To describe the association between serum 25-hydroxyvitamin D (25[OH]D) level and bone mineral density (BMD) in persons with primary knee osteoarthritis (OA).
To describe the association between serum 25-hydroxyvitamin D (25[OH]D) level and bone mineral density (BMD) in persons with primary knee osteoarthritis (OA).
We conducted a population-based survey of the Framingham Study. A total of 228 subjects with primary radiographic knee OA were identified. For vitamin D status, 25(OH)D levels ≤15 ng/ml were classified as vitamin D deficient, 25(OH)D levels 16–32 ng/ml were classified as hypovitaminosis D, and 25(OH)D levels >32 ng/ml were classified as vitamin D replete. We compared average BMD between categories of 25(OH)D levels in subjects with OA using a linear regression model while adjusting for sex, age, body mass index (BMI), knee pain, physical activity, cohort, and disease severity.
Mean age was 74.4 years and 36% were men. Of 228 individuals, 15% were vitamin D deficient, 51% had hypovitaminosis D, and 34% were vitamin D replete. Compared with subjects with vitamin D deficiency, those with hypovitaminosis D had a 7.3% higher BMD (adjusted percent difference; P = 0.02) and vitamin D replete subjects had an 8.5% higher BMD (adjusted percent difference; P = 0.02; test for trend across categories: P = 0.04).
We observed a significant positive association between serum 25(OH)D and BMD in individuals with primary knee OA, independent of sex, age, BMI, knee pain, physical activity, and disease severity. Given the high prevalence of low 25(OH)D status in persons with knee OA and the positive association between 25(OH)D and BMD, vitamin D supplementation may enhance BMD in individuals with OA.
Several investigators have suggested that osteoarthritis (OA) and osteoporosis are mutually exclusive. This hypothesis was originally based on a retrospective radiographic study by Foss and Byers in 1972 documenting the absence of osteoarthritic changes in subjects with hip fracture (1) and several reports on increased bone mineral density (BMD) in subjects with OA (2–4). However, recent studies question these earlier results. One study of monozygotic and dizygotic twins found that increased BMD is present only in the OA-affected joint and not at other sites (5). Similarly, a recent study among postmenopausal women undergoing total hip replacement for advanced OA documented that 25% of these women had occult osteoporosis (6). And most relevant, in a large prospective cohort study, osteoporotic fracture rates were not reduced in persons with OA. Thus, patients with OA with low BMD at sites other than the affected joint may be at increased risk for fractures.
One risk factor for fractures in older persons with OA may be hypovitaminosis D, as established for the general population (7–10). In addition, vitamin D deficiency may indirectly increase risk of OA progression by contributing to low bone density. Data from the Framingham Study showed that low bone density was associated with an increased risk of OA progression among persons with knee OA (11). These findings were recently corroborated by Hart and colleagues (12). Therefore, low BMD in persons with OA raises 2 concerns, increased disease progression and increased fracture risk (13–15).
We hypothesize that serum 25-hydroxyvitamin D (25[OH]D) levels are positively associated with BMD among persons with OA. If so, vitamin D supplementation may favorably affect persons with OA through 2 mechanisms. First, vitamin D may increase BMD and thereby reduce fracture risk. A recent meta-analysis found that vitamin D supplementation of at least 700–800 IU/day reduces both hip and any nonvertebral fracture by approximately one-fourth in persons ≥65 years of age (16). Second, vitamin D may decrease disease progression in OA, as previously found in the Framingham cohort study (17). The aim of this study was to evaluate the prevalence of suboptimal vitamin D status in persons with primary knee OA and whether there is a positive association between 25(OH)D levels and BMD in these individuals.
The Framingham OA study consists of 2 cohorts: the Framingham original cohort and the Framingham offspring cohort. The original cohort consisted of 5,209 subjects ages 26–62 years at the first examination, conducted between 1948 and 1951 in Framingham, Massachusetts. Subjects have been examined biannually since then. In 1971, the Framingham offspring cohort was established, which included 5,214 subjects ages 12–60 years who were the children of the original cohort and their spouses. A substudy, the Framingham Osteoarthritis Study, was conducted during the original cohort examination from 1992 to 1993 (examination 22) and the offspring study from 1993 to 1995 (18). As part of the OA substudy, standing anteroposterior fully extended weight-bearing radiographs were obtained at these examinations (19), as were serum 25(OH)D levels.
Knee OA was assessed radiographically, and a person was considered to have OA when at least 1 knee had a Kellgren/Lawrence grade 2 or more. All radiographs were read by a bone and joint radiologist and by one rheumatologist. Both readers were blinded to the exposures of interest. If there was disagreement, a second bone and joint radiologist, also blinded, was consulted, and in case the second bone and joint radiologist disagreed with the first, an adjudication panel reevaluated all individual features and the grade for overall radiographic knee OA severity. Interobserver and intraobserver reliablity expressed as intraclass correlation coefficients were assessed in a sample of the offspring and original cohort participants. The intraobserver reliablity was 0.88, and the interobserver reliability was 0.92 (18).
