Plasma B Vitamins, Homocysteine, and Their Relation with Bone Loss and Hip Fracture in Elderly Men and Women

Abstract

Context: Elevated homocysteine is a strong risk factor for osteoporotic fractures among elders, yet it may be a marker for low B-vitamin status.

Objective: Our objective was to examine the associations of plasma concentrations of folate, vitamin B12, vitamin B6, and homocysteine with bone loss and hip fracture risk in elderly men and women.

Design: This was a longitudinal follow-up study of the Framingham Osteoporosis Study.

Setting: Community dwelling residents of Framingham, MA, were included in the study.

Participants: A total of 1002 men and women (mean age 75 yr) was included in the study.

Main Outcome Measures: Baseline (1987–1989) blood samples were used to categorize participants into plasma B-vitamin (normal, low, deficient) and homocysteine (normal, high) groups. Femoral neck bone mineral density (BMD) measured at baseline and 4-yr follow-up was used to calculate annual percent BMD change. Incident hip fracture was assessed from baseline through 2003.

Results: Multivariable-adjusted mean bone loss was inversely associated with vitamin B6 (P for trend 0.01). Vitamins B12 and B6 were inversely associated with hip fracture risk (all P for trend < 0.05), yet associations were somewhat attenuated and not significant after controlling for baseline BMD, serum vitamin D, and homocysteine. Participants with high homocysteine (>14 μmol/liter) had approximately 70% higher hip fracture risk after adjusting for folate and vitamin B6, but this association was attenuated after controlling for vitamin B12 (hazard ratio = 1.49; 95% confidence interval 0.91, 2.46).

Conclusions: Low B-vitamin concentration may be a risk factor for decreased bone health, yet does not fully explain the relation between elevated homocysteine and hip fracture. Thus, homocysteine is not merely a marker for low B-vitamin status.

Among elderly men and women, lower plasma concentrations of B vitamins are associated with greater bone loss and increased hip fracture risk. Elevated plasma homocysteine also predicted hip fracture, yet this relation was not fully explained by low B vitamin status, suggesting independent affects of both B vitamins and homocysteine on the risk of hip fracture.

Mildly elevated plasma homocysteine concentration has been a strong risk factor for hip fracture in population-based cohorts of elders (1,2,3), yet the cross-sectional associations of homocysteine with bone mineral density (BMD) and markers of bone turnover have been inconsistent (4). It is generally hypothesized that homocysteine may interfere with bone collagen cross-links, thereby increasing bone fragility and susceptibility to fracture. Although two in vitro studies suggest that collagen cross-linking is impaired in homocystinuria patients (5,6), there is no evidence linking this mechanism to decreased bone strength in nonpatients. Thus, the causal link between elevated homocysteine and hip fracture risk remains unclear.

Homocysteine may instead be a marker for an associated risk factor for decreased bone strength. The B vitamins folate, B12, and B6 are important cofactors in homocysteine metabolism, and low status of these nutrients is the primary determinant of elevated plasma homocysteine concentrations in elders (7,8). Several observational studies suggest that poor dietary intakes and low blood concentrations of B vitamins may be associated with decreased BMD, greater bone loss, and higher risk of osteoporotic fracture (4,9). Furthermore, in vitro studies indicate that low B-vitamin concentrations promote osteoclast activity (10), whereas elevated concentrations may stimulate bone formation (11). Thus, if these nutrients do have direct effects on bone health, the observed association between elevated homocysteine and hip fracture risk may be due to inadequate B-vitamin status.

The objective of this study was to elucidate the relations of B vitamins and homocysteine with bone loss and risk of hip fracture among men and women enrolled in the Framingham Study. We hypothesized that low plasma B-vitamin and high homocysteine concentrations would be associated with increased bone loss, and that low plasma B-vitamin concentration is an independent risk factor for increased hip fracture risk. We further hypothesized that the association between elevated plasma homocysteine and increased hip fracture risk would be explained, at least in part, by low B-vitamin status.

