Effectiveness of a Smartphone App (e-12HR) in Improving Adherence to the Mediterranean Diet in Spanish University Students by Age, Gender, Field of Study, and Body Mass Index: A Randomized Controlled Trial

There is an urgent need to implement intervention programs to promote adherence to the Mediterranean diet (AMD) in university students to prevent non-communicable diseases. A powerful tool for this is smartphone apps. Furthermore, it is necessary to determine the subgroups that are most likely to benefit from these technologies. The objective is to evaluate the effectiveness of an app (e-12HR) at improving AMD in a sample of Spanish university students and different strata. The study method was a controlled and randomized clinical trial over a four-week follow-up period and involving 385 participants (76.9% women). The participants were in two parallel groups: the control group (CG) and the intervention group (IG), with only the IG receiving feedback to improve their AMD. There were significant statistical improvements (with higher values in the IG) at week four, after no significant statistical differences at baseline (Week One): in the whole sample: +25.7% AMD index and +74.5% percentage with moderate/high AMD index. In the subgroups, seven of eight subgroups, ranging in AMD index from +17.8% (≥20 years) to +33.0% (<20 years); and for males, in weeks two (+27.9%) and three (+23.9%), but not at week four. In conclusion, e-12HR could improve AMD among university students (in the total sample and all subgroups, except ≥25 kg/m2).


Introduction
University students are particularly vulnerable to nutritional alterations mainly due to the fact that many are faced for the first time with the opportunity to make their own dietary choices [1,2], followed by the major changes they experience during their university years, both in terms of physical and social development [3]. In addition, university students are exposed to several factors, such as stress and lack of time, combined with scant information and an unfavorable food environment (accessibility or the attractiveness and price of certain food products and advertising), which make them less likely to maintain a healthy lifestyle [3,4].
Numerous epidemiological and nutritional studies have shown that a healthy diet, such as the Mediterranean diet (MD), improves the quality of life and helps prevent several non-communicable diseases (NCDs) such as cardiovascular diseases [5], obesity [6], cancer [7], and metabolic syndrome in adults [8]. The MD is a dietary pattern that is the influence of these factors (age, gender, field of study, and BMI) on the effectiveness of a smartphone application to improve AMD among university students. In addition, a secondary objective was to establish the usability of the app in the entire sample of Spanish university students.

Design Overview
The present study is a continuation of previous research and follows the same study protocol, which has been described by Béjar et al. [31] in detail (article free for readers published in Nutrients (an open-access journal), Volume October 2022). However, in the current study, the sample and the period for obtaining it were different. In brief, the present study is a randomized and controlled clinical trial with two parallel groups: the control group (CG) and the intervention group (IG). The study took place over a four-week period, and the trial registration can be found at ClinicalTrials.gov, NCT05532137.

Setting and Participants
The Faculties of Medicine, Pharmacy and Communication at the University of Seville (Andalusia, Spain, South of Europe) were included in the study.
Confidentiality was guaranteed in accordance with the Organic Law on the Protection of Personal Data and the Law 14/2007 on Spanish biomedical research.
Inclusion criteria: Both genders, over 18 years old, students of Medicine, Pharmacy and Communication (University of Seville) and possess a smartphone.
Exclusion criteria: intolerance to any food, chronic diseases such as diabetes or pregnancy (situations which may require specific dietary recommendations).
A member of the research group explained the study to potential participants, including objectives, risks and benefits of the research; e-12HR application functionalities; and how to participate (the students had to send an e-mail to the research team).
When the research team received an e-mail from a student, they replied with another e-mail including the following documents: Document one: informed consent to be signed by the student and returned by e-mail; Document two: personal data (sex, date of birth, Faculty, weight, height, and smoking status) to be completed by the student and returned by e-mail; Document three: personal code (with numbers and letters); Document four: instructions for downloading e-12HR (a free app available on the App Store or Play Store); Document five: information for using e-12HR app; and Document six: information about the characteristics of the MD.
This protocol was selected to obtain a high participation rate, avoid unnecessary travel to complete or sign documents, as well as to avoid wasting paper.
Recruitment of participants: September-October 2022. The students who successfully completed the protocol were entered into the raffle for school materials (valued at EUR 500).

