B-vitamin status and bone mineral density and risk of lumbar osteoporosis in older females in the United States^,,

Abstract

Background: Previous data suggest that elevated serum total homocysteine (tHcy) may be a risk factor for bone fracture and osteoporosis. Nutritional causes of elevated tHcy are suboptimal B-vitamin status. To our knowledge, this is the first nationally representative report on the relation of B vitamins and bone health from a population with folic acid fortification.

Objective: The purpose of this analysis was to examine the relation between B-vitamin status biomarkers and bone mineral density (BMD), risk of osteoporosis, and biomarkers of bone turnover.

Design: We examined the relation of tHcy, methylmalonic acid (MMA), and serum/red blood cell folate and total-body and lumbar spine BMD in women aged ≥50 y participating in the NHANES 1999–2004 (n = 2806), a nationally representative cross-sectional survey. These are the only years with concurrent measurement of tHcy and whole-body dual-energy X-ray absorptiometry. We also examined B-vitamin biomarkers relative to bone turnover markers, bone alkaline phosphatase, and urinary N-terminal cross-linked telopeptide of type I collagen in a 1999–2002 subset with available data (n = 1813).

Results: In comparison with optimal concentrations, women with elevated tHcy were older with lower serum vitamin B-12, red blood cell folate, and dietary micronutrient intakes and had significantly higher mean ± SE markers of bone turnover (bone alkaline phosphatase: 15.8 ± 0.59 compared with 14.0 ± 0.25 μg/L; urinary N-terminal cross-linked telopeptide of type I collagen: 48.2 ± 2.9 compared with 38.9 ± 0.90 nmol bone collagen equivalents per mmol creatinine/L). Elevated MMA (OR: 1.88; 95% CI: 1.10, 3.18) and tHcy (OR: 2.17; 95% CI: 1.14, 4.15) were related to increased risk of lumbar osteoporosis. When examined as a continuous variable, tHcy was negatively associated, serum folates were positively associated, and MMA and vitamin B-12 were not significantly associated with lumbar and total-body BMD.

Conclusion: In this nationally representative population of older US women with high exposure to B vitamins through food fortification and dietary supplements, only elevated tHcy and MMA were independently associated with risk of lumbar spine osteoporosis.

INTRODUCTION

The risk of osteoporosis and bone fractures increases exponentially with age. Women carry a disproportionate burden of bone disorders related to aging; in fact, 1 in 3 women aged >50 y will experience a bone fracture in her lifetime, representing approximately 200 million in the United States alone (1). Hip fracture is an important cause of disability in the elderly and is associated with an increased risk of mortality. Given the demographic shift in the proportion of older adults, it is critical to identify strategies to reduce or mitigate age-related bone disorders. Sex, race-ethnicity, physical activity, alcohol, smoking, estrogen, and diet all affect the risk of fractures in the older population. Previous studies suggest that elevated serum total homocysteine (tHcy,⁹ >13 μmol/L) may be a risk factor for poor bone health (2), including low bone mineral density (BMD) (3–6), bone fracture (7–10), and osteoporosis (11–13), but the data are equivocal (14, 15). Further evidence that tHcy is related to bone health comes from the genetic disease homocystinuria in which brittle bones are seen, even in young children (16).

Elevated tHcy is caused by renal disease or suboptimal B-vitamin status, primarily in the form of inadequate folate and vitamin B-12 and, to a lesser extent, vitamin B-6 and riboflavin status. B-vitamin status may be an emerging risk factor for poor bone health in women without renal disease. No national estimates of bone health parameters in US women relative to B-vitamin status have been published since the implementation of folic acid fortification of enriched cereal grains that started in 1996 (17). Since folic acid fortification was implemented in the United States, the primary determinant of high tHcy has been vitamin B-12 (18, 19). Thus, the purpose of this analysis was to examine the association between tHcy and vitamin B-12 status and bone health of women aged ≥50 y in the era since folic acid fortification was implemented in the United States by using data from the NHANES 1999–2004. Specifically, we examined BMD, risk of osteoporosis, and biomarkers of bone turnover in the only NHANES survey years with concurrent measurement of biomarkers of B-vitamin status [i.e., serum and red blood cell (RBC) folate, tHcy, and methylmalonic acid (MMA)] and ascertainment of all bone health parameters. The study focused on women in this age range because they are a high-risk group for osteoporosis and bone fracture.