Subjects and Methods

Participants

The Framingham Study began in 1948 with the primary goal of evaluating risk factors for heart disease. Two thirds of Framingham, MA, households were sampled, and 5209 men and women, ages 28–62 yr, were recruited and have been evaluated biennially for nearly 60 yr (12). Blood samples were obtained from 1401 participants who were alive in 1987–1989 and attended the 20th biennial examination. Of these cohort members, 1256 had valid plasma measures of at least one B vitamin (folate, B12, B6) or homocysteine. Participants included in the current study are the 1002 individuals who had femoral neck BMD measured and were free of hip fracture at the blood draw. Participants included in bone loss analyses included 714 (71%) cohort members who obtained follow-up BMD measures 4 yr later (1991–1993). Details on the follow-up osteoporosis examination have been reported elsewhere (13). The appropriate institutional review boards at Hebrew SeniorLife, Boston University, and Tufts-New England Medical Center approved this study, and written informed consent was obtained for all study subjects.

Plasma measures

Nonfasting blood samples were collected in EDTA-containing tubes and promptly centrifuged at 4 C for 15 min at 2000 × g. Samples were collected over a period of 2.5 yr and stored at −70 C for up to 3 yr. Plasma folate concentration (nmol/liter) was determined by a microbial assay using a 96-well plate and manganese supplementation. Plasma vitamin B12 concentration (pmol/liter) was determined using a radioassay kit from Ciba-Corning (Medfield, MA). Plasma vitamin B6 (pyridoxal-5′-phosphate) concentration (nmol/liter) was determined by the tyrosine decarboxylase method. Interassay coefficients of variation were 13, 7, and 16% for folate, vitamin B12, and vitamin B6, respectively. Plasma homocysteine concentration (μmol/liter) was measured using HPLC with fluorometric detection (coefficient of variation, 9%) (7).

Plasma concentrations of folate, vitamin B12, and vitamin B6 were each categorized using previously published clinical cutpoints defining normal, low, or deficient vitamin status. The cutoffs for folate were more than or equal to 11 nmol/liter (normal), 7 to less than 11 nmol/liter (low), and less than 7 nmol/liter (deficient) (14). Cutoffs for B12 were more than or equal to 258 pmol/liter (normal), 148 to less than 258 pmol/liter (low), and less than 148 pmol/liter (deficient) (15). Cutoffs for B6 were more than or equal to 30 nmol/liter (normal), 20 to less than 30 nmol/liter (low), and less than 20 nmol/liter (deficient) (16). Plasma homocysteine concentration was categorized as normal (≤14 μmol/liter) or high (>14 μmol/liter) (7).

BMD

BMD of the right femoral neck was measured in grams per square centimeter (g/cm²), using a Lunar dual photon absorptiometer (DP3) at baseline and a Lunar dual x-ray absorptiometry (DPX-L) densitometer (Lunar Radiation Corp., Madison, WI) at the follow-up examination. Using standard positioning recommended by the manufacturer, the right femur was scanned at each examination unless there was a history of fracture or hip joint replacement, in which case the left side was scanned. The coefficients of variation for the DP3 and DPX-L were 2.6 and 1.7%, respectively. Because of a small but consistent shift in BMD values between the two technologies, baseline BMD was adjusted for the change in equipment from DP3 to DPX-L technology, using published corrections (17).

Hip fracture

Hip fractures were assessed by interview at each biennial examination or by telephone interview for participants unable to attend an examination (e.g. due to illness or residing out of state). In addition, medical records of hospitalizations and deaths were systematically reviewed for the occurrence of hip fractures. Reported hip fractures were confirmed by a review of medical records, including radiographic and operative reports. Hip fracture was defined as a first-time fracture of the proximal femur.

