Emotion and Wellness Profiles of Herbal Drinks Measured Using Different Questionnaire Designs

The emotion and wellness profiles of herbal drinks were assessed using six different questionnaire designs. The questionnaire designs were constructed from two formats of questionnaire items, including words and sentences, and three types of measuring scales, including a rating scale (5-point intensity; 1 = ‘not at all’, 5 = ‘extremely’), a checklist scale (check-all-that-apply, CATA), and a combination of CATA and rating scales (rate-all-that-apply, RATA; 5-point intensity; 1 = ‘slightly’, 5 = ‘extremely’). The 39 emotional terms of the EsSense Profile® and the 45 wellness terms of the WellSense ProfileTM were translated into Thai, then screened for relevance to herbal drinks. The seven positive emotional terms (active, energetic, good, happy, polite, satisfied, and warm), three negative emotional terms (bored, disgusted, and worried), and five wellness terms (comforted, healthy, invigorated, relaxed, and refreshed) were selected and included in the questionnaire. A central location test was performed to determine the emotion and wellness profiles of five herbal drinks: roselle (Hibiscus sabdariffa) drink, chrysanthemum (Chrysanthemum morifolium Ramat) drink, ginger (Zingiber officinale) drink, Jubliang (a mixture of eight herbs) drink, and Krachai Dam (Kaempferia parviflora) drink. For herbal drinks, measuring emotion and wellness with a questionnaire using full sentences did not show increased benefit over questionnaires using words alone. All three measuring methods—a rating scale, CATA, and RATA—produced similar emotion and wellness profiles. However, each method has different advantages and limitations, which researchers should carefully consider.


Introduction
In the last decade, emotional responses have been more frequently used in consumer research and food-product development because the use of liking data alone may not effectively predict product success in the market. Collecting emotional responses that offer insight into consumer experience can provide additional useful information for product development [1]. Several findings have noted that emotions influence consumers' eating behavior and decision-making processes [2][3][4]. Understanding the emotions elicited by their products could be beneficial for food industries when considering packaging design, branding, and advertising [2].
Several methods for measuring emotions associated with foods have been developed and reported, such as EsSense Profile ® [5], consumer-defined check-all-that-apply (CD-CATA) [6], EmoSemio [7], and EmoSensory ® Wheel [8]. Furthermore, the health and wellness perception of food products has become an important aspect of consumers' purchasing decisions. The WellSense Profile TM , a questionnaire measuring wellness associated with foods from the consumer perspective, was recently developed [9] and was adapted in recent research to measure organic food-related wellbeing [10] in relation to the Krachai 1 ginger drink was prepared by diluting 18 g of instant ginger powder in 150 mL of hot water.

Screening of Emotion and Wellness Terms for Relevance to Herbal Drinks
The research protocol concerned with the use of human subjects was approved by the Khon Kaen University Ethics Committee for Human Research (KKUEC), Thailand (approval code, HE611230). Emotion and wellness terms were selected from the EsSence Profile ® [5] and WellSense Profile TM tests [9]. All terms were translated into Thai by a group of experienced linguists (n = 3) and back-translated into English by another experienced group (n = 3) for meaning validation ( Table 2). Thai emotion and wellness terms relevant to five herbal drinks (Roselle drink, Chrysanthemum drink, Ginger drink, Jubliang drink, and Krachai Dam drink) were screened using online CATA questionnaires. Thai consumers (n = 1678; 43.1% males and 56.9% females) who regularly consume herbal drinks participated voluntarily. Participants were asked to select the emotion and wellness terms that they associate with when thinking about herbal drinks that they had consumed. Terms with at least 20% frequency of use were recommended as a criterion [2,5,23,24]. However, terms with least 15% frequency of use were adopted for development of the EmoSensory ® Wheel, while negative emotion terms with a lower frequency of use (≤10%) were used to obtain a balance of positive and negative terms [8]. For our current research, terms with at least 18% frequency of use were used. First, the selected terms from all of five herbal drinks were collected. Then, the duplicated terms were removed before inclusion in the questionnaire. Based on the results, seven positive emotion terms (active, energetic, good, happy, polite, satisfied, and warm) and five wellness terms (comforted, healthy, invigorated, relaxed, and refreshed) were selected. To control the balance on both of positive and negative terms, three negative emotion terms (bored, disgusted, and worried) were added to the questionnaire (Table 3). Therefore, a total of 15 emotion and wellness terms were included in the questionnaire to measure the consumer responses elicited by the five herbal drinks.

