Chapter  157:  Treatment of Primary and Secondary Osteoarthritis of the Knee

Viscosupplementation

Effectiveness and Harms in Primary OA of the Knee. Results from 42 trials (N=5,843), all but one synthesized in various combinations in six meta-analyses, generally show positive effects of viscosupplementation on pain and function scores compared to placebo. However, the evidence on viscosupplementation is accompanied by considerable uncertainty due to variable trial quality, potential publication bias, and unclear clinical significance of the changes reported.

The pooled effects from poor quality trials were as much as twice those obtained from higher quality ones. Pooled results from small trials (≤100 patients) showed effects up to twice those of larger trials, a finding consistent with selective publication of underpowered positive trials. Among trials of viscosupplementation, those that have not been published in full text comprise approximately 25 percent of the total patient population.

Most RCTs reported results as mean changes in pain and function. Interpreting the clinical significance of pooled mean effects from the meta-analyses is difficult; mean changes do not quantify proportions responding. Numbers needed to treat cannot be calculated from mean changes. It would be more informative to report response rate, i.e., comparison of the proportion of patients achieving a clinically important improvement.

Trials of hylan G-F 20, the highest molecular weight cross-linked product, generally reported larger effects than other trials.

Minor adverse events accompanying intra-articular injections are common, but the relative risk accompanying hyaluronan injections over placebo appears to be small. Pseudoseptic reactions associated with hyaluronans appear relatively uncommon, but can be severe.

Differences in Outcomes Among Subpopulations. Four RCTs were identified examining any of the specified subgroups. None examined race/ethnicity, disease duration, or prior treatment. In one trial, randomization was stratified by disease severity; all other subgroup results were obtained in post-hoc analyses. There was no evidence for differential effects according to subgroups defined by age, sex, primary/disease, body mass index/weight, or disease severity. One positive post-hoc subgroup analysis found greater efficacy among older individuals with more severe disease, but was not confirmed in a subsequent trial.

Glucosamine, Chondroitin, Alone or in Combination

Effectiveness and Harms in Primary OA of the Knee. The best evidence comes from the Glucosamine/Chondroitin Arthritis Intervention Trial (GAIT; Clegg, Reda, Harris, et al., 2006), a large (n=1,583), good quality, NIH-funded, multicenter RCT. GAIT compared glucosamine hydrochloride, chondroitin sulfate, or the combination of these agents, with placebo or celecoxib in patients with primary osteoarthritis of the knee. After 24 weeks of treatment, intention-to-treat analysis showed no significant difference in symptomatic relief between glucosamine hydrochloride, chondroitin sulfate, or glucosamine hydrochloride plus chondroitin sulfate compared to placebo. Substantiating this result was that celecoxib, the active control, was effective.

Six study-level meta-analyses (MAs) assessed glucosamine or chondroitin in OA of the knee. All but one of the MAs reported statistically significant differences between treatment and placebo. However, these MAs had limitations in the quality of the primary studies that were pooled. Limitations of the primary literature included small study size, inclusion of studies that assessed joints other than knee, and failure to report intent to treat analysis. In general, the MAs did not perform adequate quality appraisal of the primary studies.

Glucosamine sulfate has been reported to be more effective than glucosamine hydrochloride, however, the evidence is not sufficient to draw conclusions. A subgroup analysis in the largest MA (Towheed, Maxwell, Anastassiades, et al., 2006) significantly favored glucosamine sulfate. The results of GUIDE (Herrero-Beaumont, Roman, Trabado, et al., 2007), a European placebo-controlled RCT (n=318), sponsored by Rotta, a glucosamine sulfate manufacturer, report favorable results for glucosamine sulfate. While the overall results of GAIT show no benefit, in the subgroup of knee OA patients with moderate-to-severe pain at baseline, the combination of glucosamine hydrochloride and chondroitin sulfate significantly improved pain. Together, this evidence suggests an independent trial of glucosamine sulfate would be useful to definitively establish whether there is benefit.

In general, adverse events with glucosamine or chondroitin treatment were no greater than placebo. There has been some concern from in vitro and preclinical studies that glucosamine supplementation could have a deleterious effect on glucose metabolism and glycemic control. However, available clinical studies are short-term, or if longer (e.g., 3 years), excluded patients with metabolic disorders.

Differences in Outcomes Among Subpopulations. GAIT found that glucosamine plus chondroitin produced a statistically and clinically significant improvement of pain in patients with moderate-to-severe OA of the knee. Although the effect of celecoxib treatment in a similar group of patients was not statistically significant, the magnitude and direction of the response were consistent with clinical benefit. The nonsignificant statistical result in the celecoxib arm may be a function of insufficient power due to the small number of patients. Although this subgroup analysis was not explicitly prespecified in the GAIT protocol, the stratified randomization by disease severity yields statistically valid comparisons. A trial of glucosamine sulfate would be useful to definitively establish whether there is benefit.

Arthroscopic Lavage and Debridement

Effectiveness and Harms in Patients With Primary OA. The best available evidence, a single placebo-controlled RCT, found arthroscopic lavage with or without debridement was not superior to placebo. The evidence base does not definitively show that arthroscopy is no more effective than placebo. However, additional high-quality RCTs would be necessary to refute the existing trial, which suggests equivalence between placebo and arthroscopy.

No other study besides Moseley, O'Malley, Petersen, et al. (2002) addressed the potential contribution of placebo effects to apparent improvement in outcome after arthroscopy. The primary limitations of the Moseley, O'Malley, Petersen, et al. (2002) trial are lack of details describing the patient sample, the use of a single surgeon, and enrollment of patients at a single Veterans Affairs Medical Center. These concerns call into question the generalizability of this trial's findings. Since OA of the knee affects a large population, uncertainty about arthroscopy's effectiveness should be resolved with further well-conducted and well-reported RCTs.

Major methodologic shortcomings in non-placebo RCTs, an administrative database analysis and case series preclude resolution of uncertainties raised by the trial of Moseley, O'Malley, Petersen, et al. (2002).

Evidence on the harms after arthroscopic lavage and debridement comes primarily from an administrative database analysis and case series reports. Potential harms include infection, prolonged drainage from arthroscopic portals, effusion, hemarthrosis and deep vein thrombosis. To determine whether the risk of such harms is acceptable, it is important to establish whether the effectiveness of arthroscopic lavage and debridement surpasses placebo.

Differences in Outcome Among Subpopulations. Subgroup analyses for mechanical symptoms, alignment and OA stage were performed in the Moseley placebo-controlled RCT. No differences in results were observed within subgroups. Subgroup analyses were also performed in a quasi-experimental study, an administrative database and several case series. In these studies, different outcomes were observed according to age, presence of mechanical symptoms, and severity of OA. However, since these studies did not include placebo controls, it cannot be concluded that arthroscopy has greater effectiveness in specific patient subgroups.

All Interventions

Effectiveness and Harms in Secondary OA of the Knee. We identified no studies that enrolled patients with only secondary OA of the knee, or that reported separately on secondary OA of the knee.

Comparison of Interventions. We did not find any direct comparative studies in which glucosamine, chondroitin, or glucosamine plus chondroitin were compared with arthroscopy or viscosupplementation to treat OA of the knee. A single, small, underpowered, poor quality trial found no difference in outcome measures comparing intra-articular hyaluronan to arthroscopy and debridement over a 1-year followup.

Discussion and Future Research

OA of the knee is a common condition and the three interventions reviewed in this report are widely used in the treatment of OA of the knee. Yet the best available evidence reports that glucosamine/chondroitin and arthroscopic surgery are no more effective than placebo. The Glucosamine/Chondroitin Arthritis Intervention Trial (GAIT) (n=1,583) found that neither glucosamine hydrochloride, chondroitin sulfate nor the combination was superior to placebo and that all were inferior to celecoxib. The double blind randomized controlled trial by Moseley, O'Malley, Petersen, et al. (2002, n=180) found that arthroscopic lavage with or without debridement was not superior to sham arthroscopy. Results from 42 RCTs, all but one of which were synthesized in various combinations in six meta-analyses, generally show positive effects of viscosupplementation on pain and function scores compared to placebo. However, the evidence on viscosupplementation is accompanied by considerable uncertainty due to variable trial quality, potential publication bias, and unclear clinical significance of the changes reported.

For viscosupplementation, higher-quality trials are in the minority and show smaller effects; there are numerous patients lost to follow-up, and a substantial portion of studies (25 percent of total patients) have not been published as full articles. The clinical significance of reported changes in pain and function scores is uncertain, as almost all studies compare only mean difference between arms. Although the overall pooled estimate suggests that hylan G-F 20 may have a larger effect than other hyaluronans, whether this represents a meaningful clinical effect or limitations in the quality and completeness of study reporting is unknown. A rigorous RCT that showed strong evidence of improvement in pain and function would be necessary to conclude that viscosupplementation is beneficial.

While the overall results of GAIT show no benefit, a subgroup analysis found that the combination of glucosamine hydrochloride and chondroitin sulfate significantly improved pain in patients with moderate-to-severe OA of the knee. Although this subgroup analysis was not explicitly prespecified in the GAIT protocol, the stratified randomization by disease severity yields statistically valid comparisons. The nonsignificant statistical result in the celecoxib arm in the same patient subgroup may be a function of insufficient power. Given the small number of patients in the moderate-to-severe subgroup, and the large number of such patients in the general population, a further trial can be justified. However, these subgroup results do not override the overall results of GAIT, which must stand unless confirmed in a rigorous RCT.

The existing evidence does not definitively show that arthroscopic lavage with or without debridement is no more effective than placebo. However, additional placebo-controlled RCTs showing clinically significant advantage for arthroscopy would be necessary to refute the Moseley results, which show equivalence between placebo and arthroscopy. The recently published Spine Patient Outcomes Research Trial (SPORT) offers an alternative study design that could be informative, a rigorous RCT comparing surgery to conservative management, rather than sham (Weinstein, Tosteson, Lurie, et al., 2006).

Overall, our recommendations for future research reach beyond the specific treatments addressed in this report, and are intended broadly to improve the quality of research and reporting on interventions for osteoarthritis of the knee. However, our population is aging, there is increasing prevalence of obesity, and increasing burden of knee osteoarthritis, together with inconsistent evidence regarding disease treatments. Given the public health impact, research on new approaches to prevention and treatment should be given high priority.

Minimally Clinically Important Improvement in Pain and Function Should be the Measure of Success for All Trials. Clinically meaningful results require outcome measures establishing that patients experience improvement that is important to them—meaningful clinically important improvement. The range of magnitude of improvement clinically important to patients has been estimated for VAS pain and WOMAC measures, while to a lesser degree for the Lequesne Index (see Methods). Common measures and intervals for measurement will produce a more robust body of cumulative evidence and improve the ability to compare and pool results among trials.

Unpublished Studies Should be Made Available as Full Text Publications. Among RCTs of viscosupplementation, those that have not been published in full text comprise approximately 25 percent of the total patient population. Several meta-analyses of glucosamine report that trials of the Rotta product, glucosamine sulfate, show outcomes superior to trials of glucosamine hydrochloride; yet key trials have not been published as full-text studies. Existing studies should be published in full. And all trials should be registered at inception at ClinicalTrials.gov along with anticipated date for full release of results.

The Pitfalls of Meta-Analysis Should be More Widely Recognized and Acknowledged. Our evidence report draws heavily on six study-level meta-analyses of glucosamine/chondroitin and five of viscosupplementation. While we used a validated instrument to appraise the quality of the systematic reviews, the instrument does not address the question of when meta-analysis is appropriate to a systematic review. Meta-analysis is a technique with underlying assumptions that may or may not hold when a particular collection of results are pooled. Furthermore, meta-analyses may fail to convey the real uncertainty and potential bias accompanying pooled estimates.

