Unresolved Questions in Rheumatology: Motion for Debate: Osteoarthritis Clinical Trials Have Not Identified Efficacious Therapies Because Traditional Imaging Outcome Measures Are Inadequate


  • Ali Guermazi,

    Corresponding author
    1. Boston University School of Medicine, Boston, Massachusetts
    • Boston University School of Medicine, 820 Harrison Avenue, Third Floor, Boston, MA 02118Kansas University Medical Center, 5755 Windsor Drive, Fairway, KS 66205. E-mail: guermazi@bu.edu kenbrandt@yahoo.com

    Search for more papers by this author
    • Dr. Guermazi has received consulting fees, speaking fees, and/or honoraria from Serono and Stryker (less than $10,000 each) and from Genzyme, Novartis, and AstraZeneca (more than $10,000 each), and he owns stock or stock options in, and is president of, Boston Imaging Core Lab (BICL), LLC.

  • Frank W. Roemer,

    1. Boston University School of Medicine, Boston, Massachusetts
    2. University of Erlangen–Nuremberg, Erlangen, Germany
    Search for more papers by this author
    • Dr. Roemer has received consulting fees, speaking fees, and/or honoraria from Merck Serono (less than $10,000) and owns stock in, and serves as vice president of, BICL, LLC.

  • David T. Felson,

    (In Support)
    1. Boston University School of Medicine, Boston, Massachusetts
    Search for more papers by this author
    • Dr. Felson has received consulting fees, speaking fees, and/or honoraria from Knee Creations, Ltd. (less than $10,000).

  • Kenneth D. Brandt

    (In Opposition), Corresponding author
    1. Kansas University Medical Center, Fairway
    • Boston University School of Medicine, 820 Harrison Avenue, Third Floor, Boston, MA 02118Kansas University Medical Center, 5755 Windsor Drive, Fairway, KS 66205. E-mail: guermazi@bu.edu kenbrandt@yahoo.com

    Search for more papers by this author
    • Dr. Brandt has received consulting fees, speaking fees, and/or honoraria from Pfizer and Heel, Inc. (less than $10,000 each) and Janssen (more than $10,000). He owns stock in Pfizer and received royalties from Oxford University Press for the textbook Osteoarthritis.

In the face of a staggering unmet need for a therapeutic intervention that limits the progression of knee osteoarthritis (OA), there is now a small but sustained effort to develop effective disease-modifying OA drugs (DMOADs). Radiographic measures of knee OA severity, originally developed for epidemiologic research, were repurposed to measure structural progression, and they remain the primary outcome measure accepted by the Food and Drug Administration (FDA) for this indication. However, demonstrating structural modification by a pharmacologic therapy has proved to be a substantial and costly challenge in DMOAD trials, with most trials, if not all, exhibiting null results. A question arises as to whether this failure is a consequence of poor measurement properties of the outcome technology, or lack of biologic efficacy of the putative therapy. Below, experts in the fields of musculoskeletal imaging and OA clinical pathology debate this pivotal question.

  Timothy E. McAlindon, MD, MPH, MRCP

  Co-Editor, Arthritis & Rheumatism

In support, Drs. Guermazi, Roemer, and Felson: Inadequacy of radiography is one cause of the failure of clinical trials to identify a disease-modifying drug for OA


In spite of continued research efforts and millions of dollars invested in clinical trials ([2, 3]), there is still no pharmacologic agent that has been approved by regulatory agencies as a DMOAD. Three ingredients are crucial for the success of any clinical trial: an efficacious compound, inclusion of patients who are most likely to benefit from the treatment, and an end point that defines success or failure. We suggest that refocusing on the eligibility for and end points of OA clinical trials may enhance the likelihood of achieving the desired goal—a treatment that delays the structural progression of OA. In this debate-style article, we will describe some of the current problems that affect eligibility and end points in OA trials, and present the reasons we advocate a shift of the primary imaging modality from radiography to magnetic resonance imaging (MRI).

Inappropriateness of radiography for assessing eligibility

Recruitment of subjects to participate in knee OA trials is dependent on the radiographic definition of OA as defined by the Kellgren/Lawrence (K/L) scale ([4]). Knees with a K/L grade of 0 or 1 are considered not to have radiographic OA, whereas knees with evidence of definite osteophytes (K/L grade 2) or joint space narrowing (JSN) (K/L grade 3 or 4) are considered to have radiographic OA. However, these knees may have heterogeneous structural joint damage that cannot be visualized by radiography. In a recent population-based observational study ([5]), osteophytes were detected by MRI in 524 (74%) of 710 knees with normal radiographic results (K/L grade 0), suggesting that radiography lacks sensitivity for detection of early OA structural changes.

