Prevalence of the Patterns of Unhealthy Diet in the School and University Students of Iran: A Systematic Review and Meta-Analysis

Introduction The present study was conducted to investigate the pooled prevalence rate of the different patterns of unhealthy diet among the school and university students of Iran. Methods In this systematic review, the type of the main question was regarding prevalence and the effect measure was prevalence rate reported along with 95% confidence interval (CI). Data bases including PubMed, Scopus, and Web of Science as well as Google Scholar and Persian resources were used. The Newcastle–Ottawa scale (NOS) checklist was used for quality assessment of studies. Results The extracted types of unhealthy diet in the present systematic review were “breakfast skipper,” “fast food,” “hydrogenated oils consumption,” “salty snacks,” “sweetened beverages,” “breakfast skipper,” “dinner skipper,” “launch skipper,” and “sweets.” The range of pooled prevalence for different types was 0.06–0.75. The data of 16,321 subjects included in six studies were analyzed. The pooled prevalence of unhealthy diet was 0.28 (95% CI: 0.23–0.33, I2 > 99%) overall, 0.25 (95% CI: 0.20–0.31, I2 > 99%) in school students and 0.37 (95% CI: 0.12–0.62, I2 > 99%) in university students. The most prevalent pattern was breakfast skipping 0.39 (95% CI: 0.28–0.50) followed by consumption of sweetened beverages 0.31 (95% CI: 0.20–0.43). The pooled prevalence range among the patterns was 0.06–0.75 (random effects for all). Conclusion The pooled prevalence was 28% for unhealthy diet among the Iranian students (6% to 75% in different patterns). Although there was uncertainty regarding the pooled evidence, the whole of the mentioned range was clinically important for health policymakers. Decisions should be made on the basis of the patterns.


Introduction
Adolescence is a period that has a high prevalence of risky behaviors such as drug and alcohol use, inadequate sleep and lifestyle problems [1], and many unsuitable nutritional habits are formed during this period [2].Many of noncommunicable diseases that occur in middle and old age are the result of unhealthy lifestyle and improper nutritional habits in childhood and adolescence [3].
Te World Health Organization (WHO) states that proper nutrition is a cornerstone of good health [4].Proper nutrition is a determining factor in health and ensuring adequate growth.Teenagers need to follow a well-designed diet due to their active lifestyle.Skipping meals, using salty and fatty foods, and consuming sweetened beverages are among the unhealthy eating habits among adolescents and school-aged children.Unhealthy eating habits at this age, when growth is fast and nutritional needs are greater, can interfere with the growth and health of adolescents [5].In a study that examined the habit of eating breakfast in schoolaged children, it was seen that only 30.9% of participants did not intake full breakfast (six times or less) and 69.1% consumed a complete one [3].In another study, the frequency of breakfast, lunch, and dinner consumption was 37%, 56.5%, and 46.1% in six days of a week, respectively.About 7.8% of these students used carbonated sweet drinks, 13.6% of salty snacks, and 34.1% of sweets on a daily basis [6].Other studies have also been conducted in the feld of breakfast consumption in primary school children, which have similar results and show that 8 [7], 8 [8], and 2.5 percent of these children [9] do not eat breakfast.
Te importance of dietary patterns is not limited to the nutritional health.Many mental illnesses in adolescence are also related to the dietary pattern of adolescents [10].It is believed that nutritional status in children and adolescents is associated with psychiatric disorders as food intake has an efect on brain development and function.For instance, neurodevelopmental disorders may be associated with prenatal and childhood intake of essential fatty acids [11].A school-based cross-sectional study in a large population of Brazilian adolescents showed that diferent patterns of unhealthy diet are associated with common mental disorders [12].On the basis of the present evidence, it is expected that unhealthy diet in childhood, adolescence, and youth is a health issue.According to the role of social and cultural status in dietary patterns [13], the mentioned issue may become a specifc problem in some countries and populations.Knowing the frequency and patterns of unhealthy diets is an important step to resolve the stated problem.Unhealthy diet may have diferent patterns and examples, and each of them has its own prevalence.Such prevalence rates can be reported as range or pooled rates.Although the patterns are diferent, it is not without grace to pool the results.
As stated, unhealthy diet in adolescents is a major problem in the world and is categorized as a high-risk behavior.Iran as a developing country is not an exception.So far, there are some large cohort studies such as "Childhood and Adolescence Surveillance and PreventIon of Adult Non-communicable disease" (CASPIAN) conducted in Iran in this regard [14,15].However, a meta-analytic approach was needed to pool the results of diferent studies.Te present study was conducted to investigate the pooled prevalence rate of the diferent types of unhealthy diet among the students of Iran.

