- Email: [email protected]
Milk consumption and bone mineral content in Chinese adolescent girls
Bone Vol. 30, No. 3 March 2002:521–528
Milk Consumption and Bone Mineral Content in Chinese Adolescent Girls X. Q. DU,1 H. GREENFIELD,1 D. R. FRASER,2 K. Y. GE,3 Z. H. LIU,4 and W. HE3 1
Department of Food Science and Technology, University of New South Wales, Australia Department of Animal Science, University of Sydney, Sydney, Australia 3 Chinese Academy of Preventive Medicine, Beijing, China 4 Beijing China–Japan Friendship Hospital, Beijing, China 2
Key Words: Milk consumption; Calcium intake; Vitamin D intake; Chinese adolescents; Bone mineral content; Physical activity.
A cross-sectional study of a random sample of 649 girls, aged 12–14 years (mean ⴞ SD: 12.9 ⴞ 0.6 years), in the Beijing area examined the relationship between diet and bone mineral status. Food and nutrient intakes over the past year were estimated by means of a semiquantitative food frequency questionnaire. Bone mineral content (BMC) and bone width (BW) at the distal one-third and one-tenth radius and ulna were measured by single-photon absorptiometry. Results showed Beijing pubertal girls had a low mean milk consumption (fresh and powdered milk, vitamin D-fortified milk, and yogurt) at 50 g/day (95% confidence interval [CI] 44 –55 g/day whereas one-third consumed no milk at all. Mean calcium intake was 356 ⴞ 97 mg/day of which only 21% was provided by milk and milk products. Milk intake varied by region (rural, suburban, and urban: 9, 36, and 83 g/day, respectively, p < 0.0005) as did the proportion of milk consumers in the three areas (30%, 64%, and 91%, p < 0.0005). Bone mineral density (BMD) at the distal one-third and one-tenth radius and one-tenth ulna was positively associated with milk consumption (p < 0.05). Multiple regression analysis of BMC on foods and nutrients as well as confounding factors, including weight, bone age, Tanner stage, and School Physical Activity Score (SPAS), showed that milk intake was the only dietary factor included in the models for BMC at the four bone sites measured. The model explained 54%– 65% of the variation in BMC, and milk alone accounted for up to 3.2% of the variation. Milk was the only food group with significant partial correlation with BMC. SPAS, weight, bone age, and Tanner stage each accounted for a smaller variation in BMC (<1.8%). The results indicate that milk (presumably as an integrated source of nutrients) had a beneficial effect on bone mass of Beijing pubertal girls and was a better nutritional determinant of BMC than intake of any milk nutrient alone. Promotion of milk consumption should be considered for achieving optimal bone mass in this population group. (Bone 30:521–528; 2002) © 2002 by Elsevier Science Inc. All rights reserved.
Introduction Nutrition is known to be one of the environmental determinants of bone mineral content (BMC)/bone mineral density (BMD). Although it has a less striking role than genetic factors, diet is a factor with the potential for modification to achieve a maximum peak bone mass that could lead to a reduced osteoporotic fracture risk later in life. Unlike calcium intake, milk consumption has rarely been studied in relation to bone mineral status in children and adolescents. Two cross-sectional studies among white adolescents showed inconsistent results in relation to dairy calcium and bone mineral.19,29 Evidence from two milk supplementation trials showed that increased milk or dairy products consumption enhanced bone mineral acquisition in white adolescent girls aged 11–12 years.6,7 In China, average calcium intakes are low at 400 mg/capita per day (about 50% of the Chinese RDA), according to a 1992 national nutrition survey,27 with ⬃50% from plant-based foods.15 Cross-sectional studies on dietary calcium and bone mineral measures among Chinese children and adolescents have shown inconsistent results, whereas Cheng et al.8,9 reported that baseline distal BMC, lumbar spine BMD, and bone mineral accretion were not influenced by calcium intake in a 3 year prospective study of a small group of 179 Hong Kong boys and girls aged 12–13 years at baseline. On the other hand, studies of calcium intake and bone mineral measurements in Chinese children and adolescents conducted by Lee et al.,20,21 including a calcium supplementation trial, provided evidence of the positive effects of calcium intake on bone acquisition. No comprehensive research concerning nutrition and bone health has been conducted in Chinese children and adolescents. The purpose of this cross-sectional study was to investigate the relationship of intakes of foods (segregated into 13 groups) and more than 20 nutrients to bone mineral status in a random sample of Chinese adolescent girls. Materials and Methods Subjects A random sample of 649 girls, aged 12–14 years (mean ⫾ SD: 12.9 ⫾ 0.6 years), was drawn from a sample of 1277 girls from 13 middle schools in the Beijing area that had been selected by
Address for correspondence and reprints: Dr. Xueqin Du, Department of Animal Science, University of Sydney, Sydney, NSW 2006, Australia. E-mail: [email protected] © 2002 by Elsevier Science Inc. All rights reserved.
521
8756-3282/02/$22.00 PII S8756-3282(01)00698-6
522
X. Q. Du et al. Milk consumption and BMC in adolescent girls
Bone Vol. 30, No. 3 March 2002:521–528
means of a socioeconomically stratified, systematic, clustersampling procedure. Subjects had no evidence of liver, kidney, or other disorders that may have caused abnormal bone metabolism. As an important confounding variable, socioeconomic areas (urban, suburban, and rural) were defined by family income, parents’ occupation, location, and infrastructure. The study was conducted in September 1995, with written approval from the committee on experimental procedures involving human subjects of the University of New South Wales, and written consent from parents of all participants.
mated as 20:80 and 1:99, respectively.25 BMD was derived by dividing BMC by BW. The precision of the analyzer was ⬍2%. The short-term precision in vivo was 2.2% with intermediate repositioning. The correlation coefficient between the results from the SPA analyzer and a dual-energy X-ray absorptiometry (DCS-600EX, Aloka Co., Ltd., Japan) (precision ⬍1%) was 0.603 in a small group of subjects. The SPA analyzer was calibrated against an aluminum alloy model prior to use each day and at each project site. The measurements were performed by the same two technicians throughout.
Estimation of Dietary Intakes
Other Measurements
Habitual food and nutrient intakes over the past year were estimated by use of a specially designed and validated semiquantitative food frequency questionnaire (SFFQ) based on the methods of Block et al.5 The 103 item food list represented 86% of the calcium intake of Beijing residents as determined by the 1992 National Nutrition Survey.3 Particular foods rich in calcium, protein, and energy and observed to be popular among Beijing teenage girls were added to the list, whereas oil, which would be difficult for the girls to quantify, was not included in the questionnaire, but 22.4 g/day oil (the average intake of a Chinese adult in 1992) was added later for each subject for adjustment of fat intake. Chinese measures (bowl and spoons, standard size) and a set of food pictures were used to quantify food items. The SFFQ was self-administered at school by subjects under the supervision of the first author, with a verification interview in some cases. Food groups (cereals, milk, vegetables, fruits, legumes, seafood, ice confections, nuts, eggs, meat, and sugar) and 24 nutrients were calculated from the Chinese food tables,18 except for vitamin D values (not available in the Chinese food tables), which were obtained from UK food composition tables.16 A data entry and nutrient calculation program named CAVD,14 using EPI INFO (version 6, WHO/CDC, Atlanta, GA) was developed with the Information Centre, Chinese Academy of Preventive Medicine. The average frequency of consumption of each food item over the past year, in specified serving sizes, was indicated by marking one of ten frequency categories. A validity study of the SFFQ was conducted in a random subsample of 221 girls by comparison with a 6 day dietary recall (two 3 day recalls 6 months apart for seasonal differences). Results showed that correlation coefficients of calcium, vitamin D, protein, energy, and phosphorus were 0.62, 0.75, 0.53, 0.34, and 0.50, respectively, similar to those obtained for food frequency questionnaires developed by other researchers.15 The validity study showed that the questionnaire overestimated food and nutrient intakes by about 20%. The actual food and nutrient intakes were therefore calculated by the first author by adjusting the results from the SFFQ down by ⬇20%.11 This was done by means of an equation for each nutrient:
Body weight was measured by lever scales (RGT-140, Beijing) calibrated with a standard weight before use. Subjects were weighed with light clothes and without shoes. Height was measured by standing and sitting height measures (TG-III, Beijing). Date of menarche was recorded. Breast and pubic hair development were assessed by means of Tanner stage.1,26 Physical activity was assessed in two ways. The School Physical Activity Score (SPAS) for the past year was recorded because Beijing high school students spend most of the daylight hours (about 10) at school. This score is a comprehensive evaluation (1– 4, where 1 ⫽ excellent and most active), given once per semester by school physical education teachers, for performance and activity in a variety of sports and physical activities at school such as running, jumping, rope-skipping, throwing, and ball-games. Spare-time physical activities (PAs) were recorded by means of a self-administered questionnaire in minutes per week over the past year. There were ⬎15 categories of PAs during class breaks and time out of school, including walking, running, cycling, swimming, skating, table-tennis, basketball, volleyball, soccer, softball, high jump, long jump, rope-skipping, skipping, shuttlecock-kicking, and others. Bone age was estimated by the Greulich and Pyle Atlas method.12 Bone age (to the nearest 3 months) was determined by assessing development stages of metacarpals, phalanges, and carpals from a hand and wrist X-ray against a set of original photographs for the method, provided by a technician from the Beijing Anthropometry Centre. Ultraviolet (UV) exposure was measured to obtain indirect information on vitamin D supply. A semiquantitative personal UV dosimeter (John Thacher & Associates Pty, Ltd., Hong Kong)2,22 validated by the Australian Photobiology Testing Facility at the University of Sydney (Greg Murphy, 1995, personal communication) was used to obtain an average daily exposure dose to UV B waveband, 280 –320 nm (millijoules per square centimeter), required for the effective production of vitamin D in skin. It was attached to the top of the shoulder, outside the clothes, for a whole day and five badges were used for 5 consecutive days (3 weekdays and 2 weekend days). Each badge was compared against a reference comparator series of hues to obtain the exposure for each subject. The major use of the badge in this study was for comparison of UV exposure between groups and between seasons and not for absolute values. The measurements were made in early October 1995 for summer exposure and in mid-January 1996 for winter. Beijing is located at 40°N latitude.
It ⫽ a ⫹ b ⫻ Is where It is true intake, Is is intake from the SFFQ, and a and b are constants from regression analysis of data from the subsample of 221 girls.