Pathologic features of OA that have been reported to be associated with pain include bone marrow lesions, synovitis, effusion, periarticular cystic lesions, and meniscal tears ([6]). However, none of these joint pathologies can be visualized by radiography (Figure 1). In the above-mentioned study ([5]), the prevalence of bone marrow lesions, meniscal tears, and synovitis/effusion, respectively, was 52%, 24%, and 36% in radiographically normal knees. These numbers imply that knees with a K/L grade of 0 should not be automatically considered structurally normal and be excluded from OA clinical trials.

Figure 1.

A, Anteroposterior radiograph of a knee with a Kellgren/Lawrence (K/L) radiographic osteoarthritis (OA) grade of 2. A “definite” osteophyte can be seen at the margin of the lateral tibial plateau (arrow). No joint space narrowing (JSN) is seen. B, Anteroposterior radiograph of a knee with a K/L grade of 2. There are small but definite osteophytes at the medial and lateral tibial plateau (arrowheads). No JSN is seen. C, Same knee as in A, visualized by magnetic resonance imaging (MRI) on the same day. Sagittal intermediate-weighted fat-suppressed (FS) MRI shows a large subchondral bone marrow lesion in the medial femur (arrows), with adjacent focal cartilage loss. This knee represents the “osseous” OA phenotype, as marked subchondral bone changes are present. D, Sagittal intermediate-weighted FS image of same knee as in B. Marked joint effusion is observed (asterisk). Note that synovial thickening cannot be differentiated from effusion on nonenhanced scans. In addition, there are definite osteophytes at the posterior medial femur (arrow) that cannot be seen on the radiograph. This knee represents the “inflammatory” OA phenotype.

We also need to recognize that OA has multiple “phenotypes” that are not apparent on radiography ([7-9]). OA should be considered a complex and heterogeneous disease with multiple disease-triggering factors at the joint level and the systemic level, as well as a varied clinical course. Including patients with different phenotypes in the same clinical trial may increase the ambiguity of the results and hinder interpretation of results.

Limitations of radiography as an end point in OA DMOAD trials

Slowing of joint space loss is a widely used radiographic outcome measure ([10]). Indeed, reduced loss of joint space as seen on radiography is the only imaging biomarker that has been included by the FDA in its draft recommendation, and also by the European Medicines Agency (EMEA). Joint space width (JSW) measurement by radiography is still recommended by the Osteoarthritis Research Society International (OARSI) for clinical trials of structural modification ([10]). However, radiography has several limitations as an end point–defining imaging modality in OA DMOAD trials, as described below.

First, focal cartilage loss or diffuse partial-thickness loss in the tibiofemoral joint often does not produce a change in JSW (Figures 2 and 3). A study based on data from the Multicenter Osteoarthritis Study (MOST study) showed that 69 (42%) of 164 knees without JSN or definite osteophytes had focal or diffuse partial-thickness cartilage loss in the medial tibiofemoral joint as detected on MRI ([11]).

Figure 2.

Lack of sensitivity of radiography to longitudinal change. A, Baseline anteroposterior radiograph shows definite osteophytes at the medial tibial joint margin, but no JSN. B, At 24 months, visualization of osteophytes and normal joint space width is identical to that at baseline, with no progression. C, Corresponding baseline intermediate-weighted FS MRI of the medial tibiofemoral joint shows normal cartilage thickness and no additional MRI features of OA. D, Image obtained 24 months later shows diffuse partial-thickness cartilage loss at the posterior medial femoral condyle (arrowheads). Due to the posterior location of the cartilage, radiography cannot depict such changes, whereas they are clearly shown on the MRI. See Figure 1 for definitions.

Figure 3.

Focal defects seen on MRI. A, Coronal intermediate-weighted FS image shows focal full-thickness cartilage damage (arrow). This finding cannot be depicted by radiography, as joint space width is not affected by such subtle changes. B, Coronal fast low angle shot (FLASH) image of same defect exemplifies that not all MRI sequences are suited for visualizing cartilage changes in the same manner. FLASH is a high-resolution 3-dimensional gradient-echo sequence that is commonly used for cartilage segmentation because of the high signal and contrast-to-noise ratio between cartilage and the subchondral bone. However, the focal defect that was easily recognizable in A does not appear on the FLASH image. Only a circumscribed area of low signal is shown (circle), that could easily be mistaken for an intrachondral calcification and not a defect. See Figure 1 for other definitions.

