The Neurocognitive Effect of Front-of-Package Nutrition
Labels on consumer’s implicit attitudes and dietary
decisions
EliskaProchazkova
10629661
Master Brain and Cognitive Science, University of Amsterdam
Abstract
Obesity is one of the most serious concerns of modern day society. Front of package (FOP) labels have been introduced to improve dietary choices. Attention plays a crucial role in the computation of stimulus values when decisions are made, which suggests that FOP labels could be applied to improve feeding decision and enhance self-control. We utilized an fMRI food choice task to explore how different types of FOP labels- the Healthy Choice (HC) logo & the Traffic Light icon (TL) affect neural networks involved in food decision making. On the neural level, we found that HC logo increased inferior frontal gyrus (IFG) activity, which supported previous research and our hypothesis suggesting that food labels work as exogenous cues that focus attention towards the healthiness of foods. Based on previous evidence, we further tested whether FOP labels (HC/TL logo) can improve consumer’s dietary choices by increasing the level to which the dorsolateral prefrontal cortex (dlPFC) modulates the ventromedial prefrontal cortex (vmPFC). Contrary to this hypothesis, we found that the behavioral influence of HC logo (HC susceptibility) was associated with reduced dlPFC activity and decreased functional connectivity between dlPFC and vmPFC. In line with previous literature, we propose that this is due to the fact that HC provides ‘‘halo effect’’, which reduces the need for the top-down health control implementedby dlPFC. Thiscan result in positive and potentially exaggerated health evaluations and overconsumption. This study is the first to provide neurological evidence for the notion that HC logo does not improve dietary choices, instead it promotes poor dietary decisions as it works as a ‘‘seal of approval’’ signpost. These results provide implications for policy makers, manufacturers, and further nutrition research. More research needs to be done to support this finding.
1.1. Neural mechanism of food decision-making:
Obesity is one of the most frequent medical conditions and can cause diabetes strokes and heart diseases, which makes it one of the most pressing concerns of modern society. In neuroeconomics, obesity is often related to faulty decision-making, in particular to feeding decisions. Feeding decisions are based on two distinct types of attributes: Firstly, the ones with basic and immediate outcomes such as taste and secondly, those related to abstract and delayed outcomes such as health.Several behavioral studies (Armel et al., 2008; Krajbich et al., 2010), as well as psychological models of decision-making (Roe et al., 2001; Rieskamp et al., 2006), propose that the way attributes are integrated depends on how attention is deployed among various features, in this case short and long-term attributes of food. To make a good decision, value signals have to be computed in the ventromedial Prefrontal Cortex (vmPFC) that has to weight these two distinct types of attributes (health & taste) properly (Hare, Malmaud& Rangel, 2011). Hence, various factors have to be integrated in the vmPFC to compute goal values. An optimal feeding decision requires assigning values to foods in a way that properly reflects their relative contribution to the benefits they produce. Activity in the dorsolateral Prefrontal Cortex (dlPFC) is associated with incorporating higher order factors (health) in the vmPFC value signal (Hare, Camerer& Rangel, 2009). It has been suggested that self-control failures arise from a tendency of the valuation circuitry to overweight short-term relative to long-term features (Liberman& Trope, 2008). This is supported by an fMRI study on dietary choices that demonstrates that in unhealthy eaters as compared to more healthy ones the value signal in the vmPFCoverweights taste compared to health (Hare et al., 2009).
Recent research in neuroeconomics has demonstrated that attentional cues can lead to more sensitivity to long term outcomes when feeding decision are made (Hare et al., 2011). Hare et al. (2011) demonstrated that cues that guide attention to the health aspects of foods by increasing activity in inferior frontal gyrus (IFG) that increased the extent to which health was taken into consideration on a behavioral level (see Fig. 1). This area was of particular interest, since it has been previously associated with attention to salient exogenous cues, dietary self-control (Hampshire, Chamberlain, Monti, Duncan, & Owen, 2010)and gain loss (Hare, 2009). According to Hare (2009) left IFG region could potentially
affect choice by modulating valuation processes in other areas. Namely Reward, Amygdala, Insula, vmPFC, and dlPFCare neural networks that have been previously implicated to play role in (higher-order) food decision making (Bechara, et al.1999; Rolls, 2011; Rangel, Camerer, & Montague, 2008). More specifically, according to Hare’s et al. (2011) exogenous attention cues led to more responsiveness of stimulus value signals in the vmPFC. Hare et al (2009, 2011) showed that the effect of vmPFC on healthy decisions is modulated by activity in dlPFC region (Hare et al., 2011), which is related to cognitive control (Miller & Cohen, 2001) and emotion regulation (Ochsner& Gross, 2005). This was supported by their finding that the dlPFC, the IFG and the vmPFC are functionally connected during a task in which participants had to indicate whether they want to purchase food products while considering their healthiness (Hare et al., 2009 & 2011). Hence, it has been proposed that health cues can lead activation of neural mechanisms that implement successful self-control and healthy food choices.
1.2. Do FOP labels function as health attentional cues?
To embark on the worldwide issue of obesity and other nutrient related diseases, Front-of-pack (FOP) nutrition labeling has become a mandatory part of food Front-of-packages to promote healthier dietary choices and improve public health (Nishida, Uauy, Kumanyika, &Shetty, 2004). The purpose of FOP labels is to attract consumer’s attention and help them to make better choices in constructing a healthy diet (Food Standards Agency 2008).Subsequently, there has been increasing interest among researchers and policy-makers to develop the most effective types of nutrition signpost that would truly capture people’s attention and improve their food attitudes and behavior. Although numerous studies have investigated what neural mechanisms are related to obesity (Horvath, T. L., 2005; Ziauddeen, Farooqi& Fletcher, 2012) there is no research on the neurocognitive impact of FOP food labels on feeding decisions. Because attention plays a crucial role in the computation of stimulus values when decisions are made, this suggests that FOP labels, similarly as in Hare’s et al. (2011) previous research, may represent attention cues that could be applied to improve feeding decision and enhance self-control. Thus this study is the first to compare the effect of these labels on neural level. By doing so, we hope to provide direct evidence for the
notion that FOP function as a attention cues that can activate neural mechanisms that are involved in successful self-control and hence improve consumers’ dietary choices.
