Article Text

Extended report
Clinical efficacy and safety of glucosamine, chondroitin sulphate, their combination, celecoxib or placebo taken to treat osteoarthritis of the knee: 2-year results from GAIT
  1. Allen D Sawitzke1,
  2. Helen Shi2,
  3. Martha F Finco1,
  4. Dorothy D Dunlop3,
  5. Crystal L Harris4,
  6. Nora G Singer5,
  7. John D Bradley6,
  8. David Silver7,
  9. Christopher G Jackson1,
  10. Nancy E Lane8,
  11. Chester V Oddis9,
  12. Fred Wolfe10,
  13. Jeffrey Lisse11,
  14. Daniel E Furst12,
  15. Clifton O Bingham13,
  16. Domenic J Reda2,
  17. Roland W Moskowitz5,
  18. H James Williams1,
  19. Daniel O Clegg1
  1. 1University of Utah School of Medicine, Salt Lake City, Utah, USA
  2. 2Hines VA Cooperative Studies Program Coordinating Center, Hines, Illinois, USA
  3. 3Feinberg School of Medicine, Northwestern University, Chicago, Illinois, USA
  4. 4Albuquerque VA Cooperative Studies Program Clinical Research Pharmacy Coordinating Center, Albuquerque, New Mexico, USA
  5. 5Case Western Reserve University, Cleveland, Ohio, USA
  6. 6Indiana University, Indianapolis, Indiana, USA
  7. 7Cedars Sinai Medical Center, Los Angeles, California, USA
  8. 8University of California at Davis, Sacramento, California, USA
  9. 9University of Pittsburgh, Pittsburgh, Pennsylvania, USA
  10. 10Arthritis Research Center, Wichita, Kansas, USA
  11. 11Univerity of Arizona, Tucson, Arizona, USA
  12. 12University of California at Los Angeles, Los Angeles, California, USA
  13. 13Johns Hopkins University, Baltimore, Maryland, USA
  1. Correspondence to Dr Allen D Sawitzke, University of Utah School of Medicine, 30E 1900 S SOM 4B200, Salt Lake City UT 84132, USA; allen.sawitzke{at}


Background Knee osteoarthritis (OA) is a major cause of pain and functional limitation in older adults, yet longer-term studies of medical treatment of OA are limited.

Objective To evaluate the efficacy and safety of glucosamine and chondroitin sulphate (CS), alone or in combination, as well as celecoxib and placebo on painful knee OA over 2 years.

Methods A 24-month, double-blind, placebo-controlled study, conducted at nine sites in the US ancillary to the Glucosamine/chondroitin Arthritis Intervention Trial, enrolled 662 patients with knee OA who satisfied radiographic criteria (Kellgren/Lawrence grade 2 or 3 changes and baseline joint space width of at least 2 mm). This subset continued to receive their randomised treatment: glucosamine 500 mg three times daily, CS 400 mg three times daily, the combination of glucosamine and CS, celecoxib 200 mg daily, or placebo over 24 months. The primary outcome was a 20% reduction in Western Ontario and McMaster University Osteoarthritis Index (WOMAC) pain over 24 months. Secondary outcomes included an Outcome Measures in Rheumatology/Osteoarthritis Research Society International response and change from baseline in WOMAC pain and function.

Results Compared with placebo, the odds of achieving a 20% reduction in WOMAC pain were celecoxib: 1.21, glucosamine: 1.16, combination glucosamine/CS: 0.83 and CS alone: 0.69, and were not statistically significant.

Conclusions Over 2 years, no treatment achieved a clinically important difference in WOMAC pain or function as compared with placebo. However, glucosamine and celecoxib showed beneficial but not significant trends. Adverse reactions were similar among treatment groups and serious adverse events were rare for all treatments.