Second, radiography lacks sensitivity for detecting structural change over time ([12]). OA is a very slow process. A recent study based on Osteoarthritis Initiative (OAI) data showed that only a few knees in the large OAI cohort exhibited ongoing change; most were stable. In stable knees, JSW remained the same for as long as 4 years ([13]). In such knees, it would seem very difficult to test the efficacy of a drug, using radiographic outcome measures, in the usual 1- or 2-year time frame of a clinical trial in OA.

Third, joint space loss seen on radiography lacks specificity, as such change may be caused by cartilage loss, meniscal damage, or meniscal extrusion ([14]) (Figure 4). Adams and colleagues demonstrated in a small cohort that >50% of patients with radiographic JSN and meniscal extrusion (17 of 32) had no loss of cartilage as seen on MRI ([15]).

Figure 4.

Nonspecificity of radiography for longitudinal assessment. A, Baseline anteroposterior radiograph shows doubtful JSN in the medial tibiofemoral compartment (arrowheads). B, Image obtained 24 months later shows a definite increase in medial joint space loss (arrows). C, Corresponding baseline intermediate-weighted MRI shows normal cartilaginous tibiofemoral surface and no relevant meniscal extrusion. D, MRI obtained 24 months later shows definite meniscal extrusion (arrows; medial meniscus protrudes medially beyond the thin line), which is likely responsible for the JSN observed on the radiograph. See Figure 1 for definitions.

Last, due to its 2-dimensional nature, the radiographic appearance of tibiofemoral joint space is heavily dependent on knee positioning and the angle of joint flexion. Unless exactly the same positioning and angulation are used at baseline and followup, measurement error may occur (Figure 5). A recent study showed that shifting the foot forward 6 cm on a foot plate to simulate the varying extent of lower leg extension in a semiflexed knee led to a statistically significant increase in the minimum JSW (+0.07 mm/cm of shift). This was clinically relevant when compared with a mean loss of JSW of 0.11 mm in knees with radiographic progression over 2 years ([16]).

Figure 5.

Positioning. A, Baseline anteroposterior radiograph shows equivocal JSN at the medial tibiofemoral joint. Arrowhead delineates the lateral surface of a fabella, a small posterior ossicle. B, At 12 months, findings on the followup radiograph suggest a marked increase in JSN (arrows in A and B) compared to the baseline image. Note that the lateral border of the fabella now projects posteriorly to the lateral femoral condyle (arrowhead) and not laterally as in A. C, Baseline coronal intermediate-weighted MRI shows cartilage loss at the medial femur and tibia (arrows). Minimal medial meniscal extrusion is also seen (thin line). D, MRI obtained 12 months later shows identical cartilage damage and meniscal extrusion (arrows and thin line). No change between baseline and followup is observed. The difference in joint space width from baseline to followup is caused by positioning of the knee and, in this case, also malrotation. See Figure 1 for definitions.

As clearly evidenced by these multiple factors, it is time that OA research break with the practice of depending entirely on radiography to define eligibility and outcomes in OA trials.

MRI as a tool for eligibility assessment

MRI can be used to visualize not only cartilage, but also clinically relevant features of disease that are associated with pain and structural progression, including bone marrow lesions, meniscal abnormalities, and synovitis ([6, 17]). Early structural abnormalities that are visible only on MRI can be used to define trial eligibility criteria as well as to classify patients into the various phenotypes.

Using data from the MOST study, Roemer and colleagues showed that a high body mass index, meniscal damage, synovitis or effusion, or any severe baseline MRI-depicted lesions were strongly associated with an increased risk of rapid cartilage loss over a 30-month period ([18]). Generally, it is thought that measurable quantitative cartilage loss rarely occurs over a period of <6 months, but according to data from the Joints on Glucosamine Study ([19]), cartilage loss, as well as development or progression of bone marrow lesions and meniscal extrusion, does occur in this time frame, and some of these features could be used as structural outcome measures. Further, the presence of any of these features at baseline can predict rapid cartilage loss later, permitting the selection of a trial population at high risk of disease progression. Thus, depending on the length of a trial, investigators could select subjects with these MRI-based risk factors for enrollment in preventative or therapeutic OA trials.

MRI as an outcome measure

MRI allows direct visualization of cartilage 3-dimensionally, as well as quantitative, semiquantitative, and compositional assessment of cartilage. Clinical trials to evaluate repair methods for focal cartilage defects also use MRI-based end points, e.g., semiquantitative scoring of the repair zone ([20]). Two recent pilot double-blind randomized controlled trials (RCTs) using MRI-based outcome measures for cartilage have yielded encouraging results ([21, 22]).

