The clinical effectiveness of glucosamine and chondroitin supplements in slowing or arresting progression of osteoarthritis of the knee: a systematic review and economic evaluation

Article history

The research reported in this issue of the journal was commissioned and funded by the HTA programme on behalf of NICE as project number 07/61/01. The protocol was agreed in February 2008. The assessment report began editorial review in December 2008 and was accepted for publication in March 2009. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the referees for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

None

Copyright statement

© 2009 Queen’s Printer and Controller of HMSO. This monograph may be freely reproduced for the purposes of private research and study and may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NETSCC, Health Technology Assessment, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2009 Queen’s Printer and Controller of HMSO

Aims and objectives

The aim of this systematic review and economic analysis was to assess the clinical effectiveness and cost-effectiveness of glucosamine sulphate or hydrochloride and chondroitin sulphate in modifying the progression of OA of the knee.

In particular, we sought to address the following questions:

• Do glucosamine and/or chondroitin prevent or slow progression of OA of the knee?

• Is treatment with glucosamine and/or chondroitin safe?

• Is the addition of glucosamine and/or chondroitin to current clinical practice cost-effective?

• Where could research most effectively be focused to address remaining uncertainties?

Osteoarthritis of the knee: an overview

Osteoarthritis is the most prevalent joint disorder and cause of disability in the UK. 1 Although the incidence increases with age, its onset is not inevitable. Most commonly, OA affects the knees, hips and small joints of the hands. The focus of this review is on OA of the knee and the information presented below, while often applicable to OA at other anatomical sites, will concentrate on the knee joint.

Although there are different systems with which OA is classified, OA can be broadly separated into primary and secondary OA. Primary, idiopathic, OA is the most common. Secondary OA results from previous trauma, infection or congenital abnormality. 2

Clinically, OA leads to asymmetric joint swelling, joint crepitus, decreased range of movement and occasionally joint locking. Symptomatically, patients may complain of joint pain with associated short-lived early morning stiffness; however, the degree of severity of symptoms is hugely variable and does not necessarily represent radiological change.

Burden of OA

Current standard care/optimal care

Management of OA of the knee

In the UK, information about current health-care practice in the management of OA of the knee is scarce. 28 While we may not have clear details of current practice, a guideline produced by NICE in 2008,29 and based on work by the National Collaborating Centre for Chronic Conditions (NCCCC),1 provides recommendations for the care pathway for people with OA. Figure 1 summarises the recommended care. The three core treatments in the centre circle were recommended for all patients. Paracetamol and topical NSAIDs were recommended as the first line in analgesia; treatments in the outer circle then being recommended for only selected individuals. The guideline authors did, however, note a number of areas of uncertainty including: the most effective combinations of treatments; how to improve adherence to the non-pharmacological interventions; and how most effectively to treat very elderly patients.

FIGURE 1.

Targeting treatment: a summary of the treatments recommended. COX-2, cyclo-oxygenase-2; NSAIDs, non-steroidal anti-inflammatory drugs; TENS, transcutaneous electrical nerve stimulation. (Reproduced from: National Collaborating Centre for Chronic Conditions. Osteoarthritis: national clinical guideline for care and management in adults. London: Royal College of Physicians, 2008. Copyright © 2008 Royal College of Physicians. Reproduced by permission.)

Reproduced from: National Collaborating Centre for Chronic Conditions. Osteoarthritis: national clinical guideline for care and management in adults. London: Royal College of Physicians, 2008. Copyright © 2008 Royal College of Physicians. Reproduced by permission.

Studying OA treatments in trials

Studying the impact of treatments on OA within trials has been challenging. Progression of disease occurs over years and decades. The number of people deteriorating to need arthroplasty is relatively small in relation to the total population with knee OA. Thus, studying interventions that aim to modify the course of the condition requires either very long follow-up or a surrogate marker for progression.

Surrogate end points for trials are often criticised. In knee OA, two main types of end points have been adopted in trials of interventions: composite scoring systems for pain and function specific to OA; and measures of joint space.

