Relationship Between Personality Traits and Eating Speed - Results from a Large Cross-Sectional Study Among Iranian Adults

The eating speed has been hypothesized to be associated with energy intake and weight control. Recently, the effects of some factors on eating rate were investigated, however; studies about the impacts of psychological factors in this eld are rare. One of these possible factors is personality traits. This study aimed to assess the relationship between personality traits and eating speed. Methods This cross-sectional study was performed among 4763 adults in 2011. Personality trait was assessed by using the Big Five Personality Inventory Short Form. The eating speed was evaluated using a pretested questionnaire and latent class analysis (LCA) was applied to categorize participants according to the eating speed. Multiple binary logistic regression analysis was used for data analyses. Results We found that the higher levels of neuroticism (quartile 4) were associated signicantly with lower odds of normal eating speed (OR: 0.43; 95% CI: 0.31, 0.60; P<0.001) but higher levels of other personality traits were statistically signicantly associated with higher odds of normal eating speed. Conclusions The current study showed signicant associations between personality traits and eating speed; accordingly, it provides primary evidence about the potential value of personality traits for managing the eating rate. Further prospective observational and interventional studies are recommended.


Introduction
The behavioral factors related to eating, including the eating speed is one of the increasing concerns to restrict energy intake in which it became a hallmark of many weight control programs. The importance of altering eating behaviors in health promotion is broadly recognized. Eating behaviors have an imperative role in preventing diseases such as diabetes, obesity, and cardiovascular disease (Ferriday, Bosworth, et al. 2015). Some epidemiological surveys have indicated positive relations between eating rapidly and metabolic syndrome (Hsieh, Muto, et al. 2011) and obesity (Ohkuma, Fujii, et al. 2013). Studies showed that eating too speedy is associated with overeating (Zhu, Hsu, et al. 2013). There is a relationship between eating speed and satiety (Privitera, Cooper, et al. 2012, Ferriday, Bosworth, et al. 2015. The eating speed affects satiety and food intake so a slower eating speed is associated with increased satiety and decreased food intake. Two hypotheses are proposed: the rst demonstrates that slower eating gives time to start the physiological satiety signals and the second hypothesis suggests that slower eating enhances and prolongs satisfying eating so declines feelings of hunger (Privitera, Cooper, et al. 2012). However, data to support these hypotheses have been inconsistent. A systematic review of controlled trial studies has revealed that quicker eating is associated with more energy intake (Robinson, Almiron-Roig, et al. 2014), and higher body mass index (BMI) (Almiron-Roig, Tsiountsioura, et al. 2015) is clinically important. Recently, the effects of some factors such as bite-size, food viscosity, food service form , and eating method on the eating speed have been investigated. In one study by E Almiron-Roig, it was reported that larger portion size leads to bigger bite-size and quicker eating speed (de Graaf 2011, E Almiron-Roig, Tsiountsioura, et al. 2015). In another study, Zhu found that the increased food viscosity resulted in a reduced eating speed, decreased postprandial appetite, and appetite to eat (Zhu, Hsu, et al. 2013). Suh in their study showed that the eating speed was decreased when the mixed form of a meal was served and in another study, Sun showed that the eating speed and chewing time changed with diverse feeding tools (chopsticks, ngers, and spoon) (Sun, Ranawana, et al. 2015, Suh andJung 2016), However, studies about the impacts of psychological factors in this eld are rare. One of these possible factors is personality traits.
Personality traits are described as individual patterns of feeling, thinking, and acting (Sco er-Mériaux, Falzon, et al. 2015). The behaviors of persons are affected by their personality. The big ve-model of traits is mostly used to determine personality traits. This model classi es personality traits into ve categories i.e. conscientiousness, extraversion, neuroticism, agreeableness, and openness to experience (Feizi, Keshteli, et al. 2014). Conscientiousness (C) is an attempt to organize purpose and insistence to achieve goals. Extraversion (E) is the tendency toward vigorous and social activity. Neuroticism (N) is attending to experience negative actions. Agreeableness (A) is a tendency toward friendliness and sociability. Openness (O) is the liking in experiencing novel things, opinions, and individuals (Roohafza, Feizi, et al. 2016). Personality traits are associated with several health outcomes (Raynor and Levine 2009).
The aforementioned evidence provides a good base for hypothesizing that the eating speed may also be related to people's personality traits, but to our knowledge, the relation between personality and eating speed is not investigated, so this study aimed to assess the relationship between personality traits and eating speed.

