Is nicotine vaping associated with subsequent initiation of cannabis or other substances from adolescence into young adulthood?

Abstract

Prior studies estimating longitudinal associations between nicotine vaping and subsequent initiation of cannabis and other substances (eg, cocaine, heroin) have been limited by short follow-up periods, convenience sampling, and possibly inadequate confounding control. We sought to address some of these gaps using the nationally representative Population Assessment of Tobacco and Health Study (PATH) to estimate longitudinal associations between nicotine vaping and the initiation of cannabis or other substances among adolescents transitioning to adulthood from 2013 to 2019, adjusting for treatment-confounder feedback. Estimands like the longitudinal average treatment effect were not identified because of extensive practical positivity violations. Therefore, we estimated longitudinal incremental propensity score effects, which were identified. We found that reduced odds of nicotine vaping were associated with decreased risks of cannabis or other substance initiation; these associations strengthened over time. For example, by the final wave (2018-2019), cannabis and other substance initiation risks were 6.2 (95% CI, 4.6-7.7) and 1.8 (95% CI, 0.4-3.2) percentage points lower when odds of nicotine vaping were reduced to be 90% lower in all preceding waves (2013-2014 to 2016-2018), as compared with observed risks. Strategies to lower nicotine vaping prevalence during this period may have resulted in fewer young people initiating cannabis and other substances.

Keywords: adolescents; cannabis; emerging adults; nicotine; substance use; vaping.

Figure 1

Flow diagram depicting how an analytical sample size of 9571 was derived from the Population Assessment of Tobacco and Health Study baseline cohort of 13 651 youth.

Figure 2

Nicotine vaping prevalence across Population Assessment of Tobacco and Health Study waves 1 through 4 (n = 9571 baseline participants 12-16 years old).

Figure 3

Risk differences and 95% confidence intervals comparing past-year observed cannabis initiation risks and past-year observed other substance initiation risks by each follow-up wave with those risks had nicotine vaping odds been 10% to 90% lower in prior waves (= 0.9 to = 0.1, respectively), controlling for baseline confounding for effects by wave 2 and baseline and time-varying confounding for effects by waves 2 through 5 (n = 9571 baseline participants 12-16 years old). Example: Where = 0.1 on plot, risk difference shown is .

Figure 4

Observed wave 5 cannabis initiation and other substance initiation risks (and 95% confidence intervals) and those risks had nicotine vaping odds been 90% lower in prior waves, as well as differences in those risks, stratified by history of tobacco use at waves 1 through 4 and controlling for baseline and time-varying confounding (n = 9571 baseline participants 12-16 years old).