1.2.1. Different types of FOP labels:
In particular two most prominent labels will be compared: the Healthy Choice (HC) and the Traffic Light-Guidelines Daily Amounts (TL) labels (Fig. 1). The main difference between these two labeling systems is that the Healthy Choice (HC) is a single summary indicator, which compares the relative healthiness of a product within its food category. The second type of FOP labeling is the use absolute system, namely the Traffic Light–Guidelines Daily Amounts (TL) adapted by UK, which presents negative as well as positive aspects of a product by means of a color-coded scheme. The TL label indicates how much fat, saturated fat, sugar and salt are in that product by using high (red), medium (amber) and low (green) percentages for each of these ingredients (Kleef et al., 2013).
1.2.2. The Empirical Dispute: Is Simpler Always Better? 1.2.3. The ‘halo effect’ of the HC label
The introduction of the HC system to the Dutch market sparked a controversial debate and a wave of criticism from a variety of sources in the public health community (Nestle 2009; Neuman 2009; Pinkston 2009). For instance, Nestle and Ludwig (2010) in their article, published under the name ‘‘Front-of-package food labels: public health or propaganda?’’ suggested that front-of-package labels are deceptive by means of presenting information out of context. The authors suggested that a single summary icon such as HC could lead to positive but potentially misleading nutrient evaluations. This favorable effect of simplified FOP indicators has been proposed to be a result of heuristic processes where in order to reduce cognitive effort, consumers tend to focus on peripheral/attention cues which allow them to make fast and intuitive food judgments and evaluations without careful consideration of the food’s nutrient levels (Andrews et al., 1998; Pandelaere, Dewitte, &Warlop, 2007, Wansink&Chandon, 2006). According to Nesbett and Wilson (1977), this type of processing can lead to the so called ‘halo effect’ where people start to associate other aspects of the food, not explicitly identified in the FOP symbol, as positive.
Kleef&Dagevos (2013) proposed in their recent review that the ‘halo effect’ of FOP labels might bias the health evaluations which may lead to overconsumption.
1.2.4. The complexity of the TL label
On the other hand, the TL labeling system has been criticized for being too complicated. Although, current literature agrees that TL system helps consumers to identify healthier products and make healthier choices (Balcombe et al., 2010; Kelly et al., 2009). The TL labels compared to foods with a ‘seal of approval’ mark needs additional time to be processed and for that reason it has been proposed that TL may not influence the consumers attitudes as the long processing time of the TL icon may be disadvantageous.
In our study we investigate whether health cues in form of FOP-labels, either the smart choice icons (Dutch system) or the Traffic Light icons (UK system) are able to activate the neural mechanisms of self-control and consequently lead to healthier consumer behavior (Fig. 1).
(A)
Fig. 1: (A)The neural mechanism of food decision making are based on Hare et al. (2009)
and Hare et al. (2011) studies. The relative weight of health and taste can be influenced by attention cues. For instance when people are asked to consider the healthiness of the product, the attentional cue activates IFG (inferior frontal gyrus), which modulates the valuation circuitry that take place between thedlPFC (dorsolateral prefrontal cortex) the vmPFC (ventromedial prefrontal cortex). (B)Traffic Light-Guidelines Daily Amounts and (C) Healthy Choice FOP nutrition labels.
1.3. Current research:
To reconcile the empirical dispute and at the same time provide practical implications for public health officials, nutrition researchers, and food manufacturers, the current study is the first to tests the effect of front of pack labels on people’s neural responses. Until now only explicit attitudes and behavior have been assessed and the ‘halo effect’ of single summary systems (e.g. Healthy choice, Smart choice, Health check) was only tested and ‘proved’ to be true in self-reports using different types of questionnaires. However, such explicit measures pose the risk that people may conform to the intention of researchers and rate the single summary system as healthier as they are under the impression that it is expected from them. This study pioneers in investigating the neuro-cognitive effect of FOP-labels using functional magnetic resonance imaging (fMRI).By doing so this study aims at
providing explanations not only to why but also how FOP signposts impact on people’s cognition and behavior. Hence,in current research we will investigate the neural mechanisms that underpin the effect of these labels on people’s brain. Two main questions will be addressed:
1.4.1. Research Questions:
The first question isQ1:do the FOP labels function as health attentional cues? Theneural correlates of FOP labels will be investigated, with specific focus on brain regions that are known to be involved in food decision-making.
To provide a link between neural data and behavior, we will further test Q2:how is behavioral susceptibility to FOP labels related to average activity and functional connectivity between brain regions that are known to be involved in successful self-control?
1.4.2. Hypotheses:
H1: In line with previous research (Hare et al., 2011), we hypnotize that the FOP labels will function as an exogenous attention cues that activate IFG.
H2: We further hypothesize that on a neural level, if the FOP labels can improve dietary choices they should operate by enhancing the extent to which the dlPFC modulates the vmPFC so that its responsiveness to healthy food increases.
1.4.3. Implications:
If FOP labels were found to be efficient on a neurocognitive level this would lead to enhanced use of these labels. Furthermore, by addressing these hypotheses this study hopes to provide direct evidence for or against the theory that FOP labels manipulate people’s implicit attitudes. If one system proves to exert higher impact on the dlPFC-vmPFC modulation, this would suggest that this labeling condition promotes responsiveness to healthiness of food and thus exerts dietary self-control. On the other hand, reduced
dlPFC-vmPFC modulation would indicate decreased attention to healthiness of food and lower levels of dietary self-control. By understanding the underlying mechanisms, current study aims to inform policy makers and public the effect of FOP labels on people’s implicit attitudes so that obesity could be prevented to a greater extent.
2. Materials & Methods
2.1. Participants:
Twenty-one subjects (13 males; age range: 23-52 years; mean age: 34.95, SD, 9.75) conducted the MRI study. The participants were healthy Dutch participants without previous incidence of psychiatric or neurological illness and not using drugs or prescription based medication. Participants’ (self-reported) height ranged between 1.68-1.96m (M=1.8, SD=0.2), and weight was between 53-95 kg (M=76.1, SD=14.9). One additional subject participated in the experiment but did not finish the task and was consequently excluded from our sample. The review Board of the University of Amsterdam approved the study.
