O-55 - JPHC (Japan) - no association between egg intake and incident T2D in either men or women
O-55 — Japanese cohort, FFQ egg exposure, multivariable-adjusted logistic models (odds ratios). One of the Asian primaries anchoring the null/inverse Asia pole of the region contrast. (Corrects an earlier vault mis-summary that claimed a 40% lower risk in men - that inverse estimate was for dietary cholesterol in postmenopausal women, not eggs.) See
Link to originalMethodology
Prospective cohort within the Japan Public Health Center-based Prospective Study (JPHC): 27,248 men and 36,218 women aged 45-75, single validated FFQ for egg and cholesterol intake, ~5-year follow-up, 1,165 self-reported physician-diagnosed incident T2D cases. Multivariable logistic regression yielding odds ratios by intake quartile, with sex- and (in women) menopausal-status-stratified analyses.
Link to original
Why this is evidence
A null Asian cohort anchors the null-to-inverse Asia pole predicted by H-8 and H-9, and is evidence against H-51’s uniform non-null effect.
Likelihood
# E-58 (HC-3) — lone edge, one observation O-55 "JPHC (Japan) egg-T2D null in both sexes". Members in
# HC-3.hypotheses order [H-8 region-dependent-causal (Asia null-to-inverse), H-9 uniform-null-via-confounding,
# H-51 residual uniform non-null effect]. A clean Asian null is predicted about equally by H-8 and H-9 (both
# put Asia at null/inverse) and tells against H-51's uniform non-null effect (this edge's body). Anchored H-9 = 1.0.
p_jphc_H9 = 1.0 # anchor: uniform-null-everywhere predicts an exactly-null Asian cohort directly.
p_jphc_H8 = 0.95 # ~0.95x: H-8 predicts Asia "null-to-inverse", so a flat null sits at the null end of its
# range — very slightly less expected than H-9's exact null, essentially co-equal.
p_jphc_H51 = 0.55 # 0.55x: a uniform non-null effect is disfavoured by a null, but A-19 (approved) shows the
# self-reported outcome over ~5 y plus single-FFQ exposure bias toward the null, so a modest
# true effect (~US 14-18%/egg) could be masked — the null under-discriminates, kept real.
t_jphc = 0.6 # cap = trust_score of O-55's source S-9 = 0.65; docked to 0.6 for the abstract-level
# extraction (full text not accessed, per O-55) — a provenance weakness in the datum itself,
# distinct from A-19's attenuation already priced into the likelihoods.
evidence("HC-3", ["O-55"], [p_jphc_H8, p_jphc_H9, p_jphc_H51], t=t_jphc)