Interventions under investigation

Key questions and rationale

Some of the findings in this chapter have implications for the interpretation of the evidence, including the following:

• OA of the knee is diagnosed mostly on clinical grounds.

• Progression is variable; symptoms improve in a significant minority.

• Radiological changes do not usually regress, but often do not progress.

• Clinical symptoms may not correlate with radiological changes.

• Progression is slow, over many years.

• There are tried and tested instruments for assessing the impact of OA, such as WOMAC.

• Joint space is the main objective measure, but there can be some subjectivity in quantifying. The link to long-term outcomes is not well established.

In undertaking this review and economic evaluation, we were aware that a number of reviews had already been published. In general, these had included studies of short duration and had not focused on long-term outcomes. In this report, we were interested in the progression of knee OA and whether glucosamine and/or chondroitin could modify the disease process. We were, therefore, only interested in trials of sufficient duration to assess long-term outcomes. We have restricted our review to studies of a minimum of 12 months’ follow-up.

The questions we planned to address in this report are summarised below:

1. Does glucosamine prevent or slow progression of OA of the knee?

2. Does chondroitin prevent or slow progression of OA of the knee?

3. If so, does the combination of both offer any greater effect?

4. If they are useful, what dosages should be used?

5. What are the costs, in terms of both the product costs and any side effects?

6. Are there any savings to the NHS from their use, for example by reduced consumption of prescribed NSAIDs, or from avoidance of knee arthroplasty procedures?

7. Is there evidence of cost-effectiveness from well constructed economic evaluations?

From the evidence review we sought to be able to come to one of three possible conclusions:

1. The evidence base is sufficient to conclude that they are not effective.

2. The evidence base is sufficient to conclude that these products are clinically effective, in which case further research into clinical effectiveness is not required.

3. The evidence is inconclusive, but suggestive enough of benefit to justify further research.

If further research was indicated, then we planned to address several further questions, where the data permitted:

1. Are the alleged benefits of glucosamine and chondroitin on knee cartilage biologically plausible?

2. If, as suggested by the commissioning brief, the evidence suggests that there is no gain in short-term symptom relief, but only in long-term preservation of cartilage (which does imply longer-term reductions in symptoms, advanced OA and perhaps of the need for knee arthroplasty), what should the comparator in trials be?

3. If the gain is in long-term outcomes, are there short-term indicators of benefit which could be used, rather than waiting 20 years for advanced OA to become manifest?

4. Which variables or assumptions (e.g. costs, QoL, mortality, extrapolation of effectiveness over time) are most important in generating uncertainty in the cost-effectiveness estimates, and which would provide the greatest return in terms of reducing uncertainty at a reasonable cost. These variables and assumptions will be identified using value of information analysis, and will form the main basis in recommendations for further research.

Methods

Results

The results have been presented in two sections. First, we present the findings from the review of systematic reviews. Secondly, we report the results of the review of the primary RCTs that met the criteria for this review.

Review of systematic reviews

Search results

The searches of MEDLINE, EMBASE and the Cochrane Library combined (as shown in Appendix 1) retrieved 677 articles. The duplicates were removed, and the remaining abstracts checked against the inclusion criteria. This resulted in 23 articles, for which the full papers were obtained for further checking. The QUOROM flow diagram of search results is shown in Figure 2.

FIGURE 2.

QUOROM flow diagram of search results.

Of the 23 potential relevant systematic reviews, five were included in the analysis. Twelve of the remaining reviews were excluded because they were not systematic reviews (i.e. they did not include a description of their methodology and did not systematically summarise the data). The remaining six were systematic reviews, but one was a review of nutritional supplements in OA specifically excluding glucosamine and chondroitin, one only reported short-term data (up to a maximum of 12 weeks) and the remaining four did not consider structural outcomes (the bibliographies of these four reviews were searched to ensure no individual studies met our inclusion criteria for trials). Details of reasons for study exclusion are shown in Appendix 2.