2.2. Procedure:
The task used in the MRI scanner was an abbreviated version of the ‘food-choice task’ by Hare et al. (2011). In this task participants were asked whether they want to purchase the products presented on the screen. In total participants were presented to 36 products from 6 different food categories (vegetables, crisps, cereals, ready-meals, juices, spreads). Six products from the vegetable category were used as healthy stimuli and 6 products from the crisps category represented unhealthy stimuli. The remaining 24 products were classified as ambiguous stimuli and were of main interest in this experiment. Dependent on the labeling condition and the nutrition values the products were presented with a Traffic Light label (TL), Healthy Choice (HC) Icon or no label (Control). Participants had to respond on a 4-point scale from strong yes to strong no by pressing a button. The
participants had maximally 2 seconds to respond. The stimuli disappeared as soon as the participants made a response. This short time window forced the subjects to respond quickly and rendered their choice more automatic and implicit, as they did not have the time to consciously consider their choice (Fig 2). After they made their choice they got feedback on which button they pressed. The stimulus presentation and response recording was controlled through Presentation® software (Version 0.70, www.neurobs.com).
After the fMRI study, the participants were given questionnaires outside the scanner. Here subjects filled out their demographic details, along with explicit food label attitudes and health consciousness measures.In addition they had to rate the same food stimuli, but this time without food labels, on the perceived healthiness of foods. We used this questionnaire to measure individual differences in their implicit behavioral susceptibility to FOP labels (for details on FOP label susceptibility see the 2.3. Experimental Design section).
Figure 2: Trial outline for fMRI study. Participants are asked to rate food products on the purchase intention on 4-point scale (1-strong yes, 4-strong no). In total one block lasts 4.75 minutes. The BOLD was measured form the onset of the onset of the stimuli (yellow lines).
2.3. Experimental design:
Food item 2 sec(200 ms without label, 1.8 with label)
Dependent on the labeling condition products were presented with a (IV1:) Traffic Light label (IV2:) Healthy Choice Icon (IV3:) or no label (Control). Each participant viewed all 36 products (24 ambiguous, 6 healthy, 6 unhealthy) and was assigned to one of 3 groups over which the different labels presented with each product were counterbalanced. As a result each product was viewed in each labeling condition, namely Traffic Light (TL), Healthy Choice (HC) and Control (no FOP label). Within all food categories the label-product combinations were counterbalanced and half of the products were assigned to one of the labeling condition whereas the other half was assigned to another condition (see Appendix Fig 3). In group1 there were 6 subjects, in group 2 there were 7 subjects and in group 3 there were 8 subjects. In order to sustain the ecological validity of our sample we placed the HC labels on those products that also appear with the HC logo in Dutch supermarkets. In the TL-GDA icon condition absolute nutrient amounts were given, as well as their corresponding percentages of the Daily Values (DVs). The experimental conditions were consistent with current FOP information in the marketplace. Three labeling conditions were used as IV’s for subsequent fMRI analysis.
We further investigated various factors that could explain why some people are more susceptible to the FOP labels than others. In particular we wanted to test whether there is a relationship between nutrient consciousness and behavioral susceptibility to HC/TL logos? In order to do so, the behavioral susceptibility to HC logo was calculated for each subject by subtracting the mean health ratings on products with no label from the mean health ratings on products with the HC logo. Higher numbers represented higher susceptibility to the HC logo. The same applied to susceptibility to the TL logo. However, as the Traffic Light is a bidirectional indicator, the susceptibility to TL logo was calculated as a mean behavioral change (negative/positive) to TL logos compared to the control condition. After inspection of the Q-Q plots and histograms we found the susceptibility to HC/TL label to be normally distributed, where the higher values represented higher susceptibility to HC labels (see Appendix, Fig. 4). We then used a GLM to predict participant’s susceptibility using following repressors: (1) age (2) gender, (3) income (4) education (5) income (6) nutrient consciousness (6) familiarity to labels (7) attention to labels (8) influence of labels, (9) BMI
and (10) educational level. The results showed that none of the variables were significant predictors of HC or HC susceptibility (see Appendix,Table 1). These findings indicate that there was no association between self-reported nutrient consciousness or age/gender and behavioral susceptibility to HC/TL logos. In following analysis we tested whether susceptibility to FOP logos can be predicted by neural activity within dlPFC and vmPFC and connectivity between these regions.
2.4. fMRI data acquisition:
The scans were acquired on a Philips 3T Achieva TX scanner, which is located at the Spinoza Center, Roeterseiland, Amsterdam, the Netherlands. Whole brain gradient-echo echo-planar imaging (EPI) measurements (voxel size=3*3 *3 mm3 , repetition time [TR]=2000 ms, echo time [TE]=27.63 ms, flip angle=76.1°, FOV¼240 240, matrix= 80*80, slice thickness=3 mm, slice gap=0.3 mm, 38 slices per volume, sensitivity encoding factor of 2) were acquired to measure blood oxygen level-dependent (BOLD) magnetic resonance images with a 32-channel SENSE head coil. Two T1- weighted anatomical scans were acquired per session (four T1 volumes per participant, voxel size=1*1*1 mm3, TR=8229 ms, TE=3.77 ms, flip angle=8°, FOV=256*256, matrix=256*256, slice thickness=1 mm, no slice gap, 160 slices per volume). Only one T1-weighted image was used for registration purposes.The functional data were collected in six sessions and was part of a bigger fMRI study with five other tasks with in total 44 min pure scanning time. The length of our block had 158 volumes (5.26 min) and the functional data within this block were collected (2.0s TR, 27.63ms TE, 192*141.24mm2 FOV, 39 slices, 3.3mm slice thickness, 76.1° FA, sagittal orientation) covering the whole brain. fMRI data was analyzed using the FMRI Expert Analysis Tool (FEAT) in Functional MRI of the Brain (FMRIB) Software Library (FSL) version 6.0 (Oxford Centre for Functional MRI of the Brain Software Library (www.fmrib.ox.ac.uk/fsl).