Statistical Analysis Bone Mineral Measurement BMC and bone width (BW), at the distal one-third and one-tenth radius and ulna of the nondominant forearms were measured by means of a portable bone mineral analyzer (BMD-4, Beijing Broadcast Technology Institute), utilizing single-photon absorptiometry (SPA). At the distal one-tenth and one-third (proximal) radius sites, the ratio of trabecular vs. cortical bone was esti-
Descriptive statistics were performed for all variables by socioeconomic area (rural, suburban, and urban) and by milk consumption group (no milk, milk consumption less than the median, and milk consumption greater than or equal to the median). One-way analysis of variance (ANOVA) and Bonferroni post hoc multiple comparisons were used for testing differences between means and chi-square for testing differences between
Bone Vol. 30, No. 3 March 2002:521–528
X. Q. Du et al. Milk consumption and BMC in adolescent girls
523
Table 1. Characteristics and dietary intakes of adolescent girls aged 12–14 years in the Beijing area in 1995 Rural (n ⫽ 200)
Suburban (n ⫽ 150)
Urban (n ⫽ 299)
Total (n ⫽ 649)
Age (yr) Height (cm) Weight (kg) Bone age (yr) SPAS PAs (min/wk) Menarcheal status (% yes) UV summer (mJ/cm2 UVB)e UV winter (mJ/cm2 UVB)
13.2 ⫾ 0.6 154.0 ⫾ 5.7 45.0 ⫾ 7.7 13.8 ⫾ 1.3 1.8 ⫾ 0.8 372 ⫾ 478 64.5 (57.7–71.3) 59.9 ⫾ 13.5 (50) 34.2 ⫾ 11.4 (90)
13.0 ⫾ 0.6a 154.3 ⫾ 5.8 45.8 ⫾ 8.5 13.8 ⫾ 1.4 2.0 ⫾ 0.8b 527 ⫾ 587 66.7 (59.0–74.4) 43.3 ⫾ 22.3a (50) 15.9 ⫾ 4.0a (95)
12.7 ⫾ 0.4a,c 154.3 ⫾ 7.2 47.0 ⫾ 10.5 13.7 ⫾ 1.4 1.8 ⫾ 0.7c 698 ⫾ 700a,d 65.6 (60.1–71.1) 36.0 ⫾ 11.4a,d (111) 15.0 ⫾ 5.1a (190)
12.9 ⫾ 0.6 154.2 ⫾ 6.4 46.1 ⫾ 9.3 13.8 ⫾ 1.4 1.8 ⫾ 0.8 557 ⫾ 627 65.5 (61.8–69.2) 43.4 ⫾ 17.9 (211) 19.8 ⫾ 10.7 (375)
Dietary intake Milk in whole sample (g/day) Milk in consumers only (g/day) Dairy calcium (mg/day) As % of total calcium Calcium (mg/day) Vitamin D (g/day) Protein (g/day) Energy (kJ/day) Calcium: phosphorus ratio
9 ⫾ 31 32 ⫾ 51 18 ⫾ 49 5.4 ⫾ 13.4 314 ⫾ 91 0.56 ⫾ 0.55 47 ⫾ 9 6932 ⫾ 747 0.44 ⫾ 0.07
37 ⫾ 58a 57 ⫾ 64 46 ⫾ 67a 14.4 ⫾ 19.3a 349 ⫾ 89a 0.74 ⫾ 0.64 50 ⫾ 9a 6943 ⫾ 658 0.48 ⫾ 0.07a
83 ⫾ 74a,c 91 ⫾ 72a,c 120 ⫾ 100a,c 35.3 ⫾ 26.4a,c 388 ⫾ 94a,c 1.51 ⫾ 1.05a,c 51 ⫾ 9a 6953 ⫾ 658 0.51 ⫾ 0.09a,c
50 ⫾ 68 75 ⫾ 71 72 ⫾ 92 21.2 ⫾ 25.3 356 ⫾ 97 1.04 ⫾ 0.94 50 ⫾ 9 6944 ⫾ 686 0.48 ⫾ 0.09
Characteristic
Values are mean ⫾ SD or mean (95% confidence interval). Number of subjects for whom data were available in parentheses. KEY: SPAS, School Physical Activity Score (score 1– 4, where 1 ⫽ best performance and most active); PAs, spare-time physical activities; UV, ultraviolet; UVB, ultraviolet B waveband (about 280 –320 nm). a p ⬍ 0.01, bp ⬍ 0.05 (vs. rural); cp ⬍ 0.01, dp ⬍ 0.05 (vs. suburban); esignificant main effect of season, p ⬍ 0.0005.
percentage data. Bivariate and partial correlation analyses were conducted to identify variables associated with bone mineral measurements. Multiple regression analyses of BMC (for the entire sample and for each area) on possible contributing factors were performed to examine predictors of BMC and their relative importance. Factors including food and nutrient intakes, physical activity, and pubertal maturation that had a significant correlation with BMC were included in the analysis, with adjustment of bone size variables of BW, body weight, and height.24 All data were regarded as statistically significant at the 5% level (two-tailed). SPSS software was used for the analysis (SPSS, Inc., Chicago IL). Results The characteristics and dietary intakes of 649 Beijing girls by socioeconomic area are shown in Table 1. Compared with rural girls, urban girls were slightly younger. The UV dose in winter was significantly lower than in summer (repeated-measures ANOVA, p ⬍ 0.0005). UV exposure across each season, was greatest in rural ⬎ suburban ⬎ urban girls. Height, weight, and bone age were comparable between areas. Rates of Tanner breast development stages 2, 3, and 4 were 11.9%, 39.5%, and 46.3%, respectively, with 0.7% of girls at stage 1. There were no differences in Tanner stage and menarche status among girls in the three areas. Girls in suburban areas had poorer SPAS scores than their counterparts in urban and rural areas. Rural and suburban girls had lower PAs than urban girls. Bone Mineral Measurements Bone mineral density by milk consumption (three groups: nomilk; low-milk [less than median milk intake, 22 ⫾ 18 g/day]; and high-milk [greater than or equal to median milk intake, 128 ⫾ 65 g/day]) are presented as mean ⫾ SD in Table 2. Significant differences in BMD among the groups were achieved on three of the four forearm sites measured (p ⬍ 0.01). The
no-milk consumers had the lowest BMD, whereas the high-milk group had the highest BMD. Compared with the no-milk group, the percent increase in BMD at the four sites in milk consumers averaged 6.0% (range 2.5%–10.5%), with a trend of higher percent increase in the high-milk group than in the low-milk group. Differences in BMC and BW among the three groups were otherwise not significant, apart from the BMC at one-tenth radius (p ⬍ 0.05). There were no significant differences in height, Tanner stage, and SPAS between/among the three milk consumption groups. Weight and bone age, which are known to be positively related to BMD, were lower in the high-milk group than in the low-milk group (p ⬍ 0.05), making the BMD differences even more striking. Intake of nondairy calcium was the same in the no-milk and the low-milk groups. The high-milk group had an even lower intake of calcium from nondairy products than the other two groups. Nondairy vitamin D intake, protein intake, and PAs were higher in the two milk-consuming groups than in the no-milk group, which may have contributed independently to their higher BMD. This point is further addressed in what follows. Factors Associated With Bone Mass in Multiple Regression Models Multiple regression analyses of BMC on possible contributing factors were performed for the 517 girls for whom complete data were available. Ten variables correlated significantly with BMC and were included as independent variables in the analysis of model 1 for each site of BMC (i.e., BW, body weight, height, bone age, Tanner stage of breast development, SPAS, calcium and vitamin D intakes, calcium:phosphate (Ca:P) ratio, and socioeconomic area). Spare-time physical activity was also included in the analysis although it did not show a correlation that achieved statistical significance. In model 2, the milk variable was added and analyzed together with the 11 variables from model 1 because only the milk group had significant partial correlation with BMC among the 13 food groups. Intakes of
524
X. Q. Du et al. Milk consumption and BMC in adolescent girls
Bone Vol. 30, No. 3 March 2002:521–528
Table 2. Characteristics, dietary intakes, and bone mineral measurements of adolescent girls aged 12–14 years in the Beijing area in 1995, by milk consumption No-milk group (n ⫽ 222) Age (yr) Height (cm) Weight (kg) Bone age (yr) SPAS PAs (min/wk) Tanner stage (%) 1 2 3 4 5 Menarcheal status (% yes) UV summer (mJ/cm2 UVB)e UV winter (mJ/cm2 UVB) Dietary intake Milk (g/day) Dairy calcium (mg/day) As % of total calcium Nondairy calcium (mg/day) Dairy vitamin D (g/day) Nondairy vitamin D (g/day) Dairy protein (g/day) Nondairy protein (g/day) Dairy energy (kJ/day) Nondairy energy (kJ/day) Bone mineral measurements Distal one-third radius BMC (g/cm) BMD (g/cm2) BW (cm) Distal one-third ulna BMC (g/cm) BMD (g/cm2) BW (cm) Distal one-tenth radius BMC (g/cm) BMD (g/cm2) BW (cm) Distal one-tenth ulna BMC (g/cm) BMD (g/cm2) BW (cm)
13.2 ⫾ 0.6 153.8 ⫾ 5.8 46.5 ⫾ 8.6 13.8 ⫾ 1.4 1.9 ⫾ 0.8 392 ⫾ 472
Low-milk group (n ⫽ 213) 12.8 ⫾ 0.5a 154.9 ⫾ 6.2 47.1 ⫾ 9.8 13.9 ⫾ 1.4 1.8 ⫾ 0.8 591 ⫾ 645a
1.0 10.3 41.2 46.1 1.5 66.7 (60.4–73.0) 49.0 ⫾ 15.7 (76) 26.1 ⫾ 12.7 (112)
1.1 13.6 36.7 46.9 1.7 69.0 (62.7–75.3) 37.7 ⫾ 13.2a (71) 17.5 ⫾ 8.6a (123)
High-milk group (n ⫽ 299) 12.8 ⫾ 0.5a 153.9 ⫾ 7.2 44.8 ⫾ 9.2d 13.6 ⫾ 1.3d 1.8 ⫾ 0.8 698 ⫾ 710a 0 12.1 40.2 46.0 1.7 60.7 (54.0–67.4) 43.1 ⫾ 22.5 (64) 16.8 ⫾ 8.1a (140)
Total (n ⫽ 649) 12.9 ⫾ 0.6 154.2 ⫾ 6.4 46.1 ⫾ 9.3 13.8 ⫾ 1.4 1.8 ⫾ 0.8 557 ⫾ 627 0.7 11.9 39.5 46.3 1.6 65.5 (61.8–69.2) 43.4 ⫾ 17.9 (211) 19.8 ⫾ 10.7 (375)
0 0 0 293.4 ⫾ 77.0 0 0.53 ⫾ 0.59 0 46.0 ⫾ 8.3 0 6794 ⫾ 703
22.4 ⫾ 18.9a 40.8 ⫾ 37.0a 11.8 ⫾ 10.4a 299.4 ⫾ 75.3 0.12 ⫾ 0.22 0.72 ⫾ 0.55a 1.0 ⫾ 1.0a 50.0 ⫾ 9.0a 98.7 ⫾ 96.0a 6894 ⫾ 672
128.2 ⫾ 65.0a,c 176.4 ⫾ 84.7a,c 39.8 ⫾ 17.1a,c 259.9 ⫾ 79.9a,c 1.21 ⫾ 1.13a,c 0.55 ⫾ 0.58c 3.9 ⫾ 1.8a,c 48.6 ⫾ 8.0a 330.4 ⫾ 151.7a,c 6721 ⫾ 681d
49.6 ⫾ 68.1 71.6 ⫾ 92.2 17.0 ⫾ 20.3 284.3 ⫾ 79.2 0.44 ⫾ 0.86 0.60 ⫾ 0.58 1.6 ⫾ 2.0 48.1 ⫾ 8.6 141.4 ⫾ 172.7 6803 ⫾ 688
0.684 ⫾ 0.134 0.606 ⫾ 0.081 1.123 ⫾ 0.141
0.711 ⫾ 0.189 0.633 ⫾ 0.097b 1.116 ⫾ 0.213
0.708 ⫾ 0.218 0.642 ⫾ 0.114a 1.093 ⫾ 0.176
0.701 ⫾ 0.183 0.627 ⫾ 0.099 1.111 ⫾ 0.179
0.589 ⫾ 0.125 0.597 ⫾ 0.087 0.979 ⫾ 0.125
0.593 ⫾ 0.146 0.612 ⫾ 0.110 0.962 ⫾ 0.131
0.596 ⫾ 0.150 0.614 ⫾ 0.112 0.962 ⫾ 0.134
0.593 ⫾ 0.140 0.608 ⫾ 0.104 0.968 ⫾ 0.130
0.554 ⫾ 0.178 0.352 ⫾ 0.065 1.543 ⫾ 0.309
0.620 ⫾ 0.322b 0.379 ⫾ 0.113b 1.567 ⫾ 0.423
0.596 ⫾ 0.259 0.389 ⫾ 0.118a 1.499 ⫾ 0.312
0.590 ⫾ 0.260 0.373 ⫾ 0.102 1.537 ⫾ 0.352
0.335 ⫾ 0.098 0.372 ⫾ 0.062 0.880 ⫾ 0.181
0.361 ⫾ 0.162 0.393 ⫾ 0.103 0.886 ⫾ 0.242
0.359 ⫾ 0.161 0.403 ⫾ 0.115a 0.859 ⫾ 0.187
0.351 ⫾ 0.143 0.389 ⫾ 0.096 0.875 ⫾ 0.205
Values are mean ⫾ SD or mean (95% confidence interval). Number of subjects for whom data were available in parentheses. KEY: BMC, bone mineral content; BMD, bone mineral density; BW, bone width; SPAS, school physical activity score (score 1– 4, where 1 ⫽ best performance and most active); PAs, spare-time physical activities; UV, ultraviolet; UVB, ultraviolet B waveband (about 280 –320 nm). a p ⬍ 0.01, bp ⬍ 0.05 (vs. no milk); cp ⬍ 0.01, dp ⬍ 0.05 (vs. low milk); esignificant main effect of season, p ⬍ 0.0005.
protein, nondairy protein, and nondairy vitamin D showed no correlation with bone measurements and therefore were not included in the analysis. The eight models are summarized as follows and the four models with milk added are described in Table 3: Y
Y
Analyzed together with the other variables, height, calcium intake Ca:P ratio, and PAs did not meet the criteria for inclusion in any of the eight models. BW, milk intake, vitamin D intake, bone age, body weight, SPAS, Tanner stage of breast development, and socioeconomic area were predictors or determinants of BMC. About 60% of the variation in BMC (ranging 53.5%– 64.7%) could be explained by these models.