Uncertainty in the magnitude of effects pooled is influenced by factors intrinsic to the underlying trials. Among these are variable patient characteristics, trial characteristics, and the indication that a few trial results were outliers and influential on pooled estimates. The meta-analyses frequently reported high inter-trial heterogeneity. Random effects models were used in the face of high heterogeneity, but a consequence is to increase the influence of smaller trials on the pooled results. The meta-analyses did not address a threshold question, one that has not been clearly resolved by practitioners of meta-analysis: when is heterogeneity too high to justify pooling trial results. A related concern is the practice of reporting on multiple outcome measures and time intervals, which may be represented by a small portion of studies, thus potentially introducing bias.

Conclusions

Osteoarthritis of the knee is a common condition. The three interventions reviewed in this report are widely used in the treatment of OA of the knee, yet the best available evidence does not clearly demonstrate clinical benefit. Uncertainty over clinical benefit can be resolved only by rigorous, multicenter RCTs. In addition, given the public health impact of OA of the knee, research on new approaches to prevention and treatment should be given high priority.

Pathophysiology

The term “osteoarthritis” refers to a heterogeneous group of joint disorders, usually signaled by symptoms of pain and stiffness. It involves both destructive and reparative metabolic processes, with a variety of biochemical triggers in addition to mechanical injury of the joint (Mandelbaum and Waddell, 2005). It is thought that inflammation does not play a primary role in osteoarthritis, although it may be present. When inflammation occurs, it is generally mild (Hochberg, Altman, Brandt, et al., 1995b). The pathogenesis of OA is not fully understood, although multiple contributing factors are recognized including genetic, environmental, metabolic, and biomechanical factors (Kraus, 1997).

Although OA eventually involves all joint structures, it begins with damage and progressive degradation of articular hyaline cartilage structure and function (chondropenia), typically in a nonuniform, focal manner (Felson, 2006). As chondropenia progresses in localized areas, stress increases across the entire joint, further damaging and eroding cartilage. In areas with full-thickness cartilage loss, abnormal remodeling and attrition of subarticular bone commences, typically accompanied by growth of osteophytes. Synovitis, ligamentous laxity, and periarticular muscle weakness may also occur, eventually leading to joint tilting and malalignment. Malalignment is a risk factor for joint failure, hastening structural deterioration of the joint by increasing local loading forces.

The symptoms of OA result from abnormal stresses on the weight-bearing joints or normal stresses on weakened joints, becoming progressively worse and more frequent with age. The typical joints involved with osteoarthritis include the large, weight-bearing joints such as the hip and knee, as well as selected smaller joints in the hands, feet, and spine.

Classification

Osteoarthritis may be broadly categorized as primary (idiopathic) or secondary. According to the American Academy of Orthopaedic Surgeons, primary OA of the knee can be defined as a process in which articular degeneration occurs in the absence of an obvious underlying abnormality (American Academy of Orthopaedic Surgeons, 2004). Secondary OA of the knee is often the result of injury (trauma) or repetitive motion such as found in certain occupations. It can also result from congenital conditions and underlying diseases, including include systemic metabolic diseases, endocrine diseases, bone dysplasias, and calcium crystal deposition diseases. Secondary OA is more likely to manifest itself at an earlier age than primary OA, and may be an initial clue to the presence of a potentially dangerous and treatable systemic disease. While there is rationale for identifying two separate categories of OA, making a distinction between them does not alter clinical practice and therapeutic choices.

Diagnosis

The diagnosis of osteoarthritis is established using a combination of clinical information derived from history, physical examination, radiologic, and laboratory evaluation. An algorithm of diagnostic criteria for osteoarthritis of the knee has been proposed by the American College of Rheumatology (ACR) (Altman, Asch, Bloch, et al., 1986). A diagnosis of OA of the knee is defined as presenting with pain, and meeting at least five of the following criteria:

• Patient older than 50 years of age

• Less than 30 minutes of morning stiffness

• Crepitus (noisy, grating sound) on active motion

• Bony tenderness

• Bony enlargement

• No palpable warmth of synovium

• Erythrocyte sedimentation rate (ESR) <40 mm/hr

• Rheumatoid factor <1:40

• Noninflammatory synovial fluid.

The presence of clinical symptoms of OA does not always correlate well with the degree of abnormality seen on radiographs. It has been noted that approximately 40 percent of patients who have severe X-ray findings report no symptoms, and conversely, patients with clinical symptoms may show no significant radiological changes (Balint and Szebenyi, 1996; Davis, Ettinger, Neuhaus, et al., 1992; Claessens, Schouten, van den Ouweland, et al., 1990).

Treatment

Treatment for OA of the knee aims to alleviate pain and improve function in order to mitigate reduction in activity (American College of Rheumatology, 2000; Felson, 2006). However, most treatments do not modify the natural history or progression of OA, and thus are not considered curative. Nonsurgical modalities include education, exercise, weight loss, and various supportive devices; acetaminophen or nonsteroidal anti-inflammatory drugs (NSAIDs) such as ibuprofen; nutritional supplements (glucosamine and chondroitin); and, intra-articular viscosupplements.

Guidelines for the medical management of osteoarthritis emphasize the role of both nonpharmacologic and pharmacologic therapies (American College of Rheumatology, 2000; Jordan, Arden, Doherty, et al., 2003). Initial management involves nonpharmacologic therapies, including education, exercise, various appliances and braces, and weight reduction. Acetaminophen is recommended as first-line pharmacologic therapy. If pain relief is inadequate with acetaminophen, analgesic-dose NSAIDs may be used (e.g., ibuprofen, naproxen). If symptom response to a lower NSAID dosage is inadequate, higher, anti-inflammatory, doses may be used. Intra-articular corticosteroid injection may be considered when relief from NSAIDs is insufficient or the patient is at risk from gastrointestinal adverse effects. Injection of corticosteroids is frequently limited to three to four times per year per joint because of concern about the possibility of progressive cartilage damage through repeated injection in the weight-bearing joints (Neustadt, 1992).

If symptom relief is inadequate with conservative measures, invasive treatments may be considered. Operative treatments for symptomatic OA of the knee include arthroscopic lavage and cartilage debridement, osteotomy, and, ultimately, total joint arthroplasty (Day, 2005). Surgical procedures intended to repair or restore articular cartilage in the knee, including abrasion arthroplasty, microfracture techniques, autologous chondrocyte implantation, and others, are appropriate only for younger patients with focal cartilage defects secondary to injury (Clarke and Scott, 2003).

Interventions Addressed in This Report

Table 1

U.S. FDA-approved hyaluronan products and product information (more...)

Table 1

U.S. FDA-approved hyaluronan products and product information statements

Product	Regarding Treatment Course	Regarding Minimum # of Injections	Regarding Other Joints	Regarding Repeat Treatments
Hyalgan® (sodium hyaluronate); Fidia Pharmaceutical	“A treatment cycle consists of five injections given at weekly intervals. Some patients may experience benefit with three injections given at weekly intervals.”	“The effectiveness of a single treatment cycle of less than 3 injections has not been established.”	“The safety and effectiveness of the use of Hyalgan® in joints other than the knee have not been established.”	“Adverse experience data from the literature contain no evidence of increased risk relating to retreatment with Hyalgan®. The frequency and severity of adverse events occurring during repeat treatment cycles did not increase over that reported for a single treatment cycle....”
Original PMA date: 5/28/97				Hyalgan® is the only hyaluronan with demonstrated safety in a 30-month, repeat use, open-label trial in which 75 patients received a cycle of 5 weekly injections of Hyalgan® every 6 months.
MW: 0.5–0.73 million Da
Synvisc® (hylan G-F 20); Genzyme Corporation	“Synvisc® is administered by intraarticular injection once a week (one week apart) for a total of three injections.”	“The effectiveness of a single treatment cycle of less than three injections of Synvisc® has not been established.”	“The safety and effectiveness of Synvisc® in locations other than the knee and for conditions other than osteoarthritis have not been established.”	“The reactions seemed to occur more often when Synvisc® was injected into the knee as a repeat set of injections than when Synvisc® was injected as a first set of injections.”
Original PMA date: 8/08/97
MW: 6 million Da (hylan A)
Supartz® (sodium hyaluronate); Seikagaku Corporation	“Supartz® is administered by intraarticular injection once a week (one week apart) for a total of 5 injections.”	“The effectiveness of a single treatment cycle of less than 5 injections has not been established.”	“The safety and effectiveness of the use of Supartz® in joints other than the knee have not been established.”	“The safety and effectiveness of repeat treatment cycles of Supartz® have not been established.”
Original PMA date: 1/24/01
MW: 0.62–1.17 million Da
Orthovisc® (sodium hyaluronate), Anika Therapeutics, Inc.	“Orthovisc® is injected into the knee joint in a series of intraarticular injections one week apart for a total of three or four injections.”	The effectiveness of a single treatment cycle of less than 3 injections has not been established. Pain relief may not be seen until after the third injection.	“The safety and effectiveness of the use of Orthovisc® in joints other than the knee have not been established.”	“The effectiveness has not been established for more than one course of treatment.”
Original PMA date: 2/04/04
MW: 1–2.9 million Da
Euflexxa® (sodium hyaluronate), Ferring Pharmaceuticals	“A dose of 2 ml is injected intraarticularly into the affected knee at weekly intervals for three weeks, for a total of three injections.”	N/R	“Safety and effectiveness of injection in conjunction with other intraarticular injectables, or into joints other than the knee has not been studied.”	“The safety and effectiveness of repeated treatment cycles of EUFLEXXA™ have not been established.”
Original PMA date: (approved under the name Nuflexxa)
MW: 2.4–3.6 million Da

Da: Daltons; MW: molecular weight; PMA: premarket approval

Glucosamine and Chondroitin. Glucosamine is an aminomonosaccharide which is the principal component of O-linked and N-linked glycosaminoglycans, which comprise the matrix of all connective tissues, including cartilage (Biggee and McAlindon, 2004; Matheson and Perry, 2003; Hauselmann, 2001; Deal and Moskowitz, 1999). This compound historically has been derived by extraction of chitin, a component of crustacean shells, although is also is produced through fermentation of a vegetarian source. Chondroitin sulfate is a glycosaminoglycan with a polymerized disaccharide base linked to a sulfate moiety, and is a component of proteoglycans of articular cartilage. It is usually derived from bovine trachea, although other sources such as ovine or porcine trachea and shark cartilage are used. The mechanisms of action of these compounds are unknown, but it is speculated they may promote maintenance and repair of cartilage.

In the United States, glucosamine hydrochloride or sulfate and chondroitin sulfate are considered dietary supplements available in over-the-counter (OTC) products, which may vary substantially in content and purity from what is stated on the label (McAlindon, 2003). In European Union countries, glucosamine sulfate and chondroitin sulfate are regulated as prescription drugs. A number of clinical trials with positive outcomes either used glucosamine sulfate manufactured by an Italian firm, Rotta Research Laboratorium, or were financially supported by Rotta. It has been hypothesized that Rotta glucosamine sulfate has greater efficacy than the hydrochloride salt, and that the formulation is a key factor in trial outcome (Altman, Abramson, Bruyere, et al., 2006; Hochberg, 2006; McAlindon, 2003). Oral administration of glucosamine sulfate can increase serum and synovial fluid sulfate levels, whereas sodium sulfate does not. Absorbed sulfate is then used in the synthesis of proteoglycans and metabolic intermediates like coenzyme A and glutathione that are important for chondrocyte metabolism.

