Objective To determine the extent and implications of internal human electroencephalography (EEG) research conducted by the tobacco industry.
Methods This study analysed internal documents that describe the results of human EEG studies conducted by tobacco manufacturers. Emphasis was placed on documents that pertain to the application of EEG to product evaluation efforts.
Results Internal EEG research was used to determine dose-response relations and effective threshold levels for nicotine, emphasising the importance of form and mechanism of nicotine delivery for initiating robust central nervous system (CNS) effects. Internal studies also highlight the importance of human behaviour during naturalistic smoking, revealing neurophysiological markers of compensation during smoking of reduced nicotine cigarettes. Finally, internal research demonstrates the effectiveness of EEG for the evaluation of non-nicotine phenomena including smoke-component discrimination by smokers, classification of sensory characteristics and measurement of hedonics and other subjective effects.
Conclusions Tobacco manufacturers successfully developed objective, EEG-based techniques to evaluate the influence of product characteristics on acceptance and use. Internal results suggest that complex interactions between pharmacological, sensory and behavioural factors mediate the brain changes that occur with smoking. These findings have implications for current proposals regarding the regulation of tobacco products and argue for the incorporation of objective measures of product effects when evaluating the health risks of new and existing tobacco products.
- tobacco products
- smoker behaviour
- harm reduction
- nicotine products
- nicotine reduction in cigarettes
- tobacco products
Statistics from Altmetric.com
- tobacco products
- smoker behaviour
- harm reduction
- nicotine products
- nicotine reduction in cigarettes
- tobacco products
Much tobacco-related disease and premature death may be considered a side effect of tobacco addiction, which drives persistent exposure to tobacco toxicants.1–4 While nicotine is the primary pharmacological component in tobacco and is necessary to the development of tobacco addiction, nicotine also interacts with other product characteristics that contribute to tobacco use and subsequent harm by (1) encouraging experimentation and initiation, (2) facilitating nicotine delivery and exposure and (3) masking negative product characteristics that could discourage use.
The difficulty of studying the interactions among product characteristics contributes to our incomplete understanding of the factors that most influence tobacco use. Though animal models have contributed to knowledge of the role of nicotine in tobacco addiction, many aspects of tobacco use are more accurately described through human-derived data. To this end, human neurophysiological techniques, including electroencephalography (EEG), positron emission tomography (PET) and functional MRI (fMRI), have been used to assess the central nervous system (CNS) effects of product use including the biochemical and physiological correlates of nicotine dosing, dependence and withdrawal, as well as the effects of nicotine and smoking on information processing and affect.5
Relative to fMRI and PET, EEG features low cost, ease of implementation, and reduced constraints on naturalistic smoking, and has a long history of use in nicotine and tobacco research.6 The spatial resolution of EEG is limited, but its superior temporal resolution has proved useful for studying group differences and time-based manipulations. Stimulants of all classes, including nicotine, have been shown to cause EEG changes associated with brain arousal.7 For example, Pickworth et al8 reviewed literature on the relation between EEG and psychomotor and cognitive performance during nicotine withdrawal and administration. The abundance of evidence indicates that tobacco abstinence causes EEG changes in the direction of cortical hypoarousal and performance decrements, while nicotine (and tobacco) reverses these effects. Along these lines, administration of tobacco or nicotine to non-deprived smokers has very little effect on EEG and performance, further supporting a role for nicotine in maintaining the arousal state.
EEG has also provided insight into the effects of nicotine dosing and the nicotine delivery mechanism. Using EEG, Teter et al9 found that smoking was more effective than nasal nicotine in producing significant changes in brain activity, and suggested that the reduced physiological effects may explain the low success rate for nicotine nasal spray as an aid to quitting. Pickworth et al10 observed that nicotine delivery, recency of smoking and process of smoking influenced the EEG—while not ruling out that non-nicotine components of tobacco smoke may also exert EEG effects. In an earlier study, Pickworth et al11 had found that denicotinised cigarettes did not activate the EEG, although they did reduce subjective measures of tobacco craving and withdrawal.