Y
Y
Y
Among the independent variables, BW was the best predictor included in all eight models and explained most of the variation in BMC. SPAS was selected into seven of the eight models, indicating that it was a determinant of BMC, in spite of the small effect on BMC (associated with up to 1.1% change in R2). The negative regression coefficient between the SPAS and BMC was due to the coding of SPAS (1– 4, where 1 ⫽ excellent and most active). The importance of bone age, weight, and Tanner stage appeared to depend on bone site measured (distal one-third or one-tenth of forearm; radius or ulna). Weight predicted only BMC of distal one-third sites where bone is mainly cortical,
Bone Vol. 30, No. 3 March 2002:521–528
X. Q. Du et al. Milk consumption and BMC in adolescent girls
525
Table 3. Multiple regression models for BMC of 517 adolescent girls aged 12–14 years in the Beijing areaa Regression coefficient
Standard error
Standardized coefficient
BMC at distal one-tenth radius Model 1: MR ⫽ 0.770, R2 ⫽ 0.593, adjusted R2 ⫽ 0.589, SE ⫽ 0.167 BW (mm) Milk (g/day) Bone age (yr) Socioeconomic area SPAS (Constant)
0.050 0.0004 0.025 0.026 ⫺0.020 ⫺0.557
0.002 0.00009 0.006 0.010 0.010 0.081
BMC at distal one-third radius Model 2: MR ⫽ 0.805, R2 ⫽ 0.647, adjusted R2 ⫽ 0.643, SE ⫽ 0.110 BW (mm) Milk (g/day) Bone age (yr) SPAS Socioeconomic area Weight (kg) (Constant)
0.073 0.0003 0.011 ⫺0.025 0.019 0.001 ⫺0.340
BMC at distal one-tenth ulna Model 3: MR ⫽ 0.770, R2 ⫽ 0.593, adjusted R2 ⫽ 0.591, SE ⫽ 0.093 BW (mm) Milk (g/day) Tanner stage, breast (Constant) BMC at distal one-third ulna Model 4: MR ⫽ 0.732, R2 ⫽ 0.536, adjusted R2 ⫽ 0.532, SE ⫽ 0.096 BW (mm) Weight (kg) Milk (g/day) SPAS (Constant)
p value
Partial R2
0.695 0.128 0.128 0.090 ⫺0.059
0.000 0.000 0.000 0.007 0.037 0.000
0.539b 0.030 0.014 0.006 0.003
0.003 0.00006 0.004 0.006 0.006 0.001 0.056
0.725 0.127 0.081 ⫺0.105 0.096 0.070
0.000 0.000 0.001 0.000 0.002 0.022 0.000
0.582b 0.032 0.011 0.010 0.008 0.004
0.053 0.0003 0.014 ⫺0.175
0.002 0.00005 0.006 0.024
0.738 0.172 0.073
0.000 0.000 0.011 0.000
0.557b 0.031 0.005
0.070 0.002 0.0002 ⫺0.018 ⫺0.166
0.003 0.001 0.00005 0.006 0.036
0.654 0.147 0.110 ⫺0.099
0.000 0.000 0.000 0.001 0.000
0.495b 0.018 0.014 0.010
KEY: BMC, bone mineral content; BW, bone width; MR, multiple R, square root of coefficient of determination; R2, coefficient of determination; SE, standard error; SPAS, School Physical Activity Score (score 1– 4, where 1 ⫽ most active and best performance). a Stepwise method, entry criteria: p ⬍ 0.05. b Together with constant.
Y
Y
Y
whereas Tanner staging was associated with BMC of distal one-tenth sites where bone is a mixture of cortical and trabecular and bone turnover is faster. Bone age was only included into the four models for the radius. These factors were also associated with only a very small change in R2 (0.3%–1.8%). After BW, socioeconomic area accounted for the variation in BMC of up to 2.8%, with inclusion in five of the eight models. Further analysis of BMC models, stratified by socioeconomic area, is therefore presented subsequently. Vitamin D intake was an important determinant and predictor of BMC, accounting for up to 1.9% of the variation in BMC. However, milk intake was the only nutrition variable included in the model, accounting for up to 3.2% of the change in R2. This suggests milk intake was a more important determinant of BMC than vitamin D alone. When analyzed with body size and other confounding variables,
the relative importance of three milk nutrients to BMC was in the order milk vitamin D ⬎ milk calcium ⬎ milk protein, given 3.5%, 3.3%, and 2.8% changes in R2, respectively. Multiple linear regression analysis of BMC at each site of the forearm when stratified by socioeconomic area yielded a total of 12 models (Table 4). Calcium, Ca:P ratio, and PAs were not included into any of the models; BW was presented in all models; milk, SPAS, weight, and bone age were included in at least four models; predictors of BMC for each socioeconomic area varied as milk and SPAS were determinants of BMC in urban girls, whereas weight and bone age were determinants in rural and suburban girls. Milk and SPAS explained the variation of BMC up to 2.3% and 2.6%, respectively. Milk was not included in six models of the distal one-third radius and ulna BMC for rural and suburban girls, nor in the model of distal one-third ulna BMC for urban girls.
526
X. Q. Du et al. Milk consumption and BMC in adolescent girls
Bone Vol. 30, No. 3 March 2002:521–528
Table 4. Summary of multiple regression models of BMC at four sites of the forearm in Beijing girls aged 12–14 years, stratified by socioeconomic area BMC models Rural
Suburban
Urban
Independent variable
R10
R3
U10
U3
R10
R3
U10
U3
R10
R3
U10
U3
BW (mm) Milk intake (g/day) SPAS Weight (kg) Bone age (yr) Height (cm) Tanner stage VitD intake (g/day) R2 MR
⫹a
⫹
⫹ ⫹
⫹
⫹ ⫹
⫹
⫹
⫹
⫹ ⫹
⫹ ⫹ ⫹
⫹ ⫹ ⫹
⫹
⫹
⫹ ⫹ ⫹
⫹
⫹
0.70 0.84
0.67 0.82
0.57 0.75
0.65 0.81
0.57 0.76
⫹ 0.48 0.69
⫹ ⫹ ⫹ 0.60 0.77
⫹ ⫹ ⫹
0.70 0.84
⫹ ⫹ 0.76 0.87
⫹ ⫹
⫹
0.58 0.76
0.61 0.78
0.57 0.76
Partial R2 range 0.404–0.672b 0.007–0.023 0.008–0.026 0.043–0.139 0.009–0.042 0.010–0.030 0.009 0.012
KEY: BMC, bone mineral content; R2, coefficient of determination; MR, Multiple R, square root of R2; R10, one-tenth radius site; R3, one-third radius site; U10, one-tenth ulna site; U3, one-third ulna site; VitD, vitamin D. a ⫹ indicates variable is included in the model. b Together with constant.
Milk Consumption and Nutrient Intakes Cereals and vegetables formed a large proportion of the total food intake of Beijing girls. Compared with rural girls, urban girls consumed less cereal and vegetables but more animal foods (milk, seafood, and eggs) as well as more pastry, ice confections, fruits, and sugar (p ⬍ 0.0005), whereas consumption of legumes and nuts was similar. Suburban girls had the highest consumption of ice confections with all other food intakes in between rural and urban rates. Milk consumption data are presented in Tables 1 and 5. The milk group included fresh milk (pasteurized), powdered milk, vitamin D-fortified milk (15 g fortified vitamin D per liter, pasteurized), and yogurt. Sixty-six percent of girls consumed milk and milk products at an average of 75 g/day. The average level was 50 g/day for the sample as a whole, including non-milk consumers (about one-third of girls consumed no milk at all). The proportion of non-milk consumers was much higher in the rural (70%) than in the urban area (9%). As shown in Table 5, there were few problems with purchasing power or commercial supply of milk; however, milk consumption may have been affected by perceptions about milk, and food habits. Half of the girls reported that they did not like drinking milk and 5% girls reported that milk was not available at home. There were area differences in terms of milk perceptions and milk supply. The proportion reporting dislike of milk was the highest in rural girls. No problems in terms of local milk supply were reported in the urban setting. Feeling uncomfortable after drinking milk was reported at an average rate of 18%, ranging from 12% in the urban area to 29% in the rural area. The “uncomfortable feelings”
referred to stomach upset, cramps, bloating, and diarrhea, but may have included other minor negative feelings. The average intakes of calcium and vitamin D were 356 mg/day and 1.04 g/day, respectively, accounting for 30% and 10% of the Chinese recommended dietary allowance (RDA) for this age group.10 The protein intake of 50 g/day was also inadequate at 63% of the RDA of 80 g/day for a plant-based diet. The energy intake was 72% of the RDA. The Ca:P ratio was much lower than the level of 1.0 recommended in the U.S. RDAs.23 Calcium, vitamin D, and protein intakes and Ca:P ratio were higher in urban ⬎ suburban ⬎ rural girls. Table 6 shows the source of dietary calcium and vitamin D by food groups. In total, 58.9% of calcium was from plant foods, whereas 41.1% was from animal foods. About 20% of calcium was from the milk group. Ice cream and ice confections contributed 6.6% of total calcium. Dietary vitamin D was mainly from eggs (52.6%) and milk (35.4%) as well as pastry (11.4%) (traditionally made with eggs). The sources of calcium and vitamin D varied by socioeconomic area due to differences in food patterns (see earlier). The percentage of milk calcium in total calcium was 5%, 14%, and 35%, respectively, in rural, suburban, and urban areas (p ⬍ 0.01, Table 1). Discussion The positive association of BMD with level of milk consumption, the fact that no other dietary variables in this study, other than milk intake, were able to explain the difference in BMD (Table 2, the differences in BMD between milk groups were still
Table 5. Milk consumers, access to milk, and attitudes toward milk consumption in adolescent girls aged 12–14 years in the Beijing area in 1995
Item Milk consumers (%) Dislike drinking milk (%) Feeling uncomfortable after drinking milk (%)a No milk provided by parents (%) Not affordable (%) No milk supply locally (%)
Rural (n ⫽ 200)
Suburban (n ⫽ 150)
Urban (n ⫽ 299)
Total (n ⫽ 649)
29.5 63.6 29.0 6.2 6.8 8.2
64.0 54.8 16.4 6.3 1.6 2.4
91.0 39.9 12.1 3.1 2.7 0
65.8 50.6 18.0 4.7 3.6 2.8
p ⬍ 0.0005 (chi-square test) among girls in the three areas for each listed item. a “Uncomfortable” referred to symptoms such as stomach upset, cramps, bloating, and diarrhea, or any minor abnormal feeling.
