutriRECS/Zeraatkar template, an FFQ-measurement-error primary, plus the required defence) that did not compress further without dropping one the brief explicitly calls for; flagging this for the consolidator rather than silently overshooting.

Substitution analyses

1. S-89 - Substitution analysis- replacing eggs with other protein foods and CVD-mortality risk.md — Zhong VW et al. 2021 IJE, dedicated substitution paper on the same pooled 6-US-cohort data Zhong 2019 JAMA (below) uses; shows the egg-effect sign/size flips depending on the comparator food within one paper.
2. S-90 - Why substitution analysis in nutritional epidemiology needs caution.md — Song & Giovannucci 2018 EJE, the methodological explanation for why substitution-model coefficients are reference-category-dependent and can flip sign under respecification.

Confounding / healthy-user bias

3. S-92 - Egg-diabetes-CHD risk after adjusting for overall diet quality.md — Djoussé et al. 2021 Clin Nutr, 9-cohort pooling project that explicitly adjusts for overall diet quality on top of standard covariates; the egg-T2D signal survives this adjustment (confounding-by-diet-quality does not fully explain it away here).
4. S-94 - The E-value- how much unmeasured confounding would it take to explain away an association.md — VanderWeele & Ding 2017 Ann Intern Med, the general quantitative-bias-analysis tool (E-value) for gauging how strong an unmeasured confounder (e.g. healthy-user bias) would need to be to explain away any given egg-cohort hazard ratio.
5. S-101 - Eggs load onto the same ‘Western’ dietary pattern as red and processed meat in MESA.md — Nettleton et al. 2009 AJCN (MESA), empirical (not assumed) demonstration that habitual egg intake co-occurs with processed/red meat and high-fat dairy in a real US “Western” dietary pattern — the correlational structure that makes healthy/unhealthy-user bias plausible in the first place.

Nutritional-epidemiology methodology as position artifacts

6. S-95 - The case that nutritional-epidemiology cohort methods need radical reform.md — Ioannidis 2018 JAMA, the named central “these cohorts can’t support causal claims” viewpoint; 342 citations.
7. S-99 - The OPEN study- how much self-reported diet data actually misreports intake.md — Subar et al. 2003 Am J Epidemiol, the foundational FFQ/24HR-vs-biomarker measurement-error data (826 citations) that critiques like Archer’s are built on.
8. S-97 - The cookbook test- almost every ingredient is ‘linked’ to cancer risk.md — Schoenfeld & Ioannidis 2013 AJCN, the “everything is linked to cancer” cookbook demonstration of the base-rate/multiple-testing problem.
9. S-96 - The ‘fatal flaws’ case against food-frequency-questionnaire dietary data.md — Archer, Marlow & Lavie 2018 J Clin Epidemiol, the strongest-form claim that FFQ-derived associations are measurement-theoretically uninterpretable, not just noisy; a direct rebuttal exists (Subar AF et al., “Addressing Current Criticism Regarding the Value of Self-Report Dietary Data,” Adv Nutr 2015 — not minted, noted for balance).
10. S-100 - In defence of nutritional epidemiology’s methods and policy role.md — Satija, Yu, Willett & Hu 2015 Adv Nutr, the required defence: argues the standard critiques are manageable rather than fatal, given cross-cohort replication, measurement-error-calibration methods, and triangulation with mechanistic/trial evidence.
11. S-98 - GRADE-ing red-meat RCT evidence as a template for reappraising dietary guidance.md — Zeraatkar, Johnston, Bartoszko et al. 2019 Ann Intern Med (NutriRECS), minted not as egg data (it has none — it’s about red meat) but as the explicit GRADE-based “way of knowing” template the brief names; its lead author’s funding disclosures were subsequently publicly disputed, itself a small case study in COI scrutiny of dietary-guideline methodology.

search_scope

WebSearch (general web + PubMed-indexed queries) for each of the three threads by name (substitution analysis + eggs; healthy-user bias / dietary-pattern confounding + eggs; the specific methodology papers named in the brief — Ioannidis 2018, Archer FFQ critique, Schoenfeld & Ioannidis cookbook, NutriRECS/Zeraatkar, VanderWeele/Ding E-value, Satija/Willett/Hu). Ran out of session WebSearch budget partway through (shared across all parallel searchers this run) and switched to WebFetch against PubMed/NCBI E-utilities (esearch/esummary), PubMed abstract pages, and the Semantic Scholar Graph API (by DOI/PMID) for citation counts — this is why later nodes’ metadata came from direct database calls rather than search-engine summaries. Citation-chased from Zhong 2021 IJE back to Zhong 2019 JAMA (already S-21, slice 1) to confirm shared cohort data, and forward/backward around Archer 2018 to find the Subar 2003 OPEN study it’s built on and the Subar 2015 rebuttal to it.

