J. E. P. Rockell¹, S. M. Williams², R. W. Taylor¹, A. M. Grant³, I. E. Jones³ and A. Goulding³

(1)	Department of Human Nutrition, University of Otago, Dunedin, New Zealand

(2)	Department of Preventive and Social Medicine, University of Otago, Dunedin, New Zealand

(3)	Department of Medical and Surgical Sciences, University of Otago, PO Box 913 Dunedin, New Zealand

Received: 23 April 2004  Accepted: 28 September 2004  Published online: 23 November 2004

Abstract  No previous longitudinal studies of calcium intake, anthropometry and bone health in young children with a history of avoiding cow s milk have been undertaken. We report the 2-year changes of a group of 46 Caucasian children (28 girls, l8 boys) aged 8.1±2.0 years (mean ± SD) who had low calcium intakes at baseline and were short in stature, with elevated body mass index, poor skeletons and lower Z scores for both areal bone mineral density (BMD, in grams per square centimeter) and volumetric density (bone mineral apparent density, BMAD, in grams per cubic centimeter), compared with a reference population of milk drinkers. At follow-up, adverse symptoms to milk had diminished and modest increases in milk consumption and calcium intake had occurred. Total body bone mineral content (BMC) and bone area assessed by dual energy X-ray absorptiometry had increased (P<0.05), and calcium intake from all sources was associated with both these measures (P<0.05). However, although some catch-up in height had taken place, the group remained significantly shorter than the reference population (Z scores –0.39±1.14), with elevated body mass index (Z scores 0.46±1.0). The ultradistal radius BMC Z scores remained low (–0.31±0.98). The Z scores for BMD had improved to lie within the normal range at predominantly cortical sites (33% radius, neck of femur and hip trochanter) but had worsened at predominantly trabecular sites (ultradistal radius and lumbar spine), where values lay below those of the reference group (P<0.05). Similarly, although volumetric BMAD Z scores at the 33% radius had normalized, BMAD Z scores at the lumbar spine remained below the reference population at follow-up (–0.67±1.12, P<0.001). Our results demonstrate persisting height reduction, overweight and osteopenia at the ultradistal radius and lumbar spine in young milk avoiders over 2 years of follow-up.

Keywords  Bone density change - Calcium - Children - Height - Milk - Protein

Introduction

Milk is regarded as a valuable food for bone growth, being a rich source of protein, vitamins, minerals (particularly calcium) and bioactive factors stimulating bone growth [1, 2]. Trials of milk and dairy food supplementation have demonstrated augmented skeletal growth in young children and adolescents [3, 4, 5, 6, 7]. Regular consumption of cow s milk during childhood and adolescence has also been associated with higher bone density in adults [8, 9, 10, 11, 12, 13] and a lower risk of osteoporotic fracture later in life in some studies [14], though not in all [15, 16].

Information concerning the bone health of young children who rarely drink milk is scarce. Yet this behavior is not uncommon. In New Zealand l7% of 3,275 children aged 5–14 years from a nationally representative sample reported drinking milk less than once per month [17]. To date, five studies in different countries have shown that children with a chronic history of milk avoidance who do not use substitute calcium-enriched foods appropriately have low calcium intakes and low bone mineral density [18, 19, 20, 21, 22]. These children also have reduced stature, small skeletons and high adiposity [21], and they are fracture prone [2]. It is not yet known whether these factors resolve during growth, as no previous studies have investigated bone growth longitudinally in children with a history of avoiding milk. However, many of these children are able to tolerate milk when they become older and allergic reactions to cow s milk generally resolve [23].

The present observational follow-up study of a group of young Caucasian milk avoiders was therefore undertaken 24 months after baseline. Our aims were to determine whether or not the children had increased their milk and dietary calcium intakes and to find out whether they exhibited any catch-up in height, bone area, and total skeletal and regional bone mineral accrual relative to a reference population of milk-drinking children from the same community.

Methods

At baseline (between 1999 and 2000) 50 Caucasian children aged 3–10 years with a history of prolonged milk avoidance were recruited from advertisements placed in shops, schools and community well-child clinics, as has been described previously [21]. Information was collected regarding anthropometry, lifelong consumption of dairy foods, and use of alternative substitute calcium-rich beverages or mineral supplements. Current calcium intake was assessed by a validated food-frequency questionnaire (FFQ) [24], and body composition and bone mineral density were measured by dual energy X-ray absorptiometry (DXA).