For pre-processing we applied motion (Jenkinson et al., 2002)and slice time correction (Interleaved), spatial realignment, normalization to the Montreal Neurologic Institute template, spatial smoothing with a Gaussian 3mm full-width-at half-maximum kernel and high-pass temporal filtering with a cut-off of 100s. To extract voxels belonging to brain tissue from non-brain tissue voxels we used the Brain Extraction Tool (BET) (Smith, 2002). 2.5.1. Regions of interest (ROIs) selection
In addition to the whole brain analyses, we investigated how activity within specific ROIs in response to FOP labeling conditions. These ROIs were classified by inclusion masks obtained from http://neurosynth.org where they are freely available for downloading (ref). The masks are derived from meta-analyses of previous studies displaying brain regions that are consistently active in studies that include the name of the region in the abstract. Six regions of interest were selected based on previous literature: Regions associated with Reward, Amygdala, Insula, vmPFC, dlPFC, IFG (Hare et al., 2009; Hare et al. 2011). The ROIs were defined from the Neurosynth.org website, an online platform for large-scale, automated synthesis of functional magnetic resonance imaging (fMRI) data. Masks of ROIs were extracted from backward inference based on thousands of published articles reporting the results of fMRI studies. In addition to identify visual regions (right and left V1) we applied the jeulricht-bilateral atlas. The signal extracted within the ROIs, was based on average parameter estimates. The signal change scaling was determined by the whole brain mean.
3.The fMRIAnalysis
Q1: What are the neural correlates of FOP labels?
3.1. The Whole Brain Analysis:
The first step in our analysis was to indentify neural areas that are related to FOP labels occurrence. The conditions were separately modeled by convolution with a double-gamma hemodynamic response function in a general linear model. For the basic shape we
used 3 the basic 3 column format (The first column indicates the onset of stimuli, the second is duration (2s) and the third information about the condition). The GLM contained following 5 EVs: Control condition, (2) HC condition, and (3) TL condition. To control for products that are generally perceived as very healthy/ very unhealthy, apart from labeling condition in current GLM we used following five repressors of interest: (4) Vegetables and (5) Crisps food products as control categories.
We calculated the following 11 first-level single-subject contrasts: (1) HC effect versus baseline, (2) TL effect versus baseline, (3) Control versus baseline, these contrast were used to investigate the average change in neural activity when food stimuli (with TL, HC, no-label) were presented versus no food stimuli presentation (i.e. fixation cross, feedback). The main interests of current analysis were however following t-tests: (5) TL versus Control, (6) HC versus Control, (7) labels (TL/HC) versus control. These contrastswere used to test the effect of labels on brain activity as compared to effect of food stimuli alone (control-no label conditions). In addition we included (7) HC versus TL contrast to test whether there is a difference in neural activity between the two labeling condition. Finally, we includedcontrasts: (9), vegetables versus baseline, (10) crisps versus baseline, (11) vegetables versus crisps. We used these tests to investigating whether there arevoxels that are related to vegetables-‘healthy food stimuli’ and crisps-‘unhealthy food stimuli’ or whether there is a difference in neural activity in between the two. These were the simple contrasts on which basis a map of t-test values was produced for analysis in each subject.
The resulting contrast images were linearly registered to the anatomical images using FMRIB's Linear Registration Tool (FLIRT) with 7° of freedom and the full search space (Jenkinson and Smith, 2001). Subsequently, the images were spatially normalized to the T1- weighted MNI-152 stereotaxic space template (2 mm) using FMRIB's Non-Linear Registration Tool (FNIRT). For the group analysis the Mixed effect group analysis (FLAME 1+2) was used. We applied the Random field theory based correction in order to control for multiple comparisons (which z=2.3, which p=0.05).
The next step in our analysis was to compare the neural activity under different labeling conditions within selected regions of interest. After the ROI selection (for details see section 2.5.1.),for each participant and time point the average percent signal change (PSC) relative to baseline was extracted from each ROIs. The signal was averagedacross all voxels in each ROI. To get a mean Beta value for each condition (HC, TL, Control), the PSC values were then averaged across all subjects. To compare the regional activity under different conditions we used repeated measures (rm) ANOVA.
Q2:how is behavioral susceptibility to FOP labels related to average activity and functional connectivity between brain regions that are known to be involved in successful self-control?
3.3. Average vmPFC and dlPFC activity &behavioral susceptibility to FOP labels
The third step in the fMRI analysis tested the relationship between individual differences in susceptibility to FOP labels (TL/HC) and average vmPFC/dlPFC activity. To do this, susceptibility to FOP logo wascalculated for each participant. The BOLD response in dlPFC and vmPFC was averaged across all labeling conditions for each subject. The individual differences in average dlPFC and vmPFC activity were then used as predictors of individual differences in susceptibility to HC and TL labels.
3.4.Difference in dlPFC and vmPFC connectivity between the high and low HC susceptibility group
Apart from average activation, we further tested whether there is a difference between the high and low susceptibility group in their dlPFC and vmPFC connectivity, we performed a psychophysiological interaction (PPI) analysis to identify regions exhibiting functional connectivity with dlPFC (seed region) during the task. This was done for both the high and low susceptibility group in the following steps. First, subjects were divided into two groups: High and low susceptibility to HC group. We perform a median split and used this recoded
measure as an independent variable. In the low HC susceptibility group the mean level of susceptibility was -0.35 (N=10); in the higher nutrition consciousness condition the mean susceptibility level was 0.27 (N=10); (F(1,20) = 0.06, p < .0000). Second, for each individual, we extracted the BOLD time-series from the voxel within weighted mask of the dlPFC (group mask based on the backward functional inference obtained from the Neurosynth.org.). Variance associated with the six motion repressors was removed from the extracted time series. Third, for every subject, we estimated a GLM that included the following regressors: (1) an interaction between neural activity in the vmPFC and the time of HC logo onset convolved with the canonical hemodynamic response function (HRF), (2) an interaction between neural activity in the vmPFC and the time of TL logo onset convolved with the canonical hemodynamic response function (HRF), (3) an interaction between neural activity in the vmPFC and the time of control (no logo) presentation onset convolved with the canonical hemodynamic response function (HRF), (4) an indicator time of HC logo onset convolved with the canonical HRF, (5) an indicator time of TL logo onset convolved with the canonical HRF, (6) an indicator time of control condition onset convolved with the canonical HRF, and (7) the original BOLD eigenvariate from the dlPFC. In addition, we included a regressor for (8) Vegetables and (9) Crisps. Single subject contrasts were calculated following the estimation of the GLM. Third, second-level group contrasts were calculated based on the single-subject contrast values as well as between subject comparisons between high and low HC susceptibility group, using one-sample t tests.
4. Results:
Q1: What are the neural correlates of FOP labels?