Testing the Test Designs for Measurement of Emotion and Wellness Responses Elicited by Herbal Drinks
The proposed test designs for measurement of emotion and wellness responses were composed of 2 factors. Factor 1 was composed of two formats of questionnaire items, including words and full sentences (see the attached questionnaire file). The sentence format was used to reduce ambiguity within the questionnaire because including a context increases understanding of the specified emotion and wellness state. The full sentences were created using selected emotion and wellness terms, in consultation with a focus group of consumers (n = 29, average age = 45.76 ± 7.83 years old) (adapted from [7]). Factor 2 was composed of three types of measuring scales, including a rating scale (5-point intensity; 1 = 'not at all', 5 = 'extremely'), a checklist scale (CATA), and a combination (rate-all-thatapply, RATA; 5-point intensity; 1 = 'slightly', 5 = 'extremely'). Thus, the questionnaire designs comprised six test forms (Table 4).    [25], 50-300 responses should be collected per location for the central location test (CLT). In this study, about 200 Thai consumers were recruited for each test design (a total of 1252; 49.3% males and 50.7% females) from 4 regions of Thailand (Table 4), with almost equal proportions of four age groups participating in the CLT. To be selected, consumers were required to be familiar with and frequently consumers of herbal drinks. Each consumer was randomly assigned to one of the six test designs. Five herbal drinks were coded with random 3-digit numbers. Each herbal drink sample (20 mL) was poured into a clear plastic cup and served under the corresponding condition/temperature described in Table 1. All five samples were served in a balanced order. All consumers received five herbal drink samples for tasting, along with water to cleanse their palates between samples. First, they were asked to taste and rate their overall liking of each sample using a 9-point hedonic scale (1 = extremely dislike, 9 = extremely like). Secondly, they indicated their emotion and wellness responses to each sample using the scales, as described above. Consumers were seated separately and were instructed not to interact with other participants. Table 3. Selected emotional and wellness terms relevant to herbal drinks.

Statistical Analysis
Analysis of variance (ANOVA) and Duncan's multiple range test (DMRT) were performed on the overall liking scores to determine significant overall and pairwise differences among the five herbal drinks and six test designs.
For rating scores, the mean score of each emotion and wellness term was calculated from consumer responses. For CATA data, a frequency count was calculated by counting the number of consumers who chose the terms for each sample. RATA data were analyzed using one of two approaches [26]: frequency of selection (RATA frequency) or weighted frequency of selection (RATA score). RATA frequency was counted from the number of consumers who chose the terms for each sample. RATA scores were calculated by summing up the scores that the consumers rated for each selected term (0-5 scoring; 0 = 'not selected', 1 = 'slightly', and 5 = 'extremely') [27] and dividing by the number of participating consumers.
Both rating and RATA scores (when treated as a continuous outcome) were analyzed using ANOVA to determine the emotion and wellness terms with significant differences among the five herbal drinks. Next, the significant terms were analyzed using DMRT for multiple comparisons [28]. Principal component analysis (PCA) was performed to identify the emotion and wellness terms that contributed to overall differences among the herbal drinks and to classify the products based on emotion and wellness responses. Overall product differences were determined using multivariate analysis of variance (MANOVA) based on emotion and wellness responses, followed by discriminant analysis (DA) to identify the important terms of group discrimination.
CATA and RATA frequency counts were analyzed following the recommendation of Meyners et al. (2016) [27]. Cochran's Q test was carried out to determine the differences among herbal drink samples for each emotion or wellness response, and the sign test was used for pairwise comparison. Correspondence analysis (CA) based on Chi-square distances was used to identify the emotion and wellness terms that correlated to overall differences among the sampled herbal drinks. DA was performed to identify the important terms of group discrimination.
All statistical analyses were performed at 5% significance level using the SPSS 19 software (SPSS Inc., Chicago, IL, USA), except DA, which was carried out using XLSTAT ® Basic version 2019 software (Addinsoft; New York, NY, USA).

