By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Please be advised that we experienced an unexpected issue that occurred on Saturday and Sunday January 20th and 21st that caused the site to be down for an extended period of time and affected the ability of users to access content on Wiley Online Library. This issue has now been fully resolved. We apologize for any inconvenience this may have caused and are working to ensure that we can alert you immediately of any unplanned periods of downtime or disruption in the future.
York University, 343 Bethune College, 4700 Keele Street, Toronto, ON M3J 1P3, Canada. E-mail: email@example.com
Objective: To demonstrate that human overeating is not just a passive response to salient environmental triggers and powerful physiological drives; it is also about making choices. The ventromedial prefrontal cortex has been strongly implicated in the neural circuitry necessary for making advantageous decisions when various options for action are available. Decision-making deficits have been found in patients with ventromedial prefrontal cortex lesions and in those with substance dependence—impairments that reflect an inability to advantageously assess future consequences. That is, they choose immediate rewards in the face of future long-term negative consequences.
Research Methods and Procedures: We extended this research to the study of overeating and overweight, testing a regression model that predicted that poor decision making (as assessed by a validated computerized gambling task) and a tendency to overeat under stress would correlate with higher BMI in a group of healthy adult women (N = 41) representing a broad range of body weights.
Results: We found statistically significant main effects for both independent variables in the predicted direction (p < 0.05; R2 = 0.35). Indeed, the decision-making impairments across the 100 trials of the computer task were greater in those with high BMI than in previous studies with drug addicts.
Discussion: Findings suggested that cortical and subcortical processes, which regulate one's ability to inhibit short-term rewards when the long-term consequences are deleterious, may also influence eating behaviors in a culture dominated by so many, and such varied, sources of palatable and calorically dense sources of energy.
An estimated 60% of adults in the United States and Britain are currently overweight or obese—a marked increase from recent decades—and they are becoming so at an earlier age than ever before (1,2,3,4). Data from a number of reliable sources suggest that the pattern of population weight gain over the past generation has been due largely to an overconsumption of energy rather than a decrease in physical activity patterns (see (5)). However, the fattening of our society goes well beyond esthetic issues (6); 300, 000 people in the United States alone die each year of obesity-related illnesses (7), and this condition affects more people than smoking, heavy drinking, and poverty (8,9). More sobering is the prediction that, if current trends continue, 100% of Americans will be obese by the year 2230 (10).
Overeating as an Addictive Behavior
The evidence of common brain mechanisms mediating the rewarding properties of natural rewards (like eating) and addictive drugs supports the argument that food—especially when it is highly palatable—can be used for purposes that exceed basic energy requirements, and it has the potential for abuse (11). For instance, many people report using food to “self-medicate” a disturbed affect (12,13). Of relevance is research showing that sweet foods, like certain addictive drugs, can produce significant analgesic effects (14). Excessive food intake can also induce physiological responses that mimic those seen in drug addiction—viz. down-regulation, sensitization, and withdrawal (15,16,17,18). One study has demonstrated that repeated and high intake of sugar causes behavioral and neurochemical signs of withdrawal in rats when its availability is restricted (19). In addition, the fact that binge eating is often triggered by the ingestion of small amounts of a palatable food parallels the “priming” effect of drugs in addicts, whereby the initial ingestion of the drug tends to elicit a strong “craving” or compulsion for further use.
Although the tendency to consume an excess number of calories each day can be influenced by several factors, dietary variety, especially of energy-dense foods, has consistently been associated with increased body weight (20). A recent animal study has also found that overconsumption is induced in rats who have greater availability of palatable food (21). Tordoff believes that the rats in his study ate more simply because they encountered desirable food more frequently—what he calls “obesity by choice.” Given the substantial availability and variety of nutrient-poor, energy-dense foods in most Westernized countries, this finding has significance for understanding the increasing prevalence of human obesity (22,23).