Bone density at the femoral neck was measured in the offspring cohort between 1996 and 2001 using a Lunar DPX-L (Lunar, Madison, WI). At the twentieth biannual Framingham examination (1988–1989) of the original cohort, we used a Lunar DP3 dual-photon absorptiometer (Lunar) to measure BMD in the proximal femur for all participants. The coefficient of variation was 2.65% for the femoral neck. In our multivariate analysis (see below), we controlled for cohort to adjust for potential differences in bone density technology.
Blood samples for serum 25(OH)D levels were collected from study participants during examinations 19 or 20 (1987–1989) in the original cohort and between 1996 and 2001 in the offspring cohort. Fasting blood samples were collected as part of the corresponding visit for BMD testing. Plasma and serum samples were stored at –70°C for no more than 3 years and were analyzed upon the first thaw. Plasma 25(OH)D was determined by radioimmunoassay (Diasorin, Stillwater, MN) for the offspring cohort. The limit of detection for 25(OH)D using this assay was 3.8 nmoles/liter; however, no 25(OH)D samples had concentrations below the limit of detection. Total coefficients of variation for control values of 36 nmoles/liter and 137 nmoles/liter were 8.5% and 13.2%, respectively. In the original cohort, serum 25(OH)D concentrations were determined by a competitive protein-binding assay (20). Interassay and intraassay coefficients of variation were 10% and 7%, respectively. In our multivariate analysis (see below), we controlled for cohort to adjust for potential differences between 25(OH)D assays.
Commonly used cutoff levels were used to classify vitamin D status in subjects with knee OA (21–23). Subjects with 25(OH)D levels ≤15 ng/ml (≤40 nmoles/liter) were classified as vitamin D deficient (23), subjects with 25(OH)D levels between 16 and 32 ng/ml (40–80 nmoles/liter) were classified as having hypovitaminosis D (23), and subjects with 25(OH)D levels >32 ng/ml (>80 nmoles/liter) were classified as vitamin D replete (21, 22).
We extracted information on potential confounders from the Framingham Study database. This included information on age and sex. Body mass index (BMI) was calculated from height and weight assessed in examination 22 of the original cohort and examination 5 of the offspring cohort. The formula for BMI is weight in kilograms divided by height in meters squared.
Knee pain was assessed when knee radiographs were obtained in both cohorts (examination 22 of the original cohort and examination 5 of the offspring cohort). One question was asked: “On most days, do you have pain, aching, or stiffness in either of your knees?” Response options were “no,” “one knee,” or “both knees.” Levels of habitual physical activity were estimated from a validated questionnaire, described by Kannel and Sorlie (24), assessing hours spent at different levels of activity (25). The number of hours reported at each level was weighted and summed, generating a theoretical range from 24 to 120 (24).
Disease severity was defined by joint space narrowing (JSN) grade using the Framingham atlas of standard radiographic features. We used JSN as our measure of disease severity because other features, including Kellgren/Lawrence grade, include measures of osteophytes or sclerosis that might themselves be related to BMD. JSN was scored 0–3 in 2 compartments of each knee (medial and lateral tibiofemoral compartments), and the maximal score was used as the narrowing score for that knee (26).
We compared the mean BMD among the 3 previously mentioned groups based on their vitamin D status using a multiple linear regression model while adjusting for age, BMI, sex, physical activity, study cohort (Framingham original cohort study versus Framingham offspring cohort study), disease severity (highest JSN grade of both knees), and knee pain (3-level indicator variable).
To assess whether the association between 25(OH)D levels and BMD was modified by sex, age, BMI, physical activity, and disease severity, we added an interaction term in the regression model, i.e., age* vitamin D group, and tested whether the interaction terms were statistically significant (P < 0.1). The results indicated that none of the interaction terms reached our prespecified P value (0.1); therefore, the main results are presented for the total population. All analyses were performed using SAS version 8.2 (SAS Institute, Cary, NC).
Of a total of 2,105 subjects from both cohorts, 327 subjects (15.5%) with knee OA were identified. Because vitamin D levels were not obtained from all subjects, the final study sample comprised 228 subjects with complete data on serum 25(OH)D levels and femoral neck BMD. The 228 subjects with complete data did not differ significantly in sex distribution, BMI, physical activity, and disease severity from the 99 persons with incomplete data. Only age differed significantly, with a mean age of 74.4 years in the enrolled 228 persons and 66.7 years in the 99 excluded persons.
