Correlates of highly caffeinated beverage consumption among Korean adolescents

Background : Adolescents in South Korea may be particularly vulnerable to highly caffeinated beverage consumption, which may lead to psychological and physiological problems. However, correlates of highly caffeinated beverage consumption have not been rigorously examined for Korean adolescents. This study investigated socio-demographic and behavioral factors related to highly caffeinated beverage consumption, using a nationally representative sample of Korean adolescents. Methods : We used 267,907 middle and high school students from the Korea Youth Risk Behavior Web-based Survey collected from year 2014 to 2017 for the present cross-sectional study. We employed multiple logistic regression to examine the associations between highly caffeinated beverage consumption and socio-demographic and behavioral factors. Results : Highly caffeinated beverage consumption was higher in 2017 than 2014 (23.9% vs. 12.0%) among Korean adolescents. Drinkers of highly caffeinated beverages were more likely to be boys, to be overweight or obese, to overestimate their body weight, to drink soda > 1 time per week, and to drink sweetened beverages > 1 time per week ( p s < .05). Additionally, drinkers of highly caffeinated beverages were more likely to feel stressed, to feel sadness and depression, to seriously consider suicide, and to attempt suicide during last 12 months ( p s < .05). Among male participants, drinkers of highly caffeinated beverages were less likely to meet the recommended 8 to 10 hours of sleep per night ( p < .05). Conclusions : Highly caffeinated beverage consumption is a potential health concern, as drinkers of highly caffeinated beverages were more likely to report health and behavioral problems. New strategies are needed to effectively curb highly caffeinated beverage consumption in Korea.

of caffeine per day from all food sources [12], but the study used data before HCBs became popular in Korea.
Adolescents may be vulnerable to HCB abuse, as HCBs are relatively inexpensive, readily accessible, and have a sweet avor that is more appealing to younger consumers than coffee [1,3,13]. Adolescents who consume HCBs may be at higher risk of caffeine intoxication than adults [13,14]. Korean adolescents may be particularly vulnerable to HCB abuse and caffeine intoxication, as many rely on HCBs to stay awake or focus on studying [15,16]. As such, the Korean government attempted to limit adolescent's caffeine intake by (i) requiring warning labels, (ii) restricting advertisement, and (iii) banning the sale of HCBs on school grounds [9,10,[17][18][19]. However, the effectiveness of those policies remains unclear, as no studies to date have examined changes in the multi-year prevalence of HCB consumption. Moreover, only a few studies that reported the prevalence of highly caffeinated beverage consumption among Korean adolescents utilized representative samples [5,6,12], which limits the generalizability of most studies.
The correlates of highly caffeinated beverage consumption are not fully known for the Korean adolescent population. Previous studies have focused on sleep, stress, and psychological correlates of highly caffeinated beverage consumption among Korean adolescents [5,6]. Relatively less attention was given to factors that may in uence individuals' decision to use HCB, such as nutrition [20,21], physical activity level [22], and body mass index (BMI) [23]. Few of the studies that considered such factors were not generalizable to the entire Korean adolescent population [21,24]. For example, BMI, on one hand, may be positively associated with HCB consumption, as HCB often contain added sugar linked to adiposity and weight gain [25]. On the other hand, BMI may be negatively related to HCB consumption, as habitual use of caffeine can lead to weight loss [26].
More research on HCB consumption among Korean adolescents, particularly with regards to the multiyear prevalence and correlates, is needed to understand the extent of caffeine abuse. The purpose of this study was to investigate correlates of HCB consumption using a nationally representative sample of Korean adolescents.

Survey and study sample
The Korea Centers for Disease Control and Prevention has conducted the Korea Youth Risk Behavior Webbased Survey (KYRBS) every year since 2005. We have merged 4 data sets of KYRBS collected from year 2014 to 2017 because HCB was added to the questionnaire in 2014 [27]. All the responses to the survey are self-reported. For the rst stage of the sampling, middle and high schools were chosen as Primary Sampling Units (PSU). For the second stage, one classroom of each grade within a PSU was selected with the systematic sampling method. The selection probability was 14% of all middle and high schools and 2% of all middle and high school students in Korea [28]. Online consent was obtained from all participants. The current study was approved by the Touro University Institutional Review Board (IRB No. PH0219

