We performed an observational study of patterns of smoking cessation in pregnant women. Because the current analysis necessitated state-level data sharing and therefore could not be readily completed using national data, we used birth certificates files for resident live births from three states: Indiana, Kentucky and Ohio, from 2009 to 2013. These states were selected based on the overall annual number of births, the relatively high prevalence of cigarette smoking during pregnancy12 13 and their geographical proximity (the three states are geographically contiguous with overlapping Tips media markets). In addition, none had introduced new state-level tobacco control programmes at the time of the airing of Tips 2012.

Tips 2012 aired for 12 weeks, from 19 March 2012 to 10 June 2012, with sufficient media placement to reach three-quarters of US adults on multiple occasions.3 Advertisements appeared on television in all US media markets through a national buy of commercial advertising time on cable television networks. Tips 2012 had a ubiquitous national buy and a local ‘buy up’ strategy in which the campaign was broadcast through smaller local television channels in media markets with high cigarette smoking prevalence. All three states included in our analysis had multiple markets that received the additional local television ad buys. The total campaign dose was approximately 70% higher than the national average dose in Ohio and Kentucky and approximately 25% higher in Indiana. It is estimated that 80% of US cigarette smokers saw at least one Tips 2012 message and that those who saw any advertisement averaged 23 views over the 12-week period.3

Exposure to the Tips campaign was defined as temporal overlap between each woman’s pregnancy and the preceding 3 months (prepregnancy) and airing of Tips 2012. A woman’s date of conception and the beginning dates for each trimester were calculated from the obstetric estimate of the gestational age at delivery in completed weeks and from the infant’s date of birth, which was ascertained directly from the birth certificate. The 3-month period preceding her date of conception was then calculated. The date 3 months before conception and the date of delivery were compared with the dates of the Tips campaign to determine temporal overlap. Women were categorised as ‘not exposed’ to the campaign if they delivered before the first day of the Tips 2012 campaign, and women were categorised as ‘exposed’ to the campaign if any day from the first date of the 3 months before conception through the end of the second trimester overlapped with the airing of the Tips 2012 campaign. We included the 3 months before conception in our exposure window because it aligned with the period of baseline smoking status included in the birth certificate and because we felt it was reasonable to assume that women who viewed the campaign in that time period would still remember the material after becoming pregnant. We did not examine exposure that occurred only in the third trimester because the cessation status of women who quit in response to seeing the campaign very late in pregnancy might not have been captured in the birth certificate, which only notes whether a woman smoked in the third trimester or not.

Intensity of exposure to the Tips campaign was assessed using weekly media market-level Tips gross rating points (GRPs) for national and local television advertising as a continuous variable. GRPs are a standard measure of advertising ‘dose’ delivered to a given audience in a given media market and time period and are defined as the product of the percentage of the audience that is exposed (ie, audience reach) and the frequency with which that exposure occurs (ie, the number of times ads are aired). GRPs are calculated at the market level by Nielson Media Research based on TV ratings for shows on which Tips ads aired. For example, if an ad was viewed by 20% of the TV viewing audience across all shows that aired the ads during a given week and if the ad was aired five times during that week, this would yield a GRP of 100 (20×5) for that week. We used women’s cumulative Tips television GRPs in each designated market area (DMA).14 Cumulative GRPs were computed by summing the weekly GRPs in each DMA based on maternal county of residence at the time of delivery for the weeks during which the woman was exposed to the Tips campaign.

The main outcome was smoking cessation by the third trimester of pregnancy among women who smoked in the 3 months before pregnancy, as recorded on the 2003 revision of the US standard certificate of live birth.15 On the birth certificate, cigarette smoking status is determined from four questions about the average number of cigarettes smoked per day in the 3 months before pregnancy and during each trimester of pregnancy. Our study population of prepregnancy smokers were women who reported smoking >0 cigarettes per day in the 3 months before pregnancy. Cessation was defined as occurring in prepregnancy smokers who reported smoking 0 cigarettes per day in the third trimester.

Covariates were ascertained from the birth certificate and included following continuous and categorical variables: maternal age; race/ethnicity (non-Hispanic white, non-Hispanic African American, Hispanic or other); education (less than high school, General Educational Development (GED) or high school diploma, some college or college graduate); marital status (married, unmarried); parity, Special Supplemental Nutrition Program for Woman, Infants and Children (WIC) enrolment (yes/no); prepregnancy body mass index (BMI); health insurance status (Medicaid, private insurance, self-pay/uninsured or other (Indian Health Service, CHAMPUS/TRICARE, etc)); state of maternal residence; and number of cigarettes smoked per day in the 3 months before pregnancy.