Of the 228 individuals with radiographic knee OA, 83 were men and 145 were women (Table 1). Sex differences were present for several variables: women were 6 years older, had lower femoral neck BMD, were less physically active, and had lower serum 25(OH)D levels.
|Sex, no. (%)|
|Total (range)||74.4 ± 11.1 (31–94)|
|Men||70.6 ± 12.1|
|Women||76.6 ± 9.9|
|Total||28.2 ± 5.3|
|Men||28.0 ± 4.5|
|Women||28.3 ± 5.7|
|Cohort, no. (%)|
|Framingham original cohort study||181 (79)|
|Framingham offspring study||47 (21)|
|Femoral neck BMD, gm/cm2|
|Total||0.805 ± 0.168|
|Men||0.899 ± 0.171|
|Women||0.752 ± 0.141|
|Knee pain, no. (%)†|
|Yes, in 1 knee||38 (17)|
|Yes, in both knees||51 (23)|
|Physical activity score‡|
|Total||34.8 ± 6.2|
|Men||36.1 ± 7.0|
|Women||34.1 ± 5.5|
|JSN grade of most affected knee, no. (%)|
|Grade 0||81 (36)|
|Grade 1||53 (23)|
|Grade 2||94 (41)|
|25-hydroxyvitamin D levels, ng/ml|
|Total||27.8 ± 12.2|
|Men||29.9 ± 13.3|
|Women||26.5 ± 11.4|
Prevalence rates of hypovitaminosis D and vitamin D deficiency are displayed in Table 2. Including both men and women, 51% of subjects had hypovitaminosis D, defined as serum 25(OH)D levels between 16 and 32 ng/ml. Vitamin D deficiency, which was defined as serum 25(OH)D levels ≤15 ng/ml, was present in 15% of all subjects. There were approximately twice as many women compared with men with both vitamin D deficiency (34% men versus 66% women) and hypovitaminosis D (31% men versus 69% women). Only 34% of subjects were vitamin D replete, with 25(OH)D serum levels >32 ng/ml (80 nmoles/liter). Sex distribution was similar in the group that was vitamin D replete (45% men versus 55% women).
|OA phenotype||Vitamin D deficient† (n = 35)||Hypovitaminosis D‡ (n = 115)||Vitamin D replete§ (n = 78)|
|25(OH)D levels, mean ng/ml¶||11.8||23.5||41.2|
|Percent of persons in each category||15||51||34|
|Age, mean years||66.8||75.4||76.6|
|BMI, mean kg/m2||29.4||28.7||26.9|
|Physical activity, mean||35.2||34.9||34.6|
|JSN grade 0||24||43||33|
|JSN grade 1||15||58||36|
|JSN grade 2||9||58||34|
Table 2 also illustrates the distribution of important confounders between vitamin D groups included in the multivariate analysis. Subjects in the vitamin D replete group were older and had lower BMIs, on average, than individuals in the vitamin D deficient group. There appeared to be a lower prevalence of knee pain in one knee in the vitamin D replete group, when compared with both the vitamin D deficient group and the hypovitaminosis D group. The vitamin D replete group had a lower prevalence of knee pain in both knees when compared with the hypovitaminosis D group.
In all adjusted analyses, subjects in the vitamin D replete and hypovitaminosis states had higher BMDs than persons who were vitamin D deficient (Table 3). Compared with individuals in the vitamin D deficient group, those in the hypovitaminosis D group had a 7.3% higher BMD (P = 0.02), and those in the vitamin D replete group had an 8.5% higher BMD (P = 0.02). There was a significant trend between being in a higher vitamin D group and having higher BMD (test for trend: P = 0.04).
|Serum 25(OH)D (ng/ml)||Adjusted mean BMD (gm/cm2)||Adjusted mean % difference in BMD|
|Vitamin D deficient (≤15 ng/ml)†|
|Hypovitaminosis D (16–32 ng/ml)|
|Vitamin D replete (>32 ng/ml)|
This pattern was similar for men and women, as shown in a subgroup analysis in Table 3, but appeared to be more pronounced in men. Compared with men in the vitamin D deficient group, those in the hypovitaminosis D group had a 6.4% higher BMD, and those in the vitamin D replete group had a 9.8% higher BMD. Compared with women in the vitamin D deficient group, those in the hypovitaminosis D group had a 5.7% higher BMD, and those in the vitamin D replete group had a 3.7% higher BMD.
In this population-based analysis, we found a positive association between vitamin D status and BMD at the femoral neck in persons with primary radiographic knee OA independent of other important determinants of BMD such as sex, age, BMI, physical activity, knee pain, and disease severity. Compared with vitamin D deficient individuals, those with 25(OH)D serum levels between 16 and 32 ng/ml had a 7.3% higher femoral neck BMD, and those with serum levels >32 ng/ml had an 8.5% higher BMD. A subgroup analysis by sex revealed a similar pattern for men and women with a slightly stronger association in men. This may be due to higher mean 25(OH)D levels observed in men (29.9 ng/ml) compared with women (26.5 ng/ml) and a higher percentage of men in the vitamin D replete group (42% of all men and 30% of all women).
