Relations between sweetened beverage consumption and individual, interpersonal, and environmental factors: a 6-year longitudinal study in German children and adolescents

The following analyses are based on the “German Health Interview and Examination Survey for Children and Adolescents” (“KiGGS”). This longitudinal study is the first official database to generate comprehensive health data for cohorts between the ages of 0 and 17 in Germany. It was commissioned as the largest long-term study on children and adolescents in Germany by the Federal Ministry of Health and is managed by the Robert Koch Institute, Berlin (Hölling et al. 2012). To guarantee the study’s representativeness, a stratified multi-stage probability sample was selected from the official registers at local residents’ registration offices. The fieldwork for the baseline survey was carried out in 167 cities and municipalities by trained survey teams, each consisting of a physician, a key interviewer, an examiner, and a medical laboratory technician. The data quality and the nationwide representativeness of the KiGGS study were subjected to intensive internal and external quality controls (see Kamtsiuris et al. 2007 for details). For the 0–10 age group, parents reported about their children (proxy interviews), while for the 11–17 age group, both parents and adolescents were interviewed. The KiGGS baseline survey was conducted between May 2003 and May 2006. The baseline comprised a total sample of 17,641 children and adolescents; 68% of all baseline participants took part in the follow-up survey, KiGGS Wave 1. Wave 1 was carried out between June 2009 and June 2012 as standardized telephone interviews.

For the following analyses, we looked at interpersonal-environmental factors within a total longitudinal sample of all n_w = 11,691 children and adolescents aged between 0 and 17 years old (n_w = weighted; unweighted n_uw = 11,677). Of the 11,992 participants that took part in both the baseline study and Wave 1315 had to be excluded due to missing data for soft drink consumption. In addition to this, we also included individual lifestyle in addition to interpersonal-environmental correlates for a weighted sample of n_w = 5197 participants aged between 11 and 17 years old (n_uw = 4114), as individual lifestyle factors such tobacco or alcohol consumption were only available from 11 years on.

Main outcome: In KiGGS wave 1, sweetened beverage consumption was measured with two items assessing the current frequency of children’s and adolescents’ consumption and the number of glasses consumed per day or per week (Mensink et al. 2018). The question on frequency referred to “the last few weeks” and read “How often do you/does your child drink sweetened beverages like cola, lemonade, ice tea, or malt beer? Sweetened beverages refer to sugar-sweetened and artificially sweetened beverages”. In this article, we have explicitly used the phrase “sweetened beverages” when referring to our own data (in contrast to “SSBs” that were explicitly examined in several studies referred to in Introduction and Discussion). These questions and our analyses explicitly excluded the consumption of water, milk, fruit, and vegetable juices. The available ordinal categories were “daily”, “not daily but at least once a week”, “less often than once a week”, and “never” (Mensink et al. 2018). Depending on the answer to the frequency question, participants were either asked for the average number of glasses of sweetened beverages that they drank per day or per week. It was explained to participants that “a glass” represents 200 ml (6.76 oz); answers such as “half a glass per day” or “half a glass per week” (i.e. 100 ml/3.38 oz) were also possible. A dichotomous variable was created from these self-reports, indicating if participants drank sweetened beverages at least once a week (dummy value 1) or less often than once a week (dummy value 0). In addition, a continuous variable of consumed glasses per day for every child or adolescent was computed. The continuous variable was set to zero if participants answered “never” or “less often than once a week” to the question on consumption frequency.

All potential correlates were recorded as part of the KiGGS baseline survey; these included age and gender as individual variables. Health-related lifestyle behaviours were also assessed. Namely, self-reported smoking (current consumption: yes vs. no), alcohol consumption (has participant ever consumed alcohol: yes vs. no), media consumption (spending time on a computer, a mobile phone, a games console or a television; categorized into terciles of usage time), consumption of fruit, cooked vegetables, salad, and other raw vegetables (assessed using a food frequency questionnaire, which also referred to ‘‘the last few weeks''; the total of 10 frequency categories were grouped together into three categories [< 3 times a week; 3–4 times a week; > 4 times a week]) and physical activity (How often is participant physically active in his or her free time, getting sweaty or out of breath; grouped into three categories [never; 1–2 times a month up to 3–5 times a week; and daily]).