METHODS

NHANES is a cross-sectional, nationally representative survey of the noninstitutionalized US population. The National Center for Health Statistics (NCHS) of the CDC collects the NHANES data by using a complex, stratified, multistage probability cluster sampling design. NHANES participants are first interviewed in their homes; during this interview, demographic information, data on dietary supplement and prescription medication use, and some health-related data are collected. About 1–2 wk later, participants undergo a standardized physical examination, 24-h dietary recall, and a blood draw in a mobile examination center. Written informed consent is obtained from all participants; the survey protocol was approved by the NCHS Research Ethics Review Board. For this analysis, we combined data from NHANES 1999–2000, 2001–2002, and 2003–2004 (n = 31,126) because these survey years have concurrent measurements of dual-energy X-ray absorptiometry (DXA) and vitamin B-12 biomarkers. We excluded males (n = 15,184), women <50 y (n = 11,339), women missing information on tHcy (n = 581) or serum creatinine (n = 84), and those who had impaired renal function (n = 53); the analytic sample comprised 3127 women. From the analytic sample, data on bone parameters were available for 2806 women. This represents 59% of the selected sample, 81% of the interviewed sample, and 91% of the examined sample of women aged ≥50 y (20).

Plasma tHcy concentrations were measured by using a fluorescence polarization immunoassay reagent kit (Abbott Laboratories) (21). Serum vitamin B-12 concentrations were assayed by using the Quantaphase II radioassay (Bio-Rad), and plasma MMA concentrations were measured by gas chromatography–mass spectrometry after cyclohexanol derivation of the extracted MMA. Serum creatinine was measured in NHANES 1999–2000 by the Jaffe method by using the Hitachi 917 multichannel analyzer and in NHANES 2001–2004 by the Jaffe rate method (kinetic alkaline picrate) by using the Beckman Synchron LX20 modular chemistry analyzer. We adjusted serum creatinine data from NHANES 1999–2000 by using the equation, 0.147 + 1.013 × uncorrected serum creatinine (mg/dL), because the analytic note recommends harmonization with the 2001–2004 measurements (22). An estimated glomerular filtration rate was calculated for each participant based on serum creatinine by using the Modification of Diet in Renal Disease Study equation. These estimates were used to categorize and exclude those with renal disease as determined by a low estimated glomerular filtration rate (<30 mL/min · 1.73 m²) because impaired kidney function increases MMA and tHcy. Serum cotinine, a marker of tobacco exposure, was assayed via isotope dilution with liquid chromatography and tandem mass spectrometry.

Abnormal biomarkers were defined as follows: low serum vitamin B-12 as <148 pmol/L (23–26), high MMA as >271 nmol/L (25, 27–32), and elevated tHcy as concentrations >13 μmol/L (21, 33). Details and documentation for each of these methods are publicly available on the NHANES website (34–36). Frank folate deficiency was rare in the United States during this time period (37); thus, we used the following cutoffs to describe low folate status: <10 nmol/L for serum and <340 nmol/L for RBC folate based on previous NHANES studies, suggesting that folate concentrations below these points are associated with increased tHcy (38). For the estimation of the mean serum vitamin B-12 only, we also excluded data from women with serum vitamin B-12 concentrations >99th percentile (>1265 pmol/L; range: 25–148,935 pmol/L) because these extraordinarily high concentrations almost always reflect conditions that are unrelated to vitamin B-12 metabolism or status and can exert substantial influence on the mean (39).

