Meal Glycaemic Load of Normal-Weight and Overweight Hong Kong Children

ORIGINAL ARTICLE Meal glycaemic load of normal-weight and overweight Hong Kong children

LL Hui and EAS Nelson

Department of Paediatrics, Faculty of Medicine, The Chinese University of Hong Kong, Prince of Wales Hospital, Shatin, Hong Kong

Objective: To describe the pattern of meal glycaemic load of children in Hong Kong and to determine whether the meal glycaemic load is associated with childhood overweight. Method: Dietary records (3-day) of 316 Hong Kong children aged 6–7 years were collected. Glycaemic load was calculated from the estimated weight (WT), carbohydrate content (%CHO) and glycaemic index (GI) of each food taken using the equation: (% CHO Â GI Â WT/100). The meal glycaemic load was then the sum of the glycaemic loads of all food taken in each meal. Logistic regression analyses were used to compare the average meal glycaemic load (of breakfast, lunch and dinner) and other dietary parameters between overweight children and normal-weight children. Results: Breakfast, lunch, dinner and snacks provided, respectively, 17, 29, 29 and 25% of the total glycaemic load in a day. White rice (excluding congee (rice porridge) and glutinous rice) contributed almost half of the total glycaemic load. Adjusted logistic regression showed that the meal glycaemic load was not significantly associated with childhood overweight after adjusting for parental obesity, birth weight, sleeping duration, mean energy intake and paternal smoking. Conclusion: Meal glycaemic load calculated from current diet was not an independent factor associated with childhood overweight in children aged 6–7 years. Our data suggested that modifying the type of rice/staple consumed and choosing low- GI snacks could have a major influence on the total meal glycaemic load of young Hong Kong children. European Journal of Clinical Nutrition (2006) 60, 220–227. doi:10.1038/sj.ejcn.1602305; published online 9 November 2005