All women who delivered in the study period and who smoked in the 3 months before pregnancy were included in the trend analysis. Crude and standardised cessation rates were calculated and plotted by delivery date in 3-month intervals for 2009 through 2013. Data were examined for the three states individually and combined. We used joinpoint regression to identify the line segments with the best fit across the study period for the combined data.16 Cessation rates were standardised for four variables that are established in the literature as being associated with cessation (some categories were collapsed to assure adequate sample size in each stratum): maternal age (<18, 18–34 and ≥35 years of age), race/ethnicity (non-Hispanic white and other), education (<high school, high school, >high school) and parity (first or higher-order birth).8 17–19Standardisation was repeated using the number of cigarettes smoked per day before pregnancy (<10 and ≥10) in place of race/ethnicity. Standardisation was limited to four variables because of sample size constraints. Direct methods were used and were based on the 2009 first quarter population for each state for state-specific analyses and for the combined 2009 first quarter population for the combined analysis. Generalised linear models were used to generate a linear baseline trend for the precampaign period (from 1 January 2009 through 18 March 2012). Because only 9 months separated the airing of the Tips 2012 and 2013 campaigns, the exposure period for most women who delivered after 18 March 2012 (the first day of the Tips 2012 campaign) overlapped with Tips 2012 or Tips 2012 and 2013. Thus, few women who delivered after 18 March 2012 were unexposed to the Tips campaign.

Smoking cessation in women categorised as exposed or not exposed to the Tips 2012 campaign was examined using standardised rates and regression analysis in a subgroup of the study population from the trend analysis. The study population was restricted for this analysis in the following ways: women who were only exposed to Tips 2012 during the third trimester of pregnancy were excluded because of the proximity of exposure to delivery, and women who became pregnant more than 3 months after Tips 2012 aired (postcampaign) were excluded because they were considered not exposed to Tips 2012.

During our analysis, we noted that some women in our study who were exposed to Tips 2012 were also exposed to Tips 2013, which aired from 4 March 2013 to 21 June 2013, and featured content similar to Tips 2012 (neither campaign targeted pregnant women nor featured pregnancy-related health outcomes).3 5 To address this, we created a separate set of mutually exclusive exposure variables for a secondary analysis—not exposed to Tips, exposed to Tips 2012 only and exposed to Tips 2012 and Tips 2013. We did not conduct an analysis of women who were only exposed to Tips 2013 (women who were previously excluded because they became pregnant more than 3 months after Tips 2012 aired) because we did not have data for the full cohort of women, many of whom delivered in 2014.

We compared demographic characteristics and cessation rates in unexposed and exposed women using descriptive statistics. Cessation rates in exposed and unexposed women were compared using χ² tests. Standardised cessation rates were calculated using the methods previously described to standardise for maternal age, race/ethnicity, education, parity and cigarettes/day.

Logistic regression was used to generate crude and adjusted ORs (AORs) and 95% CIs for cessation in women exposed to the Tips campaign. In the unadjusted analysis, the results were stratified by maternal demographic characteristics. In multivariate regression models, results were adjusted for maternal age (continuous), race/ethnicity (non-Hispanic white, non-Hispanic African American, Hispanic or other), education (less than high school, GED or high school diploma, some college or college graduate), marital status, parity (first or higher-order birth), WIC enrolment, prepregnancy BMI (underweight or <18.5 kg/m², normal weight or 18.5–24.9 kg/m², overweight or 25–29.9 kg/m² and obese or >30 kg/m²),20 insurance status (Medicaid, private insurance, self-pay/uninsured, or other), state of residence and the number of cigarettes smoked per day in the 3 months before pregnancy (1–10, 11–20, >20). Observations with missing covariate values were excluded from logistic regression but represented less than 5% of the total. Separate models were constructed with exposure as a dichotomous variable (unexposed and exposed to Tips 2012), as a 3-level variable (unexposed, exposed only to Tips 2012 and exposed to Tips 2012 and 2013) and as a continuous variable based on cumulative GRPs for Tips 2012 and 2013 combined. We then calculated ORs and 95% CIs for every 600 GRP increase, which corresponds to approximately 6 weeks of exposure at the national level (the average cumulative exposure for the three-state area was 1857 GRPs).

Pregnant women who smoke have high rates of non-disclosure.21 22 Because we were concerned that the airing of the Tips campaign could have increased non-disclosure among smokers (if smokers felt more stigmatised after the campaign began airing), we sought to determine whether an increase in the misclassification of continuing smokers as self-reported quitters could have occurred. Infant birth weight is highly sensitive to tobacco smoke exposure,23 and we assumed that an increase in misclassification of active smokers as quitters would result in a lower mean birth weight among quitters in the Tips exposed versus the unexposed group. To assess this possibility, we compared mean birth weight among quitters (obtained from birth certificates) with singleton deliveries using analysis of covariance. Mean birth weight was adjusted for maternal age, race, education, marital status, parity, WIC enrolment, prepregnancy BMI, insurance status, state of residence and gestational age at delivery.