In addition, interpersonal information (parents’ socio-economic status, immigration background, number of siblings, age of mother, and parental tobacco consumption) and environmental information (size of town of residence, region of residence) were assessed. The socioeconomic status (SES) of the participant’s parents was defined as “low”, “medium”, or “high” based on an established and validated index (Winkler and Stolzenberg 1999). This multidimensional aggregated index comprises the dimensions of parents’ education (school education and professional qualifications), income (net household income of all household members), and occupational status. In accordance with national standards, immigration background was deduced from the participant’s country of birth and nationality and country of birth of the participant’s parents (Schenk et al. 2008). It was also assessed whether children or adolescents lived together with siblings in the same household. Participant’s parents were asked if they were currently smokers. The size of the participant’s town of residence was determined using official data on political-administrative municipality sizes. This distinguishes between rural areas, small towns, medium-sized towns, and metropolitan areas (Robert-Koch-Institut 2008). Participants region of residence was assigned to the western (former West Germany—Federal Republic of Germany) or the eastern federal states (former East Germany—German Democratic Republic, including Berlin).

We examined longitudinal associations using the individual and interpersonal-environmental variables from the baseline study (KiGGS Baseline) and the measures of sweetened beverage consumption from the follow-up study (KiGGS Wave 1). Variables were analysed using descriptive, bivariate, and multivariate methods. For the descriptive and bivariate analyses, data were weighted to adjust for dropout and make the longitudinal sample representative for the population in terms of gender, age, region, and migration background (weight made available by the Robert Koch Institute). Because the mother’s age is more strongly associated with the consumption of sweetened beverages than the father’s age, only mother’s age was included as a predictor in the analyses to avoid collinearity.

Outcome variables of interest were both dichotomous (sweetened beverage consumption ≥ once a week or not) and continuous (number of consumed glasses of sweetened beverages per day). These different variable types were analysed in separate models, to ensure the robustness of the analyses. Pearson Chi-Squared tests were used for the bivariate analyses with the dichotomous outcome variable; one-way analysis of variance were used with the continuous outcome variable.

Soft drink consumption increased with age; particularly between the ages of 3–7 and 7–11, reaching a plateau later on. Older and male adolescents consumed a particularly high volume of sweetened beverages with particularly high frequency. In terms of individual lifestyle, above average consumption of sweetened beverages in KiGGS Wave 1 was accompanied by other behaviours that are detrimental to health such as high consumption of tobacco, alcohol, and media, as well as insufficient consumption of fruit and vegetables in the baseline study. Additionally, adolescents who reported a high level of physical activity drank significantly higher volumes of sweetened beverages more frequently (Table 1).

The descriptive analyses identify clearly defined demographic subgroups with a particularly high level of consumption, for example, adolescents with a low social status drank high volumes of sweetened beverages particularly often. Other interpersonal and environmental factors that also showed bivariate associations with an increased consumption of sweetened beverages were the presence of siblings in the household, the mother’s age being comparatively high, tobacco consumption of one or both parents, and living in a more rural area (Table 2).

The bivariate correlations described above were analysed from multiple perspectives using regression models. These analyses showed that most of the independent variables previously identified as significant correlates of consumption, remained significant after adjusting for potential confounders. For example, if the consumption of sweetened beverages is operationalized as a dichotomous variable then age, gender, tobacco and media consumption, and social status remain significant correlates in an adjusted logistic regression model (Table 3).

A similar picture emerges when general linear modelling is used: in addition to the correlates previously mentioned as significant, low to moderate consumption of vegetables, high degree of physical activity, and parental tobacco consumption are also significantly associated with sweetened beverage consumption 6 years later (see supplementary materials for Table 4 and for additional analyses stratified by age below and above 11 years).

While a number of our results are in line with previous findings from other Western countries, they also deliver important new insights and observations. As previously reported for other countries, in Germany we have also found that sweetened beverage consumption increases with age (Denney-Wilson et al. 2009; Hennessy et al. 2015; Vereecken et al. 2005). However, our data show a plateau in late adolescence, while other sources tend to report a continuous increase until post-adolescence (Bere et al. 2008; Pettigrew et al. 2015; Scully et al. 2017; Van Ansem et al. 2014). A deciding factor in the higher level of consumption observed in adolescents compared with children is their increasing autonomy and greater disposable income (Bere et al. 2008; Pettigrew et al. 2015; Scully et al. 2017; Van Ansem et al. 2014). Comparable to findings from other countries, boys in our sample consume sweetened beverage more frequently and in larger amounts than girls (Bere et al. 2008; Bjelland et al. 2011; Denney-Wilson et al. 2009; Park et al. 2012, 2013; Scully et al. 2017; Sdrali et al. 2010; Vereecken et al. 2005). Although boys generally require a higher energy intake than girls, this increased energy demand is nevertheless best fulfilled with healthy food. Other reasons discussed for higher levels of SSB consumption in boys include a lower degree of health consciousness, less concern with own appearance, and gender specific socialisation (Vereecken et al. 2005).