Overall, only 34% of all individuals with primary knee OA had serum 25(OH)D levels >32 ng/ml (80 nmoles/liter) (21, 22). This high prevalence of suboptimal vitamin D levels in persons with knee OA was not unexpected because the mean age was 74 years. We have previously documented in a US national survey that only 33% of ambulatory white persons ages ≥60 years reach 25(OH)D levels ≥32 ng/ml (27, 28). This underlines the general need for vitamin D supplementation, especially in older persons, and including those with OA.
Because there is an association of hip and knee OA, it is conceivable that a small percentage of our subjects with OA had coexistent hip OA. Because the BMD scan focused on the hip, in theory, the data from such a scan could have been affected by hip OA. However, as Nevitt and colleagues have demonstrated, the artifact in the dual x-ray absorptiometry scan result is minimal, because most subjects in population studies have mild hip OA and the area of measurement, the femoral neck, is far below the joint (29). Even so, we were not focusing on whether OA was associated with BMD in this study; rather, we examined whether vitamin D status was associated with BMD among persons with OA.
Given the positive association between vitamin D status and BMD and the high prevalence of suboptimal 25(OH)D levels in this study, there may be 2 possible incentives to correct suboptimal vitamin D levels in persons with knee OA. A first incentive may be prevention of disease progression. Both higher BMD (11, 12) and higher 25(OH)D levels (17, 30) have been found to be associated with a decrease in disease progression in persons with knee OA. Zhang and colleagues, in a prospective analysis, found that high BMD as well as BMD gain decreased the risk of progression of radiographic knee OA in the Framingham study (11). In the same cohort, McAlindon and colleagues reported that the risk of progression of knee OA was increased 3 fold for persons in the middle and lower tertiles of both vitamin D intake and serum levels (17). Lane and colleagues corroborated these findings for hip OA (30).
A second incentive for the correction of suboptimal 25(OH)D levels in persons with knee OA may be the prevention of falls and fractures. As mentioned in the introduction, osteoporotic fractures are not reduced in persons with OA (14), and persons with OA of the weight-bearing joints are at increased risk of falling (13–15). Based on evidence from randomized controlled trials (RCTs) in the general older population, vitamin D supplementation of at least 700–800 IU/day is successful in reducing both nonvertebral fractures (16) and falls among older persons (31, 32).
Findings from our study, together with previous reports and evidence from RCTs performed in the general older population, support the notion of correcting suboptimal vitamin D levels in persons with knee OA, which may reduce disease progression, falls, and fractures. This suggestion, despite the lack of evidence from RCTs performed in persons with OA, is especially attractive because correcting 25(OH)D levels by vitamin D supplementation (cholecalciferol) is simple, safe, well tolerated, and inexpensive (33).
The strength of our study is its population-based design. The BMD endpoint is important in regard to disease progression in OA (11, 12) and fracture risk in older persons (9). In addition, the identified significant trend across groups of vitamin D status with a similar pattern in men and women lends credibility to the main findings. We were able to adjust the analyses for a number of important determinants of BMD: sex, age, BMI, physical activity, knee pain, and disease severity.
There are also limitations to our study. One limitation is the cross-sectional design of the analyses, which cannot be used to establish a causal relationship between vitamin D levels and BMD in persons with knee OA. However, a causal relationship between these factors has been demonstrated in several RCTs among the general older population (7, 9). In addition, 99 subjects were dropped from the analyses because they did not provide data on BMD or 25(OH)D levels. However, apart from a significant age difference (persons who were dropped had a mean age of 66.7 years versus 74.4 years in 288 included persons), the 288 analyzed subjects were representative of the total cohort of persons with primary radiographic knee OA for all other covariates. Furthermore, even with this reduced number of subjects, we were able to detect associations of interest between vitamin D status and bone density. Finally, vitamin D assays and bone density technologies differed between the 2 cohorts; however, we adjusted for cohort in our analyses to account for these potential confounders.
We conclude that femoral neck BMD in individuals with primary radiographic knee OA is positively associated with vitamin D status. Specifically, vitamin D deficiency in these individuals (25[(OH]D levels ≤15 ng/ml) is associated with a 7.3–8.5% lower BMD if compared with individuals with higher vitamin D levels. Future RCTs are needed to address the effect of vitamin D on BMD, disease progression, and fractures in persons with OA. Until then, given the high prevalence of suboptimal 25(OH)D levels in persons with knee OA and the documented positive association between 25(OH)D and BMD in these individuals, vitamin D supplementation may be warranted in persons with OA.
Supported by NIH grant AR47785 and the Harvard Hartford Foundation.