Study Design.
Te present systematic review and metaanalysis was conducted according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement.Te type of the main question was regarding prevalence, and the efect measure was the prevalence rate.Tis question consisted of the prevalence of diferent types of unhealthy diets in the context of students in Iran.Te efect measure method was prevalence rate.Te protocol of the study was approved by the Research Council of the Psychosocial Health Research Institute, Iran University of Medical Sciences, with registration number 13919.

Sources and Search Strategy.
Our general search strategy was the following which was searched without limiting the search to the titles, abstracts, or keywords.Databases such as Scopus, PubMed, and Web of Science (WOS) were used.Since it was expected that there would be some important papers in Persian sources, Persian databases including Magiran, Scientifc Information Database (SID), and Noor as well as Google Scholar were used by two independent researchers.Te limitation of Persian sources was that the exact syntaxes were not properly worked, and therefore, the researchers searched manually and screened the frst 50 results of each database.Tere was no time limitation.Te search process was done in December 2022.
Search strategy: (Student OR Pupil OR college OR school) AND (Prevalence OR Epidemiology OR Incidence OR Frequency OR Outbreak OR Occurrence OR "Population-Based") AND (Violence OR "Assaultive Behavior" OR "high risk behavior" OR "high risk behaviors") AND (Diet * OR Feeding OR Eating OR Nutrition * OR Food * ) AND (Iran).
After removal of duplicates, the remaining documents were subjected to screening based on the titles and the abstracts by two independent researchers (MRB and AM) and investigation of a third researcher (SAYA) if necessary.After screening and adding the papers from Persian sources, citation search, and manual search in Google Scholar (by SAYA), the full texts were scrutinized for investigation of the eligibility criteria.Te eligibility criteria included crosssectional studies reporting prevalence rate of unhealthy diet (in general and per type), school or university student samples, and conducting the research in Iran.Te age range was not regarded, but the participants should be school or university students.No time range was considered.Tese criteria were also subjected to data extraction.We extracted the patterns of unhealthy diet, the prevalence rates of these patterns, and demographic variables from the included studies.
2.3.Quality Assessment.Quality assessment of the studies was performed by the Newcastle-Ottawa scale (NOS) checklist for cross-sectional studies [16].Tis checklist consisted of the parts' selection (in turn, including representativeness of the sample 0-1 point, sample size 0-1 point, nonrespondents 0-1 point, and ascertainment of the exposure 0-2 points), comparability 0 and 2 points, and outcome (in turn, including assessment of outcome 0-2 points and statistical tests 0-1 point).Te total scoring categories were unsatisfactory (<5 points), satisfactory (5-6 points), good (7-8 points), and very good (8-10 points).Te process was conducted by two independent researchers.Te Scientifc World Journal analysis was performed with -by()-option of the command.Te results were reported as forest plots.Funnel plots were used for visual investigation of publication bias based on asymmetry of the distribution of the efect sizes.Sensitivity analysis was performed to investigate the efect of each study removal on the pooled prevalence.Cumulative metaanalysis and meta-regression were used to investigate the efects of the demographic variables and their trends on the pooled efect.

