Low serum vitamin D concentrations (<20 ng/mL) have been linked to hip fractures, at least in white women.1 There is little known about the relationship between circulating vitamin D and fracture in other race/ethnic groups. The prevalence of vitamin D levels less than 20 ng/mL has been reported to be three times higher in blacks than in whites.2, 3 Other racial groups, including Hispanics and Asians, have lower levels of circulating 25(OH)D.4, 5 The higher parathyroid hormone (PTH) levels observed in blacks may contribute to accelerated bone loss with aging, but small experimental studies suggest that blacks may have skeletal resistance to the bone-resorbing effects of PTH.6, 7
Vitamin D–binding protein (DBP) or group-specific component (Gc) binds to and transports vitamin D to target tissues to maintain calcium homeostasis.8 DBP has a high plasma concentration compared with 25(OH)D, with little variability observed by season.9 To our knowledge, little is known about ethnic difference in DBP and the association between DBP and fractures. One study showed that the genetic effect of the DBP gene on fracture risk depended on other environmental risk factors, for example, calcium intake.10
The objective of this analysis was to examine the association between circulating 25(OH)D, PTH, and DBP with fracture in white, black, Hispanic, Asian, and Native American women enrolled in the Women's Health Initiative Observational Study (WHI-OS).
The study population comes from the WHI-OS, a prospective cohort study that enrolled 93,676 women aged 50 to 79 years from 1994 to 1998 at 40 US clinical centers. Study methods have been described in detail elsewhere.11 Briefly, women were eligible if they were postmenopausal, unlikely to move or die within 3 years, not enrolled in the WHI Clinical Trials, and not currently participating in any other clinical trial. The study was approved by the human subjects review committees at each participating institution. All participants provided written informed consent.
Information on race or ethnicity was obtained by self-report and included whites (not of Hispanic origin), blacks (not of Hispanic origin), Hispanics/Latinos (Mexican, Cuban, Puerto Rican, Central American, or South American), Asians or Pacific Islanders (Chinese, Indo-Chinese, Korean, Japanese, Pacific Islander, or Vietnamese), Native American (including Alaskan natives), or other. All covariates were ascertained at baseline.
Women were sent questionnaires annually to report any hospitalization and other specific outcomes, including fractures. We collected information on all fractures that occurred after study entry except those of the fingers, toes, face, skull, and sternum. Hip fractures were verified by review of medical records and confirmed by blinded central adjudicators.12 Information on other clinical fractures was obtained by self-report. In the WHI, 80% of self-reported nonhip fractures were confirmed by physician review of medical records, suggesting that the self-report of such fractures is reasonably accurate.13 For this analysis, we excluded patients with hip fractures owing to cancer and women reporting clinical spine or coccyx fractures because of poor reliability.12
Nested case-control design
This study is a case-control study nested within the prospective design of the WHI-OS. All self-reported cases of fracture in the WHI-OS as of August 2006 (the last date of original follow-up) were selected as potential cases. We excluded women reporting “other” or “unknown” race/ethnicity, current hormone therapy at baseline, and use of bisphosphonates, selective estrogen receptor modulators (SERMs), or tamoxifen. We also excluded women with missing important covariates such as weight, height, history of fracture, and less than two vials of unthawed serum or less than 0.25 mL of EDTA plasma. A total of 7116 possible cases and 37,713 controls were eligible after these exclusions. We chose all fracture cases among nonwhites, including 381 black, 192 Hispanic, 113 Asian, and 46 Native American women. A random sample of 400 fracture cases was chosen from a total of 6384 fractures among the white participants.
One control was selected for each case from the risk set corresponding to the time of the case's event. All participants (cases included) were part of the risk set until they had an event or until their last recorded visit (censored if lost to follow-up). Controls were matched to cases by age at screening (±1 year), race/ethnicity (white, black, Hispanic, Asian, Native American), and blood draw date (±90 days). All matching factors were weighted equally during selection. Controls were chosen to have a fracture-free follow-up time at least as long as the cases' event times. Seventeen unmatched cases during the first match attempt were matched subsequently by age at screening (±2 years), race/ethnicity, and blood draw date (±365 days).