A limited set of peer-reviewed studies from the tobacco industry suggests that human neuroimaging, and EEG in particular, is employed by tobacco manufacturers for product evaluation. Examples include a 1992 study comparing a light cigarette and a cigarette with reduced nicotine;12 and a more recent study focused on denicotinised menthol cigarettes.13 In both cases, the authors concluded that EEG effects were primarily a function of nicotine rather than other components of smoke.
In a review of internal tobacco industry documents, Ferris Wayne et al14 identified EEG-based methods employed internally to assess the neurophysiological effects of tobacco. Among the methods used, manufacturers emphasised event-related potential (ERP) or evoked potential (EP) EEG protocols,14 15 during which subjects are presented with discrete, time-based stimuli and changes in the EEG waveform are recorded and analysed. Industry scientists characterised visual, somatosensory, auditory, nasal and even trigeminal evoked potentials to ascertain the influence of a range of stimulus characteristics on human physiology.14
Despite the promise of human EEG for the evaluation of product differences, and the evidence suggesting the use of this technology by manufacturers internally, no systematic study of industry EEG use has been conducted previously. In the current study, we review the tobacco industry's application of EEG techniques as a means to evaluate product characteristics and product differences. Our findings provide insight into the cigarette characteristics that contribute to product acceptability and underscore the value of EEG-based analysis in the evaluation of products.
During June and July 2007, the authors employed searches of the more than seven million internal tobacco industry documents made publicly available by the major US tobacco manufacturers following both state litigation and the 1998 Master Settlement Agreement (MSA). In accordance with the terms of the MSA, ongoing civil litigation results in the release of additional industry documents which are added to archival databases such as those maintained by Tobacco Documents Online (http://www.tobaccodocuments.org), the British American Tobacco Documents Archive (http://bat.library.ucsf.edu), and the Legacy Tobacco Documents Library (http://legacy.library.ucsf.edu), the website primarily used for this study. Documents were retrieved by conducting web-based searches for keywords and key phrases against both full-text and a number of indexed document characteristics including title, author, date, document type and others.
Document searches began with queries for any of the following phrases, words or abbreviations: ‘evoked potential’, ‘evoked response’, ‘event related potential’, ERP, PREP, ‘pattern reversal’, EEG, electroencephalography, electroencephalographic, EEG, electrophysiology, electrophysiological, ‘brain wave’, ‘brain activity’, psychophysiology or psychophysiological. Advanced search options were employed to reduce the number of unrelated documents generated by the initial search (eg, ‘EEG’ was excluded if present in the ‘author’ or ‘persons mentioned’ fields as it would indicate initials). Documents generated from this initial list of terms were reviewed for additional, relevant phrases, synonyms and project and people names which informed subsequent searches.
A document was considered relevant if it provided insight into the rationale for, or the conclusions drawn from, studies of electroencephalography in humans. Particular emphasis was placed on documents that discussed the potential applications of research to product development, evaluation and marketing and future research and development objectives. Internal studies and documents were not generally conducted to meet the stringent requirements of peer-review for independent scientific work; nonetheless, findings from documents were included if considered illustrative of internal research priorities and goals. From the thousands of internal documents referencing electroencephalography, these criteria resulted in a final set of approximately 140 highly relevant documents, gathered from the major tobacco manufacturers Philip Morris (PM), RJ Reynolds Tobacco Company (RJR) and British American Tobacco (BAT). Creation dates for these documents were as early as 1972 and as recent as 2002, after which no additional mention of internal EEG research was observed. Note that this is most likely because of reduced document availability rather than any observable change in industry practices.