Dietary supplement information was collected via the NHANES Dietary Supplement Questionnaire, which determines each participant’s intakes of vitamins, minerals, herbs, and other dietary supplements during the previous 30 d. The mean daily intake of micronutrients from all supplements was calculated (40) and added to micronutrient intake estimates from foods and beverages from the first 24-h dietary recall. We report intake data as means or medians rather than prevalence of low values because of the inability to correct for within-person variability necessary to characterize the tails of the intake distribution. The mean dietary intakes of calcium and vitamin D from the dietary estimates were used in our models; mean intakes do not change when within-person variability adjustment procedures are employed (41–44). Self-reported alcohol intake was collected via a questionnaire, and the mean number of drinks per day was estimated. Self-reported physical activity and the mean number of hours spent sitting in front of a computer or television screen were assessed by questionnaire. We used self-reported data from the NHANES prescription drug and reproductive health questionnaires to evaluate the use of prescription medications that may alter bone health, such as hormone replacement therapy, corticosteroids, and glucocorticoids (45).

Total-body bone density was measured by DXA by using Hologic QDR 4500A fan-beam densitometers. Details on the DXA examination protocol have been published elsewhere (46). Osteoporosis status was classified based on BMD from the lumbar spine subregion of the total-body DXA scan only because previous research indicates that BMD from the lumbar spine subregion of the total-body scan is highly correlated with BMD from a dedicated anterior-posterior lumbar spine scan and results in similar estimates of osteoporosis prevalence and similar ORs for predicting fracture (47). Dedicated scans of femurs, which are the most commonly used skeletal sites for clinical evaluation, were not performed in NHANES 1999–2004. Diagnostic criteria from the WHO were used to define lumbar spine osteoporosis and low bone mass (48). Specifically, osteoporosis was defined as a lumbar spine BMD value that fell >−2.5 SDs below the mean BMD of a young reference group, whereas low bone mass was defined as a lumbar spine BMD value between –1 and −2.5 SDs below the young reference group mean. The young reference group consisted of 30-y-old white women from the Hologic lumbar spine reference database, in accordance with recent recommendations from the International Society for Clinical Densitometry (49).

Bone turnover marker subset analysis

Bone-specific alkaline phosphatase (BAP) and urinary N-terminal cross-linked telopeptide of type I collagen (uNTx) are 2 bone turnover biomarkers that have been previously identified as independent risk factors for osteoporosis and were collected only in NHANES 1999–2002 (50). BAP, a biomarker of bone formation, was measured differently across the survey years included in this analysis by using both the Tandem-MP Ostase ImmunoEnzymetric (Hybritech) and the Access Ostase (Beckman Coulter) assays. Per NCHS guidelines, a regression equation was used to harmonize measurements obtained by different methods (51). uNTx is a marker of bone resorption measured in nanomoles of bone collagen equivalents per liter. It was assessed by using the Osteomark and Vitros ECI (Ortho-Clinical Diagnostics) on-spot urine specimens (other than the first morning void) (51). From the analytic sample, data were available on BAP for 1833 women and uNTx for 1813 women.

Statistical analysis

We performed all statistical analyses by using SAS (version 9.3; SAS Institute) and SAS-callable SUDAAN (version 10.0; Research Triangle Institute) software. All estimates were calculated by using sample weights to represent the US population, account for differential nonresponse and noncoverage, and adjust for the planned oversampling of some groups. We estimated SEs for all statistics of interest by Taylor series linearization, a standard method for variance estimation in complex surveys.