The present follow-up study was conducted 2 years after baseline, the protocol being approved by the Otago Ethics Committee. Informed consent was obtained for every participant. Study subjects still residing in Dunedin were invited to return for further studies of their general health and physical activity (by questionnaire), nutrition, bone health and body composition. Information concerning beverage consumption of the children and parental height was also collected at the follow-up visit. Pubertal status was assessed in children over 8 years of age [25, 26]. Current calcium intakes were estimated both by the same FFQ used at baseline and by 4-day diet records (4DDRs), which we collected just before the follow-up clinic appointment to avoid post-interview bias. The 4DDRs were collected on three randomly selected non-consecutive weekdays and one weekend day. The mean daily nutrient composition of the children s diets was calculated from the 4DDRs with the Diet Cruncher program (Way Down South Software, Dunedin, New Zealand) and a computerized New Zealand food composition database (version 9 of Food Files).

Each child attended the follow-up clinic visit accompanied by a parent. As at baseline, participants were weighed (electronic balance, Model 1609 N, Tanita Corp., Tokyo) and measured (Holtain stadiometer, Croswell, Crymych, UK) in light clothing without shoes. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared. Four DXA scans (total body, left hip, lumbar spine and non-dominant forearm) were performed according to the recommendations of the manufacturer (Lunar Corporation, Madison, Wis., USA) after the children had removed jewelry and any metal objects such as buckles or badges. The same scanner (Lunar DPX-L) was used for baseline and follow-up scans, which were taken and analyzed by the same person using the Lunar software package 4.7. In vivo precision for DXA scanning in adults is good [21] and quality control procedures (scanning of phantom blocks three times weekly) showed no evidence of drift in the scanner over the period of the study.

Statistical analyses were performed with Stata 7.0 (StataCorp, 2003, College Station, Tex., USA). Results for the raw data are presented as means with standard deviations and ranges. The 2-year changes in anthropometry, body composition and all bone variables were determined. Anthropometric and bone measures are expressed as Z scores derived from a contemporary reference population of 100 boys [27] and l00 girls [28] who had no history of fracture and lived in Dunedin, the city where the study was conducted.

Results

All participants still living in Dunedin (28 girls, l8 boys) completed our follow-up study: the remaining two girls and two boys from the original sample had gone overseas and could not be contacted. One participant seen at follow-up did not complete the 4DDR. At follow-up 41 children were Tanner stage l, three girls were Tanner stage 2, and two girls, Tanner stage 3. Early pubertal changes were appropriate, occurring only in the oldest (9–12 years) and heaviest girls. Thirteen children (28.3%) had a history of fracture, with five new fractures occurring during the 24 months of follow-up. Although 63% of the parents considered that their children were more active than at baseline, the time reported for vigorous physical activity (minutes per day) had decreased from a median of 46, range 8–197 at baseline, to a median of 27, range 0–197, at follow-up. Although every child participated in school physical activity classes, 12 subjects (26% of the sample) rated their physical activity for age and gender as below average. At follow-up the mean (SD) minutes per day of vigorous activity reported was lower than that in children of similar age from the reference population: 26 (22) vs 56 (65) in girls, P<0.02, and 41 (26) vs 86 (44) in boys, P<0.001, respectively.

When asked the principal reason for avoiding milk, 30 subjects stated lifestyle choice or taste dislike and only l6 stated that adverse symptoms were the reason for avoidance. However, when asked to specify any symptoms related to milk, 22 participants reported symptoms at baseline, whereas by follow-up this had dropped to ten subjects, some of whom listed more than one symptom. Thus, eight of the ten had GI symptoms, two had rhinitis or respiratory problems, four dermatitis and five problems such as headache, glue ear or malaise that were attributed to taking milk.

Overall the children had increased their mean (SD) dietary calcium intakes from 445 (236) mg at baseline to 625 (245) mg per day at follow-up (P<0.001), and more children were drinking some milk. Only seven children drank more than one cup of milk per week at baseline, whereas at follow-up 20 subjects did so (this estimate included cow, goat, soy or rice milks ). Moreover, the number of children consuming no milk whatsoever had dropped from 24 subjects at baseline to just five at follow-up, although the volume taken by most remained small, with 30 subjects (65%) still consuming less than 150 ml milk daily in total.

