Prospective cohort of 461,213 Chinese adults (China Kadoorie Biobank, ~0.5M total enrolled across 10 regions, recruited 2004-2008), free of prior cancer/CVD/diabetes, median follow-up 8.9 years. Egg consumption assessed via a non-validated qualitative food-frequency questionnaire; daily consumers averaged ~0.76 egg/day. Compared with never/rare consumers, daily egg consumption was associated with lower risk of overall CVD, ischemic heart disease, major coronary events, hemorrhagic stroke, ischemic stroke, and CVD mortality, with significant dose-response trends.
Methodology
Prospective cohort of 512,891 adults aged 30-79 from 10 regions (5 urban, 5 rural) of China, recruited 2004-2008; 461,213 eligible after excluding prior cancer/heart disease/stroke/diabetes and missing BMI. Median follow-up 8.9 y (3.9M person-years to 31 Dec 2015). Egg intake by a laptop-based single-baseline qualitative FFQ (categories never/rarely, 1-3 days/month, 1-3 days/week, 4-6 days/week, daily) covering the prior 12 months; usual amounts calibrated from a 2013-14 resurvey. Outcomes (CVD I00-I99, IHD I20-I25, haemorrhagic stroke I61, ischaemic stroke I63, major coronary events = fatal IHD or non-fatal MI) ascertained via disease/death registries, national health-insurance linkage, and active follow-up. Stratified Cox models (strata: survey site, 5-year birth cohort; attained age as timescale); fully-adjusted Model 3 covariates: age, sex, education, household income, marital status, alcohol, smoking, physical activity, BMI, waist-hip ratio, prevalent hypertension, aspirin use, family CVD history, multivitamin use, dietary pattern. Linear trend by category midpoints. Prespecified stratified analyses by sex, age, area, education, income, alcohol, smoking, physical activity, dietary pattern, BMI, hypertension.
Results
O-8 - China Kadoorie egg intake calibrated to 0.29-0.76 eggs per day across categories
Frequency categories and calibrated mean intake (egg/day): never/rarely 0.29; 1-3 days/month 0.36; 1-3 days/week 0.46; 4-6 days/week 0.56; daily 0.76. Reproducibility of the frequency measure (age/sex-adjusted Spearman r) was 0.58 over a mean 5.4-month interval (n=926). This exposure ceiling means the paper’s dose-response says nothing about intakes above ~1 egg/day.
Link to originalMethodology
Prospective cohort of 512,891 adults aged 30-79 from 10 regions (5 urban, 5 rural) of China, recruited 2004-2008; 461,213 eligible after excluding prior cancer/heart disease/stroke/diabetes and missing BMI. Median follow-up 8.9 y (3.9M person-years to 31 Dec 2015). Egg intake by a laptop-based single-baseline qualitative FFQ (categories never/rarely, 1-3 days/month, 1-3 days/week, 4-6 days/week, daily) covering the prior 12 months; usual amounts calibrated from a 2013-14 resurvey. Outcomes (CVD I00-I99, IHD I20-I25, haemorrhagic stroke I61, ischaemic stroke I63, major coronary events = fatal IHD or non-fatal MI) ascertained via disease/death registries, national health-insurance linkage, and active follow-up. Stratified Cox models (strata: survey site, 5-year birth cohort; attained age as timescale); fully-adjusted Model 3 covariates: age, sex, education, household income, marital status, alcohol, smoking, physical activity, BMI, waist-hip ratio, prevalent hypertension, aspirin use, family CVD history, multivitamin use, dietary pattern. Linear trend by category midpoints. Prespecified stratified analyses by sex, age, area, education, income, alcohol, smoking, physical activity, dietary pattern, BMI, hypertension.
