点击显示 收起
1 From the Division of Nutrition and Physical Activity (SN-O, KSS, MEC, and CG) and the Division of Diabetes Translation (BAB), National Center for Chronic Disease Prevention and Health Promotion, the CDC Drug Service, Scientific Resources Program, National Center for Infectious Disease (CA and CD), and the NHANES Laboratory, Nutritional Biochemistry Branch, Division of Laboratory Sciences, National Center for Environmental Health, (EWG), Centers for Disease Control and Prevention, Atlanta; the Division of Neonatology, Department of Pediatrics, Medical University of South Carolina, Charleston (BWH); and the Division of Health Examination Statistics, National Center for Health Statistics, Centers for Disease Control and Prevention, Hyattsville, MD (ACL).
2 Address reprint requests to KS Scanlon, Division of Nutrition and Physical Activity, NCCDPHP, Centers for Disease Control and Prevention, 4770 Buford Highway NE (MS-K-25), Atlanta, GA 30341-3717. E-mail: kscanlon{at}cdc.gov.
ABSTRACT
Background: Recent reports of rickets among African American children drew attention to the vitamin D status of these infants and their mothers. African American women are at higher risk of vitamin D deficiency than are white women, but few studies have examined determinants of hypovitaminosis D in this population.
Objective: We examined the prevalence and determinants of hypovitaminosis D among African American and white women of reproductive age.
Design: We examined 1546 African American women and 1426 white women aged 15–49 y who were not pregnant and who participated in the third National Health and Nutrition Examination Survey (1988–1994). Hypovitaminosis D was defined as a serum 25-hydroxyvitamin D concentration 37.5 nmol/L. Multiple logistic regression was used to examine the independent association of dietary, demographic, and behavioral determinants of hypovitaminosis D.
Results: The prevalence of hypovitaminosis D was 42.4 ± 3.1% (
Key Words: Hypovitaminosis D • 25-hydroxyvitamin D • vitamin D deficiency • African American women • diet • body mass index • oral contraceptive pills • rickets
INTRODUCTION
Rickets—a disease characterized by soft and deformed bones—was nearly eradicated in the United States with the vitamin D fortification of milk in the 1930s (1). However, recent reports of rickets among African American children (1–4) drew renewed attention to the vitamin D status of these infants and their mothers.
A newborn's 25-hydroxyvitamin D [25(OH)D] concentration is approximately one-half that of its mother's 25(OH)D (5). A risk factor for poor vitamin D status in early infancy is maternal vitamin D deficiency during pregnancy, resulting in inadequate maternal transfer of vitamin D to the fetus and low infant stores (6–8).
Researchers have noted a lower mean 25(OH)D concentration in African American women than in white women aged 20–40 y (9–11). However, African American women have higher levels of bone density, fewer hip fractures, and longer delays in bone loss after menopause than do white women (9,12). Nevertheless, Perry et al (12) observed that serum 25(OH)D was an independent predictor of femoral bone density in African American women.
The most important source of vitamin D is the skin's synthesis of the vitamin from sunlight (13). Factors that affect cutaneous absorption include use of sunblock, levels of sunlight exposure (eg, season, latitude, and time of day), and skin pigmentation. Indeed, African Americans may be at particular risk for vitamin D deficiency because of their high melanin concentrations (13,14). Vitamin D may also be obtained from dietary sources such as fatty fish, fish oils, fortified foods, and vitamin supplements.
The Institute of Medicine (15) currently sets the adequate intake as 5 µg vitamin D/d (200 IU/d) for women 15–50 y of age. This amount is believed to be adequate regardless of the amount of sunlight exposure. Women of reproductive age are assumed to be able to obtain the recommended intake for almost all vitamins without the use of supplements, and no national organization recommends routine vitamin D supplementation during pregnancy unless a woman is at nutritional risk (16). But in light of differences in serum 25(OH)D concentrations among women of reproductive age by race, the adequate intake may be insufficient for African American women.
We know of no study that has examined the independent association of multiple determinants of vitamin D deficiency among African American women of childbearing age. Using data from the third National Health and Nutrition Examination Survey (NHANES III), we examined the prevalence and determinants of vitamin D deficiency among African American women aged 15–49 y. We also examined the prevalence and determinants of low vitamin D among white women in this age group.
SUBJECTS AND METHODS
Source of study data
NHANES III is a national survey of the US population randomly selected from households in 81 counties across the United States. All survey participants were asked to complete an extensive interview and medical examination, which included providing blood samples. The survey was conducted in 1988–1994 and consisted of 2 phases of equal length and sample size (17). All procedures were approved by the National Center for Health Statistics Institutional Review Board; written, informed consent was obtained from all subjects.