Randomization and Masking
Four random classrooms were selected in each school (Medicine, Pharmacy and Communication): twelve classes in all. Of the four selected classrooms in each school, two were assigned to each group (CG and IG), according to the following sequence: first class: IG; second class: CG; third class: IG; and fourth class: CG. In this way, all students in the same class were assigned to the CG or the IG by probability single-stage cluster sampling.
Due to the nature of the study, the students could not be blinded. However, the statistical analysis of the data was performed by a person who remained blinded throughout the study. In addition, each participant only had access to one version of the application (CG: 'non-feedback' e-12HR version; IG; 'feedback' e-12HR version). This was possible by assigning personal alphanumeric codes (mentioned in the previous section).
The allocation sequence is detailed in Figure 1.
(CG: 'non-feedback' e-12HR version; IG; 'feedback' e-12HR version). This was possible by assigning personal alphanumeric codes (mentioned in the previous section). The allocation sequence is detailed in Figure 1.

Intervention
CG: 'non-feedback' e-12HR version. IG: 'feedback' e-12HR version. The structure and functions of the e-12HR application ('non-feedback' e-12HR and 'feedback' e-12HR versions) have been described, in detail, by Béjar et al. (in "e-12HR App" in the Section 2.4) [31]. No changes were made in the application compared to the original study protocol. In brief, the 'non-feedback' e-12HR version allowed the user to collect food consumption data; the 'feedback' e-12HR version allowed for the collection of food consumption data and, additionally, issued personalized recommendations to improve AMD.

Follow-Up and Outcome Measures
In order to assess the effect of the app ('feedback' e-12HR), follow-up was carried out at week one (baseline), week two, week three and week four of monitoring.
Main result variable: the change in the AMD score at weeks two, three and four of monitoring.
Secondary result variables: the personal information variables, and the answers to the usability rating questionnaire for e-12HR (see Section 2.7).

Intervention
CG: 'non-feedback' e-12HR version. IG: 'feedback' e-12HR version. The structure and functions of the e-12HR application ('non-feedback' e-12HR and 'feedback' e-12HR versions) have been described, in detail, by Béjar et al. (in "e-12HR App" in the Section 2.4) [31]. No changes were made in the application compared to the original study protocol. In brief, the 'non-feedback' e-12HR version allowed the user to collect food consumption data; the 'feedback' e-12HR version allowed for the collection of food consumption data and, additionally, issued personalized recommendations to improve AMD.

Follow-Up and Outcome Measures
In order to assess the effect of the app ('feedback' e-12HR), follow-up was carried out at week one (baseline), week two, week three and week four of monitoring.
Main result variable: the change in the AMD score at weeks two, three and four of monitoring.
Secondary result variables: the personal information variables, and the answers to the usability rating questionnaire for e-12HR (see Section 2.7).
Adherence to the MD Every week during the four-week study period, the AMD index score (specifically, Mediterranean Diet Serving Score (MDSS) index [32]) was calculated manually (for CG and

Sample Size Calculation
The sample size estimation was made for the main result variable: the change in the MDSS index. Considering a standard deviation of 2.7 points in the MDSS index and a dropout rate of 20.6% (from a previous study using e-12HR [31]), α = 0.05 and β = 0.20, bilateral test, 292 participants (146 per group) were needed to detect a variation of 1 point in MDSS index in the IG versus CG.
The sample size was calculated using the nQuery Advisor Release 7.0 program.

Usability Rating Questionnaire for e-12HR
After the four-week monitoring period, a member of the research group sent e-mails to students (with a usability rating questionnaire for the e-12HR app [31], included in Appendix A, Table A1).

Ethical Considerations
Participants were required to sign the informed consent prior to inclusion in the study, according to the Declaration of Helsinki.
The study was approved by the University of Seville's Research Ethics Committee on 30 March 2022 (internal identifier: 2813-N-21). Trial Registration: ClinicalTrials.gov, identifier NCT05532137.

