Racial Differences in the Effect of Early Milk Consumption on Peak and Postmenopausal Bone Mineral Density

Authors

  • Alexander R Opotowsky,

    1. Department of Medicine, College of Physicians and Surgeons, Columbia University, New York, New York, USA
    Search for more papers by this author
  • John P Bilezikian MD

    Corresponding author
    1. Department of Medicine, College of Physicians and Surgeons, Columbia University, New York, New York, USA
    2. Department of Pharmacology, College of Physicians and Surgeons, Columbia University, New York, New York, USA
    • Department of Medicine Division of Endocrinology College of Physicians and Surgeons 630 W. 168th Street New York, NY 10032, USA
    Search for more papers by this author

  • The authors have no conflict of interest.

Abstract

Studies of the effect of milk consumption on bone density have focused on white women. The applicability to other races is unclear. This study of 4316 women found a racial difference in the effect of early milk consumption on bone. Further study is required to evaluate the etiology of this difference.

Introduction: Early milk consumption has been found to influence bone mineral density positively throughout life. Previous studies leading to this finding focused on white women; there are no data to support the view that black women also gain an equivalent benefit from early milk consumption.

Materials and Methods: Using data from the Third National Health and Nutrition Examination Survey (NHANES III), we analyzed the relationship between reported childhood and teenage milk consumption and current hip bone mineral density at four sites (femoral neck, trochanter, intertrochanter, and total hip) among postmenopausal and 20- to 39-year-old non-Hispanic white (NHW) and black (NHB) women. Potential confounding variables were accounted for in the analysis.

Results: Controlling for age and body mass index, teenage milk consumption was significantly associated with higher bone mineral density at all four sites among both postmenopausal and 20- to 39-year-old white women, but was not associated with bone mineral density at any site among black women. A similar pattern was observed for childhood milk intake. The interaction between race and teenage milk consumption significantly explained postmenopausal bone mineral density at all sites, whereas the interaction between race and teenage milk consumption was significant at two of the four sites among the 20- to 39-year-old women.

Conclusion: The results show racial differences in the effect of early milk consumption on peak and postmenopausal bone mineral density.

INTRODUCTION

Osteoporosis affects 6–7 million women and results in 1.5 million fractures in the United States each year.(1,2) Several risk factors for low bone mineral density (BMD) have been described. Conversely, a number of protective characteristics have been identified. Calcium intake, especially before achieving peak bone mass, is a determinant of bone mass throughout life. Milk contributes over one-half of the calcium to the average American child's (2–11 years old) diet. All dairy products together provide ∼70% of childhood calcium intake.(3) The effects of dairy products on calcium metabolism are not limited to their high calcium content. Some dairy products also contain significant quantities of sodium and protein, both of which have been associated with urinary calcium loss.(4) Because milk has a higher ratio of calcium to sodium and protein than other dairy products, it is reasonable to consider that milk may influence bone density more positively than other dairy products.(5)

Milk consumption in children and teenagers has been associated, in many cross-sectional and a few prospective studies, with enhanced bone mineral acquisition, higher peak bone mass, and higher bone density in adulthood.(6–12) The effect of milk consumption by older women on bone mass is less well defined, but some studies have found a positive impact.(8,10) All these studies excluded blacks. The relationship between early milk consumption and BMD among black women has yet to be explored.

For several reasons, the results of studies suggesting increased bone mass in white women with high childhood and teenage milk consumption may not apply to black women. Despite clear similarities, there are significant differences between black and white women in the factors governing mineral metabolism and bone mass. Black women have significantly higher bone mass than whites.(13–15) Black women also have lower bone turnover rates, higher parathyroid hormone (PTH) levels, lower 25-hydroxyvitamin D levels [25(OH)D], and lower urinary calcium excretion than white women.(14–18) In addition, the high prevalence of lactose intolerance among blacks may impair their ability to benefit from diets high in lactose-containing products such as milk.(19) Undoubtedly, there are other undefined racial differences in calcium metabolism between white and black women. There is clearly a paucity of research on racial differences in calcium metabolism, especially dietary calcium's effect on bone mass. The National Institutes of Health consensus statement on optimal calcium intake noted the dearth of information on calcium requirements among different ethnic groups and recommended further research in this field.(20)

The current study explores the differential effects of childhood and teenage milk consumption on hip BMD in white and black postmenopausal and 20- to 39-year-old women.