View Full Text

Statistics from


Osteoarthritis (OA) of the knee results in chronic pain in a sizeable minority of the population and considerable expense1 is associated with the resultant disability and costs of treatment directed to control that pain.2,,4 Conventional drugs for chronic pain, such as non-steroidal anti-inflammatory drugs (NSAIDs), including cyclo-oxygenase-2 inhibitors, increase the risk for adverse events (AEs).5 In addition, many patients with OA are receiving drugs for long periods of time, have comorbidities and use other drugs for those illnesses, all of which may further increase the likelihood of AEs. The increase in OA-related musculoskeletal pain especially in the growing older population has created heightened interest in effective treatments with better safety profiles. The Glucosamine/chondroitin Arthritis Intervention Trial (GAIT) was designed to examine the effects of the dietary supplements glucosamine and chondroitin sulphate (CS) given alone or in combination as compared with celecoxib or placebo on pain associated with OA of the knee.

What is already known on this topic

  • ▶. It is already known that successful longer-term medical treatment of osteoarthritis (OA) of the knee is limited.

  • ▶. It is estimated that only 15–20% will continue to receive a given non-steroidal anti-inflammatory drug after 1 year of use, because of a combination of loss of efficacy and/or accumulated toxicities.

  • ▶. Relatively long-term efficacy and safety data for glucosamine and chondroitin sulphate alone, and especially in combination, are also limited.

What this study adds

  • ▶. This study adds long-term data to the efficacy and safety data not only for the nutraceuticals but also for the coxib–celecoxib.

  • ▶. All data were obtained using dosages typically used to treat OA of the knee and gathered prospectively.

  • ▶. Only one other report of celecoxib use in OA of greater than 1 year in duration has been reported.

The primary symptomatic outcome assessment of GAIT was after 24 weeks of randomised, double-blind, placebo-controlled treatment.6 An ancillary structural study describing the effects of the agents on radiographic joint space width loss for up to 2 years has also recently been reported.7 To date, very few data have been reported for the long-term treatment of OA with any agent including the supplements and celecoxib studied here. Patients in the structural study continued to have safety and clinical efficacy assessments at all scheduled visits. This paper is a subset analysis of the randomised design detailing the clinical efficacy and safety experience with these agents alone and in combination together with celecoxib as compared with placebo in patients from this subset of GAIT over 24 months of follow-up.


Study design

Study subjects enrolled in the GAIT ancillary structural study met the original GAIT inclusion criteria, summarised as at least 40 years of age with clinical evidence of painful OA of the knee for at least 6 months and radiographic evidence of OA as determined by a Kellgren and Lawrence grade 2- or grade 3-rated radiograph of the index knee. Before baseline, washout periods for excluded drugs were completed and all patients had a normalised Western Ontario and McMaster University Osteoarthritis Index (WOMAC) pain subscore between 15 and 80 at baseline. Of the 1583 original GAIT study participants, 662 were also in the structural study.7 Participants in the structural study continued to receive their originally assigned blinded treatment for up to 2 years and continued to have safety monitoring and efficacy assessments as prescribed by the protocol (approximately quarterly) throughout the 2-year follow-up period. Patients were asked to continue follow-up even if they stopped taking their assigned treatment. The study treatments were glucosamine hydrochloride (HCl) 500 mg three times daily, sodium CS 400 mg three times daily, both glucosamine and CS as above, celecoxib 200 mg once daily or placebo daily. The study included a ‘double dummy’ and ‘double placebo’ design. Up to 4 g paracetamol daily could be taken as rescue analgesia, but patients were instructed not to take this drug within 24 h of a follow-up visit to allow for accurate measurement of their current pain levels.

Outcome measures

The primary outcome measure for the original 6-month symptomatic GAIT study was the number of patients who attained a 20% decrease in their summed WOMAC pain subscale8 from baseline to week 24. For this 2-year follow-up study, we report the likelihood of reaching a 20% reduction in WOMAC Pain score from baseline modelled over 2 years in those patients who participated in this extension of the original symptomatic treatment study.6

Secondary outcome measures

Secondary measures included the actual amount of pain reduction attributable to each treatment, the WOMAC function subscale and the likelihood of achieving an Outcome Measures in Rheumatology (OMERACT)/Osteoarthritis Research Society International (OARSI) response over 24 months.9 The OMERACT/OARSI response is included as it was reported in the original GAIT report, and each component had been prospectively collected. Further, the OMERACT/OARSI response at 24 weeks was more discriminating of improvement than was a 20% reduction in WOMAC pain at 24 weeks in the original GAIT report.6

Adverse events

Information on AEs was collected for each patient at each visit in response to an open-ended query. The reported events were assessed by the investigator at each study visit and followed up until resolution or the end of follow-up. AEs or serious adverse events (SAEs) were assessed according to Food and Drug Administration (FDA) guidelines.10 Monitoring of vital signs, blood counts, serum glucose, aminotransferases, creatinine, partial thromboplastin time and urine analysis was also performed according to protocol. An independent data and safety monitoring board reviewed study performance and safety data annually.

