The MassBUILT study is a group randomized controlled trial that was delivered in collaboration with the Massachusetts Building Trades Council. The council is a collection of unions that each run apprenticeship training programs for individuals wishing to become unionized boilermakers, bricklayers, electricians, hoisting and portable engineers, ironworkers, painters, plumbers, pipefitters, sprinklerfitters, or refrigeration workers. After recruitment of apprenticeship programs, ten eligible sites that agreed to be part of the study were matched on size and randomly assigned to four intervention sites and six control sites. All apprentices at the sites were eligible to participate in the study.

We obtained survey data at all ten sites through written questionnaires at baseline (time 1), followed by a four-month intervention period in the intervention sites. Follow-up surveys were conducted one month (time 2) and at least six months (time 3) after the intervention. The Dana-Farber Cancer Institute Institutional Review Board approved all the methods and materials used in the study. Among an embedded cohort of 1213 apprentices for whom we had survey data for all three time points of the study, 490 apprentices met our definition for smoking at baseline (i.e. smoked at least 100 cigarettes in lifetime and also smoked at least once in the last 30 days).

The control sites participated in all survey periods and only received the intervention after we completed all study data collection. The apprentices in the intervention sites received a multi-pronged intervention, conducted over four months. The intervention was based on the US Public Health Service treatment guidelines for tobacco use and dependence (Fiore, Bailey, & Cohen, 2000). Also, we drew from materials and approaches of BUILT (Building Trades United to Ignite Less Tobacco)—a project of the Labor Occupational Health Program at the University of California, Berkeley and the state building and construction trades council of California (2006).

We supplemented the curriculum for all apprentices in the intervention sites to include two one-hour modules that focused on occupational hazards encountered in the building trades. These modules stressed that exposure to toxic agents can be made even more hazardous in the presence of cigarette smoking and exposure to secondhand smoke. In addition, the intervention sites had a series of five posters that reinforced key concepts from the BUILT Toxics and Tobacco curriculum modules and that reiterated the increased health risk due to exposure to both occupational hazards and smoking. The intervention also included eight weekly group counseling sessions led by state certified tobacco treatment specialists. The specialists, trained in motivational interviewing techniques for smoking cessation, led discussions of the benefits of and barriers to tobacco use and cessation with the aim of showing the participants that there are more barriers than benefits to smoking.

Apprentices self reported their race/ethnicity, age, educational attainment, gender, and income in the baseline survey. We collapsed race/ethnicity into Hispanic, Black, White and additional race/ethnicity (a group which constituted only 4.7% of the study population and included American Indians, Hawaiians, Asians and multiple races). Due to the small number of apprentices who smoke reporting that they were of Hispanic ethnicity (n=8), we further combined this group with the additional race/ethnicities category during data analyses. We categorized educational attainment as less than high school, high school or GED, some college or 2 yr degree, and 4 years or more. The less than high school and high school or GED categories were further collapsed into one category during data analysis because only four apprentices in the intervention and three in the control group reported having less than high school education. We also collapsed household income from seven $10,000 increments from under $10,000 to $75,000 or more into four categories (<$25,000, $25,000–$49,999, $50,000–$74,999 and ≥$75,000). At the time of conducting this study, the official US poverty threshold for a household of 2 adults and 2 children was $19,806 (US Census Bureau, 2006).

The primary outcome for the study is smoking cessation at time 2, the point at which the intervention was successful, defined as a 7 day point prevalence abstinence from smoking (Question: Have you smoked a cigarette, even a puff, in the last 7 days; Answer options: yes or no). The secondary outcome is prolonged abstinence from smoking for at least 6 months from the time of data collection at time 3. We opted not to conduct biochemical verification of smoking status at the apprentice sites because union leaders advised us that any biological tests would likely be misinterpreted by the apprentices as a drug test, and would in turn lead to deep mistrust of study staff and interventionists. Survey assistants stressed the confidentiality of the survey and also reiterated that truthful reporting of smoking status is important to the ability of the team to develop effective smoking cessation interventions.

Table 2 presents the results of the multivariable models of the relationship between being in the intervention group and change in scores for the psychosocial variables. Compared to the apprentices in the control group, those in the intervention group were more likely to decrease their post-intervention rating of health risk from exposure to occupational hazards and smoking compared to smoking alone (Table 2—Model 2; β=−.22, p=.006). The intervention was not associated with significant change in how the apprentices in the intervention versus control group rated their decisional balance (Table 2—Model 1; β=−.152, p=.22).

Table 3 and 4 summarize the relationship between post-intervention levels of the psychosocial variables and smoking cessation in time 2 and 3. Based on our hypothesis, we expected that those who had high scores on the psychosocial variables would be more likely to quit smoking regardless of their intervention group. A one point increase in dual hazard score was associated with more than double the odds of smoking cessation at time 2 (OR 2.45; 95% CI 1.75 to 3.43) and nearly double the odds of smoking cessation at time 3 (OR 1.76; 95% CI 1.17 to 2.64). Likewise, each point increase in decisional balance was associated with higher odds of smoking cessation at time 2 (OR 1.23; 95% CI 1.03 to 1.48) but not at time 3 (OR 1.19; 95% CI 0.93 to 1.53).