MATERIALS AND METHODS

Data source

The Third National Health and Nutrition Examination Survey (NHANES III) was conducted between 1988 and 1994 by the National Center for Health Statistics, Centers for Disease Control and Prevention, to assess the health and nutritional status of the civilian non-institutionalized population of the United States. NHANES III data are freely available for analysis.(21,22) The survey was based on a highly stratified, multistage, probability sample design. Sampling weights were developed to account for the complex survey design, survey nonresponse, and survey noncoverage.(23)

For the postmenopausal women, we selected all non-Hispanic postmenopausal (i.e., reported no menstrual period for the previous 12 months) women who responded to questions regarding childhood and teenage milk consumption. Women were excluded if information about weight, height, age at menopause, or current calcium intake data were missing. Bone densitometry had to be technically satisfactory for inclusion in the study. NHANES III included 2086 non-Hispanic white (NHW) postmenopausal women, but 167 had technically unacceptable DXA tests, 1 had incomplete weight data, 149 were unsure of the date of their last menstrual period, 58 and 77 had incomplete teenage and childhood milk consumption information, respectively, and 58 had incomplete calcium data. The survey also included 953 non-Hispanic black (NHB) postmenopausal women, but 124 had technically unacceptable DXA tests, 1 had incomplete weight information, 118 were unsure of the date of their last menstrual period, 18 and 32 had incomplete teenage and childhood milk consumption information, respectively, and 22 had incomplete calcium intake data. Among black women, those excluded were younger (p < 0.001), and they were less likely to have used oral contraceptives (p = 0.003), hormone replacement (p < 0.001), or to have been physically active in the previous month (p = 0.005). A similar pattern held for excluded white women. They were younger (p < 0.001), less likely to have used oral contraceptives (p < 0.001) or hormone replacement (p = 0.01), and less likely to have been physically active in the previous month (p = 0.03). Excluded white women were also more likely to have breastfed (p = 0.02). Other variables, including level of childhood and teenage milk consumption, were comparable between excluded and included women. The white postmenopausal women ranged in age from 22 to 90 years, and the black women ranged from 21 to 90 years of age. For both racial groups, the great majority of postmenopausal women were at least 40 years old (95.8% NHW, 91.6% NHB). Overall, 1675 NHW postmenopausal women (weighted to represent 25,350,000 women) and 735 NHB postmenopausal women (weighted to represent 2,990,000 women) were included to evaluate teenage milk consumption in this study. The corresponding figures for childhood milk consumption were 1656 (25,080,000) and 725 (2,950,000).

We also selected 20- to 39-year-old NHW and NHB women. Of the 1070 NHW women, 143 had technically unacceptable DXA tests, whereas 219 of the 1249 NHB women's DXA tests were unacceptable. In addition, 18 NHW and 30 NHB women were excluded because of incomplete current calcium intake data. One white woman and two black women had incomplete teenage milk consumption information, while five white women and six black women had incomplete information on childhood milk consumption. Excluded white women tended to be younger (28.4 versus 30.2 years old, p = 0.001) and were less likely to have used hormone replacement (0% versus 2.5%, p = 0.005) than white women included in the study. There were no differences in current calcium intake, body mass index (BMI), oral contraceptive use, smoking, drinking, physical activity, or childhood or teenage milk intake between the excluded and included white women. Among the 20- to 39-year-old black women, those excluded were less likely to use oral contraception (71.8 versus 83.3%, p < 0.01) and had higher calcium intake (687 versus 588 mg, p < 0.05). There were no differences between the excluded and included black women in terms of age, BMI, hormone replacement use, smoking, drinking, physical activity, or childhood or teenage milk intake. Overall, 908 20- to 39-year-old NHW women (weighted to represent 24,070,000 women) and 998 20- to 39-year-old NHB women (weighted to represent 4,370,000 women) were selected for the teenage milk consumption part of this study, whereas 904 NHW women (weighted to represent 23,970,000 women), and 994 NHB women (weighted to represent 4,360,000 women) were included to evaluate the effect of childhood milk consumption.