Baseline Findings.
According to the PRISMA fowchart (Figure 1), a total of six papers were eligible for meta-analysis [3,6,14,15,17,18].Te time range of publication was from 2011 to 2018.One study investigated the outcome among the university students, while the other studies were among school students.Te smallest sample size was 289, while the largest sample size was 13486 that was for a famous cohort called CASPIAN (totally, the data of 16,321 cases were analyzed).Te extracted patterns were classifed as "breakfast skipper," "dinner skipper," "fast food," "hydrogenated oil consumption," "lunch skipper," "poor nutrition," "salty snacks," "sweetened beverages," and "sweats."Te most common pattern was "breakfast skipper" as there were fve studies for this pattern, followed by "sweetened beverages" and "salty snacks" with four studies for each.
According to the quality assessment, all the studies were at least satisfactory.Te baseline information of the studies is shown in Table 1.

Pooled Analysis.
A total of nine study types of unhealthy diet groups from six studies were analyzed (including 23 study-subgroups).A signifcant heterogeneity was found between the study type of unhealthy diet group (I 2 > 99%, P < 0.001).Because of heterogeneity, large variation in sample sizes, and overlap of samples in the subgroups, a random efects model was considered.Accordingly, the pooled prevalence among all the study types of unhealthy diet groups was 0.28 (95% CI: 0.23-0.33).Pooled analysis was conducted per the type of unhealthy diet.Accordingly, among the subgroup analyses with three or more populations, fast food consumption showed the lowest prevalence 0.06 (95% CI: 0.02-0.10)followed by salty snacks 0.15 (95% CI: 0.12-0.18).Te most prevalent pattern was breakfast skipping 0.39 (95% CI: 0.28-0.50)followed by consumption of sweetened beverages 0.31 (95% CI: 0.20-0.43).Other subgroups had fewer populations.Te heterogeneity between the groups was statistically signifcant (P < 0.001) (Figure 2).In addition, subgroup analysis was performed for education.No signifcant diference was observed between university and school students (P � 0.347) (Figure 3).

Publication Bias.
Funnel plots showed asymmetry for the distribution of the efect sizes of the smaller studies.Most of the studies were outside the funnel (Figure 4).

Sensitivity Analysis.
Pooled efect is shown for removal of each study type of unhealthy diet (leave-one-out approach).Te lowest and highest possible efect sizes were 0.26 and 0.29, respectively (Figure 5).Subgroup analyses can also be interpreted as sensitivity analysis.

3.5.
Cumulative Analysis and Meta-Regression.Cumulative meta-analysis showed a negative visual trend for the time of publication.However, this trend was not statistically signifcant based on meta-regression (Figure 6).Cumulative meta-analysis showed a negative visual trend for sample size.However, this trend was not statistically signifcant based on meta-regression (Figure 7).No visual pattern of trend was observed for age (Figure 8).Cumulative meta-analysis showed a negative visual trend for male proportion.However, this trend was not statistically signifcant based on meta-regression (Figure 9).