Link to original
O-9 - Daily egg intake associated with 11% lower total CVD incidence in China Kadoorie Biobank
Fully-adjusted (stratified Cox on survey site + birth cohort; attained age as timescale; adjusted for age, sex, education, income, marital status, alcohol, smoking, physical activity, BMI, waist-hip ratio, hypertension, aspirin, family CVD history, multivitamins, dietary pattern). Category HRs (95% CI) vs never/rarely: 1-3 days/month 0.97 (0.95-1.00); 1-3 days/week 0.92 (0.90-0.94); 4-6 days/week 0.90 (0.87-0.93); daily 0.89 (0.87-0.92). Per +1 egg/week: 0.97 (0.96-0.98).
Link to originalMethodology
Prospective cohort of 512,891 adults aged 30-79 from 10 regions (5 urban, 5 rural) of China, recruited 2004-2008; 461,213 eligible after excluding prior cancer/heart disease/stroke/diabetes and missing BMI. Median follow-up 8.9 y (3.9M person-years to 31 Dec 2015). Egg intake by a laptop-based single-baseline qualitative FFQ (categories never/rarely, 1-3 days/month, 1-3 days/week, 4-6 days/week, daily) covering the prior 12 months; usual amounts calibrated from a 2013-14 resurvey. Outcomes (CVD I00-I99, IHD I20-I25, haemorrhagic stroke I61, ischaemic stroke I63, major coronary events = fatal IHD or non-fatal MI) ascertained via disease/death registries, national health-insurance linkage, and active follow-up. Stratified Cox models (strata: survey site, 5-year birth cohort; attained age as timescale); fully-adjusted Model 3 covariates: age, sex, education, household income, marital status, alcohol, smoking, physical activity, BMI, waist-hip ratio, prevalent hypertension, aspirin use, family CVD history, multivitamin use, dietary pattern. Linear trend by category midpoints. Prespecified stratified analyses by sex, age, area, education, income, alcohol, smoking, physical activity, dietary pattern, BMI, hypertension.
Link to original
O-10 - Daily egg intake associated with lower ischaemic heart disease and major coronary event incidence in China Kadoorie
IHD per +1 egg/week HR 0.97 (0.95-0.98); MCE per +1 egg/week HR 0.96 (0.93-0.99). Same fully-adjusted model as the total-CVD analysis.
Link to original
O-11 - Daily egg intake associated with 26% lower haemorrhagic stroke incidence in China Kadoorie, strongest endpoint
Category HRs vs never/rarely: 1-3/month 0.86 (0.79-0.93); 1-3/week 0.82 (0.76-0.88); 4-6/week 0.77 (0.70-0.86); daily 0.74 (0.67-0.82). This is a substantially larger effect than for coronary or ischaemic-stroke endpoints.
Link to original
O-12 - Daily egg intake associated with 10% lower ischaemic stroke incidence in China Kadoorie
Category HRs vs never/rarely: 1-3/month 0.98 (0.94-1.03); 1-3/week 0.95 (0.91-1.00); 4-6/week 0.95 (0.90-1.00); daily 0.90 (0.85-0.95). Per +1 egg/week 0.97 (0.96-0.98). Weakest of the incidence associations.
Link to original
O-13 - Daily egg intake associated with 18% lower CVD mortality in China Kadoorie
9,985 CVD deaths. CVD-death category HRs vs never/rarely: 1-3/month 0.91 (0.85-0.98); 1-3/week 0.88 (0.82-0.94); 4-6/week 0.79 (0.73-0.87); daily 0.82 (0.75-0.89). The non-significant, direction-reversed daily ischaemic-stroke-death HR (1.23) is the one endpoint hinting at possible harm rather than benefit.
Link to original
Interpretation
H-6 - Moderate egg consumption causally lowers CVD risk in Chinese adults
The authors’ preferred interpretation of the graded, multi-endpoint inverse associations that survive adjustment for major risk factors. Eggs supply high-quality protein, vitamins, and bioactive components (phospholipids, carotenoids) posited as protective.