Overall Liking of Herbal Drink Samples
Regarding the overall liking scores of the five herbal drinks from six test designs, the data were analyzed by ANOVA and DMRT (Table 5). In each test, significant differences in overall liking were observed (p < 0.05) among the five herbal drinks, indicating that consumers liked the five herbal drinks to varying degrees. In general, roselle and chrysanthemum drinks were similarly well-liked (7.11-7.49) and were found to be more acceptable than the three other herbal drinks (6.48-6.87 for ginger and Jubliang; 6.62-7.06 for Krachai Dam). For overall liking evaluation, different groups of consumers participated in each test design. Homogeneity of hedonic perception should be determined before comparing the different questionnaire designs. When comparing the overall liking score of each herbal drink among different consumers from the six test-design groups, no significant differences were found. This implies that the consumer population in this study was homogenous in terms of hedonic perception of the test samples. Thus, the subsequent emotion and wellness results of the six test designs could be compared since overall liking and emotion terms may be correlated, depending on product, product category, demographics, and psychographics [5].

Emotion and Wellness Profiles of Herbal Drinks from Different Test Designs
As seen from Tables 6 and 7, the type of herbal drink significantly affected most emotion (except "good", "happy", "bored", "worried", and "disgusted" in some cases) and all wellness responses of consumers. Generally, negative emotion responses were rated lower than positive emotion and wellness responses. These findings confirm and extend the positive asymmetry of emotions. According to the theory of positive asymmetry, previous research findings reported that consumers experience primarily positive emotions in response to food and food names [29][30][31]. In the case of beverages, Calve-Porral et al. [32] found positive asymmetry of emotions in beverage consumption, regardless of the type of beverage. Thus, our findings are in agreement with previous studies.    When comparing the two formats of questionnaire items-words vs. full sentencessome minor differences, in terms of the number of significant emotion and wellness terms, were observed, depending on the measurement scale used. The number of significant terms for test 1 (rating-words) vs. test 2 (rating-sentences) was 13 vs. 14 for test 5 (RATA/scoreswords) vs. test 6 (RATA/scores-sentences) was 15 vs. 14, for test 3 (CATA-words) vs. test 4 (CATA-sentences) was 12 vs. 13, and for test 5 (RATA/frequency-words) vs. test 6 (RATA/frequency-sentences) was 15 vs. 12. A greater difference was observed between words and sentences in the RATA frequency test format. However, overall, sentence format may not reduce ambiguity over the word-only format in the questionnaire. For terms unfamiliar to consumers (such as "polite"), including a context may help to increase understanding of the specified emotion or wellness state. However, the emotion and wellness terms used in the questionnaire in this study were pre-screened by consumers who regularly consume the herbal drink products. Consumers selected the terms that they were familiar with; therefore, a full sentence in the questionnaire may not provide much more clarity under the conditions of the current study.
The results from three types of measurement methods, including a rating scale, CATA, and RATA, were compared. The RATA data consisted of two types: RATA frequency and RATA score. Hence, the RATA scores were compared with rating scores, and RATA frequencies were compared with CATA frequencies.
To compare rating scores (test 1) and RATA scores (test 5) from the word-format questionnaire, mean scores of emotion and wellness responses were plotted on a spider web, as shown in Figure 1. The overall profiles were visually similar, but the ratings from the rating scores were higher than those from the RATA scores. The rating-words profile (test 1) in Figure 1A reveals that significant product differences were observed for 13 terms (p < 0.05), except for "worried" (1.36-1.47 scores) and "good" (2.72-2.91 scores) terms (Table 6). Interestingly, the "refreshed" term showed higher scores of four herbal drinks (roselle, Jubliang, chrysanthemum, and Krachai Dam drink) (3.47, 3.00, 3.19, and 3.24, respectively) than other emotion and wellness terms. Meanwhile, the score of the "warm" term (2.88) was higher for ginger drink than other terms, with a maximum mean-difference value of 0.98. King et al. [29] suggested that a mean emotional difference of ≥ 0.2 units (on a 5-point scale) may be of practical value. For the rata-words profile (test 5) in Figure 1B, all 15 terms had significant differences among the five products. Therefore, using the RATA scale with words (Test 5) showed better discrimination ability than the rating scale with words (15 vs. 13) in terms of improving the number of significantly different terms. To compare the CATA frequency (test 3) and RATA frequency (test 5) from the wordformat questionnaire, emotion and wellness responses, in terms of percentage of the frequency count, were plotted on a line chart, as shown in Figure 2. The two profiles were visually similar, but the frequency counts of the RATA-Words format was higher than those of the CATA-Words format. For CATA-Words in Figure 2A, the "refreshed" term was popular among consumers for roselle (74.64%), Jubliang (65.55%), chrysanthemum (68.42%), and Krachai Dam drink (54.55%), while "good", "happy", and "worried" terms were not significantly different among the five herbal drinks (p > 0.05; Table 7). For RATA-Words in Figure 2B, all 15 terms were significantly different among the five products. Thus, the RATA-Words show a slight improvement in the number of significantly different terms. Therefore, using the RATA scale with words (Test 5) showed slightly better discrimination ability than the CATA scale with words (15 vs. 12) in terms of increasing the number of significant terms. To compare the CATA frequency (test 3) and RATA frequency (test 5) from the word-format questionnaire, emotion and wellness responses, in terms of percentage of the frequency count, were plotted on a line chart, as shown in Figure 2. The two profiles were visually similar, but the frequency counts of the RATA-Words format was higher than those of the CATA-Words format. For CATA-Words in Figure 2A, the "refreshed" term was popular among consumers for roselle (74.64%), Jubliang (65.55%), chrysanthemum (68.42%), and Krachai Dam drink (54.55%), while "good", "happy", and "worried" terms were not significantly different among the five herbal drinks (p > 0.05; Table  7). For RATA-Words in Figure 2B, all 15 terms were significantly different among the five products. Thus, the RATA-Words show a slight improvement in the number of significantly different terms. Therefore, using the RATA scale with words (Test 5) showed slightly better discrimination ability than the CATA scale with words (15 vs. 12) in terms of increasing the number of significant terms.  Soares et al. [33] mentioned that the word-association (WA) method is used in consumer and marketing research to provide insights on perceptions and attitudes to achieve a deeper understanding of what consumers really think and feel about a product. Many researchers have studied the benefit of herbs, such as roselle, ginger and chrysanthemum, and found that they are a good source of vitamins and minerals and rich in phenolic compounds known to have free-radical-scavenging abilities [34][35][36][37][38]. In this study, consumers selected the "active" and "energetic" emotion terms, as well as the "invigorated", "refreshed", and "relaxed" wellness terms for these five herbal drink products. It is possible that some participants realized the important benefits of these herbal drink; thus, they chose those emotion and wellness terms based on their prior knowledge and consumption experiences. It is known that characteristics of an individual participant may affect the emotion profile of foods [2]; these characteristics include personal preference, past experience, frequency of consumption [30], and culture [39].