Bone density measurements

Bone densitometry of the hip was performed on all nonpregnant NHANES III participants 20 years of age or older. Bone density was measured at five hip sites: femoral neck, trochanter, intertrochanter, Ward's triangle, and total hip. As recommended by the NHANES III documentation and the International Society for Clinical Densitometry, our study did not use the data for Ward's triangle.(1) All measurements were performed using the Hologic QDR 1000 DXA machine (Hologic, Waltham, MA, USA). A rigorous quality control program was maintained throughout the survey as documented by Wahner et al.(24) All DXA tests were reviewed individually at the Mayo Clinic (Rochester, MN, USA) for technical evaluation before inclusion in the database.

Dietary intake, anthropometric measures, and lifestyle characteristics

NHANES III evaluated dietary intake using a 24-h dietary recall interview, and specific nutrient values were estimated. As a separate part of the NHANES III interview, participants were asked how much milk they drank as children (5–12 years old) and as teenagers (13–17 years old). Participants chose from six levels of milk intake for each age range: none, less than one glass per week, one glass per week, between one glass per week and one glass per day, one glass per day, and more than one glass per day. Because of potential misclassification and because subtle differences in milk intake are unlikely to have significant clinical effects, we stratified the responses into three categories: less than 1 glass of milk per week (low), at least 1 glass of milk per week but no more than 1 glass of milk daily (intermediate), and greater than 1 glass of milk per day (high). The breakdown of milk consumption in the intermediate group was similar for blacks and whites both in the 20- to 39-year-old group and among postmenopausal women. For example, 9.7% of NHW and 9.1% of NHB postmenopausal women consumed one glass per week, 28.3% of NHW and 27.1% of NHB consumed between one glass per week and one glass per day, and 62.0% of NHW and 63.8% of NHB consumed one glass of milk per day as teenagers. The figures were similar for childhood milk consumption, with a higher proportion drinking one glass of milk daily (71.4% overall: 71.5% NHW, 71.3% NHB).

Participants were defined as having a history of oral contraceptive pill use if they had ever taken birth control pills for any reason and as having a history of hormone replacement therapy if they had ever taken estrogen or female hormone pills other than oral contraceptive pills or had used female hormone patches. Women were defined as having a history of breastfeeding if they had breastfed any of their children for at least 1 month. Participants were categorized as having engaged in some form of physical activity in the previous month if they reported one of the following activities: jogging/running, bicycling, swimming, calisthenics/exercising, aerobics or aerobic dancing, gardening/yard work, lifting weights, or any other exercises/sports.

Participants were defined as having smoked if they smoked 100 or more cigarettes during their lifetime. Participants were categorized as having a history of heavy alcohol use if they reported that they consumed daily five or more alcoholic drinks regularly during their lifetime. Race was defined as black or white using both self-report and family questionnaires. Participants were asked their race (white, black, other), and ethnicity (Mexican-American, Other Hispanic, Not Hispanic). This study included all women who were designated as either white or black, but were not Hispanic.

Statistical analysis

We analyzed two populations: postmenopausal and 20- to 39-year-old women. The methods outlined below were applied to each population individually. Multiple linear regression analysis was performed on each study population (including both NHW and NHB) to determine if there was a racial difference in the effect of early milk consumption on BMD. Models included a categorical race variable, an interaction term between race and milk consumption, and potentially confounding variables. For the postmenopausal women, we repeated the interaction analyses, excluding potentially biased groups (previous diagnosis of osteoporosis, a history of hip fracture, or had taken thyroid hormones, anticonvulsants, systemic corticosteroids, or drugs primarily affecting calcium metabolism in the previous month). We further analyzed the effect of early milk consumption on BMD for each race individually. These analyses used the same multiple linear regression models, excluding the race and interaction terms. The least-square means method was used to calculate the mean (±SEM) BMD, by race, for given levels of childhood and teenage milk consumption at each site.