Arthroscopy. The term “arthroscopy” is often used collectively in reference to individual minimally invasive surgical procedures, joint lavage and articular debridement, which are performed using fine needles and an arthroscope (Gidwani and Fairbank, 2004; Gunther, 2001). Arthroscopic lavage is a palliative measure in which intra-articular fluid is aspirated and the joint is washed out, removing inflammatory mediators, debris, or small loose bodies from the osteoarthritic knee. Articular debridement involves removal of cartilage or meniscal fragments, but also can include cartilage abrasion, excision of osteophytes and synovectomy. Debridement is intended to improve symptoms and joint function in patients with mechanical symptoms such as locking or catching of the knee. Because lavage and debridement are often performed at the same time, it is difficult to attribute the success or failure of arthroscopy to a specific procedure.

Types of Studies

We sought systematic reviews, meta-analyses, RCTs, including abstracts of unpublished placebo-controlled RCTs, examining the clinical effectiveness of one or more of the interventions of interest among patients with primary or secondary OA of the knee; and reporting at least one outcome of interest.

RCTs had to be published either as articles in any language or English-language abstracts (if the study was only presented as an abstract). No minimum number of patients per study arm was required for RCTs. Because there were few RCTs available to address arthroscopy and Key Question 4 (comparative outcomes), we sought additional study designs. For arthroscopy, we also sought English-language articles of nonrandomized comparative trials (i.e., quasi-experimental studies), administrative database analyses, and case series with samples of 50 or more. For Key Question 4, we sought randomized and nonrandomized comparative studies.

Studies were excluded if no outcome of interest to this review was reported. Studies were also excluded if the patient population of interest was fewer than 80 percent of included patients, or, alternatively, results for the patient population of interest were not separately reported. When multiple reports were available for the same study, it was counted as a single trial and outcome data from the report with the longest followup were used.

Types of Participants

The populations of interest are patients with primary or secondary OA of the knee, as defined by the American Academy of Orthopaedic Surgeons (American Academy of Orthopaedic Surgeons, 2004):

• Primary osteoarthritis of the knee is a process in which articular degeneration occurs in the absence of any obvious underlying abnormality (unknown cause); and

• Secondary OA is often the result of injury (trauma) or repetitive motion in certain occupations, but it can also result from congenital conditions and systemic metabolic diseases, endocrine diseases, bone dysplasias, and calcium crystal deposition diseases.

Subpopulations of interest include: age, race or ethnicity, sex, disease severity and duration, weight (body mass index), and prior treatments

Types of Interventions

Types of Outcomes

Primary Outcomes. The primary outcomes of interest are:

• Pain severity or intensity

• Self-reported physical function

• Patient global assessment

• Quality of life.

Secondary Outcomes. Secondary outcomes of interest include:

• Need for or time to total knee replacement or other surgeries.

• Concomitant analgesic use.

Harms or Adverse Effects. Any adverse events reported, including:

• Hyaluron Preparations. Local: injection site redness, edema, pain, joint swelling, joint stiffness, worsened osteoarthritis, infection, pseudoseptic reactions. Systemic: severe acute inflammatory reaction or pseudosepsis, anaphylaxis, arthralgias, rash, urticaria, back pain, headache.

• Glucosamine and Chondroitin. Alterations in blood glucose, hypersensitivity reactions, and local gastrointestinal toxicities.

• Arthroscopy. Infection, prolonged drainage from arthroscopic portals, effusion, hemarthrosis and deep vein thrombosis.

Pain and Function Measurement Issues

Instruments. Pain and function should be measured by instruments with established validity and reliability. Although results are frequently reported as mean change in the intervention compared to control arms, this is not the preferred method of measuring outcomes. More informative, is a comparison of response, that is the proportion of patients achieving an improvement that is established representing a minimum clinically important improvement. (Tubach, Wells, Ravaud, et al., 2005).

Among established instruments, pain severity may be assessed by a visual analog scale (VAS) or a numeric rating scale (NRS) or from a subscale included in a knee-specific validated OA instrument. The horizontal 100-mm VAS has a left-hand or 0-mm endpoint labeled “no pain” and a right-hand or 100-mm endpoint usually labeled with a statement such as “extreme pain” or “pain as bad as it could possibly be.” While the amount of improvement required may not be definitively established (Tubach, Ravaud, Baron et al. 2005; Pham, van der Heijde, Altman, et al. 2004), the best available estimates for OA of the knee are between 20 and 40 percent improvements have been used in hyaluronan and glucosamine/chondroitin trials (Nuestadt et al. 2005, Altman et al. 2004, Clegg et al). A clinically significant change in VAS score depends on the baseline pain (Campbell and Patterson, 1998). For example, in knee OA an absolute 20 mm or 40 percent relative reduction in VAS pain score could be considered a minimal clinically important improvement (MCII) (Tubach, Wells, Ravaud, et al., 2005) and define clinically meaningful response. Accordingly, a decrease of 10–12 mm may be clinically significant from a baseline of 25 mm, while a reduction of 20–31 mm may be necessary to achieve a clinically significant reduction for patients with high baseline pain (e.g., VAS 75–100 mm).

Among 2 widely used OA instruments, the Western Ontario and McMaster University Osteoarthritis Index (WOMAC, McConnell, Kolopack, and Davis, 2001; Bellamy, Buchanan, Goldsmith, et al., 1988) evaluates 3 dimensions, pain, stiffness, and physical function with 5, 2, and 17 questions, respectively. WOMAC assesses pain using either the sum of scores from 5 items or the VAS. WOMAC outcomes can be based on the total, or a subset score. A 20- to 40-percent reduction in the WOMAC pain subscore is a positive response criterion for pain used in knee OA studies and represents achieving a MCII (Tubach, Wells, Ravaud, et al., 2005).

Another commonly used OA instrument is the Lequesne Index, a validated numerical scale in which points are assessed for various levels of pain, distance walking, and ability to perform activities of daily living (Lequesne, Mery, Samson, et al., 1987). It sums scores from 5 adjectival items, producing scores ranging from 1 to 24 points. The severity of handicap related to the knee can be categorized by point score: mild (1–4 points); moderate (5–7 points); severe (8–10 points); very severe (11–13 points); and extremely severe (>14 points) (Bellamy, 1993). What constitutes a MCII is likely approximately 20 percent (Bellamy, 1993).

Physical function may be appraised through reported difficulty performing specific daily activities affected by knee OA (Bellamy, Buchanan, Goldsmith, et al., 1988; Lequesne, Mery, Samson, et al., 1987). Patient global assessment (generally defined as the “patient's assessment of overall disease activity or improvement”) can be assessed by VAS, NRS, or other specific instruments (Pham, van der Heijde, Altman, et al. 2004). The MCII for patient global assessment on a 100 mm VAS has been suggested to be 18 mm, or a relative improvement of 40 percent.

Both generic measures and disease-specific quality of life (QOL) measures may be relevant (Salaffi, Carotti, and Grassi, 2005) assessing disease impact. The SF-36 and Arthritis Impact Measurement Scales (Meenan, 1986) are acceptable scales to assess the impact of osteoarthritis on QOL.

Pooled Outcome Measures. Meta-analyses may pool outcome measures using the metric of the original scale, or a metric related to it.

The “weighted mean difference” (WMD) combines (pools) differences between treatment and control from multiple trials on the scale of the original instrument. It can be reported as either a difference between treatment and control at some followup time or a difference in change scores. While intuitive to interpret as a difference or difference in change for some outcome measure, the WMD has doe not define proportions achieving a MCII or response (Senn 1997, page 226; Tubach, Ravaud, Giraudeau 2005).

“Relative risks” (or the approximately equal odds ratio) can be pooled for dichotomous outcome measures (e.g., patient global assessment and adverse events). It is a ratio comparing the outcome probability among treated compared placebo groups. The relative risk clearly conveys increased risk, but does not directly reflect clinical benefit in terms of response unless a comparison of meaningful clinical response rates.

“Sums of differences” in outcome measures between treatment and placebo groups (e.g., pain and function) over the course of a study can also be pooled. The measure is expressed as a percentage reflecting how much greater relief is provided by treatment compared to placebo. Although commonly used in pain research, the measure does not have direct clinical meaning with respect to response.

“Standardized effect sizes” expressed as differences or differences in change, standardized by their variability (divided by the standard deviation) can also be pooled. Standardized effect sizes are typically used when scales pooled have different metrics (e.g., a 0- to 100-mm VAS and a 25-point WOMAC scale). The clinical meaning of standardized effect sizes when different scales are pooled and variability differs across studies is difficult to intuit. While small, medium, and large referents corresponding to 0.3, 0.5 and 0.8, respectively, were suggested by Cohen (1988), they pertain to sample size calculations not clinical meaning, and were qualified substantially.^* Others have pointed out problematic aspects of standardized effect sizes including: incomparability across studies (Rothman and Greenland, 1998) and that studies with identical results may appear to differ (Greenland, Schlesselman, Criqui, 1986). Most importantly, one cannot infer individual response Senn (1997).^†

Search Strategy

Electronic Databases. The following databases were searched for citations. The full search strategy is displayed in Appendix A ^*. The search was not limited to English-language references; however, foreign-language references without abstracts were disregarded.

• MEDLINE^® (through March 29, 2007)

• EMBASE (through March 16, 2006)

• Cochrane Controlled Trials Register (through November 27, 2006).

EMBASE was updated with abbreviated searches through November 27, 2006.

Additional Sources of Evidence. The Technical Expert Panel and individuals and organizations providing peer review were asked to inform the project team of any studies relevant to the key questions that were not included in the draft list of selected studies.

We examined the bibliographies of all retrieved articles for citations to any RCT that was missed in the database searches. In addition, we sought RCTs published in conference proceedings and abstracts from the American Association of Orthopaedic Surgeons (AAOS), American College of Rheumatology (ACR) and the Osteoarthritis Research Society International (OARSI) over the past 2 years. We also consulted product inserts of U.S.-marketed viscosupplement products.

Search Screen

Figure 2. QUOROM flow diagram

   Figure 2. QUOROM flow diagram

Figure 2

Data Elements

The data elements below were abstracted, or recorded as not reported, from intervention studies. Data elements to be abstracted were defined in consultation with the Technical Expert Panel.

Data elements from intervention studies (RCTs and quasi-experimental studies) include:

• Critical features of the study design (for example, patient inclusion/exclusion criteria, number of participants, allocation method (including concealment), use of blinding)

• Patient characteristics (age, gender, race/ethnicity, body weight, primary or secondary disease. disease duration)

• Measures of disease severity

• Treatment protocols (for example, dose, frequency, duration, extent of arthroscopic surgery, other prior and concurrent treatments)

• Patient monitoring procedures (for example, followup duration and frequency, outcome assessment methods) and

• The specified key outcomes and data analysis methods

• Results

• Funding source.

Data elements from systematic reviews and meta-analyses include:

• Use of a protocol

• The study question (patients, interventions/comparisons, outcomes)

• Literature search strategy

• Study inclusion/exclusion criteria

• Data extraction methods

• Assessment of study quality

• Methods of data synthesis/analysis

• Funding source.

Data elements from case series include:

• Clinical question

• Enrollment of patients (consecutive or otherwise)

• Whether a single-center or multicenter study

• Patient selection criteria and sample characteristics

• Intervention

• Length of followup

• Validated outcome measures and independence or blinding of outcome assessment

• Statistical analyses

• Results.

Evidence Tables

Templates for evidence tables were created in Microsoft Excel^® and Microsoft Word^®. One reviewer performed primary data abstraction of all data elements into the evidence tables, and a second reviewer reviewed articles and evidence tables for accuracy. Disagreements were resolved by discussion, and if necessary, by consultation with a third reviewer. When small differences occurred in quantitative estimates of data from published figures, the values obtained by the two reviewers were averaged.

