Demeaning the Science to Support a Philosophical Position
Posted by jdsargent on 28 Aug 2011 at 11:00 GMT
There is no question that Simon Chapman does not approve of giving movies with smoking an adult rating; his essay with Matthew Farrelly in PLoS Medicine  is the third time he has made this argument in as many years. While Chapman is certainly free to hold whatever philosophical positions he wants on the movie ratings policies, it is troubling that in the latest essay  Chapman and Farrelly demean the science that logically supports the adult rating as a way to reduce youth smoking that results from exposure to onscreen smoking. Furthermore, their insistence that movie ratings systems—which were developed by the movie industry to avoid censorship (see http://www.mpaa.org/ratin…)—are a threat to free speech is misleading. Comparing movie ratings advocates to regimes like North Korea (see Chapman and Farrelly’s response to Millet’s letter in this issue ) sounds more like the hyperbole we would expect from a Rupert Murdock tabloid—not a pair of esteemed public health researchers.
Until now, it seemed that Chapman respected the science. In Tobacco Control in 2008  Chapman wrote, “The major challenge comes with adolescent-targeted movies where smoking can have a major presence. As I have argued, it is difficult to be categorical that any smoking in a movie must mean that all such movies ‘promote’ smoking. But it is undeniable that many such movies do, with the exact same consequences for the health of millions that were invoked as justification for controlling tobacco advertising.” In Addiction in 2009  Chapman again accepted the science, writing, “Repeated studies have demonstrated an association between exposure and smoking uptake, even after controlling for variables such as parental and peer smoking and permissive parenting style (which increases access to R-rated movies where smoking is more common). A 2008 US National Cancer Institute report called this association ‘causal’. Just as tobacco advertising promotes positive associations with smoking, it would seem unarguable that movie smoking also makes a major contribution to the cultural iconography and appeal of smoking.”
Despite the continued accumulation of evidence linking exposure to onscreen smoking since 2008, Chapman and Farrelly  now question whether residual confounding is a problem. They also question the specificity of the movie smoking effect, arguing that something else packaged in movies besides smoking could be inducing adolescents to smoke. Citing these concerns about confounding and specificity, Chapman and Farrelly object to attributable risk estimates—empirical estimates of the effects of movie smoking on smoking in the population.
Chapman and Farrelly state, “Movie smoking may be largely artifactual to the wider attraction that those at risk of smoking have to certain genres of films.” In other words, they are concerned that some yet unmeasured characteristic of the adolescent—that draws him to films with smoking and also causes him to smoke—is the true culprit, not the movie smoking. Their statement articulates concern about potential confounding, and Chapman and Farrelly are not the first to think about it. In fact, this possibility has been a central issue in the design of all the empirical studies in the published literature, all of which included covariates (which vary from study to study) to control for characteristics of the adolescent and his or her environment. The potential confounding variables already considered in this literature includes sociodemographics; smoking and alcohol use by friends, siblings, and parents; other media exposures (television and internet time, number of movies watched per week); personality characteristics (sensation seeking, rebelliousness, self esteem, depression, attention deficit disorder), school function, participation in sports and other extracurricular activities (like church and team sports), spending money, exposure to tobacco marketing, and authoritative parenting style. This is an extensive list of empirically tested covariates, more than one would find in a review of tobacco marketing studies . After accounting for these potential confounding influences, these studies have consistently documented a movie smoking effect on youth smoking in the United States [6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22], Mexico [23,24], Scotland [25,26], Germany [27,28], Iceland , Poland , Italy , The Netherlands , and India (where smoking in Bollywood movies has been linked with youth smoking) .
Moreover, Chapman and Farrelly ignore sensitivity analyses published in two of the studies [11,12] that demonstrate a yet to be identified, unmeasured third (confounding) variable would need to have a relationship with smoking (and movie smoking exposure) as strong as that of friend smoking to confound the adjusted relation. The plausibility that such a high powered variable exists but has totally escaped detection is low. In fact, it is likely that published research in this area has, if anything, over-controlled for other influences by including friend smoking, a variable that mediates the effect of movie smoking on youth smoking [30,31].
