Exposure to liquid sweetness in early childhood: artificially‐sweetened and sugar‐sweetened beverage consumption at 4–5 years and risk of overweight and obesity at 7–8 years

Summary Background A significant gap exists in longitudinal evidence on early exposure to artificially sweetened beverages (ASBs) and weight outcomes for paediatric populations. Objective The objective of this study is to examine the relationship between ASB/sugar‐sweetened beverage (SSB) consumption at 4–5 years and risk of overweight and obesity at 7–8 years. Methods Data from a nationally representative cohort (n = 2986) in Scotland were analysed using logistic regression to evaluate the association between exposure to ASBs/SSBs at 4–5 years and risk of overweight and obesity at 7–8 years. Results There were positive unadjusted associations between ASB consumption and risk of obesity, and following adjustment for confounders, ASB associations attenuated, and only the middle consumption category (1 to 6 times per week) remained significant (odds ratio 1.57, 95% confidence interval {CI} 1.05–2.36). For SSB consumption, there were no significant unadjusted associations, and following adjustment for confounders, only the middle consumption category was significant (odds ratio 1.65, 95% CI 1.12–2.44). There were no significant associations for risk of overweight. Conclusions Longitudinal analysis from 4–5 to 7–8 years demonstrated some evidence of associations between ASBs/SSB consumption and risk of obesity. However, non‐linear patterns and wide CIs suggest cautious interpretation and need for future studies with long‐term follow‐up.


Supporting Information
: Descriptive characteristics of final longitudinal sample (sweeps 1, 4, 5, 6 and 7) Table S8: Multivariate linear regression models for association between SSB/ASB consumption at 4-5 years and BMI at 7-8 years Table S9: Multivariate linear regression models for association between SSB/ASB consumption at 4-5 years and BMI at 7-8 years with children categorised as obese at baseline removed Data collection at each sweep was intended to be 6 weeks before the child's next birthday, and so the exposure was measured when the children were just under 5 years (i.e. 4-5 years), and the outcome when the children were just under 8 years (i.e. 7-8 years).

 Exposure Variables
For both exposures, respondents were asked to report frequency of consumption according to the following categories: 1. More than once a day 2. Once a day 3. 5 or 6 times a week 4. 2 to 4 times a week 5. Once a week 6. 1 to 3 times per month 7. Less often 8. Never For analytic purposes it was necessary to re-categorise the exposure variables 3 . Firstly, BMI at age 3-4 years and age 7-8 years was recoded because for the purposes of these analyses underweight was considered similar to healthy weight i.e. not overweight. In order to be able to consider overweight and obesity separately, the BMI outcome variable at age 7-8years was recoded into two binary variables, as shown in Table 3 below. A wide range of risk factors have been linked to obesity 4-6 . Although it was not possible to include all potential risk factors, a range of confounding variables were selected for inclusion based on some of the types variables which have been found to be significant in previous studies 3,7,8 .

 Sociodemographic variables
Equivalised income. Respondents were asked to report the "total income of your household from all sources before tax -including benefits, interest from savings and so on" ( 9 p.36). Respondents were provided with a list of 17 income bands, from "Less than £3,999 pa" to "£56,000 or more pa" ( 9 p.36). These income bands were then adjusted by the GUS project team using an equivalence scale, as show in Table S4 below, to provide "equivalised annual household income", and this was also split into quintiles ( 10 p.20). This adjustment takes into account that the standard of living that a family can achieve will differ depending on the size of the household and the ratio of adults and children 10 . Maternal educational level. At sweep 1, respondents were provided with a list of educational examinations/qualifications, and asked to identify those that applied to them. At subsequent sweeps respondents were also asked if they had achieved any new qualifications. This information was used by the project team to develop a derived variable "Highest educational level of respondent" 9 p.131. For the purposes of this study, educational level was based on a derived variable at sweep 5. Only cases where the respondent was the child's mother (including step mothers and adoptive mothers) were included. For the purposes of our study we wanted to be able to control for maternal educational level (and maternal BMI) which have been shown to be relevant determinants of child BMI 8 . Therefore we wished to include these variables as covariates. If we had included respondents who were not the child's mother this would have prevented being able to control specifically for maternal factors -the variable would instead have been 'respondent/carer educational level' (and 'respondent/carer BMI'), which would not have provided the required specificity. This variable is therefore understood as maternal educational level. The derived variable coded the qualifications that the respondent had achieved according to the Scottish Credit and Qualifications Framework as outlined in Table S5 below. These were recoded in order to simplify categories for analytic purposes.  9 p.23 and this was coded as child usually eats breakfast or does not usually eat breakfast.
Fruit and vegetable consumption: Respondents were asked: "How many different types of vegetable did ^childname eat yesterday? RANGE 0-10 and "How many different types of fruit did ^childname eat yesterday? RANGE 0 -10 9 p.24.
These data were combined to create a variable for the total number of different types of fruit and vegetables consumed in the previous day. The Scottish Government's Dietary Goals recommend that adults and children eat 5 portions of fruit and vegetables every day 11 . This variable was categorised into two groups i.e. those children who ate 5 or more different types of fruit and vegetables in the previous day, and those who ate less than 5 types of fruit and vegetables per day. It is recognised that the question asks about the number of different types of fruits and vegetables, not number of portions, and this is a limitation of the information derived from this variable.

