Epigenetic prediction of complex traits and mortality in a cohort of individuals with oropharyngeal cancer

Clinical Epigenetics

Table 7 Association of phenotypic and DNAm-based predictors of smoking, alcohol drinking, BMI and education with mortality

		Minimally adjusted^a					Fully adjusted^b
Exposure	N	HR	ll	ul	P value	N^c	HR	ll	ul	P value
Self-reported phenotype
Ever vs never smoker	364	3.29	1.75	6.18	2.22 × 10⁻⁴	364	2.21	1.14	4.30	0.019
Hazardous to harmful drinker vs not	364	1.62	1.06	2.49	0.027	364	1.34	0.86	2.09	0.202
Higher education vs school education	364	0.81	0.54	1.22	0.320	364	0.87	0.57	1.31	0.503
BMI	248	0.93	0.87	0.99	0.028	248	0.98	0.92	1.06	0.664
DNAm score
Smoking
McCartney LASSO (233 CpG sites)	364	1.53	1.24	1.88	7.89 × 10^-5	364	1.20	0.94	1.52	0.144
Trejo Bayesian (59 CpG sites)	364	1.70	1.37	2.11	1.49 × 10^-6	364	1.26	0.93	1.72	0.140
AHRR (cg05575921)	364	0.59	0.48	0.74	1.72 × 10^-6	364	0.79	0.58	1.07	0.125
Joehanes (FDR) (18,760 CpG sites)	364	1.70	1.34	2.15	1.27 × 10⁻⁵	364	1.35	0.99	1.84	0.056
Joehanes (Bonferroni) (2623 CpG sites)	364	1.67	1.36	2.05	7.57 × 10⁻⁷	364	1.38	1.04	1.83	0.025
Zhang (4 CpG sites)	364	1.48	1.16	1.88	1.48 × 10⁻³	364	1.28	1.02	1.60	0.036
Alcohol
Liu (5 CpG sites)	364	1.32	1.10	1.57	2.50 × 10⁻³	364	1.19	0.97	1.47	0.094
Liu (23 CpG sites)	364	1.26	1.04	1.52	0.019	364	1.10	0.89	1.36	0.357
Liu (78 CpG sites)	364	1.25	1.07	1.45	5.02 × 10⁻³	364	1.20	0.99	1.45	0.067
Liu (144 CpG sites)	364	1.24	1.07	1.44	5.31 × 10⁻³	364	1.21	1.00	1.46	0.052
McCartney LASSO (450 CpG sites)	364	1.28	1.03	1.60	0.024	364	1.05	0.79	1.41	0.723
BMI
Trejo Bayesian (144 CpG sites)	364	0.78	0.63	0.97	0.024	248	0.77	0.56	1.08	0.132
McCartney LASSO (1109 CpG Sites)	364	0.85	0.68	1.06	0.146	248	0.77	0.57	1.04	0.093
Education
McCartney LASSO (373 CpG sites)	364	0.76	0.61	0.96	0.021	364	0.87	0.68	1.12	0.270

N number, HR hazard ratio, ll lower confidence interval, ul upper confidence interval. ^aSelf-reported phenotypes adjusted for age and gender; epigenetic scores adjusted for age, gender, cell counts and batch effects. ^bPhenotypes additionally adjusted for clinical variables (TNM stage, HPV status and comorbidity), and a combination of smoking, alcohol intake, education and BMI, as appropriate to the model; risk scores additionally adjusted for clinical variables, the corresponding phenotype predicted by the score of interest and the remaining self-reported phenotypes (excluding BMI). ^cSample numbers vary due to missing BMI data

ISSN: 1868-7083