Keywords: glycemic index; dietary carbohydrates; obesity

Introduction of the concept of what has been termed glycaemic load (Ludwig, 2002; Brand-Miller et al., 2003) or glycaemic Glycaemic index (GI) is defined as the incremental area glucose equivalent (Monro and Williams, 2000). Both of under the blood glucose response curve following the intake these terms refer to a weighted GI based on the actual of a 50 g carbohydrate portion of a test food, expressed as a carbohydrate content of the food, the GI value of the percentage of the response to the same amount of carbo- carbohydrate and the amount of the food consumed. hydrate from a standard food (glucose or bread) taken by the A study by Wolever and Bolognesi (1996) suggested that GI same subject (Wolever et al., 1991). The use of the GI has determines the variability of the blood insulin response to a been endorsed as a method of categorizing the likely meal and the presence of protein and fat intake had metabolic effects of carbohydrates in the dietary manage- negligible effects on postprandial glucose and insulin, while ment of diabetics and impaired glucose tolerance (FAO/ recently the opposite was reported by Flint et al. (2004). A WHO, 1998). The utility of GI in managing diets and related study investigating the effects of dietary GI and glycaemic research has been further elaborated with the development load on energy metabolism and voluntary food intake in obese children showed that high-GI meals elicit a sequence of hormonal changes that might limit the availability of Correspondence: Professor EAS Nelson, Department of Paediatrics, Faculty of metabolic fuels and cause overeating in children (Ludwig Medicine, The Chinese University of Hong Kong, 6/F Clinical Science Building, et al., 1999). An increase in postpandial satiety by consuming Prince of Wales Hospital, Shatin, Hong Kong. lower dietary GI has been reported in some related studies E-mail: [email protected] Received 21 September 2004; revised 27 June 2005; accepted 12 August (Ball et al., 2003; Warren et al., 2003; Alvina and Araya, 2005; published online 9 November 2005 2004), while another study reported an inverse relationship Normal-weight and overweight Hong Kong children LL Hui and EAS Nelson 221 between appetite and postprandial blood glucose response (4120% median weight for height). From the same data set, (Anderson et al., 2002). we developed BMI cutoffs that would identify the same 8% The physiological mechanism suggested to be responsible of children who are overweight, as well as the corresponding for the relationship between satiety and dietary GI relates to BMI cutoffs that would identify the middle 10% and the the fact that insulin promotes the uptake of nutrients into lowest 8%. In this way, three groups of subjects were the liver. Therefore, a meal resulting in a high postprandial identified and recruited: an overweight group (X92nd glycaemic load, and an associated high insulin effect, will centile for BMI , weight in kg/height2 in m2), a middle- theoretically inhibit the hepatic release of glucose and weight group (45–55th centile for BMI) and a low-weight suppress lipolysis. Difficulties in accessing stored metabolic group (p8th centile for BMI). We did not intend to recruit fuels may lead to excessive hunger and overeating. a random sample for this study, but instead three distinct It is generally believed that rising intakes of dietary fat weight groups. The number of subjects in each group was the and the resultant energy imbalance has played an important maximum number we could achieve during the recruitment role in causing the obesity epidemic (WHO, 2000). However, period, and thus they did not represent a normal distribution previous studies have shown inconsistent associations in the population. A home visit was arranged for each between dietary fat intake and childhood obesity (Muecke participating family to collect dietary data and other relevant et al., 1992; Troiano et al., 2000). Moreover, the benefits of information. dietary fat restriction on weight management have been Children’s weight (to the nearest 0.1 kg) and height (to the only modest (Astrup et al., 2000). Some studies have focused nearest 0.1 cm) were measured by nursing staff at the Student on the role of other dietary factors, such as carbohydrates Health Service Centres. Students stood against the stadio- and the GI values of meals, in promoting or preventing meter for height measurements. Weight was measured by weight gain and obesity (Slabber et al., 1994; Ludwig et al., using an electronic scale. Weight and height of the parents 1999; Bessesen, 2001). The slower digestion of carbohydrates were reported by the parents themselves during the home with low-GI values has been found to be associated with visits. Adulthood obesity is defined as BMI of 25 kg mÀ2 or higher satiety (Holt et al., 1992), improved insulin sensitivity above. The BMI cutoff for childhood overweight used in (Frost et al., 1999) and other beneficial metabolic effects this study was defined as X92nd centile of growth data (Jenkins et al., 1987). A retrospective cohort study showed collected in 1993, and it was age and sex specific (Leung that body weight and body mass index (BMI) of obese et al., 1996). children reduced significantly more in those children Prior to the home interview, 3-day dietary records were receiving low-GI dietary advice as compared to the group posted to the subjects’ parents/caregivers who were asked to receiving advice to reduce fat intake (Spieth et al., 2000). The record all food consumed by their child over a 3-day period health claims of consuming low-GI diets have been popular- (two weekdays and one weekend day). Names, serving sizes, ized in a number of diet books (Brand et al., 1996; cooking methods and the brand names of the foods Montignac, 1999). While the concept of GI in body weight consumed were recorded. Food pictures and eating utensils control is moving towards public usage, controversies still such as tablespoon, teaspoons and cups were used to exist (Raben, 2002). determine the serving size. The completed 3-day dietary As part of a case–control study designed to identify record sheets were collected by the interviewer during the risk factors for childhood obesity in Hong Kong Chinese subsequent home visit. The interviewer checked the in- children aged 6–7 years (Hui et al., 2003), we set out to assess formation on the record sheets and clarified any vague the contribution by different foods to the total meal information with the family or child using coloured food glycaemic load and the pattern in glycaemic load over a pictures and standard eating utensils. day, and to determine whether the calculated meal glycae- The meal glycaemic load was calculated with SERVE mic load of the current diet was associated with childhood Nutritional Management System for Windowss (M & H overweight. Williams Pty Ltd) (Monro and Williams, 2000). The software program converts the weight (WT), the carbohydrate content (%CHO) and the GI value of each food intake to Method glycaemic load by using equation (1) (Ludwig, 2002). A food with zero carbohydrate content will be assigned with a zero A total of 343 Hong Kong young children aged 6–7 years glycaemic load. participated in a study to identify the risk factors for Glycaemic load of a food ¼ % CHOÂGIÂWTðgÞ=100 ð1Þ childhood overweight and 316 of them completed 3-day dietary records. All subjects were primary-one students who Since the glycaemic load is considered to be additive, the were recruited during February 2000–May 2000 when they glycaemic load of a meal was calculated as the sum of all the attended one of the 12 Student Health Service Centres of the individual glycaemic loads of foods consumed in the meal. Department of Health for an annual body check. In 1993, The meal glycaemic load calculated in this way is expected a local cross-sectional growth survey (Leung et al., 1996) to give an estimation of the postprandial glucose level after classified 8% of the studied children aged 6–7 as overweight the intake of the respective meal. The example in Table 1

European Journal of Clinical Nutrition Normal-weight and overweight Hong Kong children LL Hui and EAS Nelson 222 Table 1 Calculation of the meal glycaemic load for a sample breakfast (o1600 or X1600 kcal/day) and father as a current smoker Food eaten GI % CHO Intake amount/g Glycaemic load (yes or no). The odds ratios (ORs) and their 95% confidence intervals White bread 70 50% 96 33.6 (95% CI) were calculated and the significance was set at Butter 0 0 10 0 Po0.05. The SPSS version 11.0 was used in the analyses. Strawberry jam 51 100% 20 10.2 Apple juice 40 13.20% 250 13.2