Objectifying the subjective: internal validation and application of EEG
Throughout the industry, electroencephalography was recognised as a unique tool ‘to objectively quantify physiological responses to smoke constituents and tobacco flavorants’.16 Numerous internal studies applied EEG-based measures to product evaluations that, traditionally, had relied on self-report, including responses to commercial and experimental cigarettes; sensory substitutes for nicotine; and comparisons between smoke constituents and tobacco flavorants.17 A 1983 PM report cited challenges, like response bias, posed by smoking panellist evaluations and proposed EEG as a solution, being a ‘technique that was objective, non-verbal and quantifiable using standard parametric statistical techniques’.18 Researchers at RJR suggested a similar strategy, noting the growing published evidence that sensory variables such as odour pleasantness can differentially impact psychophysiological measures.19
Experimental results validated this approach: EEG was used successfully to differentiate cigarettes made with burley versus bright tobacco20; to compare flavours and smoke constituents21 22; and to evaluate responses to secondhand smoke.21 Electroencephalography allowed objective data to be paired alongside subjective evaluations18 and appeared to be more sensitive than some subjective measures. As summarised by PM scientists in 1990: ‘Our techniques are capable of detecting differences that are not apparent using normal subjective evaluations. Although subjectively undetectable, these differences most likely affect consumer acceptability and preference’.23
Acceptability and hedonics
As discussed in a 1974 BAT report, a particular appeal of EEG studies was the potential for insight into ‘the elusive factor of ‘satisfaction' in smoking'.24 In a 1989 study, PM researchers observed a strong positive correlation between positive subjective report and the amplitude of the P1–N2 component of visually evoked potentials (VEPs).25 This correlation led the authors to speculate that VEPs might be used to predict cigarette acceptability. Along these lines, a 1993 proposal from PM suggested comparing cigarettes with varying tar levels but constant nicotine concentrations; the expectation being that P1–N2 amplitude changes should reflect the acceptability of the various tar/nicotine ratios.26 The results of this proposed research could not be located among industry documents, and no published studies have linked the EEG effects of particular cigarettes with acceptability, demonstrating the need for additional research.
The complex interactions between nicotine, menthol and impact were also a topic of considerable internal interest.23 Electrophysiological measures were used as objective correlates of the subjective effects produced by menthol/nicotine interactions.22 As discussed further in a previous review of internal documents,27 P1–N2 amplitudes correlated both with perceived impact (p<0.05) as well as liking scores (p<0.001) for a variety of menthol and non-menthol cigarettes,28 29 again suggesting that EEG could be used as a tool to evaluate product acceptability. Published reports describing the pharmacological effects of menthol are limited but some evidence suggests that menthol has no influence on brain activity in the absence of nicotine19; however, these internal findings suggest that EEG could be used to evaluate synergistic effects between nicotine and menthol in cigarettes of typical yields.
Smoke component discriminability
A 1984 PM report stated that ‘one of the most important problems’ facing the company with respect to flavour evaluation was whether smokers can accurately identify product and smoke differences. The EEG was proposed as a means to address this issue: ‘We reasoned that since different compounds would probably stimulate distinct patterns of neural discharges in the chemoreceptors, the resulting cortically recorded electrical responses (ie, the EPs (evoked potentials)) should also be distinct’.30 The researchers found that the P300 component of the EEG was smaller and occurred after a greater latency when a target stimulus was difficult to distinguish from a background stimulus. The findings confirmed that information was available ‘not only about absolute discrimination, but also about degree of difference’ and that evoked potentials should enable researchers ‘to objectively determine whether, and to what degree, individuals can make flavor discriminations’.30
PM researchers also compared these EEG-based measures of odorant discriminability with traditional subjective responses. The authors concluded that evoked potentials were a more sensitive measure, noting:
The data suggest the possibility that a rather precise measure of stimulus concentration is made by the brain, but some of this precision is lost during the interpretive process involved in psychophysical judgment. Although preliminary, the results of the current study are very encouraging. It appears that EPs [evoked potentials] can provide objective information about stimulus intensity. In fact, the data suggest that in some instances, EPs may be superior to psychophysical judgments in estimating concentrations.18
PM researchers came to this conclusion after isolating a late positive component (LPC) in the EEG waveform which was ‘demonstrated to be related to the P300 wave of the auditory, visual and somatosensory (evoked potentials)’.31 The authors observed a strong correlation between the amplitude of this LPC and the discriminability of two olfactory stimuli. Specifically, as the discriminability of the stimuli increased, so did the amplitude of this EEG component. Furthermore, the measure provided a more sensitive index of stimulus discrimination, with differences in LPC amplitude observed even during trials when subjects made mistakes. Future plans sought to validate and apply this methodology to flavour evaluation,22 although no evidence confirming implementation was found.