Covariates

Covariates were obtained from data at the 20th biennial examination, and included gender, age, height, weight, weight change over the 4-yr follow-up, smoking status, consumption of caffeine and alcohol, physical activity index, fall history, and in women, current use or nonuse of estrogen. Of the 1002 participants, 808 had information on calcium intake, and 892 had valid measures of serum 25-hydroxyvitamin D [25(OH)D]. Height without shoes was measured to the nearest quarter inch. Weight (in light clothing without shoes) at baseline and follow-up was measured to the nearest pound using the same standard balance beam scale. Smoking status was assessed as whether a participant regularly smoked cigarettes over the 2 yr before baseline. Caffeine consumption in the form of tea and coffee was quantified as previously described (18). Alcohol consumption (beer, wine, spirits) was calculated as the number of ounces consumed per week, as previously reported (19). Physical activity index is a weighted sum of typical daily activities of hours spent on strenuous, moderate, and light activity, as well as at rest (20,21). Falls in the previous year (self-reported, unintended contact with the ground, regardless of whether the fall resulted in a fracture) were assessed by questionnaire (22). Calcium intake (mg/d) for the previous 12 months was assessed using a 126-item semiquantitative food frequency questionnaire that has been validated for numerous nutrients, including calcium (23). Serum 25(OH)D (ng/ml) was measured using a competitive binding protein assay (interassay coefficient of variation, 10%) (24).

Statistical analysis

The distributions of plasma folate, vitamin B12, vitamin B6, and homocysteine concentrations were skewed. Thus, analyses of bone loss and hip fracture were conducted with plasma measures modeled as both log-transformed continuous variables and as categorical variables based on the clinical cutpoints described previously. Results of both analyses revealed similar patterns, therefore, only the categorical analysis results are presented because they are more clinically relevant and easily interpreted.

To determine whether BMD should be considered as a mediator of the relations of B vitamins with hip fracture, we used multivariable linear regression to assess the association of each log-transformed continuous plasma B vitamin with baseline femoral neck BMD, adjusting for sex, age, height, weight, and estrogen use in women. We further determined the association between plasma homocysteine and baseline femoral neck BMD to confirm previous findings of no independent relation (4,25).

Femoral neck bone loss was calculated as the 4-yr annual percent change in BMD: the percent difference between baseline and follow-up BMD, divided by 4 yr. We used analysis of covariance to compare the least squares-adjusted mean bone loss in the low and deficient B-vitamin groups, and high homocysteine group, to their respective normal groups. Where there was an apparent dose-response relation across B-vitamin groups, the vitamin was modeled as a single ordinal variable in a linear regression model to test for a linear trend.

Person-years at risk for hip fracture was calculated as the time of blood sample collection to the first occurrence of hip fracture, death, last contact with the participant, or the end of follow-up (December 31, 2003). For each B vitamin, we used Cox proportional hazards regression to calculate hazard ratios (HRs) and 95% confidence intervals (CIs) for the low and deficient groups compared with their respective normal groups (referent). To determine whether BMD or homocysteine concentration may be mediating any observed associations between a B vitamin and hip fracture risk, analyses were subsequently adjusted for baseline femoral neck BMD and for homocysteine status (normal vs. high). Finally, we calculated the HR for hip fracture for the high homocysteine group, relative to the normal group, with and without adjustment for each B vitamin (modeled as a single ordinal variable) to determine whether B-vitamin status may be confounding the association between hip fracture and plasma homocysteine concentration.

Not all study participants had complete information on all plasma measures. Thus, to ensure that changes in effect estimates after adjustment for a plasma measure were not affected by missing data, analyses of baseline BMD and hip fracture were conducted within subgroups of participants with information on plasma concentrations of folate (n = 960), vitamin B12 (n = 823), and vitamin B6 (n = 909) who also had information on plasma homocysteine concentration.

All regression models were initially adjusted for a minimal set of covariates, including sex, age, height, weight, and estrogen use in women. Minimally adjusted bone loss analyses also included weight change. Models were then further adjusted for an extended set of covariates, including caffeine, alcohol, smoking, and physical activity index. Extended models for hip fracture analyses also included fall history. Finally, all models were adjusted for calcium intake and serum 25(OH)D among participants with information on these covariates. Results of models adjusted for the extended set of covariates, and for calcium intake, were similar to those of the minimally adjusted models and are, thus, not presented. All analyses were conducted separately for women and men, but because results did not differ, we present findings for the total study population only. The proportional hazards assumption was satisfied for all Cox regression analyses. We used SAS/STAT software version 9.1 (SAS Institute Inc., Cary, NC) for all analyses.