Study sample
For our study, we narrowed the NHANES III cohort to non-Hispanic African American women (n = 2097) and non-Hispanic white women (n = 1860) of reproductive age (15–49 y). Race and ethnicity were self-reported, and none of the participants were institutionalized or housebound. We excluded women with missing data on any of the following variables: serum 25(OH)D concentrations (n = 226), body mass index (BMI; n = 7), milk or cereal intake from the food-frequency questionnaire (n = 18), use of vitamin D supplements (n = 62), and oral contraceptives (n = 39). We also excluded women who were pregnant or whose pregnancy status was unknown (n = 177).
We also excluded women who were taking prescription medications (n = 388) known to influence calcium, vitamin D, or phosphorus metabolism: steroids and corticosteroids; anticoagulants; anticonvulsants; fluoride; anesthetics used for recent immobilization; antihypertensive drugs such as ß-blockers, diuretics, and calcium channel blockers (18); and medications for liver, kidney, gastrointestinal absorption, or endocrine disorders. Women missing medication data were also excluded (n = 68). Women taking vitamin and mineral supplements or oral contraceptive pills were not excluded because we wanted to examine the independent effects of these substances on serum 25(OH)D concentrations (19–22).
Our final sample included 1546 African American women and 1426 white women. These subjects were statistically similar to excluded women in age, obesity (BMI >30.0), diet (food-frequency questionnaire), and use of supplemental vitamin D or oral contraceptives.
Outcome measures
We used the serum 25(OH)D concentration to assess the vitamin D status of our study population. In NHANES III, serum 25(OH)D concentration was measured by a 2-step radioimmunoassay procedure (DiaSorin, formerly INCSTAR test kit, catalog no. 68100; INCSTAR Corp, Stillwater, MN; 23).
The lower limit of the normal range for serum 25(OH)D is 20.0–37.5 nmol/L (15); we use the term hypovitaminosis D to indicate a concentration of 37.5 nmol/L (15 ng/mL). Studies in older white populations showed that parathyroid hormone (PTH) concentrations rise when 25(OH)D concentrations fall below 37.5 nmol/L, providing support for this cutoff in determining vitamin D adequacy (24–27). A study of 35 patients aged 49–83 y indicated that PTH concentrations rise when 25(OH)D concentrations fall below 50 nmol/L (28), and a study of 1569 French adults aged 35–65 y found that serum intact PTH began to increase when serum 25(OH)D concentrations were 78 nmol/L (29). Norman (30), in a review of the study by Harris and Dawson-Hughes (10) of African American women aged 20–40 y, suggested that a rise in intact PTH above normal occurs when the 25(OH)D concentration falls below 30 nmol/L. However, we used the cutoff of 37.5 nmol/L for hypovitaminosis D because a larger sample would be needed to more precisely determine the association between serum 25(OH)D and intact PTH concentrations among the women in the study by Harris and Dawson-Hughes.
Variables
Several individual characteristics and environmental factors directly affect vitamin D concentrations, most particularly diet and exposure to sunlight. For the NHANES III nonquantitative food-frequency questionnaire, survey participants were asked to recall how often (never or the number of times per day, week, or month) a food was consumed over the past month. Participants could also respond that they "didn't know." Data from this questionnaire allowed us to estimate the usual intake of fortified milk and breakfast cereals.
For milk consumption, one question was asked: "How often did you have milk to drink or on cereal? Do not count small amounts of milk added to coffee or tea." The Food and Drug Administration standard of identity for milk states that fortified milk should contain 400 IU vitamin D/qt. Fortified milk labels read 100 IU vitamin D/8-oz serving, but studies consistently showed that the vitamin content of fortified milk is highly variable (31–33). We categorized the frequency of milk consumption as 0, >0 to <3, or 3 times/wk. We did not include responses to the question "How often did you have chocolate milk and hot cocoa?" in our frequency of milk consumption data because most brands of instant hot cocoa do not contain any vitamin D–fortified milk and only 6% of participants consumed these products 3 times/wk. Furthermore, a preliminary analysis summing the responses on the milk and chocolate milk–hot cocoa questions did not change our risk estimates.