Link to original
A-3 - Haemorrhagic-stroke-driven benefit limits transfer of China Kadoorie protective signal to Western populations
The aggregate CVD hazard ratio is a case-count-weighted mixture of the endpoint-specific hazard ratios. In this cohort the daily-vs-rare endpoint HRs span 0.74 (haemorrhagic stroke) to 0.90 (ischaemic stroke), with coronary endpoints at 0.86-0.88; the aggregate 0.89 therefore sits near the ischaemic/coronary end because those endpoints supply most events even here. Crucially, the aggregate is pulled further from 1 wherever haemorrhagic stroke - the endpoint with the largest protective HR - contributes a larger event share.
Haemorrhagic stroke is a much larger fraction of total CVD in Chinese populations than in Western ones, where ischaemic heart disease and ischaemic stroke dominate. Re-weighting the same endpoint-specific HRs to a Western event mix (heavily coronary/ischaemic, HRs 0.88-0.90) would move the aggregate HR toward null relative to the 0.89 seen here. So even granting the endpoint-specific associations at face value, the size of the aggregate protective signal is partly a compositional feature of China’s CVD profile and does not straightforwardly generalize.
This is a re-weighting (composition) inference and is valid regardless of whether the underlying endpoint associations are causal or confounded; it bounds how far the causal-protective reading can be exported to other populations but does not by itself adjudicate causation.
Validity verdict — corrected (checked)
Reconstruction. The aggregate HR is an event-count-weighted mixture of endpoint-specific HRs: haemorrhagic stroke 0.74, coronary 0.86-0.88, ischaemic stroke 0.90, aggregate 0.89. Premise: haemorrhagic stroke is a larger share of CVD in China than in the West. Conclusion (as stated): “much of the aggregate benefit is haemorrhagic-stroke reduction,” so re-weighting to a Western (ischaemic/coronary-dominated) mix moves the aggregate toward null and the benefit “need not transfer.”
Traced step. The mixture framework is valid, but I traced the arithmetic against the argument’s own numbers and two claims fail:
- “Much of the benefit is haemorrhagic-stroke.” The aggregate 0.89 sits at/above the coronary-ischaemic values (coronary 0.86-0.88, ischaemic 0.90), which is only possible if haemorrhagic stroke (0.74) carries a small event weight even in China — otherwise the aggregate would be pulled well below 0.86. So haemorrhagic stroke is not carrying “much” of the aggregate protection; its extra pull takes ~0.90 down to ~0.89 only.
- “Western re-weight moves toward null” — undercutting defeater using the argument’s own HRs. Western CVD is dominated by coronary disease, and coronary HR here is 0.86-0.88 — itself as protective as, or more than, the aggregate 0.89. A coronary-heavy Western mix would keep the aggregate protective (~0.87), not push it toward null. The body reaches “toward null” only by relabelling the Western mix as “coronary/ischaemic, HRs 0.88-0.90,” quietly dropping coronary’s protective 0.86 lower bound stated earlier — an inconsistency. Swapping haemorrhagic-stroke events for coronary events does move the aggregate marginally toward null (0.74 → 0.86), but the magnitude is small and the net Western aggregate is indeterminate/possibly still protective because coronary itself protects.
What survives (the corrected claim). The valid residual is compositional only: the aggregate HR is population-specific and shouldn’t be transferred as-is, and the haemorrhagic-stroke-specific signal applies to fewer Western events. It does not establish that overall egg-CVD benefit shrinks toward null in the West.
Original
statement: “Because the protective association is strongest for haemorrhagic stroke, which is a far larger share of CVD in China than in the West, much of the aggregate egg-CVD benefit here is haemorrhagic-stroke reduction and need not transfer to populations dominated by ischaemic/coronary disease.”
Link to original
H-7 - Inverse egg-CVD association in China Kadoorie reflects healthy-user confounding not causation
The authors flag residual confounding, a single non-validated baseline dietary measure, and the observational design as limitations that could generate the association without causation.
Link to original
relevance_note: The headline protective/null anchor for Asian populations — the paper the “eggs are fine (or good) in Asia” reading of the literature rests on; directly named in the FLF brief.