Dietary calcium intakes (in milligrams per day), assessed by FFQ at follow-up [median 588, interquartile range (IQR) = 462–779], showed good agreement with those calculated from 4 DDRs (596, IQR = 468–862), which also established reasonable intakes of energy (in kilojoules per day) (median 7,313, IQR = 6,724–8,240) and protein (in grams per day) (median 55, IQR = 46.5–66.0) by the study participants. The energy intakes were similar to those of a recent large, representative, nationwide sample that assessed nutrition by 24-h recall [17]. However, median protein intakes of boys aged 5–6 years were l4 g lower, and those of girls 7–10 years 8 g lower, than in the nationwide survey, suggesting that different food choices persisted in milk avoiders, despite every participant meeting the recommended nutrient intake (RNI) for protein. Total energy intakes were correlated with age (r=0.41, P<0.006), and 4 DDR calcium intakes with intakes of both energy (r=0.52, P<0.001) and protein (r=0.69, P<0.001).

Few children took multivitamin tablets (n=8) or calcium supplements (n=6), and use of those was intermittent. Our 4DDR analyses showed that although 15 children (33%) were consuming less than two-thirds of the age- and gender-specific RNI for calcium from all sources, 11 subjects (24%) now met the RNI for calcium, and 25 children (54%) were consuming more than 600 mg calcium daily. By contrast, at baseline the FFQ showed a higher proportion (57%) of the children were consuming less than two-thirds of the RNI, while only four participants had intakes that met the RNI for calcium [21].

Table 1 displays the raw data at follow-up for anthropometry, body composition, areal bone mineral density (BMD), bone mineral content (BMC), bone dimensions and dietary calcium intakes of the 46 subjects, with mean 24-month changes in these variables. The mean (SD) dietary calcium intakes of our milk-drinking reference populations were 1,179 (332) mg/day in the girls and 1,278 (618) mg/day in the boys, with two-thirds of this coming from dairy products.

Table 1  Characteristics at follow-up and changes observed since baseline for anthropometry, total body composition, regional bone measurements and dietary calcium intake (n=46)

Characteristic	Mean (SD)	Range	Two-year change
			Mean (95% CI)a
Age (years)	8.1 (2.0)	5.1–12.0	2.0
Height (cm)	127.4 (14.2)	101.4–163.4	12.1 (11.6, 12.6)
Weight (kg)	30.6 (11.9)	16.2–64.8	6.6 (5.6, 7.5)
Body mass index (kg/m2)	18.2 (3.2)	13.3–28.1	0.7 (0.4, 1.1)
Lean mass (kg)	21.3 (5.7)	12.4–37.9	4.1 (3.7, 4.4)
Fat mass (kg)	7.4 (6.4)	1.7–28.0	2.4 (1.7, 3.1)
Fat percentage	21.7 (9.3)	10.0–48.8	2.6 (1.4, 3.9)
Total body BMC (g)	1005 (394)	467–2139	235 (216, 273)
Total body bone area (cm2)	1152 (328)	636–2072	238 (223, 253)
Total body BMD (g/cm2)	0.85 (0.08)	0.73–1.08	0.04 (0.03, 0.05)
UD radius BMC (g)	0.44 (0.11)	0.24–0.73	0.08 (0.06, 0.09)
UD radius width (cm)	1.83 (0.23)	1.27-2.41	0.22 (0.18, 0.27)
UD radius BMD (g/cm2)	0.24 (0.04)	0.17–0.34	0.01 (0.01, 0.02)
33% radius BMC (g)	0.45 (0.10)	0.27–0.71	0.08 (0.06, 0.09)
33% radius width (cm)	1.05 (0.09)	0.85–1.24	0.05 (0.03, 0.07)
33% radius BMD (g/cm2)	0.43 (0.08)	0.31–0.62	0.06 (0.05, 0.07)
Lumbar spine (L2–4) BMC (g)	17.45 (6.59)	8.27–43.06	4.38 (3.58, 5.18)
Lumbar spine (L2–4) width (cm)	3.43 (0.32)	2.86–4.18	0.35 (0.31, 0.39)
Lumbar spine (L2–4) area (cm2)	25.4 (4.83)	17.54–37.12	4.84 (4.37, 5.30)
Lumbar spine (L2–4) BMD (g/cm2)	0.67 (0.13)	0.47–1.17	0.05 (0.03, 0.07)
Femoral neck BMC (g)	2.81 (0.73)	1.63–4.53	0.56 (0.44, 0.67)
Femoral neck area (cm2)	3.93 (0.39)	3.04–5.10	0.15 (0.04, 0.26)
Femoral neck BMD (g/cm2)	0.71 (0.13)	0.46–1.03	0.11 (0.07, 0.15)
Hip trochanter BMC (g)	3.31 (2.19)	0.62–10.60	1.73 (1.44, 2.01)
Hip trochanter area (cm2)	4.71 (2.49)	1.15 –12.74	2.25 (1.97, 2.52)
Hip trochanter BMD (g/cm2)	0.67 (0.09)	0.48–0.89	0.10 (0.07, 0.12)
Dietary calcium (mg/day)b	625 (245)	154–1149	182 (106, 258)