Four questions were used to assess consumption of breakfast cereal. Most breakfast cereals are fortified with 40–50 IU vitamin D/28 g serving (34). Survey participants were asked how often they ate 1) "All-Bran [Kellogg's, Battle Creek, MI], All-Bran Extra Fiber [Kellogg's], 100% Bran [Post, Battle Creek, MI], and Fiber One [General Mills, Minneapolis]"; 2) "Total [General Mills], Product 19 [Kellogg's], Most, and Just Right [Kellogg's]"; 3) "all other cold cereals like corn flakes, Cheerios [General Mills], Rice Krispies [Kellogg's], and presweetened cereals"; and 4) "cooked, hot cereals like oatmeal, cream of wheat, cream of rice, and grits." We summed the responses to these 4 questions to determine overall frequency of cereal consumption, which we categorized as 0, >0 to <3, or 3 times/wk.
The amount of vitamin D consumed as a vitamin supplement was assessed with the use of specific information about the product and intake level obtained by asking, "Have you taken any vitamins or minerals in the past month?" and "Please include those that are prescribed by a doctor and those that are not prescribed." The participant was asked to bring each supplement to the medical examination; the interviewer recorded the name of the product, the manufacturer's name and address, how often the product was taken over the previous month (number of times per day, week, or month, or other frequency specified by the subject), and how much was taken each time (number of capsules, teaspoons, tablespoons, fluid ounces, drops, packets, milliliters, wafers, or amount specified by the subject). Daily intake of supplemental vitamin D over the past month was then calculated, which we categorized as 0, >0 to <200, 200 to <400, or 400 IU/d.
NHANES III did not provide direct data on participants' exposure to ultraviolet B (UVB) rays. Our preliminary analysis indicated that region and season were highly associated with exposure because the survey was conducted in the South in the winter and in the North in the summer. For this reason, we retained season (winter, spring, summer, and fall) for analysis in the final model. As a surrogate for environmental pollution and smog, which can interfere with UVB exposure (35), we also dichotomized the residence of each participant as urban or rural.
We also looked at cigarette smoking, high BMI (as an indicator of obesity), and use of oral contraceptives, because previous studies indicated these factors may be associated with hypovitaminosis D. Brot et al (36), found that smoking significantly affected vitamin D metabolism, but in our preliminary analysis we found that smoking was not associated, and we did not include this factor in the final model. Wortsman et al (37) observed that obesity was associated with low serum 25(OH)D concentrations and secondary hyperparathyroidism in their study. We calculated each subject's BMI as measured weight (in kilograms) divided by measured height (in meters) squared and categorized BMI according to National Heart, Lung, and Blood Institute guidelines (38): underweight (<18.5), normal (18.5 to <25.0), overweight (25.0 to <30.0), or obese (30.0). We included oral contraceptives because numerous studies showed that women who are consuming oral contraceptives have higher concentrations of serum 25(OH)D than do women who are not (19–22).
Finally, we examined the age and poverty income ratio of survey participants to assess potential associations with serum 25(OH)D concentrations. Age was self-reported, and we categorized it into 15–24, 25–34, and 35–49 y. Each subject's poverty income ratio was classified as <1.85, 1.85–3.00, or >3.00. Preliminary analysis indicated that neither age nor poverty income ratio was associated with vitamin D status, so we did not include these variables in the final model.
Statistical analysis
We used sample weights in all statistical analyses, and we used SUDAAN version 7.5.4A (39) to account for the complex sample design of the NHANES III. We examined the mean and SE for serum 25(OH)D concentrations among African American women and white women separately by diet, vitamin D supplement use, season, location of residence, BMI, and use of oral contraceptives. We also calculated the prevalence, odds ratios, and 95% CIs of hypovitaminosis D for each variable by race. Stratified analysis and multiple logistic regression were used to assess which determinants had an independent effect on hypovitaminosis D.
We assessed season and residence location (urban, rural) for interaction because we thought an association between season and low 25(OH)D concentration would be less apparent among women living in areas where they were less prone to UVB exposure (ie, urban settings). We found no significant interaction between these 2 variables, however. Furthermore, we found no significant interactions between season and race or between residence location and race. We did not believe seasonal UVB exposure would interact significantly with other factors that are known to or that theoretically could affect 25(OH)D concentrations, so we did not assess such interactions.
RESULTS
Among African American women, the mean serum 25(OH)D concentration was 44.2 ± 1.1 nmol/L. The prevalence of hypovitaminosis D (37.5 nmol/L) was 42.4 ± 3.1%; 12.2 ± 1.7% of African American women had a concentration of <25 nmol/L. Among white women, the mean serum 25(OH)D concentration was 82.5 ± 1.5 nmol/L. The prevalence of hypovitaminosis D was 4.2 ± 0.7%, and 0.5 ± 0.2% had a concentration of <25 nmol/L.