This study enters completely new territory by conducting a comprehensive analysis addressing the “big four” (Khaw 2008) classic characteristics of health-relevant lifestyle: smoking, alcohol, nutrition, and physical activity. Previous studies have only addressed individual health-relevant lifestyle factors related to the consumption of SSBs. A few studies have reported a positive link between SSB consumption and tobacco and alcohol consumption (Park et al. 2012, 2013). Other studies have found a correlation between SSB and media consumption (Mazarello Paes et al. 2015; Park et al. 2012; Scully et al. 2017). Individual studies have also observed this behaviour among primary school children (Lopez et al. 2012; Olafsdottir et al. 2014) and even in pre-school aged children (Garnett et al. 2013; Olafsdottir et al. 2014). One particularly interesting study differentiated television consumption by weekday and found a particularly strong correlation with SSB consumption on weekends (Pettigrew et al. 2015). This could be due to influence of food advertising on adolescent viewers as well as longer TV exposition times on weekends, which lead to interruption of physiological food regulation, e.g. satiety cues (Scully et al. 2017; Verzeletti et al. 2009). For example, snacking while using a computer is also accompanied by a significant increase in SSB consumption (Verzeletti et al. 2009). Findings on the relationship between SSB consumption and eating fruits and vegetables are inconsistent (Park et al. 2012; Ranjit et al. 2010; Scully et al. 2017). In contrast to our German study population, data from the USA and Greece point to a negative correlation between SSB consumption and physical activity (Park et al. 2012; Sdrali et al. 2010).

Eating habits such as SSB consumption do not evolve in isolation, but within a micro-environment, which is primarily influenced by interpersonal relationships with family members, and within a macro-environment (Schneider et al. 2017; Van Ansem et al. 2014). We have already explored in detail how, under certain circumstances, these micro- and macro-environments can result in what is known as an obesogenic environment (Schneider et al. 2017). The home environment represents an important part of a person’s micro-environment. Parents are their children’s nutritional gatekeepers, their knowledge and choices determine what their children eat (e.g. Dallacker et al. 2016; 2018). Children have little control over purchases and generally eat the food that is available at home (Lopez et al. 2012; Elfhag et al. 2008).

One particularly striking finding from our data is that the consumption of sweetened beverages is about twice as high for children from low-status families compared to children from higher status families. This classic social gradient—usually measured using the socio-economic status of the parents—has been well documented (Bere et al. 2008; Garnett et al. 2013; Hebden et al. 2013; Lopez et al. 2012; Lundeen et al. 2018; Mazarello Paes et al. 2015; Park et al. 2014; Scully et al. 2017; Sdrali et al. 2010; Totland et al. 2013; Van Ansem et al. 2014). Parents could use SSBs as coping mechanisms for emotional discomfort in their children. Garnett et al. (2013) suggest another parental motive for this finding, namely that the introduction of SSBs may be a mechanism for coping with food insecurity and that SSBs are used as a means to quell behavioural agitation. Although the argument of limited economic resources (Vereecken et al. 2005) may play a subjective role for parents when shopping, de facto, tap water, bottled carbonated water, non-carbonated mineral water, or other alternatives such as tea and flavoured calorie-free water are generally cheaper than SSBs.