These psychosocial variables are not eligible to be mediators of the MassBUILT intervention. The intervention did not change decisional balance. It decreased scores on dual hazard but it was an increase, not a decrease, in the variable that is associated with smoking cessation. To understand what happened, we examined change in the responses to questions that capture dual hazard (table 5). These results show that the intervention did not make a difference in its key messages. Apprentices in the intervention group significantly increased their rating of the health risk from smoking between baseline and the two outcome points, but there were no differences in how they rated their risks from occupational hazard alone or the combination of occupational hazard and smoking.

Our study has several implications for current work on smoking cessation among blue-collar workers. Other mediation analyses of multi-component interventions have produced similar results where the intervention was effective but the hypothesized mediators were not found to be mediators of the effect (Calfas, Sallis, Oldenburg, & Ffrench, 1997; Haerens, Cerin, Deforche, Maes, & De Bourdeaudhuij, 2007; Lubans & Sylva, 2007). It is possible that the intervention, as a multi-component unit, had a direct effect on smoking cessation. In that case, the various components of our intervention worked synergistically to facilitate smoking cessation among the apprentices in the intervention group and their effects cannot be decomposed. On the other hand, it is possible that there are other factors that actually did fully or partially mediate the effects of the intervention. The fact that these factors were not our hypothesized mediators and thus not the focus of the intervention, could have contributed to the smoking relapse that we found at time 3.

Another important implication of our study is that it supported the theoretical basis of our intervention even as it revealed that the intervention did not achieve one of its main objectives. Using the social contextual model and formative work from the pilot study, we diagnosed that an understanding of the dual hazard from occupational hazard and smoking will lead to an increase in smoking cessation. Our results show that high scores on this concept were associated with improved smoking cessation. In fact, a one point increase in the variable was associated with double the rate of smoking cessation, regardless of intervention group. We cannot credit our study with increasing this awareness but other studies can use this information in designing smoking cessation interventions for blue-collar workers.

The main message of dual hazard is that smoking works synergistically with occupational hazard to decrease health. Our study did not succeed in driving this message home. However, we significantly increased the perception of smoking as hazardous to health among those in the intervention group. That this generic message, which the apprentices possibly get through other sources including the warning labels on cigarette packs, had such an impact on the apprentices point to a possibility that smoking was not viewed negatively among the apprentices at baseline. The high prevalence of smoking in this population along with the finding that we did not change the apprentices’ perception regarding the benefits and barrier to smoking, warrant an examination of what drives smoking behaviors in this population. It is possible that there is a need to expand on the model used in the intervention and not just hone in on the work-related aspects of smoking behavior. We recommend that future interventions devote more time before implementation of the intervention to formative research using both qualitative and quantitative methods. Through formative research, interventionist can understand how smoking is viewed and experienced by the apprentices. In addition, formative research could help interventionists understand what the smokers in the study believe would have helped them quit and what in their environment would need to be changed to encourage prolonged smoking cessation.

Interventions to improve smoking cessation among blue-collar populations have produced mixed results and many have been unsuccessful (Campbell et al., 2002; Moher et al., 2005; Sorensen et al., 2004; Willemsen et al., 1998). The MassBUILT intervention was successful; however, the effects of the intervention did not last beyond six months. The test of mediation showed that the intervention did not significantly change smoking decisional balance among the apprentices. Smoking decisional balance captures a weight of how people see the barriers versus benefits of smoking. That the study did not impact how the apprentices weigh the barriers versus benefits of smoking could have contributed to the significant relapse at the end of the study.

The perceived benefits evaluated by the decisional balance scale include smoking’s ability to relieve tension, improve concentration and induce relaxation while the barriers were embarrassment at smoking, concern about smoke from cigarette bothering other people and feeling foolish for ignoring warnings about harm of smoking. Our intervention did not change how the apprentices view these benefits and barriers. Future interventions could include more interactive sessions to address the possibility of an entrenched view of smoking as beneficial in a setting such as this where 40% of the participants were current smokers and 61% smoked in the last 30 days. For example, the Toxics and Tobacco curriculum could have been modified to start with a group discussion of smoking. This session would involve both smokers and non-smokers and could include an evaluation of the benefits and barriers of smoking as perceived by the apprentices. In addition, given the evidence of a dose-response relationship between number of smoking cessation sessions and smoking cessation (Fiore et al., 2008), the number of group sessions in the intervention could have been increased to include more class time for discussion of the benefits and barriers to smoking.

In conclusion, our study underscores the importance of examining the how interventions affect psychosocial variables through which the interventions are supposed to operate. The significant smoking relapse in our intervention could have been interpreted to indicate a failure of the theoretical basis of our intervention. However, our analysis show that the psychosocial variables were important to smoking cessation but our intervention did not have the desired effects on these variables. Few studies exist on psychosocial correlates of smoking cessation in worksite interventions with which to compare our results. Further studies are needed to test mediators of worksite based smoking cessation intervention because even negative findings can help interventionists understand what variables are related to smoking cessation and if the different components of the worksite intervention can be decomposed.