The multivariate analyses for the 20- to 39-year-old women included most of the covariates from the postmenopausal model. Years since menopause was excluded for obvious reasons, as was a history of breastfeeding. Current calcium intake was excluded because calcium intake in young adulthood is likely to correlate highly with early milk consumption. Because many studies have shown that early calcium consumption has a much greater effect on BMD than young adult calcium intake, we felt that including this variable would obscure real results, not control for a potential confounder. All analyses were repeated controlling for age, controlling for both age and BMI, and controlling for the full multivariate model. Because the results of the first two models were similar, only the results of the age and BMI model are reported.

Baseline differences in clinically relevant variables between milk consumption groups were analyzed using the χ2 test and Student's t-test. All analyses were performed using SAS-callable SUDAAN 8.0 (Research Triangle International, Research Triangle Park, NC, USA), using the appropriate sample weights and taking into account the stratified survey design.

RESULTS

Postmenopausal women

Descriptive statistics for postmenopausal women, by race and milk consumption, are shown in Table 1. For both races, those who reported drinking more milk during childhood had significantly higher current calcium intake. Black women in the higher milk intake groups were more likely to have used hormone replacement. For white women, those with higher milk intake were more likely to have breastfed a child and were more likely to have been physically active in the prior month. Higher teenage milk consumption also correlated with higher calcium intake, more physical activity, and breastfeeding among white women, in addition to a higher probability of using hormone replacement. Reported teen milk intake was not correlated with calcium intake among the postmenopausal black women, but was associated with higher probability of breastfeeding. There were also differences between the NHB and NHW postmenopausal women. NHB women had a higher prevalence of heavy alcohol use (8.1% versus 5.2%, p = 0.03), and breastfeeding (48.7% versus 41.0%, p = 0.007), but a lower prevalence of physical activity in the past month (51.8% versus 70.0%, p < 0.001) and hormone replacement use (29.3% versus 42.9%, p < 0.001). Blacks were younger by 4.6 years (p < 0.001), had been postmenopausal for a shorter period of time (by 2.3 years, p = 0.002), and had a 2.7 kg/m2 higher BMI (p < 0.001). Black women's current daily calcium intake was significantly lower than white women's (527 versus 687 mg, p < 0.001). Of note, black women in the highest childhood or teenage milk intake category had even lower current daily calcium intake than whites in the lowest milk intake group (for childhood, 519 versus 581 mg; for teenage, 548 versus 561 mg). We controlled for all of these variables in the multivariate analyses reported below, and therefore, these specific differences cannot account for our results. However, this widespread and significant variation highlights the complexity of studying racial differences. Factors unimportant in investigations of relatively homogenous populations may significantly confound studies of diverse populations.

Table Table 1. Clinical Characteristics by Childhood and Teenage Milk Consumption of White and Black Postmenopausal Women, Mean (SEM), Weighted Unless Noted*
original image

To determine if race modified the effect of early milk consumption on postmenopausal BMD, we used a regression model including an interaction term between race and reported milk intake. The results are shown in Table 2. Controlling for age and BMI, the teenage interaction term was significant at all sites (p < 0.01–0.04). This association was attenuated in the full multivariate model, but the interaction term remained significant at three of the sites (p = 0.02–0.03) and was of borderline significance at the femoral neck (p = 0.05). While the interaction term did not reach statistical significance at any site for childhood milk consumption in the limited model, it did approach significance (p = 0.06–0.11) at three of the four sites (total hip, trochanter, and intertrochanter) in the multivariate model. These results suggest that, in terms of postmenopausal BMD, whites benefit more from teenage, and possibly childhood, milk consumption than blacks do.

Table Table 2. Regression Coefficients of BMD on Reported Milk Consumption During Childhood and Teenage Years for Postmenopausal Women*
original image

All interaction analyses were repeated, excluding potentially biased groups as noted previously. The changes in statistical significance were minimal. While interactions at five sites became slightly less significant, the p values at all affected sites were still <0.10. Conversely, some borderline significant interactions (0.05 < p < 0.10) became more significant (p < 0.05) after excluding potentially biased groups. Excluding users of systemic corticosteroids enhanced the significance of the teenage milk intake by race interaction at the femoral neck (p < 0.05) and the childhood milk intake by race interaction at the total hip (p < 0.05). Excluding women who reported a previous hip fracture also improved the significance of the childhood milk intake by race interaction at the total hip (p < 0.05).