Measures
Demographic and Socio-economic Characteristics School type of the participants was categorized into middle and high school. Academic achievement was asked of students to which ve responses were given, and we condensed the ve responses into three options: (1) low (i.e., low or middle low), (2) middle, and (3) high (i.e., high or middle-high) [29]. The household income was coded into ve levels, and we regrouped ve levels into three as: (1) low (i.e., very low or low), (2) middle, and (3) high (i.e., very high or high) [29]. Weekly expenditure of the participants was originally categorized into sixteen groups from 0 -< 10,000 South Korea Won (SKW) (equivalent to 0 -< USD 8.92) to ≥ 150,000 SKW (USD 133.80) and recoded into three groups: (1) < 50,000 South Korea Won (SKW) (equivalent to < USD 44.66), (2) 50,000-99,999 SKW (USD 44.66-89.32), and (3) ≥ 100,000 SKW (≥ USD 89.33). Paternal and maternal education was grouped into three levels: (1) ≤ middle school, (2) high school, and (3) ≥ college. Living situation of the participants was categorized into four groups: (1) living with family, (2) living with relatives, (3) living alone or with friends, and (4) living in home for orphans. For work experience, we employed a dichotomous variable categorizing it as either "Yes" or "No".
Weekly physical activity for ≥ 60 minutes was coded into eight levels from none to 7 days for the last 7 days, and we reclassi ed them into three levels: (1) none, (2) 1-2 days, and (3) ≥ 3 days based on prior research [32]. Weekly vigorous exercise for ≥ 20 minutes was categorized into six levels from none to ≥ 5 days for the last 7 days and we regrouped them into three levels: (1) none, (2) 1-2 days, and (3) ≥ 3 days.
Tobacco use was dichotomized as "Never" vs. "Ever" during the participants' lifetime. Sleeping time was assessed by asking the participants "During weekdays, what time do you usually go to bed and wake up for the last 7 days?" Sleep duration was calculated from time to go to bed from time to wake up. Based on sleep duration recommendations (8-10 hours per night) for adolescents from the National Sleep Foundation [33], meeting sleep duration recommendations was classi ed as a dichotomous variable with either "Not met" or "Met". Feeling relief of fatigue after sleep was categorized into ve groups from not enough at all to very enough and we regrouped them into three levels: (1) not enough/not enough at all, (2) fair, (3) very enough/enough. Feeling stressed was classi ed into ve levels from never to very much and we reclassi ed them into three levels: (1) never/rarely, (2) a little, and (3) much/ very much. The following mental health-related variables were dichotomized as either "No" or "Yes": (1) feeling sadness and depression during last 12 months, (2) seriously considered suicide during last 12 months, and (3) suicide attempt during last 12 months.

Dietary Behaviors
For alcohol use during the participants' lifetime, a dichotomous variable was created, comparing users with nonusers. Weekly frequency of breakfast consumption was assigned into eight categories from 0 day to 7 days for the last 7 days and we regrouped them into three levels (1) 0 days, (2) 1-2 days, and (3) ≥ 3 days. Weekly frequency of soda and sweetened beverage (except soda and highly caffeinated beverages) consumption was grouped into seven levels from never to ≥ 3 times per day for the last 7 days and we regrouped them into three levels: (1) 0 times, (2) 1-2 times, and (3) ≥ 3 times per week.

Statistical Analysis
We employed the Cramer's V test to nd statistical differences in demographic, socio-economic, and health-related characteristics between the two groups (drinkers vs. nondrinkers of HCBs) [34]. We We used multiple logistic regression to identify the correlates of HCB consumption with adjustments for grade, family level household income [20], weekly expenditure [35], and academic achievement [22]. Prevalence of HCB consumption from 2014 to 2017 was calculated with 95% con dence intervals and was visualized by survey year and sex in graphical plots. We employed STATA version 14 (STATA Press, College Station, TX) to test statistical analyses with the signi cance level set at 0.05 (two-sided). We have taken into consideration sample weights from the complex survey design with the use of Svy commands in STATA.

Demographic and Health-related Characteristics of Boys
The nal sample consisted of 267,907 middle and high school students in Korea with a mean age of 15.0 (SD = 1.7) and 51.2% being boys. Among boys, 17.2% of them were drinkers of HCBs (Table 1). Boys consisted of 46.6% middle school students. The majority of boys (95.3%) lived with their family members. Only 24.1% of boys met 8-10 hours of recommended sleep duration. For the BMI weight status, the proportion of normal weight was highest (72.7%) in boys, followed by obese (10.2%), overweight (8.7%), and underweight (8.4%). Among boys, the degree of association between HCB consumption and all demographic and most health-related characteristics was negligible. However, HCB consumption was weakly associated with soda consumption (Cramer's V = 0.190) and sweetened beverage consumption (Cramer's V = 0.155). Demographic And Health-related Characteristics Of Girls As shown in Table 2, 13.1% of girls were drinkers of HCBs. Among girls, there were 46.6% middle school students. The majority of girls (87.3%) did not meet the recommended sleep duration (8-10 hours per night). About 80% of girls were normal weight (78.8%), followed by overweight (8.5%), obese (6.4%), and underweight (6.3%). Among girls, the degree of association between HCB consumption and all demographic and health-related variables was negligible, except soda consumption (Cramer's V = 0.148, weak association) and sweetened beverage consumption (Cramer's V = 0.144, weak association).