4.1. The Whole Brain Analysis
The first step in the fMRI analysis was to identify regions that are candidates for playing a role in implementing the computational changes associated with attention cues in form of
activity in HC or TL trials compared with the Control trials. In line with previous research (Hare et al., 2011), we found stronger activity during HC compared with the control condition (C) in the left and right IFG (BA 46/ 9), and occipital lobe (BA 18), (p <0.05, corrected), (Fig.4 A, Table 2). A comparison of TL minus C found more activity in the bilateral occipital cortex (p<0.05, corrected) (Fig.4 B, Table 3). Color-coded conjunction analysis indicating overlap of regions more active in label’s (HC/TL) condition minus control (no-label) condition is depicted in(Fig. 4 C, Table 4).
Table 2: Regions more active in HC than Control
Region BA Side Cluster
size x y z Peak Z score
Occipital lobe 18 L 3302 -2 -98 6 5.54*
Inferior frontal gyrus 9 L 1078 -48 20 22 3.4*
Inferior frontal gyrus 46 L 1078 -46 22 20 3.36*
Inferior frontal gyrus 9 R 489 54 16 32 3.17*
Height threshold t=2.3; (3X3X3mm). L, Left; R, right. *The activation survives whole-brain correction (p 0.05) for multiple comparisons at the cluster level (height threshold, t=2.3).
Table 3: Regions more active in TL than Control
Region BA Side Cluster size x y z Peak Z
score
Occipital Lobe 18 L 3651 -2 -96 10 6.24*
Height threshold t=2.3; (3X3X3mm). L, Left; R, right. *The activation survives whole-brain correction (p 0.05) for multiple comparisons at the cluster level (height threshold, t=2.3)
Table 4: Regions more active in label’s (HC/TL) condition than control (no-label) condition
Region BA Side Cluster
size x y z Peak Z score
Occipital Lobe 17 R/L 3823 -2 -96 8 6.17*
Precentral / Inferior frontal
gyrus WM/9 L 569 -38 6 24 3.23*
Height threshold t=2.3; (3X3X3mm). L, Left; R, right. *The activation survives whole-brain correction (p 0.05) for multiple comparisons at the cluster level (height threshold, t=2.3). WM=White Matter
(A)
(B)
Figure 4: (A) Regions more active in HC than Control, (B) Regions more active in TL than Control (C) The Neural overlap between TL (green) and HC (red) conditions versus control. The coordinates depicted in fig. 6.C. are x= -38, y=6, z=24.
Q1: Does the activity within ROIs differ in response to FOP labeling conditions?
4.2. The ROI Analysis
A Repeated-measures ANOVA showed that there was a significant difference in activation in the R/L inferior frontal gyrus (IFG) between the three labeling conditions (F (1,20) = 6.816, p=0.003). This result showed that the activity in IFG varied across condition. To further understand the source of this variation we conducted A post-hoc pairwise comparison (LSD).The post-hoc pairwise comparison (LSD) indicated a significantly higher activation (p=0.002) in the IFG when products where presented with the HC icon as compared to the Control condition.There was no significant difference in IFG activity between the TL and Control (p>0.05). In the remaining five ROIs, namely Reward, Amygdala, Insula, vmPFC, and dlPFCthat have been previously implicated to play role in (higher-order) food decision making (Bechara, et al.1999; Rolls, 2011; Rangel, Camerer, & Montague, 2008). We did not findsignificantly differ across the three labeling conditions (all p>0.05), ( see Fig. 5).
To test whether IFG activation is associated to attention switching to food labels (rather than distinct visual inputs of stimuli) we conducted a confirmatory ROI analysis. Here, we compared the neural activity of the primary visual areas (V1-R, V2-L). In the visual areas,
**
** *
we averaged the right V1-R and left V1-Lsignal. Repeated measures ANOVA showed that the level of V1 activity significantly differed across experimental conditions(F(1,20) = 12.25, p=0.002). To test how the activity differed, we used LSD Pairwise Comparisons test. LSD showed that activity was significantly higher in the V1 when products were displayed with the TL label (p=0.02) and HC label (p=0.002) compared to control condition. There was no significant difference in V1 between the HC label and TL conditions (p>0.05).
Reward Amygdala Insula vmPFC dlPFC IFG V1
-0.1 0 0.1 0.2 0.3 0.4 0.5 Control HC TL
B
et
a
Fig. 5: Beta plots displaying the amount of activity in ROIs in each labeling condition compared to baseline. Error bars represent SEM. * The mean difference is significant at the . 05 level, **at level 0.01. Adjustment for multiple comparisons: Least Significant Difference.
Q2: Is there a correlation between the average vmPFC and dlPFC activity and behavioral susceptibility to FOP labels?
A linear regression (Fig. 6A)demonstrated that activity in dlPFC during the experiment was a significant negative predictor of susceptibility to HC (F(1,19)= 6.125, p=0.024, with an R2
=0.219. The average activity in dlPFC did not predict susceptibility to TL labels (F(1,19)= 0.10, p=0.923, Fig. 6B). We further found that vmPFC was not a significant negative predictor of HC susceptibility (F(1,19)= 0.23, p=0.63) or susceptibility to TL labels (F(1,19)= 0.05, p=0.82)(Fig. 6 C-D). -0.3 4.7 9.7 14.7 19.7 24.7 -0.8 -0.4 0 0.4 0.8 R² = 0.09 DLPFC Beta Su sc e p ta b il it y to H C -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 R² = 0 DLPFC Beta Su sc e p t b il it y to T L -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 R² = 0.01 VMPFC Beta Su sc e p t b il it y to H C -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 R² = 0 VMPFC Beta Su sc e p t b il it y to T L (A) (B) (D) (C)
Fig. 6: (A) The average DLPFC activity is a significant negative predictor of susceptibility to HC label (B) The average DLPFC does not predictor of susceptibility to TL label.(C) The average vmPFC activity is not significantly predict susceptibility to HC label or (D) susceptibility to TL label.
Q2: How is behavioral susceptibility to FOP labels related to average activity and functional connectivity between brain regions that are known to be involved in successful self-control?
4.4. Difference in dlPFC and vmPFC connectivity between the high and low HC susceptibility group
In both groups the psychophysiological interactions (PPI) analysis using DLPFC as the seed region foundfunctional connectivity with a large cluster of voxels with the peak activation in the frontal medial cortex (vmPFC).(Table 5,Fig. 7).