Whether the pun in Tordoff's phrase (obesity by choice) was deliberate or not, it raises an interesting distinction between rodent and human eating behavior. Clearly, there is complex neural circuitry prompting physiological drives that regulate feeding behavior. However, in the human condition, the “choice” to eat (or not to eat) is under a degree of cognitive control not present in the rat. With ubiquitous warnings from the medical profession, it is fair to assume that, in our current culture, few adults are unaware of the health risks associated with obesity and poor nutritional status. Furthermore, the strong social stigma associated with obesity should also serve as a potent deterrent to overconsumption. In other words, unlike times in our evolutionary history when food was scarce and the next meal unpredictable, current human eating is not just a passive response to salient environmental triggers and powerful physiological drives; it is fundamentally about making a choice between short-lived and overabundant rewards in the face of a disadvantageous long-term outcome if the behavior is done in excess.
The ventromedial prefrontal cortex (VmPFC)1 has been strongly implicated in a neural system, including the amygdala-ventral striatum, that is necessary for making advantageous decisions when various options are available for action (e.g., (24,25)). In particular, its critical function in this process is to activate feelings or emotional states from “thoughts” about rewarding or punishing events that are not currently present in one's immediate environment (26,27). Much of the early research on decision making came from studying the deficits of patients with VmPFC lesions and learning that their impairment reflects an inability to advantageously assess future consequences, both positive and negative, so that their behavior is always guided by the immediate contingency (28).
Given the behavioral similarities between patients with VmPFC lesions and drug abusers (both make choices that bring immediate benefit despite long-term negative consequences), recent studies have investigated VmPFC function in drug addicts using a “gambling task,” which has demonstrated sensitivity in detecting the decision-making impairment of the VmPFC patients (29,30). Results have consistently demonstrated that many addicts show decision-making deficits (31,32,33,34). However, there is also considerable variability in the response of addicts, suggesting that their decision-making impairments could result from deficits in more than one neural process.
The present research extended these findings to the study of overeating and obesity. Based on the argument presented earlier that food—in particular, highly palatable food—can be conceptualized as a “drug,” we proposed that poor decision-making abilities also characterize many who are overweight. The purpose of the study was to test this hypothesis, employing the gambling paradigm used in the research described above, in a group of adult women representing a broad range of body weights. Our specific prediction was that those who perform poorly on the gambling task would tend to have higher body weights, mediated by their tendency to overeat. In other words, some individuals chronically overindulge, especially to ameliorate the effects of stress, despite explicit health warnings and an awareness of the adverse physical and psychological outcomes associated with excess adiposity—a clear and poignant example of poor decision making!
Research Methods and Procedures
Subjects and Procedure
As part of ongoing behavior genetics research at the Centre for Addiction and Mental Health in Toronto, Canada, 41 healthy adult women (mean age = 28.5 years, SD = 5.6) took part in this study. Subjects were recruited from advertisements posted at two large universities in Toronto and at local hospitals asking for volunteers to take part in a psychogenetics study that required the completion of a questionnaire and a simple computer task, and the giving of a blood sample. In order to recruit women with a broad range of body weight, some of the posters asked specifically for women “above average weight.” The large majority of subjects were either students (both graduate and undergraduate) or staff at these institutions. Exclusion criteria followed the recommendations for volunteers in biomedical research (see (35)). Briefly, subjects were excluded if they had a history or presence of Axis I disorder (36), if they were being treated for any serious medical condition, if they were taking any prescription medication with central nervous system effects, or if they were pregnant. They were also required to be fluent in English. All but two subjects were white of European background. Initially, subjects were contacted by telephone, at which time a detailed description of the study was given, they were screened for exclusion criteria, and a test appointment was made. On the testing day, and after giving informed consent, subjects completed the questionnaire package and the computer task and had their height and weight measured. At the end of the study, they were paid a small stipend (at the rate of $10.00/h) for their participation.
BMI [weight (kilograms)/height (meters squared)] was calculated from height and weight measurements taken with subjects standing in stocking feet. Subjects were also asked to indicate their highest ever adult weight, excluding pregnancy.
Emotional overeating was assessed by the Emotional Eating Scale (12). On the 20-item adjective checklist, respondents were asked to rate, on a 5-point Likert scale, the degree to which each mood state generates a desire to undereat (moderately or greatly), has no effect on eating, or generates a desire to overeat (moderately or greatly). The items form three subscales reflecting eating in response to anger/frustration, anxiety, and depression. The authors of the scale report good construct, discriminate, and criterion validity, as well as acceptable α coefficients.