Correlates Of Highly Caffeinated Beverage Consumption
Among all drinkers, higher odds of HCB consumption were associated with (1) being boys, (2) living with relatives, (3) having work experience in the last 12 months, (4) being overweight or obese, (5) overestimation of their body weight, (6) alcohol drinking, and (7) tobacco use (ps < .05). Additionally, drinkers were more likely (1) to have ≥ 60 minutes of physical activity 1-2 days per week, (2) to consume breakfast 1-2 days per week, (3) to consume soda ≥ 1 time per week, and (4) to consume sweetened beverages ≥ 1 time per week (ps < .05). Drinkers were more likely (1) to feel less than enough relief of fatigue after sleep, (2) to feel stressed a little or more, (3) to feel sadness and depression during last 12 months, (4) to seriously consider suicide during last 12 months, and (5) to attempt suicide during last 12 months (ps < .05). Signi cantly lower odds of HCB consumption were associated with high school educated mothers and underweight participants (ps < .05).
Among male drinkers, signi cantly higher odds of HCB consumption were associated with (1) living with relatives and (2) having work experience in the last 12 months (ps < .05). Boys drinking HCBs were more likely (1) to be obese, (2) to drink alcohol, and (3) to use tobacco (ps < .05). In addition, they were more likely (1) to have ≥ 60 minutes of physical activity 1-2 days per week, (2) to have ≥ 20 minutes of vigorous exercise ≥ 3 days per week, (3) to consume breakfast 1-2 days per week, (4) to consume soda ≥ 1 time per week, and (5) to consume sweetened beverage ≥ 1 time per week (ps < .05). Furthermore, they were more likely (1) not to meet recommended sleep duration (8-10 hours per night), (2) to feel not enough relief of fatigue after sleep, (3) to feel stressed a little or more, (4) to feel sadness and depression during last 12 months, (5) to seriously consider suicide during last 12 months, and (6) to attempt suicide during the last 12 months (ps < .05). Signi cantly lower odds of HCB consumption were associated with underweight participants and body weight underestimation (ps < .05).
Among female drinkers, signi cantly higher odds of HCB consumption were associated with (1) obesity, In addition, signi canlty lower odds of HCB consumption were associated with high school educated mothers (ps < .05).
[Insert Table 3 about here]