Of the remaining five reviews, two dealt with glucosamine only,62,73 two dealt with chondroitin only 61,74 and one dealt with both glucosamine and chondroitin. 75 All five reviews included a meta-analysis. Additionally, a recent NCCCC guideline for the management of OA was identified, which included a systematic assessment of glucosamine and chondroitin. 1 This guideline was used to inform the NICE guideline for the management of OA29 and was the only included review that considered combination therapy with glucosamine and chondroitin. Table 5 shows details of the five reviews and one guideline included.

TABLE 5 - Characteristics of included systematic reviews and guideline

NSAID, non-steroidal anti-inflammatory drug; QoL, quality of life; RCT, randomised controlled trial; VAS, visual analogue scale.

Study	Inclusion criteria and methodology	Included studies	Quality
Glucosamine
Poolsup 200573 Thailand Focus: structural and symptomatic efficacy and safety of glucosamine in knee OA	Inclusion criteria Study design: double-blind RCTs; study duration at least 1 year Participants: patients with primary knee OA Interventions: oral glucosamine Outcomes: primary outcome – joint space narrowing; additional outcomes – symptom modification, adverse events Methodology Search strategy: MEDLINE, EMBASE, BIOSIS, Evidence-Based Medicine (EBM) Reviews, Cochrane Library; search from inception of database; scanning of reference lists; search terms indicated; date of last search August 2004 Study selection: no details on study selection given Quality assessment: the methodological quality of each study was assessed using the scale developed by Jadad Data extraction: data were extracted independently by two authors; disagreements were resolved by consensus Data analysis: meta-analysis; statistical methodology appropriate Subgroups/sensitivity analyses: none	Number of included trials: 2 Number of participants: 414 (range 202–212) Trial quality: Jadad quality scores 4 and 5 Study characteristics (comment): both trials fulfil the inclusion criteria of this review; see details of the respective trials below	Inclusion criteria described: yes Details of literature search given: yes Study selection described: no Data extraction described: yes Study quality assessment described: yes Study flow shown: yes Study characteristics of individual studies described: yes Quality of individual studies given: yes Results of individual studies shown: yes Statistical analysis appropriate: yes Overall quality: high
Towheed 200562 Canada (Cochrane review) Focus: effectiveness and toxicity of glucosamine in OA	Inclusion criteria Study design: RCTs, single or double blind, placebo controlled or active control Participants: adults with primary or secondary OA at any site (not temporomandibular joint) Interventions: glucosamine vs placebo or active control; no combination therapies Outcomes: primary outcomes – pain, range of motion, functional assessments, global assessments; additional outcomes – structural benefits, adverse events Methodology Search strategy: MEDLINE, EMBASE, Cochrane Controlled Trials Register and Database of Systematic Reviews, AMED, ACP Journal Club, DARE; search from inception of databases; scanning of reference lists; letters to authors and industry for additional trial; detailed search strategy given; date of last search January 2005 Study selection: study selection was done independently by two reviewers; if randomisation was unclear, authors were contacted; agreement by consensus Quality assessment: the methodological quality of each study was assessed using the Jadad scale and a scale developed by Gøtzsche for NSAID trials in rheumatoid arthritis (maximum score 8) Data extraction: data were extracted independently by two reviewers; disagreements resolved by consensus Data analysis: meta-analysis; statistical methodology appropriate Subgroups/sensitivity analyses: allocation concealment vs no allocation concealment; Rotta preparation of glucosamine sulphate vs non-Rotta glucosamine preparation	Number of included trials: 20 Number of participants: 2596 (range 30–319) Trial quality: Gøtzsche scale mean design and analysis scores 3.7 and 6.3 (of 8, medians similar); mean total scores 10 (of 16). Jadad mean score 4.2. Inadequate allocation concealment in 10 trials Study characteristics (comment): mostly studies with durations of < 1 year (18/20); not all studies investigating only knees; not all studies investigating oral glucosamine; not all studies with placebo control	Inclusion criteria described: yes Details of literature search given: yes Study selection described: yes Data extraction described: yes Study quality assessment described: yes Study flow shown: yes Study characteristics of individual studies described: yes Quality of individual studies given: yes Results of individual studies shown: yes Statistical analysis appropriate: yes Overall quality: high
Chondroitin