Meal glycaemic load 57 Results

The mean age of the subjects was 6.770.3 years. The mean illustrates the calculation of meal glycaemic load for a BMI for the overweight (n ¼ 121), middle-weight (130) and 7 7 breakfast. low-weight (65) groups were 20.5 2.1, 15.0 0.2 and 7 2 For this study, the SERVE program was customized to 12.8 0.3 kg/m , respectively. No significant difference in incorporate a Hong Kong food composition database based BMI between male and female subjects was found in each on food composition tables from Britain, China, Taiwan and weight group (data not shown). In all, 98.5% of the children the United States (Leung et al., 1999). GI values for all the in the overweight group met the new international cutoff for 2 434 foods in the Hong Kong food composition database were overweight (equivalent to 25 kg/m in adults) derived by entered for the meal glycaemic load calculation. Food items Cole et al. (2000). All but one of the children in the low- without published GI and containing a carbohydrate con- weight group was within the normal weight range (80–120% tent of 9% or less (n ¼ 216), including most vegetables and median weight-for-height). meat, were assumed to have GI of zero. For the remaining The average distributions of glycaemic load in the three food items, that is, foods with carbohydrate contents 10% or weight groups during the morning (food eaten before 1200), more, GI values were set with references from measured GI afternoon (lunch and afternoon tea) and evening (dinner values from the international tables of GI (Foster-Powell and and food eaten before bedtime) over 3 days are shown in Miller, 1995), published books on GI values (Brand et al., Figure 1. Food consumed in the morning provided the lowest 1996; Brand and Foster-Powell, 1998; Montignac, 1999), and value of glycaemic load, while that in the afternoon was the the GI research group from the Human Nutrition Unit at the highest and this trend was the same for the three weight University of Sydney (SUGiRS, 2002). GI of carbohydrate- groups. The respective glycaemic load contributions of rich food with untested GI was estimated by their carbo- breakfast, lunch, dinner and snack (defined as food intake hydrate content, ingredients and published food GI. These out of the three main meals) were 17, 29, 29 and 25%. White estimated GI values were reviewed by researchers in the GI rice, bread (excluding bread made from whole grain), research group from the Human Nutrition Unit at the beverages (including soft drinks and juice, but excluding University of Sydney. The study was approved by the Ethics milk) and fruits (including dried fruits) contributed 41, 13, Committee of the Chinese University of Hong Kong and written informed consent was obtained from the parents of all subjects. The average meal glycaemic load of each individual was obtained by averaging the glycaemic load of the three main meals (breakfast, lunch and dinner), while the glycaemic load of snacks in-between the three main meals were not taken into account. Tertile cutoffs were used to categorize the average meal glycaemic load of individuals into three groups for subsequent statistical analyses. The relationship between characteristics of children and the tertiles of meal glycaemic load were explored by ANOVA (for continuous variables) and w2 test (for categorical variables). The meal glycaemic load and daily macronutrient consumption by different weight groups were assessed by ANOVA. Logistic regression was used to assess the effect of mean glycaemic load on being overweight. The model was adjusted for risk factors for childhood overweight determined earlier (Hui et al., 2003): paternal obesity (BMIo25 or X25 kg/m2), maternal obesity (BMIo25 or À2 Figure 1 Average distribution of glycaemic loads during the X25 kg m ), birth weight (o3.0, 3.0–3.5, X3.5 kg), sleeping morning, afternoon and evening for overweight and normal-weight duration (9, 9–11 or X11 h/day), mean energy intake (middle- and low-weight) Hong Kong children aged 6–7 years.

European Journal of Clinical Nutrition Normal-weight and overweight Hong Kong children LL Hui and EAS Nelson 223 hydrate (Po0.0005) and protein (P ¼ 0.003), but not fat (P ¼ 0.215). Increased glycaemic load was positively associated with % carbohydrates (Po0.0005), but negative associations were observed for % proteins (Po0.0005) and % fats (Po0.0005). Paternal or maternal obesity, having smoking parents, age and birth weight were not associated with meal glycaemic load. The average daily energy intake, macronutrient intake and meal glycaemic load were assessed across the three weight groups and significant difference was found in meal glycaemic load (P ¼ 0.004) and daily energy intake (P ¼ 0.026), but not in % carbohydrate (P ¼ 0.086), % proteins (P ¼ 0.724) nor % fats (P ¼ 0.079) (Table 3). Figure 2 Percentage of glycaemic load contributed by selected The unadjusted and adjusted ORs are shown in Table 4. foods rich in carbohydrate based on 3-day dietary records of Hong The third tertile (OR ¼ 1.79, 95% CI ¼ 1.00–3.08), but not Kong children aged 6–7 years. the second tertile (OR ¼ 1.27, 95% CI ¼ 0.72–2.23), of meal glycaemic load significantly increased the odds of overweight compared with the corresponding first tertile group 10 and 5% of total glycaemic load, respectively (Figure 2). (reference category). When the factors previously shown to The combined contribution of these four food items to the be associated with childhood overweight were adjusted for, total glycaemic load was 69%. the association between meal glycaemic load and overweight The characteristics of the subjects by the tertiles of meal became insignificant. glycaemic load, that is, average of the three main meals and In logistic regression only including the meal glycaemic excluding snacks, are shown in Table 2. A significant increase load and energy intake, the interaction term was borderline in BMI was observed across the three tertile groups significant (P ¼ 0.074). Further stratified analyses by energy (P ¼ 0.017). Children with higher glycaemic load were likely intake (at, below and above 1469 kcal/day) suggested that to have greater consumption of energy (Po0.0005), carbo- there might be a positive effect of meal glycaemic load