In all analyses, tests were two sided, and α-values of 0.05 were considered statistically significant. Statistical analyses were performed using SAS software, V.9.3 (SAS Institute) for Windows. This study was reviewed and approved as research by the institutional review boards of each state. This study was determined to be exempt from review as research without human subjects by the Institutional Review Board at the Centers for Disease Control and Prevention.

During the study period, there were 1 401 561 live births, of which 10 218 (0.7%) were excluded because the 2003 revised birth certificate was not used or because smoking status or gestational age was missing (figure 1). Of the 1 391 343 remaining, there were 894 258 births to women who were unexposed because they delivered before the Tips 2012 campaign; 209 053 births (23.4%) in this group were to women who smoked in the 3 months before pregnancy. There were 329 706 births to women whose pregnancies overlapped with the Tips 2012 (exposed); 73 582 (22.3%) of these smoked in the 3 months before pregnancy. Of these, 12 835 (17.4%) were exposed to Tips 2012 in the third trimester only and so were excluded from further analysis. Thus, there were 60 747 births for inclusion in the analysis of the association between Tips campaign exposure and smoking cessation. Of these, 13 610 (22.4%) were also exposed to Tips 2013.

During the precampaign period (from January 2009 to 18 March 2012), cessation rates among smokers who delivered before Tips 2012 aired remained flat (figure 2). Cessation rates then increased between the 12^th and the 15^th quarter, coinciding with the airing of Tips 2012, and reached a plateau by the end of Tips 2012. The plateau was sustained through the end of 2013. Analysis using joinpoint regression verified the number of statistically significant joinpoints at the p=0.05 level.

When states were examined individually, trends in Ohio and Indiana resembled those using the aggregated data, but an increase in cessation that began approximately one quarter before Tips 2012 was observed in Kentucky (data not shown).

Demographic characteristics of women exposed and unexposed to the Tips campaign differed slightly for all variables examined; these differences were statistically significant (table 1). Overall, 33.0% of unexposed smokers quit by the third trimester. In contrast, 34.8% of exposed smokers quit by the third trimester, an absolute increase of 1.8 percentage points (p<0.001) (table 2). Standardised cessation rates were 32.8% in unexposed women and 34.3% in exposed women, an absolute increase of 1.5 percentage points (p<0.001). Including cigarettes/day in standardisation calculations did not change these findings (data not shown). For individual states, there was a 0.9 percentage point increase in exposed versus unexposed women for Indiana, a 1.0 percentage point increase for Ohio and a 3.4 percentage point increase for Kentucky. Cessation rates were statistically significantly higher in exposed versus unexposed women for nearly all subpopulations examined. Statistically significant increases in cessation rates were observed in all age groups, in all race/ethnicity groups except Hispanic women, in all education categories, in married and unmarried women, in first and higher-order parity groups, in all BMI groups, in those enrolled in WIC and not enrolled and in all insurance categories except self-pay. Statistically significant increases were also seen in women smoking 1–10 cigarettes/day before pregnancy and in those smoking ≥21 cigarettes/day, but not in those smoking 11–20 cigarettes/day (data not shown).

In unadjusted regression models, cessation in smokers was statistically significantly associated with Tips 2012 campaign exposure (OR 1.08, 95% CI 1.06 to 1.10) (table 3). After adjustment for potential confounders, results remained statistically significant (AOR 1.07, 95% CI 1.05 to 1.10). Because the increase in cessation was substantially higher in Kentucky than the other two states, the analysis was repeated after excluding Kentucky. Results were attenuated but remained statistically significant (AOR 1.03, 95% CI 1.01 to 1.06). Including exposure to Tips 2013 in the model did not increase the odds of quitting (Tips 2012 only: AOR 1.07, 95% CI 1.04 to 1.09; Tips 2012 and 2013: AOR 1.06, 95% CI 1.01 to 1.10).

On average, exposed women in our analysis were exposed to 176.6 GRPs per week. When Tips 2012 and 2013 exposure dose was examined as a continuous variable, the adjusted odds of quitting rose for each increase of 600 GRPs and this finding was statistically significant (AOR 1.07, 95% CI 1.03 to 1.11).

The mean crude and adjusted birth weights of infants born to women who quit smoking and were not exposed to the campaign were not statistically significantly different from those of infants born to women who quit smoking and were exposed to the campaign (crude mean birth weight difference −5.2 g, 95% CI −14.1 to 3.8; adjusted mean birth weight difference −2.0 g, 95% CI −9.0 to 5.0) (table 4). These findings did not change when we restricted our analysis to term births (data not shown). Thus, we found no evidence of a substantial change in the proportion of smokers misclassified as quitters after the Tips campaigns aired.