Statistical analysis

All analyses compared placebo with each of the four treatments on a modified intention-to-treat basis. Baseline characteristics were compared across groups using a χ2 test for categorical variables and analysis of variance for continuous variables. Statistical testing of treatment differences, which were adjusted for the comparison of each of the four treatment groups with the placebo group, was performed by using multivariate t statistics (analogous to Dunnet's t test) to calculate 95% CI.11 Multiple regression analyses longitudinally compared each intervention group with placebo while controlling for design factors (weeks receiving treatment, weeks squared (to account for non-linearity)) and literature-based clinical factors (gender, baseline pain, disease duration, body mass index class, Kellgren and Lawrence grade) over 24 months. Logistic regression was used to analyse the odds of a response over 24 months of follow-up separately for 20% WOMAC pain reduction and OMERACT/OARSI outcomes. Linear regression was used to analyse the mean change from baseline over 2 years of follow-up separately for WOMAC pain and WOMAC function. Recognising that dropout before the 2-year visit in this study may not be completely at random, we applied selection models as described by Hogan et al12 using weighted generalised estimating equations to estimate the multiple regression models. This form of repeated measures analysis uses all data collected on this cohort while accounting for potentially non-random dropout. These analyses were implemented with SAS 9.1.



Since not all patients randomised to GAIT at the ancillary sites qualified for the ancillary study, comparison of baseline data was done to assess any loss of comparability of the treatment groups due to departure from randomisation. Table 1 shows that the baseline data were similar among all groups with no statistically significant differences between them. Most participants were female, most had body mass index >30 and their average age was nearly 57 years. The average duration of OA symptoms was approximately 10 years. Slightly more Kellgren and Lawrence grade 3 knees were found in patients treated with the combination of glucosamine and CS than in the other groups.

Table 1

Baseline characteristics of the structural study by group

Figure 1 shows the change in WOMAC pain over time as well as the number of participants contributing data at each visit. We tested and found no statistically significant differences in dropout by treatment group. In addition, approximately 60% of the achieved improvement occurred within the first three visits (18 weeks) for all groups, and the magnitude of improvement did not differ significantly by treatment group. Although the rate of this improvement was most rapid for celecoxib, this does not represent a statistically significant difference when compared with placebo (see online supplementary table 2 for detailed data).

Figure 1

Decline in Western Ontario and McMaster University Osteoarthritis Index (WOMAC) function and pain subscores versus week of study. The number of participants contributing data at each visit is shown in the table below the matching week.

Table 2

Odds of pain response over 24 months versus placebo by WOMAC and OMERACT/OARSI

The odds of achieving benefit over 24 months by use of an agent, whether defined as a 20% WOMAC pain response or OMERACT/OARSI response is shown in table 2. While no agent was statistically more effective than placebo, a trend toward improvement occurred with celecoxib and with glucosamine for both end points. The trends were most apparent when using the OMERACT/OARSI responder criteria.

The specific changes in WOMAC pain and function subscores by treatment after adjustment for design and clinical factors are shown in table 3. The magnitude of improvement over 24 months was greatest for celecoxib. However, when compared with placebo, no significant differences were shown for WOMAC pain or for WOMAC function for any of the interventions.

Table 3

Change in WOMAC pain and function score over 24 months versus placebo

In the original GAIT report, a subset of participants with more severe baseline pain appeared to benefit by use of the combination of glucosamine and CS. This question was examined in our study by including interactions between severe pain and treatment group; none of these interactions was statistically significant.