Interestingly, it seems that mothers play a more important role in SSB consumption than fathers—a stronger correlation was found between SSB consumption and the mother’s behaviour than with that of the father (Bjelland et al. 2011); this finding holds equally true for both, girls and boys (Elfhag et al. 2008). Furthermore, estimations regarding the availability of SSBs within the family home tend to correspond more closely between mother and child than between father and child, indicating that mothers have better knowledge than fathers of the food selection actually available in the home (Totland et al. 2013). Other publications also emphasize the role of the mother as a role model (Bjelland et al. 2011; Denney-Wilson et al. 2009) for their children’s dietary behaviour. A study on regulatory food parenting has shown that mothers play a greater role in monitoring, regulating, and controlling their children’s eating habits than fathers (Melbye et al. 2016). We did not replicate previous findings in which a higher level of SSB consumption was observed among the children of younger mothers (Garnett et al. 2013; Mazarello Paes et al. 2015); on the contrary, in our data set, having an older mother was associated with higher consumption. As in previous studies, our data are also inconclusive with regard to the significance of siblings (Mazarello Paes et al. 2015; Park et al. 2014) and the region of residence (Garnett et al. 2013).

When interpreting the KiGGS study, it is important to consider the definition of sweetened beverages, a potential social desirability bias, and residual confounding factors.

There is currently no internationally agreed standard definition for SSBs (Mazarello Paes et al. 2015). As in most studies, the method we use to operationalize our data also meant that other drinks such as sweetened milk, sweetened coffee-based drinks, and sweetened vegetable juice were excluded. Likewise, as done in many comparable studies, we also excluded various types of sweetened fruit juice and sports drinks. Flavoured sports beverages usually contain large amounts of added sugar (Ranjit et al. 2010); however, studies have shown that adolescent consumers and their parents tend to subjectively view the health effects of this special type of sweetened beverage differently than other soft drinks, resulting in positive associations between consuming sports beverages and several healthy lifestyle practices (Hennessy et al. 2015; Ranjit et al. 2010).

Unfortunately, the food frequency questionnaire used here did not allow for the extraction of specific data regarding the consumption of calorie-reduced drinks (e.g. low calorie, practically calorie-free, “diet” soft drinks). However, given that in Germany more than 80% of all soft drinks consumed by adolescents are classic SSBs (Mensink et al. 2018), this limitation likely has little impact on our findings. Furthermore, several of the health risks mentioned in the Introduction, such as development of metabolic syndrome, type 2 diabetes, or the displacement of milk and calcium in the diet are still equally applicable when sugar-free versions of soft drinks are consumed. Lastly, prospective cohort studies have demonstrated a positive correlation between the consumption of artificial sweeteners and weight gain. It is proposed that the regular exposure to sweet flavours, regardless of their origin, leads to a persisting preference for sweet things and thus a preference for high-calorie food. This is accompanied by the fact that artificial sweeteners (e.g. aspartame) also function as appetite stimulants (Elfhag et al. 2007). It should be mentioned that the role of artificially sweetened beverages for the development of obesity is controversially discussed, particularly in industry-funded studies (e.g. Peters et al. 2014).

With a total of 16 independent variables, we were able to consider more variables than most other comparable studies. Nevertheless, a risk remains that the regressions presented here may be affected by residual confounding factors from unmeasured but relevant influences. We also tested for multicollinearity: the highest correlation between the independent variables (included in Table 3 and in Table 4 (see supplementary materials), models 2–4) was r = .56 and thus well below common thresholds for multicollinearity.

Finally, as with any epidemiological questionnaire-based study, typical social desirability bias cannot be completely eliminated in the reporting of eating habits. The main strengths of this study are its large, representative sample size, the extensive quality assurance measures implemented in the field work, its longitudinal design, and the comprehensive consideration of health-relevant lifestyle practices.

Health and health inequalities start to develop very early in life due to the complex interactions of cumulative risks and resources at the individual, meso-, and macro-level (Seabrook and Avison 2012). Sweetened beverages are one piece of this puzzle. We show that various dimensions of an unhealthy lifestyle occur accumulatively. Our findings call for complex, multi-level interventions. Next to the role of individual risk behaviours such as smoking or low vegetable intake related to sweetened beverage consumption, our data also highlight the central role played by a person’s social context—their home environment in particular. Because dietary habits are formed in childhood, food skills and the learning of healthy choices are particularly important early in life. For example, one recent study reports that the participation of young people in the preparation of family meals was the top predictor of food skills later in life (Seabrook et al. 2019). Interventions need to aim not only at children and adolescents, but especially at their parents and the (interpersonal) environments inhabited by children and adolescents (Park et al. 2014). In addition to reducing the consumption of unhealthy SSBs, interventions also need to address other unhealthy lifestyle habits that go hand in hand with drinking SSBs, such as eating little fruits and vegetables, media use, and physical activity (Bogart et al. 2017).