Cessation of Dual Use of Cigarettes and E-cigarettes During Pregnancy

Study overview. We propose a pilot randomized controlled trial (RCT) to develop and test a cessation program for dual users during pregnancy. We will enroll 45 Western New York pregnant mothers (≤20 weeks gestational age) with low household incomes who are currently dual-using CCs and nicotine-containing ECs. These mothers will be randomized into one of three groups: 1) simultaneous cessation intervention (N=15), 2) stepwise cessation intervention (N=15), and 3) control group (N=15). Both intervention groups will receive a multi-component intervention, including behavior counseling, biomarker feedback, and contingent financial incentives for biochemically verified abstinence. However, the two intervention groups will have different orders of cessation targets: the simultaneous cessation group will target CCs and ECs at the same time, whereas the stepwise cessation group will first target CCs and then target ECs after quitting CCs. The control group will only receive behavior counseling. Mothers will be followed throughout pregnancy until delivery. The primary outcome is nicotine abstinence at 8 weeks after randomization.

Randomization. After the pre-test visit, participants who are still dual using CCs and ECs will be randomized into one of three groups: 1) simultaneous cessation intervention group, 2) stepwise cessation intervention group, and 3) control group. A computer will be used to generate balanced-permuted blocks with a block size of 9 and a 3-3-3 ratio, which can ensure equal numbers of participants across the three groups throughout the project. The study biostatistician will create the sequences a priori and then apply them using a workflow and database software system.

Sample size. Based on the literature and our previous intervention, we estimate the dual-use nicotine abstinence rate at 8 weeks after randomization (the primary maternal outcome) to be 30% in the intervention groups versus 10% in the control group. Accordingly, we need 62 participants per group to detect a significant difference in nicotine abstinence with 80% power at a significance level of 0.05 (Aim 1). Therefore, 15 participants per group for this pilot study should provide a sufficient sample size to test study feasibility, recruitment, and retention and yield preliminary data on intervention efficacy and the magnitude of group differences in health outcomes to support an NIH R01 grant application for a large project.

Statistical analysis. We will first conduct descriptive analyses to determine if transformations are necessary, and then run bivariate analyses. We will compare the distributions of covariates above across the 3 randomized groups. Only the unbalanced covariates that predict smoking/vaping abstinence will be controlled in data analyses. We will use Intention-to-Treat analysis, i.e., including all women as assigned at randomization. Multiple imputations with 20 replicates will be used to impute missing data on outcomes or covariates.36 Sensitivity analysis will be conducted by conservatively assuming participants lost in the follow-up to be current dual users.

We will use Chi-square tests to compare categorical outcomes (e.g., maternal nicotine abstinence rates) and analysis of variance to compare continuous outcomes (e.g., infant birth-weight-for-gestational-age z-score) across groups. Then, we will use multivariable logistic and linear regression models to fit categorical and continuous outcomes, respectively. A propensity score approach will be used to control confounders, including socio-demographics, pregnancy-related characteristics, and other substance use. The model will include group assignment (simultaneous cessation intervention vs. control, stepwise cessation intervention vs. control, or simultaneous vs. stepwise cessation intervention), fidelity of intervention delivery, adherence to intervention, and a propensity score derived from all unbalanced covariates. We will estimate the group differences in nicotine abstinence rates and odds ratios from the regression model.