The 2009 Family Smoking Prevention and Tobacco Control Act (FSPTCA) gives the Food and Drug Administration (FDA) the authority to limit, but not eliminate, the nicotine content of cigarettes, if such action is likely to improve public healt. In response, the FDA and National Institutes of Health (NIH) have funded several randomized trials to evaluate the impact of Very Low Nicotine Content (VLNC) cigarettes on tobacco product use behavior. The presence of non-compliance to randomized treatment assignment (i.e., smoking commercially available non-study product) precludes generalizing the change experienced by subjects in these trials to the change in tobacco use in the entire population if the nicotine content of cigarettes was limited by regulation and normal nicotine content cigarettes were no longer legally available. In recent randomized trials of VLNC cigarettes, approximately 75% of subjects reported non-compliance to their randomized treatment assignment. These non-compliant subjects are problematic because they did not receive the full intervention (i.e., nicotine reduction) and their measures of product use behavior are likely to be different than if they had only smoked the VLNC cigarettes they were randomly assigned. A number of approaches to estimating the causal effect of VLNC cigarettes, i.e., the effect if no subjects were noncompliant, from randomized clinical trials have been proposed in the statistical literature. However, all rely on the assumption that the compliance status can be measured with certainty. In randomized trials of VLNC cigarettes, self-reported compliance status is not accurate so compliance must be estimated using biomarkers of nicotine exposure. We propose to develop statistical methods for identifying and accounting for non-compliance in randomized trials of VLNC cigarettes.
In Aim 1, we will develop statistical methods for estimating the probability that a subject was compliant given their levels of biomarkers of nicotine exposure. This will allow us to properly account for the misclassification due to using biomarkers of nicotine exposure to detect non-compliance.
In Aim 2, we will develop a statistical framework for estimating the causal effect of treatment when noncompliance is imprecisely measured. The development of these methods will result in consistent estimators of the causal effects of VLNC cigarettes, while accounting for the error associated with using biomarkers to identify non-compliance. Our application is directly relevant to the goals of the FDA Center for Tobacco Products (CTP). The estimation of the causal effect of nicotine reduction on tobacco product use behavior would represent a significant contribution to tobacco regulatory science. We will accomplish this goal through the development of innovative statistical methods that will allow us to identify non-compliance using biomarkers of nicotine exposure and estimate the causal effects that are most relevant for informing future FDA regulations

Public Health Relevance

In response to FDA's new authority to limit the nicotine content of cigarettes, the FDA and National Institutes of Health (NIH) have funded a number of randomized clinical trials to evaluate the impact of Very Low Nicotine Content (VLNC) cigarettes on tobacco product use behavior. The generalizability of the comparison of smokers randomized to VLNC versus normal nicotine cigarettes (NNC) in these trials to a scenario in which nicotine content is limited by regulation is complicated by the presence of non-compliance to randomized treatment assignment. The goal of this project is to develop innovative statistical methods that will allow regulatory tobacco researchers to detect non-compliance using biomarkers of nicotine exposure and to develop a statistical framework for estimating the causal effects of VLNC cigarettes when non-compliance is measured imprecisely using a biomarker.

Agency
National Institute of Health (NIH)
Institute
National Institute on Drug Abuse (NIDA)
Type
Small Research Grants (R03)
Project #
1R03DA041870-01
Application #
9127535
Study Section
Special Emphasis Panel (ZRG1)
Program Officer
Lin, Yu
Project Start
2016-04-01
Project End
2018-03-31
Budget Start
2016-04-01
Budget End
2017-03-31
Support Year
1
Fiscal Year
2016
Total Cost
Indirect Cost
Name
University of Minnesota Twin Cities
Department
Biostatistics & Other Math Sci
Type
Schools of Public Health
DUNS #
555917996
City
Minneapolis
State
MN
Country
United States
Zip Code
55455
Koch, Brandon; Vock, David M; Wolfson, Julian (2018) Covariate selection with group lasso and doubly robust estimation of causal effects. Biometrics 74:8-17
Boatman, Jeffrey A; Vock, David M; Koopmeiners, Joseph S et al. (2018) Estimating causal effects from a randomized clinical trial when noncompliance is measured with error. Biostatistics 19:103-118