Discriminating Herbal Drinks Based on Emotion and Wellness Responses
To determine the overall product differences and which emotion and wellness attributes were mainly responsible for product discrimination, MANOVA and DA were performed (Table 8). When considering the MANOVA results of rating scores and RATA scores, significant differences were observed among the five herbal drinks in terms of emotion and wellness responses (p < 0.0001). Based on the first canonical dimension (Can 1), DA identified the "warm" emotion and the "refreshed" wellness terms as the two most discriminating terms among the herbal drinks based on questionnaire responses. However, one additional wellness term ("healthy") was identified in Can 1 for the CATA frequency questionnaire. When considering Can 2, "active" and "comforted" terms significantly contributed to overall product differences. However, RATA scores identified one more wellness term ("relaxed"), while rating scores and CATA frequencies identified two more wellness terms ("relaxed" and "invigorated"). In summary, DA of CATA data revealed the greatest number of terms that discriminated the products, including two positive emotion terms ("warm" and "active") and all five wellness terms ("refreshed", "healthy", "comforted", "relaxed", and "invigorated") with greatest explained variance (93.6%). This finding (Table 8) is not in agreement with that previously reported based on the results from Tables 6 and 7. One possible explanation is that the former data analysis took into account intercorrelations among emotion and wellness attributes, while the latter did not.
PCA was performed to investigate the relationship between the five herbal drinks and emotion and wellness responses. PCA biplots of rating scores and RATA scores were compared ( Figure 3A,B) with the first two PCs (PC1 and PC2), explaining 78.98% and 77.50% of total variance, respectively. For rating-words data, the PCA biplots identified four groups of herbal drinks: I-ginger drink, II-Krachai Dam drink, III-roselle drink, and IVchrysanthemum-Jubliang drink ( Figure 3A). Ginger drink and roselle drink corresponded with PC1 but in the opposite quadrant, while Krachai Dam drink and chrysanthemum-Jubliang drink corresponded with PC2 but in the opposite quadrant as well. Ginger drink correlated with "warm", "healthy", "disgusted", and "bored". Roselle drink correlated with "refreshed", "invigorated", and "active". Krachai Dam drink correlated with "energetic" and "worried". Chrysanthemum and Jubliang drinks correlated with "polite" and "comforted".
For RATA-words data, the PCA biplots revealed three groups of herbal drinks: Iginger drink, II-Krachai Dam drink, and III-roselle, chrysanthemum, and Jubliang drinks ( Figure 3B). Ginger drink and roselle drink corresponded with PC1, and Krachai Dam drink and Chrysanthemum drink corresponded with PC2, in opposite directions. Jubliang drink corresponded with PC3 (not shown in Figure 3), as it could not be explained using PC1 and PC2. Ginger drink correlated with "warm", "healthy", and "disgusted". Roselle drink correlated with "refreshed", "invigorated", "happy", "satisfied", and "polite". Krachai Dam drink correlated with "energetic" and "active". Chrysanthemum drink correlated with "polite", "comforted", and "bored". Obviously, the "warm" emotion term and the "refreshed" wellness term were the most important terms differentiating the hot ginger drink from the cold herbal drinks (Table 8 and Figure 3). Table 8. Canonical structure (r's) 1 describing group differences among the five herbal drinks based on emotion and wellness responses obtained from the rating, CATA, and RATA questionnaires. The CA symmetric plots of CATA frequency and RATA frequency were compared ( Figure 4A,B). The first two dimensions (Dim1 and Dim2) accounted for 94.79% and 89.60% of variance, respectively. CA configuration revealed four groups of herbal drinks: I-ginger drink, II-Krachai Dam drink, III-roselle drink, and IV-chrysanthemum-Jubliang drink. The ginger drink and the other four herbal drinks were clearly separated, on the opposite sides of Dim1 in both CA configurations. However, the CA configurations of CATA data vs. RATA data revealed some slightly different relationships between herbal drinks and emotion and wellness profiles. For CATA data ( Figure 4A), ginger drink correlated with "warm", "healthy", and "disgusted". Roselle drink correlated with "refreshed" and "invigorated". Krachai Dam drink correlated with "active". Chrysanthemum and Jubliang drinks correlated with "polite", "relaxed", and "comforted". For RATA data ( Figure 4B), ginger drink correlated with "warm" and "disgusted". Roselle drink correlated with "refreshed" and "invigorated". Krachai Dam drink correlated with "energetic" and "active". Chrysanthemum and Jubliang drinks correlated with "polite", "comforted", "relaxed", and "happy".