Table 5: Regions showing task related functional connectivity with the DLPFC Group: Low susceptibility
Region BA Side Cluster
size
x y z Peak Z score
Frontal Medial Cortex+ 10 R/L 2783 2 48 -14 6.83*
Group: High susceptibility
Region BA Side Cluster
size x y z Peak Z score
Height threshold: 4.6 (3X3X3mm). L, Left; R, right. *The activation survives whole-brain correction (p 0.05) for multiple comparisons at the cluster level (height threshold, z=4.6), +referred to as vmPFC.
Fig. 6: (A)Sagital view: Task-related Functional connectivity with DLFC in low susceptibility groups, VMPFC refers to Frontal Medial Cortex. For the visualization purposes that threshold height is 4.6. The coordinates used: X=2, Y=48, Z=-14. (B) The Coronal view of the same image.
In order to determine whether there is a difference between the high and low susceptibility group in the functional connectivity between the vmPFC and dlPFC we computed the partial correlation between vmPFC and dlPFC BOLD time series, while controlling for common input from left V1-L and right V1-R BOLD time series. Independent sample t-test indicatedthat the high HC susceptibility group showed lower functional connectivity between vmPFC and dlPFC(M= 0.34)than the low susceptibility group (M= 0.16), t(20)=0.308, p =0.031 ( one-tailed), (Fig. 8).
Fig. 8:Partial correlation between dlPFC and vmPFC time series for high and low susceptibility group *p<0.05.
5. Discussion:
Obesity is one of the most serious concerns of modern day society. Front of package (FOP) labels have been introduced to improve dietary choices. Obesity studies suggests that attention plays a crucial role in the computation of stimulus values when decisions are made, which suggests that FOP labels could be applied to improve feeding decision and enhance self-control. We utilized an fMRI food choice task to explore how different types of FOP labels- the Healthy Choice (HC) logo & the Traffic Light icon (TL) affect neural networks involved in food decision making. In particular, current research tested how different labelingsystems affect the neural mechanisms. First of all, In line with previous research (Hare et al., 2011), we hypnotized (H1:) that the FOP labels will function as an exogenous attention cues that activate left IFG.
We found that HC label presentation was associated with increased activation in the left IFG (BA 46/ 9) and Occipital areas (Fig. 3A).Visual inspection of the neural overlap (Fig. 3C)indicated that compared to the HC label, the TL label was associated mainly with increased activity in the occipital cortex, whereas the HC label was more strongly associated with activity in left IFG regions. With regard to the occipital areas this is an expected result as the amount of visual information is likely to increase with complexity and amount of information the label provides (TL provides more information).
In the HC label condition the left IFG activation is consistent with Hare’s (2011) previous research where the left IFG (BA46/9) was activated when participants were asked to pay attention to the health aspect of the food presented. Previous studies associated IFG to increased attention to important task-related informationand immediate response inhibition1(Hare et al., 2009; Hampshire, Chamberlain, Monti, Duncan, & Owen, 2010;
Tabibnia et al., 2011). Considering thatattention to health iinformation and responseinhibitionarecognitiveprocessesnecessaryformaintaining a healthydiet, it has been proposed that the IFG facilitates attentional switching to enhance voluntary self control processes (Dove et al., 2000, Cools et al., 2002). In line with this theory, previous research showed that high left IFG response is related to dietary self-control and gain loss (Hare, 2009).
To further understand the impact of food labels on neural circuits involved in food decision making. We have conducted a ROI analysis. In accordance with previous analysis, our ROI data showed that IFG activity significantly differs across labeling conditions, whereas the level of activation significantly increased when the HC logo was presented compared to no-label condition. There was no significant difference in IFG activity between the TL and control condition or the TL and the HC condition. This supported previous research suggesting that left IFG could potentially be a region that implements the computational changes associated with health cues (Hare et al, 2011). In our task, the activity in the IFG
1 In particular Hampshire et al. (2010) found that the IFG is recruited when task relevant
cues are detected, regardless of whether that detection is followed by the inhibition of a motor response or the generation of a motor response.
might be associated with increased attention to health aspect of food product presented induced by FOP label.
To test whether IFG activation is associated to attention switching to food labels (rather than distinct visual inputs of stimuli) we conducted a confirmatory analysis with visual areas. Here, we found that while in visual areas the pattern of activation corresponds to the level of visual input (control<HC<TL). The decision-making areas including IFG, showed the highest activation during HC presentation as compared to control at the significance level of (p=0.05, see Fig.4) However, as there was no significant difference in IFG and V1 activation between the TL and HC, the main effect of visual input cannot be ruled out. In line with previous research (Hare et al., 2011), current findings supported our first hypothesis suggesting that FOP labels function as an exogenous attention cues that activate IFG. However, further studies are needed to prove that IFG is in fact directly related to attention switches to health aspect of the food and not only the change in visual input.
We further hypothesize that on a neural level, if the FOP labels can improve dietary choices they should operate by enhancing the extent to which the dlPFC modulates the vmPFC so that its responsiveness to healthy food increases. Prior research has shown that dietary evaluations are modulated by the level of interaction between the vmPFC and dlPFC (Hare et al, 2011). We thus tested whether the effect of FOP labels on food perceptions is associated with participant’s average vmPFC and dlPFC activity as well as functional connectivity between these two regions. In support, of our hypothesis, we found that average activation of DLPFC predicts how susceptible people are in terms of changing their behaviour in response to HC labels. Furthermore, similarly as in Hare’s et al. (2011) PPI analysis with DLPFC as seed region, we found that in both groups the DLPFC showed strong functional connectivity with vmPFC.These results confirmed the hypothesis that the level to which people are influenced by HC logo is likely to be related to individual differences in their DLPFC and vmPFC activity and the functional connectivity between these areas. Yet contrary to the assumption that the HC logo improves health evaluations via recruitment of DLPFC, we found that higher dlPFC activity predicted lower susceptibility to the HC logo. In addition by correlating the BOLD time series of the vmPFC and dlPFC over time (controlling
high susceptibility group as compared to low HC susceptibility group. The susceptibility to TL labels was not associated to neural changes in these areas.