This research was also applied to comparisons of product constituents. For example, a PM proposal described plans to test subject responses (n=10) to natural and synthetic menthol, as well as other menthol-like compounds, in order to identify differences in discrimination.21 32 Ultimately, this research was used to evaluate which of several alternate menthol sources were most like that of the menthol sources in use by the company.23 Taste discrimination studies were developed for other product characteristics, such as tobacco types and flavours.31 An example of the latter—mapping of trigeminal and olfactory stimulants—is described in greater detail in a review by Megerdichian et al.15
Behaviour and nicotine interact to influence the CNS effects of smoking
EEG provided a direct approach to evaluate the thesis that smokers use smoking to control brain physiology. As a 1974 BAT report explains: ‘If we can understand why individuals or groups self-administer particular amounts of nicotine at particular times, then our chances of optimising our product will be increased’.33 Interestingly, CNS responses to smoking did not depend on nicotine concentrations alone. Instead, internal studies provided evidence that nicotine dosing, nicotine form and smoking behaviour each influenced the neurophysiological effects of tobacco use, consistent with published reports.34 35
Nicotine and dose response
As a primary objective, internal EEG research sought to define the relation between brain response and nicotine delivery levels.36 37 For example, a 1975 BAT funded study considered whether stimulant and depressant effects on EEG response could be obtained in the same subject by altering the dose of nicotine. In a preliminary trial (n=4), dose-response curves were constructed with intravenous nicotine. Researchers observed that lower doses (≤0.15 mg) produced a stimulant effect whereas higher doses (≥0.20 mg) produced depressant effects. The findings suggested a biphasic response to nicotine, although it was noted that each subject crossed the control line at different levels of nicotine.38
Industry studies also confirmed that human CNS response varied as a function of nicotine delivery during controlled smoking. In a 1981 PM trial, (n=10) a no nicotine, low nicotine (0.14 mg) and high nicotine (1.34 mg) cigarette were compared. The sham cigarette produced no pre-post differences in the amplitude of the N1 component of the VEP, while statistically significant amplitude decreases were observed for only the high nicotine condition (p<0.05).39 Another study (n=20) compared six nicotine levels ranging from 0.12 mg nicotine to 1.10 mg using a controlled smoking procedure (figure 1). The electrophysiological effect of smoking the lowest nicotine cigarette (0.12 mg) were indistinguishable from the nicotine-free cigarette, in contrast to the cigarettes of 0.21 mg or greater. The authors commented that if the EEG latency effects seen were related to the reinforcing properties of cigarettes, then smoking cigarettes at or above 0.21 mg may sustain use, whereas smoking of the 0.12 mg cigarette may promote cessation.36 Interestingly, this threshold concentration of 0.21 mg is comparable to nicotine levels predicted to promote long-term use.41
PM researchers compiled data from a collection of dose-response studies (28 subjects, investigating five experimental cigarettes) in order to construct a theoretical best-fit curve relating the latency of the P1 component to nicotine delivery.42 P1 latency was found to be sensitive to nicotine delivery and to decrease in a dose-dependent fashion as nicotine was increased. In addition, the curve suggested that latency decrease, as a function of nicotine concentration, was largest up to 0.4 mg per cigarette, and that that no further latency shifts occurred beyond approximately 1.4 mg per cigarette, suggesting a dynamic range for the brain effects of nicotine that matches published observations.43
Nicotine form and mechanisms of delivery
A series of VEP studies were conducted at PM to assess CNS effects resulting from changes to the chemical form of nicotine (ie, free vs bound). For example, a 1989 study compared the effects of cigarettes made from denicotinised tobacco oversprayed with nicotine as either the base or the citrate. The CNS effects, as reflected by P1 latency, obtained using cigarettes oversprayed with base nicotine were comparable to those obtained using cigarettes of the same mainstream nicotine delivery, while the CNS effects obtained using the citrate were approximately half the magnitude.44 In another study, three prototype cigarettes were tested that were made with filler containing zero, 1% or 2% calcium hydroxide (a base). Despite equivalent smoke nicotine deliveries, VEP latencies decreased for treated cigarettes versus control.45 The studies taken together indicated that larger CNS effects could be obtained with more basic (higher pH) filler and smoke.46 In a separate report the authors concluded:
…it might be possible to optimise electrophysiological and subjective responses to cigarettes by modifying filler pH….findings suggest that cigarettes produced from such fillers would be perceived as having greater impact than would be predicted, based upon nicotine delivery. Also, our findings suggest that electrophysiological responses would be enhanced relative to cigarettes with normal filler pH.47
These results inspired a 1998 proposal for future experiments that included comparing the effects of Marlboro Light-type cigarettes containing added ammonia (a strong base) versus the same cigarettes without added ammonia; as well as comparing the puff-by-puff effects of cigarettes made from calcium hydroxide-treated filler.48 The results of these proposed studies could not be located among available documents; however, additional internal document research describes evidence linking smoke pH modifications and subjective evaluations,49 suggesting that these modifications are likely to influence EEG.