Two other potential MRI-based outcome measures are bone marrow lesions and synovitis. Although not all published studies have demonstrated an association between bone marrow lesions and pain, a recent study in the MOST cohort by Zhang et al showed that fluctuations of bone marrow lesions between baseline and followup are directly associated with fluctuations in pain ([23]). Thus, a reduction in the size of bone marrow lesions (assessed either semiquantitatively [24] or quantitatively [25]) might be a suitable biomarker for assessing the efficacy of a therapeutic agent. Indeed, a recent RCT demonstrated that, compared with placebo, zoledronic acid reduced both knee pain and the size of bone marrow lesions ([26]). It has also been shown that MRI-detected synovitis is associated with pain ([27]). Thus, a reduction in synovitis is another potential outcome target.

In a recent systematic review by Hunter and colleagues ([28]), the magnitude of the standardized response mean was 0.86 for quantitative cartilage morphometry in the medial tibiofemoral joint and 0.55 for semiquantitative cartilage scoring in the same joint. For bone marrow lesions and synovitis, respectively, the standardized response means for semiquantitative scoring were 0.43 and 0.52, indicating adequate responsiveness. Thus, both quantitatively and semiquantitatively assessed cartilage morphometry, as well as semiquantitatively evaluated bone marrow lesions and synovitis, were shown to be effectively responsive in longitudinal studies. These findings strongly support the notion that MRI-based outcome measures should be included in updated regulatory guidance statements for OA clinical trials.


The failure of clinical trials to demonstrate treatment efficacy of OA drugs to date may be due, in part, to the failure of radiography-based eligibility criteria and end points. The use of MRI can help investigators select subjects who are most suitable for a specific aim of the trial, taking into account disease phenotypes and potential treatment targets. MRI seems to be the modality of choice on which to define imaging-based criteria for inclusion in and end points of OA trials.

In opposition, Dr. Brandt: Clinical trials of disease-modifying drugs for OA have failed to identify efficacious therapies because they have inadequately addressed the biomechanical etiopathogenesis of common, garden-variety OA


Because RCTs of putative DMOADs have not shown them to be unequivocally superior to placebo in patients with knee OA and radiographic limitations can affect patient eligibility and end points for DMOAD RCTs, a shift from radiography to MRI as the primary imaging modality for DMOAD RCTs in patients with knee OA has recently been advocated. For some anatomic structures of the knee, clearly, MRI is much more sensitive in detecting structural abnormalities than radiography. But in some cases, might MRI be too sensitive? How often do abnormalities seen on MRI not progress within the time constraints of a DMOAD RCT? Or regress? Or have no clinical significance? Furthermore, in the knees of patients with K/L grade 3 OA ([39]), Lyon schuss radiography showed greater sensitivity to change in articular cartilage thickness at 12 and 24 months than did quantitative MRI (qMRI) ([40]). It is unclear how often the results of a DMOAD RCT, or its cost, would differ significantly if MRI, rather than radiography, had been used as the primary outcome measure.

Reduction in JSW of the medial tibiofemoral compartment is widely considered to represent a surrogate for the thickness of the articular cartilage at that site and is the only imaging marker recommended for this purpose by the FDA, EMEA, and OARSI. However, although the most striking pathologic changes of OA are often seen in articular cartilage, OA is not a cartilage disease. In addition to articular cartilage, OA involves all of the tissues of the joint—the subchondral bone, synovium, capsule, meniscus, ligaments, periarticular muscle, and nerves. In the knee, JSN may be due to pathology in the meniscus, rather than in articular cartilage ([41]). OA represents the failure of an organ—the diarthrodial joint. Just as the heart can fail because of a primary disorder in the endocardium, myocardium, or epicardium, the joint can fail because of a primary abnormality in any of the tissues mentioned above, and any of these may be the first to fail ([42]).

