Introduction

Joint destruction from various pathologies, most notably rheumatoid arthritis (RA) and osteoarthritis, leads many individuals to elect total joint replacement. Worldwide, more than 1.3 million total joint arthroplasties are performed each year [1]. This number can be expected to increase dramatically in the 21st century. While total joint replacement is remarkably effective in relieving pain and improving function and mobility, it it not without complications. Up to 20% of patients so treated will show evidence of osteolysis within 10 years [2,3,4]. This osteolysis usually leads to implant failure and need for revision arthroplasty, which has a poorer clinical result and a shorter duration of survival than primary total joint replacement [5,6]. Because of such failures, many younger people who would otherwise be excellent candidates for surgery are told to wait, because they might need two or three revisions in their lifetime. Therefore, a clinical intervention to prevent prosthetic implant loosening is greatly needed.

Prosthesis failure results from multiple factors, including those relating to materials, biomechanics, and host responses. The quest for more durable and wear-resistant materials, as well as for better implant designs, and the study of the forces involved in implant integration and prosthesis failure continue to be areas of active investigation. Several groups, however, have focused on the host response to debris produced by wearing of the joint, postulating that wear-debris-induced osteolysis in the main cause of failure of prosthetic implants [7]. In this model, wear debris generated from the prosthesis is phagocytosed by macrophages and initiates an inflammatory response that leads to the recruitment of activated osteoclasts and to osteolysis at the bone-implant interface. Several lines of evidence support this model. First, as many as 109 particles per gram of tissue can be recovered from the inflamed membrane attached to the failed prosthesis after revision surgery [8]. Second, ingestion of wear-debris particles induces cytokine production by mononuclear phagocytes in vitro [9]. Third, high concentrations of cytokines, including TNF-α, that are produced by macrophages are found in the fluid and tissue surrounding loose implants [10,11,12]. Fourth, conditioned medium from monocytes stimulated by wear debris can stimulate increased bone resorption in vitro [13]. Fifth, animal models of wear-debris-induced osteolysis have demonstrated the importance of cytokines in this process [14,15].

This wear-debris-induced osteolysis, which is associated with aseptic loosening, is very different from the phenomenon of stress shielding. In stress shielding, an implant takes on a portion of the mechanical load transmitted to the skeleton and shields bone from this stress [16,17,18]. Since bone metabolism is dependent upon mechanical load, bone density decreases in the affected area. Stress shielding is different in several ways from the inflammatory bone loss that occurs in response to particulate debris. First, stress shielding occurs in the absence of inflammation [18]. Second, it occurs around implants (such as rods, plates and screws) that do not release particles [19]. Third, it is not influenced by polyethylene or the bearing surface, but is reduced by using implants that have a lower modulus of elasticity so that bone takes on more of the mechanical load [16,17]. Fourth, like disuse osteopenia or osteoporosis, stress shielding increases the general porosity of bone, whereas aseptic loosening is associated with localized endosteal bone erosions [20]. Fifth, and most importantly, stress shielding has not been associated with mechanical loosening of the implant [17,18,21,22].

The first clinical manifestation of prosthesis failure is pain with associated radiographic evidence of osteolysis (Fig. 1a). If the volume of osteolysis is small (up to 2 mm in diameter), osteolysis often does not progress and the implant remains fixed. However, when the lesion is greater than 2 mm, osteolysis usually continues rapidly, leading to implant failure. In these lesions, bone is resorbed by osteoclasts and is replaced by a fibro-inflammatory membrane containing lymphocytes, macrophages, and fibroblasts (Fig. 1b) [7]. Although the histopathology and initiating mechanisms differ from those for RA, the tissue reaction in peri-implant osteolysis resembles the pannus of RA in its tendency to produce localized cytokine-mediated bone loss. Thus, a central aim in developing a therapeutic intervention for aseptic loosening is to identify a drug that will eliminate or dramatically reduce inflammation in the periprosthetic synovium-like membrane.

TNF-α has been identified as a drug target in aseptic loosening for many of the same reasons it has been a focus in RA. First, since addition of anti-TNF-α antibodies inhibits the production of other pro-inflammatory cytokines such as IL-1, IL-6, IL-8, and GM-CSF (granulocyte-macrophage colony-stimulating factor) by synovial tissue, it has been proposed that this factor is at the apex of the pro- inflammatory cytokine cascade in the synovium [23,24,25]. Another reason is that TNF-α can induce joint inflammation and proliferation of joint cells [26]. Also, it can stimulate bone resorption by inducing osteoclastogenesis and activating mature osteoclasts [27]. A fourth reason is that TNF receptor I knockout mice have virtually no osteolytic response to polymethylmethacrylate [15] or titanium [14]. And finally, in animal models, the TNF-α antagonist etanercept has been used to prevent wear-debris-induced osteolysis [28,29].