Bana 200674 France Focus: symptomatic and structural efficacy and safety of chondroitin sulphate	Inclusion criteria Study design: randomised, double-blind, placebo-controlled trials; study duration at least 4 weeks Participants: patients with OA of the knee or hip Interventions: oral chondroitin sulphate Outcomes: symptoms – Lequesne index, WOMAC, investigator’s or patient’s global assessment, VAS pain, walking time, NSAID or analgesic consumption; structure – mean joint space width, joint surface area, minimum joint space narrowing; adverse events Methodology Search strategy: MEDLINE/PubMed, Cochrane Controlled Trials Register; scanning of reference lists; search terms indicated; limited to studies published in English or French; search limited to studies published in peer-reviewed journals between 1980 and 2005; date of last search 2005 Study selection: no details on study selection given Quality assessment: 14-item quality instrument used by two reviewers; differences resolved by consensus Data extraction: no details on data extraction given Data analysis: meta-analysis; statistical methodology appropriate; however, in two cases the authors included two parts of a single study in their meta-analyses as though they were separate studies – bias? Subgroups/sensitivity analyses: none	Number of included trials: 7 Number of participants: 909 Trial quality: demographic baseline data well matched; quality scores: 6 for one study, 7 for two studies, 9 for one study, 11 for two studies, 13 for one study Study characteristics (comment): three studies lasted a year or longer; authors noted that pain evaluation was biased owing to concomitant consumption of analgesics	Inclusion criteria described: yes Details of literature search given: yes (but limited) Study selection described: no Data extraction described: no Study quality assessment described: yes Study flow shown: yes Study characteristics of individual studies described: yes Quality of individual studies given: yes Results of individual studies shown: yes Statistical analysis appropriate: unclear Overall quality: moderate
Reichenbach 200761 Switzerland/Germany/UK Focus: effects of chondroitin on pain in patients with OA	Inclusion criteria Study design: randomised or quasi-RCTs Participants: patients with OA of the knee or hip Interventions: chondroitin vs placebo or no treatment; trial groups given low doses excluded (< 400 mg/day administered orally) Outcomes: primary – pain [at end of trial or at a maximum of 3 months after termination of chondroitin therapy (whichever came first)]; when study provided data on more than one pain scale, previously described hierarchy of pain-related outcomes referred to; secondary – changes in mean and minimum joint space width; adverse events (numbers/withdrawals because of/serious) Methodology Search strategy: Cochrane CENTRAL (1970–2006), MEDLINE (1966–2006), EMBASE (1980–2006), CINAHL (1982–2006); detailed search strategy shown; no language restrictions; extensive supplemental searches; date of last search 30 November 2006 Study selection: two reviewers independently evaluated reports for eligibility; disagreements resolved by discussion Quality assessment: two reviewers independently assessed concealment of treatment allocation, blinding and adequacy of analyses (handling of missing data, intention-to-treat analysis); disagreements resolved by discussion with a third reviewer and consequent consensus; table presented listing study design, allocation concealment, reported to be double blind, adequate blinding of patients, adequate blinding of therapists, described as placebo controlled, withdrawal rate in chondroitin group, withdrawal rate in control group, intention to treat analysis performed, method to handle missing data Data extraction: in duplicate, using standardised form; only first halves of crossover trials used; authors contacted for additional information where necessary; disagreements resolved by discussion with a third reviewer and consequent consensus Data analysis: meta-analysis; statistical methodology appropriate Subgroups/sensitivity analyses: analyses stratified by – concealment of allocation, use of a placebo control, patient blinding, adequacy of analyses in accordance with the intention-to-treat principle, trial size (cut-off 200 patients), funding, route of administration, length of follow-up (cut-off 26 weeks), difference in use of co-interventions in trial groups; included in univariable meta-regression – chondroitin dosage (in trials with oral administration), treatment duration, length of follow-up	Number of included trials: 22 (20 included in meta-analysis) Number of participants: 4056 (range 17–631) Trial quality: recently performed trials tended to be larger and of higher quality; generation of allocation sequence and allocation concealment unclear for 19 trials; only three trials used intention-to-treat analysis; most trials did not describe approaches for handling missing data; withdrawal rates reported in 13 trials (between 0 and 0.7, most between 0.02 and 0.27); all but three reported to be double blind Study characteristics (comment): most studies had a short study duration; some studies of non-oral chondroitin and non-knee applications included; noted that unequal consumption of analgesics between comparison groups was an unlikely reason for small effects found	Inclusion criteria described: yes Details of literature search given: yes Study selection described: yes Data extraction described: yes Study quality assessment described: yes Study flow shown: yes Study characteristics of individual studies described: yes Quality of individual studies given: yes Results of individual studies shown: yes Statistical analysis appropriate: yes Overall quality: high
