Conceptualizing Health Behaviors as Acute Mood-Altering Agents: Implications for Cancer Control

A collection of highly prevalent health risk behaviors (e.g., low physical activity, high sedentary behavior, unhealthy diet, tobacco, and alcohol use) jointly accounts for a massive portion of cancer burden. It is estimated that over half of cancer cases are attributable to a small set of modifiable behaviors including tobacco use, excess body weight and obesity, poor diet, and lack of exercise (1). Despite their known health risks, engagement in health-damaging behaviors remains common in the population (2–5), including a sizeable group who engages in multiple health-compromising behaviors, which compounds cancer risk (6). Identifying modifiable determinants of cancer risk and prevention behaviors (i.e., cancer control behaviors) is critical for the development of effective interventions and policies that substantially reduce and eliminate the excess numbers of behaviorally determined and preventable cases of cancer in the population.

When approaching the topic of behavior change, many cancer prevention scientists and interventionists often overlook the fact that the products driving various cancer risk behaviors, such as cigarettes and processed food, are engineered to be emotionally attractive and that many cancer prevention behaviors, such as exercise, can be emotionally unpleasant. Rather, the field has traditionally focused on methods to increase knowledge about the health risks of behaviors and modifying cognitively based beliefs and perceptions about what a rationally thinking person “should” do to limit cancer risk. Although many of the resulting interventions and policies showed initial success, advancements in cancer control science have stalled in recent years. Interventions have been effective at increasing intention to adopt cancer-reducing behaviors, yet have not universally had large effects on actual behavior change (7, 8). Furthermore, scientific understanding has stagnated, as the development of disconnected single-behavior programs of research (e.g., interventions focused on exercise, but not tobacco or diet) has led to a fragmented literature base. The opportunity to leverage basic behavioral science principles to identify common mechanisms that explain and predict engagement in numerous types of cancer risk and prevention behaviors has been largely overlooked. Consequently, integrated and parsimonious frameworks for cancer control behavioral research are lacking.

In this commentary, we propose that most cancer control behaviors act as “acute mood-altering agents,” which motivate individuals to pursue or avoid those behaviors in the future. We first describe a conceptual and scientific framework for understanding and studying affective mechanisms underlying cancer control behaviors. Next, we selectively review previous studies of the nature and phenomenology of mood-altering effects of different cancer risk and prevention behaviors, as well as their reinforcement and predictive value from three distinct, yet interrelated disciplines: addiction science, exercise science, and behavioral nutrition. We then propose central research questions that can address the validity and utility of the framework. Finally, we discuss how this framework can be applied to address critical scientific questions in cancer prevention and control.

The basis of the current conceptual framework illustrated in Fig. 1 is that most cancer control behaviors (and their accompanying consumer products) act as robust “acute mood-altering agents,” which signal neural systems that monitor the ongoing homeostasis of the body and lead to the manifestation of an emotional response that contributes to further pursuit or avoidance of that behavior. This framework builds off tenants of incentive salience (9), psychologic hedonism (10), affect processing (11), classical conditioning (12), and the affect heuristic (13), which suggest that exposures to stimuli can lead to feeling states that drive future behavior, and extends these theories to cancer control behaviors. We propose that cancer risk and prevention behaviors produce a range of affective responses (occurring during the behavior) in reaction to the experience of “doing” the behavior or “missing out” on alternative behaviors. Engaging a behavior (i.e., smoking a cigarette and exercising) can lead to biological reactions (e.g., metabolic, cardiovascular, hormonal, or immune), which may cause neurotransmitter-mediated physiologic sensations (e.g., pain or pleasure) that may interact with external or internal stimuli (e.g., being alone and lack of sleep the prior night), leading to emotions and conscious feelings (e.g., excitement, joy, fear, and frustration; ref. 14). Linkages between engagement in cancer risk/prevention behaviors and the emotional responses to those behaviors become learned through associative memory processes (15). Future exposure to behavioral cues (e.g., seeing other people smoke and being asked by a friend to go exercise) triggers these learned behavior–affect associations, which lead to affectively charged motivations (e.g., desires, urges, craving, and dreading) to pursue or evade the behavior. Similar to the concepts of hedonic motivation (16) and Pavlovian motivation (17), affectively charged motivations involve dopamine-mediated, evolutionary-hardwired drives or impulses to approach or avoid a target behavior that are distinct from cognitive forms of motivation such as expected utility and value. Following this framework, affective response to cancer control behaviors can vary widely between people based on individual differences (e.g., age, body weight, poverty, and diagnosed depression) and within people based on features of the external context (e.g., pleasantness of the physical setting), and background internal states (e.g., fatigue, hunger, and stress).