Responses
King et al. [31] evaluated the impact of rating scale and CATA on measurement of emotional responses. They found that the rating scale provided differentiation for more attributes at lower levels of emotional response, while CATA provided greater differentiation at higher levels of emotional frequency for a few select emotions. The rating scale was more sensitive than CATA, but both were found to be acceptable approaches, depending on the objective of the test [31]. In addition, Ng et al. [6] evaluated the effectiveness of the CATA approach compared to intensity scaling for the EsSense ® Profile. They reported that the qualitative nature of the data obtained from CATA limited the extent of statistical analysis, making it difficult to draw clear inferential conclusions from the data obtained with the EsSense Profile ® . Therefore, using a combination of both approaches, RATA was proposed. The CA symmetric plots of CATA frequency and RATA frequency were compared ( Figure 4A,B). The first two dimensions (Dim1 and Dim2) accounted for 94.79% and 89.60% of variance, respectively. CA configuration revealed four groups of herbal drinks: I-ginger drink, II-Krachai Dam drink, III-roselle drink, and IV-chrysanthemum-Jubliang drink. The ginger drink and the other four herbal drinks were clearly separated, on the

Overall Discussion
Overall, all types of data (rating, RATA, and CATA) showed similar results in terms of emotion and wellness profiles of the five herbal drinks studied. The "warm" and "healthy" terms were important and relevant to ginger drink, and the negative emotion term "disgusted" was elicited. "Refreshed" and "invigorated" were the important well ness terms for roselle drink. "Active", "energetic", and "worried" emotion terms were  Meyners et al. [27] mentioned that compared to CATA questions, the use of a RATA variant was found to increase the number of attribute terms selected to describe samples and led to a slight increase in the percentage of terms for which significant differences among samples were identified. Jaeger et al. [40] compared the CATA and RATA question formats using an emoji questionnaire. They found that neither CATA nor RATA emojiquestions were regarded by consumers as difficult or tedious. Their recommendation for method selection was to use CATA emoji questions when samples have distinct emotional associations, whereas RATA seems to better discriminate between samples with more similar emotional profiles. However, in this study, the CA configurations of CATA data vs. RATA data revealed only slightly different relationships between herbal drinks and emotion and wellness profiles (Figure 4).