In summary, there were 1132 cases and 1132 controls for a total of 2264 women. The average follow-up time from enrollment to the last follow-up date as of August 31, 2006, was 8.60 (SD = 1.61) years among cases and 8.65 (SD = 1.61) among controls.
Ancillary grant funding was obtained in 2008 to measure the biochemical measures. A 12-hour fasting blood sample was obtained at the baseline visit, processed, and stored at −80 °C according to strict quality-control procedures.14 Previously unthawed serum samples were sent to have 25(OH)D, PTH, and DBP concentrations measured at the Reproductive Endocrine Research Laboratory at the University of Southern California. For 25(OH)D, a radioimmunoassay was used with DiaSorin reagents (Stillwater, MN, USA). The sensitivity of the assay was 3.75 nM. The interassay coefficients of variation (CV, %) were 11.7%, 10.5%, 8.6%, and 12.5% at 14.0, 56.8, 82.5, and 122.5 nM, respectively.
Plasma PTH was measured using the Scantibodies (Santee, CA, USA) whole PTH(1–84) specific assay kit, designed for the quantitative determination of whole PTH without cross-reaction with PTH(7–84) fragments.15 This is an immunoradiometric (IRMA) assay using polyclonal PTH(1–84) antibodies, one that tends to bind in the N-terminal regions (label antibody) and one in the C terminal region (capture antibody) of PTH(1–84), respectively. The use of these antibodies guarantees that only whole PTH is detected using this whole PTH assay; values greater than 39 are considered indicative of hyperparathyroidism. The interassay CVs were 2.3% and 1.30% for the low (39 pg/mL) and high (359 pg/mL) quality control samples, respectively. The sensitivity of the assay was 2 pg/mL.
DBP was measured with a sandwich-type enzyme-linked immunosorbent assay (ELISA) using reagents from a commercial kit (ALPCO, Salem, NH, USA). The assay is intended for the quantitative determination of free and non actin-bound DBP. It uses a polyclonal DBP antibody that is coated on microtiter plate wells and a polyclonal peroxidase-labeled DBP antibody that is added after addition of different concentrations of the standard and aliquots of serum samples. The interassay CV was 8.8%, and assay sensitivity was 1.2 ng/mL.
Participants were excluded if their vitamin D, PTH, or DBP level was greater than 3 standard deviations (SDs) above the mean. The final sample included 390 white cases and controls, 379 black cases and controls, 191 Hispanic cases and controls, 112 Asian cases and controls, and 44 Native American cases and controls.
All covariates were ascertained at baseline and included education, living arrangements, personal history of fracture, parental history of fracture after age 40, reproductive and medical history, alcohol intake, cigarette smoking, physical activity, general health status, diet, and falls. Current use of prescription medications was recorded by clinic interviewers by direct inspection of medicine containers. Prescription names were entered into the WHI database and assigned drug codes using Medispan software (Wolter-Kluwer Health, Indianapolis, IN, USA). Dietary intakes of calcium and vitamin D were assessed using a semiquantitative food-frequency questionnaire.16 We also collected information on the use of calcium and vitamin D supplements (ie, amount and frequency of intake).
Physical activity was classified on the basis of frequency and duration of walking and mild, moderate, and strenuous activities in the prior week. Kilocalories of energy expended in a week were calculated (metabolic equivalent [MET] score = kcal-hours/week/kg).17 Physical function was measured using the 10-item RAND-36 physical function scale.18 We compared women with a score of greater than 90 versus 90 or less; this cutoff corresponded to the median score. Weight was measured on a balance-beam scale and height using a stadiometer. BMI was calculated as weight (kg)/height (m2).
We used chi-square and t tests to compare baseline characteristics between cases and matched control participants. For continuous measures that were skewed (eg, PTH), Wilcoxon rank-sum tests were used. Separate analyses were carried out within each race/ethnicity. To further assess confounding, we compared baseline characteristics across categories of 25(OH)D (<20 ng/mL, 20 to <30 ng/mL, and ≥30 ng/mL)19 in the combined cohort of cases and controls. For normally distributed variables, a test of linear trend was performed by treating vitamin D clinical cutoffs as category integer-valued steps. For nonnormally distributed or skewed variables, the Jonckheere-Terpstra test of trend was performed. The Cochran-Armitage test for trend was used for dichotomous variables.
