Research Article

*Correspondence to David T. Felson, Boston University School of Medicine, 715 Albany Street, Suite A203, Boston, MA 02118

Funded by:
  National Heart, Lung, and Blood Institute's Framingham Heart Study; Grant Number: N01-HC-25195
  Osteoarthritis Biomarkers grant from the Arthritis Foundation
  NIH; Grant Number: AR-47785, AG-18393, AG-14759
  USDA; Grant Number: 58-1950-4-401

Received: 30 June 2006; Accepted: 19 September 2006

10.1002/art.22292  About DOI

Osteoarthritis (OA) is characterized by loss of cartilage and concurrent changes in subchondral bone, and there is evidence that subchondral bone has a major influence on the development of OA and its worsening ([1]). The OA process includes rapid remodeling of subchondral bone ([2][3]), and the capacity of bone to respond to various stresses including loading may alter the trajectory of OA. Further, high or low levels of bone density may influence the expression of disease and affect the likelihood of incident or progressive disease ([4-6]). Healthy remodeling of bone depends on adequate availability of vitamin D ([7]).

Vitamin D sufficiency may also influence cartilage metabolism. Hypertrophic chondrocytes in OA cartilage redevelop vitamin D receptors like cells in growth plate cartilage ([8]). Vitamin D may influence these chondrocytes and somehow alter their metabolic processes and products. Thus, vitamin D may have effects on bone, and possibly also cartilage, with implications regarding disease.

Two longitudinal epidemiologic studies have shown that low vitamin D levels worsen the course of OA. In one, using data from the original Framingham Study cohort, McAlindon and colleagues ([9]) showed that subjects who had vitamin D levels in the lowest and middle tertiles had a 3-fold increased risk of radiographic worsening of preexisting knee OA. In a subsequent article, Lane et al ([10]) reported that levels of vitamin D in the lowest and middle tertiles were associated with incident hip OA, defined as development of joint space loss. Risk was similar to that described in the Framingham Study subjects. As a result of these reports, investigators have begun clinical trials to test the therapeutic efficacy of vitamin D in OA. While the results of these studies seem to replicate each other, in each of them, effects of vitamin D on disease were inconsistent. In one ([9]), vitamin D insufficiency increased the risk of disease incidence but did not influence progression in those with preexisting OA, and in the other ([10]), only effects on incidence were reported.

Vitamin D deficiency is extremely common, especially in northern latitudes ([7]). Any effect of vitamin D deficiency on OA or its worsening would have public health implications and would warrant testing of vitamin D levels in all patients with OA or at risk of developing OA. Thus, a careful characterization of the association of vitamin D deficiency with the course of OA is needed. The inconsistent findings within prior studies suggest the need for further evidence of the association of low vitamin D levels with disease. Using 2 different longitudinal studies of knee OA, we reevaluated 25-hydroxyvitamin D (25[OH]D) levels and their association with knee OA worsening (consisting of disease incidence and progression). In one study, of a community-based cohort selected without regard to the presence of knee OA, we evaluated the association of 25(OH)D levels with joint space loss over time. In the second cohort of patients who had symptomatic knee OA, we evaluated the association of 25(OH)D levels with loss of joint space seen on radiographs and with cartilage loss seen on magnetic resonance imaging (MRI).

SUBJECTS AND METHODS

Study populations.

Framingham Offspring cohort.

The Framingham Offspring cohort assembled in the early 1970s consists of the sons and daughters of the original Framingham Study cohort and the spouses of these offspring ([11]). (The study by McAlindon et al [[9]] focused on the original cohort, not the offspring cohort.) Offspring subjects were recruited as part of a study of the inheritance of OA. During a callback visit after examination 5 in the Offspring study (1993-1994), we obtained a weight-bearing radiograph of both knees at full extension, using a standardized protocol that included outlines of the feet to keep constant the rotation of the knee at followup radiography. Radiographs were obtained at 0° or at 6° caudad, and the best view of these 2 was selected for the followup examination. Weight-bearing semiflexion lateral radiographs were also obtained, using a protocol recently described ([12]).

In 2002-2005, the same subjects were called back for a followup examination, which included exactly the same assessments. The same knee radiographs were obtained on all subjects. At both baseline and followup examinations, subjects were weighed without heavy clothing or objects in their pockets, using a balance-beam scale.

Boston Osteoarthritis of the Knee Study (BOKS).

