A bit of a long post, but it would be nice to put this somewhere where a few people will actually read it….
Comment regarding Experimental Study of Graphic Cigarette Warning Labels
OMB Control Number 0910-NEW
Docket No. FDA-2010-N-0079
as appears in Federal Register, v.75 no.164, 25 August 2010, p.52352
by Carl V. Phillips, MPP PhD
on behalf of myself and TobaccoHarmReduction.org (other contributing members of TobaccoHarmReduction.org including Paul Bergen, Karyn Heavner, and Catherine Nissen.
I am a researcher and public educator focused primarily in the area of tobacco harm reduction, as well as various other aspects of epidemiology, health and policy ethics, health communication, and scientific epistemology. I spent most of my career as a professor of public health and now direct an independent scientific research group. My CV can be found here: tobaccoharmreduction.org/team.htm. This year I have done a great deal of secondary research that focused on the effects, politics, and ethics of graphical warning labels, particularly for tobacco products.
The following comments are based primarily on my concerns about good and honest science and health communication, and about policy ethics. It might be that ignoring some of the cautions contained here might actually benefit the cause of tobacco harm reduction (e.g., seeking graphic labels for cigarettes that fail as warnings because they mislead and merely inflict emotional violence on consumers might drive more consumers to low-risk alternative products). But that end does not justify conducting misleading science or research that is aimed at supporting unethical public policies. Moreover, bad science in the area of tobacco policy has a tendency to be particularly damaging to harm reduction efforts, which depend on the science to push back against the political power.
1. For a graphic to be a warning, it needs to primarily convey information.
There seems to be some confusion about what warning labels should do and how they can be assessed. While there are those within the anti-tobacco activist community who are willing to inflict almost any cost on scientific integrity and good public policy in order to discourage tobacco use, the United States government should not condone this, let alone contribute to it.
It is possible to alter any product’s packaging to lower the overall quality of the experience of buying, having, and consuming the product, and thus reduce consumption (see Thrasher et al 2007 for an illustration of how graphical warnings lower the perceived value of cigarettes). Options range from requiring ink that rubs off and stains consumers’ hands, to implanting an audio chip that periodically emits an unpleasant noise, to printing offensive graphics on the product. In general, these could not be called warnings, except if the sound or graphic actually communicates accurate information about risks. Moreover, context matters. A picture of a diseased lung might legitimately constitute a graphical warning on a package of cigarettes (at least if it were used in a non-literate community that was somehow not already aware of the risk), but would clearly not be a warning on a package of carrots. Yet the repulsive picture would lower the quality of the package of carrots to the consumer, even if no consumer were so ill-informed as to believe that carrots cause lung disease.
All of these hypothetical package alterations, including the ones that are not actually warnings, will reduce consumption of legally marketed cigarettes. (They would increase the demand for black market cigarettes, though the net effect on total consumption should be negative.) Perhaps a particular alteration would make such a small reduction that it would not be possible to measure it, but welfare economics tells us that any alteration that lowers quality product will result in less demand. Most of these reductions, however, would not be the result of successfully warning someone about a risk. For the case of shocking graphics that have not been shown to communicate accurate risk information, but merely to repulse consumers, the correct characterization would not be “warning label”, but rather something along the lines of “emotional assault label”, something that manipulates people by triggering an emotional reaction, in particular by inflicting emotional violence.
The FDA has a long history of successfully and appropriately using legitimate warning labels, and has a mandate regarding graphics-based warning labels for cigarettes. However, the FDA does not have a mandate to engage in emotional violence via labeling, and moreover, it would be ethically inappropriate for the U.S. government to intentionally inflict emotional violence on a large portion of its citizens, particularly if it does not communicate accurate risk information. Thus, evaluation of proposed graphical warnings should distinguish between warning and emotional violence.
A graphic that produces distress, but not by way of increasing awareness of a previously unrecognized or underestimated risk, is inflicting emotional violence. Thus, a study that finds a negative or distressed reaction to a label without an accompanying (accurate) change in the estimate of its risk is merely violence, not warning (see Hammond et al 2004 for an example of a study that measures emotional reactions such as fear and disgust in response to graphical warnings, rather than accurate risk perceptions). In order to do proper scientific research on this topic and to inform ethical public policy, it is necessary that the evaluation of the effects of the graphics be able to distinguish between conveyance of information and mere behavioral manipulation through the infliction of emotional violence.