We assessed the association between serum 25(OH)D, PTH, and DBP concentrations and incident fracture in conditional logistic regression models, models appropriate for the matched case-control design. We calculated odds ratios (ORs) and 95% CIs from the conditional logistics models across categories of 25(OH)D (<20 ng/mL, 20 to <30 ng/mL, and ≥30.0 ng/mL) and across quartiles of PTH, DBP, and the ratio 25(OH)D/DBP. Quartile cutoffs were determined from the distribution among control individuals. We defined the lowest category as the referent group.
For 25(OH)D, we first examined the unadjusted association. For the multivariable adjusted model, weight, height, physical activity, total calcium intake, and personal history of fracture were forced into the model. Backwards elimination was used, and the following variables were removed from the model: alcohol, education, smoking, health status, diabetes, past hormone therapy use, glucocorticoid use, and parental history of fracture. The final multivariable model included weight, height, physical activity, calcium intake, and history of fracture. Based on examination of Lowess smoothers for fracture rates versus 25(OH)D curves, threshold effects were evaluated by identifying potential inflection points using a single change-point linear spline model. We performed a test of equality versus difference of slopes to determine the 25(OH)D concentration where the slopes above and below the cut point were significantly different and most different. If more than one cutoff was identified (p < 0.05), we chose the 25(OH)D value with the lowest p value as the threshold. Waist circumference was added to the multivariable model to test whether the effect of 25(OH)D on fracture is mediated by body fat. For PTH, DBP, and 25(OH)D/DBP multivariable models, we also used backwards elimination to identify covariates and forced inclusion of covariates weight, height, physical activity, total calcium intake, and personal history of fracture. To explore whether the results are modified by other calcitropic hormones, we subsequently adjusted the 25(OH)D models for PTH and DBP, the PTH models for 25(OH)D and DBP, and the DBP models for 25(OH)D and PTH.
The average age of the subjects varied across race/ethnic groups and ranged from 62.6 years among black women to 66.2 years among white women (Table 1). Among white women, fractures cases were more likely to report diabetes, a positive personal and parental fracture history, and have a higher waist circumference. Among black women, fracture cases were more likely to report past smoking, diabetes, and a positive personal fracture history. Hispanic cases were more likely to report a previous fracture and have a higher waist circumference than control individuals. Among Asian women, women with a fracture were more likely to report fair to poor health status, although the difference did not reach statistical significance (p = 0.06). There were no significant differences between cases and controls among Native American women except a borderline significant lower history of past hormone therapy among the cases.
Table 1. Baseline Characteristics of Fracture Cases and Controls by Ethnicity
Among white women, the percentage of those with fair or poor health status was lower with increasing serum concentrations of 25(OH)D (Table 2). Physical function, physical activity, past use of hormone therapy, and calcium intake were higher with increasing concentrations of 25(OH)D while the average body mass index (BMI) and waist circumference were lower. Both the mean and median PTH was lower among women with 25(OH)D ≥30 ng/mL. There was a modest relationship with DBP: women with 25(OH)D D ≥30 ng/mL had about a 6% higher DBP than women in the deficient range 25(OH)D <20 ng/mL.
Table 2. Baseline Characteristics Across Categories of 25(OH)D Within Each Ethnic Group
Statistical tests for linear trend across category of 25(OH)D.
