Osteoarthritis (OA) is the most prevalent form of arthritis and the leading cause of disability among older adults in the US.1 The disease is characterised by pathology of cartilage, bone and periarticular structures. Spinal disc degeneration is often considered as spinal OA and is characterised by disc space narrowing and presence of anterior vertebral osteophytes.2 3 The degree of osteophyte formation appears to be related directly to the degree of narrowing at the intervertebral disc space.4 In addition, osteophytes in the spine are accompanied by osteophytes in peripheral joints and form through similar processes.5 6 Thus, spinal osteophytes may provide evidence about the tendency to develop or have progressive osteophytes elsewhere.

Bisphosphonates have been demonstrated to have chondroprotective effects, to reduce the incidence and progression of osteophytes in animal models, and to modify osteoblast function in vitro.7^–12 In an observational epidemiological study, compared to non-users, women on alendronate had less severe knee pain and less subchondral bone attrition and bone marrow lesions.13 One uncontrolled study of etidronate demonstrated an improvement in subjective pain in individuals with spondylosis and knee OA,14 in keeping with the analgesic effects of bisphosphonates seen in other conditions involving hyper-resorptive states such as Paget disease, multiple myeloma and bone metastatic disease. Although these studies have focused on pain symptoms, it has been postulated that antiresorptive agents may have salutary effects on cartilage and bone metabolism.8 15 There have been no human trials to date examining alendronate’s structural effects on OA or spinal disc disease.

The objective of this exploratory study was to examine the effects of alendronate on the progression of spinal disc degeneration in postmenopausal women, a group at risk for OA, in a secondary analysis of data from women who received alendronate vs placebo over a 3–4 year interval as part of a large randomised study, the Fracture Intervention Trial (FIT).16^–18 Specifically, we evaluated whether alendronate had effects on progression of spinal osteophytes and disc space narrowing. We also evaluated whether prevention of vertebral fractures or presence of such fractures at baseline relates to osteophyte progression.

METHODS

Study participants

The FIT study was a randomised placebo-controlled trial carried out in 6457 postmenopausal women with prevalent vertebral fractures or low bone mass that assessed the efficacy of alendronate vs placebo in preventing vertebral fractures and clinical fractures.16^–18 Women with prevalent vertebral fractures (n = 2023) were followed for 3 years, while women without prevalent vertebral fractures (n = 4434) were followed for 4 years. Alendronate was dosed at 5 mg/day in the first 2 years of the study and then 10 mg/day thereafter.

For the purposes of this study, we randomly selected 200 participants from the parent trial to re-read their radiographs for features of spinal OA as an exploratory analysis. We were limited to 200 participants due to limited funding for reading radiographs. To select subjects, we sampled randomly using random number tables within strata defined by age (<65 vs ⩾65 years old) and presence of vertebral fractures at baseline in an attempt to ensure that the prevalence of osteophytes was roughly equal between the treatment arms in this sample. We did not stratify on treatment arm as we expected to obtain roughly even numbers from each arm given that the parent trial was randomised 1:1 treatment vs placebo.

Radiographic assessments

Lateral lumbar and thoracic spinal radiographs were obtained at baseline and at follow-up in the parent study.17 18 We read all radiographs blinded to treatment, paired and in sequence using a standard atlas19 that was revised in 2004. One reader scored baseline and follow-up lateral lumbar and thoracic spinal x rays for: (1) anterior osteophytes, which were scored 0–3 on superior and inferior aspects of each vertebral body from the levels T4–T12 and L1–L5, for which the ICC was 0.95; (2) disc-space narrowing (DSN), which was scored 0–3 at each vertebral level from the level of T4/5 to T12/L1 and L1/2 to L5/S1, for which the ICC was 0.65. There was very little DSN in the thoracic spine, resulting in a low ICC due to its low prevalence at that site.20 The ICC for the lumbar spine was 0.81.

Statistical analyses

Using films from each of baseline and follow-up exams, we computed the sum of OST scores from T4 to L5 and the sum of DSN scores from T4 to S1. The difference between the sum at follow-up and the sum at baseline yielded a change in each summary score. After assuring normality, we assessed the mean difference in change of each summary score between treatment arms using multiple linear regression. Since there was very little disc-space narrowing in the thoracic spine, we additionally performed post hoc analyses for the change in disc-space narrowing summary scores in the lumbar spine. All analyses were adjusted for age, body mass index (BMI), presence of baseline vertebral fractures, presence of incident vertebral fractures (to ensure the treatment effect was not due to prevention of fractures alone), baseline kyphosis score and presence of diabetes at baseline since diabetes may be associated with osteophytosis.21 To determine whether prevention of fractures by alendronate could account for any differences noted, we also assessed interactions between treatment assignment and presence of baseline or incident vertebral fracture. All analyses were performed using SAS V 9.1 (SAS Inc, Cary, North Carolina, USA) and p values ⩽0.05 were considered statistically significant.