The multiple imputed total-body DXA data were used in all analyses, as recommended by the NCHS; complete information on the imputation procedures can be accessed on the NCHS website (51). Imputation is used to correct biases that may occur in estimates, and it increases the precision of estimates (52). Descriptive statistics were estimated and compared by using proc descript with diff var statements. Least squares means and SEs were estimated for the fully adjusted models by using proc regress. The median dietary intakes are more descriptive because some very high values skewed the means and, thus, their interpretability; while medians are presented in the tables, statistical comparison of the log-transformed mean intakes was assessed by using proc descript, diffvar statements. Logistic regression was used to estimate the ORs and 95% CIs, and χ² (via proc crosstab) was used to examine categorical variables. Rank-normal inverse transformations were applied to the lumbar and total-body BMD scores to satisfy the normality assumption before examination of biomarkers of B-vitamin status through multiple linear regression analysis. Statistical significance was set at P < 0.05. Confounders were selected based on a priori knowledge of variables that are known to relate to BMD and biomarkers of B-vitamin status and are not on the causal pathway (53, 54).

RESULTS

Women with elevated tHcy were older and were more likely to be non-Hispanic black and to have lower estimated glomerular filtration rate, serum vitamin B-12, and RBC folate concentrations (Table 1). Dietary intakes of vitamin B-12, vitamin B-6, and calcium were significantly higher in those with normal tHcy concentrations, and the prevalence of use of calcium (mean ± SE; 68% ± 1.4% vs. 51% ± 3.6%) and vitamin D (mean ± SE; 56% ± 1.4% vs. 37% ± 3.8%) supplements was higher than in those with elevated tHcy. Women with elevated tHcy also reported less alcohol consumption, more use of hormone therapy, and less physical activity. In the 1999–2002 subgroup analysis, women with elevated tHcy had significantly higher BAP and uNTx concentrations than did women with normal tHcy (Table 1).

TABLE 1.

Descriptive data on demographic, clinical, diet, and lifestyle characteristics; medication use; and bone markers by homocysteine classification in US women aged ≥50 y, 1999–2004¹

	Normal tHcy (n = 2482)	Elevated tHcy (n = 324)
Age, y	63.5 ± 0.32	71.7 ± 1.0*
Race-ethnicity, %
Non-Hispanic white	79.3 ± 1.8	81.9 ± 2.2
Non-Hispanic black	7.7 ± 1.1	10.4 ± 1.5*
Mexican American	3.5 ± 0.7	2.8 ± 0.6
Clinical
BMI, kg/m2	28.4 ± 0.2	28.9 ± 0.5
Glomerular filtration rate, mL/min · 1.73 m2	76.7 ± 0.5	57.7 ± 1.8*
Red blood cell folate, nmol/L	804 ± 8	735 ± 32*
Serum vitamin B-12, pmol/L	439 ± 15	288 ± 10*
Bone3
Bone alkaline phosphatase, μg/L	14.0 ± 0.25	15.8 ± 0.59*
uNTx, nmol bone collagen equivalents per mmol creatinine/L	38.9 ± 0.90	48.2 ± 2.9*
Diet
Total vitamin B-12 intake, μg/d	7.6 (3.0–25.6)4	3.7 (1.6–7.6)*
Total vitamin B-6 intake, mg/d	2.5 (1.3–4.7)	1.5 (0.9–3.3)*
Total calcium intake, mg/d	954 (590–1481)	616 (415–930)*
Supplemental vitamin D,5 μg/d	9.91 ± 0.15	9.73 ± 0.48
Alcohol6
Nondrinkers	45.8 ± 2.0	56.9 ± 3.2
1 drink/d	54.1 ± 2.0	42.7 ± 3.1
>2 drinks/d	<0.5	<0.5
Physical activity
Daily hours sitting, n	2.7 ± 0.04	2.9 ± 0.10
Moderate physical activity, %	50 ± 1.4	33 ± 3.1*
Medication use, %
Hormones	6.7 ± 0.7	0.5 ± 0.4*
Glucocorticoids	0.9 ± 0.2	0

¹The analytic sample comprised 3127 women, of whom 2806 had information on bone parameters. Individuals who identified as “other” for race-ethnicity are not presented, and the percentages do not add to 100 as a result. *P ≤ 0.05 (these comparisons are not controlled for any variables). tHcy, serum total homocysteine; uNTx, urinary N-terminal cross-linked telopeptide of type I collagen.