Glucosamine/chondroitin
NCCCC guideline 20081 UK Focus: management of OA	Inclusion criteria Study design: systematic reviews, RCTs, (observational studies); trials with sample size < 40 excluded Participants: adults with OA Interventions: glucosamine or chondroitin alone or in compound form vs placebo Outcomes: symptoms, function, QoL, structural changes Methodology Search strategy: MEDLINE (1966 to April 2007), EMBASE (1980 to April 2007), CINAHL (1982 to April 2007), Cochrane Library (April 2007), AMED (1985 to April 2007); search terms not indicated; papers published/accepted by peer-reviewed journals included; conference paper abstracts and non-English language papers excluded Study selection: exclusion lists generated with reasons for exclusions; full papers obtained where relevant Quality assessment: critical appraisal checklists compiled for each paper by one researcher; compliant with NICE technical manual; evidence considered by guideline development group for accuracy and completeness Data extraction: data extracted by one researcher; evidence considered by guideline development group for accuracy and completeness Data analysis: evidence tables and evidence synthesis; evidence statements graded Subgroups/sensitivity analyses: none	Number of included trials: two systematic reviews (Towheed 2005, Reichenbach 2007), six additional RCTs Number of participants: 10,342 (4504 glucosamine trials, 3999 chondroitin trials, 1839 combination glucosamine /chondroitin) Trial quality: all glucosamine trials described as high quality/methodologically sound; all chondroitin and trials of combined glucosamine/chondroitin trials described as methodologically sound Study characteristics (comment): Towheed 2005 and Reichenbach 2007 summarised above; of the additional RCTs included, all have been considered by the present review and were either included or not considered eligible	Criteria described: yes Details of literature search given: partially Study selection described: yes Data extraction described: partially Study quality assessment described: partially Study flow shown: no Study characteristics of individual studies described: partially Quality of individual studies given: yes Results of individual studies shown: yes Statistical analysis appropriate: yes Overall quality: high/moderate
Richy 200375 Belgium/France Focus: structural and symptomatic efficacy of glucosamine sulphate and chondroitin sulphate in knee OA	Inclusion criteria Study design: randomised, double-blind, parallel, placebo-controlled trials performed between January 1980 and March 2002; treatment period at least 4 weeks Participants: patients with OA of the knee or hip Interventions: oral glucosamine or chondroitin Outcomes: at least one of the following outcomes: joint space narrowing, Lequesne index, WOMAC, VAS pain, VAS for mobility assessment, responders to treatment, adverse effects Methodology Search strategy: MEDLINE, PreMEDLINE, BIOSIS Previews, HealthSTAR, EMBASE, Cochrane Controlled Trials Register, Current Contents, EBM Reviews; search for studies published between January 1980 and March 2002; search terms indicated; no language restrictions; extensive supplemental searches; date of last search March 2002 Study selection: potentially relevant publications identified were reviewed by two independent authors for eligibility; in the case of disagreement, a third author was consulted; authors’ names and sources were blinded for the selection process Quality assessment: trials fulfilling the inclusion criteria were blindly scored by two reviewers using the Jadad scale; differences were resolved by consensus; authors’ names and sources were blinded for quality assessment Data extraction: data for predefined outcomes were extracted blindly by two authors using a standardised form; a third reviewer was consulted in case of disagreement Data analysis: meta-analysis; statistical methodology appropriate; data for glucosamine and chondroitin were analysed together after checking for all outcomes that there were no significant differences between the two compounds Subgroups/sensitivity analyses: none	Number of included trials: 15 (seven glucosamine trials, eight chondroitin trials) Number of participants: 1775 (1020 glucosamine trials, 775 chondroitin trials) (range 24–319) Trial quality: glucosamine trials – three scored 4/5 and four 5/5; chondroitin trials – five scored 3/5, one 4/5, one 5/5, one had insufficient detail; individual demographic baseline data well matched, demographic baseline data for glucosamine and chondroitin trials also well matched Study characteristics (comment): many short-term studies; all studies of knee OA	Criteria described: yes Details of literature search given: yes Study selection described: yes Data extraction described: yes Study quality assessment described: yes Study flow shown: yes Study characteristics of individual studies described: yes Quality of individual studies given: yes Results of individual studies shown: yes Statistical analysis appropriate: yes Overall quality: high

