The futility of current approaches to chondroprotection

Introduction

While affecting all structures within a joint, osteoarthritis (OA) has generally been defined as a disease characterized by loss of hyaline articular cartilage. Research has focused on factors that affect chondrocyte metabolism and the mechanism of cartilage matrix degradation. This has yielded creative approaches to cartilage protection, including using agents active on bone turnover, antibiotics that inhibit enzymes that break down cartilage, and other drugs that abrogate cytokine effects on cartilage. In addition to using agents that modify the behavior of native tissue, another active area of research is focused on developing cartilage transplantation techniques, which encourage cartilage matrix growth and the development of a functional matrix via gene expression manipulation or via cytokine stimulation. These types of therapies are termed chondroprotective treatments because they are focused on preventing the loss or encouraging the regrowth or healing of damaged articular cartilage. This is in contrast to current treatments, which focus on relieving the symptoms of OA without necessarily affecting the progression of joint damage.

Despite advances in our understanding of cartilage metabolism and disease process and the development of new agents and treatment strategies, chondroprotective agents have not been added to the therapeutic armamentarium. This is generally because such treatments have not demonstrated clearcut efficacy in human disease. This statement in itself is controversial, because a number of treatments have shown possible efficacy. Even so, the Food and Drug Administration has not approved any therapy as being chondroprotective or structure modifying in OA, and major efforts by industry to develop or test chondroprotective therapies have generally been frustrating and unsuccessful.

The goal of this article is not to critically evaluate the in vitro or ex vivo evidence regarding the chondroprotective effectiveness of agents, but rather, to argue that in vivo chondroprotection will likely be disappointing even in the face of real in vitro or ex vivo efficacy. We suggest that agents already developed may, under optimal circumstances, show efficacy for structure modification in OA. The point of this article is to argue that current approaches to developing structure-modifying treatment for OA that focus on chondroprotection only are flawed and that trials testing new agents are likely to fail, just like previous attempts. We suggest that a new conceptual approach is needed before therapies that protect against structural progression in OA can be successfully developed.

Failure of trials of putative chondroprotective agents in OA

Risedronate and doxycycline have been among the possible chondroprotective agents tested in recent years. Risedronate, a bisphosphonate approved for the treatment of osteoporosis, showed promise as a chondroprotective drug in animal studies (1, 2). Procter and Gamble undertook 2 large multicenter trials to test the efficacy of risedronate in human OA. The British Study of Risedronate in Structure and Symptoms of Knee Osteoarthritis (3), a 1-year trial performed in England with radiograph outcomes (fluoroscopically positioned), showed favorable effects of risedronate compared with placebo in terms of pain relief, but there were too few cases of progressive radiographic OA to definitively evaluate chondroprotection. Using fluoroscopically positioned knee radiographs, a large multinational trial failed to show any efficacy on structural outcomes (4).

Doxycycline, an antibiotic, has been shown to have broad inhibitory effects on cartilage matrix metalloproteinases (MMPs) (5, 6). A large multicenter National Institutes of Health trial that tested doxycycline as a chondroprotective agent showed effects in one OA knee in a person but failed to show an effect even in the direction of efficacy in the contralateral knee also affected by OA (7). The results are difficult to interpret and may represent a lack of chondroprotection. An alternate explanation is that cartilage MMP inhibition induced by doxycycline may be relevant only for joints in later disease stages and not for joints with early OA. This trial was not designed to examine effects on symptoms, and the trial's results leave questions about efficacy for a compound that animal studies suggested was likely to be chondroprotective.

There are 2 industry-based trials of glucosamine, which, using radiographic joint space loss as the outcome, suggest chondroprotective effects (8, 9). These trials, both with sample sizes generally considered too small to be likely to detect chondroprotective effects, used radiographs of fully extended knees that do not provide reproducible or accurate joint space measurement. The results should be regarded with skepticism. The history of the development of OA therapeutics is filled with other examples of unclear success or outright failures.

If a chondroprotective agent blocked all cartilage degradation, its efficacy might be detected using any testing strategy. But we believe that 100% efficacy is not possible given the biology of the disease. OA is a complex disease in which both biochemical and mechanical factors influence disease progression. Furthermore, it is a heterogeneous disorder where, at some point, matrix regeneration or preservation is simply impossible. Therefore, the current method of staging the disease for clinical trials may be insufficient to adequately identify suitable candidates for chondroprotective therapies. Additionally, the staging method must be sensitive enough to identify early disease states where the potential for disease modification is still present. We suggest that new approaches to drug development in OA are necessary before clearcut success is achieved. Given the advances in treatment among a broad group of chronic degenerative diseases of aging, OA is frustratingly among the few remaining where efficacious preventive strategies or treatments are generally unavailable.

Why has there been a failure of chondroprotection?

Attempts to develop pharmacologic chondroprotective treatments have failed in part because they have assumed that pharmacologic therapies would work regardless of the mechanical environment. We propose that an appreciation of the mechanical environment in a joint, at both the organ and tissue levels, is a prerequisite to determining whether a particular joint will benefit from drugs that may protect cartilage.