²Mean ± SE (all such values).

³The bone turnover markers were available in only 1648 women with normal tHcy and 187 women with elevated tHcy in 1999–2002.

⁴Median; IQR in parentheses (all such values). Medians (IQRs) are presented because of skewed distribution; statistical comparison was performed on the log-transformed dietary intakes by using proc descript.

⁵Data are presented for dietary supplement users only; data on dietary vitamin D were not available in these survey years.

⁶Significant group-level difference determined by χ². For alcohol, 1 drink contains 10 g ethanol and is equivalent to 12 oz beer, 4 oz wine, or 1 oz distilled spirits.

Women with normal lumbar spine BMD had significantly lower tHcy and significantly higher RBC folate than did women with low bone mass or osteoporosis in both the unadjusted and fully adjusted models (Table 2). Results for the other variables were less consistent between the unadjusted and fully adjusted models. For example, mean MMA differed significantly by lumbar spine status in the unadjusted model but no longer differed after all the variables in the full model were adjusted. The converse was seen for serum vitamin B-12 and serum folate; means did not differ by lumbar spine BMD group in the unadjusted models but did differ in the fully adjusted models. The magnitude of the differences in means between the BMD groups was roughly 4–6% for serum vitamin B-12 and 6–8% for serum folate (Table 2).

TABLE 2.

Biomarkers of B-vitamin status by lumbar spine bone classification in US women aged ≥50 y, 1999–2004¹

	Mean ± SE
	Normal (n = 1478)	Low bone mass (n = 1020)	Osteoporosis (n = 308)
Basic models2
Total homocysteine, μmol/L	8.9 ± 0.1a	9.5 ± 0.2b	10.1 ± 0.3b
Serum vitamin B-12, pmol/L	395 ± 7	385 ± 7	408 ± 14
Methylmalonic acid, μmol/L	172 ± 5a	205 ± 17b	220 ± 14b
Red blood cell folate, nmol/L	817 ± 11a	777 ± 11b	750 ± 20b
Serum folate, nmol/L	42.1 ± 0.9	41.1 ± 0.8	43.5 ± 1.8
Fully adjusted models3
Total homocysteine, μmol/L	8.9 ± 0.1a	9.4 ± 0.2b	9.6 ± 0.3b
Serum vitamin B-12, pmol/L	403 ± 6a	380 ± 7b	389 ± 15a,b
Methylmalonic acid, μmol/L	171 ± 5	208 ± 20	202 ± 15
Red blood cell folate, nmol/L	815 ± 8a	774 ± 11b	728 ± 24c
Serum folate, nmol/L	43.2 ± 0.8a	40.7 ± 0.9b	40.0 ± 1.9a,b

¹Means ± SEs within a row not sharing a common superscript letter are significantly different at P ≤ 0.05 based on simple or multiple regression. The analytic sample comprised 3127 women, of whom 2806 had information on bone parameters.

²These comparisons are not controlled for any variables. P ≤ 0.05.

³Full model adjusted for age, race, BMI, physical activity, glomerular filtration rate, hormone use, serum cotinine, and dietary intakes of calcium and vitamin D.

Both elevated tHcy and MMA were associated with increased odds of having lumbar osteoporosis both when the outcome was defined as osteoporosis vs. normal BMD only and when it was defined as osteoporosis vs. normal and low bone mass combined (Table 3); similar results were obtained in unadjusted models (data not shown). Low serum vitamin B-12 and RBC folate concentrations were not significantly associated with odds of lumbar osteoporosis in either model; however, the number of subjects who had low vitamin B-12 and low RBC concentrations was smaller. Because total-body BMD cannot be classified into clinical bone mass categories, we examined the risk of being in the lowest total BMD quartiles vs. being in quartiles 2–4 by abnormal vitamin B biomarker status (Supplemental Table 1). Results were similar to those observed for lumbar spine for tHcy and MMA. However, results for serum vitamin B-12 were significant when total-body BMD was considered, with those having low serum vitamin B-12 being roughly twice as likely to be in the lowest BMD quartile.