Age, mean (SD), years
66.5 ± 7.1
66.2 ± 7.4
65.8 ± 7.7
62.4 ± 7.6
63.5 ± 8.0
62.0 ± 8.1
63.5 ± 7.5
62.1 ± 7.3
62.5 ± 7.3
64.4 ± 8.0
65.5 ± 7.0
65.5 ± 8.0
62.6 ± 8.0
64.7 ± 6.4
65.1 ± 5.5
Health status, fair to poor, %
Parental history of fracture, % yes
RAND 36-Physical Functioning Score > 90, %
Physical activity, median (range), METs/week
Hormone therapy use, past, %
BMI, mean (SD), kg/m2
29.1 ± 6.5
27.2 ± 5.2
25.6 ± 4.3
30.9 ± 6.8
29.6 ± 5.6
29.6 ± 6.7
29.4 ± 6.6
28.5 ± 7.1
28.7 ± 5.2
25.3 ± 5.2
24.9 ± 5.0
23.5 ± 6.9
29.9 ± 6.8
28.9 ± 7.9
28.7 ± 5.5
Waist circumference, median (SD), cm
89.4 ± 14.1
84.9 ± 12.3
81.5 ± 11.0
92.1 ± 14.3
90.2 ± 13.2
86.7 ± 13.3
88.7 ± 13.0
85.7 ± 12.5
86.7 ± 13.5
81.1 ± 12.5
78.8 ± 11.4
75.6 ± 10.8
90.9 ± 14.7
88.9 ± 17.1
86.7 ± 14.1
Total calcium intake, median (IQR), mg/day
784.9 (563 –1107)
PTH, median (IQR), pg/mL
PTH, mean (SD), pg/mL
22.2 ± 10.0
19.1 ± 11.2
18.0 ± 8.5
25.0 ± 12.4
21.1 ± 10.4
20.1 ± 14.0
21.4 ± 8.6
19.9 ± 9.3
19.8 ± 9.1
20.6 ± 11.1
17.9 ± 7.9
16.6 ± 8.4
23.4 ± 9.7
17.2 ± 12.1
25.2 ± 36.9
DBP, median (SD), mg/dL
25.3 ± 7.7
26.3 ± 7.9
26.8 ± 6.4
24.9 ± 5.7
26.1 ± 9.8
26.8 ± 6.5
25.0 ± 5.3
27.2 ± 10.8
27.2 ± 5.6
23.8 ± 6.2
24.7 ± 5.5
26.8 ± 5.4
26.2 ± 5.9
26.3 ± 3.7
25.6 ± 4.6
Among black women, the patterns generally were the same, although the associations were not consistently statistically significant. Physical activity and calcium intake were higher with increasing 25(OH)D level, whereas BMI and waist circumference were lower. The mean and median PTH decreased and DBP increased with increasing 25(OH)D.
We saw similar patterns of higher 25(OH)D with higher physical activity among Hispanic women. Of note, there was little evidence of an association between BMI and 25(OH)D in Hispanic women, although the average waist circumference tended to be lower with higher 25(OH)D (p = 0.12). A higher proportion of Hispanic women with 25(OH)D levels of less than 20 ng/ml reported a parental history of fracture. PTH tended to decrease with increasing 25(OH)D, although it was borderline significant; DBP increased with increasing 25(OH)D.
For Asian women, the patterns were similar for physical activity, past use of hormone therapy, BMI, waist circumference, PTH, and DBP. Among Native American women, the patterns were similar, although they were not consistently significant.
25(OH)D and fracture risk
Among whites, the mean and median 25(OH)D levels were lower among women with a fracture compared with controls (Table 1). About 39% of fracture cases had vitamin D levels of less than 20 ng/mL compared with 31% of controls. Among the other race/ethnic groups, the mean, median, or distribution of 25(OH)D did not differ between cases and controls (Table 1). The proportion of women with vitamin D levels of less than 20 ng/mL was greatest among blacks and Native American; approximately 60% of these women had 25(OH)D levels of less than 20 ng/mL. The mean or median PTH did not differ by case/control status in any ethnic group (Table 1). Black women tended to have the highest PTH levels and Asian women the lowest. There also was no difference in DBP by case/control status in any ethnic group, and levels were similar across ethnicity. Of interest, the ratio of 25(OH)D to DBP differed significantly across ethnic groups (p < 0.001). The ratio was approximately 1.0 in white and Asian women but was lower in blacks, Hispanics, and Native Americans. Within ethnic groups, there was no difference in this ratio by case/control status.
The association between 25(OH)D and fractures is summarized in Table 3. White women with higher 25(OH)D levels (≥30.0 ng/mL) had a reduced risk of fracture (p trend = 0.008). In multivariable adjusted models, women with 25(OH)D levels of 30.0 to 65.1 ng/mL had a 44% lower risk of fracture (OR = 0.56, 95% CI 0.35–0.90) than women with 25(OH)D levels of less than 20 ng/mL. Further adjustment for waist circumference, PTH, or DBP had little effect on these results.