Definition of Ratings Based on Criteria

In consultation with the AHRQ Task Order Officer and Technical Expert Panel, the general approach to grading evidence developed by the U.S. Preventive Services Task Force (Harris, Helfand, Woolf, et al. 2001) were applied to primary studies. The quality of the abstracted studies was assessed by two independent reviewers. Discordant quality assessments were resolved with input from a third reviewer, if necessary.

Primary RCTs and Quasi-Experimental Studies

The quality of RCTs and quasi-experimental studies were assessed on the basis of the following criteria:

• Initial assembly of comparable groups: adequate randomization, including concealment and whether potential confounders (e.g., other concomitant care) were distributed equally among groups

• Maintenance of comparable groups (includes attrition, crossovers, adherence, contamination)

• Important differential loss to followup or overall high loss to followup

• Measurements: equal, reliable, and valid (includes masking of outcome assessment)

• Clear definition of interventions

• All important outcomes considered

• Analysis: adjustment for potential confounders, intention-to-treat analysis.

Definition of Ratings Based on Above Criteria. The rating of intervention studies encompasses the three quality categories described here:

• Good: Meets all criteria: Comparable groups are assembled initially and maintained throughout the study (followup at least 80 percent); reliable and valid measurement instruments are used and applied equally to the groups; interventions are spelled out clearly; all important outcomes are considered; and appropriate attention is given to confounders in analysis. In addition, for RCTs, intention-to-treat analysis is used.

• Fair: Studies were graded “fair” if any or all of the following problems occur, without the fatal flaws noted in the “poor” category below: In general, comparable groups are assembled initially but some question remains whether some (although not major) differences occurred with followup; measurement instruments are acceptable (although not the best) and generally applied equally; some but not all important outcomes are considered; and some but not all potential confounders are accounted for. Intention-to-treat analysis is done for RCTs.

• Poor: Studies were graded “poor” if any of the following fatal flaws exists: Groups assembled initially are not close to being comparable or maintained throughout the study; unreliable or invalid measurement instruments are used or not applied at all equally among groups (including not masking outcome assessment); and key confounders are given little or no attention. For RCTs, intention-to-treat analysis is lacking.

Systematic Reviews and Meta-Analyses

Assessment of the quality of systematic reviews and meta-analyses were guided by a quality rating method reported by Oxman and Guyatt (1991; Overview Quality Assessment Questionnaire).^* Oxman and Guyatt tool results in a quality score, based on the answers to ten questions that provide information on the content of a review in terms of how it was conducted, as follows:

• 1

  Were the search methods used to find evidence on the primary question(s) stated?

• 2

  Was the search for evidence reasonably comprehensive?

• 3

  Were the criteria used for deciding which studies to include in the overview reported?

• 4

  Was bias in the selection of studies avoided?

• 5

  Were the criteria used for assessing the validity of the included studies reported?

• 6

  Was the validity of all the studies referred to in the text assessed using appropriate criteria?

• 7

  Were the methods used to combine the findings of the relevant (to reach a conclusion) reported?^*

• 8

  Were the findings of the relevant studies combined appropriately relative to the primary question of the overview?

• 9

  Were the conclusions made by the author(s) supported by the data and/or analysis reported in the overview?

• 10

  What was the overall scientific quality of the overview? Use the following scoring scale:

The following guidelines are used to apply the Oxman and Guyatt rating:

It should be noted that a new quality assessment tool for systematic reviews and meta-analyses was recently developed (Shea, Grimshaw, Wells, et al., 2007). It was based, in part, on the work of Oxman and Guyatt, but differs in significant ways. In particular, the Oxman and Guyatt tool does not adequately address whether quality concerns of the underlying literature were incorporated into conclusions. The tool by Shea, Grimshaw, Wells, et al. (2007) more clearly assesses whether conclusions took appropriate account of the quality of included studies and the potential for publication bias. The recently developed tool was unavailable during the time when ratings of meta-analyses were performed for this evidence report.

Case Series

Table 2

Carey and Boden case series quality assessment tool (more...)

Table 2

Carey and Boden case series quality assessment tool

Clearly Defined Question	Well-Described Study Population*	Well-Described Intervention	Use of Validated Outcome Measures	Appropriate Statistical Analysis	Well-Described Results	Discussion/Conclusions Supported by Data	Funding/Sponsorship Source Acknowledged
Question should be appropriate to study design;	Case definition (diagnostic criteria); type of criteria (clinical, radiographic); whether criteria used before (reference); explicit inclusion/exclusion criteria;	Sufficiently clear that another center could replicate study; if not identified in detail, should provide references;	Reference to previous validation;	Statistical tests and power calculations aimed at improvement over time; prepost analysis should take into account paired nature of data;	Utilize only validated outcome measures;	Conclusion should be supported by the data in the article	Funding source should be disclosed in addition to consulting or board relationship with manufacturer
should not be stated in terms of effectiveness;	includes standard information (age; sex; socioeconomic status; stage and duration of disease; comorbidities; n; time to accrual; exclusions and reasons; loss to followup; refusal)	cointerventions should be described in reasonable detail	ideally individual assessing patient's outcome should be masked to specific intervention; alternatively, assessor who is not in direct employ of clinical office;	comparisons with historical controls should take into account differences in cointerventions between time periods;	description of adequacy of followup (number lost to followup, number who switch to another provider or pursue other treatments, number who die from other causes);	where other information is used to buttress conclusions, should be explicitly stated and referenced;
best when focused;			standardized length and intervals of observation and of sufficient duration to be clinically meaningful; justification for the duration of followup	attention to nonspecific effects and inability to distinguish procedure's effect from spontaneous improvement;	[adaptation: inclusion of both potentially beneficial outcomes (symptom/function/quality of life) and adverse events]	limitations should be made explicit;
				avoids over-reliance on those variables showing improvement;		description of specific next research steps (e.g., need for RCT, details of RCT) [adaptation: this element disregarded]
				analysis should address multiple comparisons

*

OA criteria noted; minimum set of characteristics: age, sex, disease duration and preop severity described.

• Clearly defined question

• Well-described study population

• Well-described intervention

• Use of validated outcome measures

• Appropriate statistical analyses

• Well-described results

• Discussion and conclusion supported by data

• Funding source acknowledged.

Literature Overview

Five study-level meta-analyses comparing intra-articular hyaluronans with placebo (e.g., arthrocentesis and saline injection) for osteoarthritis (OA) of the knee have been published. One patient-level meta-analysis of a single product was also identified.^* The quality of the meta-analyses was appraised with a validated tool (Oxman and Guyatt, 1991; Oxman, Guyatt, Singer, et al., 1991)—the Overview Quality Assessment Questionnaire.

These meta-analyses included outcome measures from 41 relevant randomized, controlled trials (RCTs). One additional placebo-controlled trial (Rolf, Engstrom, Ohrvik, et al., 2005) identified by our literature search^† was not included in any meta-analysis (42 trials, therefore, included in this review). RCTs pooled by the meta-analyses overlap considerably; their quantitative results and limitations also overlapped. Owing to the broad scope of the meta-analyses, they were judged to effectively capture existing evidence and formed the primary basis for evaluating hyaluronans' effectiveness. Important details relevant to the evidence, or inconsistently reported in the meta-analyses, were abstracted from the primary literature (e.g., sample size and power calculations, use of intention-to-treat or per protocol analyses, industry involvement, quality appraised according to our protocol).

Adverse Events

The meta-analyses examining adverse events described small relative increased risk. Wang, Chen, Huang, et al. (2004) reported a pooled relative risk for minor events of 1.2 (95 percent CI: 1.01 to 1.41) and Arrich, Piribauer, Mad, et al. (2005) 1.08 (95 percent CI; 1.01 to 1.15). Bellamy, Campbell, Robinson, et al. (2006) estimated a pooled relative risk for local reactions accompanying hylan G-F 20 (five RCTs) of 1.9 (95 percent CI: 0.51 to 7.3, five RCTs) and other hyaluronans (5 RCTs) of 1.6 (95 percent CI: 0.54 to 5.6).

Six articles or abstracts were identified addressing adverse event occurrence. Hamburger, Lakhanpal, Mooar, et al. (2003) reviewed hyaluronan product safety profiles from a MEDLINE^® search through July 2002 and the FDA Manufacturer and Device Experience Database (MAUDE).^† The review noted rare occurrence of serious reactions to both Hyalgan^® and hylan G-F 20.

Waddell (2003) described adverse event rate accompanying hylan G-F 20 from a retrospective review in a single clinical practice. He reported a local adverse event rate of 2.1 percent (82/3,931) per injection—1 percent (34/3,367) for those receiving a single course and 8.5 percent (48/564) accompanying a second course.

Maheu and Bonvarlet (2003) surveyed French rheumatologists to explore the occurrence of acute pseudoseptic arthritis post-hyaluronan injection—a severe hyaluronan-related adverse event reportedly uncommon. A questionnaire was sent to 81 rheumatologists of whom 26 responded. Sixteen reported 33 cases or pseudoseptic arthritis, possibly more frequently associated hylan G-F 20. The authors concluded acute pseudoseptic arthritis is “not so rare.” Limitations of the survey included the absence of a denominator to quantify risk and the low survey response rate.^‡

Kemper, Gebhardt, Meng, et al. (2005) reported a 5.3 percent adverse event rate accompanying hylan G-F 20 injections in 4,253 patients. Arthropathy was most common occurring in 3.1 percent of patients. The most severe event reported was a large effusion and synovitis in one patient. Those with previous hyaluronan treatments had a two-fold increased risk of adverse events. Lussier, Cividino, McFarlane, et al. (1996) reported adverse events among 336 patients receiving 1,537 injections of hylan G-F 20. Local adverse events occurred at a rate of 2.7 percent per injection and in 1 of 12 patients.

Finally, a search of MAUDE for hyaluronan products (code MOZ) from January 1, 2005 through January 1, 2007 identified 236 records reporting adverse events following knee injection. Nine reports mentioned pseudosepsis or pseudoseptic reaction—four associated with Synvisc^® (hylan G-F 20), one with Euflexxa^©, and four with Hyalgan^®. In 85 adverse events patients were hospitalized.

Generally, severe adverse events associated with hyaluronan-based products have been reported as uncommon in trials. In contrast, local minor adverse events appear common, although the risk appears not substantially different compared to placebo injection. The true risk of pseudoseptic reactions may be small, but one study suggests they could be more common than generally thought.

Supplementary Analyses Performed by the Evidence-Based Practice Center

We performed supplementary analyses to address three key issues:

• 1

  Heterogeneity—clinical and statistical

• 2

  Publication bias

• 3

  Hylan G-F 20.

The majority of these analyses rely upon data abstracted by Bellamy, Campbell, Robinson, et al. (2006) which included the largest number of trials. However, trial quality ratings we performed and cited throughout this reported were used for all analyses.

Clinical and Statistical Heterogeneity/Sensitivity Analyses. All study-level meta-analyses found high heterogeneity and appropriately employed random effects models. Four of the five identified hylan G-F 20 and trial quality issues as factors affecting pooled estimates. Using post-test VAS pain as the outcome at 5–13 weeks (Bellamy, Campbell, Robinson et al. 2006, Comparison 50, 16 pooled studies), we performed sensitivity analyses exploring factors suggested by the meta-analyses and our own review of evidence:

• Trial quality (good/fair versus poor)^*

• Hylan G-F 20 versus other hyaluronans

• Sample size (≤100 or >100) or reported power calculations (these attributes were correlated; differences according to sample size was found to explain more heterogeneity)

• Industry involvement

• Use of rescue analgesia

• Primary intention-to-treat analyses.^*

Table 22

Results of sensitivity analyses for Bellamy, Campbell, (more...)