Table 2 Characteristics of subjects by tertiles of mean meal glucose load

Mean meal glycaemic load P-value

First tertile (n ¼ 105) Second tertile (n ¼ 106) Third tertile (n ¼ 105)

Meal glycaemic load, mean7s.d. 20.574.1 29.472.3 38.575.7 — Age (years), mean7s.d. 6.970.3 6.970.3 6.970.4 0.720 BMI, mean7s.d. 15.972.9 16.773.5 17.373.6 0.017 Paternal obesitya, n (%) 29 (29.0) 25 (25.8) 18 (19.6) 0.311 Maternal obesitya, n (%) 14 (13.7) 12 (12.0) 22 (22.7) 0.091

Birth weight (kg), n (%) o3.0 46 (43.8) 46 (44.2) 31 (29.5) 3.0–3.5 35 (33.3) 37 (35.6) 44 (41.9) X3.5 24 (22.9) 21 (20.2) 30 (28.6) 0.166

Father being a current smoker, n (%) 40 (38.1) 38 (36.5) 39 (37.9) 0.969 Mother being a current smoker, n (%) 5 (4.8) 7 (6.7) 9 (8.7) 0.530 Energy intake (kcal/day), mean7s.d. 13187237 14957286 16117238 o0.0005

Carbohydrate intake (g/day), mean7s.d. 166732 197733 226732 o0.0005 % of total energy, mean7s.d. 50.576.0 53.476.0 56.475.9 o0.0005

Protein intake (g/day), mean7s.d. 57714 62718 63716 0.009 % of total energy, mean7s.d. 17.372.8 16.572.6 15.772.8 o0.0005

Fat intake (g/day), mean7s.d. 49712 52716 52713 0.215 % of total energy, mean7s.d. 33.574.6 31.074.5 28.974.6 o0.0005 aBMIX25.

European Journal of Clinical Nutrition Normal-weight and overweight Hong Kong children LL Hui and EAS Nelson 224 Table 3 Dietary composition by the three weight groups of children

Weight groups P-value

Overweight (n ¼ 121) Middle (n ¼ 130) Low (n ¼ 65)

Meal glycaemic load, mean7s.d. 31.279.2 29.177.8 27.077.6 0.004 Average energy intake/day, mead7s.d. 15197306 14707259 14037261 0.026 % Carbohydrates, mean7s.d. 52.576.3 54.276.5 53.976.4 0.086 % Protein, mean7s.d. 16.672.8 16.372.9 16.572.6 0.724 % Fat, mean7s.d. 31.875.0 30.474.9 31.275.1 0.079

Table 4 ORs (95% CI) for overweight by meal glycaemic load

Overweight n (%) Middle/low weight n (%) Unadjusted OR (95% CI) Adjusted a OR (95% CI)

Meal glycaemic load First tertile 34 (28) 71 (37) 1.00 1.00 Second tertile 40 (33) 66 (34) 1.27 (0.72–2.23) 1.01 (0.51–2.01) Third tertile 48 (39) 57 (29) 1.79 (1.00–3.08) 1.08 (0.52–2.26) P-for-trend 0.048 0.832

aAdjusted for paternal obesity (BMIo25 or X25 kg/m2), maternal obesity (BMIo25 or X25 kg/m2), birth weight (o3.0, 3.0–3.5, X3.5 kg), sleeping (9, 9–11 or X11 h/day), mean energy intake (o1600 or X1600 kcal/day) and father as a current smoker (yes or no).