Table 3. Odd Ratio (95% CI) of Fracture Across Categories of 25(OH)D Within Each Ethnic Group
Categories of Vitamin D
20 to <30 ng/mL
MV = multivariable model: weight, height physical activity, total calcium intake, and history of fracture. Cases and controls matched on age, race/ethnicity, and date of blood draw.
In contrast, black women with higher 25(OH)D levels had a higher risk of fracture (p trend = 0.043). In the multivariable models, the ORs of fracture were 1.48 (95% CI 1.05–2.10) and 1.33 (95% CI 0.73–2.43) in women with 25(OH)D levels of 20 to less than 30 ng/mL and 30 ng/mL or greater, respectively, compared with women with a 25(OH)D level of less than 20 ng/mL. Black women with a 25(OH)D level of 20 ng/mL or greater, compared with Black women with a circulating 25(OH)D level of less than 20 ng/mL had a statistically significant 45% increased risk of fracture (OR = 1.45, 95% CI 1.06–1.98). Further adjustment for waist circumference and PTH had little effect on these results. The addition of DBP attenuated the association slightly (p trend = 0.059).
There was no association between 25(OH)D and fracture in Hispanic women. Native American women with a 25(OH)D level greater than 20 ng/mL had a lower risk of fracture than women with a 25(OH)D level of less than 20 ng/mL, but this association was not significant.
The risk of fracture tended to increase with increasing 25(OH)D level in Asian women, but the results were not significant in multivariable models or after adjusting for waist circumference or PTH. However, adjustment for DBP markedly increased the OR. Asian women with a 25(OH)D level of 30.0 ng/mL or greater had an OR of 2.78 of fracture (95% CI 0.99–7.80) compared with women with a 25(OH)D level of less than 20 ng/mL. The trend of increasing fracture risk with increasing 25(OH)D level was statistically significant in models including DBP.
In the spline analysis, there was no evidence of a threshold in white women (Fig. 1). The association was linear across the range of 25(OH)D levels in this study. For black women, there was a significant threshold around 17 ng/mL, where the risk of fracture above this level increased with increasing 25(OH)D level.
There was no association between PTH and fracture in any ethnic group (Table 4). Further adjustment for 25(OH)D or DBP had no effect on these results.
Table 4. The Association Between PTH Quartiles and Fracture in Different Ethnic Groupsa
0.19–14.01 pg/mL (n = 574)
14.02–19.00 pg/mL (n = 546)
19.01–28.85 pg/mL (n = 570)
25.86–141.65 pg/mL (n = 542)
Quartile cutoffs were based on the distribution in the controls. Cases and controls matched on age, race/ethnicity, and date of blood draw.
Adjusted for weight, height, physical activity, calcium, and history of fracture, which were forced into the model.
White women with the higher DBP levels tended to have a lower risk of fracture, but the trend was not significant (Table 5). On the other hand, black women with higher DBP levels tended to have an increased risk of fracture in multivariable adjusted models (p trend = 0.047). Asian women with higher DBP levels tended to have a reduced risk of fracture, but results were not significant. There was no association between DBP and fracture in Hispanic or Native American women. There also was no consistent pattern between the ratio of 25(OH)D to DBP and fracture in any group (data not shown).
Table 5. The Association Between Vitamin D–Binding Protein Quartiles and Fracture in Different Ethnic Groupsa
2.48–21.53 mg/dL (n = 560)
21.54–25.17 mg/dL (n = 561)
25.18–28.99 mg/dL (n = 547)
29.0–54.40 mg/dL (n = 564)
Quartile cutoffs were based on the distribution in the controls. Cases and controls matched on age, race/ethnicity, and date of blood draw.
Adjusted for weight, height, physical activity, calcium, and history of fracture, which were forced into the model.
In our prospective, nested case-control study, we found that white women with 25(OH)D levels between 30.0 and 65.1 ng/mL at study entry had a significantly reduced risk for fracture during the next 8 years compared with women with the lowest 25(OH)D levels (<20 ng/mL). The overall trend between increasing 25(OH)D level and lower risk of fracture was statistically significant with no evidence of threshold. This association was independent of many factors, including body weight, total calcium intake, waist circumference, parental history of hip fracture, and PTH and DBP levels.