Among both African American women and white women, mean 25(OH)D concentrations were higher with consumption of milk or cereal 3 times/wk, the fall season, rural residence, and use of oral contraceptives (Table 1). For both races, mean concentrations were lowest during the spring months and highest during the fall months and were lower in women residing in urban than in rural areas. Mean serum 25(OH)D concentrations were also lower in women with a BMI 30.0 (obese) than in women whose BMI was normal (18.5 to <25.0). Among both groups of women, mean serum 25(OH)D concentrations were higher in those who reported current use of oral contraceptive pills than in nonusers.
View this table:
TABLE 1 . Mean serum 25-hydroxyvitamin D concentrations by select characteristics for African American and white women aged 15–49 y in the third National Health and Nutrition Examination Survey1
For African American women, we observed significant independent associations between hypovitaminosis D and all variables in the final model (Table 2). Compared with no consumption of milk or cereal, consumption 3 times/wk gave an odds ratio for hypovitaminosis D of 0.5 and 0.6, respectively. Compared with no vitamin D from supplements, consumption of 200 to <400 IU/d from supplements gave an odds ratio of 0.5 and consumption of 400 IU/d gave an odds ratio of 0.1. Women examined in the fall, winter, or spring were 2.4–3.6 times as likely as women examined in the summer to have hypovitaminosis D. Women residing in urban locations were more likely than those residing in rural locations to have this condition (odd ratio: 1.7), as were women with a BMI <18.5 (underweight) compared with women whose BMI was 18.5 to <25.0 (odds ratio: 2.3). Use of oral contraceptives decreased the likelihood of hypovitaminosis D (odds ratio: 0.6 compared with no oral contraceptives).
View this table:
TABLE 2 . Prevalence and adjusted odds ratios (ORs) for low serum 25-hydroxyvitamin D [25(OH)D] concentration (37.5 nmol/L) by select characteristics for African American and white women aged 15–49 y in the third National Health and Nutrition Examination Survey1
For white women, only 3 determinants in the final model were significantly associated with hypovitaminosis D (Table 2). Compared with no consumption of vitamin D in the form of supplements, consumption of 200 to <400 IU/d gave an odds ratio for hypovitaminosis D of 0.3. Women examined during the winter were 5.4 times as likely to have hypovitaminosis D as were women examined in the summer. Women with a BMI 30.0 (obese) were 3.3 times as likely to have hypovitaminosis D as were women with a normal BMI (18.5 to <25.0). Although the associations of other variables were not significant, the magnitude and direction of all associations except BMI were similar to those among African American women. The lack of statistical significance for some variables is likely related to inadequate statistical power as a result of the small number of white women with hypovitaminosis D.
DISCUSSION
Our study is unique in that it examined the independent associations of multiple determinants of hypovitaminosis D among African American and white women of reproductive age in a nationally representative sample. Our findings suggest that 10 times as many African American as white women are vitamin D deficient. Several other studies found lower mean serum 25(OH)D concentrations in African American women than in white women (9–11). Thomas et al (26) examined determinants of low 25(OH)D concentrations among older, predominantly white, hospitalized patients and found that vitamin D intake, winter season, and being housebound were independent predictors of hypovitaminosis D. We found that vitamin D use and season—as well as residence, BMI <18.5, and use of oral contraceptives—predicted hypovitaminosis D among African American women.
Correlations between serum 25(OH)D concentrations and dietary intake of vitamin D were reported previously for African American women (9) and white women (22). We found that hypovitaminosis D was common even among African Americans who consumed the adequate intake for vitamin D through supplements. Among women who consumed 200 to <400 or 400 IU vitamin D/d, 30.1 ± 3.0% and 10.6 ± 6.0%, respectively, were deficient. Among an overall 243 African American women who consumed 200 IU vitamin D/d, 28.2 ± 2.7% were deficient. These results suggest that 200 IU/d may not meet these women's needs. Further analysis showed that even among African Americans who consumed 200 IU vitamin D/d in supplements, milk 3 times/wk, and cereal 3 times/wk (n = 52), the prevalence of hypovitaminosis D was 19.4 ± 6.0%.