Bone Vol. 30, No. 3 March 2002:521–528
X. Q. Du et al. Milk consumption and BMC in adolescent girls
527
Table 6. Sources of calcium and vitamin D for adolescent girls aged 12–14 years in the Beijing area in 1995a Food group Cereals and products Milk and milk products Vegetables Legume Seafood Ice cream and ice confections
Ca (%) 29.4 19.9 11.0 8.5 6.8 6.6
VitD (%) c
11.4 35.4 0 0 0 0.1
Food group
Ca (%)
VitD (%)
Nuts Eggs Meat and meat products Fruits Chocolate and sugar Total
6.1 4.5 3.3 2.8 1.1 100.0
0 52.6 0.5 0 0 100.0
KEY: Ca, calcium; VitD, vitamin D. Data from the whole sample of Beijing girls (n ⫽ 1248) from which subjects in present study were drawn. From pastry.
a
b
present after stratification by Tanner stage [data not presented]), and the fact that milk intake was the only dietary determinant of forearm BMC (Tables 3 and 4) all suggest a favorable effect of milk consumption on bone mass acquisition in Beijing pubertal girls. No comparable studies of Asian children have been reported; however, the results are comparable with a cross-sectional study in Icelandic white girls aged 13 and 15 years,19 in which calcium intake from dairy products was found to be significantly correlated with distal and ultradistal BMD in the forearm of the 15-year-old group after adjustment for menarcheal age and weight. The researchers also reported that the correlation coefficients between BMD and calcium from dairy products were significantly greater within the lowest tertile of ⬍800 mg/day. These Icelandic girls had a much higher dairy calcium intake (1082 mg/day) than our subjects (70 mg/day). In contrast, no association was found between calcium intake from dairy products (1200 –1300 mg/day) and BMC at diaphyseal and metaphyseal sites in 1359 Dutch boys and girls aged 7–11 years.29 The investigators suggested that this result might be explained by the fact that about 80% of the children in the study had calcium intakes around 1400 mg/day, which is probably a ceiling value for children of this age. The effect of milk on BMC/BMD was also indicated by our finding that nutrients such as calcium, vitamin D, and protein from milk correlated well with BMC/BMD, whereas these nutrients from nondairy sources were not correlated with the bone measurements. This is consistent with a finding in Chinese adult women that BMC and BMD were correlated positively with total calcium, dairy calcium, and to a lesser extent with nondairy calcium.17 The positive effect of milk on bone mineral is probably an integrated consequence of its provision of several nutrients, particularly vitamin D, calcium, and protein, because all of these showed positive correlations with the bone measurements. The results may also be due to growth factors known to be present in milk.4,28 Furthermore, the relative importance of the three milk nutrients to BMC was milk vitamin D ⬎ milk calcium ⬎ milk protein, as suggested by their contributions to R2 in the multiple regression equations. One major concern about milk consumption promotion in Chinese children by Western experts is the high lactose intolerance rate. Yang et al. reported that incidence of lactose intolerance, after taking 25 g of lactose, was 30.5% in urban children aged 11–13 years, with a lower rate of 14% in children with intolerance symptoms after consuming 50 g of milk powder (12 g of lactose).30 Incidence of feeling uncomfortable after drinking milk was reported to be 18% in the present study (Table 2), with values of 12% in urban, 16% in suburban, and 29% in rural areas (p ⬍ 0.0005). This rate is similar to Yang’s result with the 50 g milk powder test and to the 17.8% rate in Hong Kong Chinese women aged 21– 40 years with symptoms of lactose intolerance, as reported by Ho et al.15 Ability to tolerate milk is being further tested in our current 2 year milk supplementation trial in which
500 urban Beijing girls aged 11 years consumed 330 mL of milk per school day for 2 years.31 Physical activity (PA) expressed as SPAS was another determinant of BMC identified for Beijing adolescent girls in this study. SPAS was included in seven of the eight final models with a small contribution to the variation in BMC (up to 1.1% change in R2), suggesting that SPAS was an independent determinant of BMC. There have been no studies using similar school PA scores in children and adolescents. On the other hand, the result that spare-time PA was not a predictor of BMC from this study is consistent with data of Gunnes et al.,13 who found that weightbearing PA was not a predictor of forearm BMD gain in 150 Norwegian girls aged ⱖ11 years, and those of Cheng et al.9 whose 3 year longitudinal study of a small sample of 179 Hong Kong boys and girls aged 12–16 years showed that bone mineral accretion was not influenced by exercise, or level of physical fitness. The SPAS used in this study is a comprehensive rating of each subject’s performance over the past year in a variety of school physical education activities and as such may reflect innate ability, but may also be assumed to be positively associated with the activity level of the girls, whereas PAs quantitatively accounted for all physical activities conducted during class breaks and out-of-school hours. Physically demanding activities after school (e.g., assistance in farm work) was not included in the questionnaire, because this kind of activity was assumed to be rare in young girls given the long school hours and the protection afforded to the sole child in Chinese families. The role of PA assessed in a variety of ways (type, frequency, and intensity of activity; physical fitness; performance level) as a determinant of bone mineral status in this population needs further study. As a cross-sectional study of this nature has limitations, further studies (e.g., milk intervention trials) are needed to provide more definitive evidence of the effects of milk on bone mass in the Chinese adolescent population. Because few problems appeared to be associated with milk drinking in this crosssectional study, such trials may be feasible in this group.
Acknowledgments: The authors thank Dr. D. Mackerras for her advice on dietary survey methods and Professor A. Baumann for his advice on sample size. The authors are indebted to Professor G. S. Ma, X. W. Li, G. S. Liu, X. X. Li, J. J. Tan, J. M. Zhu and his staff, Dr. L. Huang, B. M. Guo, J. X. Gao, H. Wang, Dr. X. J. Shi, Dr. Z. H. Liu, and E. Emmerson for their help and technical assistance with the project. The authors are grateful to all principals, staff, and health workers involved for their effort and contribution to the fieldwork and all subjects and their parents for their cooperation. This research was supported in part by the Dairy Research and Development Corporation, Australia.
528
X. Q. Du et al. Milk consumption and BMC in adolescent girls
Bone Vol. 30, No. 3 March 2002:521–528
References 1. Adelaide Children’s Hospital (ACH). Stages of Puberty. Sydney: Pharmacia & Upjohn; 1989. 2. Bannard, J. and Maguire, D. Suncheck patches and monitoring of UVR exposure. Br J Dermatol 131:911–912; 1994. 3. Beijing Institute of Food Inspection and Examination (BIFIE). Data on Dietary Survey Among Beijing Residents in 1992. Beijing: BIFIE; 1992. 4. Belford, D. A., Rogers, M. L., Francis, G. L., Payne, C., Ballard, F. J., and Goddard, C. Platelet-derived growth factor, insulin-like growth factors, fibroblast growth factors and transforming growth factor beta do not account for the cell growth activity present in bovine milk. J Endocrinol 154:45–55; 1997. 5. Block, G., Hartman, A. M., Dresser, C. M., Carroll, J. G., and Gardner, L. A data-based approach to diet questionnaire design and testing. Am J Epidemiol 124:453– 469; 1986. 6. Cadogan, J., Eastell, R., Jones, N., and Barker, M. E. Milk intake and bone mineral acquisition in adolescent girls: Randomised, controlled intervention trial. BMJ 315:1255–1260; 1997. 7. Chan, G. M., Hoffman, K., and McMurry, M. Effects of dairy products on bone and body composition in pubertal girls. J Pediatr 126:551–556; 1995. 8. Cheng, J. C. Y., Leung, S. S. S. F., Lee, W. T. K., Lau, T. F., Maffulli, N., Cheung, A. Y. K., and Chan, K. M. Determinants of axial and peripheral bone mass in Chinese adolescents. Arch Dis Child 78:524 –530; 1998. 9. Cheng, J. C., Maffulli, N., Leung, S. S., Lee, W. T., Lau, J. T., and Chan, K. M. Axial and peripheral bone mineral acquisition: A 3-year longitudinal study in Chinese adolescents. Eur J Pediatr 158:506 –512; 1999. 10. Chinese Nutrition Society (CNS). Recommended Dietary Allowance for Chinese. Beijing: CNS; 1988. 11. Du, X., Greenfield, H., Fraser, D. R., Ge, K., Trube, A., and Wang, Y. Vitamin D deficiency and associated factors in adolescent girls in Beijing. Am J Clin Nutr 74:494 –500; 2001. 12. Greulich, W. W. and Pyle, S. I. Radiographic Atlas of the Skeletal Development of the Hand and Wrist, 2nd Ed. Stanford, CA: Stanford University; 1959. 13. Gunnes, M. and Lehmann, E. H. Physical activity and dietary constituents as predictors of forearm cortical and trabecular bone gain in healthy children and adolescents: A prospective study. Acta Pediatr 85:19 –25; 1996. 14. He, W., Du, X., and Greenfield, H. CAVD, A Survey System Using EPI INFO. Beijing: Chinese Academy of Preventive Medicine; 1997 (software). 15. Ho, S. C., Leung, P. C., Swaminathan, R., Chan, C., Chan, S. S. G., Fan, Y. K., and Lindsay, R. Determinants of bone mass in Chinese women aged 21– 40 years. II. Pattern of dietary calcium intake and association with bone mineral density. Osteopors Int 4:167–175; 1994. 16. Holland, B., Welch, A. A., Unwin, I. D., Buss, D. H., Paul, A. A., and Southgate, D. A. T. McCance and Widdowson’s The Composition of Foods, 5th Ed. London: Royal Society of Chemistry and Ministry of Agriculture; 1991. 17. Hu, J. F., Zhao, X. H., Jia, J. B., Parpia, B., and Campbell, T. C. Dietary
18. 19.
20.
21.
22.
23. 24.
25. 26. 27. 28.
29.
30.
31.
calcium and bone density among middle-aged and elderly women in China. Am J Clin Nutr 58:219 –297; 1993. Institute of Nutrition and Food Hygiene (INFH). Food Composition Tables. Beijing: People’s Health; 1991. ¨ ., Valdimarsson, O ¨ ., Steingrimsdottir, L., and Sigurdsson, G. Kristinsson, J. O Relation between calcium intake, grip strength and bone mineral density in the forearms of girls aged 13 and 15. J Intern Med 236:385–390; 1994. Lee, W. T. K., Leung, S. S. F., Ng, M. Y., Wang, S. F., Xu, Y. C., Zeng, W. P., and Lau, J. Bone mineral content of two populations of Chinese children with different calcium intakes. Bone Miner 23:195–206; 1993. Lee, W. T. K., Leung, S. S. F., Wang, S. H., Xu, Y. C., Zeng, W. P., Lau, J., Oppenheimer, S. J., and Cheng, J. C. Y. Double-blind, controlled calcium supplementation and bone mineral accretion in children accustomed to a low-calcium diet. Am J Clin Nutr 60:744 –750; 1994. Moseley, H., Mackie, R. M., and Ferguson, J. The suitability of Suncheck patches and Tanscan cards for monitoring the sunburning effectiveness of sunlight. Br J Dermatol 128:75–78; 1993. National Research Council (NRC). Recommended Dietary Allowance, 10th Ed. Washington, DC: National Academy; 1989. Prentice, A., Parsons, T. J., and Cole, T. J. Uncritical use of bone mineral density in absorptiometry may lead to size-related artifacts in the identification of bone mineral determinants. Am J Clin Nutr 60:837– 842; 1994. Rosen, C. J. Osteoporosis: Diagnostic and Therapeutic Principles. Totowa, NJ: Humana; 1996. Tanner, J. M. Growth at Adolescence. Oxford: Blackwell; 1962. Technical Committee of National Nutrition Survey (TCNNS). Dietary Intakes of Chinese People. Beijing: Institute of Nutrition and Food Hygiene; 1994. Toba, Y., Takada, Y., Yamamura, J., Tanaka, M., Matsuoka, Y., Kawakami, H., Itabashi, A., Aoe, S., and Kumegawa, M. Milk basic protein: A novel protective function of milk against osteoporosis. Bone 27:403– 408; 2000. VandenBergh, M. F. Q., DeMan, S. A., Witteman, J. C. M., Hofman, A., Trouerbach, W. T., and Grobbee, D. Physical activity, calcium intake, and bone mineral content in children in The Netherlands. J Epidemiol Commun Health 49:299 –304; 1995. Yang, Y., He, M., Cui, H., Bian, L., Liu, J., Cheng, W., Xu, H., and Feng, W. Study on the incidence of lactose intolerance of children in China. J Hygiene Res 28(1):44 – 46; 1999. Zhu, K., Du, X., Greenfield, H., and Fraser, D. R. Milk supplementation and bone mineral acquisition in Chinese adolescent girls: Preliminary results after one year. Proc Nutr Soc Austral 24:123; 2000.