Role of sponsor

The sponsor was not involved in the study design, sample selection, data analysis, writing, or approval of this manuscript.

RESULTS

There were 84 subjects in the alendronate group and 116 subjects in the placebo group. The participant characteristics are shown in table 1.

The mean age of the two groups was similar (65.8 years in alendronate group, 66.1 years in placebo group), with similar BMI, prevalence of diabetes and prevalence of baseline fractures. As expected, based on the efficacy of alendronate in preventing fractures in this trial, the placebo group had a higher frequency of incident vertebral fractures, although given our small sample in this study (3% of the total trial population), the effect estimate is imprecise. The mean baseline osteophyte and disc-space narrowing scores were also similar between the two groups. Most of the disc-space narrowing scores arose from the lumbar spine. No subjects met criteria for diffuse idiopathic skeletal hyperostosis.

The adjusted mean change in osteophyte score in the alendronate group was 3.2 (95% CI 2.4 to 4.1), meaning that over the course of the trial, the osteophyte score progressed by an average of 3.2 in the alendronate group. The adjusted mean change in the osteophyte score in the placebo group was 4.7 (95% CI 3.7 to 5.7, p = 0.04). Thus, the placebo group had greater progression of osteophytes over the course of the trial (table 2).

The adjusted mean change in disc-space narrowing score at all disc levels was 0.4 in the alendronate group and 0.7 in the placebo group (p = 0.2), suggesting that the progression of disc-space narrowing in the alendronate group was almost half that of the placebo group, although the difference was not statistically significant. When analyses were limited to the lumbar levels, the adjusted mean change in disc-space narrowing score was 0.3 (95% CI 0.2 to 0.5) in the alendronate group and 0.6 (95% CI 0.4 to 0.8) in the placebo group (p = 0.04) (table 3).

We found no significant interactions between treatment assignment and baseline or incident vertebral fractures, with the mean change summary scores remaining unchanged for osteophyte and disc-space narrowing scores when the interaction terms were included. The β estimate was −1.07, p = 0.7 for the baseline fracture and treatment assignment interaction term in the change in osteophyte analyses, with mean change in the osteophyte score in alendronate group being 3.2 and in placebo group being 4.7, unchanged from analyses without the interaction term. Similarly, for the incident fracture and treatment assignment interaction term, the β estimate was −1.49, p = 0.3, with the mean change in osteophyte score in the alendronate and placebo groups remaining unchanged (3.2 and 4.7, respectively). Likewise, the interaction terms were not significant in the change in the disc-space narrowing analyses for either baseline or incident vertebral fractures, leaving the mean change in disc-space narrowing scores unchanged. This suggests that the effect of treatment on osteophytes and disc-space narrowing was no different in those with and without incidence or prevalent vertebral fractures.

DISCUSSION

In this secondary and exploratory analysis of a sample from a randomised trial, progression of osteophyte score was significantly less in the alendronate group compared to placebo. Alendronate also lessened the progression of disc-space narrowing compared with placebo in the lumbar spine. Although focusing on disc-space narrowing in the lumbar spine was a post hoc analysis, because the bulk of disc-space narrowing occurred here, it was appropriate to limit our analyses to this region in order to appropriately detect a difference in the change measure. Prevention of vertebral fractures did not affect the progression of osteophytes or disc-space narrowing. Although the differences in the change scores were small and may not be clinically meaningful, these results support the possibility that there is an effect on a biological process. Thus, alendronate may have a role in altering the process of osteophyte formation and possibly disc-space narrowing progression. Given the association between osteophytes in the spine and elsewhere, the implication of the effects of alendronate on spinal osteophyte progression may be important for other joints such as the knee where OA is common and debilitating. Any effect of alendronate may be due to direct effects on endochondral ossification or bone remodelling. Alternatively, this may be due to other effects on spinal disc disease, such as through chondroprotection. Similarly, it is unclear if effects on disc-space narrowing are direct or via osteophyte effects. In either case, subchondral bone integrity is likely important for cartilage integrity.

Although much focus has been placed on cartilage degeneration in OA research, there is a growing interest in the potential role of bone pathology in the initiation and progression of OA.15 22 Increased subchondral bone turnover, as measured by bone scintigraphy, has been associated with progression of radiographic OA and with pain upon joint loading.23 24 Bone loss defined by bone densitometry has also been associated longitudinally with osteoarthritis progression.25 Subchondral bone attrition may be more prevalent than previously thought in radiographically normal or early OA knees.26 Thus, agents that alter bone metabolism may offer one therapeutic strategy in the management of OA.

