Estimating the Effect of a Low Nicotine Standard on Smoking Behavior in the U.S. Population
In January of 2025, the FDA announced a proposed rule to set a nationwide nicotine standard of 0.7 mg/cigarette that would reduce nicotine levels in cigarettes by 95$\%$. Past trials comparing very-low nicotine content (VLNC) cigarettes and normal nicotine content (NNC) cigarettes consistently found randomization to receive VLNC cigarettes reduces cigarette consumption. However, these results are not intrinsically generalizable, as the trials exhibited partial treatment non-adherence and were not designed as representative samples of the United States (U.S.) population of people who smoke (PWS). Existing transportability methods do not account for non-adherence or target a different estimand. Additionally, non-adherence cannot be directly observed and must be estimated using biomarkers of nicotine exposure, and different biomarkers were collected across trials. To estimate the effect of full adherence to a low nicotine standard in the U.S. population of PWS, we develop a novel inverse odds weighting estimator for transporting an intervention's effect to a fully adherent target population, and propose a bivariate mixture model approach for estimating adherence using biomarkers that allows partial missingness in one of the biomarkers. Finally, we use our method to transport inferences from seven trials investigating VLNC cigarettes to a representative sample of U.S. PWS.