Limitations of radiography of the OA knee are well recognized, but it is less well appreciated that the rate of JSN and sensitivity to change in JSW in protocols that aim to achieve standardized, reproducible radioanatomic positioning of the knee can vary markedly. Nonetheless, eligibility criteria for DMOAD RCTs are often based on the severity of OA at baseline, as defined by the K/L grade. In longitudinal studies the K/L grade is widely accepted as a surrogate for progression of structural joint damage in OA. However, greater flexion of the knee, e.g., in a fixed flexion or Lyon schuss view, may “convert” a K/L grade 2 classification of a knee in a conventional standing anteroposterior view into a K/L grade 3 designation. In a direct comparison of the Lyon schuss and fixed flexion protocols, the Lyon schuss view showed a more rapid rate of JSN and greater sensitivity to change in JSW than a concurrent fixed flexion view, because of smaller differences in alignment of the x-ray beam with the plane of the medial tibial plateau ([43, 44]). In serial fixed flexion radiographs, the error of measurement of JSW in normal knees was as large as the mean rate of JSN at 12 months in OA knees, or larger ([45]).

The above-described change in JSW, furthermore, is not random. Depending on the direction of change in the magnitude of the anterior and posterior margins of the medial tibial plateau (intermargin distance), the rate of JSN seen on serial Lyon schuss radiographs will increase or decrease ([44, 45]). When intermargin distance on the followup radiograph was larger than that on the baseline radiograph, the rate of JSN in OA knees was relatively slow, but when intermargin distance on the followup image was smaller than that in the baseline image, the rate of JSN was significantly more rapid, and sensitivity to change in JSN significantly greater ([42]). Unless precautions are used to assure a small intermargin distance (e.g., <1.0 mm) (5) in serial knee radiographs used in DMOAD RCTs, there is a concern not only with respect to the possibility of false-negative results in measurements of JSW change, but also with respect to false-positive results. It is important to differentiate RCTs yielding false-negative results from RCTs that fail to demonstrate treatment efficacy due to, e.g., inadequate sample size.

The importance of excess mechanical stress as a cause of much of the damage in the OA joint has led some investigators to conclude that “ … if this is not addressed, drugs that slow the structural progression [of OA] may not be identified” ([46]). This, however, does not take into account a point of great etiopathogenetic importance: the data provide powerful evidence that common, garden-variety OA is not a disease, but represents mechanically induced damage to joint tissues and the failed repair of that damage ([47, 48]). Furthermore, if the increased level of intraarticular stress that caused the joint to get into trouble in the first place is normalized, the OA joint can heal ([47, 48]). Even if a DMOAD did not reverse, but only arrested, structural damage that had accrued before the treatment was instituted, concurrent normalization of the excess intraarticular stress can facilitate reversal of the existing OA pathology. Under those circumstances, prescription of a DMOAD might even prove superfluous. The greater sensitivity of MRI than radiography for assessment of OA joints is not in dispute. I suggest, however, that our failure to find a DMOAD is due, most importantly, to the fact that we have not paid sufficient attention to the importance of abnormal intraarticular stress in initiating and driving the progression of structural damage in OA, which has led to selection of inappropriate candidates for DMOAD RCTs.

Evidence that knee OA is mechanically induced

Peak adductor moment

The peak adductor moment reflects the magnitude of the intrinsic compressive load on the medial compartment of the knee in stance ([49]). Varus–valgus alignment is a major determinant of peak adductor moment. Varus further increases the medial compartment load during gait; valgus acts similarly to increase stress in the lateral compartment ([50]). The peak adductor moment predicts radiographic progression in patients with medial compartment OA ([51]) and development of knee pain in asymptomatic elderly subjects ([52]). A greater degree of toe-out walking, which reduces the peak adductor moment, diminished the risk of radiographic progression of OA ([53]). Interventions that reduce the peak adductor moment improve both symptoms and structural abnormalities of OA, providing a proof-of-concept of their etiopathogenetic importance.

Repetitive impulsive loading

Physiologically, the viscoelasticity of articular cartilage and subchondral bone serves to maximize the contact area in a loaded joint, minimizing the stress (force per unit area) within the articular cartilage and transmitting load to the underlying bone, sparing the cartilage from damage ([47]). If, however, the ability of the articular cartilage to deform with loading is restricted so that it cannot conform completely to the load, the size of the contact area will be reduced and high stress will be generated within the articular cartilage. It is relevant, therefore, that the congruity of joints increases with age (as does the prevalence of OA), so that they become less flexible under load ([54]). The subchondral bone deforms less, or becomes stiffer, when load is applied rapidly than when loading is more gradual, limiting the chondroprotective effect of the shock-absorbing capacity of the bone. Joint damage caused by excessive loading may be related, therefore, not only to the magnitude of the load, but also to the rate of loading. Rapid repetitive impulsive loading does not permit sufficient time for the flow of interstitial fluid needed to absorb the energy that is transmitted, thereby protecting the articular cartilage matrix and cells ([47]). The data on the magnitude of the peak adductor moment in knee OA and the data on the rates of loading require reconciliation, but in any case, they strongly suggest that mechanical abnormalities play a principal role in the etiopathogenesis of structural damage and the symptoms experienced by patients with knee OA.