In contrast, among black women, those with higher 25(OH)D levels (≥20 ng/mL) had an increased risk of fracture. The overall trend between 25(OH)D level and increasing fracture was statistically significant in multivariable models adjusting for the important covariates identified earlier. Our spline analyses suggested a significant threshold for black women at 17 ng/mL, where women with a 25(OH)D level above this threshold had a significantly increased risk of fracture.
There was no association between 25(OH)D level and fracture among Hispanic women. Results in Asian women were similar to the results among black women, where higher values of 25(OH)D were associated with an increased risk of fracture, but these results were not statistically significant until we adjusted for DBP.
For Native American women, higher levels of 25(OH)D were associated with a lower risk of fracture, but these results were not significant, probably reflecting low power because we were limited to 44 fracture cases and controls, and only about one-third of both cases and controls had a 25(OH)D level above 20 ng/mL.
Our results in white women are consistent with our previous study of 25(OH)D level and hip fracture in the WHI, where low levels of 25(OH)D were associated with an increased risk of hip fracture.1 In this analysis, we were unable to examine hip fracture because there were few hip fractures among the ethnic minority groups (blacks, n = 16; Hispanics, n = 4; Asians, n = 4; and Native American, n = 3). Nevertheless, our findings extend the protective association of 25(OH)D to include not only hip fractures but also all nonspine fractures at least in white women. In the Osteoporotic Fractures in Older Men Study, we found a protective association between 25(OH)D level and hip fracture, but there was no association between 25(OH)D level and all nonspine fractures in men.20 To our knowledge, only one previous study examined the association between 25(OH)D level and all fractures and found no association.21 However, the authors of that study did not stratify by gender, and hence it is possible that the association been 25(OH)D level and nonhip fractures may differ in men and women.
The target range for 25(OH)D has been proposed to be 30 ng/mL or higher22 by some experts, although the recent Institute of Medicine (IOM) report noted that a level greater than 20 ng/mL is needed for bone health in most individuals.23 Our findings, consistent with our previous reports on hip fracture in predominately white women1 and men,20 suggest that it is only individuals with 25(OH)D levels of less than 20 ng/mL who have an increased risk of fracture. Thus our findings, at least in white women, are consistent with the current recommendations by the IOM.
The results in black women are divergent from those in white women and need replication in other cohorts of black women. If our results are confirmed, it suggests that optimal 25(OH)D levels may differ by race/ethnic group. Previously, it was thought that the lower risk of osteoporotic fracture among blacks, despite their lower serum 25(OH)D levels, was a paradox.24 Our results suggest that this may indeed not be a paradox but a rather a contributing factor for their lower risk of fractures. The threshold for optimal 25(OH)D among blacks may be lower than for whites. If true, then clinical interventions to correct 25(OH)D levels in blacks for fracture prevention may not be appropriate and even may be counterproductive. It is also interesting to note that other investigators have found a different relationship between 25(OH)D and diabetes in blacks.4 Finally, lower (<20 ng/ml) and higher (≥30 ng/ml) levels of 25(OH)D among older white women both were associated with higher odds of frailty.25 It is possible that race/ethnic differences in frailty could have contributed to our differential findings in whites and blacks. However, although physical functioning scores were lower in blacks than in whites, there was no evidence that they differed by case/control status or across 25(OH)D levels in blacks. Further adjustment for physical function had no effect on our results.
On the other hand, if our results are not confirmed, what else might account for our findings in black women? We considered selection bias: black women in WHI with higher 25(OH)D levels have different characteristics than black women with low 25(OH)D levels. For example, black women with 25(OH)D levels of 30 ng/mL or greater reported higher physical activity levels, weighed on average 4 kg lower, and reported a total daily calcium intake 160 mg higher than black women with 25(OH)D levels of less than 20 ng/mL. We adjusted for all these factors in our models. However, inherent to observational studies such as ours, unmeasured factors may have caused residual confounding and could have contributed to our findings.
