Obesity
Original Article PEDIATRIC OBESITY
Relationship Between Raised BMI and Sugar Sweetened Beverage and High Fat Food Consumption Among Children Lynne Millar1, Bosco Rowland2, Melanie Nichols1, Boyd Swinburn1,3, Catherine Bennett4, Helen Skouteris2 and Steven Allender1
Objective: Longitudinal evidence of relationships between unhealthy diets and BMI in children is crucial for appropriately targeting obesity prevention activities. The objective was to determine the relationship between frequency of consumption of sugar sweetened beverages (SSBs) and high fat foods (HFFs) and body weight in Australian children aged from 4 to 10 years. Methods: Data from 4,164 children participating in four waves (wave 1, 2004; wave 2, 2006; wave 3, 2008; and wave 4, 2010) of the Longitudinal Study of Australian Children were analyzed. A multi-level growth model tested relationships between consumption of SSB and HFF and BMI z-scores. Results: BMI z-scores were associated with daily consumption of HFF, SSB and maternal BMI independent of BMI z-scores at wave 1 (baseline); with each additional occurrence of SSB and HFF consumption intake per day, BMI z-score increased by 0.015 U (P < 0.01) and 0.014 U (P < 0.001), respectively. With each additional maternal BMI unit, BMI z-score increased by 0.032 (P < 0.001). Conclusions: Higher BMI z-scores were strongly associated with the consumption of SSBs and HFFs. Future efforts to prevent obesity should consider urgent action to address the impact of the consumption of SSBs and HFFs in childhood. Obesity (2014) 22, E96-E103. doi:10.1002/oby.20665
Introduction The level of obesity for children in developed countries, such as Australia, has reached epidemic proportions, and is now a major preventative health priority for most countries. Low levels of physical activity, high levels of television and computer screen time and high consumption levels of energy dense and nutrient poor food have all been identified as significant contributors to this epidemic (1-3). In Australia, the reduction of sugar sweetened beverages (SSB) and high fat foods (HFF) has been the focus of many prevention measures (4-6) but as yet there is not clear evidence about the relationship between high consumption of SSB and HFF and childhood obesity. Studies in developed and developing countries have demonstrated that increases in SSB are associated with increased body mass index (BMI) (7-10). These studies have principally examined BMI at two time points where higher levels of SSB consumption at the earlier time point was found to be predictive of greater levels of BMI at a later time point. Systematic reviews of cross sectional and two-time point studies have identified associations between consumption of SSBs and increasing body weight in adults and children (11-13).
Critics of the focus on reducing SSB to prevent obesity argue that the existing studies are limited because they are often cross sectional or examine risk at baseline against outcomes at time two without measures across multiple time points (14,15). In this context, it remains clear that stronger evidence from studies that allows for temporal influences is required, especially if preventative measures targeting SSB are to be sustained and supported in the long-term by funders and policy makers (11,16). Longitudinal studies whereby data on food and drink consumption, and individual, family and environmental are collected over several time points (>2) is one way of developing such evidence. Research from Barclay and Brand-Miller (17) point to a possible paradox; they identify decreasing sugar consumption and SSB consumption in Australia alongside increasing prevalence of obesity in Australia. Many have argued that this indicates no association between sugar consumption and obesity patterns at population level. Others have found that industry funding of nutrition-related scientific articles may bias conclusions in favor of sponsors’ products (12,18) and so potential conflicts of interest exist in the three key reviews demonstrating no association between high levels of SSB
1
WHO Collaborating Centre for Obesity Prevention, Deakin University, Geelong, Australia. Correspondence: Lynne Millar ([email protected]) School of Psychology, Deakin University, Melbourne, Australia 3 Population Nutrition and Global Health, University of Auckland, Auckland, New Zealand 4 Deakin Epidemiology, Deakin University, Melbourne, Australia
2
Disclosure: the authors have no competing interests. Author contributions: LM developed the major concepts, LM & BR conceived the analysis plan and analyzed the data. All authors were involved in writing the article and had final approval of the submitted and published versions. Received: 11 March 2013; Accepted: 21 November 2013; Published online 28 November 2013. doi:10.1002/oby.20665
E96
Obesity | VOLUME 22 | NUMBER 5 | MAY 2014
www.obesityjournal.org
Original Article
Obesity
PEDIATRIC OBESITY