The structural and clinical improvement that may be seen after distraction of OA hips, knees, and ankles with external fixation ([19, 20, 55]), and after hip and knee osteotomy ([58, 59]), further support this view. In this country, however, osteotomy has, to a large extent, been superceded by arthroplasty as a surgical treatment for symptomatic OA. In the hands of most orthopedic surgeons it is a more reliable procedure and, for the patient, is associated with a much more rapid return to load-bearing activities than osteotomy. Parenthetically, with respect to the adequacy of radiography as an outcome measure for RCTs of putative DMOADs, plain radiography has clearly demonstrated structural improvement after both osteotomy and joint distraction ([19-22, 55]).

Several presumptive DMOADs studied in placebo-controlled RCTs have been aimed at blocking the effect(s) on articular cartilage of, e.g., a matrix-degrading protease or cytokine or toxic oxygen radical, or increasing the concentration of cartilage matrix macromolecules ([60]). In none of them, however, were attempts made to correct existing local biomechanical abnormalities. It is important to differentiate the pathogenesis of joint breakdown, due to, e.g., proteases, matrix-degrading cytokines, or toxic oxygen radicals, from the etiopathogenesis of OA, which, in most cases, is mechanically induced and mechanically driven.

Although no DMOAD has received approval for use in humans with OA, not all RCTs of putative DMOADs have yielded negative results. Examination of studies in which positive results were obtained can be informative, and examples are presented below.

In a placebo-controlled trial ([61]), doxycycline significantly slowed the rate of JSN in the index knee, but not the contralateral knee, of patients in whom unilateral radiographic OA was seen on standing anteroposterior radiography, whereas the mean rates of medial tibiofemoral JSN in the contralateral and index knees of the placebo-treated patients were comparable. Lateral and/or sunrise views, however, commonly indicated the presence of patellofemoral or posterior tibial OA in the contralateral knee ([62]), leading to speculation that the target(s) for doxycycline in knees with early structural damage may differ from that in knees with more severe OA.

Although the efficacy of glucosamine hydrochloride, alone or in combination with chondroitin sulfate, was not significantly different from that of placebo with regard to effects on the rate of JSN in the Glucosamine/Chondroitin Arthritis Intervention Trial (GAIT) ([63]), the efficacy of chondroitin sulfate as a DMOAD remains debated by some. Significant limitations in the experimental design and power of the GAIT study, however, severely limit the conclusions that may be drawn with respect to the reported results ([64]).

Licofelone, a dual inhibitor of cyclooxygenase and lipoxygenase, exhibited DMOAD activity, based on JSW measurements from Lyon schuss radiographs obtained at baseline and at 6, 12, and 24 months ([65]). Quantitative change in articular cartilage volume was assessed by MRI at the same intervals. The authors compared the results with those in a group of OA patients who received naproxen 500 mg twice daily. The mean loss of articular cartilage volume with licofelone was greater than that with naproxen at both 12 months and 24 months. The reduction in radiographic JSW in patients who received licofelone was less than that in the naproxen arm, but the difference between treatment groups in this respect was not significant. The relative insensitivity of Lyon schuss radiography in comparison with qMRI in assessing articular cartilage thickness in that study contrasts with recent findings reported by Le Graverand et al ([40]).

In a recent phase III RCT using radiography with a fixed flexion view, strontium ranelate, an osteoporosis treatment that dissociates bone remodeling processes, inhibits resorption of subchondral bone, and stimulates chondrogenesis in vitro, exhibited DMOAD activity ([66, 67]). In the strontium ranelate treatment group the rate of JSN was slower, and the proportion of patients with radiographic progression smaller, than in the placebo group. As noted above, however, the fixed flexion view does not assure a high degree of alignment of the medial tibial plateau with the central x-ray beam, which has been shown to significantly affect the appearance of JSW (and, hence, the rate of JSN in serial images) ([7, 8, 43]). Also, given the important contribution of subchondral bone to the structural changes and symptoms of OA ([47]) and the hypothetical worsening of OA by dissociation of bone formation and bone resorption, the results of the strontium ranelate RCT require confirmation.