2. The measure of the success of a risk warning label is whether it moves people’s perceptions closer to the best scientific estimates.
Some warning labels are intended as immediate alerts rather than warnings about risk. Such warnings – such as “do not step here or you will fall to your death”, “do not open this case or you will be electrocuted”, or “do not drink any of this because it is a deadly poison”, as well as “absolutely no unauthorized persons should open this door” – are designed to be substitutes for the impossible act of prohibiting an action and physically enforcing the prohibition. These contrast fundamentally with warnings that are intended to convey decision-relevant information about risk, like the labels that FDA is already very familiar with. These alert-type labels can be reasonably evaluated based on whether they help ensure that the particular action is not taken, and since the goal is to be the best available substitute for physically prohibiting the action (walking near a cliff, opening the wrong door, etc.) it does not necessarily matter whether people understand the probability of the risks or anything else.
While some anti-smoking activists want people to believe that smoking a cigarette results in immediate peril, this is obviously not a correct characterization. Obviously smoking is not like accidentally swallowing poison or stepping off a tall cliff, which everyone would want to avoid, and thus be grateful for any signage that prevented it. Rather, smoking is a voluntary act with a complicated collection of benefits and costs, and even someone who cannot understand why many people feel that the benefits exceed the costs must recognize that the costs do not include peril of sudden death that would warrant alert-type warnings. Moreover, the FDA’s mandate with regard to tobacco products explicitly excludes prohibition as an option. Thus, the use of graphics and methods of evaluation that are appropriate for alert warnings (whose intentions are to substitute for a physically enforced prohibition), is inappropriate in the present case since the peril is not imminent and prohibition is not allowed, and thus neither is crypto-prohibition. (If prohibition is prohibited then extreme measures that are functionally equivalent are also prohibited. A hypothetical package graphic that was so horrifically emotionally violent that it discouraged almost everyone from buying cigarettes would basically be prohibition, much the same way that imposing a $1000/pack price or restricting cigarette sales to one store in Montana would be.)
If a smoker believes that his lifetime risk of lung cancer is 90% (a gross overestimate but one that many smokers believe (Slovic 2001)), and the label causes him to revise that estimate to 95%, then the warning label has failed in the particular case (see Magnan et al 2009 for an example of a study that indicates that negative emotional reaction and increased risk perception are measures of success of graphical warning labels; whether or not these risk perceptions are accurate is ignored). Employing a graphic that systematically caused smokers who overestimated a particular risk to increase their estimate would be a form of lying to the public. Not only would this be a failure of a risk warning, making people’s understanding worse rather than better, but it would arguably be unacceptable even for an alert warning. Even in the case of legal and ethically-defensible prohibitions, it is generally concerned unethical for the government to lie to people to manipulate their behavior (e.g., the U.S. government prohibits cocaine use, but should not tell people that using it will inevitably kill them in a year, both because it is unethical and illegal for the government to lie, and because it ruins the credibility of honest health messages the government might want to convey; note also the example of anti-condom rhetoric by the U.S. government).
There are three goals stated for evaluating the effects of the labels on consumers. The first is “conveying information about various health risks of smoking.” This accurately describes the proper role of a warning label. However, the other two goals, “encouraging cessation” and “discouraging initiation” either pre-declare what the data will show or declare that the intention is coercion rather than warning. Conveying information will only encourage cessation if the information corrects underestimates about risks, either in terms of probability or severity. In the above example, the smoker who now overestimates his risks even more is more likely to quit smoking, but this represents a success of manipulative propaganda, not a successful ethical warning. In the Federal Register entry to which these comments respond, FDA defends survey questions about intention to quit against critics who complain that they are not effective measures of actual future quitting. But both the comments and the defense overlook that measures of intention to quit, whether accurate or not, are not measures of whether the warnings effectively communicated about risk, at least not without substantial further information.
In sum, if this study were intended to evaluate, say, a project to convince children to stop smoking, then a behavioral outcome measure would be appropriate. But this is explicitly stated as a study of warning labels, and as such needs to be evaluated based on accuracy of communication, not successful manipulation of people’s behavior through any means. In the Federal Register is the statement, “FDA agrees that it is important to ensure that the graphic health warnings convey accurate information about smoking risks to consumers,” which makes clear that the FDA is aware of this distinction.