As shown in Fig. 1, cancer risk behaviors such as sedentary behavior, poor diet, tobacco and alcohol use, sunbathing, and risky sexual behaviors persist because they robustly and immediately (and temporarily) improve affect (i.e., trigger pleasant or alleviate aversive affective states), which is desirable. Through positive reinforcement (i.e., presenting a positive consequence), these desirable affective states experienced during the behavior trigger appetitive affectively charged motivations (e.g., desires, drives, urges, and cravings) when exposed to future behavioral cues, causing individuals to approach those behaviors. In contrast, people may avoid cancer prevention behaviors such as exercise, healthy diet, sun-safe behaviors, cancer prevention screening, and practicing safe sex because they worsen affect (i.e., trigger unpleasant or attenuate positive affective states). Experiencing these aversive affective states during behavior may have a positive punishment effect (i.e., presenting a negative consequence), which induces aversive affectively charged motivations (e.g., dread, fear, and avoidance) that deters participation in those behaviors when exposed to future behavioral cues.

Following propositions made by dual-process models (18), affective processes may interact in various ways with cognitive processes (e.g., intentions, goals, and plans) to jointly contribute to engagement in cancer control behaviors. Affectively charged avoidance motivations may attenuate the influence of intentions to perform a cancer prevention behavior such as exercise. However, enjoyment of a cancer prevention behavior (e.g., eating fruits and vegetables) can augment the effect of careful menu planning on the likelihood that those cancer preventing foods are consumed.

A critical piece of framework depicted in Fig. 1 is the operationalization of what constitutes an affective response that elicits the corresponding affectively charged motivation. Operationalizing affective responses is critical to informing research agendas aiming to test and apply the framework. While affect-based mechanisms and the affective responses involved in health behaviors has been described previously (19, 20), the paradigms have been fairly nonspecific. Studies on cancer risk and prevention behaviors typically characterize affective response to those behaviors using simple difference score paradigms (i.e., pre-postengagement in the behavior), measures that only tap into the subjective (i.e., self-report) channel of affect, and unidimensional instruments (i.e., simple bipolar continuums ranging from very positive to very negative). Insights from basic psychologic science have been under-leveraged in previous conceptualizations of how to characterize what constitutes a reinforcing or punishing affect response to a cancer control behavior. The proposed conceptual framework aims to break new ground by defining and characterizing, with detail and precision, what constitutes a reinforcing or punishing affective response profile, in real time while engaging in a cancer control behavior. Our proposed state of the science approach to characterizing affective response to behavior builds off advancements in basic behavioral science, including assessment of micro-temporal signatures, multichannel manifestations, and multidimensional phenotyping. As described below, we hypothesize that a wealth of empirically and conceptually germane data to explaining and predicting cancer risk and preventive behavior may be derived by exploring variation in the shape, typology, and quality of affective response profiles.

Our proposed framework focuses on behavior-specific affective responses occurring during the experience of engaging in a target behavior. As shown in Fig. 2, when individuals engage in cancer risk or prevention behavior, they may demonstrate different affective response profiles based on variations in time course (x-axis) and affective intensity (y-axis) after exposure to the behavioral stimulus. These responses may be characterized by differences in rapidity of response (i.e., ascending slope) and recovery (i.e., descending slope), volume (i.e., peak change from baseline or total AUC), duration (i.e., time period between initial response and return to baseline), and consistency (i.e., likelihood of experiencing a similar response profile with similar behavioral exposures over time). Affective responses may also be characterized in terms of the affect intensity occurring at standardized time points such as at 10%, 30%, or 70% into the total duration of the behavior. Separate response profiles may be observed for different dimensions of affect (e.g., valence vs. arousal). Capturing the microtemporal signatures of affective response profiles requires repeated and timed assessments across the duration of engagement in the behavior or its psychopharmacologic effects either in controlled laboratory environments (21) or in naturalistic settings through real-time assessment methods such as ecological momentary assessment (EMA; ref. 22).