Overall Discussion
Overall, all types of data (rating, RATA, and CATA) showed similar results in terms of emotion and wellness profiles of the five herbal drinks studied. The "warm" and "healthy" terms were important and relevant to ginger drink, and the negative emotion term "disgusted" was elicited. "Refreshed" and "invigorated" were the important wellness terms for roselle drink. "Active", "energetic", and "worried" emotion terms were often elicited by Krachai Dam drink. Additionally, "comforted", "relaxed", and "bored" were significant terms for describing the chrysanthemum and Jubliang drinks. Although chrysanthemum drink had the highest overall liking scores (Table 5), some consumers expressed a "bored" emotion. It is a challenge of product developers to reduce the negative emotions associated with products. Elicited negative emotion terms should be carefully considered when developing or improving products. In the case of herbal drinks, ginger drink was associated with "disgusted" by some consumers, possibly because of its natural spicy flavor. Krachai Dam drink was associated with "worried" because it is not well-known by many Thai consumers and it normally imparts a pungent and sour flavor. "Bored" was elicited by chrysanthemum, possibly because it is a well-known traditional product that has not changed for a long time (i.e., not much product innovation applied to this drink).
When the three types of measuring methods-a rating scale, CATA question, and RATA question-were compared, they all produced similar emotion and wellness profiles, though there were advantages and limitations for each method. For a rating scale, the participants needed more time to complete the questionnaire than for other methods. However, a rating scale is appropriate for measuring intensity of attributes. The PCA and CA of rating scores were able to differentiate five herbal drinks into four groups with a higher percentage explained variance (78.98%) when compared with the RATA score (three groups; 77.50%). The CATA questionnaire was the most consumer-friendly method and was found to be appropriate for use for measurement of the presence or absence of the selected attributes. DA of CATA data identified the important terms discriminating the overall differences among the herbal tested drinks with the highest percentage of explained variance (93.6% total for Can1 and Can2). Additionally, CA of CATA data separated the five herbal drinks into four groups, similar to those of the rating data, with a higher percentage of explained variance (94.79%) when compared with the RATA frequency (four groups, 89.60%). However, the RATA score and RATA frequency from the RATA-Words question seemed to show better discrimination ability in term of improving the number of significant terms among the five herbal drinks.

Conclusions
In this study, we demonstrated the effects of two factors in designing a questionnaire for measurement of emotion and wellness responses elicited by herbal drinks consumed by Thai consumers. The first factor was questionnaire items in the form of words vs. full sentences. Results indicate that measuring emotion and wellness using full sentences did not provide a clear benefit over using words alone. When using familiar terms clearly understood by consumers, a full sentence in the questionnaire is not needed. For the second factor, three types of measuring methods, including a rating scale, CATA question, and RATA question were compared for their ability to discriminate among five herbal drinks. All three measuring methods produced similar emotion and wellness profiles. However, each method has advantages and limitations that researchers should carefully consider. Overall, this study provides some useful options in terms of questionnaire-item formats and measurement methods in the design of questionnaire to measuring consumers' emotion and wellness responses to herbal drinks. The information elucidated in this study is applicable to other food products and beverages. In the future, the impact of health-benefit statements on consumer perception and purchase intent with regard to herbal drinks formulated in our lab will be assessed by product tasting and with an appropriate sample size representing the target population. We also plan to collaborate with a medical school for human clinical trials to determine the health benefits of these herbal drinks. Funding: This research receives funding from Sakon Nakhon Rajabhat University (grant no. 0213).

Institutional Review Board Statement:
The study was reviewed and approved by the Ethics Committee of Khon Kaen University (HE611230, approved on 7 June 2018).

Informed Consent Statement:
Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data presented in this study are available in the manuscript.