In conclusion, the genes whose identification may be required for prevention and treatment of common, garden-variety OA may not be those that regulate metabolism of the chondrocyte. Rather, they may be genes that control congenital and developmental deformities of the joint (thereby reducing the habitually loaded area of the joint surface) or that underlie micro-incoordination of the patient (resulting in concentration of the peak dynamic load on the joint, with ensuing microdamage and remodeling of joint tissues that are detrimental to joint function). Possibly, had the treatments that failed to exhibit DMOAD activity been studied under conditions in which the effects of measures to alleviate the elevated levels of intraarticular stress were also evaluated, e.g., in a factorial design, some may have demonstrated efficacy or been shown to be unnecessary.

In support, Drs. Guermazi, Roemer, and Felson: Rebuttal

We appreciate the comments of Dr. Brandt, who has contributed enormously to the understanding of OA and its assessment. In the first part of his remarks, Dr. Brandt describes the currently known limitations of conventional radiography as a tool for imaging OA pathologic features in the context of clinical trials of structure-modifying agents. He also states that OA is not a cartilage disease but a process that involves the whole joint. We fully agree with both of these positions—that conventional radiography is limited and that OA is a disease process involving far more than just hyaline cartilage. These are reasons we are advocating use of an imaging modality that, unlike radiography, can allow investigators conducting OA clinical trials to visualize all joint tissues.

Dr. Brandt notes that radiography is the only imaging modality that is approved by the FDA, but this is not quite true. As we noted, the FDA recommends the use of radiography in its published draft guidance for industry regarding OA clinical trials ([29]), but no imaging modality has yet received official FDA approval.

One study ([30]) is cited to put forth the argument that radiography is more sensitive than MRI with regard to assessment of change in articular cartilage. However, there are other studies that show evidence to the contrary. Using data from the OAI, a recent study demonstrated that MRI-based measurement of cartilage thickness displays greater sensitivity to change in knee OA than does JSW measurement on radiography ([31]). Another study that Dr. Brandt quotes demonstrated that MRI-based outcome (cartilage volume) was more sensitive to change than radiography-based outcome (JSW) when the length of observation period and the number of participants were the same ([32]). Moreover, a meta-analysis of evidence from 42 published reports showed that MRI has adequate responsiveness using both quantitative and semiquantitative techniques ([28]).

Dr. Brandt mentions the challenges associated with knee positioning in radiographic assessment. Indeed, radiographic studies in which the knee flexion position is determined using fluoroscopic guidance are so challenging ([33]) that one wonders whether they are worth doing. We fully agree with his opinion and would like to point out that knee positioning is much less of a challenge when using MRI. Granted that costs of MRI are higher and imaging time required to complete an MRI scan is longer than for obtaining a radiograph. However, by using a tailored MRI pulse sequence protocol designed to best achieve specific study aims, imaging time may be markedly reduced, and this can lower costs associated with image acquisition. Moreover, new technological developments allow for faster image acquisition than is achieved with standard pulse sequences. For example, use of 3-dimensional turbo spin-echo sequences allows for acquisition of images within a shorter scan time (5 minutes) compared to standard 2-dimensional turbo spin-echo sequences (8 minutes) ([34]). Thus, we trust there are ways to make MRI-based OA clinical trials practically feasible.

Because radiographic OA may progress slowly ([13]), a long followup period with a large number of participants is necessary in order to demonstrate statistically significant outcomes in an OA clinical trial. Dr. Brandt cites one such example ([35]). The SEKOIA (Strontium Ranelate Efficacy in Knee Osteoarthritis Trial) used change in JSW as the primary end point and examined the efficacy of strontium ranelate in slowing the progression of JSN. The study included 1,371 participants, with a followup period of 3 years. By targeting MRI-detected bone marrow lesions and synovitis, which are clinically relevant because of their association with pain and can show short- to medium-term (6 weeks to 30 months) fluctuation ([22, 36, 37]), we believe a clinical trial can be conducted with fewer participants and a shorter followup period compared to a study using radiographic outcome as the primary end point. This decrease in the number of participants will reduce study costs. To evaluate bone marrow lesions and synovitis by imaging, MRI is essential since radiography does not allow their visualization. Inflammation may be a target for therapy but it remains unclear, as Dr. Brandt notes, whether any treatment can be successful without targeting or correcting the abnormal mechanics that are so central to disease pathogenesis and progression.

In conclusion, we largely agree with Dr. Brandt with regard to the understanding of OA being a primarily mechanically driven disease process and the currently known limitations of conventional radiography. However, we stand firm in our view that the failures of OA clinical trials to demonstrate efficacy of any putative DMOAD to date are attributable in part to the dependence on radiography for defining eligibility criteria and its use for primary outcome measures. It seems unwise not to use MRI to detect cartilage change and, in addition, to evaluate changes in MRI-detectable pathologic features (e.g., bone marrow lesions, synovitis) in OA clinical trials.