consumption and high body weight as they were industry funded (15,16,19). Moreover, all data used in the analysis to support the proposed Australian paradox were cross sectional. Longitudinal designed studies can assist to untangle and better understand this seemingly paradoxical relationship. High fat food (HFF) consumption has also been implicated in the obesity epidemic among children and adolescents. Australian studies examining HFF consumption and obesity among children have shown mixed results; for example, cross-sectional studies have found positive relationships between more frequent fast food consumption at home and a person being overweight (20). Conversely, lower fat intake in girls and boys and lower consumption of energy dense snacks in boys has been shown to be associated with a greater risk of overweight/obesity in a sample of Australian adolescents (21). Caution needs to be applied when interpreting cross-sectional studies. For example, without temporal data they cannot establish causality nor consider the possibility of reverse causality (that is being overweight causing a reduction in HFF consumption). Overall, studies examining the relationship between children’s body weight and both SSB and HFF have been inconclusive. These studies have found weak relationships between consumption of HFF and increased body weight but more robust findings between SSB consumption and increased body weight. This has been found in one study within Australia (22,23) and other high income countries (8,24,25). The mixed findings may be due to the differences in foods included in studies and the limitations of dietary assessment methods. Measurement and interpretation of SSB consumption is less complex than measurement of HFFs as the kilojoule content of SSBs is quite uniform and the beverages are usually presented in standard sizes so measurement can be more precise. Conversely HFF includes a range of diverse food items, and studies have variously defined HFFs according to frequency of consumption, calculated energy density or number of convenience foods (not fast foods) (24,26,27). Stronger evidence is needed to adequately support government decisions on investing in strategies that reduce HFF and SSB consumption. Longitudinal data with more than two time points, using national representative samples, and which uses strong and reliable measures will be vital in providing this information. Such data provide the opportunity for analysis that allows for trajectory modeling, and the assessment of influences at a variety of levels. Such an opportunity is provided by the Longitudinal Study of Australian Children (LSAC); a large, high quality, observational longitudinal cohort study over multiple time-points that includes robust data collected every 2 years on a representative cohort of Australian children. Such a data set combined with an analysis method, such as Multilevel modeling, that allows individual growth trajectories and multiple influences over time to be modeled will assist to provide stronger and more robust evidence. This article determines the longitudinal relationship between frequency of consumption of SSBs and HFFs and change in body weight over 6 years (four waves) in Australian children aged from 4 to 10 years. It uses a multilevel growth model to identify whether SSB and HFF consumption is changing over time, and whether this consumption is linked to changes in BMI z-scores.
www.obesityjournal.org
Methods Sample and study design The LSAC is an ongoing nationally representative cross-sequential longitudinal survey study which aims to examine Australian children’s development and wellbeing and how this relates to social, economic, and cultural aspects of their environment. There are two cohorts; the B cohort (birth) and K cohort (kindergarten). The sample for this analysis comprised all children in the K-cohort (4-5years old at wave 1—baseline) with complete height and weight data from the four waves of LSAC (wave 1—baseline, 2004; wave 2—first follow-up, 2006; wave 3—2nd follow-up, 2008; and wave 4—3rd follow-up, 2010) (28). Briefly, LSAC employed a two-stage clustered sampling design stratified by a state and capital city statistical division /rest of state division and clustered by postcode within each stratum. Children born between March 1999 and February 2000 were randomly selected to achieve a cohort aged between 4.3 and 5.2 years at interview with all birth months represented. There were 4,169 children who completed all four waves of data collection and this cohort was representative of the Australian population on most demographic features (29). Trained professional interviewers conducted face-to-face interviews in the child’s home with the study child’s primary caregiver (“Parent 1,” usually the biological mother), who also completed a written questionnaire. There were 4,983 respondents who participated in the baseline (wave 1) survey, 4,464 at wave 2, 4,331 at wave 3 and 4,169 at wave 4 (Table 1). Compared to the baseline, the attrition rate for waves two, three, and four were, 10.4, 13.1, and 16.3%, respectively. Full details of the characteristics of responders are available from The LSAC technical papers (29-31). Common characteristics that were associated with continued participation in the study included; if Parent 1 was female, had a bachelor degree and the study child lived in a home that was being paid off rather than rented (29-31). Additionally, t test analysis of the key variables used in this study showed that the BMI z-scores at baseline of the children who remained in the study (mean 5 0.59) were similar to those who left the study (mean 5 0.58; P 5 0.65), average SSB consumption was lower for those who remained in the study (mean 5 1.4) compared to those who left (mean 5 1.7; P < 0.001) as was average HFF consumption (mean 5 2.0: mean 5 2.1; P 5 0.005).