The conceptual premise for detailed microtemporal characterization derives from understanding that absolute affect intensity (i.e., how someone feels in a certain moment irrespective of their previous affective state) provides incomplete information about motivational potential without incorporating the relative dynamics of affect (i.e., how one feels relative to how they recently felt; ref. 23). Consider, for example, the doubling of one's affective intensity that occurs gradually across a period of hours after engaging in a behavior relative a doubling of intensity occurring within minutes of a behavior. The latter situation is much more likely to generate a psychologic appraisal that the change in affect is due to the behavior, and therefore lead to a motivational tendency to approach or avoid the behavior. The former situation may generate a much more ambiguous psychologic appraisal about the behavior's motivational relevance, even if the affect is indeed caused by the behavior.

Affective responses to cancer control behaviors are manifested through a variety of channels. Emotional reactions to threat and reward can be captured through physiologic indicators (e.g., brain activity, heart rate, skin conductance, eye tracking, facial and expressions; ref. 24). Emotional responses may also involve implicit processes, which are automatic, instinctual, and operate outside of conscious self-awareness (e.g., Implicit Associations Task and Stroop Task; ref. 25). Furthermore, affective responses can be expressed as subjective states, which involve conscious awareness of physical sensations and feelings that can be self-reported through questionnaires (e.g., Positive and Negative Affect Scale; ref. 26 and Profile of Mood States; ref. 27).

At the basic level, subjective feelings can occur as core affective states that quantitatively vary along a valence dimension, which refers to the inherent pleasantness or unpleasantness of an experience (i.e., negative/bad to positive/good) and an arousal dimension, which refers to the intrinsic level of activation during an experience (i.e., deactivated/fatigued to activated/energetic; ref. 28). These broad-spanning core affective states may be further differentiated into narrower, qualitatively discrete feelings based on individual interpretations of stimuli or events (e.g., elation, serenity, tension, and dejection; ref. 29). Subjective affective states also are expressed as social feelings (e.g., pride and embarrassment; ref. 30), and physical sensations (e.g., pleasure and pain; ref. 31).

A comprehensive empirical characterization of affective response to cancer control behaviors in terms of their temporal signature across multiple channels and multidimensional phenotyping of specific feelings has yet to be performed. We suspect that many of these different parameters may each provide nonredundant information that allows scientists to predict future likelihood of engaging in the behavior and a precision target for cancer prevention interventions designed to modify affective responses to cancer control behaviors (see Conclusion section below for further details on potential applications of the framework). For instance, it is plausible that although two cancer risk behaviors may generate equivalent AUC values for a positive affective state, yet one behavior has a steeper ascending slope and peak change (i.e., reactivity) from baseline than the other (see Fig. 2). The behavior that produces the faster onset and more robust spike is expected to be more reinforcing.

Emerging research to support the proposed framework on affective response to cancer risk and prevention behaviors comes from distinct yet interrelated disciplines including addiction science, exercise science, and behavioral nutrition. Research in these fields shows that people respond emotionally to the experience of engaging in cancer control behaviors, and affective responses are predictive of future behavior. While the extant research reviewed predominately utilizes cursory characterizations of affective response and lacks many of the parameter assessments described above, they provide preliminary evidence supporting the premise of the current framework, as summarized below.