Participants
This cross-sectional study was performed within the framework of the Study on the Epidemiology of Psychological, Alimentary Health, and Nutrition (SEPAHAN) in Iran. This project was a community-based plan designed to investigate the epidemiological impressions of functional gastrointestinal disorders (FGIDs) and their association with mental and lifestyle determinants on non-academic members of Isfahan University of Medical Sciences. SEPAHAN was conducted in 2010 in two separate waves. Brie y, 10087 non-academic members of the staff of Isfahan University of Medical Sciences (IUMS), who were working in hospitals, university campus, and health centers a liated with IUMS, were invited to participate in the rst study. The rst phase focused on evaluating the basic characteristics, lifestyle, and dietary habits, and intakes of participants. The response rate in this phase was 86.16%, and 8691 completed questionnaires were returned. At the second phase, other questionnaires, which were designed to collect information on gastrointestinal, psychological, and somatoform symptoms, were distributed and 6239 completed questionnaires were returned (response rate: 64.64%). After merging data from these two phases, complete information was available for 4763 people. Ethics Committee of the University of Medical Sciences has approved the protocol of the study (Project numbers: #189069, #189082, and #189086), and written informed consent was obtained from all participants after clarifying the protocol. The details about the design, conducting, instruments, and all aspects of study could be found elsewhere (Adibi, Keshteli, et al. 2012).

Assessment of eating speed
The eating speed was investigated by using a questionnaire that included the following questions: 'how thoroughly do you chew foods? (Not very well, well, very well)', 'how long does it take you to eat lunch? (never eat lunch, < 10 min, 10-20 min, > 20 min)', 'how long does it take you to eat dinner? (never eat dinner, < 10 min, 10-20 min, > 20 min)'. We applied Latent class analysis to determine classes of people in terms of their eating speed (Hagenaars, Mccutcheon, 2002). All questions, provided in Table 1, were variables that entered into LCA. Eating speed was considered as a latent construct that could not be examined directly. Therefore, LCA as an advanced statistical method was used for the extracting of the eating speed habit indirectly based on the series of eating speed variables. Eating speed was assessed using LCA by examining the pattern of relations between observed eating speed variables and classifying individuals with similar pro les (i.e., latent classes). Therefore, each constructed class plays a role in a speci c level of eating speed. In this analysis, different indicators of eating rate described in Table 1, were examined. During LCA tting to the data, at rst, a one-class model was applied. Then, we sequentially increased the number of latent classes to determine the most parsimonious and interpretable model. For example, since a high percentage of subjects responded that they would chew meals more than 20min, such a pattern of eating rate was considered as normal eating rate class and when the majority of people expressed their chewing duration of each meal as "lower than 10min" the related class was considered as speedy. Finally, based on the mentioned questions, using latent class analysis (LCA) eating speed was de ned as three classes: 'normal', 'relative speedy' and 'speedy (Saneei, P et al. 2016)

Assessment of personality traits
The big ve personality inventory short form was used to evaluate personality traits. The ve factors include Neuroticism (N): the tendency to experience affective instability (hostility, anxiety, depression, angry, vulnerability, and impulsivity) and negative affect; Extraversion (E): dispositions toward sociability and energetic activity (assertiveness, positive emotion, warmth, gregariousness, seeking, and excitement); Openness (O): re ects individual's like interest in novel people, and ideas, as well as esthetic and intellectual propensities (values, feelings, fantasy); Agreeableness (A): a propensity toward amiability, (modesty, trust, altruism, compliance, and straightforwardness); and Conscientiousness (C): qualities such as fastidiousness, goal-orientation, and dependability (order, self-discipline, competence, achievement striving, and deliberation) (Chapman B et al.2007). These factors include the principal axes of behavioral and psychological variations in people and each element is associated with several prominent health-related behaviors and outcomes, which contain higher levels of overall morbidity and self-rated health (Neeleman J et al. 2002). This 60-item scale consists of 12 items for each subscale. These items are scored from one to ve. One is for strongly disagree, and ve for strongly agree. Some items are reverse scored. Higher scores point to higher levels of that particular personality trait. The reliability of the entire scale (α = 0.70) and subscales (αs > 0.68), and the internal consistency of the subscales in Iranian have been established previously(Anisi, Majdiyan, et al. 2011).