Statistical Analysis
The results were displayed as numbers (percentages) for qualitative variables and as means (standard deviations) for quantitative variables.
The data were tested for normality using the nonparametric Kolmogorov-Smirnov test.
In order to compare proportions, the chi-square test (or Fisher exact test) was carried out as appropriate, and Student's t-test (or the nonparametric Mann-Whitney U-test) was used for the comparison of continuous variables.
A p-value < 0.05 was considered significant. Statistical analyses: using the SPSS statistical software package version 26.0 (SPSS Inc., Chicago, IL, USA).

Sample and Adherence to the Study
A total of 491 students signed the informed consent forms; however, two students were excluded for being diabetic (one student from the IG and one from the CG), failing to meet the selection criteria. Of those who signed, 104 (64 in the CG and 40 in the IG) were considered to be non-responsive, as they did not complete the study's 28-day follow-up period ( Figure 1). The data for these individuals were not included in the later statistical analysis.
Overall, the study response rate was 78.  Table 1 shows the personal information of the participants (CG and IG).  No significant statistical differences were observed in the variables studied (CG versus IG) except in the "field of study", although, in both groups, Health Science students exceeded 70% of the sample (Table 1).

Personal Information of the Participants
No significant statistical differences were observed in the variables studied between responsive (those who completed the study) and non-responsive (those who did not) participants.
The responsive participants registered their daily consumption for the 19 food groups included in the study (fruits, vegetables, cereals -breakfast cereals, pasta, rice, and bread-, olive oil, milk and dairy products, nuts, fermented beverages -wine and beer-, potatoes, legumes, eggs, fish, white meat, red meat, processed meats and sweets [31,32]) for 10,780 days altogether (385 participants and a 28-day follow-up period). This value represents a collected total of 204,820 data points on daily consumption for the food groups.

MDSS Index
For both groups (CG and IG), scoring for the MDSS was calculated manually by the research team [31]. In this process, the research team modified the obvious errors made by participants during data entry (as it was considered that the data must have been introduced as milliliters or grams instead of standard servings). For example, on several occasions, a value between 50 and 70 was introduced for the question, "How many servings of pasta have you consumed today?" The research team considered that these values indicated a consumption between 50 and 70 g, which is the equivalent of one serving. In any case, only 692 data points were modified by the research team (out of a total of 204,820 registered data points: 0.31%). There were significant statistical differences for both the MDSS index and the percentage of participants with moderate/high (≥9) MDSS index (CG versus IG) in weeks two, three and four, with higher values in the IG (no significant differences in week one): for the MDSS index with 1.25, 1.78 and 1.93 points of improvement, respectively, and for the percentage of participants with moderate/high (≥9) MDSS index with increases of 19.2, 20.7 and 24.2 percentage points, respectively ( Table 2).  Table 3 shows the percentage of participants that meet the consumption criteria for each food group, for the CG and the IG, throughout the four weeks of follow-up in the whole study sample. In addition, Table 3 shows the MDSS index throughout the four weeks of follow-up (CG and IG) in the whole study sample (these data have already been previously collected in Table 2; however, they have been included again for easy comparison by readers with the data from food groups).

Effect of the Intervention
No statistically significant differences were observed at week one in any of the food groups except fish. It should be noted that all food groups that, according to the MDSS index, present a daily consumption recommendation [32] (except fermented beverages) showed statistically significant differences (CG versus IG) throughout the four weeks of follow-up: for fruits, cereals, olive oil and nuts in weeks two, three and four; for vegetables in weeks three and four; and for milk and dairy products in week four. Additionally, regarding the food groups with weekly recommendations [32], there was a statistically significant difference (CG versus IG) for legumes at week four. In those subgroups in which statistically significant differences were observed at week four (and no significant differences at week one), CG versus IG, the IG showed higher percentages, specifically: +14.7 percentage points for fruits, +11.2 percentage points for vegetables, +17.2 percentage points for cereals, +15.7 percentage points for olive oil, +10.8 percentage points for milk and dairy products, +7.1 percentage points for nuts and, finally, +10.4 percentage points for legumes (Table 3).