AEs were reported as the percentage of patients who reported the AE at least once during follow-up. They were characterised according to standard FDA reporting language as AEs or SAEs.10 AEs statistically divergent from placebo were mild, showed no important differences and were evenly distributed among treatment groups. There were 84 SAEs occurring in 64 patients, of which five were assessed as possibly related to the study drugs. These included myocardial infarction (in a patient receiving the glucosamine/CS combination, hereafter referred to as ‘combination’), coronary angioplasty (in the placebo group) and hip arthroplasty, cerebrovascular accident and abdominal wall abscess (all receiving celecoxib). Other SAEs reported regardless of relatedness are one death, which occurred as a completed suicide (placebo group), two myocardial infarctions (one glucosamine, one combination), two cerebrovascular accidents (one glucosamine, one celecoxib), two hypertension (one combination, one placebo), one patient with palpitations (combination) and one transient ischaemic attack (combination). No serious adverse gastrointestinal bleeding events were reported.

Compliance was evaluated by pill counts at each visit. While patients continued in the trial, the percentage of prescribed drug taken was 90%. The use of rescue paracetamol averaged 570 mg daily. The lowest use was in the celecoxib (465 mg) group and the highest use in the placebo group (645 mg). Interestingly, the rank order of rescue drug use (least to greatest) exactly paralleled that of the primary efficacy outcome for this study (see online supplementary tables 1 and 2).


Literature examining the use of the dietary supplements, especially glucosamine and CS, to treat OA has increased substantially over the past 5 years. Reviews describe their effects on symptoms13,,15 and to a lesser extent on structural modification.7 16 17 Unfortunately, the duration of most of the studies for the symptomatic treatment of OA is relatively short, especially when compared with the duration of use expected for these agents in patients affected by OA of the knee. Systematically collected data on the safety of these agents are also limited by short study durations. The previous report from GAIT examined the primary outcome in the overall study population of 1583 patients at 24 weeks.6 This report describes the efficacy and safety over 2 years of exposure to the treatments.

The patients reported on in this 2-year cohort from GAIT are, as would be expected of patients with OA in general, predominantly female, overweight and over 45 years of age. Because this sample represents a subset analysis of a randomised design, bias related to non-random dropout was considered; we found that the baseline characteristics of the treatment group subsets were balanced. As shown in table 3, improvement in pain and function over 24 months was seen in all groups, and no treatment was statistically better than placebo. However, patients treated with celecoxib and with glucosamine monotherapy had the greatest improvement. Table 2 shows an improved sensitivity to change relative to placebo for the OMERACT/OARSI outcome measure as compared with a 20% reduction in WOMAC pain score. Overall, our results are in agreement with the Cochrane review of glucosamine, which found glucosamine ineffective for treatment of pain in studies that used a WOMAC pain outcome.15 Interestingly, in the Cochrane analysis, studies which used the Lequesne index as an outcome measure showed that glucosamine sulphate was superior to placebo. Two trials, including the parent to this study (GAIT), reported that use of glucosamine HCl, did not show a benefit compared with placebo.6 18 At the time GAIT was developed, glucosamine HCl was the only preparation available that met pharmaceutical grade manufacturing specifications that were required for the Investigational New Drug application. While meta-analyses have supported benefit for OA pain from use of CS,19 20 we found no evidence for such a benefit. A protocol for a Cochrane review of CS for OA has been defined, but a completed analysis has not been reported.

Remarkably few published reports of efficacy data exist in OA of the knee beyond 6-month duration. Those studies that are published typically use patient global pain measures rather than validated composite indexes. For example, naproxen was shown equal to diclofenac for 26 weeks based on continuation of use and patient reports of pain at rest,21 while 2-year data for naproxen using a five-point Likert scale showed it to be statistically similar to paracetamol.22 Sustained-release diclofenac was also studied for 2 years and shown to be similar to placebo as assessed by patient report on a five-point Likert scale.23 Use of celecoxib to treat OA for 1 year has been reported.24 Celecoxib has demonstrated efficacy as compared with placebo at 12 weeks as a mean reduction in WOMAC pain score,25 and was effective at 24 weeks in the primary GAIT study.6 The ancillary study reported here suggests a waning of benefit with longer usage as has been described for NSAIDs in general.7