Ethics Written informed consent was obtained for each participating child, and the LSAC study was approved by the Australian Institute of Family Studies Ethics Committee.
Anthropometric measurement Children’s weight was measured in light clothing to the nearest 50 g using glass bathroom scales (Salter Australia, Code 79985; Springvale, Victoria, Australia) and height to the nearest 0.1 cm using a portable rigid stadiometer (Invicta, (Leicester, UK), Model IPO955). The averages of two height measurements were used in analyses; where the two differed by more than 0.5 cm a third measurement was taken and the average of the two closest was used.
Outcome measure BMI z-scores were the outcome measures and these were calculated using the age- and sex-specific WHO Growth Standard for 6- to 60-
Obesity | VOLUME 22 | NUMBER 5 | MAY 2014
E97
Obesity
Sweet Drinks and High Fat Food and BMI Millar et al.
TABLE 1 Participant characteristics and differences between child BMI-z scores, frequency of consumption of SSB and HFF and mother BMI at the first wave (2004) and the subsequent waves (2006, 2008, and 2010, respectively)
Characteristic Child
Male, n (%)
Sugar sweetened beverage (mean number of times consumed in the 24 hours prior to survey)
High fat foods (mean number of times consumed in the 24 hours prior to survey)
Age (years) BMI z-scoresa All
Males Females All
Males Females Age (years) BMI, (kg m22)
Mother
Wave 1; N 5 4,983
Wave 2; N 5 4,464
Wave 3; N 5 4,331
Wave 4; N 5 4,169
2,537 Mean 4.8 0.65 1.7
2,277 Mean 6.9 0.51** 1.4**
2,212 Mean 8.8 0.59* 1.3**
2,132 Mean 10.9 0.59* 1.5**
(50.9) (SD) (0.2) (1.00) (1.2)
1.7 (1.2) 1.6 (1.2) 1.9 (1.2)
2.0 1.9 34.6 25.3
(1.3) (1.2) (5.2) (5.2)
(51.0) (SD) (0.2) (1.1) (1.2)
1.5** (1.2) 1.4** (1.1) 1.9 (1.2)
2.0 1.9 36.8 25.4
(1.3) (1.2) (5.1) (5.2)
(51.1) (SD) (0.2) (1.17) (1.1)
1.3** (1.1) 1.2** (1.1) 1.8* (1.2)
1.9 1.8* 38.9 25.8**
(51.1) (SD) (0.3) (1.19) (1.2)
1.6** (1.2) 1.5** (1.2) 2.5** (1.8)
(1.2) (1.1) (5.2) (5.3)
2.6** 2.3** 41.0 26.5**
(1.9) (1.7) (5.2) (5.7)
a WHO Reference 2007 (43); *significant at P < 0.05; **significant at P < 0.001. Bold text indicates a significant difference from wave1, differences tested using linear regression.
month-old children (32) and Growth Reference for 5- to 19-year-old children (33), where applicable.
Independent variables Diet. Both the SSB and HFF were derived variables from the LSAC dataset. They were both reported by parent 1 during the face to face interview. The stem of the items was identical, “In the last 24 h has your child had the following foods or drinks once, more than once or not at all? Frequency was coded 0 for “not at all,” 1 for “once” and 2 for “more than once.” Responses were summed and a final score for frequency of SSB and HFF consumption was allotted. SSBs comprised two parent-reported survey questions including the frequency of consumption of: (1) fruit juice; (2) soft drink or cordial (not diet), in the 24 h prior to the survey. HFFs was measured by the reported frequency of consumption of: (1) meat pie, hamburger, hot dog, sausage or sausage roll; (2) hot chips or French fries; (3) potato chips or savoury snacks such as “Twisties”; 4) biscuits, doughnuts, cake, pie or chocolate, in the 24 h prior to the survey.
Covariates.
The LSAC composite household socio-economic position (SEP) variable was constructed by Blakemore et al. and was derived from standardized scores for: combined annual household income (with natural log transformation); parents’ years of
E98
Obesity | VOLUME 22 | NUMBER 5 | MAY 2014
education; and parents’ occupations (main occupation and occupational status) (34). Demographic information such as child age and gender and mother age and self-reported anthropometry was also collected. Mother’s BMI was calculated using weight (kg)/height2 (m2). All continuous independent predictors were centered at the sample mean.