Assessment of other variables
A self-administered questionnaire was used to gather information on age, sex, educational attainment (less than diploma (12-year formal education), diploma, and university graduate), smoking habits (current smoker, never smoking, ex-smoker), marital status (single, married), body weight and height of the participants and BMI (kg/m 2 ). The general physical activity questionnaire (GPAQ) was used to assess their physical activity(Physical activity levels of participants were categorized as never, less than 1 hour, 1-3 hour, and more than 3 hours per week) (Ahmad, Harris, et al. 2015). The reliability of this questionnaire was assessed using Cronbach's alpha coe cient (α = 0.84). The functional gastrointestinal disorders (FGIDs) were de ned as suffering from at least one of the following: irritable bowel syndrome (IBS), gastroesophageal re ux disorder (GERD), constipation, and dyspepsia. FGIDs were determined using a modi ed Persian version of the Rome III questionnaire (The split-half test reliability of whole items value was 0.72), as a section of the main comprehensive questionnaire and Talley  Participants were categorized into three groups based on the eating speed behavior using latent class analysis ('normal', 'relative speedy', and 'speedy'). To determine signi cant differences in demographic characteristics, personality traits, lifestyle variables, and gastrointestinal disorders, and psychological distress across categories of eating speed, we used analysis of variance (ANOVA) and chi-square test for continuous and categorical variables, respectively.
Odds ratios (OR), and 95% con dence intervals (CIs) for eating speed as dependent variable across quartiles of different personality traits scores were calculated using logistic regression in crude and adjusted models. Model 1 was adjusted for age, sex, marital status, educational level. Additional adjustments for BMI, smoking, and physical activity were made in model 2. Model 3 was additionally adjusted for FGIDs and psychological distress. P for linear trend was determined by considering the frequency of eating speed as linear continuous variables in the logistic regression model. Holm-Bonferroni method (Hommel, G,1988) was used to adjust the type one error rate for multiple comparisons, and adjusted p-value thresholds for signi cance were reported in tables. P < 0.05 was considered as a statistical signi cance (2-sided) level.
The majority of people in the normal eating speed class had a long chewing duration, while in a speedy class, people had a short chewing duration (Table 1).  Table 2 presents the characteristic of study participants across classes of eating speed identi ed by latent class analysis. Participants in the rst class of the eating speed were older. The eating speed was signi cantly different between men and women so that 40.2 % of individuals in the second class of eating speed were female and 29.2 % were male (P < 0. 001). Also, physical activity and educational level were signi cantly different (P < 0. 001), but marital status and smoking were not different between the three classes of eating speed. Individuals with 'faster eating' had higher BMI, higher neuroticism scores, and had lower extraversion, openness, agreeableness, and conscientiousness scores. The prevalence of functional gastrointestinal disorders (FGIDs) was not signi cantly different between classes of eating speed. According to the General health questionnaire (GHQ), 76.9% of individuals had no psychological distress. The prevalence of psychological distress was signi cantly different across classes of eating speed (P < 0. 001, Table 2). *Derived from one-factor ANOVA and chi-square test for continuous and categorical variables, respectively. † FGIDs: functional gastrointestinal disorders. ‡ Quantitative variables were reported as mean ± SD while qualitative variables as a percentage. All p-values were adjusted using the false discovery rate controlling procedure developed by the Holm-Bonferroni method. Table 3 presents the distribution of eating speed classes and other characteristics of study participants across quartiles of personality traits' scores. The eating speed was signi cantly different across quartiles of ve personality traits; in which people with higher neuroticism scores (fourth quartile) showed higher speeding of eating while people in the fourth quartile of other personality traits had lower speeding of eating. Genders were signi cantly distributed across quartiles of all personality traits (P < 0.001) except Conscientiousness (P = 0.27), while there was a signi cant educational distribution across quartiles of all personality traits (P < 0.001). BMI was not different across the quartile of neuroticism and extraversion but it was different across other personality traits.
The prevalence of FGIDs was different across the quartile of some personality traits, and the prevalence of psychological distress was different across quartiles of all personality traits (Table 3). Crude and multivariable-adjusted ORs (95% CIs) for eating speed classes across quartiles of personality traits are presented in Table 4. In the crude model, higher levels of neuroticism (quartile 4) were associated with lower odds of normal eating (OR = 0.43; 95% CI: 0.31, 0.60; P < 0.001). Further adjustment for various confounders such as age, gender, marital status, and educational level, although weakened the associations but strongly remained practically signi cant. In the nal model, we observed that higher scores of neuroticism were associated with a 50.0% decrease in the odds of being in the normal eating speed compared with the reference category (OR = 0.50; 95%CI:0.31,0.81; P = 0.01). But in the crude model, higher levels of extraversion (quartile 4) were associated with higher odds of being in normal eating class (OR = 1.89; 95% CI: 1.39, 2.57; P < 0. 001). Further adjustment for age, gender, marital status, educational level, physical activity, smoking, and BMI weakened the associations, and in the nal model, the association was not remained statistically signi cant after adjusting the effect of FGIDs, and psychological distress. Also, in the crude model, higher levels of openness (quartile 4) were statistically signi cantly associated with higher odds of being in normal eating class (OR = 1.59; 95% CI: 1.17, 2.15; P < 0.001). Further adjustment for various confounders, weakened the associations. In the nal model, we observed that the higher scores of openness were associated with a 70.0% increase in the odds of being in the normal eating class compared with the reference category (OR = 1.70; 95%CI:1.15,2.54; P = 0.004). In the crude model, higher levels of agreeableness (quartile 4) were associated with higher odds of normal eating (OR for the class of normal eating speed = 1.82; 95% CI: 1.33, 2.50; P < 0.001). Further adjustment for various confounders although attenuated the associations, it strongly remained signi cant. In the nal model, we observed that the higher scores of agreeableness were associated with a 55.0% increase in the odds of being in the normal eating speed compared with the reference category (OR = 1.55; 95%CI: 1.03, 2.34; P = 0.002). Finally, in the crude model, higher levels of conscientiousness (quartile 4) were associated with higher odds of normal eating (OR for the class normal eating speed = 2.00; 95% CI: 1.45, 2.77; P < 0.001). Further adjustment for various confounders, weakened the associations. In the nal model, we observed that higher scores of conscientiousness were associated with a 95.0% increase in the odds of being in the normal eating speed class compared with the reference category (OR = 1.95; 95%CI:1.26,3.03; P = 0.002).