Results

Baseline characteristics of the study sample are described in Table 1. Among the 1002 participants, 60% were female, and mean age (±sd) was 75 yr (5). The proportions (percentage) of participants with below normal plasma concentrations of folate (<11 nmol/liter), vitamin B12 (<258 pmol/liter), and vitamin B6 (<30 nmol/liter) were 57, 39, and 23, respectively, whereas 26% had high homocysteine concentration (>14 μmol/liter).

Table 1.

Descriptive characteristics of men and women in the Framingham Study at baseline who had measures of at least one plasma measure (folate, vitamin B12, vitamin B6, or homocysteine) and BMD (1987–1989), according to whether they had a follow-up BMD measurement (1991–1993)

Characteristica	All participants	Baseline and follow-up	Baseline only
No.	1002	714	288
% Women	60.2	63.8	53.8
Age (yr)	75.3 ± 4.9	74.5 ± 4.5	77.3 ± 5.6
Height (in.)	63.7 ± 3.8	63.7 ± 3.7	63.8 ± 75.3
Weight (lb)	154.8 ± 31.4	155.8 ± 31.9	152.4 ± 29.8
% Current cigarette smoker	10.3	9.9	11.5
Physical activity index	33.5 ± 59.5	33.9 ± 5.4	32.5 ± 5.2
Alcohol (oz/wk)	2.3 ± 3.7	2.3 ± 3.7	2.2 ± 3.7
Caffeine (% more than two cups/d)	24.3	24.1	24.7
% Current estrogen use in women	4.6	5.6	1.9
Calcium intake (mg/d)	800.9 ± 428.5	798.6 ± 423.7	807.2 ± 442.3
25(OH)D (ng/ml)	29.6 ± 12.3	30.0 ± 12.0	28.7 ± 12.8
Baseline femoral neck BMD (g/cm2)	0.780 ± 0.142	0.787 ± 0.139	0.765 ± 0.148
Folate (nmol/liter)
% Normal (≥11)	43.3	43.0	43.9
% Low (7–10.9)	21.5	22.7	18.7
% Deficient (<7)	35.2	34.3	37.4
Vitamin B12 (pmol/liter)
% Normal (≥258)	61.0	60.2	62.9
% Low (148–257.9)	27.8	29.1	24.6
% Deficient (<148)	11.2	10.7	12.5
Vitamin B6 (nmol/liter)
% Normal (≥30)	76.6	78.4	72.2
% Low (20–29.9)	11.8	11.2	13.1
% Deficient (<20)	11.6	10.4	14.7
Homocysteine (μmol/liter)
% Normal (≤14)	73.7	76.7	66.6
% High (>14)	26.3	23.3	33.4

^aMean ± sd unless otherwise noted. 

Femoral neck BMD was directly associated with log-transformed plasma folate (P = 0.02), vitamin B12 (P < 0.01), and vitamin B6 (P < 0.01), and was thus considered as a potential mediator of the relations of these nutrients with hip fracture risk. Log-transformed plasma homocysteine concentration was inversely related to baseline femoral neck BMD (P = 0.03), yet the association was no longer statistically significant (P ranged from 0.10–0.35) after adjustment for each individual plasma B vitamin.

Bone loss

Table 1 lists the characteristics of the 714 cohort members with longitudinal BMD, and the 288 members who attended the baseline examination and did not obtain follow-up. Participants without follow-up tended to be male, older, and have lower baseline femoral neck BMD. The distribution of plasma B-vitamin concentrations did not differ between participants with follow-up data and those without, although a higher proportion of participants without follow-up had high plasma homocysteine (33 vs. 23%).

There was a significant trend (P = 0.01) for greater mean femoral neck bone loss with decreasing plasma vitamin B6 concentration after adjustment for covariates (Fig. 1). Furthermore, participants with deficient plasma vitamin B6 concentration had significantly greater mean annual bone loss compared with those with normal B6 concentrations (−1.16 vs. −0.60%; P = 0.02). Mean bone loss did not differ among categories of plasma folate or vitamin B12. Among the 684 participants with plasma homocysteine information and follow-up BMD, mean percent annual bone loss in the high homocysteine group (−0.93, se 0.15) was not statistically different from that in the normal group (−0.63, se 0.08; P = 0.07). Results were similar after adjustment for serum 25(OH)D.

Figure 1.