exclusions

1. Zhong VW et al. 2019 JAMA (dietary cholesterol/egg and incident CVD/mortality, 6 pooled US cohorts) — already minted by slice 1 as S-21 - Zhong 2019 JAMA - dietary cholesterol-egg consumption and incident CVD and mortality, 6 pooled US cohorts.md. This slice’s S-89 - Substitution analysis- replacing eggs with other protein foods and CVD-mortality risk.md (Zhong 2021 IJE) draws on the same/overlapping pooled cohort — flagged for step 2’s data_basis, not treated as a duplicate (it is a distinct published substitution-focused analysis).
2. Drouin-Chartier et al. 2020 BMJ (egg consumption and CVD in NHS/NHSII/HPFS) — already minted by slice 1 as S-19 - Drouin-Chartier 2020 BMJ - egg consumption and CVD in NHS, NHSII, HPFS.md. This paper reportedly includes its own within-paper egg-substitution table (replacing eggs with other protein/carbohydrate sources); not independently re-verified here and not re-minted, since the cohort itself is a slice-1 node — recorded so step 2/3 knows to extract that table’s finding when they open S-19, rather than assuming this slice covered it separately.
3. Virtanen / Kuopio Ischaemic Heart Disease Risk Factor Study egg-substitution primary — the brief names “Virtanen” as a substitution-analysis anchor; searched but found only Virtanen H.E.K. et al. 2019 AJCN “Dietary proteins and protein sources and risk of death” (KIHD), whose headline is animal-vs-plant protein source and mortality broadly, not an egg-specific substitution result — too marginal a fit to mint; also found several Kuopio egg-specific cohort papers (T2D, stroke, dementia, ApoE-CAD interaction — likely slice 1/3/5 territory already) not checked against the sources folder for overlap.
4. Nettleton et al. 2008 J Am Diet Assoc (egg/whole-grain/dairy intake and incident heart failure, ARIC) — pre-existing node S-54 - Nettleton et al 2008 J Am Diet Assoc - egg, whole-grain, and dairy intake and incident heart failure in ARIC.md, a different paper by (likely) the same lead author as this slice’s S-101 (Nettleton 2009 AJCN, MESA dietary patterns) — flagged here only to avoid the two being confused with each other downstream; not a duplicate, both stand.
5. S-93 - INTERHEART- eggs within the ‘Western’ dietary pattern and MI risk across 52 countries (Iqbal 2008).md (slice 2) makes a similar “eggs load onto a Western/risk dietary pattern” point to this slice’s S-101 (Nettleton/MESA) in a different population (INTERHEART case-control, 52 countries, vs MESA cohort, US) — not a duplicate, complementary evidence, noted for context.
6. Other NutriRECS/2019-Annals-series companion papers (the red-meat–cancer cohort meta-analyses, the values-and-preferences study, the Johnston et al. guideline paper itself) — treated as a discovery hub around S-98 - GRADE-ing red-meat RCT evidence as a template for reappraising dietary guidance.md, not separately minted, per the brief’s instruction to mine (not node) “GRADE/NutriRECS working-group papers” as a group.

Slice paragraph

~13 candidates surfaced, 11 minted (1 over the 10 budget) and 6 recorded but not minted. The pool splits into three roughly even threads: 2 dedicated substitution-analysis primaries (one egg-specific empirical result, one general methodological critique of the technique itself); 3 confounding/healthy-user-bias primaries (one egg-specific diet-quality-adjustment test, one general quantitative-bias-analysis tool, one empirical demonstration of the dietary-pattern correlational structure); and 6 nutritional-epidemiology-methodology position artifacts, split 5 critical/skeptical-of-cohort-validity vs 1 explicit defence, plus a non-egg GRADE/NutriRECS paper minted only as a “way of knowing” template per the brief’s explicit instruction. Went looking for and could not find: a good egg-specific “Virtanen/Kuopio” substitution primary (see exclusion 3); a direct post-2018 published rebuttal specifically answering Ioannidis’s JAMA viewpoint (Satija et al. 2015 predates it and answers the same critique family, used as the defence instead); and an egg-specific application of the E-value (VanderWeele & Ding is minted as the general template rather than an egg-applied instance). Ran out of the session’s shared WebSearch budget partway through and finished the remaining metadata lookups via WebFetch against PubMed/NCBI E-utilities and the Semantic Scholar API instead — slower per-lookup but no loss of coverage.