Disentangling the discordance between epidemiological associations and physiological mechanisms