What the analysis says
The cluster asks the net effect of adding egg or egg-derived cholesterol on the atherogenic blood-lipid profile in controlled feeding. Four members: H-34 - Egg-derived dietary cholesterol raises fasting LDL-total cholesterol linearly, roughly +1.4-1.5 mg-dL per 100 mg a linear absolute LDL/total-cholesterol rise of ~+1.4-1.5 mg/dL per 100 mg dietary cholesterol borne by LDL and apoB not HDL; H-20 - In dysglycemic adults, replacing high-carb breakfast foods with eggs modestly worsens LDL-C a comparator-relative LDL worsening in dysglycemic adults (LDL still falls, just less than a high-carb breakfast); H-40 - Whole eggs on a carbohydrate-restricted diet produce a net-favorable-neutral lipid response (HDL up, LDL-HDL preserved) even in hyper-absorbers a net-neutral-to-HDL-favourable response on a carbohydrate-restricted background even in hyper-absorbers; and residual H-53 - Some other egg lipid-profile effect (a particle-quality-only or subgroup-confined effect). The prior [H-20, H-34, H-40, H-53] = [0.179, 0.448, 0.269, 0.103] moved only modestly to posterior [0.201, 0.432, 0.297, 0.069]: the linear-rise reading leads throughout but the neutral reading gains. Four witnesses did the work, each anchored on the member it defines. CG-21 - HC-6 joint over O-75+O-76+O-77+O-78+O-80 (Ginsberg 1994, trust 0.8) is the strongest: a 0/1/2/4-egg crossover in healthy young men gave TC +1.47 and LDL +1.38 mg/dL per 100 mg with HDL/TG flat and apoB up ~10% (A-27 - The parallel ~10% apoB rise means added egg cholesterol raises atherogenic LDL-particle number, not just cholesterol per particle: added particles, not just cholesterol per particle) — H-34’s exact fingerprint, scored 0.12x against the favourable H-40. Against it, CG-22 - HC-6 joint over O-97+O-98 (Mutungi 2008, trust 0.5) and CG-19 - HC-6 joint over O-4+O-5 (DIABEGG RCT, trust 0.66) anchor H-40: 3 eggs/day (640 mg) raised HDL with LDL:HDL preserved on a carb-restricted diet, and a 3-month prediabetes RCT saw a flat panel. CG-20 - HC-6 joint over O-50+O-51 (Maki 2020, trust 0.56) anchors H-20: eggs lowered LDL less than a high-carb breakfast (-2.9% vs -6.0%, P=0.023) but LDL still fell. The near-standoff — one strong pro-rise witness offset by two weaker neutral ones — is why H-34 stays on top without dominating.
What the model may not capture
The exclusivity here is a labelling convention, not a fact of biology. H-34 (a mean effect in healthy young men on a mixed-fat background) and H-40 (a response on a carb-restricted background) describe different physical regimes; both can be literally true in scope, yet the cluster forces “the single best characterization,” so the posterior split reads like a factual disagreement that may not exist. The same collapse hides the mean-vs-spread relationship logged to HC-7: a linear average rise and large inter-individual variance are the first two moments of one trait, not rival answers. The deeper carving risk is the biomarker itself — LDL-C bulk is a proxy for atherogenicity, and apoB / particle number (measured in only one of the four trials) is the better one; if particle quality (size, oxidation) carries the risk, the whole member set is scored on the wrong quantity and H-53’s ~7% understates it. And the favourable pole is motivated: H-40’s defining study is the egg-industry-funded Fernandez lab (already docked to trust 0.5), so the counter-evidence to the harm reading is exactly the evidence most likely to be biased. Is the true answer on the list? A plausible unlisted one — “a small real rise that is clinically negligible and swamped by background fat quality” — straddles H-34 and H-40 and would be more consequential for the CVD verdict than any listed member, because it makes the biomarker nearly irrelevant rather than adjudicating its sign.
What would help
Head-to-head controlled feeding in one population varying only the carbohydrate background, to test whether H-34 and H-40 are contradictory or merely context-scoped — does not exist. ApoB / LDL-particle-number and LDL size across all four trials, not just Ginsberg — exists, unread where measured, does not exist where not. Independent (non-Fernandez) replication of the carb-restricted HDL rise — unclear.
Confusions and contradictions
Ginsberg (LDL rises linearly, HDL flat) and Mutungi (LDL flat, HDL up) point opposite ways on the same exposure, and the analysis reconciles them by background diet and a trust discount rather than resolving which is the real egg effect — a genuine, source-limited standoff. Alongside it, A-15 - Observed egg LDL-C change was far below the dietary-cholesterol dose-response prediction, consistent with yolk phospholipids blunting absorption notes real-world egg LDL changes run far below the clean dose-response Ginsberg reports (yolk phospholipids may blunt absorption), so the model simultaneously holds “linear +1.38/100 mg” and “eggs under-deliver that in practice.” Shipped unresolved; the honest fix is prose, not a fifth estimate.