TABLE 3.

Abnormal biomarkers of B-vitamin status and lumbar spine osteoporosis in US women aged ≥50 y, 1999–2004¹

	Lumbar spine osteoporosis vs. normal (n = 1652)2			Lumbar spine osteoporosis vs. normal and low bone mass (n = 2806)3
	n	OR (95% CI)	P value	n	OR (95% CI)	P value
Total homocysteine (>13 μmol/L)	324	2.17 (1.14, 4.15)*	0.009*	324	1.95 (1.24, 3.07)*	0.003*
Serum vitamin B-12 (<148 pmol/L)	75	1.07 (0.41, 2.80)	0.89	75	1.15 (0.43, 3.11)	0.77
Methylmalonic acid (>271 μmol/L)	283	1.88 (1.10, 3.18)*	0.016*	283	1.54 (0.99, 2.39)*	0.054*
Red blood cell folate (<340 nmol/L)	114	2.10 (0.78, 5.60)	0.14	114	1.91 (0.88, 4.18)	0.10

¹The analytic sample comprised 3127 women, of whom 2806 had information on bone parameters. Full model adjusted for age, race, BMI, physical activity, glomerular filtration rate, hormone use, serum cotinine, and dietary intakes of calcium and vitamin D. *P ≤ 0.05 (significant difference determined by logistic regression).

²This analysis compares those with osteoporosis of the lumbar spine with those with normal lumbar spine (i.e., low bone mass is not included in this analysis).

³This analysis compares those with osteoporosis of the lumbar spine with those with normal or low bone mass of the lumbar spine.

To assess which vitamin had the strongest influence on tHcy, we examined the proportion of variation in tHcy that was explained by MMA, serum vitamin B-12, serum folate, and RBC folate in age-, race-, and eGFR-controlled regression models. Although all biomarkers were significant predictors (P < 0.001), MMA explained much more of the variation in tHcy concentrations (r² = 0.43) than did serum vitamin B-12 (r² = 0.19) or serum (r² = 0.13) or RBC (r² = 0.14) folate concentrations (data not shown).

We also examined the relation between B-vitamin status and BMD quartiles for the lumbar spine as well as for the total body. For both total body and lumbar spine, women in the lowest quartiles had significantly higher proportions of elevated tHcy and MMA. For example, the percentage with elevated tHcy or MMA was roughly 3 times higher in the lowest vs. highest quartile of total-body BMD and 2 times higher in the lowest vs. highest quartile of lumbar spine BMD. However, no differences were observed for serum vitamin B-12 or RBC folate (Table 4).

TABLE 4.

Abnormal biomarkers of B-vitamin status by BMD quartiles in US women aged ≥50 y, 1999–2004¹

	Quartile
	1	2	3	4
Total-body BMD
Participants, n	880	680	646	600
Total homocysteine (>13 μmol/L)*	44 ± 42	24 ± 3	19 ± 3	13 ± 3
Serum vitamin B-12 (<148 pmol/L)	37 ± 6	17 ± 6	23 ± 6	23 ± 5
Methylmalonic acid (>271 μmol/L)*	47 ± 3	25 ± 2	13 ± 2	15 ± 3
RBC folate (<340 nmol/L)	22 ± 5	29 ± 4	27 ± 6	22 ± 4
Lumbar spine BMD
Participants, n	865	634	666	641
Total homocysteine (>13 μmol/L)*	37 ± 3	25 ± 3	20 ± 4	18 ± 2
Serum vitamin B-12 (<148 pmol/L)	34 ± 5	16 ± 5	32 ± 6	18 ± 5
Methylmalonic acid (>271 μmol/L)*	38 ± 3	21 ± 3	20 ± 3	21 ± 3
RBC folate (<340 nmol/L)	29 ± 5	24 ± 6	27 ± 5	20 ± 3