Chapman and Farrelly also conveniently ignore the consistent observation that the movie smoking effect is stronger for lower vs. higher risk adolescents; for example, adolescents with lower sensation seeking scores  or whose parents do not smoke [7,28] have a significantly larger response to movie smoking. If residual confounding on other smoking risk factors was an issue, there would be a larger estimated movie effect among the higher risk adolescents. Thus, Chapman and Farrelly would have the reader believe that researchers have paid little attention to confounding or that they used dated measures. In fact, attention to possible confounders has been a central theme in the empirical research on movie effects.
In questioning the specificity of the observed link between exposure to smoking in movies and smoking behavior Chapman and Farrelly argue that there are many risky behaviors in addition to smoking depicted in movies and that these behaviors cluster, making it difficult to empirically disentangle the direct modeling effect (seeing movie smoking prompting adolescents to smoke) from the effect of seeing other behaviors (seeing people killed might prompt adolescents to smoke) or the movie as a whole (seeing movies high in excitement might prompt adolescents to smoke).
While everyone recognizes that movies are complex stimuli, Chapman and Farrelly fail to start with the most parsimonious theoretical explanation for the consistently observed association between movie exposure and youth smoking—the social modeling effect. The conclusion that children and adolescents model behaviors they observe is backed by years of behavioral research that supports the social cognitive theory of Bandura , and what is known about how adolescents assimilate the social images they see and eventually model them [31,33]. Social cognitive theory also emphasizes that, although children see many other images of smoking in their lives, images transmitted by people important to them (friends, parents, siblings) and people they admire (movie stars) will have much more impact than people they see smoking on the street outside smoke free bars and restaurants [34,35]. We have discussed in several places the concept that social influences are complex and that persons do not necessarily reproduce any behavior they observe in a knee-jerk fashion [30,31]. Nevertheless, social modeling and consensus effects are one of the most reliable phenomena for psychology in general and adolescent smoking in particular .
The sole piece of empirical evidence supporting the Chapman and Farrelly essay is a draft manuscript Farrelly and colleagues  posted on the internet. Farrelly and colleagues assessed smoking onset in a cohort of 1511 youth followed up annually between 2005 and 2008 and found an association between exposure to smoking in movies and youth smoking. Exposure was assessed to a fixed sample of 71 films over that period, some of which had smoking in them and some of which did not, using data from the Thumbs Up! Thumbs Down! Project of Breathe California of Sacramento-Emigrant Trails (www.SceneSmoking.org), whereby adolescents watched movies and counted screen shots of smoking, to create a continuous movie smoking index. Other adult content was obtained from a site developed for parents (Kids-in-Mind.com) and captured in three categories—sex/nudity, violence/gore, and profanity. They tested the relation between seeing smoking in moves and smoking onset, controlling for several covariates, including sensation seeking propensity (i.e., “I like to do frightening things”, a measure the Chapman and Farrelly now describe as “rather dated and of dubious validity”), whether the parents restricted access to R-rated movies, friend smoking, and use of other substances.
Surprisingly, given Chapman and Farrelly’s arguments , Farrelly and colleagues  reported that, consistent with the peer-reviewed scientific literature, “we find that exposure to smoking in the movies is correlated with youth initiating smoking and becoming susceptible to smoking. Although the magnitude of these relationships diminished after controlling for an extensive set of potential confounders, the results were robust and suggest that smoking initiation would have been roughly 20% lower if this cohort had no exposure to smoking in the movies.” Even though seeing movie smoking is the most parsimonious explanation for the findings (particularly since controls for confounding included parental oversight of adult movie content) Farrelly and colleagues  rejected their own conclusion, citing the high correlation between movie smoking and other adult content.
Previous studies have controlled for propensity of youths to seek out movies by including youth’s self-report of how many movies they watch each week as a covariate [11,16]. Other studies , including the one by Farrelly and colleagues , have controlled for youth access to adult movie content. Morgenstern and colleagues  improved on studies that rely on self-report by controlling the number of movies that a youth has seen, derived from the Beach method  assessment of exposure to movie smoking. They found an association between seeing smoking in movies and youth smoking even after controlling for number of movies the adolescent had seen. Morgenstern and colleagues were able to separate the smoking effect from movie exposure because the sample was adequately powered to test that hypothesis, despite the higher correlation between movie smoking and number of movies seen; they had data from over 16,000 youths from 6 European Union countries (compared to just 632 at the end of the Farrelly et al  study).