^aP<0.001 versus baseline (paired t-test) for all variables

^bEstimated by food frequency questionnaire

Table 2 displays results for anthropometry and total body composition as Z scores. While all the children grew well, their height remained low relative to the reference population, although some catch-up in the original height deficit had occurred since baseline. While Z scores for height were positively correlated with age at baseline (r=0.44, P<0.002), this association was no longer significant at follow-up (r=0.26, P<0.08), which suggested that younger study children were shorter relative to their peers than older ones. Additionally, BMI Z scores of the group remained higher than in our reference population (Table 2), and 15 (33%) of the children had BMI values indicative of overweight or obesity at follow-up [29]. Both total body BMC and bone area had increased, so that, although values were lower than in the reference population at baseline, this was no longer the case at follow-up.

Table 2  Baseline, follow-up and 2-year changes in Z scores for anthropometry and whole body composition (n=46)

Characteristic	Baseline	Follow-up	Two-year change
	Mean (SD)	Mean (SD)	Mean (95% CI)
Height (cm)	–0.74 (1.33)aa	–0.39 (1.14)a	0.35 (0.20, 0.51)b
Weight (kg)	0.01 (1.14)	0.18 (1.22)	0.16 (0.06, 0.27)b
BMI (kg/m2)	0.51 (0.90)aa	0.46 (1.00)aa	–0.06 (-0.20, 0.09)
Lean mass (kg)	–0.18 (1.13)	–0.02 (1.08)	0.16 (0.03, 0.28)b
Fat mass (kg)	0.09 (1.03)	0.29 (1.14)	0.19 (0.02, 0.36)b
Total body BMC (kg)	–0.44 (1.11)a	–0.19 (1.06)	0.25 (0.14, 0.37)b
Total body area (cm2)	–0.58 (1.27)a	–0.26 (1.21)	0.32 (0.14, 0.50)b

^aP<0.05, ^aaP<0.01 significantly different from reference population (z-test)

^bP<0.05 significant change from baseline (paired t-test)

Table 3 shows that although areal BMD values had improved significantly in most cortical regions of the skeleton (33% radius, neck of femur and hip trochanter), this pattern was not seen at predominantly trabecular sites (ultradistal radius and lumbar spine). Thus, whereas at baseline the group had significantly lower Z scores than the reference population in the forearm, hip and spine, by follow-up BMD values were lower than in the reference population only at the ultradistal radius and lumbar spine. Moreover, volumetric changes in bone mineral density confirmed this pattern (Fig. 1), with significant improvement in bone mineral apparent density (BMAD) occurring at the 33% radius but not at the lumbar spine. These results occurred in both prepubertal and pubertal children. Table 4 shows that Z scores for regional BMC and bone dimensions tended to increase at most sites, particularly the hip. However, BMC Z scores in the ultradistal radius did not improve and remained below those of the reference population, while spinal BMC did not keep pace with increases in spinal areas.

Table 3  Baseline, follow-up and 2-year changes in Z scores of areal bone mineral density (aBMD) values in different regions of the skeleton (n=46)

aBMD (g/cm2) Z scores	Baseline	Follow-up	Two-year change
	Mean (SD)	Mean (SD)	Mean (95% CI)
UD radius	–0.23 (0.90)	–0.58 (0.97)aa	–0.35 (–0.61, 0.21)b
33% radius	–0.63 (1.33)aa	–0.25 (0.18)	0.38 (–0.10, 0.67)b
Lumbar spine (L2–4)	–0.45 (1.02)aa	–0.66 (1.07)aa	–0.22 (–0.39, –0.05)b
Femoral neck	–1.20 (2.24)aa	–0.34 (1.33)	0.86 (0.20, 1.51)b
Hip trochanter	–0.57 (1.57)a	0.13 (0.86)	0.69 (0.23, 1.15)b
Total body	0.15 (0.78)	–0.13 (0.80)	–0.28 (–0.40, –0.12)b

^aP<0.05, ^aaP<0.001 significantly different from reference population (z-test)

^bP<0.05 significant change from baseline (paired t-test)

Fig. 1  The Z scores for volumetric bone mineral density measurements improved over time in cortical bone (33% radius BMAD) but remained low in trabecular bone (L2–4 BMAD). Values are means (SE), n= 46; P<0.05 or P<0.001 versus the reference population, which has a mean Z score of zero.