¹The analytic sample comprised 3127 women, of whom 2806 had information on bone parameters. Percentiles of the population distribution were used to construct the quartiles. For total-body BMD: quartile 1 (<0.95), quartile 2 (≥0.95 to <1.03), quartile 3 (≥1.03 to <1.11), and quartile 4 (≥1.11). For lumbar spine BMD: quartile 1 (<0.87), quartile 2 (≥0.87 to <0.96), quartile 3 (≥0.96 to <1.08), and quartile 4 (≥1.08). *P ≤ 0.05 (significant difference by χ² and Wald’s test). BMD, bone mineral density; RBC, red blood cell.

²Mean ± SE percentage (all such values).

Finally, we explored the lumbar and total-body BMD as transformed linear terms with regard to all B-vitamin biomarkers. tHcy, RBC, and serum folate were significant predictors of BMD as a continuous variable for both lumbar spine and total body (Supplemental Table 2). The analysis was performed on BMD transformed by the inverse normal rank transformation; thus, the β-coefficients are not biologically relevant but represent the magnitude and strength of association assessed by multiple linear regressions.

DISCUSSION

In this nationally representative sample of older US women without renal disease, having an elevated tHcy or MMA was associated with roughly twice the odds of having lumbar spine osteoporosis. Furthermore, the percentage of women with elevated tHcy was 3 times higher in the lowest vs. highest quartiles of total-body BMD and 2 times higher in the lowest vs. highest quartiles of lumbar spine BMD. Although serum vitamin B-12 was not related to bone status directly, the significant findings for tHcy and MMA provide indirect evidence of a specific role for this vitamin in bone health because they are both functional indicators of vitamin B-12 status. Specifically, elevated MMA reflects vitamin B-12 status exclusively. tHcy can be influenced by both vitamin B-12 and folate, so determining which of these 2 vitamins is responsible for elevated tHcy requires examining its relation with MMA and serum vitamin B-12 (which reflects vitamin B-12 status), serum folate (which reflects folate status only), and RBC folate (which reflects both vitamin B-12 and folate status) (55, 56). Because MMA is specific to vitamin B-12 status and is a better functional indicator of vitamin B-12 status than serum vitamin B-12, these results indirectly support a role for vitamin B-12 in skeletal health because they suggest that vitamin B-12, and not folate, is the primary driver of elevated tHcy in our sample. Other studies in the postfortification era have also indicated that elevated tHcy is more closely linked with vitamin B-12 status than folate status (18). It should be noted, however, that when BMD was examined as a continuous variable, folate concentrations were positively associated with lumbar and total-body BMD. During these survey years, folic acid fortification was part of the US food supply, and previous studies have determined that fortification coincided with the highest serum and RBC folates in US history. Thus, it is surprising that folate was found to be significantly associated with BMD, even in this population of women who would be generally considered replete.

Most clinical trials have examined vitamin B-12 and folic acid in combination, and thus, the independent effect of each vitamin on bone health is largely unknown. In unfortified populations, clinical trial and cohort data suggest that tHcy has an independent role in bone health (2, 4–7, 11, 12, 14, 15, 57–68). However, the mechanism by which tHcy could influence bone health remains incompletely resolved. tHcy could exert an influence on bone health through modulation of collagen or by altering osteoblast and/or osteoblast activity (12, 63, 69). Our findings support the association of elevated tHcy and higher concentrations of markers of bone turnover. Cross-sectional analysis of the B-PROOF (B-Vitamins for the Prevention of Osteoporotic Fractures) study and of the Rotterdam study elderly cohorts I and II (n = 6040) indicate that elevated tHcy was associated with poorer measures from quantitative ultrasound of the bone, reflecting microarchitecture of bone independently of BMD (3). Although the B-PROOF clinical trial reduced tHcy concentrations (−4.4 μmol/L), only treated subjects >80 y of age showed a lower risk of any fracture (HR: 0.26; 95% CI: 0.10, 0.71; P = 0.01) and osteoporotic fracture (HR: 0.28; 95% CI: 0.10, 0.74; P = 0.01) (70). B-PROOF did not assess BMD categorized as osteoporosis or low bone mass, but no significant difference in mean BMD was observed between treatment and placebo groups (71). Furthermore, there are 2 noteworthy differences between the 2 studies. Our population was exposed to folic acid fortification and was younger, yet we found a tHcy effect. The B-PROOF findings that lowering tHcy in the oldest age group reduced the fracture risk, along with our data, support the need for additional study, including randomized controlled trials, in nations with folic acid fortification.