Missing data. There was very little missing data for most variables; BMI z-scores
Original Article PEDIATRIC OBESITY
Relationship Between Raised BMI and Sugar Sweetened Beverage and High Fat Food Consumption Among Children Lynne Millar1, Bosco Rowland2, Melanie Nichols1, Boyd Swinburn1,3, Catherine Bennett4, Helen Skouteris2 and Steven Allender1
Objective: Longitudinal evidence of relationships between unhealthy diets and BMI in children is crucial for appropriately targeting obesity prevention activities. The objective was to determine the relationship between frequency of consumption of sugar sweetened beverages (SSBs) and high fat foods (HFFs) and body weight in Australian children aged from 4 to 10 years. Methods: Data from 4,164 children participating in four waves (wave 1, 2004; wave 2, 2006; wave 3, 2008; and wave 4, 2010) of the Longitudinal Study of Australian Children were analyzed. A multi-level growth model tested relationships between consumption of SSB and HFF and BMI z-scores. Results: BMI z-scores were associated with daily consumption of HFF, SSB and maternal BMI independent of BMI z-scores at wave 1 (baseline); with each additional occurrence of SSB and HFF consumption intake per day, BMI z-score increased by 0.015 U (P < 0.01) and 0.014 U (P < 0.001), respectively. With each additional maternal BMI unit, BMI z-score increased by 0.032 (P < 0.001). Conclusions: Higher BMI z-scores were strongly associated with the consumption of SSBs and HFFs. Future efforts to prevent obesity should consider urgent action to address the impact of the consumption of SSBs and HFFs in childhood. Obesity (2014) 22, E96-E103. doi:10.1002/oby.20665
Introduction The level of obesity for children in developed countries, such as Australia, has reached epidemic proportions, and is now a major preventative health priority for most countries. Low levels of physical activity, high levels of television and computer screen time and high consumption levels of energy dense and nutrient poor food have all been identified as significant contributors to this epidemic (1-3). In Australia, the reduction of sugar sweetened beverages (SSB) and high fat foods (HFF) has been the focus of many prevention measures (4-6) but as yet there is not clear evidence about the relationship between high consumption of SSB and HFF and childhood obesity. Studies in developed and developing countries have demonstrated that increases in SSB are associated with increased body mass index (BMI) (7-10). These studies have principally examined BMI at two time points where higher levels of SSB consumption at the earlier time point was found to be predictive of greater levels of BMI at a later time point. Systematic reviews of cross sectional and two-time point studies have identified associations between consumption of SSBs and increasing body weight in adults and children (11-13).
Critics of the focus on reducing SSB to prevent obesity argue that the existing studies are limited because they are often cross sectional or examine risk at baseline against outcomes at time two without measures across multiple time points (14,15). In this context, it remains clear that stronger evidence from studies that allows for temporal influences is required, especially if preventative measures targeting SSB are to be sustained and supported in the long-term by funders and policy makers (11,16). Longitudinal studies whereby data on food and drink consumption, and individual, family and environmental are collected over several time points (>2) is one way of developing such evidence. Research from Barclay and Brand-Miller (17) point to a possible paradox; they identify decreasing sugar consumption and SSB consumption in Australia alongside increasing prevalence of obesity in Australia. Many have argued that this indicates no association between sugar consumption and obesity patterns at population level. Others have found that industry funding of nutrition-related scientific articles may bias conclusions in favor of sponsors’ products (12,18) and so potential conflicts of interest exist in the three key reviews demonstrating no association between high levels of SSB
1
WHO Collaborating Centre for Obesity Prevention, Deakin University, Geelong, Australia. Correspondence: Lynne Millar ([email protected]) School of Psychology, Deakin University, Melbourne, Australia 3 Population Nutrition and Global Health, University of Auckland, Auckland, New Zealand 4 Deakin Epidemiology, Deakin University, Melbourne, Australia
2
Disclosure: the authors have no competing interests. Author contributions: LM developed the major concepts, LM & BR conceived the analysis plan and analyzed the data. All authors were involved in writing the article and had final approval of the submitted and published versions. Received: 11 March 2013; Accepted: 21 November 2013; Published online 28 November 2013. doi:10.1002/oby.20665
E96
Obesity | VOLUME 22 | NUMBER 5 | MAY 2014
www.obesityjournal.org
Original Article
Obesity
PEDIATRIC OBESITY