We relied primarily on self-report of fractures, and black race was shown previously to be a predictor of unconfirmed self-reported fractures.13 It is possible that there may have been underascertainment of fractures in women with lower 25(OH)D levels, but there is no evidence to suggest that ascertainment differed across 25(OH)D levels. We found the expected associations between self-reported fractures with established risk factors for fracture, including history of smoking, diabetes, and personal history of fracture, suggesting that self-report of fracture is reasonably accurate in black women.
Estimates of European ancestry in African Americans range from 20% to 25%.26, 27 Significant associations have been reported between the proportion of European ancestry and BMD.26 Blacks with a higher proportion of European ancestry have lower BMD values. Thus it is possible that the black women in our study with higher 25(OH)D levels have a greater degree of European admixture, which accounts for both their higher 25(OH)D levels and their higher risk of fracture. There also could be other genetic factors that contribute to this finding.
Results among Asian women were similar to those among black women, where Asian women with higher 25(OH)D levels had a higher risk of fracture, but this association was not statistically significant until we adjusted for DBP. This is a bit surprising because we found no difference in DBP by case/control status in any ethnic group. Furthermore, the 25(OH)D/DBP ratio was about 1.0 in Asians, similar to the ratio in white women, suggesting no excess of DBP between these groups.
We found no associations between PTH and fracture in any ethnic group. This result is consistent with a recent meta-analysis of 905 hip fractures cases and 924 controls from population-based case-control studies.28 In our study, PTH levels were within the normal range, and there may be no association with fracture within this normal range. Only 5% of women had clinically elevated levels of PTH, limiting our power to see an association among these women.
Our study has several limitations. Although we matched each case to a control whose blood draw date was within 90 days, seasonal variability in 25(OH)D concentrations may have confounded our results. We could not test whether the association between low 25(OH)D level and fracture was independent of BMD because only three WHI clinics measured BMD. Our study included few women with 25(OH)D concentrations greater than 30 ng/mL, so we could not test whether even higher concentrations influence fracture risk. Finally, although we included all clinical fracture cases in the minority groups, we had limited power in Asians, Hispanics, and Native Americans.
There are also a number of strengths to our study. Most previous studies of ethnic differences compared a single ethnic/race group with whites. We studied a large, racially diverse sample of well-characterized women from five different race/ethnic groups. We matched on age, race/ethnicity, and blood draw date. We adjusted for a comprehensive set of risk factors, including clinical factors found to be associated with fractures and 25(OH)D level in each ethnic group.20 We also adjusted for PTH and DBP levels.
In conclusion, our results suggest divergent relationships between 25(OH)D level and fracture by race/ethnicity. Higher levels of 25(OH)D were associated with a reduced risk of fracture among white women but an increased risk among black women. Further research to confirm and elucidate the basis of these findings could enhance our understanding of ethnic differences in fracture. Our results suggest that the optimal level of 25(OH)D for skeletal health may differ in white and black women.
JAC has received research support and consulting fees from Novartis Pharmaceuticals. DCB has received research support from Novartis, Amgen, and Merck. JEM and colleagues at Brigham and Women's Hospital, Harvard Medical School, are recipients of funding from the National Institutes of Health to conduct the VITamin D and OmegA-3 TriaL (VITAL), a large-scale randomized trial of vitamin D and omega-3s in the prevention of cancer and cardiovascular disease. All the other authors state that they have no conflicts of interest.
The Women's Health Initiative program is funded by the National Heart, Lung and Blood Institute, National Institutes of Health, US Department of Health and Human Services, through contracts N01WH22110, 24152, 32100-2, 32105-6, 32108-9, 32111-13, 32115, 32118-32119, 32122, 42107-26, 42129-32, and 44221. The sponsor played a role in the design and analysis of the WHI. Additional funds were obtained for these assays and analyses under contract N01-WHI-7-4318.
Authors' roles: Study concept and design: JAC and JW-W. Acquisition of data: JAC, JEM, JW-W, JMS. Analysis and interpretation of data: JAC, RB. Drafting of the manuscript: JAC. Critical revision of the manuscript for important intellectual content: MED, RB, KEB, KE, JEM, JW-W, RJ. Statistical analysis: RB, MJH, DCB, KEB. Obtained funding: JAC, MED, JW-W, RJ. Administrative, technical, or material support: JAC, KE, RJ. Study supervision: JAC.