O-17 - Cholesterol responsiveness correlates with HDL2-total-cholesterol rise but not with habitual intake or body weight
Link to original
Why this is evidence
H-8’s saturating dose-response predicts that people with high habitual cholesterol intake sit on the flat part of the curve and are therefore less responsive - i.e. a negative correlation between habitual intake and responsiveness. None was found, which disfavours H-8. H-1 and H-5 make no such prediction.
Likelihood
# E-6 (HC-1) - lone edge, one observation O-17 "responsiveness correlates with the HDL2/total-cholesterol
# rise but NOT with habitual cholesterol intake or body weight". Members in HC-1.hypotheses order:
# [H-1 large ApoB rise, H-5 poor bioavailability, H-8 real-but-saturating, H-11 residual].
# This edge cuts AGAINST H-8 (see "Why this is evidence"): a saturating curve implies high-habitual-intake
# subjects sit on the flat part and respond less, i.e. a negative intake-responsiveness correlation.
# Anchored on H-1 = 1 (rule 7).
lik_nocorr_H1 = 1.0 # anchor: a non-saturating account makes no prediction tying responsiveness to
# habitual intake, so a null correlation is unremarkable
lik_nocorr_H5 = 1.0 # same: absorption-limited effects predict nothing about habitual-intake gradients
lik_nocorr_H8 = 0.45 # 0.45x: H-8's corollary predicts the negative correlation that was not found.
# Not lower because habitual-intake measurement is self-reported and noisy, the
# sample is small, and the observed habitual range may be narrow - all of which
# attenuate a real correlation toward zero. So this disfavours H-8 but weakly
lik_nocorr_H11 = 0.85 # residual, unconstrained (rule 3): most unlisted characterizations impose no
# intake-responsiveness link, so a null is largely expected
t_nocorr = 0.50 # cap = trust_score of O-17's source S-5 = 0.68. Docked substantially: this is an
# absence-of-correlation in a small responder-typing study, the weakest kind of
# inference, and habitual intake is a self-reported exposure. A-5 (approved) notes
# responder classes are stable but unpredictable, which is consistent with the
# measured covariates simply being poor proxies rather than the curve being flat
evidence("HC-1", ["O-17"], [lik_nocorr_H1, lik_nocorr_H5, lik_nocorr_H8, lik_nocorr_H11], t=t_nocorr)