Discussion
According to the stated problem and the rationale, the present study was conducted to achieve the patterns of unhealthy diet in the literature, as well as the range and the pooled prevalence of diferent types of unhealthy diet in the context of Iranian students.A total of 16321 persons from six studies were investigated.Eight studies were found from the CASPIAN cohort, and we selected only one of which was the most complete and most recent.Studying the characteristics of overall diet resulted in the examination of the combined nutrient efects on health and better prediction of dietdisease relationships [19].To our knowledge, this is the frst systematic review and meta-analysis study to obtain prevalence of the diferent types of unhealthy diet among the students of Iran.Te existing studies in Iran are mainly conducted on adults [20,21].
In terms of risk of bias, all the included studies had at least a satisfactory quality.Te comparability section of the checklist had a two-point item to adjust for confounders.Since the question of our study was prevalence, this item was    Te Scientifc World Journal diferences in subgroups (types of unhealthy diet).In terms of publication bias investigation on the basis of funnel plots, smaller studies showed a higher prevalence.It seemed that the gap of small studies with low prevalence indicated publication bias.However, it did not seem to have much efect on the pooled result.
In the worst scenario of the sensitivity analysis, by removing one study, the total frequency decreases from 0.28 to 0.26.Te results of this part were also in line with the results of the funnel plot and indicated the possibility of a slight drop in the pooled frequency in the absence of publication bias.None of the baseline factors had a signifcant efect on the pooled results.Although age is logically subjected as an infuencing factor, it could not afect the pooled result (P � 0.385).Among these factors, the most efective one was the sample size, which showed a negative trend on the prevalence in the cumulative meta-analysis and funnel plot; however, no signifcant relationship was observed in the meta-regression (P � 0.111).
Diferent patterns of unhealthy diet among the studies may be a potential source of heterogeneity.Comparing diet patterns between studies was somewhat difcult due to diferences in dietary assessment, food groups, and the number of patterns that should be retained for analyses and statistical techniques.Despite inconsistencies between the results derived from diferent studies, there were some 8 Te Scientifc World Journal similarities regarding dietary patterns identifed among children [22].In the present meta-analysis, the consumption of fast food and snacks had a low prevalence.Tis fnding was notable and justifed only in primary schools.
A meta-analysis conducted by Babashahi et al. [23] investigated the pooled daily servings of the most common processed food consumed by children.Teir meta-analysis consisted of 10 papers including 67,093 participants.Te highest consumption was related to the sugars and sweets group followed by oils and then biscuits and cakes [23].However, the present meta-analysis investigated unhealthy diet patterns and also in students who might not necessarily be children.
Te limitations of the study were high heterogeneity and the possibility of publication bias.In spite of existing these limitations, subgroup analysis showed a lower heterogeneity.However, this study presented pooled prevalence rates for the frst time in Iran for each type of unhealthy diet among the students.

Conclusions
Te present systematic review found "breakfast skipper" as the most commonly investigated pattern of unhealthy diet, and as well, the meta-analysis part reported a pooled prevalence of 28% for unhealthy diet among the Iranian students.According to the diferent types of unhealthy diet patterns, the pooled prevalence rate varied from 6% to 75%.Although there was uncertainty regarding the pooled evidence, the whole of the mentioned range was clinically important.Terefore, planning for the reduction of such behaviors is strongly recommended in the present target population.Decisions and plans should be on the basis of the patterns.Te Scientifc World Journal

Figure 5 :
Figure 5: Sensitivity analysis for efect of each study type of unhealthy diet removal.

Figure 6 :
Figure 6: Cumulative meta-analysis and meta-regression for the trend of publication year efect.

Figure 9 :
Figure9: Cumulative meta-analysis and meta-regression for the trend of male proportion efect.

Table 1 :
Baseline characteristics of the eligible studies (six studies with nine unhealthy diet groups).
. For meta-analysis, Stata 17 (Stata Corp. LLC, TX, US) was used with the latest version of packages metan, metafunnel, and metareg.Pooled prevalence rates were calculated for unhealthy diet divided by the pattern with their asymptotic 95% confdence intervals (CI).I 2 test was used to investigate heterogeneity and random efects model was used if I 2 > 50% with P < 0.05.Subgroup 2 Te Scientifc World Journal

Table 1 :
Continued.Figure 2: Forest plot for prevalence of unhealthy diet with subgroup analysis based on the types of unhealthy diet.Figure 3: Forest plot for prevalence of unhealthy diet with subgroup analysis based on education.
not relevant, and as a result, none of our studies received these two points.Instead, studies that had subgroups based on the type of unhealthy diet were given one point.However, all the studies gave a satisfactory score or higher in spite of the mentioned limitation.In terms of heterogeneity, the source of heterogeneity could be attributable to the Figure7: Cumulative meta-analysis and meta-regression for the trend of sample size efect.Cumulative meta-analysis and meta-regression for the trend of age efect.