There are 2 such reasons why abnormal mechanics may overwhelm the chondroprotective ability of an agent. First, by the time OA is symptomatic in many patients, organ-level mechanical abnormalities that either caused or developed with disease have progressed to include joint malalignment, bony remodeling, and ligamentous stretching, so that any effect of a drug on cartilage will have only a minor, and probably reversible, effect on the joint as a whole.

Second, tissue-level dynamic stresses on cartilage in OA joints may exceed thresholds that could be reversed by any effective pharmacologic agent. For example, even in a normally aligned joint with early OA, there will be focal areas with early damage. The initial changes would be loss of proteoglycans, which would decrease the mechanical stiffness of the tissue. In more degenerated tissue, the collagen fibrils unwind, leading to tissue swelling and further loss of tissue mechanical stiffness. In these focally damaged areas, even normal joint loading can lead to a concentrated area of increased mechanical strain that may be beyond the chondroprotective ability of the intervention. In this situation, we suggest that no pharmacologic agent is likely to work as a sole therapy.

There is some proof of the importance of organ-level mechanics in the progression of OA, although the actual focal loads have not been quantified. For example, we have recently shown that obesity, a cause of a marked increase in loading across the knee and of new- onset disease in joints that start off neutrally aligned, does not increase the risk of disease progression when limbs are severely malaligned (10). These findings suggest that severe malalignment confers such a focal overloading of cartilage during daily activities that additional loading from obesity is immaterial and that all joints in this circumstance experience cartilage loss whether obesity exists or not. According to this theory, pharmacologic agents would not work in a severely malaligned joint. Further, malalignment not only produces cartilage loss focally, but leads to bone damage, evidenced as bone marrow lesions (11) on magnetic resonance imaging (MRI), suggesting that the focal transmitted load is great.

In addition to the organ- or limb-level and tissue-level biomechanical factors, at some point, articular cartilage damage may simply be irreversible by any means. This may have already occurred at the point at which joint space loss is detected on radiographs or when cartilage volume loss is noted on MRI. By the time actual tissue is lost, there will certainly be collagen and proteoglycan damage at the gross and microscopic levels. Therefore, depending on the chondroprotective mechanism of the therapy, the disease state of the cartilage tissue may dictate whether a positive effect is seen. For example, a protease inhibitor that blocks proteoglycan loss may be ineffective in tissue with a compromised collagen matrix. The proteoglycans will simply diffuse out. To demonstrate the efficacy of such drugs, joints at a much earlier stage of disease need to be studied. Functional imaging techniques, such as delayed gadolinium-enhanced MRI of cartilage (dGEMRIC) and T1ρ imaging, which provide histology-like images of cartilage structure, may help us appropriately stage these early disease states.

While animal studies of chondroprotective agents suggest that the efficacy of chondroprotective agents can be detected in the presence of surgical lesions and biomechanical abnormalities, animals that are studied have incipient disease with only cartilage affected (12). These animals therefore do not provide good models for fully developed OA in humans.

Many researchers believe that inflammatory processes and/or an imbalance between anabolic and catabolic processes in cartilage are the predominant reasons for the development of the disease and its progression (13); treatments have attempted to manipulate these processes. Mechanical and biochemical pathways are not mutually exclusive.

Measurement of cartilage loss

To date, the amount of OA seen on radiographs has been both the primary method of staging and the primary outcome of trials testing chondroprotective agents. In OA, radiographic progression is remarkably slow and difficult to predict in individual patients. This means that to test treatments for efficacy as chondroprotective agents, trials need to be large and of long duration, at least 2 years. While these barriers are not insurmountable, even well-designed and expensive radiographic studies may fail to show chondroprotection because of the noise of measurement of joint space loss. This noise, or variability, can be introduced by tiny changes in the beam angle from one knee radiograph to a later one. Consistency in beam angle and joint imaging on radiographs, while achievable, necessitates rigorous quality by many technologists at multiple farflung clinical sites. The vulnerability of large-scale expensive efforts to evaluate chondroprotection to trivialities like minor beam angle changes on radiographs imposes an almost insurmountable barrier to testing these agents. In hip OA, it may be easier to evaluate joint space loss on radiographs, but loss is still slow, necessitating the study of large numbers of patients for at least 2 years.

MRI assessment of cartilage loss longitudinally is likely to be the solution to the challenges of radiographic studies. MRI can detect cartilage loss in 3 dimensions, not just a 2-dimensional plane, and has been shown to be more sensitive to change than radiography (14). It is not dependent on the positioning of the beam angle, but rather, permits reconstruction of the knee's cartilage with registration of images. Unlike radiographs, where cartilage thickness is assumed to be equal to the distance between 2 bones (an assumption that is probably incorrect in many knees [15]), MRI directly visualizes cartilage. However, even assessing cartilage loss by MRI has its challenges. While initial longitudinal studies suggest that rates of cartilage loss are more than sufficient for chondroprotection to be demonstrated in a year or 2 (16-18), recent unpublished data from the Osteoarthritis Initiative suggest that cartilage loss may occur much more slowly than previously believed.