Patients were recruited to participate in a natural history study of symptomatic knee OA, the Boston Osteoarthritis of the Knee Study. The recruitment for this study, based at a Department of Veterans Affairs Medical Center, has been described in detail elsewhere ([10]). To be eligible, subjects had to have frequent knee pain and radiographic OA in that knee. A series of posteroanterior (PA) knee radiographs (lateral and skyline) was obtained on each subject to determine whether radiographic OA was present. If subjects had a definite osteophyte on any view in the symptomatic knee, they were eligible for the study. Based on frequent knee symptoms and radiographic OA, all subjects met American College of Rheumatology criteria for symptomatic knee OA ([13]).

The study included a baseline examination and followup examinations at 15 months and 30 months. We obtained a fluoroscopically positioned PA radiograph according to the protocol of Buckland-Wright ([14]) and the same lateral semiflexed radiograph as used in the Framingham Study ([12]) on all subjects who attended the clinic examination. At baseline, subjects who did not have contraindications had an MRI of the more symptomatic knee. MRIs of the same knee were also performed at followup visits. At each visit, pain over the past week was assessed using a visual analog scale (0-100), and subjects were weighed, without shoes, on a balance-beam scale.

The Institutional Review Board (IRB) at Boston University Medical Center approved the Framingham Osteoarthritis Study. Both the IRB at Boston University Medical Center and the IRB of the Veterans Administration Boston Health Care System approved the BOKS.

Radiographic assessments.

Framingham Study.

After subjects had completed the followup, both baseline and followup radiographs were read, paired and without blinding for sequence, by 2 study readers, one a bone and joint radiologist (PA), and the other a rheumatologist (BS). After assigning a Kellgren/Lawrence (K/L) grade ([15]), we focused on joint space narrowing and graded each medial and lateral tibiofemoral joint space on a semiquantitative scale from 0 (normal) to 3 (bone on bone) at each time point. We defined worsening of joint space narrowing as worsening by  1 grade in the tibiofemoral compartment on either the PA or lateral weight-bearing radiograph. Features were scored using the Osteoarthritis Research Society International (OARSI) atlas ([16]) for the anteroposterior (AP) view. For the lateral view, we scored features using an atlas created for the Framingham Study.

After each reader had evaluated the radiographs, we assessed the readings and determined whether there were important disagreements in scoring of joint space loss from baseline to followup in the tibiofemoral joint on either the AP or the lateral view. For those knees on which there was important disagreement, we held adjudication sessions to arrive at a final reading. Adjudication sessions were attended by the 2 readers and a third experienced reader (DTF), and each knee was scored by consensus. All evaluations of radiographic worsening in the tibiofemoral joint were based either on unanimity of opinion or on adjudication of readings.

For K/L grades, we used readings by the bone and joint radiologist, but if adjudication of joint space change altered the appropriateness of the K/L score, it was also adjudicated. For the bone and joint radiologist, the kappa statistic for intrareader reliability in determining K/L grade was 0.82 (P < 0.001), and the kappa statistic for interreader reliability was 0.74 (P < 0.001).

We focused on tibiofemoral joint space loss. Joint space loss was defined as a 1-grade change (e.g., from 0 to 1 or from 2 to 3) from baseline to followup on either the lateral view or the AP view. A recent study has validated use of the lateral view to identify joint space worsening ([12]). In addition, because the study by McAlindon et al ([9]) suggests that osteophyte growth is more likely in persons with low vitamin D levels, we evaluated growth in osteophytes using the same approach as was used in the previous study ([9]). We scored osteophytes on a 0-3 scale based on size using the OARSI atlas and characterized the score as an increase if the largest osteophyte in the knee (on AP or lateral view) had an increase in score. Knees with grade 3 osteophytes at baseline were excluded, since their scores could not increase.

BOKS.

As in the Framingham Study readings, we focused on the joint space to evaluate worsening and used the same radiographs and same rules to define worsening. One reader (DTF) read all radiographs. While all radiographs were read without blinding with regard to sequence, a subsample of radiographs was read with blinding for sequence, in order to test the reproducibility of the measurement of worsening and to evaluate possible bias in characterizing worsening based on the known sequence. The kappa statistic for intraobserver agreement in reading worsening in radiographs with blinding for sequence versus radiographs without blinding for sequence was 0.81 (P < 0.001), and discrepancies between blinded and unblinded readings were in no particular direction (i.e., there was no greater tendency for unblinded readings to be read as showing worsening). We also included an examination of osteophyte growth using the same scoring approach and definition of increase as in the Framingham Study, as described above.

MRI assessment of cartilage loss.