Milk and water consumption:
Respondents were asked to report on the child's consumption of other beverages: "How often does ^childname drink milk, not including milkshakes or other flavoured milks? INTERVIEWER: include soya/goat's milk" 9 p. 26, and "How often does ^childname drink unflavoured water, for example, from the tap, a water cooler or a bottle of water?" 9 p.26.
Respondents were asked to choose from the eight categories of frequency as outlined above and these were recoded into the binary category of either daily or less than daily consumption.

Sweets and crisps consumption:
Respondents were asked: "How often does ^childname eat sweets or chocolates? INTERVIEWER: Include only whole packets of sweets or a chocolates/chocolate bar, not individual sweets" and "How often does ^childname eat crisps?" 9 p.25.
These variables were recoded into binary variables i.e. whether the child ate sweets/chocolates once a day or more, versus less than once a day. The same categorisation was used for the crisps variable. These variables were then combined into three categories: 0 = "Eats sweets OR crisps less than every day of the week" 1 = "Eats sweets OR crisps once a day or more" 2 = "Eats sweets AND crisps once a day or more" Processed meals consumption: Respondents were also asked to report on the child's consumption of different types of meals. Three of these questions were chosen for their relevance to overall diet quality: "Can you tell me on how many days in the last week ^childname has had each of the following things for ^his main meal? By 'the last week', I mean the last 7 days. (RANGE 0-7) … a ready meal? … a take-away meal, for example, from a fish and chip shop or an Indian or Chinese takeaway? … a fast-food meal, for example, from McDonald's?" 9 p.22.
The scores for each of these three variables were then summed to give the frequency of total consumption. A new variable was created according to the following categories: 0 = "Has not had processed meal in last 7 days" 1 = "Has had processed meal once in last 7 days" 2 = "Has had processed meal twice or more in last 7 days"  Activity variables Respondents were asked a range of questions regarding the child's physical activity, television and computer use.
Television viewing: Several questions asked respondents about the child's television viewing, including patterns for both weekday and weekends. The following question was chosen as a measure of TV viewing time: "How long would ^childname usually watch television for in total on an average weekday? INTERVIEWER PLEASE ENTER TIME IN HOURS (Range 0-24) 9 p. 84.
In line with previous analysis of GUS data, TV viewing was categorised into those that watched less than 3 hours of TV per day or more than 3 hours per day 12 . The weekday variable was chosen because the distributions for the weekday and weekend variables were similar, and therefore only one was required.
Physical activity: Respondents were asked a range of questions regarding the child's physical activity. Firstly respondents were asked to report on whether a child had engaged in a range of different activities in the previous week. Then respondents were asked to report the time spent on each activity: "Now looking at this card, in the last week, how much time did ^childname spend doing activity? This question was asked in relation to the time spent engaging in the following activities: riding a bicycle; throwing or kicking a ball; dancing or gymnastics; running and/or jumping; playing on a trampoline; swimming; playing at a soft play area or ball swamp; playing at a play/swing park; walking and doing something else active.
In order that a judgement could be made regarding a child's overall physical activity, these variables were combined. To generate a composite measure of physical activity, a decision was made to treat each category as the maximum number of minutes for that category, as shown in Table S6 below. If it was recorded that this item was not applicable (i.e. the child had not taken part in that activity in the past week) or there was no time, then it was recorded as zero minutes. The minutes for each activity were then summed to generate a total number of minutes of physical activity for the previous week. The UK Government recommends that children and young people are active for 60 minutes per day, on every day of the week, and that children under 5 are active for 180 minutes per day 13 . The guideline for under 5s includes activities "of any intensity" 13 whilst the guideline for children and young people are focused on "moderate to vigorous intensity physical activity" 14 . The activities measured in GUS arguably fit with moderate to vigorous intensities, and so the guidelines for children and young people were used to inform interpretation of the data. It was not possible to calculate a meaningful measure of daily activity. Therefore the total number of minutes per week was compared to what would be achieved if a child was active for 60 minutes on every day of the week i.e. 420 minutes per week. This generated a binary variable: 0 = "Does not meet weekly activity guidelines (i.e. 420 minutes per week)" 1 = "Meets weekly activity guidelines of 60 mins per day (i.e. 420 minutes per week)" It is acknowledged that this variable is a less than optimal measure of physical activity and it is likely that it does not represent actual activity levels. However it was reasoned that this was the best option for a proxy measure based on available data.

 BMI factors
Heights and weights were measured at sweep 4 (when the children were approximately 3 years), and the National (UK) BMI percentiles were calculated by the GUS project team 15 . National BMI percentile classification at age 3-4 years were used as a measure of baseline BMI.
Mothers' heights and weights were measured by trained interviewers at sweep 6 according to a protocol. Maternal BMI was then derived. It must be noted that estimated weights where the mother's weight was over 130kg were also included (At Sweep 6 this was 9 cases) (Growing Up in Scotland Main Carer Questionnaire Sweep 6).