Date Received: July 18, 2001 Date Revised: October 4, 2001 Date Accepted: October 24, 2001
Milk Consumption and Bone Mineral Content in Chinese Adolescent Girls X. Q. DU,1 H. GREENFIELD,1 D. R. FRASER,2 K. Y. GE,3 Z. H. LIU,4 and W. HE3 1
Department of Food Science and Technology, University of New South Wales, Australia Department of Animal Science, University of Sydney, Sydney, Australia 3 Chinese Academy of Preventive Medicine, Beijing, China 4 Beijing China–Japan Friendship Hospital, Beijing, China 2
Key Words: Milk consumption; Calcium intake; Vitamin D intake; Chinese adolescents; Bone mineral content; Physical activity.
A cross-sectional study of a random sample of 649 girls, aged 12–14 years (mean ⴞ SD: 12.9 ⴞ 0.6 years), in the Beijing area examined the relationship between diet and bone mineral status. Food and nutrient intakes over the past year were estimated by means of a semiquantitative food frequency questionnaire. Bone mineral content (BMC) and bone width (BW) at the distal one-third and one-tenth radius and ulna were measured by single-photon absorptiometry. Results showed Beijing pubertal girls had a low mean milk consumption (fresh and powdered milk, vitamin D-fortified milk, and yogurt) at 50 g/day (95% confidence interval [CI] 44 –55 g/day whereas one-third consumed no milk at all. Mean calcium intake was 356 ⴞ 97 mg/day of which only 21% was provided by milk and milk products. Milk intake varied by region (rural, suburban, and urban: 9, 36, and 83 g/day, respectively, p < 0.0005) as did the proportion of milk consumers in the three areas (30%, 64%, and 91%, p < 0.0005). Bone mineral density (BMD) at the distal one-third and one-tenth radius and one-tenth ulna was positively associated with milk consumption (p < 0.05). Multiple regression analysis of BMC on foods and nutrients as well as confounding factors, including weight, bone age, Tanner stage, and School Physical Activity Score (SPAS), showed that milk intake was the only dietary factor included in the models for BMC at the four bone sites measured. The model explained 54%– 65% of the variation in BMC, and milk alone accounted for up to 3.2% of the variation. Milk was the only food group with significant partial correlation with BMC. SPAS, weight, bone age, and Tanner stage each accounted for a smaller variation in BMC (<1.8%). The results indicate that milk (presumably as an integrated source of nutrients) had a beneficial effect on bone mass of Beijing pubertal girls and was a better nutritional determinant of BMC than intake of any milk nutrient alone. Promotion of milk consumption should be considered for achieving optimal bone mass in this population group. (Bone 30:521–528; 2002) © 2002 by Elsevier Science Inc. All rights reserved.
Introduction Nutrition is known to be one of the environmental determinants of bone mineral content (BMC)/bone mineral density (BMD). Although it has a less striking role than genetic factors, diet is a factor with the potential for modification to achieve a maximum peak bone mass that could lead to a reduced osteoporotic fracture risk later in life. Unlike calcium intake, milk consumption has rarely been studied in relation to bone mineral status in children and adolescents. Two cross-sectional studies among white adolescents showed inconsistent results in relation to dairy calcium and bone mineral.19,29 Evidence from two milk supplementation trials showed that increased milk or dairy products consumption enhanced bone mineral acquisition in white adolescent girls aged 11–12 years.6,7 In China, average calcium intakes are low at 400 mg/capita per day (about 50% of the Chinese RDA), according to a 1992 national nutrition survey,27 with ⬃50% from plant-based foods.15 Cross-sectional studies on dietary calcium and bone mineral measures among Chinese children and adolescents have shown inconsistent results, whereas Cheng et al.8,9 reported that baseline distal BMC, lumbar spine BMD, and bone mineral accretion were not influenced by calcium intake in a 3 year prospective study of a small group of 179 Hong Kong boys and girls aged 12–13 years at baseline. On the other hand, studies of calcium intake and bone mineral measurements in Chinese children and adolescents conducted by Lee et al.,20,21 including a calcium supplementation trial, provided evidence of the positive effects of calcium intake on bone acquisition. No comprehensive research concerning nutrition and bone health has been conducted in Chinese children and adolescents. The purpose of this cross-sectional study was to investigate the relationship of intakes of foods (segregated into 13 groups) and more than 20 nutrients to bone mineral status in a random sample of Chinese adolescent girls. Materials and Methods Subjects A random sample of 649 girls, aged 12–14 years (mean ⫾ SD: 12.9 ⫾ 0.6 years), was drawn from a sample of 1277 girls from 13 middle schools in the Beijing area that had been selected by
Address for correspondence and reprints: Dr. Xueqin Du, Department of Animal Science, University of Sydney, Sydney, NSW 2006, Australia. E-mail: [email protected] © 2002 by Elsevier Science Inc. All rights reserved.
521
8756-3282/02/$22.00 PII S8756-3282(01)00698-6
522
X. Q. Du et al. Milk consumption and BMC in adolescent girls
Bone Vol. 30, No. 3 March 2002:521–528
means of a socioeconomically stratified, systematic, clustersampling procedure. Subjects had no evidence of liver, kidney, or other disorders that may have caused abnormal bone metabolism. As an important confounding variable, socioeconomic areas (urban, suburban, and rural) were defined by family income, parents’ occupation, location, and infrastructure. The study was conducted in September 1995, with written approval from the committee on experimental procedures involving human subjects of the University of New South Wales, and written consent from parents of all participants.
mated as 20:80 and 1:99, respectively.25 BMD was derived by dividing BMC by BW. The precision of the analyzer was ⬍2%. The short-term precision in vivo was 2.2% with intermediate repositioning. The correlation coefficient between the results from the SPA analyzer and a dual-energy X-ray absorptiometry (DCS-600EX, Aloka Co., Ltd., Japan) (precision ⬍1%) was 0.603 in a small group of subjects. The SPA analyzer was calibrated against an aluminum alloy model prior to use each day and at each project site. The measurements were performed by the same two technicians throughout.
Estimation of Dietary Intakes
Other Measurements
Habitual food and nutrient intakes over the past year were estimated by use of a specially designed and validated semiquantitative food frequency questionnaire (SFFQ) based on the methods of Block et al.5 The 103 item food list represented 86% of the calcium intake of Beijing residents as determined by the 1992 National Nutrition Survey.3 Particular foods rich in calcium, protein, and energy and observed to be popular among Beijing teenage girls were added to the list, whereas oil, which would be difficult for the girls to quantify, was not included in the questionnaire, but 22.4 g/day oil (the average intake of a Chinese adult in 1992) was added later for each subject for adjustment of fat intake. Chinese measures (bowl and spoons, standard size) and a set of food pictures were used to quantify food items. The SFFQ was self-administered at school by subjects under the supervision of the first author, with a verification interview in some cases. Food groups (cereals, milk, vegetables, fruits, legumes, seafood, ice confections, nuts, eggs, meat, and sugar) and 24 nutrients were calculated from the Chinese food tables,18 except for vitamin D values (not available in the Chinese food tables), which were obtained from UK food composition tables.16 A data entry and nutrient calculation program named CAVD,14 using EPI INFO (version 6, WHO/CDC, Atlanta, GA) was developed with the Information Centre, Chinese Academy of Preventive Medicine. The average frequency of consumption of each food item over the past year, in specified serving sizes, was indicated by marking one of ten frequency categories. A validity study of the SFFQ was conducted in a random subsample of 221 girls by comparison with a 6 day dietary recall (two 3 day recalls 6 months apart for seasonal differences). Results showed that correlation coefficients of calcium, vitamin D, protein, energy, and phosphorus were 0.62, 0.75, 0.53, 0.34, and 0.50, respectively, similar to those obtained for food frequency questionnaires developed by other researchers.15 The validity study showed that the questionnaire overestimated food and nutrient intakes by about 20%. The actual food and nutrient intakes were therefore calculated by the first author by adjusting the results from the SFFQ down by ⬇20%.11 This was done by means of an equation for each nutrient:
Body weight was measured by lever scales (RGT-140, Beijing) calibrated with a standard weight before use. Subjects were weighed with light clothes and without shoes. Height was measured by standing and sitting height measures (TG-III, Beijing). Date of menarche was recorded. Breast and pubic hair development were assessed by means of Tanner stage.1,26 Physical activity was assessed in two ways. The School Physical Activity Score (SPAS) for the past year was recorded because Beijing high school students spend most of the daylight hours (about 10) at school. This score is a comprehensive evaluation (1– 4, where 1 ⫽ excellent and most active), given once per semester by school physical education teachers, for performance and activity in a variety of sports and physical activities at school such as running, jumping, rope-skipping, throwing, and ball-games. Spare-time physical activities (PAs) were recorded by means of a self-administered questionnaire in minutes per week over the past year. There were ⬎15 categories of PAs during class breaks and time out of school, including walking, running, cycling, swimming, skating, table-tennis, basketball, volleyball, soccer, softball, high jump, long jump, rope-skipping, skipping, shuttlecock-kicking, and others. Bone age was estimated by the Greulich and Pyle Atlas method.12 Bone age (to the nearest 3 months) was determined by assessing development stages of metacarpals, phalanges, and carpals from a hand and wrist X-ray against a set of original photographs for the method, provided by a technician from the Beijing Anthropometry Centre. Ultraviolet (UV) exposure was measured to obtain indirect information on vitamin D supply. A semiquantitative personal UV dosimeter (John Thacher & Associates Pty, Ltd., Hong Kong)2,22 validated by the Australian Photobiology Testing Facility at the University of Sydney (Greg Murphy, 1995, personal communication) was used to obtain an average daily exposure dose to UV B waveband, 280 –320 nm (millijoules per square centimeter), required for the effective production of vitamin D in skin. It was attached to the top of the shoulder, outside the clothes, for a whole day and five badges were used for 5 consecutive days (3 weekdays and 2 weekend days). Each badge was compared against a reference comparator series of hues to obtain the exposure for each subject. The major use of the badge in this study was for comparison of UV exposure between groups and between seasons and not for absolute values. The measurements were made in early October 1995 for summer exposure and in mid-January 1996 for winter. Beijing is located at 40°N latitude.
It ⫽ a ⫹ b ⫻ Is where It is true intake, Is is intake from the SFFQ, and a and b are constants from regression analysis of data from the subsample of 221 girls.