Internal studies also demonstrated that mechanism of nicotine delivery, as well as nicotine form, can influence physiological responses in a manner that is independent of nicotine concentration. For example, a 1985 PM study measured responses after puffing on an early model of a non-burning smoking article, yet no EEG differences were observed at any nicotine dose level.31 An additional experiment (n=10) assessed a collection of commercial cigarettes under controlled smoking conditions and determined that nicotine concentration strongly predicted evoked-potential latency (figure 2). An exception was the controversial Barclay brand, which was designed to facilitate human smoke delivery at a level greater than indicated by standard machine testing.50 The researchers observed ‘…the results obtained with Barclay are greatly at odds with what would be predicted on the basis of F.T.C. (ie, machine-measured) nicotine delivery data. Our data are in agreement with other tests conducted at R&D and elsewhere which indicate that Barclay's actual delivery level is close to that of a Merit’.42
Behavioural influences on smoking effects
Numerous internal studies demonstrated that smoking behaviour is highly capable of influencing the CNS effects of smoking. Providing evidence of this, an early PM study examined whether the effects of a higher nicotine delivery cigarette (0.9 mg) could be replicated by having subjects smoke three lower delivery cigarettes (0.3 mg nicotine). The latency of the N1 waveform component was mimicked by the three lower nicotine cigarettes while the amplitude of the P1 component was unaffected, suggesting that ‘the N1 amplitude effect depends on a single, relatively large intake of nicotine over a short time interval’. When three 0.1 mg nicotine cigarettes were compared to a single 0.3 mg cigarette, the latency effects were no longer similar (p<0.05). The authors concluded that the neurophysiological effects of nicotine exhibit ‘a threshold […] somewhere between 0.1 and 0.3 mgs’: a result consistent with nicotine dosing studies described above.42
Another PM study (n=10) assessed the P1 latency effects of various commercial cigarettes during controlled versus ad-lib smoking, in order to ‘gain insight into how individuals modify their smoking behaviour in response to the unique characteristics of a particular cigarette’.42 The cigarettes ranged in nicotine delivery from 0.11 to 1.04 mg/cig; and included the aforementioned Barclay cigarette. Findings indicated that smokers can achieve CNS effects from ultra-low ‘tar’ cigarettes comparable to that of full flavour cigarettes; and that nicotine delivery is less variable under ad-lib conditions than under controlled conditions: a result consistent with some published reports.35 These observations also confirmed the unique design of the Barclay cigarette under ad-lib smoking conditions (figure 2).
During the same study, the P1 latency shift for the lowest nicotine delivery cigarette (0.11 mg/cig) was far greater during ad-lib smoking than during controlled smoking, suggesting that subjects had greatly modified their smoking behaviour. The study authors concluded:
These findings, we feel, are important in several respects. First, the results obtained with the ad-lib smoking of cigarette A [the low nicotine cigarette] strongly suggest that smokers can achieve central nervous system (CNS) effects with ultra-low delivery cigarettes comparable to those obtained with high-delivery cigarettes… Second, the data indicate the possibility that smokers might modify their smoking behavior in order to obtain some optimal CNS levels of nicotine.42
Internal EEG research also evaluated the relation between CNS effects and onset of smoking.16 In one 4-hour ad-lib smoking study, PM researchers analysed the time course of the P1 latency in an attempt to isolate predictors of cigarette smoking rates. Although not reaching statistical significance, the findings suggested that P1 latency returned to the pre-smoking baseline approximately 50–60 minutes after smoking, just before subjects began smoking their next cigarette. The authors concluded that smokers ‘may be prompted to smoke their next cigarette when the CNS effects of the previous cigarette have begun to wane’; and that ‘although the effects of smoking a single cigarette are short-lived, they are maintained by repeated smoking’.18 From a separate dose-response study that utilised VEPs (n=10), it was concluded that ‘when an individual has been smoking regularly throughout the day, the interval between cigarettes necessary to achieve CNS effects may be lengthened’.51
Regulatory efforts supported by the WHO Framework Convention on Tobacco Control, as well as by legislation recently signed into law within the US granting authority to the US Food and Drug Administration over tobacco products52–55 open the door to new product standards and represent opportunities for the public health community to influence cigarette design. The recent cessation of commercial cigarette testing by the US Federal Trade Commission (FTC) highlights the conspicuous absence of effective public measures of tobacco product risk. For regulators to understand the health risk posed by tobacco products, they must apply new methods that are able to identify and assess those characteristics that support addiction and long-term use.