Effect of the Intervention in Terms of Variation in MDSS Index in Different Subgroups of the Study Sample
The differences were statistically significant considering the MDSS index (CG versus IG): in weeks two, three and four for the subgroups <20 years, female, Health Science and <25 kg/m 2 ; in weeks two and three for male; and in week four for ≥20 years and Non-Health Science. However, there were no statistically significant differences at any week in the subgroup ≥25 kg/m 2 . In those subgroups in which statistically significant differences were observed at week four (no significant differences at week one in any of the subgroups), CG versus IG, the IG showed higher MDSS index values, specifically: +2.34 points for <20 years, +1.43 points for ≥20 years, +2.03 points for female, +1.86 points for Health Science, +1.39 points for Non-Health Science, and, finally, +2.18 for <25 kg/m 2 ( Table 4).

Usability Rating Questionnaire for e-12HR
This questionnaire was answered by 127 students (66 from the CG and 61 from the IG).
The responses of the users are shown in Table 5.
Considering the 127 participants who answered the questionnaire, all (100%) reported that e-12HR was easy to complete, and most of them indicated that: (1) the application contained understandable questions (97.0% CG and 91.8% IG); (2) the app contained understandable feedback (only for the IG, 85.2%); (3) they would be willing to complete the e-12HR app again, (53.0% CG and 63.9% IG); and (4) the time to complete the task was 2 min or less (56.1% CG and 57.4% IG). There were no statistically significant differences (CG versus IG) for any of the questions on the questionnaire (Table 5).

