Skip to main content
. Author manuscript; available in PMC: 2020 Aug 24.
Published in final edited form as: Cancer Epidemiol Biomarkers Prev. 2018 Mar 21;27(5):531–540. doi: 10.1158/1055-9965.EPI-17-0649

Table 3.

Results from the mediation analyses, with ORs and their associated 95% CIs for the NDE, NIE, and TE using residuals based on country (X- and M-sets) and batch (M-set only)

Modelsa NDE q value NIE q value TE q value % Mediated
BMI 0.85 (0.60–1.20) 4.81E–01 1.56 (1.24–1.96) 1.20E–03 1.23 (0.93–1.62) 1.74E–01 100
Lifetime alcohol 1.31 (1.06–1.61) 4.20E–02 1.09 (1.03–1.15) 4.67E–03 1.40 (1.14–1.72) 3.50E–03 24
Diet score 0.77 (0.54–1.11) 3.92E–01 0.85 (0.74–0.97) 2.52E–02 0.66 (0.47–0.92) 3.27E–02 38
Physical activity 0.98 (0.72–1.35) 9.18E–01 0.97 (0.87–1.09) 6.17E–01 0.98 (0.71–1.34) 8.84E–01 b
Smoking 1.17 (0.77–1.77) 5.36E–01 1.22 (1.04–1.44) 2.52E–02 1.42 (0.99–2.03) 1.02E–01 57
Hepatitis infection 17.99 (5.15–62.80) 4.11E–05 0.94 (0.83–1.06) 3.48E–01 16.70 (4.82–57.84) 6.24E–05 0
Diabetes 0.46 (0.11–1.93) 4.82E–01 5.11 (1.99–13.14) 2.45E–03 2.45 (0.84–7.18) 1.41E–01 100

NOTE: Values reported in bold are statistically significant.

a

Models were mutually adjusted for all HLI variables. Cases and controls were matched on age at blood collection (±1 year), sex, study center, date (±2 months), and time of day at blood collection (±3 hours), fasting status at blood collection (<3/3–6/>6 hours); women were additionally matched on menopausal status (pre/peri/postmenopausal) and hormone replacement therapy. The mediator models were linear. The outcome models were computed through conditional logistic regressions. In the mediation analysis, the exposure was the original modified HLI lifestyle factor, the mediator was the associated M-score (metabolic profile), and the outcome was HCC.

b

As the associations were null for direct and indirect effects, the proportion mediated was not computed. NDE and NIE and their 95% CIs computed from formulae are detailed in Materials and Methods.