Least squares-adjusted mean annual percent change in femoral neck BMD according to B-vitamin groups for men and women in the Framingham Study. The number of participants included in folate, vitamin B12, and vitamin B6 analyses was 672, 581, and 633, respectively. Means adjusted for sex, age, height, weight, weight change, and estrogen use in women. Bars denote ses. *P = 0.02 vs. normal.

Hip fracture

Characteristics of participants according to whether they experienced a hip fracture are listed in Table 2. Participants who fractured tended to be female, of lighter weight, had lower baseline femoral neck BMD, and had greater proportions of individuals with deficient plasma B-vitamin concentrations and high plasma homocysteine concentrations.

Table 2.

Descriptive characteristics of men and women in the Framingham Study at baseline who had measures of at least one plasma measure (folate, vitamin B12, vitamin B6, or homocysteine) and BMD (1987–1989), according to whether they had a hip fracture during follow-up (through 2003)

Characteristica	Fracture	No fracture
No.	111	891
% Women	81.1	57.6
Age (yr)	76.9 ± 5.9	75.1 ± 4.8
Height (in.)	62.6 ± 3.8	63.9 ± 3.7
Weight (lb)	140.2 ± 31.8	156.6 ± 30.8
% Current cigarette smoker	11.7	10.1
Physical activity index	33.7 ± 5.2	33.5 ± 5.4
Alcohol (oz/wk)	1.5 ± 2.7	2.4 ± 3.8
Caffeine (% more than two cups/d)	26.1	24.0
% Current estrogen use in women	2.2	5.1
Calcium intake (mg/d)	816.3 ± 481.0	799.2 ± 422.4
25(OH)D (ng/ml)	26.7 ± 12.7	30.0 ± 12.2
Baseline femoral neck BMD (g/cm2)	0.691 ± 0.117	0.791 ± 0.141
Folate (nmol/liter)
% Normal (≥11)	43.0	43.3
% Low (7–10.9)	15.9	22.3
% Deficient (<7)	41.1	34.4
Vitamin B12 (pmol/liter)
% Normal (≥258)	55.7	61.7
% Low (148–257.9)	29.6	27.5
% Deficient (<148)	14.8	10.8
Vitamin B6 (nmol/liter)
% Normal (≥30)	69.6	77.5
% Low (20–29.9)	13.7	11.5
% Deficient (<20)	16.7	11.0
Homocysteine (μmol/liter)
% Normal (≤14)	67.6	74.5
% High (>14)	32.4	25.5

^aMean ± sd unless otherwise noted. 

Plasma folate was not significantly associated with hip fracture risk in adjusted analyses (Table 3). Lower plasma concentrations of both vitamins B12 (P for trend = 0.02) and B6 (P for trend = 0.04) were associated with increased hip fracture risk, with participants in the deficient vitamin B12 and vitamin B6 groups at 89 and 73% increased risk, respectively, compared with their respective normal groups. When adjusted for BMD and homocysteine status, the increasing trends were no longer statistically significant, and the HRs for the deficient vitamin B12 and vitamin B6 groups were somewhat attenuated, ranging from 1.54–1.62. Controlling for serum 25(OH)D did not substantially change the association between vitamin B12 and hip fracture (data not shown), although the HR for the deficient vitamin B6 group was reduced to 1.47 (95% CI 0.92–2.28).

Table 3.

HRs (95% CI) for hip fracture according to the plasma B-vitamin group among men and women in the Framingham Study (1987–2003)

Vitamin	No.	Status group	Multivariable adjusteda	Multivariable and baseline BMD adjusted	Multivariable and homocysteine adjusted
Folate	960	Normal	1.00	1.00	1.00
		Low	0.76 (0.43, 1.32)	0.73 (0.42, 1.28)	0.72 (0.41, 1.26)
		Deficient	1.38 (0.91, 2.09)	1.31 (0.86, 1.98)	1.20 (0.78, 1.86)
		P for trend	n/a	n/a	n/a
Vitamin B12	823	Normal	1.00	1.00	1.00
		Low	1.48 (0.90, 2.42)	1.53 (0.93, 2.50)	1.39 (0.84, 2.29)
		Deficient	1.89 (1.01, 3.52)	1.62 (0.86, 3.04)	1.60 (0.83, 3.10)
		P for trend	0.02	0.06	0.10
Vitamin B6	909	Normal	1.00	1.00	1.00
		Low	1.31 (0.74, 2.34)	1.19 (0.66, 2.12)	1.16 (0.64, 2.09)
		Deficient	1.73 (1.02, 2.95)	1.54 (0.90, 2.63)	1.56 (0.90, 2.68)
		P for trend	0.04	0.11	0.12

n/a, Not applicable. 