Statistical Analysis Bone Mineral Measurement BMC and bone width (BW), at the distal one-third and one-tenth radius and ulna of the nondominant forearms were measured by means of a portable bone mineral analyzer (BMD-4, Beijing Broadcast Technology Institute), utilizing single-photon absorptiometry (SPA). At the distal one-tenth and one-third (proximal) radius sites, the ratio of trabecular vs. cortical bone was esti-
Descriptive statistics were performed for all variables by socioeconomic area (rural, suburban, and urban) and by milk consumption group (no milk, milk consumption less than the median, and milk consumption greater than or equal to the median). One-way analysis of variance (ANOVA) and Bonferroni post hoc multiple comparisons were used for testing differences between means and chi-square for testing differences between
Bone Vol. 30, No. 3 March 2002:521–528
X. Q. Du et al. Milk consumption and BMC in adolescent girls
523
Table 1. Characteristics and dietary intakes of adolescent girls aged 12–14 years in the Beijing area in 1995 Rural (n ⫽ 200)
Suburban (n ⫽ 150)
Urban (n ⫽ 299)
Total (n ⫽ 649)
Age (yr) Height (cm) Weight (kg) Bone age (yr) SPAS PAs (min/wk) Menarcheal status (% yes) UV summer (mJ/cm2 UVB)e UV winter (mJ/cm2 UVB)
13.2 ⫾ 0.6 154.0 ⫾ 5.7 45.0 ⫾ 7.7 13.8 ⫾ 1.3 1.8 ⫾ 0.8 372 ⫾ 478 64.5 (57.7–71.3) 59.9 ⫾ 13.5 (50) 34.2 ⫾ 11.4 (90)
13.0 ⫾ 0.6a 154.3 ⫾ 5.8 45.8 ⫾ 8.5 13.8 ⫾ 1.4 2.0 ⫾ 0.8b 527 ⫾ 587 66.7 (59.0–74.4) 43.3 ⫾ 22.3a (50) 15.9 ⫾ 4.0a (95)
12.7 ⫾ 0.4a,c 154.3 ⫾ 7.2 47.0 ⫾ 10.5 13.7 ⫾ 1.4 1.8 ⫾ 0.7c 698 ⫾ 700a,d 65.6 (60.1–71.1) 36.0 ⫾ 11.4a,d (111) 15.0 ⫾ 5.1a (190)
12.9 ⫾ 0.6 154.2 ⫾ 6.4 46.1 ⫾ 9.3 13.8 ⫾ 1.4 1.8 ⫾ 0.8 557 ⫾ 627 65.5 (61.8–69.2) 43.4 ⫾ 17.9 (211) 19.8 ⫾ 10.7 (375)
Dietary intake Milk in whole sample (g/day) Milk in consumers only (g/day) Dairy calcium (mg/day) As % of total calcium Calcium (mg/day) Vitamin D (g/day) Protein (g/day) Energy (kJ/day) Calcium: phosphorus ratio
9 ⫾ 31 32 ⫾ 51 18 ⫾ 49 5.4 ⫾ 13.4 314 ⫾ 91 0.56 ⫾ 0.55 47 ⫾ 9 6932 ⫾ 747 0.44 ⫾ 0.07
37 ⫾ 58a 57 ⫾ 64 46 ⫾ 67a 14.4 ⫾ 19.3a 349 ⫾ 89a 0.74 ⫾ 0.64 50 ⫾ 9a 6943 ⫾ 658 0.48 ⫾ 0.07a
83 ⫾ 74a,c 91 ⫾ 72a,c 120 ⫾ 100a,c 35.3 ⫾ 26.4a,c 388 ⫾ 94a,c 1.51 ⫾ 1.05a,c 51 ⫾ 9a 6953 ⫾ 658 0.51 ⫾ 0.09a,c
50 ⫾ 68 75 ⫾ 71 72 ⫾ 92 21.2 ⫾ 25.3 356 ⫾ 97 1.04 ⫾ 0.94 50 ⫾ 9 6944 ⫾ 686 0.48 ⫾ 0.09
Characteristic
Values are mean ⫾ SD or mean (95% confidence interval). Number of subjects for whom data were available in parentheses. KEY: SPAS, School Physical Activity Score (score 1– 4, where 1 ⫽ best performance and most active); PAs, spare-time physical activities; UV, ultraviolet; UVB, ultraviolet B waveband (about 280 –320 nm). a p ⬍ 0.01, bp ⬍ 0.05 (vs. rural); cp ⬍ 0.01, dp ⬍ 0.05 (vs. suburban); esignificant main effect of season, p ⬍ 0.0005.
percentage data. Bivariate and partial correlation analyses were conducted to identify variables associated with bone mineral measurements. Multiple regression analyses of BMC (for the entire sample and for each area) on possible contributing factors were performed to examine predictors of BMC and their relative importance. Factors including food and nutrient intakes, physical activity, and pubertal maturation that had a significant correlation with BMC were included in the analysis, with adjustment of bone size variables of BW, body weight, and height.24 All data were regarded as statistically significant at the 5% level (two-tailed). SPSS software was used for the analysis (SPSS, Inc., Chicago IL). Results The characteristics and dietary intakes of 649 Beijing girls by socioeconomic area are shown in Table 1. Compared with rural girls, urban girls were slightly younger. The UV dose in winter was significantly lower than in summer (repeated-measures ANOVA, p ⬍ 0.0005). UV exposure across each season, was greatest in rural ⬎ suburban ⬎ urban girls. Height, weight, and bone age were comparable between areas. Rates of Tanner breast development stages 2, 3, and 4 were 11.9%, 39.5%, and 46.3%, respectively, with 0.7% of girls at stage 1. There were no differences in Tanner stage and menarche status among girls in the three areas. Girls in suburban areas had poorer SPAS scores than their counterparts in urban and rural areas. Rural and suburban girls had lower PAs than urban girls. Bone Mineral Measurements Bone mineral density by milk consumption (three groups: nomilk; low-milk [less than median milk intake, 22 ⫾ 18 g/day]; and high-milk [greater than or equal to median milk intake, 128 ⫾ 65 g/day]) are presented as mean ⫾ SD in Table 2. Significant differences in BMD among the groups were achieved on three of the four forearm sites measured (p ⬍ 0.01). The
no-milk consumers had the lowest BMD, whereas the high-milk group had the highest BMD. Compared with the no-milk group, the percent increase in BMD at the four sites in milk consumers averaged 6.0% (range 2.5%–10.5%), with a trend of higher percent increase in the high-milk group than in the low-milk group. Differences in BMC and BW among the three groups were otherwise not significant, apart from the BMC at one-tenth radius (p ⬍ 0.05). There were no significant differences in height, Tanner stage, and SPAS between/among the three milk consumption groups. Weight and bone age, which are known to be positively related to BMD, were lower in the high-milk group than in the low-milk group (p ⬍ 0.05), making the BMD differences even more striking. Intake of nondairy calcium was the same in the no-milk and the low-milk groups. The high-milk group had an even lower intake of calcium from nondairy products than the other two groups. Nondairy vitamin D intake, protein intake, and PAs were higher in the two milk-consuming groups than in the no-milk group, which may have contributed independently to their higher BMD. This point is further addressed in what follows. Factors Associated With Bone Mass in Multiple Regression Models Multiple regression analyses of BMC on possible contributing factors were performed for the 517 girls for whom complete data were available. Ten variables correlated significantly with BMC and were included as independent variables in the analysis of model 1 for each site of BMC (i.e., BW, body weight, height, bone age, Tanner stage of breast development, SPAS, calcium and vitamin D intakes, calcium:phosphate (Ca:P) ratio, and socioeconomic area). Spare-time physical activity was also included in the analysis although it did not show a correlation that achieved statistical significance. In model 2, the milk variable was added and analyzed together with the 11 variables from model 1 because only the milk group had significant partial correlation with BMC among the 13 food groups. Intakes of
524
X. Q. Du et al. Milk consumption and BMC in adolescent girls
Bone Vol. 30, No. 3 March 2002:521–528
Table 2. Characteristics, dietary intakes, and bone mineral measurements of adolescent girls aged 12–14 years in the Beijing area in 1995, by milk consumption No-milk group (n ⫽ 222) Age (yr) Height (cm) Weight (kg) Bone age (yr) SPAS PAs (min/wk) Tanner stage (%) 1 2 3 4 5 Menarcheal status (% yes) UV summer (mJ/cm2 UVB)e UV winter (mJ/cm2 UVB) Dietary intake Milk (g/day) Dairy calcium (mg/day) As % of total calcium Nondairy calcium (mg/day) Dairy vitamin D (g/day) Nondairy vitamin D (g/day) Dairy protein (g/day) Nondairy protein (g/day) Dairy energy (kJ/day) Nondairy energy (kJ/day) Bone mineral measurements Distal one-third radius BMC (g/cm) BMD (g/cm2) BW (cm) Distal one-third ulna BMC (g/cm) BMD (g/cm2) BW (cm) Distal one-tenth radius BMC (g/cm) BMD (g/cm2) BW (cm) Distal one-tenth ulna BMC (g/cm) BMD (g/cm2) BW (cm)
13.2 ⫾ 0.6 153.8 ⫾ 5.8 46.5 ⫾ 8.6 13.8 ⫾ 1.4 1.9 ⫾ 0.8 392 ⫾ 472
Low-milk group (n ⫽ 213) 12.8 ⫾ 0.5a 154.9 ⫾ 6.2 47.1 ⫾ 9.8 13.9 ⫾ 1.4 1.8 ⫾ 0.8 591 ⫾ 645a
1.0 10.3 41.2 46.1 1.5 66.7 (60.4–73.0) 49.0 ⫾ 15.7 (76) 26.1 ⫾ 12.7 (112)
1.1 13.6 36.7 46.9 1.7 69.0 (62.7–75.3) 37.7 ⫾ 13.2a (71) 17.5 ⫾ 8.6a (123)
High-milk group (n ⫽ 299) 12.8 ⫾ 0.5a 153.9 ⫾ 7.2 44.8 ⫾ 9.2d 13.6 ⫾ 1.3d 1.8 ⫾ 0.8 698 ⫾ 710a 0 12.1 40.2 46.0 1.7 60.7 (54.0–67.4) 43.1 ⫾ 22.5 (64) 16.8 ⫾ 8.1a (140)
Total (n ⫽ 649) 12.9 ⫾ 0.6 154.2 ⫾ 6.4 46.1 ⫾ 9.3 13.8 ⫾ 1.4 1.8 ⫾ 0.8 557 ⫾ 627 0.7 11.9 39.5 46.3 1.6 65.5 (61.8–69.2) 43.4 ⫾ 17.9 (211) 19.8 ⫾ 10.7 (375)
0 0 0 293.4 ⫾ 77.0 0 0.53 ⫾ 0.59 0 46.0 ⫾ 8.3 0 6794 ⫾ 703
22.4 ⫾ 18.9a 40.8 ⫾ 37.0a 11.8 ⫾ 10.4a 299.4 ⫾ 75.3 0.12 ⫾ 0.22 0.72 ⫾ 0.55a 1.0 ⫾ 1.0a 50.0 ⫾ 9.0a 98.7 ⫾ 96.0a 6894 ⫾ 672
128.2 ⫾ 65.0a,c 176.4 ⫾ 84.7a,c 39.8 ⫾ 17.1a,c 259.9 ⫾ 79.9a,c 1.21 ⫾ 1.13a,c 0.55 ⫾ 0.58c 3.9 ⫾ 1.8a,c 48.6 ⫾ 8.0a 330.4 ⫾ 151.7a,c 6721 ⫾ 681d
49.6 ⫾ 68.1 71.6 ⫾ 92.2 17.0 ⫾ 20.3 284.3 ⫾ 79.2 0.44 ⫾ 0.86 0.60 ⫾ 0.58 1.6 ⫾ 2.0 48.1 ⫾ 8.6 141.4 ⫾ 172.7 6803 ⫾ 688
0.684 ⫾ 0.134 0.606 ⫾ 0.081 1.123 ⫾ 0.141
0.711 ⫾ 0.189 0.633 ⫾ 0.097b 1.116 ⫾ 0.213
0.708 ⫾ 0.218 0.642 ⫾ 0.114a 1.093 ⫾ 0.176
0.701 ⫾ 0.183 0.627 ⫾ 0.099 1.111 ⫾ 0.179
0.589 ⫾ 0.125 0.597 ⫾ 0.087 0.979 ⫾ 0.125
0.593 ⫾ 0.146 0.612 ⫾ 0.110 0.962 ⫾ 0.131
0.596 ⫾ 0.150 0.614 ⫾ 0.112 0.962 ⫾ 0.134
0.593 ⫾ 0.140 0.608 ⫾ 0.104 0.968 ⫾ 0.130
0.554 ⫾ 0.178 0.352 ⫾ 0.065 1.543 ⫾ 0.309
0.620 ⫾ 0.322b 0.379 ⫾ 0.113b 1.567 ⫾ 0.423
0.596 ⫾ 0.259 0.389 ⫾ 0.118a 1.499 ⫾ 0.312
0.590 ⫾ 0.260 0.373 ⫾ 0.102 1.537 ⫾ 0.352
0.335 ⫾ 0.098 0.372 ⫾ 0.062 0.880 ⫾ 0.181
0.361 ⫾ 0.162 0.393 ⫾ 0.103 0.886 ⫾ 0.242
0.359 ⫾ 0.161 0.403 ⫾ 0.115a 0.859 ⫾ 0.187
0.351 ⫾ 0.143 0.389 ⫾ 0.096 0.875 ⫾ 0.205
Values are mean ⫾ SD or mean (95% confidence interval). Number of subjects for whom data were available in parentheses. KEY: BMC, bone mineral content; BMD, bone mineral density; BW, bone width; SPAS, school physical activity score (score 1– 4, where 1 ⫽ best performance and most active); PAs, spare-time physical activities; UV, ultraviolet; UVB, ultraviolet B waveband (about 280 –320 nm). a p ⬍ 0.01, bp ⬍ 0.05 (vs. no milk); cp ⬍ 0.01, dp ⬍ 0.05 (vs. low milk); esignificant main effect of season, p ⬍ 0.0005.
protein, nondairy protein, and nondairy vitamin D showed no correlation with bone measurements and therefore were not included in the analysis. The eight models are summarized as follows and the four models with milk added are described in Table 3: Y
Y
Analyzed together with the other variables, height, calcium intake Ca:P ratio, and PAs did not meet the criteria for inclusion in any of the eight models. BW, milk intake, vitamin D intake, bone age, body weight, SPAS, Tanner stage of breast development, and socioeconomic area were predictors or determinants of BMC. About 60% of the variation in BMC (ranging 53.5%– 64.7%) could be explained by these models.