In the BOKS, all MRIs were performed with the same magnet, a Signa 1.5T system (General Electric, Milwaukee, WI), using a phased-array knee coil. A positioning device was used to ensure uniformity among patients. Coronal, sagittal, and axial images were obtained. Fat-suppressed fast spin-echo (FSE) proton-density and T2-weighted images were obtained with the following acquisition parameters: repetition time 2,200 msec, echo time 20-80 msec, slice thickness 3 mm, interslice gap 1 mm, 1 excitation, field of view 11-12 cm, and matrix 256 × 128 pixels.

Cartilage morphology was assessed by musculoskeletal radiologists (led by AG) using a semiquantitative, multifeature scoring method, the Whole-Organ Magnetic Resonance Imaging Score (WORMS), for whole-organ evaluation of the knee that is applicable to MRI readings obtained with conventional techniques ([17]). Intraclass correlation coefficients of agreement in scoring cartilage readings ranged from 0.72 to 0.97.

Tibiofemoral cartilage on MRI was scored on all 5 plates (central and posterior femur and anterior, central, and posterior tibia) in both the medial and lateral tibiofemoral joints. The anterior femur was not included in this analysis, since it is part of the patellofemoral joint. Patellofemoral cartilage was scored on 4 plates (medial and lateral patella, and medial and lateral anterior femur). These were read using the fat-suppressed, T2-weighted FSE images and graded on a 7-point scale, as follows: 0 = normal thickness and signal, 1 = normal thickness but increased signal on T2-weighted images, 2 = partial-thickness focal defect <1 cm in greatest width, 3 = multiple areas of partial-thickness (grade 2) defects intermixed with areas of normal thickness or a grade 2 defect wider than 1 cm but <75% of the region, 4 = diffuse ( 75% of the region) partial-thickness loss, 5 = multiple areas of full-thickness loss wider than 1 cm but <75% of the region, and 6 = diffuse ( 75% of the region) full-thickness loss.

In the WORMS, grade 1 does not represent a morphologic abnormality but rather a change in signal in cartilage of otherwise normal morphology. Grades 2 and 3 represent similar types of cartilage abnormality, focal defects without overall thinning. Scores of 1 and 2 were exceedingly unusual. Therefore, to create a consistent and logical scale for evaluation of cartilage morphologic change, we collapsed the WORMS cartilage score to a 0-4 scale, where the original WORMS scores of 0 and 1 were collapsed to 0, the original scores of 2 and 3 were collapsed to 1, and the original scores of 4, 5, and 6 were considered 2, 3, and 4, respectively, in the new scale. We defined a lesion as occurring in either the medial or lateral tibiofemoral compartment if it was present in the femur or tibia of that compartment.

Measurement of vitamin D levels.

In the Framingham Offspring Study, blood was drawn from subjects participating in Offspring examinations from 1996-2000, which were held between the 2 OA examinations. In the BOKS, blood was drawn at the baseline clinic visit. For some subjects, blood was not drawn until the second visit. We used blood obtained at the first visit when available.

Levels of 25(OH)D were determined by radioimmunoassay (Diasorin, Stillwater, MN), and samples from both studies were measured at the Jean Mayer United States Department of Agriculture Human Nutrition Research Center at Tufts University. In the Framingham Offspring Study, plasma was used and in the BOKS, serum. The limit of detection using this assay is 1.5 ng/ml (3.8 nmoles/liter); however, no samples had concentrations at this limit. Coefficients of variation in control samples were between 8.5% and 13.2%.

Other measurements.

In both studies, body mass index (BMI) was assessed at baseline as weight (measured on a balance-beam scale with heavy clothes and shoes off) divided by height squared.

Statistical analysis.

To examine 25(OH)D levels and their association with OA worsening and with osteophyte growth, we tested different categorizations of vitamin D levels. First, we used cut points for tertiles of vitamin D from the Framingham Study, since it is a community-based study that should provide a  normative  range of levels. We could not use the exact cut points used in the earlier study by McAlindon et al ([9]) because the present study used a different assay ([18]). Second, we used a threshold recently derived to represent vitamin D deficiency (20 ng/ml) ([7][19]) to determine whether subjects had sufficient vitamin D.

For joint space loss, we used a dichotomous knee-specific outcome and assessed the association of 25(OH)D levels with joint space loss using logistic regression, with generalized estimating equations to adjust for the correlation between knees. These analyses were adjusted for age, baseline BMI, change in weight from baseline to followup, sex, and baseline K/L score, and were modeled the same in the Framingham Study and the BOKS.

For cartilage loss, we used an ordinal logistic regression with knee-specific cartilage loss as the dependent variable (0 plates, 1 plate, and >1 plate of cartilage loss defined using the WORMS score). In cartilage analyses, we adjusted for age, BMI, sex, and baseline cartilage score.