Table 22

Results of sensitivity analyses for Bellamy, Campbell, Robinson, et al. (2006) 5–13 week pain

		Random-Effects Model*
Study or Sample Characteristic		WMD VAS 100 mm	95% CI	I2
Study Quality	Good/Fair	-8.8	-12.4 to -5.2	61.0%
	Poor	-23.2	-37.2 to -9.3	89.7%
Hylan	G-F 20	-20.8	-31.3 to -10.4	83.8%
	Others	-9.3	-13.4 to -5.1	68.3%
Sample Size	≤ 100	-17.0	-20.8 to -13.2	26.3%
	> 100	-7.3	-14.6 to 0.4	89.2%
ITT	Yes	-12.8	-18.8 to -6.8	84.6%
	No	-13.5	-22.1 to -4.9	80.2%
Power Calculation	Yes	-9.1	-16.5 to -1.8	86.5%
	No	-16.2	-22.7 to -9.8	78.5%
Rescue Analgesia	Yes	-11.4	-16.3 to -6.6	82.5%
	No	-24.2	-34.6 to -13.7	38.1%
Industry Involvement	Yes	-12.9	-18.5 to -7.3	85.4%
	No	-13.7	-18.4 to -9.0	0.0%*

*

A fixed-effects model.

Add P-values

Characteristics found to influence results next examined in a hierarchical Bayes linear model (DuMouchel, 1994) with a vague prior for τ^{2 †}specified. Study quality and hylan G-F 20 were retained in the model based on these findings and conclusions from the meta-analyses. Of the remaining attributes, only sample size was found independent and statistically significant.^‡ In the model including study quality, use of hylan G-F 20, and sample size all were statistically significant (respective probabilities of .006, .049, and .01) and between-study variability in the model (τ²) was reduced by 38 percent. In the model pooled weighted mean differences in VAS pain varied from -3.0 mm (good/fair study quality, non G-F 20 hyaluronan, sample size >100) to -29.6 mm (poor study quality, hylan G-F 20, sample size ≤100).

Although analyses must be considered exploratory, in subgroup analyses and meta-regressions results were sensitive to study characteristics and use of hylan G-F 20. Industry involvement had no effect on pooled estimates. While the use of rescue analgesia in subgroup analyses influenced results, it was not independent of study quality and use of hylan G-F 20 and only three trials did not allow rescue analgesia. Study quality, hylan G-F 20, and sample size were independently associated with the trial effects explaining a sizeable proportion of between-study variability.

Publication Bias. Three findings suggest the presence of publication bias:

• 1

  Funnel plot asymmetry

• 2

  Small trial bias

• 3

  Unpublished trials.

Funnel Plot Asymmetry. Two meta-analyses found funnel plot asymmetry (Lo, LaValley, McAlindon, et al., 2003; Modawal, Ferrer, Choi, et al., 2005); using sample size as the ordinate Wang, Chen, Huang, et al., (2004) suggested no evidence of asymmetry. Arrich, Piribauer, Mad, et al. (2005) found no evidence of publication bias while Bellamy, Campbell, Robinson et al. (2006) did not report examining potential publication bias.

Funnel plots constructed with precision as the ordinate using data from Wang, Chen, Huang, et al. (2004) showed asymmetry for SPID% (p=0.038) and peak PID% (p=.015) although not for ASPID% (p=.56) which as an average measure could be anticipated.^* In Bellamy, Campbell, Robinson et al. (2006), Egger tests calculated for pooled VAS pain at rest, 1 to 4 weeks, 5 to 13 weeks, and 14 to 26 weeks yielded p-values of .9, <.001, .017, and .086, respectively.^† While other factors could explain these test results (Lau, Ioannidis, Terrin, et al., 2006) those reported in the meta-analyses and those we performed are consistent with publication bias.

Small Trial Bias. An apparent small trial bias was noted by Wang, Chen, Huang, et al. (2004) and shown in our sensitivity analyses. The average size of trials reporting sample size calculations was 204 compared to 60 for those without. The effect magnitude in clearly adequately powered trials was 44 percent lower than in those not reporting sample size calculations—consistent with concluding positive underpowered studies were more often published than negative ones.

Hylan G-F 20. The five study-level meta-analysis suggested hylan G-F 20 has greater effects than other hyaluronans. To extend results from the meta-analyses and explore how the potential effect of hylan G-F 20 might differ, we examined pooled trial results further.

Pooling. Eight trials of hylan G-F 20 assessed outcome measures at different time points using different instruments (Cubukcu, Ardic, Karabulut, et al., 2004; Dickson, Hosie, and English, 2001; Karlsson, Sjogren, and Lohmander, 2002; Kotevoglu, Iyibozkurt, Hiz, et al., 2006; Moreland, Arnold, Saway, et al., 1993; Rolf, Engstrom, Ohrvik, et al., 2005; Scale, Wobig, and Wolpert, 1994; Wobig, Dickhut, Maier, et al., 1998). For consistency and to allow comparison with other meta-analyses, we adopted the general approach taken by Bellamy, Campbell, Robinson, et al. (2006) pooling weighted mean differences between treatment and placebo arms at follow-up. Data extracted by Bellamy, Campbell, Robinson, et al. (2006) at 5 to 13 weeks post-injection (near the time of maximum anticipated benefit) were used.

Results from two trials could not be included in the pooled result. Follow-up in the Moreland, Arnold, Saway, et al. (1993) trial was limited to four weeks. Rolf, Engstrom, Ohrvik, et al. (2005) did not report a pain outcome measure amenable to pooling with the other trials. Five of the remaining six RCTs reported pain on a VAS scale (Dickson, Hosie, and English, [2001] as part of WOMAC 100-mm VAS). Cubukcu et al. (2006) assessed WOMAC pain on a 20 point scale (which we rescaled to 100 for pooling). From Karlsson, Sjogren, and Lohmander (2002) only the hylan G-F 20 and placebo arms were included. Random-effects models were fitted in all but one instance due to heterogeneity.

Figure 3. Oxman and Guyatt rating

   Figure 3. Oxman and Guyatt rating

Figure 4. Forest plot of hylan G-F 20 studies reporting (more...)

   Figure 4. Forest plot of hylan G-F 20 studies reporting VAS pain at 5 to 13 weeks—weight-bearing or WOMAC (95% confidence intervals)

Figure 4

These results can be summarized as follows:

• 1

  The pooled effect magnitude from the available hylan G-F 20 RCTs appears larger than for other hyaluronans.

• 2

  Due to trial quality, drop-out rates, heterogeneity, considerably larger effects in the Wobig, Dickhut, Maier, et al. (1998) and Scale, Wobig, and Wolpert (1994), and between-trial variability, the pooled effect estimate must be considered accompanied by greater uncertainty than reflected in the confidence interval.

Summary and Appraisal

Table 23

Summary pain result closest to 8–12 weeks and (more...)

Table 23

Summary pain result closest to 8–12 weeks and key characteristics of study-level viscosupplementation meta-analyses

	Lo et al., 2003	Wang et al., 2004	Arrich et al., 2005	Modawal et al., 2005	Bellamy et al., 2006
Trials pooled at 8–12 weeks	22	20	5	6	16
Sample size: mean (range) *	134 (24–108)	117 (12–347)	250 (49–408)	181 (80–347)	131 (24–407)
Total patients	2,927	2,345	1,251	1086	2,090
Pooled pain outcome cited†	Hierarchy‡	With/without Activities	During or After Exercise	During Activity or Rest	Weight Bearing
Comparison/Effect Measure	Difference in Change	Differences	Difference	Difference in Change	Difference
	(standardized)	(in pain intensity summed)	(at follow-up)	(unstandardized)	(at follow-up)
	(effect size)	(0–100%)	(mm VAS pain)	(mm VAS pain change)	(mm VAS pain)
Overall pooled effect	-0.32	7.90%	-4.3 mm	-18.1 mm change	-13.0 mm
95% CI	(-0.47 to -0.17)	(4.1% to 11.7%)	(-7.6 to -0.9)	(-29.9 to -6.3)	(-18.0 to -7.9)
p Value	<.001	NR	.013	NR	<.001
Sensitivity Analyses
Trial quality
Good ± Fair)	NR	Reported NS in meta-regression §	-6.2 mm (-15.9 to 3.5) **	-7.1 mm (-11.3 to 3.0)	-8.8 mm (-12.4 to -5.2 )††
Poor	NR		NR	NR	-23.2 mm (-37.2 to -9.3)††
Trial size
Large	NR	3.6% (0.9 to 6.3)	NR	NR	-7.3 mm (-14.6 to -7.7)††
Small	NR	6.0% (2.1 to 10.1)	NR	NR	-17.0 mm (-20.8 to -13.2)††
Molecular weight
G-F 20	NR	23.6% (CI not reported)	Did not include any G-F 20 trials	-33.0 mm ( -50.5 to -17.5) ‡‡	-20.8 mm (-31.3 to -10.4)††
Non G-F 20	-0.19 (-0.27 to -0.10)	5.4% (2.6 to 19.9)		-19.2 mm (-30.5 to -7.9)	-9.3 mm (-13.4 to -5.1)††
Heterogeneity
I2	NR	NR	0%	95%	83%
Other	Cochran Q: P<.001	Cochran Q: p<.001*		Cochran Q: p<.001
Explored/Explained	Yes/Yes†	Yes/No	NA/NA ‡	Yes/Partially§	No/No
Results consistent with publication bias	Yes	No**	No	Yes	Yes††
					(EPC analysis)

*

If not reported in the meta-analysis, figures calculated from original trial publications using patients randomized (not knees).

†

While Arrich et al. (2005) and Bellamy et al. (2006) pooled a similar effect measure, the other meta-analyses chose different approaches detailed in the Methods chapter.

‡

Pain reported from one of the following instruments in order of decreasing preference: global knee pain score; knee pain on walking; WOMAC index; Lequesne Index; knee pain during activities other than walking.

§

Also reported that elements characterizing studies of lower methodologic quality were associated with higher effect estimates.

**

Result from a single high quality trial.

††

From supplementary EPC analyses; not reported in Bellamy et al. (2006).

‡‡

Calculated from meta-regression model also including study quality and pain with activity or at rest, not presented in publication.

CI: confidence interval; NA: not applicable; NR: not reported; NS: not significant (p<.05); VAS: Visual Analog Scale.

Drawing conclusions requires considering the clinical meaning of pooled results, strengths and limitations of the meta-analysis and trial evidence, heterogeneity in pooled results, potential publication bias, and the uncertainty contributed by each.

Clinical Meaning. Important effects, regardless of statistical considerations, must be accompanied by a minimal clinically important improvement patients can identify. While the amount of improvement required may not be definitively established (Tubach, Ravaud, Baron et al., 2005; Pham, van der Heijde, Altman, et al., 2004), between 20 and 40 percent improvements have been used in recent hyaluronan trials (Nuestadt, Caldwell, Burnette, et al., 2005, Altman, Akermark, Beaulieu, et al., 2004). In this respect, pooled results from the meta-analyses are limited due to a primary literature not generally reporting results quantifying proportions responding or achieving likely minimal clinically important improvements for the various outcome measures. Few trials reported response rates and an insufficient number from which to draw conclusions or to combine.

Strengths of the Meta-Analyses. Lo, LaValley, McAlindon, et al. (2003) attempted to acquire intention-to-treat data even if not reported, conducted sensitivity analyses supporting their conclusions, and were able to explain between-trial variability by excluding two outlier results. Wang, Chen, Huang, et al., (2004) reported extensive subgroup results and meta-regressions. Arrich, Piribauer, Mad, et al. (2005) examined effects at different time periods and carefully explored between-trial variability. Bellamy, Campbell, Robinson et al. (2006) examined the greatest breadth of literature. Strand, Conaghan, Lohmander, et al. (2006) was able to examine patient-level data.