AEs were mild and occurred in all groups over the 2-year period. Only five SAEs were thought to be related to study agents. While there were concerns that glucosamine might exacerbate diabetes or asthma, 2–3-year studies from Europe and these data do not substantiate those concerns.26 27 One- and 2-year safety data for celecoxib come from a trial on prevention of adenomatous polyps,28 and an ankylosing spondylitis trial.29 Reports of the relative safety of celecoxib are supported by our findings, which show no difference from the safety profile of placebo. While the number of participants using celecoxib in this study is modest, it is noteworthy that no AEs related to gastrointestinal bleeding were reported and no difference in cardiovascular events as compared with use of placebo was demonstrated.

Perhaps the most interesting finding is the pronounced and persistent clinical improvement in WOMAC pain score for all treatment groups, including placebo. The improvement is seen early and persists throughout the 2-year study period. This response cannot easily be dismissed as regression to the mean as it persists throughout the entire 24-month duration of follow-up. Patients in flare studies may demonstrate regression toward the mean, but this study did not require flare criteria for entry, and patients did not systematically discontinue drugs before entry. If the agents indeed had therapeutic benefit to account for the change, then the placebo group's improvement is difficult to explain, as it is identical to the other arms. It is well recognised that many variables affect placebo response in OA.30 Factors relating to this study that may have affected the placebo responsive include the increased placebo effect observed with use of therapeutic agents of low effect size and also the increased placebo effect with large trial size. While a clinical benefit of being closely followed up by a study team is reported, it is not typically of this magnitude or of this extended duration.

In another 2-year study in OA evaluating risedronate as a possible structure-modifying agent in OA, there was a very similar benefit demonstrated in all treatment groups as well as in the placebo group, which was detected at the first efficacy time point at 6 months, and persisted through the 2 years of the study.31 Risedronate decreases biochemical markers of cartilage degradation, but does not decrease symptoms or slow radiographic progression in patients with medial compartment OA of the knee.32 While the study allowed patients to take background analgesics, all patients were required to discontinue NSAIDs and analgesics in a proscribed manner before all clinical assessments to allow for a possible flare to better detect a treatment effect of the study drugs. While in this study, regression to the mean is a possibility, baseline levels of pain were also low (mean WOMAC pain 42.4), making it more difficult to detect a treatment effect due to a floor effect of the outcome measurement. Also, a significant expectation bias on the part of participants as reflected in the very high completion rate at 2 years (76.4%) might have contributed.

As shown by both of these longer-term studies demonstrating similarly sustained placebo benefits,6 30 33 further evaluation of the factors involved is warranted and will be important in designing future OA trials. This study is limited by its design as an ancillary preplanned continuation of a randomised controlled trial in which all of the initially randomised patients were not eligible to participate. There was significant dropout manifested as discontinuation of drug use as well as missed assessments. Recognising that selective dropout can bias study findings, specific analytical methods were applied to mitigate this problem.

In conclusion, findings from this study provide important longer-term safety information on the use of glucosamine and CS given alone or in combination and the use of celecoxib in a 2-year placebo-controlled trial for treatment of OA of the knee. All the tested treatments appeared to be generally safe and well tolerated over a 2-year period. A clinically detectable symptomatic benefit was seen as early as 24 weeks in all groups, including placebo, and was sustained over 2 years of study. Although none of the agents were statistically better than placebo, celecoxib and glucosamine gave the highest odds of benefit for improved pain and function, but with widely overlapping CI for all treatments.


The authors gratefully acknowledge statistical advice from Dr Donald Hedeker, University of Illinois.


View Abstract

Supplementary materials

  • Web Only Data ard.2009.120469

    Files in this Data Supplement:


  • Funding Supported by NIH (National Institute of Arthritis and Musculoskeletal and Skin Diseases and National Center for Complementary and Alternative Medicine, contract N01-AR-2236.

  • Competing interests None.

  • Ethics approval This study was conducted with the approval of the University of Utah and each participating centre.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.