Behavior Change Techniques to Promote Smoking Cessation During Pregnancy: A Theory-Based Meta-Analysis

Despite significant progress, smoking during pregnancy remains one of the leading preventable causes of adverse fetal and maternal health outcomes. Using the current best practice standard of psychosocial counseling, only about one out of every 20 pregnant women quits smoking, and relapse rates are very high. Developing more effective interventions to promote smoking cessation during pregnancy is a critical public health priority that requires a thorough understanding of behavior change and its complex pathways and determinants. As such, the purpose of this three-part study was to conduct the first systematic theory-based evidence synthesis of smoking cessation interventions during pregnancy, and to quantify the effectiveness of specific behavior change techniques and behavioral theories used in these interventions, with the long-term goal of informing the development of more effective interventions to reduce smoking during pregnancy. The first aim was to conduct a meta-analysis to produce quantitative estimates of intervention effect sizes and to identify factors that may explain the observed heterogeneity in intervention effectiveness. A search of six major bibliographic databases for prenatal smoking cessation interventions published between 1995 and 2015 yielded 1,223 unique articles, of which 38 met criteria for inclusion and 34 were randomized controlled trials where the primary outcome was late-pregnancy biochemically-validated smoking cessation and the unit of randomization was the individual. The results of a random effects meta-analysis of the 34 randomized controlled trials of prenatal smoking cessation interventions yielded a significant risk ratio for the primary outcome of late-pregnancy smoking cessation, such that women in the treatment groups were 1.53 times as likely to achieve smoking cessation before giving birth than women in the respective control groups (RR = 1.53; 95% CI: 1.30-1.79). Several study-level variables emerged as potential moderators of intervention effectiveness. Treatment-group participants in contingent rewards interventions were 2.82 times as likely to achieve late-pregnancy smoking abstinence than control group participants. In comparison, treatment-group participants in counseling interventions were 1.3 times as likely to achieve late pregnancy smoking abstinence than their control group counterparts. Intensity level was not associated with effectiveness in this sample. Interventions in this review also yielded promising (significant) results for many secondary outcomes of interest, including additional measures of smoking behavior as well as perinatal outcomes. Specifically, treatment group participants were 1.44 times as likely as control group participants to significantly reduce (by at least 50%) their cigarette consumption, 1.54 times as likely to be smoke free in the early postpartum period, and 1.99 times as likely to be smoke free in the late postpartum period. The results also revealed that smoking cessation interventions reduced the risk of two very common adverse perinatal health outcomes: low birthweight and preterm birth. Specifically, treatment group participants had 73% less risk of delivering a low birthweight or very low birthweight infant and 67% less risk of preterm birth compared to control group participants. The second aim was to evaluate the use of the health behavior theory in intervention design, implementation, and evaluation, and to assess whether the use of theory was associated with intervention effectiveness. Of the 26 published trials that explicitly mentioned theory in the introduction or methods, only nine were based on a single theoretical framework. Five of these studies utilized the learning-based theory of operant conditioning, two studies utilized the transtheoretical/stages of change model, one study used social cognitive theory, and one study used social learning theory. Even among these nine trials, theory was used primarily in a descriptive manner, as opposed to an explanatory or predictive manner. The results of the subgroup analyses and meta-regression models were counter to the hypothesis that use of theory would be positively associated with intervention effectiveness. Scores on two categories of the theory coding scheme (“Was theory tested?” and “Was theory used to tailor or select participants?”) were significantly associated with the primary outcome of late-pregnancy smoking abstinence, but both of the associations were negative, indicating that greater use of theory was associated with a lower likelihood of smoking abstinence during the late-pregnancy period. However, this may reflect the limited use of theory in intervention planning and design among trials included in this meta-analysis, rather than the contribution of theory when it is used optimally. The third aim was to isolate the “active” ingredients in prenatal smoking cessation programs by applying a standardized taxonomy of behavior change techniques to identify the techniques, and then quantifying the effectiveness of each individual technique. We first used Abraham and Michie’s (2008) 26-item taxonomy to identify theory-derived behavior change techniques in published descriptions of intervention content, then performed a meta-regression analysis to determine whether interventions utilizing more techniques were more likely to be effective, and then used subgroup and moderator analyses in order to quantify the effectiveness of each technique. The results revealed that the total number of behavior change techniques used was not associated with late pregnancy smoking abstinence, indicating that more is not necessarily better. Effect sizes were significantly larger for the treatment group than the control group for subsets of interventions that 1) provided information about the link between smoking and health (RR = 1.68; 95% CI: 1.26-2.12); 2) provided information about the negative consequences of smoking (RR = 1.38; 95% CI: 1.08-1.77); 3) prompted the formation of intentions to quit smoking (RR = 1.24; 95% CI: 1.00-1.53); 4) provided instructions (RR = 1.51; 95% CI: 1.21-1.89); 5) prompted specific goal setting (RR = 1.48; 95% CI: 1.17-1.88); 6) provided contingent rewards (RR = 2.82; 95% CI: 2.05-3.88); 7) taught participants to use prompts and/or cues (RR = 1.63; 95% CI: 1.03-2.59); and/or 8) had participants agree to a behavioral contract (RR = 2.14; 95% CI: 1.29-3.56). Several key findings emerged from this study. First, behavior change theory is not being utilized to its full capacity in the development and evaluation of prenatal smoking cessation interventions, with only half of the studies in this review (n = 19) reporting an explicit link between at least one behavior change technique and at least one targeted predictor of behavior change. Secondly, many of the most common behavior change techniques used in prenatal smoking cessation interventions were not associated with better intervention outcomes, nor was the quantity of techniques used associated with effectiveness. Third, the current review identified contingent rewards as the most effective behavior change technique for promoting smoking cessation during pregnancy and into the postpartum period when tangible rewards were no longer offered. While previous meta-analyses have assessed whether or not prenatal smoking cessation interventions were effective, this review expanded on existing findings by using a recently developed taxonomy to identify, isolate, and quantify the effectiveness of individual behavior change techniques used in interventions, as well as applying a coding scheme to evaluate how theory is being used in the literature and whether the use of theory is associated with the effectiveness of interventions. The results provide a framework for evaluating not only if an intervention worked, but also why, how, and under what conditions, marking an important step towards a new set of standards in evidence synthesis and theory-testing in smoking cessation research and beyond