Key Limitations of Primary Literature. Trial quality was the fundamental limitation of the primary literature—noted in four of five study-level meta-analyses. The second key limitation was the lack of reported response rates from intention-to-treat samples. This limits applying results to individual patients.

Heterogeneity among trials results was high for pooled outcome measures in all study-level meta-analyses; use of hylan G-F 20 and trial quality were found to influencing pooled effect magnitude and heterogeneity. Supplementary analyses suggested trial size also to account for some heterogeneity.

Potential Publication bias was consistent with Egger test results in three of the meta-analyses (Lo, LaValley, McAlindon, et al., 2003; Modawal, Ferrer, Choi, et al., 2005; Bellamy, Campbell, Robinson et al., 2006), and in Wang, Chen, Huang, et al., (2004), dependent on the choice of ordinate. Lo, LaValley, McAlindon, et al. (2003) also reported larger effect sizes in unpublished trials. Small trial size was associated with larger effects and less often accompanied by sample size calculations; a substantial number of patients were participants in unpublished trials. This evidence supports the presence of publication bias.

Results, Part I: Key Question 3 (Subgroup Analyses)

Four RCTs examined subgroups specified by our protocol including age, sex, primary/secondary OA of the knee, body mass index (BMI)/weight, and disease severity. None examined ethnicity, disease duration, or prior treatment. In one trial a subgroup comparison was preceded by stratified randomization. No other subgroup comparisons were prespecified—results obtained in post-hoc analyses.

Lohmander, Dalen, Englund, et al. (1996) noted the subgroup aged 60 to 75 years with Lequesne Index scores over 10 (worse disease severity) experienced greater reduction in VAS pain compared to placebo (-23 mm versus -7 mm respectively at 13 weeks). However, in a confirmatory trial (Karlsson, Sjogren, and Lohmander, 2002) no benefit was found for that subgroup. This was the only subgroup result tested in a confirmatory study.

Table 24

Results by subgroups from Altman and Moskowitz (1998) (more...)

Table 24

Results by subgroups from Altman and Moskowitz (1998)

		Mean Reduction Walking VAS Pain (mm) Compared to Placebo (and 95% CI; from figure)
Age	<65	-12.0 (-20 to -4)
	≥65	-5.5 (-16 to 6)
Sex	Women	-17.0 (-17to 0)
	Men	-16.0 (-22 to -2)
BMI	≤ 30.5	-6.0 (-13 to 2)
	> 30.5	-16.0 (-25 to -7)
Disease Severity	“Moderate”	-6.0 (-12.5 to 1.5)
	“Severe”	-10.5 (-25 to 2.0)
	KL2	-9.0 (-17 to -1)
	KL3	-7.0 (-13 to 1)

Dahlberg, Lohmander, Ryd, et al. (1994) reported no beneficial effect of hyaluronan in the presence of previous trauma (secondary disease). Henderson, Smith, Pegley, et al. (1994) concluded that “hyaluronan offers no significant benefit over placebo during a five week treatment period...” but also reported effects among those classified as Kellgren-Lawrence grade 2 and grades 3–4—each with separate control groups. At 5 weeks, the VAS pain score in the Kellgren-Lawrence grade 2 hyaluronan arm improved -15.6 mm compared to -14.2 mm for placebo arm; in the Kellgren-Lawrence grade 3–4 hyaluronan arm -8.7 mm, compared to -18.0 mm for placebo. Finally, Petrella, DiSilvestro, and Hildebrand (2002) reported no significant differences within subgroups defined by age, sex, and BMI but estimates were not stated.

Comment. There is no evidence of differential effect of intra-articular hyaluronan according to subgroups defined by age, sex, primary/secondary OA of the knee, BMI/weight, or disease severity. However, the subgroup evidence is limited. The single positive subgroup finding subsequently examined in a confirmatory RCT was not substantiated.

Results, Part I: Key Question 4 (Comparative Outcomes)

The single study comparing the interventions of interest to this Evidence Report was conducted by Forster and Straw (2003). Forster and Straw (2003) randomized patients to arthroscopic lavage and debridement or intra-articular Hyalgan^®. It should be noted that the Forster and Straw trial is the only study meeting selection criteria for this Evidence Report's Key Question 4, concerning the comparative short-term and long-term outcomes of viscosupplements, glucosamine and chondroitin, or arthroscopic lavage and debridement. The trial by Forster and Straw will be discussed separately, in Results, Part III, Key Question 4.

Conclusions: Part I

• 1

  What are the Clinical Effectiveness and Harms of Intra-Articular Hyaluronic Acid/Hyaluron Preparations Injections in Patients With Primary OA of the Knee?

  • Results from 42 trials (N=5,843), all but one synthesized in various combinations in six meta-analyses, generally show positive effects of viscosupplementation on pain and function scores compared to placebo. However, the evidence on viscosupplementation is accompanied by considerable uncertainty due to variable trial quality, potential publication bias, and unclear clinical significance of the changes reported.

    • The pooled effects from poor-quality trials were as much as twice those obtained from higher-quality ones.

    • There is evidence consistent with potential publication bias. Pooled results from small trials (≤100 patients) showed effects up to twice those of larger trials consistent with selective publication of underpowered positive trials. Among trials of viscosupplementation, those that have not been published in full text comprise approximately 25 percent of the total patient population.

    • Interpreting the clinical significance of pooled mean effects from the meta-analyses is difficult; mean changes do not quantify proportions responding. Numbers needed to treat cannot be calculated from mean changes.

  • Trials of hylan G-F 20, the highest molecular weight cross-linked product, generally reported better results than other trials.

  • Minor adverse events accompanying intra-articular injections are common, but the relative risk accompanying hyaluronan injections over placebo appears to be small. Pseudoseptic reactions associated with hyaluronans appear relatively uncommon but can be severe.

• 2

  What are the Clinical Effectiveness and Harms of the Interventions of Interest in Patients With Secondary OA of the Knee?

  • We identified no studies enrolling patients with only secondary disease, or that stratified randomization by primary and secondary disease. There is insufficient evidence to draw conclusions about treatment outcomes in patients with secondary disease.

• 3

  How do the Short-Term and Long-Term Outcomes of the Interventions of Interest Differ by the Following Subpopulations: Age, Race/Ethnicity, Gender, Primary or Secondary OA, Disease Severity and Duration, Weight (Body Mass Index), and Prior Treatments?

  • Four RCTs were identified examining any of the specified subgroups. None examined race/ethnicity, disease duration, or prior treatment. In one trial, randomization was stratified by disease severity; all other subgroup results were obtained in post-hoc analyses. There was no evidence for differential effects according to subgroups defined by age, sex, primary/disease, BMI/weight, or disease severity. One positive post-hoc subgroup analysis found greater efficacy among older individuals with more severe disease, but was not confirmed in a subsequent trial.

• 4

  How do the Short-Term and Long-Term Outcomes of the Interventions of Interest Compare for the Treatment of: Primary OA of the Knee; and Secondary OA of the Knee?

  • No trials were identified comparing intra-articular hyaluronan to glucosamine and/or chondroitin. A single, small, underpowered, poor quality trial found no difference in outcome measures comparing intra-articular hyaluronan to arthroscopy and debridement over a 1-year followup. There is insufficient evidence to draw conclusions regarding comparative efficacy of the interventions.

Literature Overview

This section of the Evidence Report includes six MAs^* and five RCTs not included in the MAs.^† In this section, we provide a brief descriptive overview of the MAs and identify the additional RCTs. Our systematic review of the literature did not identify any patient-level MAs on these interventions.

Table 25

Summary description of meta-analyses of glucosamine (more...)

Table 25

Summary description of meta-analyses of glucosamine and chondroitin in knee osteoarthritis

MA Author, Year	Industry Funding of MA	Key Question(s) Addressed				Included RCT Design				No. of RCTs Included (total pts)		Outcomes Reported
		1	2	3	4	DB	SB	PC	AC	C	G	Pain	Func	Struc	AEs
Bjordal et al., 2006	NR	X	X			X		X	X	6 (362)	7 (401)	X
Towheed et al., 2006	NR	X	X			X		X	X	NA	20 (2,596)	X	X		X
Poolsup et al., 2005	NR	X				X		X		NA	2 (414)	X	X	X	X
Richy et al., 2003	NR	X	X			X		X		8 (855)	7 (1,203)	X	X	X	X
Leeb et al., 2000	NR	X	X			X		X	X	7 (703)	NA	X	X		X
McAlindon et al., 2000	NR	X	X			X		X	X	9 (799)	6 (1,118)	X
No. RCTs Pooled (Total in Literature)										12	21

AC: active-controlled; AEs: adverse events; C: chondroitin; DB: double-blind; G: glucosamine; Func: function; NR: not reported; PC: placebo-controlled; pts: patients; SB: single-blind; Struc: structural; RCT: randomized controlled trial;

Table 26

Additional RCTs not included in glucosamine and chondroitin (more...)

Table 26

Additional RCTs not included in glucosamine and chondroitin meta-analyses

Study	No. Pts per Study Arm					Duration (wks)	Outcomes Reported
	G	C	G/C	Pl	Act		Pain	Func	Struc	AEs
Herrero-Beaumont et al., 2007	106			104	108	24	X	X		X
Clegg et al., 2006	317	318	317	313	318	24	X	X		X
Michel et al., 2005		150		150		104	X	X	X	X
Uebelhart et al., 2004		54		56		52	X	X	X	X
Das and Hammad, 2000			46	47		24	X	X	X	X

Act: active; AEs: adverse events; C: chondroitin; G: glucosamine; G/C: glucosamine plus chondroitin; Pl: placebo; Func: function; Struc: structural; wks: weeks

Results, Part II: Key Questions 1 and 2

Table 27

Oxman and Guyatt method quality evaluation of glucosamine (more...)

Table 27

Oxman and Guyatt method quality evaluation of glucosamine and chondroitin meta-analyses

Evaluation Criteria	Bjordal et al., 2006	Towheed et al., 2006	Poolsup et al., 2005	Richy et al., 2003	Leeb et al., 2000	McAlindon et al., 2000
Were the search methods used to find evidence (primary research) on the primary question(s) stated?	Y - clearly stated	Y - clearly stated	Y - clearly stated	Y - clearly stated	Y - clearly stated	Y - clearly stated
Was the search for evidence reasonably comprehensive?	Y - clearly stated, comprehensive, but language restricted to English, German, Scandinavian	Y - clearly stated, comprehensive, no language restrictions	N - did not specify language restrictions, did not seek unpublished data	Y - clearly stated, comprehensive, no language restrictions	P - search strategy not specified, language restrictions unclear, scope unclear	P - electronic search did not include EMBASE but did include Cochrane database
Were the criteria used for deciding which studies to include in the overview reported?	Y - clearly stated	Y - clearly stated	Y - clearly stated	Y - clearly stated	Y - clearly stated	Y - clearly stated
Was bias in the selection of studies avoided?	Y - comprehensive search, published and unpublished data sought	Y - comprehensive search, published and unpublished data sought	N - Unpublished data not sought or included, language restrictions not specified	Y - comprehensive search, published and unpublished data sought	N - Unpublished data not sought or included, language restrictions not specified	P - electronic search did not include EMBASE
Were the criteria used for assessing the validity of the included studies reported?	Y - numerical score provided according to Jadad et al.	Y - quality scores provided according to Gotzsche and Jadad et al.	Y - numerical score provided according to Jadad et al.	Y - numerical score provided according to Jadad et al.	Y - unclear, no method cited	Y - clearly stated
Was the validity of all studies referred to in the text assessed using appropriate criteria (either in selecting studies for inclusion or in analyzing the studies that are cited)?	Y - validated methods clearly stated	Y - validated methods clearly stated	Y - validated methods clearly stated	Y - clearly stated in tables	Y - clearly stated in tables	Y - validated methods clearly stated
Were the methods used to combine the findings of the relevant studies (used to reach a conclusion) reported?	Y - clearly stated	Y - handling of dichotomous and continuous outcomes clearly stated	Y - clearly stated	Y - handling of dichotomous and continuous outcomes clearly stated	Y - clearly stated	Y - clearly stated
Were the findings of the relevant studies combined appropriately relative to the primary question the overview addresses?	Y - clearly stated	Y - clearly stated	Y - only used 2 studies because of very strict inclusion criteria	P - combined data from studies of both compounds based on the absence of efficacy differences, also mixed in some data from hip pts	Y - clearly stated	Y - clearly stated
Were the conclusions made by the author(s) supported by the data and/or analysis reported in the overview?	Y - analysis within parameters was adequate, but went further in putting results into a “clinical” context for pain perception	Y - thorough analyses broken down according to outcomes scored and adverse events	Y - but limited number of studies reduces the impact of the MA	P - combined data from studies of both compounds based on the absence of efficacy differences, yet stated they were individually efficacious	Y - authors stated MA only “suggests that CS may be useful in OA”.	P - combined enteral and parenteral administration data, made reference to “safety” even though adverse events weren't compiled or analyzed
How would you rate the scientific quality of the overview?*	7	7	3	5	3	4

Y: Yes; P: Partially or can't tell; N: No

*

1&2: extensive flaws; 3&4: major flaws; 5&6: minor flaws; 7 minimal flaws

Table 28

Methodologic characteristics of glucosamine and chondroitin (more...)