The best way to assess cartilage loss and, hopefully, cartilage recovery is currently unknown. Change in cartilage volume may be due in part to expansion of bone ends at the joint margin, with cartilage covering a larger area over time. Cartilage can be measured in several plates in each knee. Should all plates be used to assess treatment effects or just those most loaded? How should data be interpreted if 1 cartilage plate appears to reveal an effect of treatment while others show no such effects? What if both knees are diseased and 1 knee shows a treatment effect, whereas the other knee does not? Problems of multiplicity (multiple cartilage plates in 2 knees followed longitudinally at multiple time points, each assessed for volume and thickness in multiple ways) will certainly arise.

Pain relief and chondroprotection

OA is a painful disorder. However, cartilage is aneural; hence, there is no reason to expect that pain or discomfort would accompany cartilage loss. Indeed, of 2 recent longitudinal MRI studies correlating pain with cartilage loss, 1 found no association (19), whereas the other found an inverse relation (the more pain, the less cartilage loss) (20). Even if a chondroprotective agent were to be identified, it is hard to imagine it being prescribed if it did not lessen pain or discomfort.

Suggested new approaches to the development of structure-modifying treatments in OA

An extension of our argument is that chondroprotective drugs will need to be targeted to joints in which the mechanical environment is favorable. If not, it will be difficult to demonstrate their efficacy. Without a new approach that incorporates an appreciation for mechanical stresses within a joint, there will continue to be a huge opportunity to develop treatments for OA, yet a frustrating failure of effort.

Biochemical changes in cartilage are likely to be reversible (21) and may precede loss of cartilage substance. MRI techniques such as dGEMRIC and T1ρ, which provide a histology-like assessment of the articular cartilage structure, may be an answer. One, dGEMRIC, can be effective in staging hips for pelvic osteotomies (22). In broader terms, we need improved ways of properly staging OA for clinical trials and improved ways of detecting the structure-modifying effect to demonstrate that a chondroprotective agent is efficacious.

If tissue loss is the primary outcome, then quantitative MRI may be sufficient. However, if an early stage of the disease is needed to identify joints with the appropriate potential for response, then a functional imaging technique that may allow for detection of the reversal of the structural tissue damage may be preferable (21). The structural outcome must be appropriate for the structural change and stage of the disease being tested. Without a new approach that incorporates all of these factors, there will continue to be a huge opportunity to develop treatments for OA.

Structure modification must be tied to symptom improvement. We make the assumption that an efficacious chondroprotective agent will improve symptoms, but this may not necessarily be true. How can we protect cartilage and lessen pain? There are, in fact, persuasive examples of the 2 processes occurring in parallel. In knee OA, tibial osteotomy, a surgical treatment that unloads the knee, both reduces pain and prevents further cartilage loss in the affected compartment. Similarly, pelvic osteotomies, designed to correct acetabular dysplasia, also lessen pain and improve proteoglycan concentrations within cartilage (23, 24), suggesting chondroprotection.

These 2 examples suggest not only that loading has relevance for structural deterioration of the joint, but that treatments that alleviate pain can be best developed if the effects of loading are optimally understood. Medical realignment therapy with knee braces has been shown to lessen knee pain (25), presumably as a consequence of the redistribution of load that occurs in patients wearing braces (26). A similar effect may be attained by modifying loading across the knee by redistributing weight-bearing forces in the foot (27) or even by taping the patella so that its excursion avoids areas of painful loading (28). These examples suggest that lowering focal stress enough will not only prevent cartilage loss but will also alleviate pain. Recent studies (29) tying malalignment to incident knee pain confirm that this risk factor for cartilage loss, which increases focal stress in the joint (30), also causes pain.

How does malalignment cause pain, and how does realignment (bracing, osteotomy) alleviate it? This can be done in several ways. First, in damaged cartilage, canals containing nerves and vessels extend through the subchondral plate into the tidemark in OA joints, and the unmyelinated neurons there could sense stress in cartilage, generating nociceptive signals. Second, damage to cartilage parallels damage to underlying bone, seen as bone marrow lesions on MRI and histologically as areas of bone damage with necrosis, fibrosis, and microcracks. These bone lesions could generate pain. Third, mechanical stress to cartilage could induce synovitis through as-yet-unspecified mechanisms, and the synovitis could cause pain.

In summary, we argue that the current approach to testing potential chondroprotective agents in OA is destined to continue to fail because it does not take into account irreversible pathology outside of the cartilage and fails to deal with mechanical aberrancies at the organ and tissue levels that propel disease progression. Opportunities may exist for structure-modifying treatments, but their testing and use need to be targeted toward subsets with nonmechanical disease etiologies and, for weight-bearing joints, may need to be combined with unloading treatments.

AUTHOR CONTRIBUTIONS

Analysis and interpretation of data. Felson.

Manuscript preparation. Felson, Kim.