Discussion
This study was conducted aimed at assessing the relationship between personality traits and eating speed in a large sample of the Iranian adult population.
The ndings of our study showed that the higher scores of neuroticism were associated with higher odds of faster eating rates whereas higher scores of other personality traits were associated with higher odds of normal/slower eating speed. To our knowledge, the current study is the rst study to assess the relationship between personality traits and eating speed. In line with our results, some previous studies showed that neuroticism was associated with other dietary habits such as pickiness, neophobia, breakfast skipping, and promotion of other unhealthy food choices (Keller C et al. 2015, Intiful, F.D et al. 2019).
Another study reported that lower scores of neuroticism were associated with making healthy dietary choices (Kikuchi Y, 1999 Finally, some important limitations of our study need to be mentioned. First, we used self-administered questionnaires for evaluating eating speed and personality, so due to the high volume of the sample in the rst stage, we did not directly control them. Second, the cross-sectional plan of our study does not allow us to extract any causal conclusion. Third, our study was carried on IUMS personnel; therefore, the generalization of our ndings may be limited. To date, there have been only a few studies concentrating on the association between personality traits and dietary habits, and our study as the rst one contributed to the scarce research by evaluating its association with eating speed. The other major strengths of this study are high response rate, the application of validated tools, and a large sample size with various demographic data and detailed data on all ve personality domains.

Conclusion
The association between eating speed and people's personality traits is supported by the current study ndings. Our study provides primary evidence about the potential value of personality traits for managing the eating rate. Personality traits are in association with numerous diseases that this association is mediated by health behaviors particularly eating habits accordingly knowledge of the impacts of personality trait on dietary habit is of relevance for an effective behaviors' modi cation in eating habits through conducting health promotion and individualizing dietary health care programs. Further prospective observational and interventional studies are recommended.

Declarations
Ethics approval and consent to participate Written informed consent was obtained from all study participants and the ethic committee of Isfahan University of Medical sciences approved the SEPAHAN study protocol.

Consent for publication
Not applicable

Competing interests
The authors declare that they have no con icts of interest. Funding: SEPAHAN study has been supported by Isfahan University of Medical Sciences in part (Study project number: #189069, #189082, and #189086). The current secondary study did not receive nancial supports.
Data Availability: Data and materials supporting the results of this article are available from the corresponding author on reasonable request.