The current study describes how the industry developed objective, EEG-based techniques to evaluate the influence of product characteristics, both nicotine-based and those separate from nicotine, on acceptance and use. Internal research successfully translated established EEG-based paradigms of stimulus discriminability56 to the evaluation of cigarettes, determining smoker discrimination thresholds for product differences. In addition, industry researchers isolated EEG correlates of sensory characteristics and subjective evaluations of product acceptability, enabling accurate prediction of smoker responses to flavorants, smoke components, and other design differences. These applications of EEG to product evaluation are unique to the published literature and suggest new research strategies for exploring the mechanisms of product acceptability and use.
Beyond providing a framework for future research, industry EEG results provide evidence that the neurophysiological consequences of smoking are mediated by sophisticated interactions between pharmacological, sensory and behavioural factors, supporting the existing literature.12 57–59 Internal documents demonstrate that the CNS effects of nicotine are non-linear and reveal distinct thresholds of action. In addition, form and mechanism of nicotine delivery, as well as interactive effects between nicotine and other compounds, each influence the CNS response to smoking. Indeed, EEG provided manufacturers with more information about the ‘impact’ of a cigarette on human neurophysiology than nicotine concentrations alone. Internal research also highlights the profound influence of human behaviour on the neurophysiology of smoking. Smokers are able to compensate for reduced nicotine concentrations in order to achieve desired CNS effects: findings that agree with published studies demonstrating compensation during naturalistic smoking.35 59 60
Although many tobacco harm reduction efforts have focused on lowering toxic emissions, the high level of toxicity and the uniform range of toxins across cigarettes after adjusting for compensatory smoking behaviour have made such efforts largely unsuccessful.61 Another approach to reducing harm would be a product use reduction approach, which would depend on identifying and regulating product characteristics that are intended to increase impact and acceptability, support smoking behaviours and sustain use. While such an approach could have greater impact on health if it results in reduced dependence and use, it requires a better understanding of products and the effects of specific product characteristics (such as additives and flavours) than is currently available.
The internal findings described in the current study highlight the value of objective techniques, such as EEG, during the evaluation of new and existing tobacco products and suggest that EEG methods are at least as informative as subjective reports in predicting acceptability. If the public health community and tobacco regulators are to understand products, emphasis should be placed on objective, reliable predictors of product adoption and ongoing use. Identification of characteristics that correlate with product addictiveness would bring an effective regulatory strategy closer to reality and would serve as empirical measures of health risk.
What this paper adds
By allowing measurement of the central nervous system (CNS) effects of tobacco products, human electroencephalography (EEG) can provide insights into the physiological correlates of tobacco dependence. Previous reviews of tobacco industry documents reported the use of EEG research internally as a means to understand the neurobiological underpinnings of tobacco use; however, no systematic survey of industry EEG applications has been conducted.
In this study, we review industry documents and find that results of internal EEG studies suggest that the central nervous system effects of smoking are influenced by complex interactions between three primary factors: nicotine dosage, nicotine form and smoker behaviour. The findings also highlight the value of EEG-based measures for product evaluation and argue for the use of human neurophysiological techniques to characterise the health risks of new and existing tobacco products.
Funding Funding for this research was provided through the National Cancer Institute, grant RO1-CA87477-08. Other funders: National Cancer Institute, grant RO1-CA87477-08.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.