among children with higher energy consumption, but not this was one of the risk factors being adjusted for, it was among the lower energy consumption group (data not possible that the unadjusted relationship between glycaemic shown). load and overweight was confounded by energy intake. Meal glycaemic load (average glycaemic load of three main Further analysis suggested that the effect of meal glycaemic meals and excluding snacks) was used in these analyses. load might be modified by energy consumption and might The main analyses were repeated using total glycaemic load only be associated with increased risk of overweight in the (all meals and snacks), snack glycaemic load and breakfast higher energy consumers. glycaemic load, and none of these were found to be Dietary glycaemic load was found to increase the risk of significantly associated with childhood overweight in the coronary heart disease, but has no association with BMI in a unadjusted model (data not shown). large prospective cohort study (Liu et al., 2000). Risk factors for overweight might be influenced by both genetic and environmental factors (Heitmann et al., 1995; Perusse and Discussion Bouchard, 2000). Genetically predisposed children might be more vulnerable to the unfavourable metabolic con- A number of studies have indicated that carbohydrates and sequences of consuming high-GI foods, whereas normal the GI values of meals can play a role in promoting weight children may not gain weight when subjected to the same gain and obesity (Slabber et al., 1994; Ludwig et al., 1999; high glycaemic load. In view of the fact that some Bessesen, 2001; Anderson et al., 2002). A low-GI diet has intervention studies in children have found an increase in been advocated for the treatment and prevention of over- satiety and a reduction in food consumption after lowering weight and obesity in the general public in the US (Ludwig, the glycaemic load of meals (Spieth et al., 2000; Warren et al., 2000; Ebbeling, 2001; Pawlak et al., 2002). While the concept 2003), and beneficial outcome has also been observed in of GI in body weight control is moving into the public arena, intervention studies in adults (Wolever and Mehling, 2003; the present study has provided relevant information on the Sloth et al., 2004), more long-term intervention studies are pattern of glycaemic load and the relationship between meal needed to delineate the potential causal relationship glycaemic load and childhood overweight in a Chinese between GI and obesity in childhood. population. In this case–control study of 6-year-old children, While the role played by dietary GI in obesity develop- an increase in the mean BMI across the three tertile groups of ment and treatment is still inconclusive (Pawlak et al., 2002; meal glycaemic load and a significant positive association Raben, 2002), the evidence for GI in diabetes management in unadjusted logistic regression was observed. However, seems to be substantial (Buyken et al., 2001; Gilbertson et al., when risk factors for overweight previously published were 2001; Jimenez-Cruz et al., 2003). The practical utility of GI adjusted, the meal glycaemic load was not significantly has been further suggested by the joint FAO/WHO expert associated with childhood overweight. Since meal glycaemic consultation on carbohydrates, which had directly addressed load was positively associated with energy intake, and since the usage of GI in the dietary management of diabetes