Chapman and Farrelly question the reliability of attributable risk estimates in the case of complex processes, like onset of smoking, despite the fact that Farrelly et al concluded that “smoking initiation would have been roughly 20% lower if this cohort had no exposure to smoking in movies” . Attributable risk estimates, like the one Farrelly conducted and others already published [7,11,21], are useful because they allow epidemiologists to assess the public health importance of a risk factor. Once a judgment of causality has been rendered, as it has been by the National Cancer Institute for the effects of smoking in movies , attributable risk estimates are routine, including for complex and poorly understood medical outcomes. For example, even though we do not know all the causes of breast cancer, and these unmeasured risk factors could confound known risk factors, the National Cancer Institute provides an attributable risk estimate based on known risk factors to help women determine their own risk for breast cancer using the Gale Model (see http://www.cancer.gov/bcr…). Combining estimates from peer reviewed studies on movie smoking, Millett et al found that the attributable risk is 44% (95 percent confidence intervals 34% to 58%), meaning that youth smoking would be 44% lower if they were not exposed to smoking in movies. The excess risk attributable to smoking in movies is this high because the adjusted relative risk for individuals is close to 2.0 and the exposure almost universal.
Finally, it is important to emphasize that, in addition to the epidemiological studies discussed here, there is a body of experimental literature supporting the effect of smoking in the movies on smoking behavior. For example, one study investigated whether exposure of young adult smokers to images of smoking in films stimulated subsequent smoking behavior . Viewing smoking scenes stimulated immediate smoking behavior. Another study compared the fMRI response to images of smoking in movies in smokers vs. nonsmokers . The response in smokers indicated that smokers have higher activity in reward centers and also spontaneous action representation while watching smoking. In other words, they not only crave, but they prepare to light up. Because these are experiments in which the environment is controlled, confounding is less of an issue and movie prompts isolate the smoking as the specific element of the exposure.
In summary, Chapman and Farrelly would have us believe that the truth is so complicated that we could never unravel it with a scientific questionnaire or an experimental study. We disagree. Any theoretical confounder is irrelevant unless it can be measured, and researchers have already shown that the specificity concern can be addressed in studies that are adequately powered and employ reliable measures of movie characteristics. To demonstrate that their concerns about confounding are real, Chapman and Farrelly need to actually develop the set of items that reliably assess the measures of character—“irreverence,” “attitude,” and “fashion sense”—that they demand and, more importantly, to demonstrate with actual data that these constructs a) predict youth smoking, b) are related also to movie smoking exposure, and c) explain relation between movie smoking exposure and youth smoking, something no potential confounder studied to date has been able to do.
In Chapman’s 2009 commentary in Addiction , to argue against the science supporting an adult content rating, he posed the question, “If movie smoking is such a problem, then why is adolescent smoking declining?” He dropped that argument after it was shown that the decline in youth smoking over the last decade parallels a decline in movie smoking . It is time to acknowledge the strength and consistency of all the studies linking onscreen smoking to adolescent smoking and to stop denigrating the science in an effort to promote a philosophical position.
James D. Sargent, M.D.
Norris Cotton Cancer Center, Dartmouth Medical School
Mike Stoolmiller, Ph.D.
Department of Education, University of Oregon
Thomas A. Wills, Ph.D.
University of Hawaii Cancer Research Center
1. Chapman S, Farrelly MC (2011) Four arguments against the adult rating of movies with smoking scenes. PLoS Medicine 8: e1001078. doi:1001010.1001371/journal.pmed.1001078
2. Millett C, Polanksy JR, Glantz SA (2011) Smoking in movies: Arguments vs. evidence. Comment in PLoS Medicine, http://www.plosmedicine.o... Accessed 28 August, 2011.
3. Chapman S (2008) What should be done about smoking in movies? Tob Control 17: 363-367.
4. Chapman S (2009) With youth smoking at historic lows, how influential is movie smoking on uptake? Addiction 104: 824-825.
5. Dalton M, Beach M, Adachi-Mejia AM, Longacre MR, Matzkin AL, et al. (2009) Early exposure to movie smoking predicts established smoking by older teens and young adult. Pediatrics 123: e551-e558.