Table 28

Methodologic characteristics of glucosamine and chondroitin meta-analyses

Study	Bjordal et al., 2006	Towheed et al., 2006	Poolsup et al., 2005	Richy et al., 2003	Leeb et al., 2000	McAlindon et al., 2000
Heterogeneity
Assessed	Yes	Yes	Yes	Yes	Yes	Yes
Test used	Cochran Q	Chi-square	Cochran Q	Cochran Q	95% CIs of Glass scores	p value reported, but test used not stated
Result	Outcome measures during first 4 weeks of treatment were not heterogeneous	For GS or GH vs. placebo: reduction in pain and LI scores were heterogeneous	Disease progression:	Outcome measures including JSN (p=.95), LI (p=.68), WOMAC (p=.83), mobility (p=.73) showed no heterogeneity	NR	Heterogeneity (p<.001) among chondroitin trials but attributable to a single study (Rovetta 1991)
	GS Q = 1.3	Pain:	Q=0.35	VAS pain likely heterogeneous as RE model was used to combine data (p value not provided)
	CS Q = 1.8	I2 = 88.5%	(p>.1)
	(p>.05 for either)	LI:	Pain:
	I2 = 0 for both comparisons	I2 = 89.4%	Q=0.003
	(due to critically low Q)		(p>.1)
			WOMAC function:
			Q=0.0009
			(p>.1)
			I2 = 0 for all comparisons
			(due to critically low Q)
Meta-Regression
Conducted	Yes	Yes	NR	NR	NR	NR
Factors explored	Drug types within the same class	Pain and function in studies that used Rotta Research Laboratorium preparation of glucosamine versus those that used non-Rotta preparation(s)	NR	NR	NR	NR
	Patient selection criteria
	Missing data in ITT analyses
Sensitivity analysis**	Yes - planned using same subgroups if Q values indicated heterogeneity was present, not necessary for GH/Gs or CS	Yes - Pain, function, radiologic measures in studies with adequate allocation concealment	NR	Yes	NR	Yes for trial size, quality
Funnel plot/publication bias	NR	NR	NR	Funnel Plot (asymmetric)	Yes	Funnel plot (asymmetric, p<.01)
				Egger Test (p=.08)	Non-central t-distribution revealed a relative error of about 30%
Included studies and compounds assessed	CS = 6 single- or double-blind placebo-controlled RCTs	20 double-blind RCTs, GS/GH	2 double-blind placebo-controlled RCTs of GS	15 double-blind placebo-controlled RCTs of GS and CS	7 double-blind placebo-controlled RCTs of CS	CS=6 double-blind placebo-controlled RCTs
	GS = 7 single- or double-blind placebo-controlled RCTs					GS/GH = 9 double-blind placebo-controlled RCTs
Industry sponsored	5 of 6 CS trials industry funded	15/20 connected to Rotta to some degree	NR in meta-analysis, but both studies were funded by Rotta	NR in meta-analysis	NR in meta-analysis	13/15 trials had some connection with a product manufacturer

CS: chondroitin sulfate; GS: glucosamine sulfate; GH: glucosamine hydrochloride; ITT: intent to treat; NR: not reported; RCT: randomized controlled trial

*

If study subgroups examined eliminating those likely to influence or bias results

Table 29

Primary randomized trials included in glucosamine meta-analyses (more...)

Table 29

Primary randomized trials included in glucosamine meta-analyses

Primary Study	Study Design	Route of Administration		Type of Control Used		≥80% Knees	Publication Type		Meta-Analysis* (Year)
		O	P	Pl	Act		Art	Abs	Bjordal (2006)	Towheed (2006)	Poolsup (2005)	Richy (2003)	McAlindon 2000)
Cibere et al., 2004	DB	X		X		X	X			X
McAlindon et al., 2004	DB	X		X		X	X		X	X
Usha and Naidu, 2004	DB	X		X		X	X		X	X
Hughes and Carr, 2002	DB	X		X		X	X		X	X		X
Pavelka et al., 2002	DB	X		X		X	X			X	X	X
Zenk et al., 2002	DB	X		X		NR	X			X
Reginster et al., 2001	DB	X		X		X	X			X	X	X
Rindone et al., 2000	DB	X		X		X	X		X	X		X
Houpt et al., 1999	DB	X		X		X	X		X	X
Houpt et al., 1998	DB	X		X		X		X					X
Qiu et al., 1998	DB	X			X	X	X			X
Rovati, 1997	DB	X		X		X		X		X		X	X
Muller-Fassbender et al., 1994	DB	X			X	X	X			X
Noack et al., 1994	DB	X		X		X	X		X	X		X	X
Reichelt et al., 1994	DB		X	X		X	X			X			X
			(IM)
Lopes Vaz, 1982	DB	X			X	X	X			X
D'Ambrosio et al., 1981	DB		X		X	NR	X			X
			(IV/IM)
Vajaradul, 1981	DB		X	X		X	X			X			X
			(IA)
Crolle and D'Este, 1980	DB		X		X	NR	X			X
			(IM/IA)
Drovanti et al., 1980	DB	X		X		NR	X			X
Pujalte et al., 1980	DB	X		X		X	X		X	X		X	X
No. RCTs Pooled (Total 21 in Literature)									7	20	2	7	6

Abs: abstract; Act: active; Art: article; DB: double-blind; IA: intra-articular; IM: intramuscular; IV: intravenous; NR: not reported; O: oral; P: parenteral; Pl: placebo;

*

Bold face type and shading indicates study that meets Evidence Report selection criteria (see Methods section)

One MA included primary studies that used a reference control, pooling them with studies that used placebo controls (Towheed, Maxwell, Anastassiades, et al., 2006). Glucosamine was administered orally in 17 RCTs and parenterally in four. Two MAs combined data from studies in which glucosamine was administered parenterally with those in which it was given orally (Towheed, Maxwell, Anastassiades, et al., 2006; McAlindon, LaValley, Gulin, et al., 2000). Seventeen studies reported at least 80 percent of patients had knee OA. Four RCTs did not specify the knee as the primary affected joint (Zenk, Helmer, Kuskowski, et al., 2002; D'Ambrosio et al. 1981; Crolle and D'Este, 1980; Drovanti, Bignamini, and Rovati, 1980).

To assess the MAs as a means to address the Key Questions of this Evidence Report, we applied study selection criteria outlined in the Methods chapter to the primary studies in each MA. Two MAs contained RCTs that do not match the criteria specified in our Evidence Report (Towheed, Maxwell, Anastassiades, et al., 2006; McAlindon, LaValley, Gulin, et al., 2000). Ten of 20 RCTs included by the MA by Towheed and colleagues (2006) are not relevant to the aims of this Report, as will be outlined in the Results section for each MA. However, Towheed, Maxwell, Anastassiades, et al. (2006) includes all 10 trials that we have determined are applicable to our Report, whereas Bjordal, Klovning, Ljunggren, et al. (2006) and Richy, Bruyere, Ethgen, et al. (2003) excluded 3 of the 10.

The MA by Poolsup, Suthisisang, Channark, et al. (2005) examined the effect of glucosamine on structural progression of OA of the knee. Only two RCTs report such data (Pavelka, Gatterova, Olejarova, et al., 2002; Reginster, Deroisy, Rovati, et al., 2001). The earliest MA includes only 3 RCTs that meet our selection criteria, but publication chronology may be the key factor in that situation (McAlindon, LaValley, Gulin, et al., 2000). The primary literature on glucosamine comprising the other three MAs is consistent with our selection criteria (Bjordal, Klovning, Ljunggren, et al., 2006; Poolsup, Suthisisang, Channark, et al., 2005; Richy, Bruyere, Ethgen, et al., 2003).

Table 30

Primary randomized trials included in chondroitin meta-analyses (more...)

Table 30

Primary randomized trials included in chondroitin meta-analyses

Primary Study	Study Design	Route of Administration		Type of Control Used		≥80% Knees	Publication Type		Meta-Analysis* (Year)
		O	P	Pl	Act		Art	Abs	Bjordal (2006)	Richy (2003)	Leeb (2000)	McAlindon (2000)
Mazieres et al., 2001	DB	X		X		X	X		X	X
Bourgeois et al., 1998	DB	X		X		X	X		X	X	X	X
Bucsi and Poor, 1998	DB	X		X		X	X		X	X	X	X
Conrozier, 1998	DB	X		X		X	X			X		X
Pavelka et al., 1998	DB	X			X	X		X		X		X
Uebelhart et al., 1998	DB	X		X		X	X		X	X	X	X
Morreale et al., 1996	DB	X			X	X	X		X		X
Conrozier and Vignon, 1992	DB	X		X			X				X
L'Hirondel, 1992	DB	X		X		X	X			X	X	X
Mazieres et al., 1992	DB	X		X			X		X	X	X	X
Rovetta, 1991	DB		X		X	X	X					X
			(IM)
Kerzberg et al., 1987	DB/CO		X		X	X	X					X
			(IM)
No. RCTs Pooled (Total 12 in Literature)									6	8	7	9

Abs: abstract; Act: active; Art: article; DB: double-blind; CO: crossover; IM: intramuscular; O: oral; P: parenteral; Pl: placebo;

*

Bold face type and shading indicates study that meets Evidence Report selection criteria (see Methods section)

Three MAs included RCTs that used reference controls (Bjordal, Klovning, Ljunggren, et al., 2006; Leeb, Schweitzer, Montag, et al., 2000; McAlindon, LaValley, Gulin, et al., 2000). Chondroitin was administered orally in ten trials and parenterally in two. One MA pooled data from RCTs that used either route (McAlindon, LaValley, Gulin, et al., 2000). Ten studies included only patients with OA of the knee. Two included patients with OA of the knee and of the hip (Conrozier and Vignon, 1992; Mazieres, Loyau, Menkes, et al., 1992). The latter 2 RCTs were pooled with OA of the knee patient data in one MA (Leeb, Schweitzer, Montag, et al., 2000).

Our study selection criteria excluded primary studies from each of the four MAs. This is particularly evident with one MA of nine primary studies, five of which would be allowed by our criteria (McAlindon, LaValley, Gulin, et al., 2000).

Outcomes Measured in Randomized Trials That Meet Protocol Selection Criteria. A number of health outcomes reported in primary RCTs provide relevant information to address Key Questions 1 and 2. To facilitate this presentation, where appropriate we have included the studies from the MAs with the additional studies in the summary tables.