European Journal of Clinical Nutrition Normal-weight and overweight Hong Kong children LL Hui and EAS Nelson 225 (FAO/WHO, 1998). Other studies also revealed the potential in this study indicated that it may be possible to modify application of GI in preventing or treating cardiovascular nearly three-quarters of the total glycaemic load by swapping disease (Liu et al., 2000), undesirable lipid profile (Jenkins the types of rice/staple and snacks consumed to low-GI et al., 1987; Frost et al., 1999; Liu et al., 2001) and cancer alternatives. (Augustin et al., 2001; Franceschi et al., 2001). However, it This study had some limitations. Dietary intakes vary day- is practically difficult to plan a low-GI diet (Raben, 2002) by-day and thus the total glycaemic load calculated from because there is a shortage of low-GI foods in the market and the 3-day dietary records might not be representative of the most staple foods have high GI values (Bjorck et al., 2000). usual nutrient intake of the subjects. Underreporting, This is also the case in Hong Kong, where steamed white rice especially for overweight children of unhealthy or ‘bad’ is the staple for most local people. Different from European foods, might also underestimate the actual intake as well and American community where, respectively, bread and (Milner and Allison, 1999). It should also be recognized that pasta (Buyken et al., 2001) and potatoes and cold breakfast the dietary habits of the children might have changed as a cereals (Liu et al., 2000) were the most important contribu- result of their weight status, so that the current diet may not tors to dietary glycaemic load, our data suggested that the reflect the diet taken when the child was becoming over- contribution of steamed white rice to the total dietary weight. glycaemic load was the highest. It contributed 41% of the The average meal glycaemic load was used in the analysis total meal glycaemic load of our 6-year-old subjects, which of this study; therefore, fluctuation in glycaemic load in each was relatively higher than bread (13%) and beverages (10%) individual among different meals in a day and the meal-to- (Figure 2). It also explained the higher glycaemic load during meal difference in glycaemic load were not considered in this the afternoon and evening, as rice is commonly consumed study. Although we gave all carbohydrate foods a GI value for at lunch and dinner (Figure 1). GI of rice, ranging from calculating the glycaemic load, this was in many cases a ‘best 35 (parboiled high-amylose white rice) to 139 (boiled low- estimate’ from experts in the field, as precise GI values for amylose white rice) (Foster-Powell et al., 2002), is determined many local Chinese foods are not known. Allocating GI by its ratio of amylose-amylopectin and cooking method values to foods is further complicated by the findings that used. The rice commonly consumed in Hong Kong is the same types of food may have different GI values in Jasmine rice from Thailand, whose GI value as stated in the different studies (Foster-Powell et al., 2002). Rice, the most international table is 109. The Jasmine rice available in Hong important staple in the Chinese diet, has different GI values Kong has not been tested for its GI, so we used a modest for different brands. It was thus likely that there would be estimated GI of 74 for the glycaemic load calculation to discrepancies between the estimated GI of these foods with avoid overestimation. Therefore, the contribution from rice their real values. It is also possible that underestimation on in the total glycaemic load was possibly higher than what the meal glycaemic load could occur if large amounts of our results showed. By substituting Jasmine rice with other foods containing less than 10% carbohydrate were ingested, varieties with lower GI, such as Basmati rice (GI ¼ 58) and as the GI of such foods was assumed to be zero. Although it Bangladeshi rice (GI ¼ 28), or other types of staple such as has been suggested that the concept of GI can be applied to wholemeal spaghetti (GI ¼ 37) and mung bean noodle mixed meals (Wolever et al., 1991), there is evidence that the (GI ¼ 39), it would be possible to make a significant change same kinds of food have a different glycaemic response when in the meal glycaemic load in this population group. ingested at different times of the day (Pi-Sunyer, 2002) and Snacks, defined as food intake between any two main calculated dietary GI in mixed meals cannot predict post- meals in this study, contributed to a quarter of the total prandial glycaemic and insulinaemic responses (Flint et al., glycaemic load, showing that snacks are an important part of 2004). Some factors that might affect the true meal these young children’s diet. Ludwig et al. (1999) showed that glycaemic load were unable to be considered in this study, voluntary energy intake in obese adolescents was greater such as physical form of the food, cooking method, meal following the consumption of meals with higher GI values. A combinations and eating rate. The application of the formula crossover study has reported a lower lunch intake in a group we used to calculate the glycaemic load was probably an of young children who consumed low-GI breakfast com- oversimplification of the real postprandial situation. pared with those after high-GI breakfast (Warren et al., 2003). Ball et al. (2003) demonstrated a prolongation of satiety following an intervention with a low-GI meal replacement, Conclusions while Alvina and Araya (2004) observed significantly less satiety in obese children after consumption of a carbo- Our study provides information on the pattern of dietary hydrate meal with a rapid digestion rate. Although some glycaemic load in a population of young children and studies did not suggest an increase postprandial satiety by emphasizes the importance of particular key foods in increasing dietary GI (Heini et al., 1998; Anderson et al., designing more practical low-GI diets. Although glycaemic 2002), providing snacks with lower GI values probably cause load was not found to be an independent risk factor for no harm and possibly reduce the frequency of snacking due childhood overweight in this study, our data suggested that to hunger. The contribution of different foods in the meal GI glycaemic load may play a role in childhood overweight