consumption and high body weight as they were industry funded (15,16,19). Moreover, all data used in the analysis to support the proposed Australian paradox were cross sectional. Longitudinal designed studies can assist to untangle and better understand this seemingly paradoxical relationship. High fat food (HFF) consumption has also been implicated in the obesity epidemic among children and adolescents. Australian studies examining HFF consumption and obesity among children have shown mixed results; for example, cross-sectional studies have found positive relationships between more frequent fast food consumption at home and a person being overweight (20). Conversely, lower fat intake in girls and boys and lower consumption of energy dense snacks in boys has been shown to be associated with a greater risk of overweight/obesity in a sample of Australian adolescents (21). Caution needs to be applied when interpreting cross-sectional studies. For example, without temporal data they cannot establish causality nor consider the possibility of reverse causality (that is being overweight causing a reduction in HFF consumption). Overall, studies examining the relationship between children’s body weight and both SSB and HFF have been inconclusive. These studies have found weak relationships between consumption of HFF and increased body weight but more robust findings between SSB consumption and increased body weight. This has been found in one study within Australia (22,23) and other high income countries (8,24,25). The mixed findings may be due to the differences in foods included in studies and the limitations of dietary assessment methods. Measurement and interpretation of SSB consumption is less complex than measurement of HFFs as the kilojoule content of SSBs is quite uniform and the beverages are usually presented in standard sizes so measurement can be more precise. Conversely HFF includes a range of diverse food items, and studies have variously defined HFFs according to frequency of consumption, calculated energy density or number of convenience foods (not fast foods) (24,26,27). Stronger evidence is needed to adequately support government decisions on investing in strategies that reduce HFF and SSB consumption. Longitudinal data with more than two time points, using national representative samples, and which uses strong and reliable measures will be vital in providing this information. Such data provide the opportunity for analysis that allows for trajectory modeling, and the assessment of influences at a variety of levels. Such an opportunity is provided by the Longitudinal Study of Australian Children (LSAC); a large, high quality, observational longitudinal cohort study over multiple time-points that includes robust data collected every 2 years on a representative cohort of Australian children. Such a data set combined with an analysis method, such as Multilevel modeling, that allows individual growth trajectories and multiple influences over time to be modeled will assist to provide stronger and more robust evidence. This article determines the longitudinal relationship between frequency of consumption of SSBs and HFFs and change in body weight over 6 years (four waves) in Australian children aged from 4 to 10 years. It uses a multilevel growth model to identify whether SSB and HFF consumption is changing over time, and whether this consumption is linked to changes in BMI z-scores.
www.obesityjournal.org
Methods Sample and study design The LSAC is an ongoing nationally representative cross-sequential longitudinal survey study which aims to examine Australian children’s development and wellbeing and how this relates to social, economic, and cultural aspects of their environment. There are two cohorts; the B cohort (birth) and K cohort (kindergarten). The sample for this analysis comprised all children in the K-cohort (4-5years old at wave 1—baseline) with complete height and weight data from the four waves of LSAC (wave 1—baseline, 2004; wave 2—first follow-up, 2006; wave 3—2nd follow-up, 2008; and wave 4—3rd follow-up, 2010) (28). Briefly, LSAC employed a two-stage clustered sampling design stratified by a state and capital city statistical division /rest of state division and clustered by postcode within each stratum. Children born between March 1999 and February 2000 were randomly selected to achieve a cohort aged between 4.3 and 5.2 years at interview with all birth months represented. There were 4,169 children who completed all four waves of data collection and this cohort was representative of the Australian population on most demographic features (29). Trained professional interviewers conducted face-to-face interviews in the child’s home with the study child’s primary caregiver (“Parent 1,” usually the biological mother), who also completed a written questionnaire. There were 4,983 respondents who participated in the baseline (wave 1) survey, 4,464 at wave 2, 4,331 at wave 3 and 4,169 at wave 4 (Table 1). Compared to the baseline, the attrition rate for waves two, three, and four were, 10.4, 13.1, and 16.3%, respectively. Full details of the characteristics of responders are available from The LSAC technical papers (29-31). Common characteristics that were associated with continued participation in the study included; if Parent 1 was female, had a bachelor degree and the study child lived in a home that was being paid off rather than rented (29-31). Additionally, t test analysis of the key variables used in this study showed that the BMI z-scores at baseline of the children who remained in the study (mean 5 0.59) were similar to those who left the study (mean 5 0.58; P 5 0.65), average SSB consumption was lower for those who remained in the study (mean 5 1.4) compared to those who left (mean 5 1.7; P < 0.001) as was average HFF consumption (mean 5 2.0: mean 5 2.1; P 5 0.005).