Table 31

Clinical outcomes in RCTs of glucosamine that meet (more...)

Table 31

Clinical outcomes in RCTs of glucosamine that meet protocol selection criteria

Study	VAS Pain			WOMAC				Global Assessment		LI	Walking Time
	Motion	Rest	Overall	Pain	Function	Stiffness	Total	Phys	Pat
Studies Included in Meta-Analyses
McAlindon et al, 2004				X	X	X	X
Usha and Naidu, 2004		X								X
Hughes and Carr, 2002	X	X	X	X	X	X
Pavelka et al., 2002				X	X	X	X
Reginster et al., 2001				X	X	X	X
Rindone et al., 2000	X	X
Houpt et al., 1999				X	X	X	X
Rovati, 1997										X
Noack et al., 1994										X
Pujalte et al. 1980			X							X
Additional Studies not Included in Meta-Analyses
Herrero-Beaumont et al., 2007				X	X		X	X	X	X
Clegg et al., 2006				X	X	X	X	X	X

LI: Lequesne Index; VAS: visual analog scale; WOMAC: Western Ontario and McMaster index;

Table 32

Clinical outcomes in RCTs of chondroitin that meet (more...)

Table 32

Clinical outcomes in RCTs of chondroitin that meet protocol selection criteria

Study	VAS Pain			WOMAC				Global Assessment		LI	Walking Time
	Motion	Rest	Overall	Pain	Function	Stiffness	Total	Phys	Pt
Studies Included in Meta-Analyses
Mazieres et al., 2001	X	X						X	X	X
Bourgeois et al., 1998		X								X
Bucsi and Poor, 1998		X								X	X
Conrozier, 1998										X
Uebelhart et al., 1998		X
L'Hirondel, 1992		X								X
Additional Studies not Included in Meta-Analyses
Clegg et al. (2006)				X	X	X	X	X	X
Michel et al. (2005)				X	X	X	X
Uebelhart et al. (2004)		X								X	X

LI: Lequesne Index; pt: patient; VAS: visual analog scale; WOMAC: Western Ontario and McMaster index;

Table 33

Clinical outcomes in RCTs of glucosamine plus chondroitin (more...)

Table 33

Clinical outcomes in RCTs of glucosamine plus chondroitin that meet protocol selection criteria

Study	VAS Pain			WOMAC				Global Assessment		LI	Walking Time
	Motion	Rest	Overall	Pain	Function	Stiffness	Total	Phys	Pt
Clegg et al., 2006				X	X	X	X	X	X
Das and Hammad, 2000							X		X	X

LI: Lequesne Index; pt: patient; VAS: visual analog scale; WOMAC: Western Ontario and McMaster index;

Table 34

Baseline characteristics of randomized trials of glucosamine (more...)

Table 34

Baseline characteristics of randomized trials of glucosamine that meet protocol selection criteria*

Study	Dose (Type)	N Tx/Pl	Mn Age Tx/Pl (yrs)	Female Pts (%) Tx/Pl	BMI (kg/m2) Tx/Pl	OA Diag†	OA Stage (%Tx/%Pl)	Mn Dis Duration Tx/Pl (yrs)	Mn VAS Movement (mm) Tx/Pl	Mn VAS Rest (mm) Tx/Pl	Mn WOMAC Pain Tx/P	Mn WOMAC Function Tx/Pl	Mn WOMAC Stiffness Tx/Pl	Mn WOMAC Total Tx/Pl
Studies Included in Meta-Analyses
McAlindon et al., 2004	1,500 mg/day (GS)	101/104	Rng <54–95/<54–84	57/71	31.0 ± 7.6/34.1 ± 9.0	?	NR	NR			Likert 8.8/9.1	Likert 4.2/4.1	Likert 30.2/31.6	Likert 43.2/44.8
				p=.04	p=.01
Usha and Naidu, 2004	1,500 mg/day (inferred GS)	30/28	52/50	60/57	26.6/25.4 (calculated)	?	K-L	3.2/2.9		58/NR
							1–3 most
Hughes and Carr, 2002	1,500 mg/day (GS)	40/40	All: 62	All: 68	NR	?	K-L 1 (all, 9)	All: 7.6	All: 60.7	All: 35.0	Likert All: 9.2	Likert All: 32.9	Likert All: 4.4
							2 (all 31)
							3 (all 37)
							4 (all 23)
Pavelka et al., 2002	1,500 mg/day (GS)	101/101	61/64	79/76	25.7 ± 2.1/25.7 ± 1.8	1°	K-L 2 (54/53)	10.1/11.0			Likert 6.6/6.3	Likert 21.8/22.0	Likert 2.2/2.2	Likert 30.7/30.5
							3 (46/47)
Reginster et al., 2001	1,500 mg/day (GS)	106/106	66/66	75/78	27.3 ± 2.6/27.4 ± 2.7	1°	K-L2 (71/70)	8.0/7.6			194.1/172.2	740.1/670.8	96.0/96.7	1030/940
							3 (29/30)
Rindone et al., 2000	1,500 mg/day (unclear)	49/49	63/64	4/6	NR	?	K-L 1 (40/30)	12/14	(0–10) 6.4/6.4	(0–10) 3.9/3.6
							K-L 2(18/19)
							K-L 3 (35/35)
							K-L 4 (7/16)
Houpt et al., 1999	1,500 mg/day (GH)	58/60	64/65	64/60	NR	1°	NR	8.3/8.3			Likert 8.8/8.4	Likert 33.4/30.1	Likert 4.1/4.0	Likert 46.4/42.4
Rovati (1997)	1,500 mg/day (GS)	NR	NR	NR	NR	?	NR	NR	NR (used LI)‡	NR	NR	NR	NR	NR
Noack et al., 1994	1,500 mg/day (GS)	126/126	55/55	59/62	26.6/26.2 (calculated)	1°	NR	All: rng <6 mo to >10 yr
Pujalte et al., 1980	1,500 mg/day (GS)	10/10	59/65	80/90	NR	?	NR	NR
Additional Studies not Included in Meta-Analyses
Herrero-Beaumont et al., 2007 [GUIDE]	1,500 mg/day (GS)	106/108/104	GS: 63.4 ± 6.9	91/93/89	GS: 27.7 ± 2.3	1°	K-L 2: 50/56/50	GS: 7.4 ± 6.0			GS: 7.8 ± 3.0	GS: 27.8 ± 11.4		GS: 38.3 ± 15.2
			Acet: 63.8 ± 6.9		Acet: 27.9 ± 2.3		K-L 3: 41/31/39	Acet: 6.5 ± 5.3			Acet: 8.0 ± 2.9	Acet: 29.4 ± 11.0		Acet: 40.4 ± 14.8
			Pl: 64.5 ± 7.2		Pl: 27.6 ± 2.4		K-L 2/3: 9/12/11	Pl: 7.2 ± 5.8			Pl: 7.9 ± 3.0	Pl: 27.2 ± 10.9		Pl: 37.9 ± 14.3

*

All values are mean ± SD unless otherwise noted;

†

ACR criteria;

‡

Outcomes are generally those that are denoted in the paper as being the primary study outcomes;

Acet: acetaminophen; ACR: American College of Rheumatology; BMI: body-mass index; Dis: disease; GS: glucosamine sulfate; GH: glucosamine hydrochloride; K-L: Kellgren-Lawrence criteria; LI: Lequesne Index; mn: mean; NR: not reported; Pl: placebo; rng: range; Tx: treatment; VAS: visual analog scale; WOMAC: Western Ontario and McMaster Universities Osteoarthritis Index;

The mean age of patients ranged between 50 and 66 years, with females comprising 4–90 percent of the study samples. In nine of 11 trials, females made up 60 percent or more of the enrolled patients. Five RCTs of glucosamine reported on patients with primary OA according to ACR criteria. None of the glucosamine studies reported patients specifically with secondary OA. Six reports did not specify whether patients had primary or secondary OA. The mean duration of OA of the knee ranged from 6 months or less to more than 10 years. Most patients in the RCTs had Kellgren-Lawrence grade 2–3 OA of the knee. One study included subjects who had Kellgren-Lawrence grade 4 disease (Hughes and Carr, 2002). No significant differences were reported between the composition of the treatment and placebo groups or their baseline characteristics, with the exception of a slight variation in sex distribution and BMI reported in one study (McAlindon, Formica, LaValley, et al., 2004).

Table 35

Baseline characteristics of randomized trials of chondroitin (more...)

Table 35

Baseline characteristics of randomized trials of chondroitin treatment that meet protocol selection criteria*

Study	Dose (Type)	N Tx/Pl	Mn Age Tx/Pl (yrs)	Female Pts (%) Tx/Pl	BMI (kg/m2) Tx/Pl	OA Diag†	OA Stage (%Tx/%Pl)	Mn Dis Duration Tx/Pl (yrs)	Mn VAS Movement (mm) Tx/Pl	Mn VAS Rest (mm) Tx/Pl	Mn WOMAC Pain Tx/P	Mn WOMAC Function Tx/Pl	Mn WOMAC Stiffness Tx/Pl	Mn WOMAC Total Tx/Pl	Mn LI Tx/Pl
Studies Included in Meta-Analyses
Mazieres et al., 2001	1,000 mg/day (CS)	63/67	67/67	71/78	29.2 ± 5.1/28.9 ± 4.8	1°	K-L	NR	54.4/53.0	29.9/27.7					8.8/8.9
							2 (59/54)
							3 (41/46)
Bourgeois et al., 1998	Daily 1,200 mg/day (CS 4&6)	Daily/3X daily//Pl 40/43/44	63/63/64	65/79/84	NR	1°	ACR	By L,R 6,5/4,5/6,6		58/54/56					11/10/10
	3X daily						All: 1–3 (100)
	400 mg/day (CS 4&6)
Bucsi and Poor, 1998	800 mg/day (CS)	39/46	61/59	56/63	29.2/29.1 (estimated)	1°/2°	K-L	NR		56/56					R,L
							All: 1–3 (100)								12.8,12.0/11.8,11.5
Conrozier, 1998	800 mg/day (CS 4&6)	All: 104	NR	NR		?	NR	NR							~9.0/~9.1
Uebelhart et al., 1998	800 mg/day (CS 4&6)	23/23	60/57	48/56	25.5/27.2 (estimated)	1°/2°	K-L	NR		56/64
							1 (44/48)
							2 (48/44)
							3 (9/9)
L'Hirondel, 1992	1200 mg/day (CS)	63/62	All: 63	32.6	NR	?	NR	NR		(0–5) 4.03/3.90					10.73/11.02
Additional Studies not Included in Meta-Analyses
Michel et al., 2005	800 mg/day (CS 4&6)	150/150	62/63	51/52	27.7 ± 5.2/28.1 ± 5.5	1°	K-L	NR			(0–10) 2.5/2.7	(0–10) 2.1/2.5	(0–10) 3.0/3.5	(0–10) 2.3/2.6
							All: 1–3 (100)
Uebelhart et al., 2004	800 mg/day (CS 4&6)	54/56	63/64	80/82	NR	1°	K-L	4.2/4.4		58.8/61.1					9.0/9.1
							1 (7/6)
							2 (32/33)
							3 (15/17)

*

All values are mean ± SD unless otherwise noted;

†

ACR criteria;

ACR: American College of Rheumatology; BMI: body-mass index; CS: chondroitin sulfate; Dis: disease; K-L: Kellgren-Lawrence criteria; LI: Lequesne Index; mn: mean; NR: not reported; Pl: placebo; rng: range; Tx: treatment; VAS: visual analog scale; WOMAC: Western Ontario and McMaster Universities Osteoarthritis Index;