We performed several sensitivity analyses, the results of which were the same as those reported. These included analyses in which we created a level of seasonally adjusted vitamin D by regressing vitamin D on season and using the residual vitamin D level as a predictor of joint space loss. In an attempt to replicate the findings of Lane et al ([10]), we used exactly the same cut points of vitamin D tertiles as were used in that study. In the BOKS we assessed 25(OH)D levels at all examinations and calculated average 25(OH)D levels over all visits attended.

RESULTS

In the Framingham Study, 715 subjects underwent both vitamin D measurement and a longitudinal radiographic followup. These subjects had a mean age of 53.1 years at baseline, and more than half were women (Table 1). Most knees at baseline were graded as not showing evidence of OA (87% had a K/L score <2). Over the 9.5 average years of followup, 20.3% of these knees showed tibiofemoral joint space loss on radiography.

In the BOKS, subjects were older (mean age 66.2 years) and heavier (mean BMI 31.2) (Table 1) and, based on enrollment criteria, were much more likely to have radiographic OA at baseline than subjects in the Framingham Study. Of those with radiographic followup, only 21% had K/L grades <2 at baseline, and most of these had grade 1 disease. (Those with a K/L grade of 0 had isolated patellofemoral OA, which would make a subject eligible for the BOKS.) Over the 30-month followup period in the BOKS, 23.6% of the knees with radiographic followup and 27.4% of the knees with MRI followup showed radiographic joint space loss in the tibiofemoral joint.

We found that in Framingham Study subjects, high levels of vitamin D (high tertile) were associated with slightly greater rates of worsening than were low levels (Table 2). After adjustment for other risk factors, there was no significant association of vitamin D level with disease worsening. Similarly in the BOKS (Table 2), those with high levels of vitamin D had a higher risk of worsening than those with lower levels, and the adjusted risk of worsening showed a modest but not significant protective effect of low and medium levels of vitamin D on worsening, the opposite direction of findings of prior studies.

We then examined whether levels of vitamin D thought to represent deficiency increased the risk of worsening of radiographic OA (Table 3). In the Framingham Study, knees of subjects with vitamin D deficiency (<20 ng/ml) showed a slightly higher risk of joint space loss than those without deficiency, but the adjusted risk was not elevated. In the BOKS, the risk of worsening was actually slightly higher in persons without vitamin D deficiency, and the adjusted risk showed a modest but not significant protective effect of low levels of vitamin D, once again the opposite effect of the one anticipated.

With regard to osteophyte growth, in the Framingham Study, those with low vitamin D levels had a modestly decreased risk (odds ratio [OR] 0.71, 95% confidence interval [95% CI] 0.50-1.00) compared with those with sufficient levels. In the BOKS, we found no association of low vitamin D levels with osteophyte growth (among subjects with low levels, the OR for osteophyte growth was 0.99 [95% CI 0.37-2.66]).

When we examined cartilage loss on MRI in the BOKS, our results were similar (Table 4). Those with vitamin D deficiency actually experienced a slightly lower risk of cartilage loss than those with sufficient levels of vitamin D (9.9% of those with 25[OH]D levels <20 ng/ml showed worsening of cartilage score versus 13.1% of those with 25[OH]D levels  20 ng/ml), and this translated into a modest protective effect of vitamin D deficiency on cartilage loss. We also explored whether vitamin D levels evaluated as a continuous measure affected the risk of loss, and we found no significant association (Table 4).

DISCUSSION

Contrary to expectations, in 2 longitudinal studies examining knee OA we found no association of vitamin D levels with structural disease worsening, defined as joint space loss on radiography or as worsening cartilage score on MRI. Findings are occasionally null when real associations are missed as a consequence of an excessively small sample size (Type II error). To determine whether our null results were a consequence of inadequate power, we looked at vitamin D deficiency using an accepted threshold. We found that the confidence limits were narrow, with the upper limit being 1.27 in the Framingham Study and 1.14 in the BOKS. This suggests that vitamin D deficiency could not increase the odds of joint space loss by more than 27% and 14%, respectively. Such an increased risk, even though small, could have public health implications given the safety and low cost of vitamin D. Our data suggest that vitamin D has no effect on the structural worsening of OA, at least with regard to cartilage loss. Furthermore, our point estimates of most likely risk tended to suggest, if anything, a protective effect of low vitamin D levels. Although this finding did not reach statistical significance, it suggests that, in contrast to previously reported findings ([9][10]), vitamin D is not likely to increase risk.