European Journal of Clinical Nutrition Normal-weight and overweight Hong Kong children LL Hui and EAS Nelson 226 among children with higher energy consumption. The value FAO/WHO (1998). Carbohydrates in human nutrition. Report of a of low-GI diets in weight management and prevention needs Joint FAO/WHO Expert Consultation. FAO Food Nutr Paper 66, further study. It will be necessary to obtain GI values of more 1–140. Flint A, Moller BK, Raben A, Pedersen D, Tetens I, Holst JJ et al. Chinese foods to better estimate the glycaemic load of (2004). The use of glycaemic index tables to predict glycaemic Chinese diets. Further validation of the calculated meal index of composite breakfast meals. Br J Nutr 91, 979–989. glycaemic load with actual glycaemic response is also Foster-Powell K, Holt SH, Brand-Miller JC (2002). International table needed. of glycemic index and glycemic load values: 2002. Am J Clin Nutr 76, 5–56. Foster-Powell K, Miller JB (1995). International tables of glycemic index. Am J Clin Nutr 62, 871S–890S. Acknowledgements Franceschi S, Dal Maso L, Augustin L, Negri E, Parpinel M, Boyle P et al. (2001). Dietary glycemic load and colorectal cancer risk. Ann We thank Professor J Brand Miller and Dr S Holt of the Oncol 12, 173–178. Human Nutrition Unit, University of Sydney for their advice Frost G, Leeds AA, Dore CJ, Madeiros S, Brading S, Dornhorst A (1999). Glycaemic index as a determinant of serum HDL- on GI values of foods, and Ms LM Yu (Centre for Clinical cholesterol concentration. Lancet 353, 1045–1048. Trial and Epidemiological Research, Chinese University of Gilbertson HR, Brand-Miller JC, Thorburn AW, Evans S, Chondros P, Hong Kong) for her statistical advice. We are also grateful to Werther GA (2001). The effect of flexible low glycemic index Dr LY Tse, Dr G Tong and the doctors and nurses in the dietary advice versus measured carbohydrate exchange diets on glycemic control in children with type 1 diabetes. Diabetes Care Student Health Service Centres, Department of Health, for 24, 1137–1143. helping to recruit the subjects for this study. Heini AF, Kirk KA, Lara-Castro C, Weinsier RL (1998). Relationship between hunger-satiety feelings and various metabolic parameters in women with obesity during controlled weight loss. Obes Res 6, References 225–230. Heitmann BL, Lissner L, Sorensen TI, Bengtsson C (1995). Dietary fat intake and weight gain in women genetically predisposed for Alvina M, Araya H (2004). Rapid carbohydrate digestion rate obesity. Am J Clin Nutr 61, 1213–1217. produced lesser short-term satiety in obese preschool children. Holt S, Brand J, Soveny C, Hansky J (1992). Relationship of satiety to Eur J Clin Nutr 58, 637–642. postprandial glycaemic, insulin and cholecystokinin responses. Anderson GH, Catherine NL, Woodend DM, Wolever TM (2002). Appetite Inverse association between the effect of carbohydrates on blood 18, 129–141. glucose and subsequent short-term food intake in young men. Am Hui LL, Nelson EA, Yu LM, Li AM, Fok TF (2003). Risk factors for J Clin Nutr 76, 1023–1030. childhood overweight in 6- to 7-y-old Hong Kong children. Int J Astrup A, Ryan L, Grunwald GK, Storgaard M, Saris W, Melanson E Obes Relat Metab Disord 27, 1411–1418. et al. (2000). The role of dietary fat in body fatness: evidence from Jenkins DJ, Wolever TM, Collier GR, Ocana A, Rao AV, Buckley G et al. a preliminary meta-analysis of ad libitum low-fat dietary interven- (1987). Metabolic effects of a low-glycemic-index diet. Am J Clin tion studies. Br J Nutr 83 (Suppl 1), S25–S32. Nutr 46, 968–975. Augustin LS, Dal Maso L, La Vecchia C, Parpinel M, Negri E, Jimenez-Cruz A, Bacardi-Gascon M, Turnbull WH, Rosales-Garay P, Vaccarella S et al. (2001). Dietary glycemic index and glycemic Severino-Lugo I (2003). A flexible, low-glycemic index Mexican- load, and breast cancer risk: a case–control study. Ann Oncol 12, style diet in overweight and obese subjects with type 2 diabetes 1533–1538. improves metabolic parameters during a 6-week treatment period. Ball SD, Keller KR, Moyer-Mileur LJ, Ding YW, Donaldson D, Jackson Diabetes Care 26, 1967–1970. WD (2003). Prolongation of satiety after low versus moderately Leung SS, Chan SM, Lui S, Lee WTK, Davies DP (1999). Growth and high glycemic index meals in obese adolescents. Pediatrics 111, nutrition of Hong Kong children aged 0–7 years. J Paediatr Child 488–493. Health 36, 56–65. Bessesen DH (2001). The role of carbohydrates in insulin resistance. Leung SS, Lau JT, Tse LY, Oppenheimer SJ (1996). Weight-for-age and J Nutr 131, 2782S–2786S. weight-for-height references for Hong Kong children from birth to Bjorck I, Liljeberg H, Ostman E (2000). Low glycaemic-index foods. 18 years. J Paediatr Child Health 32, 103–109. Br J Nutr 83 (Suppl 1), S149–S155. Liu S, Manson JE, Stampfer MJ, Holmes MD, Hu FB, Hankinson SE Brand MJ, Foster-Powell K (1998). The Glucose Revolution Pocket Guide et al. (2001). Dietary glycemic load assessed by food-frequency to the Top 100 Low Glycemic Foods. Marlowe & Company: New questionnaire in relation to plasma high-density-lipoprotein York, NY. cholesterol and fasting plasma triacylglycerols in postmenopausal Brand MJ, Foster-Powell K, Colagiuri S, Leeds A (1996). The GI Factor: women. Am J Clin Nutr 73, 560–566. The Glucose Revolution. Griffin Press Pty Ltd: Netley, South Liu S, Willett WC, Stampfer MJ, Hu FB, Franz M, Sampson L et al. Australia. (2000). A prospective study of dietary glycemic load, carbohydrate Brand-Miller JC, Thomas M, Swan V, Ahmad ZI, Petocz P, Colagiuri S intake, and risk of coronary heart disease in US women. Am J Clin (2003). Physiological validation of the concept of glycemic load in Nutr 71, 1455–1461. lean young adults. J Nutr 133, 2728–2732. Ludwig DS (2000). Dietary glycemic index and obesity. JNutr130, Buyken AE, Toeller M, Heitkamp G, Karamanos B, Rottiers R, Muggeo 280S–283S. M, et al., EURODIAB IDDM Complications Study Group (2001). Ludwig DS (2002). The glycemic index: physiological mechanisms Glycemic index in the diet of European outpatients with type 1 relating to obesity, diabetes, and cardiovascular disease. JAMA 287, diabetes: relations to glycated hemoglobin and serum lipids. Am J 2414–2423. Clin Nutr 73, 574–581. Ludwig DS, Majzoub JA, Al Zahrani A, Dallal GE, Blanco I, Roberts SB Cole TJ, Bellizzi MC, Flegal KM, Dietz WH (2000). Establishing a (1999). High glycemic index foods, overeating, and obesity. standard definition for child overweight and obesity worldwide: Pediatrics 103, E26. international survey. BMJ 320, 1240–1243. Milner JA, Allison RG (1999). The role of dietary fat in child nutrition Ebbeling CBL (2001). Treating obesity in youth: should dietary and development: summary of an ASNS workshop. American glycemic load be a consideration? Adv Pediatr 48, 179–212. Society for Nutritional Sciences. J Nutr 129, 2094–2105.