Figure 1
figure 1

Radiographic and histologic findings in periprosthetic osteolysis and loosening of the prosthesis. (a) The radiograph demonstrates periprosthetic bone erosions along both the medial and lateral endosteal bone surfaces. The femoral head is eccentrically placed in a superior position in the acetabular cup, indicating polyethylene wear and the generation of particles. (b) The bone in the osteolytic lesions is replaced by fibro-inflammatory tissue (arrow) consisting of a background of fibroblasts with a diffuse infiltrate of inflammatory cells (lymphocytes, plasma cells, and macrophages), which is most intense in the top left-hand quadrant of this micrograph. Released particles of wear debris accumulate in this tissue, which acts as a reservoir for them and thus enhances the progression of the bone loss and further loosening. This patient underwent a revision arthroplasty.

Full size image

Therapies for aseptic loosening

There are currently no drugs specifically approved for the treatment of aseptic loosening of prostheses. However, the above paradigm for loosening (ie wear-debris-induced, TNF-α-mediated inflammation resulting in osteoclast activation) suggests that three categories of drugs should be tested for their ability to prevent or treat loosening of prosthetic joints. The first category is the bisphosphonates. These drugs inhibit osteoclasts, are effective, and are widely used to prevent or treat osteoporosis. A small, recent clinical study has shown that alendronate can reduce the periprosthetic bone loss that develops soon after total hip replacement [30]. However, as the authors of that study pointed out, this early bone loss is probably secondary to stress shielding rather than to wear-debris-induced inflammation. Indeed, patients who had had a total hip replacements more than 5 years previously or who were awaiting revision surgery for loosening did not have a similar increase in periprosthetic bone density when treated with alendronate. Unfortunately, periprosthetic osteolysis was not an end point in that study. The effect of bisphosphonates on inflammation-induced osteolysis has also been evaluated in patients with RA. In three studies, the effects of bisphosphonates on radiographically evaluated erosions have varied but have been predominantly negative. In one small study (a total of 27 patients randomized to either pamidronate, 1000 mg/day by mouth, or a placebo, for a year), erosions in the treated group progressed less rapidly [31]. However, no such effect was found in two larger studies (a total of 40 patients given pamidronate, 30 mg, intravenously per month or placebo by monthly infusion, and a total of 105 patients given pamidronate, 300 mg/day by mouth, or a placebo for 3 years) [32,33]. In this last study, spinal and femoral bone mineral density significantly improved in the treated group, even though the erosions progressed. Although it is possible that the doses used were inadequate to block osteolysis, these studies in humans suggest that bisphosphonates may be less effective for use against inflammation-induced osteolysis than against generalized osteoporosis. On the other hand, a report that zoledronate blocks bone resorption in a rabbit/carrageenan model of inflammatory arthritis [34] indicates that bisphosphonates may be effective in some types of inflammation-induced osteolysis.

A second category of drugs for the treatment of prosthetic loosening are those designed to inhibit TNF (etanercept and infliximab). Both of these agents are potent inhibitors of synovial inflammation [35,36] and both have been approved worldwide for the treatment of RA. Importantly, recent studies have shown that both can block erosions in this disease [37,38]. Because of their effects on erosions in RA and on wear-debris-induced asteolysis in animals [28,29], these anti-TNF agents are the most promising medications already available for the treatment of established loosening. However, they are also remarkably expensive and therefore should not be used to treat loosening until their efficacy is proven in clinical trials.

Finally, a third category of drugs to treat prosthetic loosening are the biologics being developed that interfere with RANK/RANK-ligand signaling. RANK (receptor activator of NFκB) [39] is a receptor on osteoclasts and osteoclast precursors that transmits a signal required during osteoclast and lymph node development [40]. RANK ligand (also known as OPGL, ODF, and TRANCE) is an agoinst for RANK, and is expressed on osteoblasts and activated T cells [41]; it provides the essential signal for osteoclast differentiation and survival [27]. Osteoprotegerin (OPG) is a natural decoy receptor that binds to RANK ligand and prevents its interaction with RANK [42]. The biologics being developed to inhibit osteoclasts include recombinant OPG and a soluble form of RANK. The potency of these molecules is best illustrated by the phenotype of transgenic mice that overexpress these factors and suffer from severe osteopetrosis [42,43]. Preliminary studies in an animal model indicate that a soluble chimeric RANK:Fc molecule has no effect on inflammation but completely inhibits osteoclast induction and wear-debris-induced osteolysis in vivo [29]. Thus, these new RANK-based biologics, which are even more potent inhibitors of osteoclasts than the bisphosphonates, may offer another future approach to the treatment of established loosening or to its prevention.

Based on the animal studies summarized above, a strong case can be made for the involvement of TNF in at least some models of wear-debris-induced osteolysis. We believe that clinical studies with TNF inhibitors will be a direct way of testing the validity of anti-TNF therapy in preventing prosthetic hip loosening.