In pursuit of this stated goal of the study, it is important to ask the following about any question posed to subjects: If the answer is X, will we know whether this represents successful accurate warning of risks or mere emotional manipulation of the subject? Obviously the answer will be “no” for some individual question in isolation, but if the answer is “no” for most questions, and especially for combinations of answers from multiple questions, then the evaluation instrument will be a failure. A survey design that creates a situation where a particular combination of answers leaves us knowing only “either Y or Z is true”, when the difference between Y and Z is important, is a bad survey design. Previous studies relating to this topic have generally (perhaps intentionally) failed to distinguish between successful warning and infliction of emotional violence, declaring to have found the former based on data that equally (or better) supports the latter.
3. It would be very useful to test whether graphical labels can effectively communicate that other nicotine products are far less risky than cigarettes.
The Federal Register entry includes not only the previous quote from FDA that warning labels should provide accurate information, but an outside comment (which FDA reports and implicitly agrees with) that warning labels should communicate about cessation methods. Since switching to a low risk alternative nicotine product is a very effective method of smoking cessation for many users (many of us conclude from the substantial evidence that exists that it is the most effective method), but the major barrier to switching is inaccurate risk perceptions, learning whether graphic warnings could inform users about that option would make an extremely valuable contribution to public health. Indeed, if the graphic warnings prove ineffective at improving accuracy of perceptions about smoking risk (which seems quite possible given previous studies), and even emotionally violent graphics have little long-term effect other than inflicting distress (which seems quite likely given previous studies), then learning about ways to communicate harm reduction information could be the only contribution of these labels to public health. Unlike the risk for lung disease, which is known by basically the entire U.S. population, only a small minority knows about the low risk of other nicotine products.
Furthermore, it is possible for label graphics (particularly the ones that are not legitimate warnings) to convince current smokers that they are already doomed, creating a fatalism that discourages cessation. Research might show that labels that encouraged switching to low-risk products can help counter such unfortunate messages.
We proposed simple analog quantitative graphics that could communicate this information in Bergen and Heffernan (2010). Other candidates could easily be created. We are not aware of any tests of any graphic warnings that communicate this information. Such tests would be very feasible within the context of the proposed study, and would not substantially take away from other study goals. If the FDA is genuinely interested in reducing the health costs of smoking and learning about major questions that we do not already have good answers to, it should include one or more graphics that try to communicate comparative risks and test their effectiveness.
Since learning about these labels would in no way create an obligation to use them, debates about the wisdom of promoting tobacco harm reduction have no bearing on the present point: This is a research study, not a proposed implementation. It would be useful to know whether the right graphical label could effectively educate people about the potential for harm reduction, even if it is later decided that such is not among the goals for the labels. Since much of the anti-harm-reduction rhetoric is based on the claim that it is not possible to accurately communicate the relevant information or that smokers are not interested in learning about comparatives risks, only if we learn what the label would accomplish can there be a legitimate informed discussion of whether it is appropriate. Failure to take the opportunity to confirm or deny those claims, by testing a label that communicated about the option of cessation by switching products, will be interpreted as intentionally avoiding finding out the truth about how effective that could be.
Hammond D.; Fong, G. T.; McDonald, P. W.; Brown, K. S.; and Cameron, R. (2004). Graphic Canadian Cigarette Warning Labels and Adverse Outcomes: Evidence from Canadian Smokers. American Journal of Public Health, 94(8).
Magnan, R.E.; Köblitz, A.R.; Zielke, D.J.; & McCaul, K.D. (2009). The Effects of Warning Smokers on Perceived Risk, Worry, and Motivation to Quit. Annals of Behavioral Medicine 37, 46–57
Slovic P. (2001). Smoking: risk, perception, policy. Sage Publications.
Thrasher J. F.; Rousu, M. C.; Anaya-Ocampo, R.; Reynales-Shigematsu, L. M.; Arillo-Santillán, E.; Hernández-Ávila, M. (2007). Estimating the impact of different cigarette package warning label policies: The auction method. Addictive Behaviors 32, 2916–2925