6. DiFranza JR, Wellman RJ, Sargent JD, Weitzman M, Hipple BJ, et al. (2006) Tobacco promotion and the initiation of tobacco use: assessing the evidence for causality. Pediatrics 117: e1237-1248.
7. Dalton MA, Sargent JD, Beach ML, Titus-Ernstoff L, Gibson JJ, et al. (2003) Effect of viewing smoking in movies on adolescent smoking initiation: a cohort study. Lancet 362: 281-285.
8. Distefan JM, Gilpin EA, Sargent JD, Pierce JP (1999) Do movie stars encourage adolescents to start smoking? Evidence from California. Prev Med 28: 1-11.
9. Distefan JM, Pierce JP, Gilpin EA (2004) Do favorite movie stars influence adolescent smoking initiation? Am J Public Health 94: 1239-1244.
10. Jackson C, Brown JD, L’Engle KL (2007) R-rated movies, bedroom televisions, and initiation of smoking by white and black adolescents. Arch Pediatr Adolesc Med 161: 260-268.
11. Sargent JD, Beach ML, Adachi-Mejia AM, Gibson JJ, Titus-Ernstoff LT, et al. (2005) Exposure to Movie Smoking: Its Relation to Smoking Initiation Among US Adolescents. Pediatrics 116: 1183-1191.
12. Sargent JD, Beach ML, Dalton MA, Mott LA, Tickle JJ, et al. (2001) Effect of seeing tobacco use in films on trying smoking among adolescents: cross sectional study. BMJ 323: 1394-1397.
13. Sargent JD, Dalton MA, Beach ML, Mott LA, Tickle JJ, et al. (2002) Viewing tobacco use in movies: does it shape attitudes that mediate adolescent smoking? Am J Prev Med 22: 137-145.
14. Sargent JD, Gibson J, Heatherton TF (2009) Comparing the effects of entertainment media and tobacco marketing on youth smoking. Tob Control 18: 47-53.
15. Sargent JD, Hanewinkel R (2009) Comparing the effects of entertainment media and tobacco marketing on youth smoking in Germany. Addiction 104: 815-823.
16. Sargent JD, Stoolmiller M, Worth KA, Dal Cin S, Wills TA, et al. (2007) Exposure to Smoking Depictions in Movies: Its Association With Established Adolescent Smoking. Arch Pediatr Adolesc Med 161: 849-856.
17. Song AV, Ling PM, Neilands TB, Glantz SA (2007) Smoking in movies and increased smoking among young adults. Am J Prev Med 33: 396-403.
18. Tanski SE, Stoolmiller M, Dal Cin S, Worth K, Gibson J, et al. (2009) Movie character smoking and adolescent smoking: who matters more, good guys or bad guys? Pediatrics 124: 135-143.
19. Tickle JJ, Hull JG, Sargent JD, Dalton MA, Heatherton TF (2006) A Structural Equation Model of Social Influences and Exposure to Media Smoking on Adolescent Smoking. Basic Appl Soc Psych 28: 117-129.
20. Tickle JJ, Sargent JD, Dalton MA, Beach ML, Heatherton TF (2001) Favourite movie stars, their tobacco use in contemporary movies, and its association with adolescent smoking. Tob Control 10: 16-22.
21. Titus-Ernstoff L, Dalton MA, Adachi-Mejia AM, Longacre MR, Beach ML (2008) Longitudinal Study of Viewing Smoking in Movies and Initiation of Smoking by Children. Pediatrics 121: 15-21.
22. Wilkinson AV, Spitz MR, Prokhorov AV, Bondy ML, Shete S, et al. (2009) Exposure to smoking imagery in the movies and experimenting with cigarettes among Mexican heritage youth. Cancer Epidemiol Biomarkers Prev 18: 3435-3443.
23. Thrasher JF, Jackson C, Arillo-Santillan E, Sargent JD (2008) Exposure to smoking imagery in popular films and adolescent smoking in Mexico. Am J Prev Med 35: 95-102.
24. Thrasher JF, Sargent JD, Huang L, Arillo-Santillan E, Dorantes-Alonso A, et al. (2009) Does film smoking promote youth smoking in middle-income countries?: A longitudinal study among Mexican adolescents. Cancer Epidemiol Biomarkers Prev 18: 3444-3450.