Racial subgroup analysis was performed to investigate the effect of early milk consumption on postmenopausal BMD for each race individually. The regression coefficients of multiple linear regression models of postmenopausal BMD on reported teenage milk consumption controlling for age and BMI are shown in Table 2. At all sites, white women with the highest milk intake had significantly higher BMD than those in the lowest category of milk intake, and there was a significant linear trend toward higher BMD with higher milk intake at all sites. Interestingly, among black women childhood and teenage milk consumption had a negative, albeit statistically insignificant, effect on postmenopausal BMD. Once again, controlling for the nine additional variables listed previously (see right side of Table 2) lessened the statistical significance of these findings, but the higher milk intake groups still tended to have higher BMD among white women (all p < 0.10). The linear effect became less pronounced, and there seemed to be a plateau effect of milk consumption, with similar BMD values for the intermediate and highest milk intake groups. The effect of childhood milk consumption on BMD among white women was somewhat less clear than the effect of teenage milk consumption. The values of the coefficients for childhood milk consumption were similar to those seen with teenage milk consumption, but the statistical significance was lower, likely because of the smaller number of subjects consuming little or no milk during childhood. Only two sites showed a significant linear relationship among whites (trochanter and total hip) after controlling for age and BMI. At all sites, however, white women with the highest milk intake had significantly higher BMD than those in the lowest group. Once again, the signs of the coefficients for whites and blacks were reversed, with milk consumption having a positive effect on BMD among white women and a negative, although statistically insignificant, effect among black women.

Least-square means were calculated for white and black postmenopausal women by reported teenage milk consumption and are shown in Fig. 1. For white women, the mean BMD, adjusting for covariates, seems to increase with higher milk intake, whereas this is not the case among the black women.

Figure FIG. 1..

Least-square means of bone mineral density, by race and teenage milk consumption, for postmenopausal women at the (A) total hip, (B) femoral neck, (C) trochanter, and (D) intertrochanter with approximate regression lines. The analysis controlled for age, BMI, hormone replacement use, oral contraceptive use, smoking, alcohol consumption, physical activity, years since menopause, history of breastfeeding and pregnancy, and current daily calcium intake. Among the postmenopausal women, 208 of 1675 white women and 87 of 735 black women consumed less than one glass of milk per week as teenagers. p values represent the significance of the race by teenage milk intake interaction for each site. Error bars represent SEM.

Younger women (20–39 years old)

We repeated our analyses on the 20- to 39-year-old women participating in NHANES III. Descriptive statistics for this population, by race and childhood and teenage milk intake, are shown in Table 3.

Table Table 3. Clinical Characteristics by Childhood and Teenage Milk Consumption of White and Black 20- to 39-Year-Old Women, Mean (SEM), Weighted Unless Noted*
original image

There were no racial differences in the prevalence of heavy alcohol use (black, 6.4% versus white, 6.3%, p = 0.95), hormone replacement use (1.6% versus 2.5%, p = 0.31), or oral contraceptive use (83.3% versus 85.3%, p = 0.31) between black and white women in this population. However, there were other racial differences. Blacks had a lower prevalence of physical activity in the past month (69.5% versus 84.6%, p < 0.001) and smoking (38.9% versus 52.0%, p < 0.001). Blacks were younger by 0.7 years (p = 0.02) and had higher BMI (27.5 versus 24.7, p < 0.001). Black women's current daily calcium intake was significantly lower than white women's (588.4 versus 788.0 mg, p < 0.001).

The interaction between childhood milk consumption and race was not significant at any site in either the limited or full multivariate models, but the corresponding interaction for teenage milk consumption trended toward significance (p < 0.10) at all hip sites in both models and at p < 0.05 at two of four sites, after controlling for age and BMI, and at one site in the full multivariate model (see Table 4). White women in the highest milk intake group had significantly higher BMD than women in the lowest intake group at all sites. Similar to our results from postmenopausal women, milk consumption had no significant effect on BMD at any site analyzed among the black women.

Table Table 4. Regression Coefficients of BMD on Reported Milk Consumption During Childhood and Teenage Years for 20- to 39-Year-Old Women*
original image

Least-square means were calculated for white and black women by reported teenage milk consumption and are shown in Fig. 2. For white women, the mean BMD, adjusting for covariates, at each of these sites is significantly higher among those drinking more milk. There is no such difference among the black women. Once again, white women seems to benefit from drinking more milk as teenagers, whereas blacks do not. Of note, blacks had higher peak and postmenopausal BMD at all sites, independently of childhood or teenage milk consumption status.