Decision making was assessed by the computerized version of the Iowa Gambling Task (29). This is a paradigm designed to simulate real-life decision making in terms of reward and punishment and has been shown to be sensitive to the decision impairment found in patients with bilateral lesions of the VmPFC and those with substance dependence (32,33,34). Because a full explanation of the task has been given elsewhere (e.g., refs. (26) and (31)), only a relatively brief description follows.
Subjects had to choose among four decks of cards presented on a computer screen, two of which (A and B) yielded high immediate gain (but larger future losses), and two of which (C and D) yielded lower immediate gain but a smaller future loss. Subjects were informed that the goal of the game was to accumulate as much (play) money as possible by picking one card at a time from any of the four decks across 100 trials. Trials were scored in 5 blocks of 20 (see (31)), in which the number of cards from the disadvantageous decks (A and B) was subtracted from the cards in the advantageous decks (C and D). Therefore, a positive score reflected a tendency to make good decisions. To reflect improvements in performance (i.e., learning from experience), the net “good decisions” in the first block of 20 were subtracted from the net good decisions in the last block—higher scores reflecting overall better decision making.
Table 1 presents the means, SDs, and minima and maxima for all variables used in the analyses. To test the mediational model described in the introduction, we followed the procedures described by Baron and Kenny (37). Mediation is present when the following conditions are met: the independent variable (decision making) significantly predicts the proposed mediator (emotional overeating); the proposed mediator (emotional overeating) significantly predicts the dependent variable (BMI); and the independent variable (decision making) significantly predicts the dependent variable (BMI). However, this relationship is substantially minimized when the proposed mediator (emotional overeating) is added as an independent variable in the multiple regression model. The mediational model was tested in four steps according to the requirements outlined above. Table 2 presents a summary of the regression results for this set of analyses.
Table 1. Means, SDs, and minima and maxima of all variables used in the analyses
In the first analysis, the three emotional overeating subscales were regressed separately on the gambling task score. Results showed a positive trend in the relationship between the gambling task and emotional overeating (depression subscale; p = 0.074) but not with the anger/frustration or anxiety subscales. Although the relationship only approached significance, there was a suggestion that those who tended to make bad decisions also tended to overeat when feeling depressed more than those who were characterized by better decision making.
In the second analysis, BMI was regressed on emotional overeating. However, given our findings in the analyses described above, we regressed BMI only on the emotional overeating-depression subscale. Results confirmed a highly significant and positive relationship between these two variables.
In the third analysis, BMI was regressed on the gambling task score, and a highly significant inverse relationship was found. Those with gambling task scores reflecting poor decision making tended to have higher BMIs than those who made good decisions.
To illustrate more clearly the BMI-related differences in the decision-making learning curve across the 100 gambling task trials—similar to the graph shown by Bechara and Damasio (31) for VmFRC lesion patients, substance-dependent individuals, and normal controls—we dichotomized our sample, according to conventional BMI cutoff values (see (38)), into normal weight (BMI < 25) and overweight and obese (BMI = > 25). Figure 1 plots the mean scores for each group for each block of 20 trials. The graph shows a clear improvement in good decisions across the trials for normal-weight subjects (N = 26). On the other hand, the overweight and obese subjects (N = 15) showed no improvement in decision making across the trials. Indeed, the mean scores in each block show a net deficit in advantageous choices—performance that was even worse than that demonstrated by substance-dependent individuals in the study by Bechara and Damasio (31).
In the final step of the analysis, the model described in the third step was repeated with the proposed mediator (emotional overeating-depression subscale) added as an independent variable. The gambling task remained a statistically significant predictor in the model, and its parameter estimate was only moderately reduced in size. Therefore, we are obliged to conclude that, although emotional overeating did mediate the relationship between decision making and BMI, its effect in this regard was relatively small. A more appropriate conclusion is that both decision-making and emotional overeating had relatively independent relationships with BMI.2
Results from this study provided support for our hypothesis that poor decision making, as assessed by the gambling task, tends to characterize those who are overweight and obese. Indeed, as a group, the high BMI subjects showed greater performance impairments than the substance-dependent patients in the study by Bechara and Damasio (31). Poor performance on the gambling task has been linked to VmPFC malfunction because patients with lesions in this brain area have pronounced decision-making deficits compared with healthy subjects. Substance-dependent individuals also score poorly on the task, which is consistent with functional neuroimaging studies showing abnormal activity in the VmPFC cortices of various groups of addicts (e.g., (39,40)). However, in the case of addicted subjects, causality is difficult to establish in studies using cross-sectional data because it is unknown whether the VmPFC malfunction occurred as result of chronic and excessive substance use or whether it existed as a developmental, predisposing factor. Also, and importantly, Bechara and Damasio (31) emphasize that a developmental VmPFC malfunction does not cause addiction; rather, it provides a phenotypic characteristic that tends to render individuals “myopic” for future consequences and, therefore, more likely to be guided in the direction of immediately gratifying behaviors such as drug taking.