European Journal of Clinical Nutrition Normal-weight and overweight Hong Kong children LL Hui and EAS Nelson 227 Monro JA, Williams M (2000). Concurrent management of post- Spieth LE, Harnish JD, Lenders CM, Raezer LB, Pereira MA, Hangen SJ prandial glycaemia and nutrient intake using glycemic glucose et al. (2000). A low-glycemic index diet in the treatment of equivalents, food composition data and computer-assisted meal pediatric obesity. Arch Pediatr Adolesc Med 154, 947–951. design. Asia Pac J Clin Nutr 9, 67–73. Sydney University GI Research Service (SUGiRS) (2002). The Human Montignac M (1999). Eat yourself Slimy and Stay Slim. Nutrimont/ Nutrition Unit, Department of Biochemistry, The University of Montignac Publishing UK Ltd: UK. Sydney,http://www.glycemicindex.com/. Muecke L, Simons-Morton B, Huang IW, Parcel G (1992). Is child- Troiano RP, Briefel RR, Carroll MD, Bialostosky K (2000). Energy and hood obesity associated with high-fat foods and low physical fat intakes of children and adolescents in the United States: data activity? J School Health 62, 19–23. from the national health and nutrition examination surveys. Am J Pawlak DB, Ebbeling CB, Ludwig DS (2002). Should obese patients be Clin Nutr 72, 1343S–1353S. counselled to follow a low-glycaemic index diet? Yes Obes Rev 3, Warren JM, Henry CJ, Simonite V (2003). Low glycemic index 235–243. breakfasts and reduced food intake in preadolescent children. Perusse L, Bouchard C (2000). Gene–diet interactions in obesity. Am J Pediatrics 112, e414–e419. Clin Nutr 72, 1285S–1290S. WHO (2000). Obesity: preventing and managing the global epidemic: Pi-Sunyer FX (2002). Glycemic index and disease. Am J Clin Nutr 76, report of a WHO Consultation. World Health Organization: Geneva. 290S–298S. Technical Report Series 894, p 101. Raben A (2002). Should obese patients be counselled to follow a low- Wolever TMS, Bolognesi C (1996). Prediction of glucose and insulin glycaemic index diet? No. Obes Rev 3, 245–256. responses of normal subjects after consuming mixed meals Slabber M, Barnard HC, Kuyl JM, Dannhauser A, Schall R (1994). varying in energy, protein, fat, carbohydrate and glycemic index. Effects of a low-insulin-response, energy-restricted diet on weight J Nutr 126, 2807–2812. loss and plasma insulin concentrations in hyperinsulinemic obese Wolever TMS, Jenkins DJ, Jenkins AL, Josse RG (1991). The glycemic females. Am J Clin Nutr 60, 48–53. index: methodology and clinical implications. Am J Clin Nutr 54, Sloth B, Krog-Mikkelsen I, Flint A, Tetens I, Bjorck I, Vinoy S et al. 846–854. (2004). No difference in body weight decrease between a low- Wolever TMS, Mehling C (2003). Long-term effect of varying the source glycemic-index and a high-glycemic-index diet but reduced LDL or amount of dietary carbohydrate on postprandial plasma glucose, cholesterol after 10-wk ad libitum intake of the low-glycemic-index insulin, triacylglycerol, and free fatty acid concentrations in subjects diet. Am J Clin Nutr 80, 337–347. with impaired glucose tolerance. Am J Clin Nutr 77, 612–621.

European Journal of Clinical Nutrition