Ethics Written informed consent was obtained for each participating child, and the LSAC study was approved by the Australian Institute of Family Studies Ethics Committee.
Anthropometric measurement Children’s weight was measured in light clothing to the nearest 50 g using glass bathroom scales (Salter Australia, Code 79985; Springvale, Victoria, Australia) and height to the nearest 0.1 cm using a portable rigid stadiometer (Invicta, (Leicester, UK), Model IPO955). The averages of two height measurements were used in analyses; where the two differed by more than 0.5 cm a third measurement was taken and the average of the two closest was used.
Outcome measure BMI z-scores were the outcome measures and these were calculated using the age- and sex-specific WHO Growth Standard for 6- to 60-
Obesity | VOLUME 22 | NUMBER 5 | MAY 2014
E97
Obesity
Sweet Drinks and High Fat Food and BMI Millar et al.
TABLE 1 Participant characteristics and differences between child BMI-z scores, frequency of consumption of SSB and HFF and mother BMI at the first wave (2004) and the subsequent waves (2006, 2008, and 2010, respectively)
Characteristic Child
Male, n (%)
Sugar sweetened beverage (mean number of times consumed in the 24 hours prior to survey)
High fat foods (mean number of times consumed in the 24 hours prior to survey)
Age (years) BMI z-scoresa All
Males Females All
Males Females Age (years) BMI, (kg m22)
Mother
Wave 1; N 5 4,983
Wave 2; N 5 4,464
Wave 3; N 5 4,331
Wave 4; N 5 4,169
2,537 Mean 4.8 0.65 1.7
2,277 Mean 6.9 0.51** 1.4**
2,212 Mean 8.8 0.59* 1.3**
2,132 Mean 10.9 0.59* 1.5**
(50.9) (SD) (0.2) (1.00) (1.2)
1.7 (1.2) 1.6 (1.2) 1.9 (1.2)
2.0 1.9 34.6 25.3
(1.3) (1.2) (5.2) (5.2)
(51.0) (SD) (0.2) (1.1) (1.2)
1.5** (1.2) 1.4** (1.1) 1.9 (1.2)
2.0 1.9 36.8 25.4
(1.3) (1.2) (5.1) (5.2)
(51.1) (SD) (0.2) (1.17) (1.1)
1.3** (1.1) 1.2** (1.1) 1.8* (1.2)
1.9 1.8* 38.9 25.8**
(51.1) (SD) (0.3) (1.19) (1.2)
1.6** (1.2) 1.5** (1.2) 2.5** (1.8)
(1.2) (1.1) (5.2) (5.3)
2.6** 2.3** 41.0 26.5**
(1.9) (1.7) (5.2) (5.7)
a WHO Reference 2007 (43); *significant at P < 0.05; **significant at P < 0.001. Bold text indicates a significant difference from wave1, differences tested using linear regression.
month-old children (32) and Growth Reference for 5- to 19-year-old children (33), where applicable.
Independent variables Diet. Both the SSB and HFF were derived variables from the LSAC dataset. They were both reported by parent 1 during the face to face interview. The stem of the items was identical, “In the last 24 h has your child had the following foods or drinks once, more than once or not at all? Frequency was coded 0 for “not at all,” 1 for “once” and 2 for “more than once.” Responses were summed and a final score for frequency of SSB and HFF consumption was allotted. SSBs comprised two parent-reported survey questions including the frequency of consumption of: (1) fruit juice; (2) soft drink or cordial (not diet), in the 24 h prior to the survey. HFFs was measured by the reported frequency of consumption of: (1) meat pie, hamburger, hot dog, sausage or sausage roll; (2) hot chips or French fries; (3) potato chips or savoury snacks such as “Twisties”; 4) biscuits, doughnuts, cake, pie or chocolate, in the 24 h prior to the survey.
Covariates.
The LSAC composite household socio-economic position (SEP) variable was constructed by Blakemore et al. and was derived from standardized scores for: combined annual household income (with natural log transformation); parents’ years of
E98
Obesity | VOLUME 22 | NUMBER 5 | MAY 2014
education; and parents’ occupations (main occupation and occupational status) (34). Demographic information such as child age and gender and mother age and self-reported anthropometry was also collected. Mother’s BMI was calculated using weight (kg)/height2 (m2). All continuous independent predictors were centered at the sample mean.
Missing data. There was very little missing data for most variables; BMI z-scores