It has been suggested that two types of abnormality in the extended VmPFC system could explain the so-called “loss of control” seen in addictive behaviors (33). Primary inducers (stimuli), or conditioned inducers, automatically evoke a somatic state, like fear or pleasure, when they are present in the immediate environment—a process regulated by amygdala-striatal brain circuitry. Secondary inducers (stimuli), on the other hand, are generated by “thoughts” about rewarding and punishing events not currently present in the immediate environment, which also elicit positive or negative states when they are brought into conscious memory. The VmPFC is a critical substrate in the origination of the somatic states linked to secondary inducers (see (41)). Studies have shown that some addicts closely resemble VmPFC lesion patients; that is, they display an insensitivity to future consequences, both positive and negative (33). Other addicts, however, reflect a phenotype that generates a strong somatic response to primary inducers of reward but a weak somatic response to secondary inducers of punishment, so that their behavior is strongly guided by reward when it is immediately available, irrespective of the long-term consequences.
The results of the present study also showed, similar to other research (e.g., ref. (42)), that self-reports of overeating during periods of negative mood were significantly related to higher BMI. However, contrary to our expectations, we did not find that emotional overeating mediated the relationship between decision making and BMI in any substantial way. In other words, there was little support for the proposal that poor decision making contributes to an increase in BMI by fostering overeating during periods of negative emotion. It behooves us, therefore, to consider other means whereby poor decision making may lead to increased BMI. It is also important to acknowledge that although we imply a directional relationship between emotional overeating and BMI, by proposing that the former influences the latter, it is entirely possible that the relationship is bidirectional. An increased BMI could lead to more frequent periods of depressed mood and, therefore, more overeating.
Similar to the studies with substance-dependent subjects (31,33) it will be important in future research to establish whether overweight and obese individuals behave more like VmPFC patients who have diminished ability to assess future consequences or whether they are more likely to be hypersensitive to primary reward and hyposensitive to secondary punishment so that their behavior is dominated by the presence of immediately available rewards. Some evidence suggests that the latter is a stronger possibility. In one study linking reward sensitivity to eating behavior, adolescent girls with inherently high sensitivity to cues of reward were more likely to engage in bouts of binge eating (43). Clinical research has also found that patients with bulimia nervosa have significantly higher sensitivity to reward than the general population, and especially compared with patients with anorexia nervosa, who have very low sensitivity to reward (44).
There is some evidence that high-fat, more calorie-dense, diets promote overeating and overweight more than their low-fat equivalents (45). One reason may be that palatability and the energy density of food are closely correlated with human food preferences (46). In addition, fat has the lowest satiation effect among all of the micronutrients, even though it is the most energy dense (47). Marketing studies also show that food choices are more likely to be made on the basis of taste, cost, and convenience than for health-related reasons (48). It follows, therefore, that those who are highly sensitive to the rewarding properties of food and who are “myopic” about the future negative consequences of a diet with poor nutritional status are more likely to make poor food choices, especially in an environment where there are so many, and such varied, delectable and highly caloric foods available at relatively little cost. More than ever before in our history, the maintenance of a healthy body weight seems dependent on the strength of the cortical inhibitory processes necessary for making good decisions overriding the strength of the drive to consume calories.
Finally, despite the statistical strength of our results, it is important to acknowledge that our sample size was relatively small and that replication is necessary before we can draw firm conclusions about these findings.
The four mediational analyses described above were repeated with highest ever BMI as the dependent variable. Results indicated that the relationship with the gambling task was not different from the analyses which used current BMI. However, the relationship with emotional overeating was substantially stronger. In the 2nd and 4th models, the R2 increased from 0.25 to 0.30 and from 0.35 to 0.40, respectively.