Figure FIG. 2..

Least-square means of bone mineral density, by race and teenage milk consumption, for 20- to 39-year-old women at the (A) total hip, (B) femoral neck, (C) trochanter, and (D) intertrochanter with approximate regression lines. The analysis controlled for age, BMI, oral contraceptive use, smoking, alcohol consumption, history of pregnancy, and physical activity. Among the 20- to 39-year-old women, 85 of 904 white women and 102 of 998 black women consumed less than one glass of milk per week as teenagers. p values represent the significance of the race by teenage milk intake interaction for each site. Error bars represent SEM. *p < 0.05 vs. lowest milk intake group.

DISCUSSION

Blacks have lower calcium intake during both childhood and adulthood than whites.(25) However, despite chronically lower calcium intake than white women, black women have a much lower prevalence of osteoporosis and osteopenia.(1,26) Until the results of this study, however, there was no information in blacks on the relationship between childhood and teenage milk consumption and BMD in their adult years. The results of this analysis indicate that the expected positive relationship between early milk consumption and BMD is not demonstrable among black women.

Our results show a significant racial difference in the effect of early, especially teenage, milk consumption on peak and postmenopausal bone density. As demonstrated by previous studies, white women benefit significantly from childhood and teenage milk consumption, but our observations suggest that black women derive no such benefit. Several plausible mechanisms to explain these observations are worthy of comment.

Lactose intolerance is common among blacks. Milk consumption in lactose intolerant individuals may decrease transit time, thereby reducing calcium absorption and the absorption of other nutrients essential for skeletal health. However, multiple studies have demonstrated that lactose-intolerant individuals absorb calcium from milk normally, or at least sufficiently.(27–29) Furthermore, there is specific evidence that lactose-intolerant black female adolescents adapt to high-lactose diets with a reduction in symptoms associated with lactose consumption.(30) However, milk consumption over a prolonged or developmentally critical period of time among lactose-intolerant subjects may affect calcium absorption or metabolism in entirely different ways from the published reports. Additionally, investigation of calcium absorption in lactose-intolerant individuals has focused on white women, raising the possibility that blacks with lactose intolerance may adapt differently.

Genetic diversity may play a role in explaining the demonstrated racial differences. For example, polymorphisms in the vitamin D receptor are thought by some to be associated with a marked effect on bone density among white women(31) and may affect calcium absorption among children.(32) The same polymorphisms, while present in a large proportion of black women, do not seem to be correlated with bone density among black women.(31,33) Undoubtedly, other genetic distinctions will be of interest when they are elucidated.

Dietary interactions between calcium and other minerals and nutrients(34–36) are unlikely to account for the negative correlation between early milk consumption and bone density in black women, but differences in calcium homeostasis may provide a clue toward explaining our results. Black children and teenagers seem to absorb calcium more efficiently.(37,38) Furthermore, blacks retain more calcium, as demonstrated by lower 24-h urinary calcium excretion.(14–18) Vitamin D and PTH have a different physiology in blacks, reflecting a picture similar to secondary hyperparathyroidism. Serum 25(OH)D lower and serum 1,25-dihydroxyvitamin D and PTH concentrations are lower among blacks.(14–16,18) Nevertheless, bone turnover is lower among blacks.(39) Lower bone turnover and greater calcium retention may tend to negate the importance of low calcium intake among blacks. Alternatively, high calcium intake early in life may interfere with the development of this highly efficient calcium conservation metabolism, and thus put these black women at greater risk for age-related and postmenopausal bone loss. Of course, milk is a complex food and contains nutrients besides calcium, which may affect bone mass, such as phosphorus, protein, and vitamin A. Plausibly, racial variation in the responsiveness to or metabolism of these or other constituents in milk might play a role in explaining these results.

