Cigarette smoking contributes to one-third of cancer deaths. Approximately 14% of adults in the United States
are current tobacco smokers. Though several Food and Drug Administration (FDA)-approved smoking cessation
pharmacotherapies exist [e.g., varenicline, bupropion, nicotine replacement therapy (NRT)], utilization rates
remain low and a substantial portion of smokers do not respond to existing treatments. A personalized treatment
recommendation in which smokers are provided with a smoking cessation pharmacotherapy based on their
individual characteristics may improve both utilization of FDA-approved smoking cessation pharmacotherapies
and quit success among smokers. Our goal is to develop an algorithm, based on demographic and clinical data
assessed prior to treatment, to estimate individual smokers' likely response to FDA-approved pharmacotherapies
for smoking cessation, including varenicline, bupropion, and nicotine replacement therapy (NRT). Models will
account for the likelihood of adverse effects of medication and non-adherence. Individual estimates of treatment
response will be obtained through sophisticated analytic modeling (e.g., machine learning techniques) of existing
data from a single, large-scale randomized controlled trial (EAGLES trial conducted by Pfizer and
GlaxoSmithKline, United States sample, N=4207). The EAGLES trial provides a rich dataset comparing three
FDA-approved medications head-to-head in a large and clinically representative sample. In the EAGLES trial,
participants were randomly assigned to receive varenicline (1 mg twice daily), bupropion (150 mg twice daily),
NRT patch (21 mg per day with taper), or placebo pill capsules/patches for 12 weeks. Smoking cessation
outcomes at weeks 9 through 12 were measured. We propose to use multiple statistical techniques (e.g.,
machine learning) to optimize a model for predicting an individual's likelihood of specific smoking cessation
success in response to each treatment. Consistent with the primary analyses in the EAGLES trial, we will define
treatment success as carbon monoxide-confirmed continuous abstinence during weeks 9 through 12.
Secondarily, we will also examine continuous abstinence during weeks 9 through 24. We will develop a patient
and provider-facing mobile app prototype that implements the best-fitting algorithm and prospectively predicts
new patients' likelihood of smoking cessation with various pharmacotherapies. The mobile app will allow a new
patient to complete a reduced set of assessments based on the predictors deemed relevant in the final model.
The development of an app prototype will position us to complete user testing and refinement in a future study.
Finally, we will develop a R package to facilitate implementation of similar models by statisticians working with
other disease data.