Y
Y
Y
Among the independent variables, BW was the best predictor included in all eight models and explained most of the variation in BMC. SPAS was selected into seven of the eight models, indicating that it was a determinant of BMC, in spite of the small effect on BMC (associated with up to 1.1% change in R2). The negative regression coefficient between the SPAS and BMC was due to the coding of SPAS (1– 4, where 1 ⫽ excellent and most active). The importance of bone age, weight, and Tanner stage appeared to depend on bone site measured (distal one-third or one-tenth of forearm; radius or ulna). Weight predicted only BMC of distal one-third sites where bone is mainly cortical,
Bone Vol. 30, No. 3 March 2002:521–528
X. Q. Du et al. Milk consumption and BMC in adolescent girls
525
Table 3. Multiple regression models for BMC of 517 adolescent girls aged 12–14 years in the Beijing areaa Regression coefficient
Standard error
Standardized coefficient
BMC at distal one-tenth radius Model 1: MR ⫽ 0.770, R2 ⫽ 0.593, adjusted R2 ⫽ 0.589, SE ⫽ 0.167 BW (mm) Milk (g/day) Bone age (yr) Socioeconomic area SPAS (Constant)
0.050 0.0004 0.025 0.026 ⫺0.020 ⫺0.557
0.002 0.00009 0.006 0.010 0.010 0.081
BMC at distal one-third radius Model 2: MR ⫽ 0.805, R2 ⫽ 0.647, adjusted R2 ⫽ 0.643, SE ⫽ 0.110 BW (mm) Milk (g/day) Bone age (yr) SPAS Socioeconomic area Weight (kg) (Constant)
0.073 0.0003 0.011 ⫺0.025 0.019 0.001 ⫺0.340
BMC at distal one-tenth ulna Model 3: MR ⫽ 0.770, R2 ⫽ 0.593, adjusted R2 ⫽ 0.591, SE ⫽ 0.093 BW (mm) Milk (g/day) Tanner stage, breast (Constant) BMC at distal one-third ulna Model 4: MR ⫽ 0.732, R2 ⫽ 0.536, adjusted R2 ⫽ 0.532, SE ⫽ 0.096 BW (mm) Weight (kg) Milk (g/day) SPAS (Constant)
p value
Partial R2
0.695 0.128 0.128 0.090 ⫺0.059
0.000 0.000 0.000 0.007 0.037 0.000
0.539b 0.030 0.014 0.006 0.003
0.003 0.00006 0.004 0.006 0.006 0.001 0.056
0.725 0.127 0.081 ⫺0.105 0.096 0.070
0.000 0.000 0.001 0.000 0.002 0.022 0.000
0.582b 0.032 0.011 0.010 0.008 0.004
0.053 0.0003 0.014 ⫺0.175
0.002 0.00005 0.006 0.024
0.738 0.172 0.073
0.000 0.000 0.011 0.000
0.557b 0.031 0.005
0.070 0.002 0.0002 ⫺0.018 ⫺0.166
0.003 0.001 0.00005 0.006 0.036
0.654 0.147 0.110 ⫺0.099
0.000 0.000 0.000 0.001 0.000
0.495b 0.018 0.014 0.010
KEY: BMC, bone mineral content; BW, bone width; MR, multiple R, square root of coefficient of determination; R2, coefficient of determination; SE, standard error; SPAS, School Physical Activity Score (score 1– 4, where 1 ⫽ most active and best performance). a Stepwise method, entry criteria: p ⬍ 0.05. b Together with constant.
Y
Y
Y
whereas Tanner staging was associated with BMC of distal one-tenth sites where bone is a mixture of cortical and trabecular and bone turnover is faster. Bone age was only included into the four models for the radius. These factors were also associated with only a very small change in R2 (0.3%–1.8%). After BW, socioeconomic area accounted for the variation in BMC of up to 2.8%, with inclusion in five of the eight models. Further analysis of BMC models, stratified by socioeconomic area, is therefore presented subsequently. Vitamin D intake was an important determinant and predictor of BMC, accounting for up to 1.9% of the variation in BMC. However, milk intake was the only nutrition variable included in the model, accounting for up to 3.2% of the change in R2. This suggests milk intake was a more important determinant of BMC than vitamin D alone. When analyzed with body size and other confounding variables,
the relative importance of three milk nutrients to BMC was in the order milk vitamin D ⬎ milk calcium ⬎ milk protein, given 3.5%, 3.3%, and 2.8% changes in R2, respectively. Multiple linear regression analysis of BMC at each site of the forearm when stratified by socioeconomic area yielded a total of 12 models (Table 4). Calcium, Ca:P ratio, and PAs were not included into any of the models; BW was presented in all models; milk, SPAS, weight, and bone age were included in at least four models; predictors of BMC for each socioeconomic area varied as milk and SPAS were determinants of BMC in urban girls, whereas weight and bone age were determinants in rural and suburban girls. Milk and SPAS explained the variation of BMC up to 2.3% and 2.6%, respectively. Milk was not included in six models of the distal one-third radius and ulna BMC for rural and suburban girls, nor in the model of distal one-third ulna BMC for urban girls.
526
X. Q. Du et al. Milk consumption and BMC in adolescent girls
Bone Vol. 30, No. 3 March 2002:521–528
Table 4. Summary of multiple regression models of BMC at four sites of the forearm in Beijing girls aged 12–14 years, stratified by socioeconomic area BMC models Rural
Suburban
Urban
Independent variable
R10
R3
U10
U3
R10
R3
U10
U3
R10
R3
U10
U3
BW (mm) Milk intake (g/day) SPAS Weight (kg) Bone age (yr) Height (cm) Tanner stage VitD intake (g/day) R2 MR
⫹a
⫹
⫹ ⫹
⫹
⫹ ⫹
⫹
⫹
⫹
⫹ ⫹
⫹ ⫹ ⫹
⫹ ⫹ ⫹
⫹
⫹
⫹ ⫹ ⫹
⫹
⫹
0.70 0.84
0.67 0.82
0.57 0.75
0.65 0.81
0.57 0.76
⫹ 0.48 0.69
⫹ ⫹ ⫹ 0.60 0.77
⫹ ⫹ ⫹
0.70 0.84
⫹ ⫹ 0.76 0.87
⫹ ⫹
⫹
0.58 0.76
0.61 0.78
0.57 0.76
Partial R2 range 0.404–0.672b 0.007–0.023 0.008–0.026 0.043–0.139 0.009–0.042 0.010–0.030 0.009 0.012
KEY: BMC, bone mineral content; R2, coefficient of determination; MR, Multiple R, square root of R2; R10, one-tenth radius site; R3, one-third radius site; U10, one-tenth ulna site; U3, one-third ulna site; VitD, vitamin D. a ⫹ indicates variable is included in the model. b Together with constant.
Milk Consumption and Nutrient Intakes Cereals and vegetables formed a large proportion of the total food intake of Beijing girls. Compared with rural girls, urban girls consumed less cereal and vegetables but more animal foods (milk, seafood, and eggs) as well as more pastry, ice confections, fruits, and sugar (p ⬍ 0.0005), whereas consumption of legumes and nuts was similar. Suburban girls had the highest consumption of ice confections with all other food intakes in between rural and urban rates. Milk consumption data are presented in Tables 1 and 5. The milk group included fresh milk (pasteurized), powdered milk, vitamin D-fortified milk (15 g fortified vitamin D per liter, pasteurized), and yogurt. Sixty-six percent of girls consumed milk and milk products at an average of 75 g/day. The average level was 50 g/day for the sample as a whole, including non-milk consumers (about one-third of girls consumed no milk at all). The proportion of non-milk consumers was much higher in the rural (70%) than in the urban area (9%). As shown in Table 5, there were few problems with purchasing power or commercial supply of milk; however, milk consumption may have been affected by perceptions about milk, and food habits. Half of the girls reported that they did not like drinking milk and 5% girls reported that milk was not available at home. There were area differences in terms of milk perceptions and milk supply. The proportion reporting dislike of milk was the highest in rural girls. No problems in terms of local milk supply were reported in the urban setting. Feeling uncomfortable after drinking milk was reported at an average rate of 18%, ranging from 12% in the urban area to 29% in the rural area. The “uncomfortable feelings”
referred to stomach upset, cramps, bloating, and diarrhea, but may have included other minor negative feelings. The average intakes of calcium and vitamin D were 356 mg/day and 1.04 g/day, respectively, accounting for 30% and 10% of the Chinese recommended dietary allowance (RDA) for this age group.10 The protein intake of 50 g/day was also inadequate at 63% of the RDA of 80 g/day for a plant-based diet. The energy intake was 72% of the RDA. The Ca:P ratio was much lower than the level of 1.0 recommended in the U.S. RDAs.23 Calcium, vitamin D, and protein intakes and Ca:P ratio were higher in urban ⬎ suburban ⬎ rural girls. Table 6 shows the source of dietary calcium and vitamin D by food groups. In total, 58.9% of calcium was from plant foods, whereas 41.1% was from animal foods. About 20% of calcium was from the milk group. Ice cream and ice confections contributed 6.6% of total calcium. Dietary vitamin D was mainly from eggs (52.6%) and milk (35.4%) as well as pastry (11.4%) (traditionally made with eggs). The sources of calcium and vitamin D varied by socioeconomic area due to differences in food patterns (see earlier). The percentage of milk calcium in total calcium was 5%, 14%, and 35%, respectively, in rural, suburban, and urban areas (p ⬍ 0.01, Table 1). Discussion The positive association of BMD with level of milk consumption, the fact that no other dietary variables in this study, other than milk intake, were able to explain the difference in BMD (Table 2, the differences in BMD between milk groups were still
Table 5. Milk consumers, access to milk, and attitudes toward milk consumption in adolescent girls aged 12–14 years in the Beijing area in 1995
Item Milk consumers (%) Dislike drinking milk (%) Feeling uncomfortable after drinking milk (%)a No milk provided by parents (%) Not affordable (%) No milk supply locally (%)
Rural (n ⫽ 200)
Suburban (n ⫽ 150)
Urban (n ⫽ 299)
Total (n ⫽ 649)
29.5 63.6 29.0 6.2 6.8 8.2
64.0 54.8 16.4 6.3 1.6 2.4
91.0 39.9 12.1 3.1 2.7 0
65.8 50.6 18.0 4.7 3.6 2.8
p ⬍ 0.0005 (chi-square test) among girls in the three areas for each listed item. a “Uncomfortable” referred to symptoms such as stomach upset, cramps, bloating, and diarrhea, or any minor abnormal feeling.