In opposition, Dr. Brandt: Rebuttal

Drs. Guermazi, Roemer, and Felson review the well-recognized limitations of plain radiography that may significantly affect outcomes in RCTs of putative DMOADs. In support of their argument they contend that by affecting the eligibility of subjects being recruited and the selection of radiographic end points, these limitations can result in the failure of DMOAD RCTs to demonstrate treatment efficacy.

Guermazi et al appropriately emphasize the high dependence on reproducible positioning of the joint in measurement of JSW, which is widely considered to be a surrogate for the thickness of articular cartilage. Lack of reproducibility in positioning of the knee can lead to significant error in JSW measurement. On serial radiographs, the rate of JSN and the sensitivity to change in JSW can vary markedly among protocols that aim to achieve standardized, reproducible positioning of the knee.

The K/L grade is widely used as an index of the radiologic severity of OA and is commonly used as an eligibility criterion in the recruitment of subjects for DMOAD trials. Some potential subjects may erroneously be excluded from RCTs, and other subjects may be erroneously included, because of artifactual changes in JSW due to lack of strict control of the intermargin distance, a measure of alignment of the central beam of the x-ray with the plane of the medial tibial plateau.

For some anatomic structures of the knee, MRI is much more sensitive than radiography in detecting structural abnormalities. That is not a debatable issue. However, as I noted above in stating my position in this debate, in knees with K/L grade 3 OA, Lyon schuss radiography exhibited greater sensitivity to change in articular cartilage thickness at 12 months and 24 months than did qMRI. The reason for this difference between the two techniques is unclear, but may be related to softening of articular cartilage in OA joints and the fact that the knee radiographs were obtained with the patient in a standing, weight-bearing position, i.e., under conditions that afforded greater compression of the OA articular cartilage under load, whereas the MRIs were obtained with the patient in a horizontal, i.e., non–weight-bearing, position in the magnet.

As I have noted above, it is unclear how the results and the costs of previous DMOAD RCTs would have differed significantly if MRI, rather than radiography, had been used as the primary outcome measure. Guermazi et al provide examples, but not comparative quantitative data from RCTs, to support their contention that this would have appreciably affected the outcomes of DMOAD trials.

Furthermore, in response to the fundamental question that is implicit in this debate, i.e., “Why have clinical trials of putative DMOADs not demonstrated DMOAD activity?” Guermazi and colleagues note the limitations of radiography for DMOAD RCTs, but fail to consider an alternative—and, in my view, more likely—possibility, i.e., that although the agents tested for DMOAD activity target pathogenetic mechanisms involved in breakdown of articular cartilage in OA, none of the RCTs in which they were tested took into account the fundamental etiopathogenetic importance of the abnormal intraarticular stress (i.e., force per unit area) that is the proximate cause of the joint damage in OA and the failure of that damage to repair. If, however, the increased intraarticular stress that got the OA joint into trouble in the first place is normalized, the damage may not (merely) be arrested; under such conditions, the damage can be reversed and the OA joint can heal. The structural and clinical improvements documented after distraction of OA joints and after osteotomy support this view, serving as a proof-of-concept.

I agree with Guermazi et al that state-of-the art imaging methods are needed to optimize the chances of recognizing a DMOAD effect in RCTs. I disagree, however, with their narrow definition of DMOADS as drugs that “delay” the structural progression of OA. In my view, as long as we continue to ignore the mechanical etiopathogenesis of OA, improved imaging methods alone are not likely to demonstrate efficacy of putative DMOADs, if it exists. Had drugs that previously failed to exhibit DMOAD activity been studied under conditions under which intraarticular stress in the OA joint was alleviated, some may have shown efficacy or, indeed, proved to be superfluous.

Finally, for both the patient and the physician, the main problem with OA does not lie in its inherent imaging problems but in the fact that it is painful. “X-rays don't weep” ([68]). Patients weep. Manufacturers and clinicians often comment that it is unlikely that a DMOAD will ever be approved by regulatory authorities if it does not offer symptom relief as well as structural modification. The salutary effects on joint pain that can be seen after osteotomy and after distraction of the OA joint suggest that normalization of the structural abnormalities can be accompanied by symptomatic relief.


All authors drafted their individual section of the article, revised it critically for important intellectual content, and approved the final version to be published.