Laboratory-based human behavior pharmacologic studies show that individual differences in the affect reported immediately following administration of an addictive agent predicts continued use in the future. For example, greater positive affect following alcohol administration in the laboratory predicts the likelihood of binge alcohol use at a 2-year follow-up (32). In addition, greater reports of negative affect during tobacco smoking abstinence in the laboratory predicts earlier relapse during a quit attempt outside of the laboratory (33). Addiction studies have also found that affective response to drug use can vary across contexts and individuals. For example, laboratory research has demonstrated that social contexts (e.g., being in a room with other drinkers) can enhance the positive (and reduce the negative) affective effects of alcohol (34). Also, individuals with mental health vulnerabilities experience stronger positive affect enhancement and negative affect reduction while smoking, which aligns with observed disparities in the prevalence of smoking in the general population when stratified by mental health status (35). In addition, key theories and corresponding from addiction science data suggest that the ascending and descending slope of drugs' responses are key sources of variation that associate with underlying physiology and frequency of substance use, such as the “rate hypothesis,” which links the speed of drug absorption with its abuse liability and the “biphasic alcohol effects model,” which ties ascending and descending limbs of alcohol absorption with corresponding affect-enhancing and affect-reducing (sedating) effects (36–40). Although these data and theory support the premise that different components of temporal signatures provide unique information about the effects of motivationally relevant behaviors, a head-to-head comparison of whether rapidity, recovery, volume, duration, and consistency of affective response are all important in determining future cancer control behaviors is yet to be tested and remains an empirical question.

Physical activity may influence affect through a variety of mechanisms including the sensation of pain, the release of endorphins (41), modulating levels of brain-derived neurotrophic factors (42), or perceptions of the value of certain activities (43). The dual-mode model of affective responses to exercise (44) suggests that individuals experience homogeneous pleasure responses at laboratory-based low intensity physical activity and homogenous displeasure responses at high intensity physical activity, which has been replicated in naturalistic settings through EMA studies (45). However, there is considerable variability in positive and negative affective response at moderate intensity levels (46), which may be explained by contextual features of the physical activity setting (e.g., being with others or outdoors; ref. 22) or individual differences. Studies have shown that positive affective response during moderate intensity exercise is associated with higher levels of subsequent engagement in physical activity (47).

Laboratory and EMA studies in behavioral nutrition show that people have positive emotional responses to eating high-sugar and healthy foods (48, 49), and these responses may vary by amount of food consumed (50) and one's subjective experience of control over eating. Affective response to eating has been shown to reliably vary across individuals based on mental health and situational factors such as level of awareness or mindfulness (51). Studies on how emotional responses to eating predict future behavior or weight gain are lacking.

While there is evidence implicating affective responses in cancer control behaviors more generally noted above, detailed data on the validity and utility of obtaining detailed affective response profiles to cancer control behaviors aligned with the current framework is currently lacking. Such data are needed to inform necessary refinement of the framework (e.g., eliminating certain aspects of the temporal signature that do not provide incremental predictive value for identifying the future behavior). Key steps to test the validity and utility of the framework are outlined below.

We believe that developing standardized “population-level” estimates of the typical affective response profiles of prominent cancer control behaviors with known prevalence of engagement in these behaviors in the population (e.g., sugar sweetened beverage consumption, smoking standard combustible cigarette, and engaging in aerobic exercise) is an important first step in validating the framework. It would be expected that cancer control behaviors with higher prevalence of initiation and persistence in the population would show affective response profiles consistent with a desirable response. Behaviors with the lowest prevalence are predicted to show the least desirable (and most aversive) affective response profiles.

Characterizing variability in affective response to cancer control behaviors across different subpopulations would also add information about the influence of individual and group differences in personality, mental health, genetics, and environments on affective response. As a validation step, subgroups with higher prevalence of a cancer-risk behavior relative to the general population would be expected to show a more desirable affective response profile relative to the general population estimate.

It is also valuable to characterize cross-context differences in affective responses to cancer control behaviors to understand sources of within-subject variation. Individuals may differ in their affective responses to behavior depending on features of the setting or situation (e.g., social influences and physical environment). Understanding patterning of cross-context differences in affective response to cancer control behaviors may help explain why the prevalence of behaviors is higher in one type of setting over another.

Testing the validity and utility of the proposed framework also involves determining the extent to which affective response profiles predict subsequent engagement in cancer control behaviors. The reinforcing or punishing value of desirable and undesirable affective responses, respectively, can be tested by examining their associations with future behavior (e.g., next-day or next-year). Demonstrating the predictive validity of affective response profiles can help to identify who is at higher risk and how they can be prevented from experiencing cancer incidence, progression, and recurrence.