^aAdjusted for sex, age, height, weight, and estrogen use in women. 

Participants with high homocysteine concentrations had 69% greater hip fracture risk (95% CI 1.12–2.55) compared with those with normal concentrations (Table 4). After adjustment for plasma folate and vitamin B6, HRs were similar. However, adjustment for plasma vitamin B12 attenuated the HR to 1.49 (95% CI 0.91–2.46). Further adjustment for serum 25(OH)D did not change results.

Table 4.

Hip fracture HR (95% CI) for high plasma homocysteine concentration (>14 μmol/liter) vs. normal (≥14 μmol/liter), with and without adjustment for plasma B-vitamin concentration among men and women in the Framingham Study (1987–2003)

Vitamin adjusted for	No.	Homocysteine group	Multivariable adjusteda	Multivariable and B vitamin adjusted
None	979	Normal	1.00
		High	1.69 (1.12, 2.55)
Folate	960	Normal	1.00	1.00
		High	1.74 (1.15, 2.64)	1.66 (1.08, 2.56)
Vitamin B12	823	Normal	1.00	1.00
		High	1.70 (1.06, 2.73)	1.49 (0.91, 2.46)
Vitamin B6	909	Normal	1.00	1.00
		High	1.75 (1.15, 2.66)	1.63 (1.06, 2.51)

^aAdjusted for sex, age, height, weight, and estrogen use in women. 

Discussion

In our study of community dwelling older men and women, lower plasma vitamin B6 concentration was associated with greater bone loss, whereas elevated plasma homocysteine concentration was not. Lower concentrations of vitamins B12 and B6 were associated with increased risk of hip fracture. Our findings also suggest that vitamins B12 and B6 do have modest effects independent of BMD, serum vitamin D, and homocysteine because adjustment for these factors did not substantially attenuate these associations. Lower plasma vitamin B12 concentration may, to some degree, explain the association between elevated plasma homocysteine and increased hip fracture risk, although there are still homocysteine effects that are not explained by vitamin B12 status.

The observational nature of our study precludes conclusions of causality, yet our results provide some insight into the role of BMD in the associations of B vitamins and homocysteine with hip fracture risk. Our findings are not consistent with a previous study that reported increased bone loss at the total femur among elderly women in the lowest quintile of serum B12 concentration (26). However, in our hip fracture analysis, the association between vitamin B12 and hip fracture was attenuated after adjusting for baseline BMD, suggesting that vitamin B12 may have effects on fracture risk that are mediated through BMD. This mechanism is supported by several cross-sectional studies (4), including previous work by our group in the younger Framingham offspring cohort (27), that have found a direct association between vitamin B12 status and BMD. In our current study, participants in the deficient vitamin B12 group tended to lose more bone than those with higher concentrations, and we may have underestimated the true magnitude of the association in our cohort if participants who did not return for follow-up BMD assessment had the greatest bone loss (28).

We found that lower vitamin B6 was associated with greater bone loss, and adjustment for BMD attenuated the association with hip fracture, indicating that insufficient vitamin B6 may contribute to greater bone fragility through reductions in bone mass. Although few studies have examined vitamin B6 in relation to bone health (4), our results are congruent with a recent investigation of men and women in the Rotterdam Study in which dietary intake of vitamin B6 was directly related to femoral neck BMD and inversely related to risk of osteoporotic fracture risk (9).

However, homocysteine does not seem to affect fracture risk through an influence on bone mass. In line with our results, several cross-sectional studies have been unable to link elevated homocysteine with decreased BMD (4,25), and van Meurs et al. (2) found that the relation between homocysteine and fracture was unchanged after accounting for baseline BMD. Furthermore, clinical trial data indicate that although lowering homocysteine with B-vitamin supplementation may reduce fracture risk in elderly stroke patients (29), this intervention had no effect on BMD in these patients or on markers of bone turnover in healthy older persons (30).

