Whether habitual egg consumption influences long-term mortality remains uncertain, with cohort studies and meta-analyses reporting heterogeneous associations across populations [1-3]. Most evidence relies on a single baseline dietary measure, despite diet being a sustained and time-varying behaviour; standard regression may also inadequately address time-varying confounders such as body mass index and hypertension that are themselves affected by prior diet. We therefore estimated the 18-year effects of sustained egg consumption strategies on all-cause mortality using a target trial emulation. We analysed eight waves (1991-2011) of the China Health and Nutrition Survey (n=7,903 adults) and used parametric g-formula to compare sustained strategies of no eggs, <50 g/day, ≥50 g/day (approximately one egg/day), and ≥100 g/day with the no intervention (natural) course. We also estimated a substitution strategy in which eggs added to reach ≥50 g/day displaced an equal quantity of meat. Models incorporated baseline and repeated time-varying confounders; 95% confidence intervals (CIs) were obtained from 500 community-level cluster bootstrap resamples. Over 18 years, 1,124 deaths occurred (observed risk 18.6%). Estimated mortality risk under no intervention was 19.8% (95% CI 18.4, 21.4). Risk increased monotonically as assigned egg intake decreased. Sustained elimination of eggs increased estimated risk to 21.9% (20.2, 23.9), corresponding to a risk ratio (RR) of 1.11 (1.06, 1.15) and a risk difference of +2.1 percentage points (1.2, 3.0). Sustaining ≥50 g/day lowered estimated risk to 18.2% (16.2, 20.7; RR 0.92, 0.83-1.01), while ≥100 g/day gave an RR of 0.90 (0.74-1.05). Substituting eggs for meat produced a similar estimate (RR 0.93, 0.83-1.02), indicating that meat displacement did not explain the gradient. Estimates were stable across sex, age, residency and interim follow-up horizons. In this non-Western cohort, sustained elimination of eggs was associated with higher long-term mortality, whereas higher sustained intake was compatible with a modest reduction. These findings support framing nutritional epidemiology questions as explicit, sustained interventions and accounting for time-varying confounding, rather than interpreting baseline exposure contrasts as policy-relevant causal effects.
References