To establish a plausible causal mechanism through which affective responses influence future behavior, there need to be measurable alterations in affective processes that are proximal predictors of behavior including motivations, drives, desire, dread, urges, cravings, affective attitudes, and anticipated emotions. For example, functional MRI (fMRI) research may confirm the role played by mesolimbic dopaminergic systems for reward in mediating the effects of affective responses on future cancer control behaviors. A recent study found that when people saw pictures of food, fMRI signal in the nucleus accumbens, a key region in the brain's reward circuitry, predicts weight gain several months in the future (52).

Whether the proposed parameters of affective response profiles (e.g., rapidity, recovery, and duration) are influenced by experimental interventions that attempt to enhance or diminish affective response provides an important indicator of the degree to which they represent the intended construct. For instance, the nicotinic receptor partial antagonist smoking cessation medication, varenicline, would be expected to suppress the many parameters within the acute affective response profile to combustible cigarette smoking. Further utility of the proposed framework can be assessed by the extent to which an intervention that successfully experimentally manipulates and changes the affective response, in turn decrease the likelihood of future cancer risk behaviors and increase the likelihood of future cancer prevention behaviors.

Pending data on the validity and utility, the current framework could yield several key applications anticipated to directly impact the population's cancer burden.

The aforementioned affective response profiling strategy can be applied to new behaviors or products that are likely to influence cancer risk or prevention. Depending on the results of an affect profiling screen of such behaviors or products, we can anticipate their likely appeal (and population-level prevalence) prior to their being widely adopted or avoided in the population. Such information is vital to policies aimed at protecting the population from exposures that cause cancer. When a novel product arrives to the market place that has high cancer risk, rapid screening of affective response profiles to the product can inform policy makers whether regulations and other population-based public health measures are needed and should be a priority.

Novel interventions may reduce cancer risk behavior by changing their affective response profile to be less desirable. Some interventions for tobacco, alcohol, or poor diet exert their influence by reducing the affect-enhancing properties of these cancer risk behaviors. By developing a more precise method for characterizing components of the affect profile that are relevant to predicting future behavior, new cancer risk behavior reduction interventions can be screened in the laboratory or natural environment to determine whether administration of such interventions reduce the desirable affective response to the risk behavior.

Affective response profiles may explain population-level differences and perhaps disparities in the engagement in behaviors that affect cancer risk. By targeting environments and contexts that alter the affective response profile and differ across population strata (e.g., type of product, dose, and setting), policies, and personalized interventions may be able to module affective response parameters and reduce future cancer risk behaviors. For example, increasing the availability of nutrient-rich healthy foods that also generate desirable affective response parameters might be priorities in communities that lack access to such foods and have disproportionately high cancer rates.

This article presented a conceptual framework and summarized interdisciplinary evidence for how cancer control behaviors serve as acute mood-altering agents that reinforce or repel subsequent behavioral engagement. Pending the necessary evidence, the proposed conceptual framework and resulting evidence generated has the potential to promote a paradigm shift that advances prediction of the likelihood that new products are likely to be widely consumed and significantly alter cancer control behaviors and outcomes in the population; and development of individual, population, and public policy interventions that target affective mechanisms to prevent cancer risk behavior and promote cancer prevention behavior.

G.F. Dunton reports grants from University of Southern California and grants from National Heart Lung and Blood Institute during the conduct of the study; grants from NIH, personal fees from University of Maryland, personal fees from University of Connecticut, personal fees from Arizona State University, and other from National Academies of Science, Engineering and Medicine outside the submitted work. R.D. Pang reports grants from NIH and grants from TRDRP outside the submitted work. S.P. Eckel reports grants from NIH during the conduct of the study. No potential conflicts of interest were disclosed by the other authors.

Conception and design: G.F. Dunton, J.T. Kaplan, A.M. Leventhal

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.P. Eckel

Writing, review, and/or revision of the manuscript: G.F. Dunton, J.T. Kaplan, J. Monterosso, R.D. Pang, T.B. Mason, M.G. Kirkpatrick, S.P. Eckel, A.M. Leventhal

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G.F. Dunton

Study supervision: G.F. Dunton