Discussion
Few randomized and controlled clinical trials have analyzed the effectiveness of promoting the MD (in Spanish adults) of smartphone apps with certain similarities to e-12HR [33][34][35][36]. The smartphone applications EVIDENT II [33][34][35] and SalBi Educa Nutrition [36] shared functionalities with e-12HR; all of them allowed entry of food intake and offered personalized dietary advice. According to Recio-Rodríguez et al. [34], future research should clarify the possible effects certain factors (such as age, gender, or educational level) might have on the success of using diet applications; in this way, it would be possible to determine the subgroups that are most likely to benefit from the support of these technologies. This work is pioneering since, to the best knowledge of the research team, it was the first to evaluate the influence of certain factors (such as age, gender, field of study, and BMI) on the ability of an application to improve AMD in adults (specifically, university students).
The main findings of this work were (significant intergroup statistical modifications): First: in relation to the main objective, in the entire sample of Spanish university students, the increase in AMD from week two throughout the follow-up of the study (at week four the increase was favorable to IG versus CG by +25.7% for MDSS index and by +74.5% for the percentage of participants with moderate/high (≥9) MDSS index) ( Table 2); with improvements observed in all food groups with daily consumption recommendations [32] (except fermented beverages), as well as for legumes (Table 3). In the subgroups, the increase for MDSS index throughout the follow-up of the study in seven of the eight subgroups considered (at week four, the increase was favorable to IG versus CG by +33.0% for <20 years, +17.8% for ≥20 years, +27.1% for female, +23.5% for Health Science, +21.0% for Non-Health Science, and, finally, +28.8% for <25 kg/m 2 ); for male, there were higher data of MDSS index in IG versus CG in weeks two (+27.9%) and three (+23.9%), but not at week four (Table 4). Second: regarding the secondary objective, the responses of the participants to the usability rating questionnaire for e-12HR have shown that the app has good usability in the whole of the sample and the different strata considered (Table 5). Usability is an important aspect in the field of applications. In fact, for healthcare professionals, the three principal criteria for selecting a "Nutrition and Diet" app for their clients/patients were [37]: ease of use (e-12HR obtained very positive responses to the usability rating questionnaire), apps being free of charge (download of e-12HR is free) and validated (e-12HR is a pre-validated app [30,[38][39][40][41]).
The increase in MDSS index (CG versus IG) with the use of e-12HR can be considered moderate (with higher values in the IG): in the entire sample of university students, +25.7%; in the subgroups, ranging from +17.8% (for ≥20 years) to +33.0% (for <20 years). However, this moderate increase among university students could be considered, at the same time, promising. University students are not characterized by being in high motivation stages to change a lifestyle factor (preparation or action, following the model of Prochaska and Velicer [42]), i.e., they are, in general, participants with little motivation for diet change. In this study, the motivation stage for the change of the participants was not formally collected; however, in the informal contact with the students, they stated that, for the most part, they were not in phases of high motivation. Therefore, the research team hypothesizes that the use of e-12HR could lead to greater increases in AMD with a sample of participants with greater motivation to change their diet (new studies will be performed to test this hypothesis, see "Future Research Related to the Current Study" Section).
As previously mentioned, some randomized and controlled clinical trials have used applications with certain similarities to e-12HR to improve AMD in Spanish adults [33][34][35][36]; however, these studies differed from this study and other ones by the research team [31] (both carried out among university students, using MDSS index and with a four-week follow-up period), in the participants selected (patients of healthcare centers [33,34,36] or patients with diabetes mellitus type 2 [35]), the length of the follow-up period (three months [33][34][35], only the study by Gonzalez-Ramirez et al. [36] had a similar duration of four weeks), and the AMD index used (Mediterranean Diet Adherence Screener (MEDAS) [33][34][35][36]). These differences between studies should be considered when comparing the results. In the intergroup comparisons (CG versus IG), in line with the results of this work and those of a previous study by the research team [31] (both with similar results in the entire sample -previous study: +17.4% for MDSS index and +61.9% for the percentage of participants with moderate/high (≥9) MDSS index for the IG-), statistically significant differences were observed (with higher values in the IG) in the study by Alonso-Domínguez et al. [35]: with moderate increases in MEDAS index and percentage of participants with adequate AMD (MEDAS score ≥ 9 points) and, according to food groups, with improvements in the consumption of several food groups, such as olive oil, vegetables, fruits, fish, commercial baked, nuts, sofrito sauce, and white meats after three months. It should be noted that e-12HR's effect was noticeable after only four weeks of the intervention. It must be considered that, in the study by Alonso-Domínguez et al. [35], the intervention combined a food workshop, five walks and a smartphone application (EVIDENT II). Due to the multifactorial nature of the intervention, it is not possible to know which component produced the change in the IG in comparison to the CG [35]. No intergroup modifications (CG versus IG) were shown in the rest of the studies (for the MEDAS index, for the percentage of participants with adequate AMD MEDAS score ≥ 9 points, or for food groups) throughout the periods of use of the applications [33,34,36]. However, this should not be interpreted as these applications not being effective at improving the quality of the diet; in fact, EVIDENT II and SalBi Educa Nutrition apps proved to be useful for significantly increasing carbohydrate intake and decreasing total fat intake (CG versus IG, with higher values in the IG) [34,36]. Intakes of macro and micronutrients have not been measured in the current study.
This study presents several limitations. First, all the data collected were self-administered: the participants completed the daily nutrition questionnaire using the app (e-12HR being a self-reporting method, it presents the inherent limitations of this type of tool, described in detail in the bibliography [43-49]), and, at the end of the study, answered the usability rating questionnaire for e-12HR. In addition, the intervention was short (four weeks), and the long-term AMD index (once the use of the app has finished) is unknown. Furthermore, there was a relatively small number of individuals in some of the subgroups analyzed: male (n = 89), Non-Health Science (n = 92), and ≥25 kg/m 2 (n = 60). Finally, the nature of this study made it impossible to blind the participants or to guarantee that the participants were not using another nutrition app during the follow-up period. Regarding the latter, a multifactorial intervention (combining the simultaneous use of e-12HR and another application to improve AMD) would mean that it would be impossible to know which component produced the change in the IG.

Future Research Related to the Current Study
In those subgroups of the population that have improved AMD with the use of e-12HR, the research team intends to evaluate the possible increase in the MDSS index by combining the use of the app with counseling (counseling focused on food groups that have not improved consumption with the use of e-12HR, such as eggs, white meat, red meat, etc.): CG ('non-feedback' e-12HR version) versus IG ('feedback' e-12HR version + counseling) [35]. Additionally, the research team intends to evaluate the effectiveness of e-12HR at improving AMD among participants truly motivated to change their diet (i.e., who are in the preparation or action stages of change [42]).

Conclusions
The results of this study support recommending the use of e-12HR in university students as a tool to improve AMD in the short term, in the total sample and in all its subgroups, except ≥25 kg/m 2 (group in which no improvement was observed in AMD throughout the follow-up period). Additionally, the application presents satisfactory usability in the whole of the sample and the different strata considered.