The HC logo has been criticized for manipulating consumer’s food choices by exaggerating the foods’ health value (Nestle and Ludwig, 2010). Current findings suggest that the perception influence of HC logo is associated with reduced DLPFC activation and weaker functional connectivity dlPFCthese vmPFC. We propose that this could be due to the fact that the HC logo is not a health cue that makes people evaluate the healthiness of a food par se, but rather provides immediate approval of a food’s healthiness. The DLPFC plays a critical role in the deployment of self-control. Hence, the reduced activity in dlPFC and functional connectivity to vmPFC might be the underlying mechanism of HC logo’s ‘halo effect’ associated with positive and potentially exaggerated health evaluations.This is because reduces the need for the top-down health control via dlPFC (conscious nutrient evaluation)make people believe that food is more healthy -as shown by higher HC logo susceptibility. This could have potentially aversive effect on consumers’ dietary choices .and lead to overconsumption.
The fact that susceptibility to TL labels was not associated to neural changes in dlPFC or vmPFC areas, could be because a)TL logo is too complex to work as an exogenous cue that activates neural mechanisms implemented in successful self-control and healthy food choices. Or due to the fact that size of the neuroimaging study was relatively small especially with regard to the short duration of the task (approx 6 min.), which lead to a loss of power. Further, each participant only viewed 36 products as compared to Hare et al. (2011) who used 180 products, this may additionally decrease power and reduce the generalizability of the fMRI study. Hence the second option is that b) the strength of TL effects may be underestimated or not found at all due to loss of power.More research needs to be done to support this hypothesis.For instance, future direction may be multi-voxel pattern analysis (MVPA), which provides a higher sensitivity than the conventional univariate analysis we used in this study (Kriegeskorte, Goebel &Bandettini 2006). By means of focusing on the analysis and comparison of distributed patterns of active voxels stronger and more complex effects of FOP labels on consumer behavior may be found. For
instance, the strength of similarity of patterns for HC between brain areas may be compared to the similarity for TL across brain areas. This more sophisticated analysis may lead to valuable insights in decision-making in the supermarket.
6. Conclusion:
In conclusion, current research allowed us to tap into the neurocognitive effects of food labels on people’s dietary decisions and provided insights into their underlying mechanisms. Here we demonstrated that an increase in health perception associated to the HC logo is related to increases in the IFG signalling. Moreover, the susceptibility to HC logo was related to reduced DLPFC activity and its functional connectivity to vmPFC. We propose that this is due to the fact that HC provides a ‘seal of approval’ that the food is healthy, which reduce the need to consider and evaluate the foods healthiness.With regards to TL logo, current study provide inconclusive findings, on one hand the TL logo as a bi-directional indicator might be too complex to work as an exogenous health cue, on the other hand TL logo may have relatively weaker effect on consumer’s health ratings compare to HC logo and future larger fMRI studies may revile more about its neurological impact on consumer’s food behaviour. We conclude that neuromarketing is a promising new field that can provide additional valuable insights about consumer’s food choices beyond and above behavioral studies. Here we propose that neuromarketing research could be a new potential way for public health improvement. The current study brings implications for policy makers as well as clinical research interested in healthy diet and obesity.
Andrews, J. C., Burton, S., &Kees, J. (2011). Is simpler always better? Consumer evaluations of front-of-package nutrition symbols.Journal of Public Policy & Marketing, 30(2), 175-190. Armel, K. C., Beaumel, A., Rangel,A., (2008). Biasing simple choicesbymanipulating relative visualattention. JudgmentDecision Making 3:396 – 403.
Bechara, A., Damasio, H., Damasio, A. R., & Lee, G. P. (1999). Different contributions of the human amygdala and ventromedial prefrontal cortex to decision-making. The Journal of
Neuroscience, 19(13), 5473-5481.
Blokstra, A., Vissink, P., Venmans, L. M. A. J., Holleman, P., Schouw, Y. T., van der Smit, H. A., et al. Nederland de Maat Genomen, 2009-2010. Monitoring van risicofactoren in de algemenebevolking. RIVM-rapport nr. 260152001/2011. Bilthoven, 2011.
Burton, S., Garretson, J. A., &Velliquette, A. M. (1999). Implications of accurate usage of nutrition facts panel information for food product evaluations and purchase intentions. Journal of the Academy of Marketing Science, 27(4), 470-480.
Cools, R., Clark, L., Owen, A.M., Robbins, T.W., 2002. Defining the neural mechanisms of probabilistic reversal learning using event-related functional magnetic resonance imaging. J. Neurosci. 22, 4563–4567.
Dove, A., Pollmann, S., Schubert, T., Wiggins, C.J., von Cramon, D.Y., 2000. Prefrontal cortex activation in task switching: an event-related fMRI study. Cogn. Brain Res. 9, 103–109. Federal Register (1993), “Food Labeling; General Provisions; Nutrition Labeling…Final Rules,” 21 CFR Parts 5 and 101, U.S. Dept. of Health & Human Services, Food & Drug Administration, (January 6), 2302–2964.
Feunekes, G. I., Gortemaker, I. A., Willems, A. A., Lion, R., & Van den Kommer, M. (2008). Front-of-pack nutrition labelling: testing effectiveness of different nutrition labelling formats front-of-pack in four European countries. Appetite,50(1), 57-70.
Food Standards Agency (2008), “Signpost Labelling Research,” (September 10), [available at http://www.food.gov.uk/ foodlabelling/signposting/siognpostlabelresearch].
Hampshire, A., Chamberlain, S. R., Monti, M. M., Duncan, J., & Owen, A. M. (2010). The role of the right inferior frontal gyrus: inhibition and attentional control. Neuroimage, 50(3), 1313-1319.
Hare, T. A., Camerer, C. F., & Rangel, A. (2009). Self-control in decision-making involves modulation of the vmPFC valuation system. Science,324(5927), 646-648.
Hare, T. A., Malmaud, J., & Rangel, A. (2011). Focusing attention on the health aspects of foods changes value signals in vmPFC and improves dietary choice. The Journal of
Neuroscience, 31(30), 11077-11087.
Hollands, Gareth J., Andrew Prestwich, and Theresa M. Marteau. "Using aversive images to enhance healthy food choices and implicit attitudes: an experimental test of evaluative conditioning." Health Psychology 30.2 (2011): 195.
Jenkinson, M., & Smith, S. (2001). A global optimisation method for robust affine registration of brain images. Medical image analysis, 5(2), 143-156.
Jenkinson, M., Bannister, P., Brady, M., & Smith, S. (2002). Improved optimization for the robust and accurate linear registration and motion correction of brain images. Neuroimage, 17(2), 825-841.