25. Hunt K, Henderson M, Wight D, Sargent JD (2011) Exposure to smoking in films and own smoking among Scottish adolescents: a cross-sectional study. Thorax. ;thoraxjnl-2011-200095 Published Online First: 15 July 2011. Available at http://thorax.bmj.com/con...
26. Morgenstern M, Poelen EA, Scholte RH, Karlsdottir S, Jonsson SH, et al. (2011) Smoking in movies and adolescent smoking: cross-cultural study in sex European countries. Thorax. doi:10.1136/thoraxjnl-2011-200489. Published Online First: 26 August 2011. Available at http://thorax.bmj.com/con....
27. Hanewinkel R, Sargent JD (2007) Exposure to Smoking in Popular Contemporary Movies and Youth Smoking in Germany. Am J Prev Med 32: 466-473.
28. Hanewinkel R, Sargent JD (2008) Exposure to Smoking in Internationally Distributed American Movies and Youth Smoking in Germany: A Cross-cultural Cohort Study. Pediatrics 121: e108-117.
29. Arora M, Mathur N, Gupta VK, Nazar GP, Reddy KS, et al. (2011) Tobacco use in Bollywood movies, tobacco promotional activities and their association with tobacco use among Indian adolescents. Tob Control.
30. Wills T, Sargent J, Stoolmiller M, Gibbons F, Gerrard M (2008) Movie smoking exposure and smoking onset: A longitudinal study of mediation processes in a representative sample of U.S. adolescents. Psychol Addict Behav 22: 269-277.
31. Wills TA, Sargent JD, Stoolmiller M, Gibbons FX, Worth KA, et al. (2007) Movie exposure to smoking cues and adolescent smoking onset: a test for mediation through peer affiliations. Health Psychol 26: 769-776.
32. Bandura A (1986) Social Foundations of Thought and Action. A Social Cognitive Theory. Englewood Cliffs, NJ: Prentice Hall. 617 p.
33. Wills TA, Cleary SD (1999) Peer and adolescent substance use among 6th-9th graders: Latent growth analyses of influence versus selection mechanisms. Health Psychol 18: 453-463.
34. Gibbons, F. X., Gerrard, M., & Lane, D. J. (2003). A social reaction model of adolescent health risk. In J. M. Suls & K. A. Wallston (Eds.), Social psychological foundations of health and illness (pp. 107-136). Oxford, UK: Blackwell.
35. Gibbons FX, Houlihan AE, Gerrard M (2009) Reason and reaction: The utility of a dual-focus perspective on prevention of adolescent health risk behavior. British Journal of Health Psychology 14: 231-248.
36. Wills, T. A., Ainette, M. G., & Walker, C. (2007). Social influence. In M. Gerrard & K. D. McCaul (Eds)., Health Behavior Constructs: Theory , Measurement and Research. National Cancer Institute Website:http://cancercontrol.canc.…
37. Farrelly M, K. K, Nonnemaker J, Crankshaw E (2011) Movie smoking and youth initiation: parsing smoking imagery and other adult content. Social Science Research Network.
38. Sargent JD, Worth KA, Beach M, Gerrard M, Heatherton TF (2008) Population-based assessment of exposure to risk behaviors in motion pictures. Communication Methods and Measures 2: 1-18.
39. National Cancer Institute. The Role of the Media in Promoting and Reducing Tobacco Use. Tobacco Control Monograph No. 19. Bethesda, MD: U.S. Department of Health and Human Services, National Institutes of Health, National Cancer Institute. NIH Pub. No. 07-6242, June 2008.
40. Millett C, Glantz SA (2010) Assigning an ’18’ rating to movies with tobacco imagery is essential to reduce youth smoking. Thorax 65: 377-378.
41. Shmueli D, Prochaska JJ, Glantz SA (2010) Effect of smoking scenes in films on immediate smoking: a randomized controlled study. Am J Prev Med 38: 351-358.
42. Wagner DD, Dal Cin S, Sargent JD, Kelley WM, Heatherton TF (2011) Spontaneous action representation in smokers when watching movie characters smoke. J Neurosci 31: 894-898.
43. Sargent JD, Heatherton TF (2009) Comparison of trends for adolescent smoking and smoking in movies, 1990-2007. Jama 301: 2211-2213.