Milk consumption in childhood may not only directly help in the accrual of peak bone mass. It may also be beneficial by establishing a pattern of milk consumption that lasts throughout life. Our results support this assertion. Milk consumption in childhood correlates well with milk consumption and calcium intake later in life for both black and white women. Nevertheless, it is interesting to note that despite much lower average calcium intake, black women have higher bone density than white women. This suggests that calcium intake in general may not play as pivotal a role in bone health for black women as it does for whites. Other physiological factors, some of which have been listed, may be dominant and thereby limit the importance of early milk intake among blacks on bone mass.

This study has several limitations. The history of milk consumption was self-reported, and for the postmenopausal women, were recalled behaviors of 45–55 years before the survey. Nevertheless, distant dietary recall has consistently proven superior to current diet as a predictor of historical diet.(40–44) Furthermore, recall of milk intake correlates well with historical intake, with correlation coefficients ranging from 0.47 to 0.58. In addition to recall history, recall bias is another potential limitation of our study. In this case, however, a recall bias would have to correlate milk consumption with BMD differentially by race, positively for whites and negatively for blacks, to produce the reported results. Bone density testing was in its infancy at the time of this study and few women in the population had previously undergone bone density testing. To affect the results, the few women previously tested would have had to remember their bone densities and associate these values with early milk consumption, a most unlikely scenario.

While we were able to control for concurrent potentially confounding variables, we could not address associations between early milk consumption and potentially confounding factors during childhood or teenage years. For example, childhood milk consumption may be positively associated with childhood physical activity in one racial group but negatively correlated in the other. If these associations are consistent throughout life, our results would be valid, but there is no way to test this hypothesis using the NHANES III data set.

BMD is the most widely used and best validated predictor of fracture risk, but it admittedly does not give a perfect measure of bone strength. Bone turnover, microarchitecture, geometry, and size may all play significant and independent roles in determining bone strength. It is possible that milk intake has a disproportionately larger effect on these other properties among black women. Such changes may not necessarily be reflected in higher bone density but would certainly improve the biomechanical characteristics of bones. Thus, while our results show racial variation in the effect of milk on bone density, these findings cannot be directly related to bone strength or fracture risk. Nor do they infer that milk doesn't have important effects on these other parameters in black and white women.

A final noteworthy limitation refers to the possibility that the classification of regular milk drinkers differed between blacks and whites. It is conceivable that white children in the highest category of milk consumption ingested much more milk than the black children in the same, highest, category of milk consumption. This concern would also have to assume the ingestion of milk, and therefore calcium, in the black children, even in this highest intake category, was below the threshold for optimal calcium absorption.

This study has several major advantages over previous studies addressing this issue. First, the sample size is large, and the survey population, as appropriately weighted, is representative of the entire United States population. Previous studies recruited subjects from smaller geographic areas, limiting the generalizabilty of their results. Second, NHANES III used rigorous quality control measures to ensure the validity of BMD measurements. In fact, these data were used to develop the universal reference values for hip BMD measurements. Third, data on multiple potential confounding factors were collected, allowing us to control for all six important confounders noted by Weinsier.(5)

Furthermore, two adult populations were used to support the validity of our findings: young and postmenopausal women. Any confounding factor, such as childhood physical activity, unaccounted for in the analyses above, would have to be similar for these two generations of women. These consistent results suggest that the racial difference in response to early milk consumption seen in postmenopausal women have their origins much earlier and persist throughout life.

In conclusion, there are racial differences in the effect of early milk consumption on peak and postmenopausal bone density. Our results do not diminish the well-documented and essential role of adequate calcium intake in a variety of skeletal health issues. In fact, this study supports an important role for early milk and calcium consumption in promoting lifelong bone health among white women. While the physiological mechanisms underlying the differences among whites and blacks remain to be identified, the public health implications are clear. Universal recommendations for dairy product consumption are based on incomplete data and should be reconsidered pending rigorous investigation of the effects of dairy products on bone health among racial and ethnic minorities.

Acknowledgements

We are grateful to Beverly Diamond for assistance with the statistical aspects of this study, Raymond Arons for advice regarding the analysis of NHANES III, and Robert P Heaney for insightful comments on this manuscript. ARO designed the study, analyzed the data, and wrote the first draft of the manuscript. JPB participated in the analysis and interpretation of the data, provided intellectual input, and helped in the writing of the paper.

Ancillary