Alternately, elevated homocysteine may increase the risk of fracture by influencing the propensity for falls. High homocysteine is a risk factor for cardiovascular disease and cognitive dysfunction (31), and may be associated with greater disability and decline in physical function (32,33), all of which can increase the risk for falls. However, this mechanism is not supported by the clinical trial among stroke patients, which found similar fall rates in the treatment and placebo groups (29), or by our current study and that of van Meurs et al. (2) in which adjustment for recent falls did not change the association between homocysteine and hip fracture.

Also of note was that adjustment for BMD did not completely attenuate the associations of vitamins B12 and B6 with hip fracture risk, suggesting effects through alternate mechanisms. Vitamin B12 deficiency, which is common among elders (34), can lead to neurological complications characterized by paresthesia, loss of proprioception, and reduced vibration sense in the lower extremities (15), conditions that may increase the propensity for falls. Vitamin B6 is an essential coenzyme for lysyl oxidase, a precursor to collagen cross-links (35). Thus, insufficiency of this nutrient may impair cross-link formation and subsequently contribute to bone fragility. Although an animal study (36) and a small case-control study of hip fracture patients (37) support this vitamin B6 mechanism, to our knowledge the relation between vitamin B12 deficiency and falls has not been investigated. Further work is needed to determine the true causal mechanisms linking homocysteine and B vitamins with fracture risk.

Our study has some limitations. The generalizability of our results is limited to older white men and women. Only plasma concentrations of B vitamins were available, which may not provide optimum assessment of vitamin status. Although red blood cell folate may better indicate long-term status (14), low plasma folate in the Framingham cohort was predictive of elevated homocysteine, which itself may be considered a functional indicator of poor folate status (7). Manifestations of vitamin B12 deficiency may occur at plasma concentrations more than 148 pmol/liter (38), thus the functional marker methylmalonic acid is considered a more sensitive and specific indicator of deficiency (39). In our study, participants with deficient vitamin B12 tended to have greater bone loss and higher fracture risk. Thus, misclassification of deficiency based only on plasma vitamin B12 would have biased our results toward the null. Urinary excretion of vitamin B6 metabolites is the most sensitive indicator of vitamin B6 status in controlled studies (40), yet these measures are not practical for use in large observational studies. Only a single measure of B vitamins and homocysteine concentration was available, making our study susceptible to regression dilution bias. Bias may also have been introduced by losses to follow-up. In both cases, our results would be biased toward the null, underestimating the true effects. We assessed variables as potential mediators of an association by adjusting our analyses for the mediator. When doing so, it must be considered that any effect remaining after adjustment may represent a residual association resulting from misclassification of the mediator or the exposure. Finally, it is possible that low B-vitamin status is a marker for an unbalanced diet, and the observed associations reflect a deficiency in other unmeasured nutrients that are important for bone health.

Nonetheless, this study has several unique strengths. To the best of our knowledge, this is the first study to examine the associations of bone loss and hip fracture with plasma concentrations of homocysteine and its main nutritional determinants in a large, population-based, elderly cohort that includes both men and women. Vitamin status was determined by plasma concentrations rather than assessment by questionnaire. Folate status was assessed before nationwide folic acid fortification (1998), allowing us to examine low-plasma folate concentrations that are presently uncommon in the United States. Finally, reported hip fractures were rigorously confirmed by medical records abstraction, reducing the potential for misclassification of the outcome.

This study suggests that low vitamin B6 status, but not elevated homocysteine, is an important determinant of bone loss in community dwelling elders. Low status of vitamin B12 and vitamin B6 may be independent risk factors for hip fracture, and the relation between homocysteine and hip fracture is largely independent of B-vitamin status. Supplementation or changes in diet are easy and effective methods for controlling B-vitamin and homocysteine concentrations, and may be considered as potential novel measures for reducing fracture rates. Further research is needed to determine how homocysteine and B vitamins influence fracture risk, and whether these interventions may help maintain bone mass and reduce the risk of fracture in elderly men and women.