Table 1 Impact of adjusting for common liability factors on the apparent gateway association: Examples from published studies

Sun et al. [13],

PATH Waves 4.5-5

Model: Lifetime e-cigarette use and P12M smoking

Adjusted for:

socio-demographics

4.07 (2.86–5.81), p  < .001

Adjusted for:

socio-demographics family tobacco use, secondhand smoke, friends’ tobacco use

2.28 (1,60–3.25), p  < .001

Adjusted for:

socio-demographics, family tobacco use, secondhand smoke, friends’ tobacco use, cigarette susceptibility, lifetime use of other tobacco products, P12M alcohol and marijuana use

1.35 (0.84–2.16), p = .22

Kim & Selya [11], MTF 2015-16

Model: Lifetime e-cigarette use and P30D smoking

Unadjusted association

35.86 (15.85–81.11), p  < .001

Standard logistic regression, adjusting for: sociodemographics, P30D alcohol and marijuana use, lifetime use of other illicit substances, peer smoking, exposure to health warnings, experienced discipline, risk-seeking behavior, mood, attitutes towards smoking

4.45 (1.73– 11.40), p .002

Same set of covariates as in “intermediate adjustment,” but using inverse propensity-weighted regression

2.17 (0.62–7.63), p = .228

Leventhal et al. [39], longitudinal survey of California high school students

Model: lifetime e-cigarette use and lifetime smoking

Unadjusted association

2.65 (1.73–4.05), p < .001

(NA; no intermediate model presented in this article)

Adjusted for: sociodemographics, lifetime substance use, family smoking history, peer smoking, depressive symptoms, impulsivity, delinquent behavior, smoking susceptibility, smoking expectancies, time

1.75 (1.10–2.77), p = .02