Other evidence exists supporting a possible favourable effect of antiresorptive agents on OA. For example, many but not all epidemiological studies of OA have shown that women on postmenopausal hormonal therapy have a lower than expected prevalence of OA, and longitudinal studies of knee OA have shown that women on postmenopausal hormonal therapy experience less disease progression than women not on them, likely due to effects on osteophytes.27 However, although most studies have shown a reduction in the risk of osteophytes in knee and hip OA in those women using postmenopausal hormonal therapy, its effects on joint space narrowing and pain in hip and knee OA have been conflicting.28^–30 In a dog model with anterior cruciate ligament transection, dogs treated with calcitonin had less severe OA lesions compared to those not treated with calcitonin.31 In this model, rates of bone resorption occurred early, and were felt to contribute to cartilage degradation. Ovariectomised Sprague–Dawley rats treated with oral calcitonin had less cartilage erosions than those not treated, similar to the effects noted with 17β-oestradiol supplementation.32 Whether such cartilage effects are related to underlying changes in bone is not known.15

Human clinical trial data have been sparse, with a recent reanalysis of data from a strontium ranelate trial demonstrating reduced radiographic spinal OA progression and back pain in osteoporotic women with prevalent spinal OA.33 In terms of other bisphosphonate studies, one trial of risedronate did not demonstrate any benefit for structural progression or symptoms of knee OA, although a subgroup analysis of joint-space narrowing progressors seemed to indicate that in that subgroup, risedronate may preserve the structural integrity of subchondral bone in knee OA.34 35 Another trial of risedronate did demonstrate an improvement in patient global assessment, but not in structural progression.36 Of note, analysis conducted in the original FIT trial demonstrated that alendronate reduced back pain independently of its effects on vertebral fractures, raising the question of whether the effect on back pain was perhaps due to effects on spinal disc degeneration.37 Although the results of the two risedronate trials were null, there was a suggestion of a possible favourable benefit, and they may not have been able to definitively demonstrate a benefit due to length of study. The FIT study provided a unique opportunity of a longer follow-up than those studies, and an anatomical site, the spine, which includes numerous points of observation that may add to the ability to detect subtle differences over and above that which can be seen in the knee. Nonetheless, the exact mechanism of action for bisphosphonates in OA is not yet known and may reflect effects on chondrocytes and/or on bone remodelling so as to limit cartilage degeneration through promotion of chondrocyte proliferation, inhibition of matrix metalloproteinases (MMPs) and apoptosis and possibly anti-inflammatory effects.8 15 Further, different bisphosphonates may have different effects.8 38

A limitation to this study, however, is that direct implication for OA at sites outside the spine can only cautiously be drawn, although spinal osteophytes do correlate with osteophytes in other sites.5 Repetitive mechanical loading in rabbit spines has been demonstrated to cause abnormalities in the annulus fibrosus accompanied by early osteophytes at the same disc level.39 In humans, increasing severity of osteophytes in the spine have been associated with increasing disc-space narrowing.40 It is not clear that the pathophysiology of spinal disc degeneration is directly relevant to OA in peripheral joints. In rabbit knees that underwent osteotomies to induce altered mechanical loading, the overloaded condyles demonstrated cartilage damage and associated osteophytes.41 Although hyaline cartilage in the knee and hip are obviously distinct from the fibrocartilage of the intervertebral disc, in experimental rabbit models, the inducement of intervertebral disc degeneration by ventral nuclear herniation reliably produces vertebral osteophytes at sites adjacent to cartilage loss.42 These osteophytes arise from proliferating inner annular fibres that undergo metaplasia into cartilage, calcify and proceed through an endochondral ossification sequence,42 which is the same process by which osteophytes in the knee, hip and hand form; however, in addition, osteophytes can also arise directly from periosteum through intramembranous ossification.

It is unclear what the effect we found signifies clinically. The treatment effects we found are similar to findings from a recent study of strontium ranelate on spinal osteoarthritis.33 When, like the strontium ranelate study, we evaluated the proportion of individuals in each treatment arm whose osteophyte score increased by at least 1 in the lumbar spine, we obtained a similar risk ratio of 0.78 (compared to 0.84 in their study), while the same analysis for disc-space narrowing yielded a risk ratio of 0.74 (compared to 0.67 in their study). Nonetheless, we may have only found a small change in progression of osteophytes and disc-space narrowing because we studied postmenopausal women with low bone mineral density, while osteoarthritis, particularly presence of osteophytes, is more commonly cross-sectionally associated with higher bone mineral density. However, Zhang et al demonstrated an increased risk of osteoarthritis progression in older women with low bone mineral density.25 Regardless, that we found an effect of treatment on spinal osteoarthritis supports the likelihood that the prevalence of osteoarthritis in this postmenopausal group was sufficient, and we may have found an even larger effect were we to study a higher risk population.

In conclusion, this study raises the intriguing possibility that bisphosphonates may have a structural effect on pathological processes seen in OA, lending further support to evaluating the role of bone and antiresorptive agents in OA.