e., NRT or placebo) were also available from the trial database. Outcome Variables For analyses in this paper, we used validated cessation at 1-month postquit date and at delivery as outcome variables. At 1 month, cessation was defined as continuous abstinence from quit date to 1 month, validated by an exhaled CO reading of ��8 ppm; and, at delivery, cessation was defined as continuous selleck chemicals Axitinib abstinence from a quit date until delivery, validated by an exhaled CO reading of ��8 ppm and/or a saliva cotinine level of <10ng/ml. Participants who were lost to follow-up were coded as continuing smokers. Analysis Strategy This analysis investigated associations between baseline characteristics of participants and cessation at 1 month after initiating treatment (i.e., from quit date) and at delivery.
Two multivariable logistic models were built. Initially, for both models, variables were identified which had significant univariate associations (p �� .05) with validated cessation at each timepoint. Secondly, these variables were all entered into a multivariable model using stepwise backwards elimination to remove variables found to have nonsignificant associations with outcome (p > .05). Finally, variables which showed no association at the univariate level were entered into the models individually, to determine if they were subsequently associated with validated cessation. Treatment assignment was included as an a-priori confounder. To maximize the number of participants included in the analysis, where possible, a missing category was created for categorical variables with missing data and imputation was planned for continuous variables (baseline cotinine) where >10% of cases had missing data.
Where all missing values for an exposure occurred among people in the same outcome category (i.e., continuing smokers) or a small percentage of data was missing for a continuous variable, a univariate sensitivity analysis comparing all participants against those with complete data was conducted, to verify that they did not differ in their baseline characteristics. All analyses were conducted using Stata 11.2 (College Station, TX). RESULTS In the 1 month and delivery multivariate analyses, missing data for the categorical variable ��age full time education finished�� could not be included as all those with missing data were smokers at follow-up and inclusion as an extra category would perfectly predict the outcome.
Furthermore, imputation for the continuous variable ��baseline cotinine level�� was not carried out due to data being missing for only 80 participants (7.6%). As a result, analysis was undertaken Cilengitide on 957/1,050 participants (91.1%), for whom complete exposure data were available. Including the 93 participants for whom some baseline data were missing in the final multivariable model, did not alter the results. An analysis based on achieving validated cessation was conducted and characteristics of the women included are detailed in Table 1.