Table 4.

Estimates of Linear Trend in Lung Cancer Relative Risks From Fixed-Effects and Random-Effects Meta-Analysis According to Alcohol Intake^a

Model and Method	Relative Riskb	95% Confidence Interval	P for Trend	P for Nonlinearity	P for Heterogeneity
Fixed-effects model
Primary data
Age-adjusted	1.21	1.16, 1.26	<0.0001	0.22	<0.0001
Multivariatec	1.07	1.03, 1.12	0.001	0.17	0.11
Floating absolute risk	1.07	1.03, 1.12	<0.001	0.19	0.11
Greenland and Longnecker	1.07	1.02, 1.12	0.002	0.22	0.11
Hamling	1.07	1.03, 1.12	0.002	0.17	0.10
Zero correlationd	1.03	0.98, 1.08	0.28	0.15	0.02
Random-effects model
Primary data
Age-adjusted	1.20	1.08, 1.34	0.0009
Multivariatec	1.07	1.01, 1.14	0.03
Floating absolute risk	1.07	1.01, 1.14	0.03
Greenland and Longnecker	1.07	1.00, 1.13	0.04
Hamling	1.07	1.01, 1.14	0.03
Zero correlationd	1.02	0.94, 1.11	0.58

^aAnalyses were based on pooled primary and summarized published dose-response data (Am J Clin Nutr. 2005;82(3):657–667) and assumed either nonzero or zero covariances across relative risks within each study.

^bRelative risk for a 12-g/day increase in alcohol intake.

^cAdjusted for smoking status (never, past, or current), smoking duration for past and current smokers (years), number of cigarettes smoked daily for current smokers (continuous), educational level (less than high school graduate, high school graduate, or postsecondary education), body mass index (weight (kg)/height (m)²; <23, 23–<25, 25–<30, or ≥30), and energy intake (kcal/day).

^dCovariances among log relative risks within each study were set to zero.