Krajbich, I. ,Armel, C,, Rangel, A., (2010). Visual
fixationsandthecomputationandcomparisonofvalue in goal-directedchoice. NatNeurosci 13:1292–1298.
Kriegeskorte, N., Goebel, R., &Bandettini, P. (2006). Information-based functional brain mapping. Proceedings of the National Academy of Sciences of the United States of
America, 103(10), 3863-3868.
Nestle, M., & Ludwig, D. S. (2010). Front-of-package food labels: public health or propaganda?.JAMA, 303(8), 771-772.
Nishida, C., Uauy, R., Kumanyika, S., &Shetty, P. (2004). The joint WHO/FAO expert consultation on diet, nutrition and the prevention of chronic diseases: process, product and policy implications. Public health nutrition, 7(1a), 245-250.
Kleef, E. V., &Dagevos, H. (2013). The growing role of front-of-pack nutrition profile labelling: A consumer perspective on key issues and controversies. Critical Reviews in Food Science and Nutrition, (just-accepted).
Lupton, Joanne R., Douglas A. Balantine, Richard M. Black, Regina Hildwine, Barbara J. Ivens, Eileen T. Kennedy, Patricia T. Packard, Bradley R. Sperber, Daniel Steffen, and Mary Story
Development of Nutrition Criteria,” American Journal of Clinical Nutrition, 91 (Supplement), 1078S–1089S.
Miller, E. K., & Cohen, J. D. (2001). An integrative theory of prefrontal cortex function. Annual review of neuroscience, 24(1), 167-202.
Nestle, Marion (2009), “Not-So-Smart Choice Labels,” The Daily Green, (September 23), [available at http://www.thedaily 190 Is Simpler Always Better? green.com/healthy-eating/blogs/healthy-food/smart-choicelabels-50092309].
Nestle, M., & Ludwig, D. S. (2010). Front-of-package food labels: Public health or propaganda?. JAMA, 303(8), 771-772.
Neuman, William (2009), “For Your Health, Froot Loops,” The New York Times, (September 4), [available at http://www. nytimes.com/2009/09/05/business/05smart.html?_r=1] Nesbett, Richard E. and Timothy DeCamp Wilson (1977), “The Halo Effect: Evidence for Unconscious Alteration of Judgments,” Journal of Personality and Social Psychology, 35 (4), 250–56.
Ochsner, K. N., & Gross, J. J. (2005). The cognitive control of emotion. Trends in cognitive
sciences, 9(5), 242-249.
Pinkston, Randall (2009), “Experts Say Industry Food Labels Deceptive,” CBS Evening News,
(September 6), [available at
http://www.cbsnews.com/stories/2009/09/06/eveningnews/main 5291352.shtml? tag=contentMain;contentBody].
Rangel, A., Camerer, C., & Montague, P. R. (2008). A framework for studying the neurobiology of value-based decision making. Nature Reviews Neuroscience, 9(7), 545-556. Richetin, Juliette, et al. "The IAT as a predictor of food choice: The case of fruits versus snacks." International Journal of Psychology 42.3 (2007): 166-173.
Rieskamp J, Busemeyer JR, Mellers BA (2006) Extendingtheboundsofrationality: evidenceandtheoriesofpreferentialchoice. J EconLit 44:631– 661.
Roodenburg, A. J. C., B. M. Popkin, and J. C. Seidell. "Development of international criteria for a front of package food labelling system: the International Choices Programme." European
Roe, Brian, Alan S. Levy, and Brenda M. Derby (1999), “The Impact ofHealth Claims on Consumer Search andProduct Evaluation Outcomes: Resultsfrom FDA Experimental Data,” Journal of Public Policy& Marketing, 18 (Spring), 89–105.
Roe RM, Busemeyer JR, Townsend JT (2001) Multialternative decisionfieldtheory: a dynamicconnectionistmodelofdecisionmaking. PsycholRev 108:370 –392.
Rolls, E. T. (2011).Taste, olfactory and food texture reward processing in the brain and obesity. International Journal of Obesity, 35(4), 550-561.
Smith, S. M. (2002). Fast robust automated brain extraction. Human brain mapping, 17(3), 143-155.
Tabibnia, G., Monterosso, J. R., Baicy, K., Aron, A. R., Poldrack, R. A., Chakrapani, S., ... & London, E. D. (2011). Different forms of self-control share a neurocognitive substrate. The
Journal of Neuroscience, 31(13), 4805-4810.
vanHerpen, Erica, Sophie Hieke, and Hans van Trijp. "Inferring product healthfulness from nutrition labelling. The influence of reference points." Appetite 72 (2014): 138-149.
van Rijswijk, W., Frewer, L. J., Menozzi, D., &Faioli, G. (2008). Consumer perceptions of traceability: A cross-national comparison of the associated benefits. Food Quality and
Preference, 19(5), 452-464.
Vyth, E. L., Hendriksen, M. A. H., Roodenburg, A. J. C., Steenhuis, I. H. M., Van Raaij, J. M. A., Verhagen, H., Seidell, J. C. (2011). Consuming a diet complying with front-of-pack label criteria may reduce cholesterol levels: a modeling study. European Journal of Clinical Nutrition, 1-7.
Wansink, B., &Chandon, P. (2006). Can "low-fat" nutrition labels lead to obesity? Journal of Marketing Research, 43(4), 605-617.
Whelan, R. (2010). Effective analysis of reaction time data.The Psychological Record, 58(3), 9.
Ziauddeen, H., Farooqi, I. S., & Fletcher, P. C. (2012). Obesity and the brain: how convincing is the addiction model?. Nature Reviews Neuroscience, 13(4), 279-286.
Fig.3: Counterbalanced placement of FOP labels across groups. HC susceptibility TL susceptibility Explanatory Variable Mean Squar
e F Sig. MeanSquare F Sig.
Gender .075 .512 .492 .015 .232 .640 Income .204 1.390 .269 .036 .448 .518 Nutrient Consciousness .136 .924 .362 .126 1.554 .241 Familiarity to lables .505 3.437 .097 .042 .520 .487 Attention to labels .293 1.992 .192 .052 .643 .441 Influence of Labels .061 .413 .536 .034 .420 .531 BMI .063 .430 .528 .005 .064 .805 Education level 0.02 0.148 0.71 0.15 0.165 0.694