Within each of the earlier studies ([9][10]), there were actually inconsistencies in findings. In the study by McAlindon et al ([9]), which used another study group at Framingham, the strongest effect on OA risk was vitamin D intake, and intake had a weak association with blood levels of 25(OH)D (r = 0.24). The radiographic data used consisted of full-extension AP radiographs only. This would probably now be unacceptable as a method for defining progression because such radiographs without fluoroscopic positioning cannot be used to accurately assess joint space loss over time ([20]), leading to concerns over whether OA progression in fact occurred in knees labeled as showing progression. It is nonetheless possible that the association of vitamin D with OA in the original cohort differed from that in the offspring cohort because of the younger age (mean age 53.1 years versus 70.3 years) and commensurate healthier status of the subjects in the Offspring cohort. In the former study, low levels of vitamin D were not related at all to OA incidence (OR 0.92, 95% CI 0.45-1.87).

Lane and colleagues ([10]) did not examine the effects on disease progression (when disease is present at baseline), and, contrary to the results of the Framingham Study, they found that low vitamin D levels increased the risk of disease incidence. In both studies, therefore, vitamin D levels were not consistently related to change, and many analyses testing their association with disease change yielded null results. We suggest that these null findings were likely correct.

Our longitudinal cohorts included persons with incident disease (most subjects in the Framingham Offspring cohort started with no disease) and those with extant disease at baseline (all subjects in the BOKS), thereby allowing us to test effects of vitamin D at different disease stages. Vitamin D levels are known to vary by season ([19]). In additional analyses, we created, for each subject, a level of vitamin D adjusted for season (a residual of vitamin D level regressed on season). Testing this residual value produced the same findings as the analyses presented here, showing no increase in risk of OA worsening by seasonally adjusted vitamin D level. We also carried out analyses of physical activity and found it to be unrelated to disease incidence or progression in Framingham Study subjects ([21]). Thus, physical activity did not confound the associations described here. We also investigated whether one measure of vitamin D captures a person's long-term vitamin D status, by testing average vitamin D levels over all visits in the BOKS. Results did not differ in this analysis. It should be noted that the measure of 25(OH)D used in the present study was the same as that used in prior studies that showed an association with disease.

We read radiographs paired and without blinding for sequence (although in a subsample in the BOKS we read radiographs with blinding for sequence and obtained the same results). In a longitudinal radiographic study of vertebral fractures, Ross et al ([22]) found that, compared with reading with blinding for sequence, reading without blinding for sequence led to better detection of known risk factors for fracture and fewer errors in characterizing fractures. We contend that reading all radiographs without blinding for sequence is a more accurate way of assessing progression and is valid, especially if the reader does not know the risk factor status of the subject.

There were a few limitations to our study. First, the number of subjects was not huge, and a small increased risk of vitamin D deficiency could have been missed. However, the effects we found were in the opposite direction of those previously reported. Also, MRIs were not acquired in a way that permits evaluation of change in cartilage volume, so we might have missed such an effect. In addition, there was loss to followup and selective participation in the vitamin D substudy in the Framingham Study. The total number of subjects followed up was  1,200, but vitamin D levels were measured in only a few more than 700 subjects. There was little loss to followup (13%) in the BOKS.

While vitamin D may have little effect on cartilage loss, as identified on MRI or radiograph, it is quite possible that vitamin D deficiency could importantly affect other elements of disease, including pain and weakness, which are critical to a patient's experience with OA. Low vitamin D levels have been shown to be associated with muscle weakness ([23]) and with an increased risk of generalized pain in persons with painful disorders. It is also possible that effects of vitamin D on bone ultimately influence the structural progression of disease, although there was an interval of 9 years between baseline and followup in our studies, suggesting that any effect of vitamin D on structure, at least as visible radiographically, would likely have been detected.

In conclusion, findings from 2 longitudinal studies have not confirmed that low vitamin D levels predispose to worsening of knee OA. However, other effects of vitamin D that are not visible as cartilage loss or as radiographic progression may be important to our understanding of disease pathogenesis and may be associated with a possible therapeutic or preventive role of vitamin D in OA.

AUTHOR CONTRIBUTIONS

Dr. Felson had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Dr. Felson.

Acquisition of data. Dr. Felson, Ms Clancy, and Drs. Sack, Guermazi, Rogers, and Booth.

Analysis and interpretation of data. Drs. Felson, Niu, Guermazi, Amin, and Booth.

Manuscript preparation. Drs. Felson, Hunter, Amin, and Booth.

Statistical analysis. Dr. Niu.

Image interpretation. Drs. Aliabadi, Guermazi, and Sack.

Manuscript review. Dr. Guermazi.