We assessed exposure to UVB radiation by season and location of residence. In our study, among African American and white women, mean serum 25(OH)D concentrations were lowest in spring and highest in fall. Previous studies reported the lowest mean concentrations in winter and the highest in summer (9,10). The lower mean 25(OH)D we observed for urban areas than for rural areas may have been due to less outdoor activity among urban residents or to greater environmental pollution in urban settings. Persons living in polluted environments may have decreased cutaneous synthesis of vitamin D, thereby placing them at risk of deficiency (35).
In our study, a BMI 30.0 was significantly associated with hypovitaminosis D in white women but not in African American women. An association between obesity and hypovitaminosis D was reported previously for young white adults (37,40,41). Wortsman et al (37) suggested that obese persons have decreased bioavailability of vitamin D from cutaneous and dietary sources because of a tendency of vitamin D to deposit in adipose tissue. If this hypothesis is true, then the higher proportion of lean body mass in African American women than in white women of similar weight-for-height status (42) may account for the lack of association between high BMI and hypovitaminosis D we noted among African American women. An alternative explanation was proposed by Bell et al (40), who suggested that the vitamin D endocrine system in obese persons is altered, with increased production of 1,25-dihydroxyvitamin D exerting negative feedback control on the hepatic synthesis of 25(OH)D. Epstein et al (43) found that obese African Americans did not show additional changes in the vitamin D endocrine system over alterations already present in nonobese African Americans.
In agreement with previous studies (19–21), we observed a higher mean serum 25(OH)D concentration in oral contraceptive users than in nonusers of both races. In addition, among African American women, use of oral contraceptives was associated with a lower prevalence of hypovitaminosis D. The estrogen component of oral contraceptives may alter the relative proportion of free and protein-bound 25(OH)D by increasing concentrations of vitamin D binding protein (20,22).
Our findings are subject to some limitations. The assessment of vitamin D intake was restricted. We chose not to use data from the 24-h food recall available in NHANES III because these data may misrepresent usual individual intakes (44). The food-frequency questionnaire in NHANES III should represent usual frequency of intake over the past month for the 2 most commonly consumed vitamin D–containing foods (milk and breakfast cereals), but it does not provide quantitative data on portion size or vitamin D content, nor does it include all vitamin D– containing foods (eg, fatty fish and fish oils). Daily vitamin D supplement intake was estimated indirectly from total monthly intake; longer-term and short-term use of this dose of supplement could not be distinguished. Although many determinants of hypovitaminosis D were examined, we did not have direct data on outdoor activity, clothing habits, or use of sunscreen. Finally, PTH and 1,25-dihydroxyvitamin D concentrations were not measured in NHANES III, and therefore we are unable to provide information on these calciotropic hormones.
For practical reasons, data for NHANES surveys were collected in the winter in the South and in the summer in the North. Serum 25(OH)D data did not include women living in the North during the winter, an important risk group for vitamin D deficiency. Two other high-risk groups that were not examined were institutionalized individuals (because of survey design) and homebound persons (serum vitamin D was not one of the assays performed on this group). Differences in the geographic distribution of African American subjects and white subjects made it difficult to directly compare the rates of hypovitaminosis D in these groups, but stratification by race showed a higher prevalence of hypovitaminosis D among African Americans in each season and by location of residence, indicating that racial differences were not accounted for by differences in residence. In addition, the interaction between season and residence was not significant in either group.
Nonresponse bias is a potential limitation, but this bias was reduced somewhat because the data of those who received physical examinations in the mobile examination centers were adjusted by a nonresponse adjustment factor (17). Approximately 7.8% of African American women and 4.6% of white women who came to the mobile examination center did not have data on 25(OH)D concentration; this bias was not addressed by sample weight adjustments, but women with missing data for determinants added to our models had concentrations of serum 25(OH)D that were similar to those included in the final study sample.
In summary, we found that nearly one-half of African American women aged 15–49 y had a serum 25(OH)D concentration of 37.5 nmol/L. Even among African American women who consumed 200 IU vitamin D/d from supplements, 30% had a low concentration. By contrast, the overall prevalence of hypovitaminosis D among white women was one-tenth that of African Americans. Our data suggest further examination of dietary recommendations for women at risk of hypovitaminosis D. Furthermore, the American College of Obstetricians and Gynecologists does not recommend a daily prenatal vitamin supplement unless the adequacy of the woman's diet is questionable or she is at high nutritional risk (16). Our data suggest that nearly one-half of the African American women in the United States could enter pregnancy with serum 25(OH)D 37.5 nmol/L. Our data further highlight important determinants of hypovitaminosis D among both African American and white women. These data should be considered when advising women on dietary and supplemental intakes of vitamin D.
REFERENCES