Table 4  Baseline, follow-up and 2-year changes in Z scores for BMC, width and area in different regions of the skeleton (n=46)

Z scores	Baseline	Follow-up	2-yr change
	Mean (SD)	Mean (SD)	Mean (95% CI)
UD radius BMC	–0.30 (0.92)a	–0.31 (0.95)a	–0.01 (–0.22, 0.21)
UD radius width	–0.01 (1.17)	0.07 (0.84)	0.09 (–0.21, 0.38)
33% radius BMC	–0.27 (1.17)	–0.05 (1.01)	0.22 (–0.03, 0.48)
33% radius width	0.24 (1.19)	0.21 (0.95)	–0.03 (–0.29, 0.23)
Lumbar spine (L2–4) BMC	–0.16 (0.95)	0.02 (0.95)	0.18 (0.01, 0.34)b
Lumbar spine (L2–4) width	0.07 (1.03)	0.98 (0.96)aa	0.90 (0.70, 1.11)b
Lumbar spine (L2–4) area	0.08 (1.17)	0.76 (0.94)aa	0.67 (0.47, 0.88)b
Femoral neck BMC	–0.59 (1.57)a	0.08 (1.03)	0.51 (0.09, 0.93)b
Femoral neck area	0.52 (0.94)aa	0.19 (0.88)	–0.33 (–0.63, –0.03)b
Hip trochanter BMC	–0.68 (2.60)	0.58 (1.03)aa	1.27 (0.47, 2.07)b
Hip trochanter area	–0.56 (2.48)	0.64 (1.05)aa	1.19 (0.44, 1.95)b

^aP<0.05, ^aaP<0.001 significantly different from reference population (z-test)

^bP<0.05 significant change from baseline (paired t-test)

At follow-up current calcium intakes from all sources were positively correlated with the Z scores for total body BMC (r=0.34, P<0.023), total bone area (r=0.33, P<0.025), ultradistal radial BMD (r=0.36, P<0.014) and 33% radial BMD (r=0.30, P<0.045). Our results indicated that every additional l00 mg of calcium consumed was commensurate with a change of approximately 0.1 unit of the Z score for each of these. Calcium intakes from dairy products alone were also associated with ultradistal radial BMD (r=0.34, P<0.022) but not with any other bone variable, which suggested that this site may be influenced particularly by milk consumption. Total protein intake was correlated with follow-up Z scores for total bone area (r=0.35, P<0.02) and total BMC (r=0.33, P<0.03) but not with any BMD measurement. Total energy intake was correlated with Z scores for follow-up body weight (r=0.34, P<0.03) and BMD at the spine only (r=0.34, P<0.03) and with BMC of the total body (r=0.40, P<0.006), ultradistal radius (r=0.31, P<0.04) and L2–4 spine (r=0.37, P<0.01) but not the Z scores for BMC at the 33% radius or hip.

Measures of physical activity were not associated with Z scores for either total or regional BMC at follow-up, and only at the hip trochanter was there any significant association with BMD (r=0.32, P<0.03).

Discussion

This study documents some improvement over time in the intakes of milk and dietary calcium, total body BMC and skeletal size in our young milk avoiders. Nevertheless, despite some catch-up in height, the group remained short in stature, with BMI values that were higher than the reference population. We observed site-specific persistence of osteopenia at the ultradistal radius and lumbar spine, where Z scores for BMD and BMC or BMAD were significantly lower than in the reference population. Indeed, areal BMD Z scores of the ultradistal radius and lumbar spine had worsened in 2 years, relative to the reference population, which suggests that bone mineral accrual was lagging behind expansion of bone area at these sites. These findings are a concern, given evidence of high fracture rates in young milk avoiders, particularly in the forearm [2]. Forearm fracture rates of children and adolescents have increased sharply in the USA over recent years [30] and nutritionists are worried that falling milk consumption and displacement of milk by carbonated drinks and fruit juices may affect bone health adversely and promote obesity [31, 32].