Bone Vol. 30, No. 3 March 2002:521–528
X. Q. Du et al. Milk consumption and BMC in adolescent girls
527
Table 6. Sources of calcium and vitamin D for adolescent girls aged 12–14 years in the Beijing area in 1995a Food group Cereals and products Milk and milk products Vegetables Legume Seafood Ice cream and ice confections
Ca (%) 29.4 19.9 11.0 8.5 6.8 6.6
VitD (%) c
11.4 35.4 0 0 0 0.1
Food group
Ca (%)
VitD (%)
Nuts Eggs Meat and meat products Fruits Chocolate and sugar Total
6.1 4.5 3.3 2.8 1.1 100.0
0 52.6 0.5 0 0 100.0
KEY: Ca, calcium; VitD, vitamin D. Data from the whole sample of Beijing girls (n ⫽ 1248) from which subjects in present study were drawn. From pastry.
a
b
present after stratification by Tanner stage [data not presented]), and the fact that milk intake was the only dietary determinant of forearm BMC (Tables 3 and 4) all suggest a favorable effect of milk consumption on bone mass acquisition in Beijing pubertal girls. No comparable studies of Asian children have been reported; however, the results are comparable with a cross-sectional study in Icelandic white girls aged 13 and 15 years,19 in which calcium intake from dairy products was found to be significantly correlated with distal and ultradistal BMD in the forearm of the 15-year-old group after adjustment for menarcheal age and weight. The researchers also reported that the correlation coefficients between BMD and calcium from dairy products were significantly greater within the lowest tertile of ⬍800 mg/day. These Icelandic girls had a much higher dairy calcium intake (1082 mg/day) than our subjects (70 mg/day). In contrast, no association was found between calcium intake from dairy products (1200 –1300 mg/day) and BMC at diaphyseal and metaphyseal sites in 1359 Dutch boys and girls aged 7–11 years.29 The investigators suggested that this result might be explained by the fact that about 80% of the children in the study had calcium intakes around 1400 mg/day, which is probably a ceiling value for children of this age. The effect of milk on BMC/BMD was also indicated by our finding that nutrients such as calcium, vitamin D, and protein from milk correlated well with BMC/BMD, whereas these nutrients from nondairy sources were not correlated with the bone measurements. This is consistent with a finding in Chinese adult women that BMC and BMD were correlated positively with total calcium, dairy calcium, and to a lesser extent with nondairy calcium.17 The positive effect of milk on bone mineral is probably an integrated consequence of its provision of several nutrients, particularly vitamin D, calcium, and protein, because all of these showed positive correlations with the bone measurements. The results may also be due to growth factors known to be present in milk.4,28 Furthermore, the relative importance of the three milk nutrients to BMC was milk vitamin D ⬎ milk calcium ⬎ milk protein, as suggested by their contributions to R2 in the multiple regression equations. One major concern about milk consumption promotion in Chinese children by Western experts is the high lactose intolerance rate. Yang et al. reported that incidence of lactose intolerance, after taking 25 g of lactose, was 30.5% in urban children aged 11–13 years, with a lower rate of 14% in children with intolerance symptoms after consuming 50 g of milk powder (12 g of lactose).30 Incidence of feeling uncomfortable after drinking milk was reported to be 18% in the present study (Table 2), with values of 12% in urban, 16% in suburban, and 29% in rural areas (p ⬍ 0.0005). This rate is similar to Yang’s result with the 50 g milk powder test and to the 17.8% rate in Hong Kong Chinese women aged 21– 40 years with symptoms of lactose intolerance, as reported by Ho et al.15 Ability to tolerate milk is being further tested in our current 2 year milk supplementation trial in which
500 urban Beijing girls aged 11 years consumed 330 mL of milk per school day for 2 years.31 Physical activity (PA) expressed as SPAS was another determinant of BMC identified for Beijing adolescent girls in this study. SPAS was included in seven of the eight final models with a small contribution to the variation in BMC (up to 1.1% change in R2), suggesting that SPAS was an independent determinant of BMC. There have been no studies using similar school PA scores in children and adolescents. On the other hand, the result that spare-time PA was not a predictor of BMC from this study is consistent with data of Gunnes et al.,13 who found that weightbearing PA was not a predictor of forearm BMD gain in 150 Norwegian girls aged ⱖ11 years, and those of Cheng et al.9 whose 3 year longitudinal study of a small sample of 179 Hong Kong boys and girls aged 12–16 years showed that bone mineral accretion was not influenced by exercise, or level of physical fitness. The SPAS used in this study is a comprehensive rating of each subject’s performance over the past year in a variety of school physical education activities and as such may reflect innate ability, but may also be assumed to be positively associated with the activity level of the girls, whereas PAs quantitatively accounted for all physical activities conducted during class breaks and out-of-school hours. Physically demanding activities after school (e.g., assistance in farm work) was not included in the questionnaire, because this kind of activity was assumed to be rare in young girls given the long school hours and the protection afforded to the sole child in Chinese families. The role of PA assessed in a variety of ways (type, frequency, and intensity of activity; physical fitness; performance level) as a determinant of bone mineral status in this population needs further study. As a cross-sectional study of this nature has limitations, further studies (e.g., milk intervention trials) are needed to provide more definitive evidence of the effects of milk on bone mass in the Chinese adolescent population. Because few problems appeared to be associated with milk drinking in this crosssectional study, such trials may be feasible in this group.
Acknowledgments: The authors thank Dr. D. Mackerras for her advice on dietary survey methods and Professor A. Baumann for his advice on sample size. The authors are indebted to Professor G. S. Ma, X. W. Li, G. S. Liu, X. X. Li, J. J. Tan, J. M. Zhu and his staff, Dr. L. Huang, B. M. Guo, J. X. Gao, H. Wang, Dr. X. J. Shi, Dr. Z. H. Liu, and E. Emmerson for their help and technical assistance with the project. The authors are grateful to all principals, staff, and health workers involved for their effort and contribution to the fieldwork and all subjects and their parents for their cooperation. This research was supported in part by the Dairy Research and Development Corporation, Australia.
528
X. Q. Du et al. Milk consumption and BMC in adolescent girls
Bone Vol. 30, No. 3 March 2002:521–528
References 1. Adelaide Children’s Hospital (ACH). Stages of Puberty. Sydney: Pharmacia & Upjohn; 1989. 2. Bannard, J. and Maguire, D. Suncheck patches and monitoring of UVR exposure. Br J Dermatol 131:911–912; 1994. 3. Beijing Institute of Food Inspection and Examination (BIFIE). Data on Dietary Survey Among Beijing Residents in 1992. Beijing: BIFIE; 1992. 4. Belford, D. A., Rogers, M. L., Francis, G. L., Payne, C., Ballard, F. J., and Goddard, C. Platelet-derived growth factor, insulin-like growth factors, fibroblast growth factors and transforming growth factor beta do not account for the cell growth activity present in bovine milk. J Endocrinol 154:45–55; 1997. 5. Block, G., Hartman, A. M., Dresser, C. M., Carroll, J. G., and Gardner, L. A data-based approach to diet questionnaire design and testing. Am J Epidemiol 124:453– 469; 1986. 6. Cadogan, J., Eastell, R., Jones, N., and Barker, M. E. Milk intake and bone mineral acquisition in adolescent girls: Randomised, controlled intervention trial. BMJ 315:1255–1260; 1997. 7. Chan, G. M., Hoffman, K., and McMurry, M. Effects of dairy products on bone and body composition in pubertal girls. J Pediatr 126:551–556; 1995. 8. Cheng, J. C. Y., Leung, S. S. S. F., Lee, W. T. K., Lau, T. F., Maffulli, N., Cheung, A. Y. K., and Chan, K. M. Determinants of axial and peripheral bone mass in Chinese adolescents. Arch Dis Child 78:524 –530; 1998. 9. Cheng, J. C., Maffulli, N., Leung, S. S., Lee, W. T., Lau, J. T., and Chan, K. M. Axial and peripheral bone mineral acquisition: A 3-year longitudinal study in Chinese adolescents. Eur J Pediatr 158:506 –512; 1999. 10. Chinese Nutrition Society (CNS). Recommended Dietary Allowance for Chinese. Beijing: CNS; 1988. 11. Du, X., Greenfield, H., Fraser, D. R., Ge, K., Trube, A., and Wang, Y. Vitamin D deficiency and associated factors in adolescent girls in Beijing. Am J Clin Nutr 74:494 –500; 2001. 12. Greulich, W. W. and Pyle, S. I. Radiographic Atlas of the Skeletal Development of the Hand and Wrist, 2nd Ed. Stanford, CA: Stanford University; 1959. 13. Gunnes, M. and Lehmann, E. H. Physical activity and dietary constituents as predictors of forearm cortical and trabecular bone gain in healthy children and adolescents: A prospective study. Acta Pediatr 85:19 –25; 1996. 14. He, W., Du, X., and Greenfield, H. CAVD, A Survey System Using EPI INFO. Beijing: Chinese Academy of Preventive Medicine; 1997 (software). 15. Ho, S. C., Leung, P. C., Swaminathan, R., Chan, C., Chan, S. S. G., Fan, Y. K., and Lindsay, R. Determinants of bone mass in Chinese women aged 21– 40 years. II. Pattern of dietary calcium intake and association with bone mineral density. Osteopors Int 4:167–175; 1994. 16. Holland, B., Welch, A. A., Unwin, I. D., Buss, D. H., Paul, A. A., and Southgate, D. A. T. McCance and Widdowson’s The Composition of Foods, 5th Ed. London: Royal Society of Chemistry and Ministry of Agriculture; 1991. 17. Hu, J. F., Zhao, X. H., Jia, J. B., Parpia, B., and Campbell, T. C. Dietary
18. 19.
20.
21.
22.
23. 24.
25. 26. 27. 28.
29.
30.
31.
calcium and bone density among middle-aged and elderly women in China. Am J Clin Nutr 58:219 –297; 1993. Institute of Nutrition and Food Hygiene (INFH). Food Composition Tables. Beijing: People’s Health; 1991. ¨ ., Valdimarsson, O ¨ ., Steingrimsdottir, L., and Sigurdsson, G. Kristinsson, J. O Relation between calcium intake, grip strength and bone mineral density in the forearms of girls aged 13 and 15. J Intern Med 236:385–390; 1994. Lee, W. T. K., Leung, S. S. F., Ng, M. Y., Wang, S. F., Xu, Y. C., Zeng, W. P., and Lau, J. Bone mineral content of two populations of Chinese children with different calcium intakes. Bone Miner 23:195–206; 1993. Lee, W. T. K., Leung, S. S. F., Wang, S. H., Xu, Y. C., Zeng, W. P., Lau, J., Oppenheimer, S. J., and Cheng, J. C. Y. Double-blind, controlled calcium supplementation and bone mineral accretion in children accustomed to a low-calcium diet. Am J Clin Nutr 60:744 –750; 1994. Moseley, H., Mackie, R. M., and Ferguson, J. The suitability of Suncheck patches and Tanscan cards for monitoring the sunburning effectiveness of sunlight. Br J Dermatol 128:75–78; 1993. National Research Council (NRC). Recommended Dietary Allowance, 10th Ed. Washington, DC: National Academy; 1989. Prentice, A., Parsons, T. J., and Cole, T. J. Uncritical use of bone mineral density in absorptiometry may lead to size-related artifacts in the identification of bone mineral determinants. Am J Clin Nutr 60:837– 842; 1994. Rosen, C. J. Osteoporosis: Diagnostic and Therapeutic Principles. Totowa, NJ: Humana; 1996. Tanner, J. M. Growth at Adolescence. Oxford: Blackwell; 1962. Technical Committee of National Nutrition Survey (TCNNS). Dietary Intakes of Chinese People. Beijing: Institute of Nutrition and Food Hygiene; 1994. Toba, Y., Takada, Y., Yamamura, J., Tanaka, M., Matsuoka, Y., Kawakami, H., Itabashi, A., Aoe, S., and Kumegawa, M. Milk basic protein: A novel protective function of milk against osteoporosis. Bone 27:403– 408; 2000. VandenBergh, M. F. Q., DeMan, S. A., Witteman, J. C. M., Hofman, A., Trouerbach, W. T., and Grobbee, D. Physical activity, calcium intake, and bone mineral content in children in The Netherlands. J Epidemiol Commun Health 49:299 –304; 1995. Yang, Y., He, M., Cui, H., Bian, L., Liu, J., Cheng, W., Xu, H., and Feng, W. Study on the incidence of lactose intolerance of children in China. J Hygiene Res 28(1):44 – 46; 1999. Zhu, K., Du, X., Greenfield, H., and Fraser, D. R. Milk supplementation and bone mineral acquisition in Chinese adolescent girls: Preliminary results after one year. Proc Nutr Soc Austral 24:123; 2000.
Date Received: July 18, 2001 Date Revised: October 4, 2001 Date Accepted: October 24, 2001