In addition, some meta-analyses have addressed the role of B vitamins and bone health. One review and meta-analysis with limited clinical trial (n = 1) and multiple observational studies (n = 27) relied primarily on data obtained from countries without folic acid fortification (72). The van Wijngaarden et al. (72) meta-analysis included both men and women and included only studies of bone as the primary outcome; this work suggests a borderline significant association of higher serum vitamin B-12 with lower fracture risk and higher tHcy with higher fracture risk (RR: 1.04; 95% CI: 1.02, 1.07), but there was no significant association appearing for either biomarker and BMD in women (no data on men). Two smaller, lower quality meta-analyses are also available. The first failed to find an association with B-vitamin supplementation for fracture risk or bone turnover but did not evaluate tHcy (73); however, significant heterogeneity was observed in the trials included in this meta-analysis because bone was a secondary endpoint that the cardiovascular trials included, and these trials were primarily in males. The second meta-analysis examined 6 case-control trials in postmenopausal women with osteoporosis and indicated cases had higher serum vitamin B-12 and tHcy than did controls (74). Overall, these meta-analyses provide some useful insights, but the role of B vitamins in bone health remains unclear.

Study limitations include the cross-sectional nature of the analyses; as such, our results cannot establish causality between biomarker concentrations and bone outcomes. Another limitation is the lack of scientific consensus regarding which cutoffs to use to determine “low” status; for this reason, we used the recommended, optimal cutoffs to define vitamin B-12 status based on an extensive analysis of the NHANES 1999–2004 data as previously published (75, 76). The cutoffs selected may not be accurately identifying subclinical vitamin B-12 deficiency, which is more common than the more severe clinical deficiency (77). Also, vitamin B-6 may be related to both tHcy concentrations and bone health, and data on this nutrient were not available during these survey periods. Riboflavin was not included in the analysis, but low riboflavin is not an issue in the United States because there is mandatory fortification of the grains (78). Previous work by Yazdanpanah et al. (79) indicates that both vitamin B-6 and riboflavin are associated with risk of bone fracture and that vitamin B-6 is also associated with BMD. We did not have genetic data in this analysis; previous work suggests that riboflavin may modify fracture risk in women who are homozygous for the MTHFR 677 T allele (80) and that the Sp1 polymorphisms in the collagen type I α1 gene are associated with BMD and risk of bone fracture in the Rotterdam study (81). Finally, lumbar spine BMD assessment is typically based on a site-specific lumbar spine DXA scan rather than on the lumbar spine subregion from a total-body DXA scan because the former has less measurement error; however, the 2 scan types are highly correlated (47). Finally, vitamin D information was available only for intake from dietary supplements during all of the survey years in which bone measures and B-vitamin biomarker status measures were available.

In summary, our analysis suggests that elevated tHcy and MMA are significantly associated with increased likelihood of osteoporosis in a nationally representative sample of US women. Both vitamin B-12 status indicators and folate status (as a linear term only) were associated with poorer total-body and lumbar BMD in a population exposed to food fortified with folic acid. Given the public health burden of osteoporosis and age-related bone disorders together with the rapidly aging US population, we hope this work will serve to stimulate additional research aimed at addressing the role of vitamin B-12 and folate in bone health in older women in the United States.