Milk allergies tend to diminish as children become older [33]. Higher calcium intakes at follow-up may have been related to increased energy consumption due to greater energy requirements of advancing age, as well as to greater selection of calcium-rich foods. However, the calcium intakes of many participants remained poor, perhaps because lifelong patterns of milk consumption are initiated early in life [34, 35]. Our children have had prolonged dairy avoidance, and their increases in calcium intakes have been gradual, which suggest that changes in bone mineralization are not due to sharp changes in remodeling transients [36]. Milk in New Zealand is not supplemented with vitamin D, so effects of avoidance on vitamin D status were not an issue. However, lack of vitamin D is detrimental to bone development [37, 38], and many countries supplement milk with vitamin D. In such countries milk avoidance in childhood could jeopardize vitamin D status importantly.

In milk avoiders osteopenia is associated with low milk intake, not with the presence or absence of adverse symptoms to milk [21, 39]. Children who drink little milk and make no compensatory nutritional adjustments may have intakes of calcium and protein that are insufficient to support optimal height gain and bone growth. Subsequent amelioration of osteopenia over time may be associated, at least in part, with better nutrition and improved intakes of milk, calcium and protein [35].

Correlations between total energy and bone variables may reflect higher calcium intakes associated with greater energy intakes. In our study Z scores for total skeletal BMC, bone area and radial BMD values at both the ultradistal and 33% sites were correlated, albeit weakly, with total current dietary calcium intakes, which suggests that these bore some relationship. In particular, low intakes of calcium from dairy foods were also associated with low ultradistal radius BMD values in the study children. Bone width at the ultradistal radius had expanded significantly faster than width at the 33% radius (2-year increases of 15.3% vs 5.2% at these sites, respectively). BMD values in the ultradistal radius decreased because gains of bone mineral by trabecular bone at this site did not keep pace with increases in bone area. A similar pattern occurred in the spine.

In contrast, the 2-year improvement of BMD at predominantly cortical regions was striking, especially at the 33% radius, where Z scores rose to within the normal range. BMD rose at this site because mineral accretion exceeded bone area increase, which suggests that considerable thickening of the cortical bone within the periosteal envelope had occurred. This may have involved both endosteal and periosteal gains in mineral. When dietary calcium intakes are suboptimal cortical bone may be resorbed to supply the high needs for new mineral at the growth plate [40]. Rising calcium intakes, by inhibiting this resorption and enhancing apposition of bone, could explain the rise of BMD at the 33% radius, where BMD values were positively correlated with total calcium intake.

Although we have no hormonal measurements, we consider it unlikely that changing levels of sex hormones accounted for observed bone changes, because our children were predominantly prepubertal. Rising milk consumption may have improved bone mineralization and increased bone size by supplying essential minerals for new bone [41], altering bone cell function directly [42, 43, 44] and raising plasma levels of insulin-like growth factor-1 (IGF-1) [4]. IGF-1 plays a critical role in augmenting lean mass [45] and stimulating bone elongation and periosteal expansion [46, 47]. Regular dairy consumption during growth is associated with greater height, bone size and bone mineralization [13], whereas low milk consumption is linked to reduced height [41, 48]. Children with cow s milk allergies and food intolerance are shorter than their peers and show no catch-up growth [49, 50, 51]. In children following macrobiotic diets dairy supplementation, but not rising meat consumption, increased height gain [52], which suggests that dairy foods may have special value for early bone development. Our children did exhibit some catch-up in height as they grew older, which suggests that improved nutrition, rather than an inherent genetic predisposition to shortness, may have contributed to this. The children remained heavy for their height, and this was due to greater body fat, not greater lean mass. Though some research indicates that low calcium intakes favor the accumulation of body fat [53, 54], we did not find any association between BMI values and calcium intakes.

Low physical activity may have contributed to the propensity of our young milk avoiders to be overweight and osteopenic. Intermittent load bearing is critically important in stimulating osteogenesis, and randomized control trials demonstrate that increased weight-bearing exercise raises hip BMC and BMD substantially in prepubertal children [55, 56]. Physical activity is hard to assess by questionnaire in young children, and although we found only a weak association at follow-up with our measures of physical activity and hip density, study participants were less active than the reference