Discussion
This study examined the prevalence and recent trends of HCB consumption among Korean adolescents.
This study also investigated socio-demographic and behavioral factors related to HCB consumption in a nationally representative sample of Korean adolescents. To the best of our knowledge, this is the rst study to comprehensively examine HCB consumption using multiple years of KYRBS data.
In this study, boys were more likely to consume HCBs, consistent with ndings in previous studies [6,7].
However, correlates of HCB consumption were generally similar for boys and girls. For both sexes, adolescents who consumed HCB were more likely to be obese, to drink soda at least once a week, and to drink sweetened beverage at least once a week. This not only con rms previous ndings that the sugar content and the subsequent sweet avor are reasons why adolescent consume HCBs [36,37], but it also suggests that for adolescents consuming HCBs, the sugar content and its subsequent impact on weight may mask any weight-loss effect expected from the caffeine content [23,26]. Hence, HCB abuse should be considered a concern for obesity as well as caffeine abuse.
Both boys and girls who consumed HCBs were more likely to engage in risky behaviors (alcohol and tobacco use) than nondrinkers, although the effect sizes were negligible. This is generally consistent with previous ndings that adolescents with risky behaviors are more likely to consume HCBs [5,6,20,38].
Similarly, adolescents who consumed HCBs were more likely to feel stressed, sad and depressed, feel unrested after sleep, and have suicidal thoughts and suicide attempts. Adolescents who feel stressed and unrested may choose to consume HCB for temporary relief, though it may lead to the development of long-term sleep problems [5,6,13,24].
Some sex differences were observed in the present study. Our regression analyses showed that among boys, HCB drinkers were more likely to have ≥ 20 minutes of vigorous exercise ≥ 3 days per week than nondrinkers. This phenomenon can be explained by how boys are more likely than girls to perceive higher energy and athletic performance after acute administration of caffeine [39]. Because boys perceive higher energy and athletic performance from caffeine [39], they may also expect HCBs to improve their workout, which leads to HCB consumption when engaged in vigorous-intensity exercise. Such association was not observed in girls, as girls were less likely to associate caffeine with athletic performance [39].
Another characteristic unique for boys was how HCB consumption was higher for boys who ate breakfast for low number of days. This may be attributed to how adolescents sometimes consume HCBs as snack supplements when skipping breakfast [40]. However, it is unclear why this phenomenon was observed in boys only. It may be that, in girls, the phenomenon was masked by other characteristics unique to girls, such as maternal education. Our results indicate that girls with mothers having high school education were less likely to consume HCBs than those with mothers having middle school education or less. It may be that mothers who received high school education were aware of potential harms of HCBs, and at the same time, managed intimate mother-daughter relationship. Previous studies have suggested that mother-daughter relationship may be inversely associated with body image and health behavior [41,42], as mothers with intimate relationship with their daughters are able to function as effective advisors or role models. Similarly, in our study, mothers with intimate relationship with their daughters may have been more effective at discouraging their daughter's HCB consumption, even when the daughters were tempted to consume HCBs as snack supplements after skipping breakfast. Future strategies to curb HCB use among Korean adolescents should consider such possible sex differences.
The prevalence of HCB consumption was 26.7% for boys and 20.9% for girls in 2017. This is higher than 11.9-14.1% reported in earlier studies [6,7]. One of our unique ndings was the change in prevalence by year, most notable between years 2016 and 2017. We observed a sharp (13.8-23.9%) increase in HCB consumption between years 2016 and 2017, despite Korean government's policies intended to reduce adolescent's access to HCBs [17,18]. This suggests that Korea's HCB policies may be ineffective, as asserted by a number of critics. For example, one of the existing policies that restrict television advertisements has been criticized for not understanding adolescent's media consumption, as it only prohibits advertisements between 5-7 pm in which many Korean adolescents are attending Hagwon learning academies and are away from television [43,44]. A different policy that prohibits the sale of HCBs in school grounds has been criticized for its loopholes, as numerous venues outside school grounds remain free to sell HCBs [44,45]. Although longitudinal analysis comparing HCB use before and after policy implementation is needed to evaluate the extent of effectiveness of each speci c policy, this study shows that overall, existing policies intended to curb adolescent's HCB use are not as effective as initially envisioned. New strategies are needed to address HCB use, before HCB becomes even more popular among Korean adolescents.

Limitations
This study is not without limitations. First, although efforts were made to include and control for known confounders, there are unmeasured confounders such as the type of HCB, nutrient content of HCB, and access to HCB that could potentially in uence the association between independent variables and HCB consumption. Second, this study relied on self-reported data, and given how adolescents have shown to under-report their body weight and over-report their height [46], BMI percentiles computed in this study may have differed slightly from those using direct anthropometric measurements. Third, this study used BMI percentiles to assign weight categories. While BMI percentile is commonly used to predict adiposity in children and adolescents, they cannot account for individuals' body composition (e.g., amount of fat mass vs. fat-free mass) or fat distribution (e.g., abdominal vs. non-abdominal fat). Lastly, though this study used multiple years of data to best identify factors associated with HCB consumption, this study was not a longitudinal analysis. Numbers reported re ect correlation, but not necessarily causation.

Conclusions
Despite the limitations, this study provides important information on HCB consumption and some of its correlates among Korean adolescents. This study found various factors associated HCB consumption, such as obesity, alcohol/tobacco use, stress, and sleep problems. This study found sex differences in HCB consumption, including differences by frequency of vigorous-intensity exercise, frequency of breakfast consumption, and maternal education. This study also found a signi cant growth in HCB consumption between 2016 and 2017, despite Korean government's efforts to curb adolescent's caffeine abuse. New strategies are needed to effectively curb HCB consumption among Korean adolescents.
Future studies may utilize longitudinal and qualitative methods to further investigate sex differences in HCB consumption.

Declarations
Ethics approval and consent to participate Ethics approval was given by the Institutional Review Board of Touro University, CA, USA (IRB No. PH0219). All participants provided online consent to participate in the survey. This study also obtained consent from their school principals and homeroom teachers.

Consent for publication
All participants have provided online consent for their personal data to be utilized.

Availability of data and material
Data can be made available through the authors.
Competing interests