Quality of included reviews

Four of the five included reviews were classified as being high quality reviews, with three61,62,75 fulfilling all the quality criteria and one73 failing to give details of study selection. One review74 was rated as moderate quality, as neither study selection nor data extraction were described and details on statistical analysis were lacking. The guideline1 described certain methodological aspects rather briefly, but referred to methods manuals which appeared to assure a high quality systematic reviewing process.

Inclusion criteria

Key inclusion criteria and review characteristics are shown in Table 6. The inclusion criteria varied substantially between reviews. One review73 had narrow inclusion criteria, looking at double-blind RCTs lasting at least 12 months and studying the effect of oral glucosamine in primary knee OA. All the studies included in this review, therefore, met with our inclusion criteria of 12 months’ duration/follow-up. The other four reviews and one guideline included at least one study of a minimum of 12 months’ duration (so were included here), but had wider inclusion criteria and, therefore, also included a number of studies of shorter duration.

TABLE 6 - Summary of review characteristics

QoL, quality of life; RCTs, randomised controlled trials.

Poolsup 200573	Towheed 200562	Bana 200674	NCCCC guideline 20081	Reichenbach 200761	Richy 200375
Inclusion RCTs, double blind, duration at least 1 year Primary knee OA Glucosamine vs placebo Joint space narrowing, symptoms, adverse events Quality: high Included trials: 2 (414 participants)	Inclusion RCTs, single or double blind Primary or secondary OA Glucosamine vs placebo or active intervention Pain, range of motion, functional assessment, structural benefits, adverse events Quality: high Included trials: 20 (2596 participants)	Inclusion RCTs, double blind, duration at least 4 weeks OA of the knee or hip Oral chondroitin sulphate vs placebo Symptoms, pain, structure, adverse events Quality: moderate Included trials: 7 (909 participants)	Inclusion Systematic reviews, RCTs OA in adults Glucosamine or chondroitin alone or in compound form vs placebo Symptoms, function, QoL, structural changes Quality: moderate Included trials: 2 systematic reviews, 6 additional RCTs (10,342 participants)	Inclusion Randomised or quasi-randomised controlled trials OA of the knee or hip Chondroitin vs placebo or no treatment Pain, joint space width, adverse events Quality: high Included trials: 22 (4056 participants)	Inclusion RCTs, double blind, duration at least 4 weeks OA of the knee or hip Oral glucosamine or chondroitin vs placebo Joint space width, function, pain, adverse events Quality: high Included trials: 15 [7 glucosamine (1020 participants), 8 chondroitin (775 participants)]

Trials included in systematic reviews

The five reviews and one guideline described a total of 45 controlled trials – 21 studying glucosamine, 20 studying chondroitin – and four studying a combination of chondroitin and glucosamine (one of which compared combination therapy with glucosamine or chondroitin monotherapy). Table 7 summarises the characteristics of the 45 primary studies. The NCCCC OA guideline1 included two systematic reviews61,62 and six additional RCTs. For glucosamine, only two RCTs (the long-term studies by Reginster and colleagues76 and Pavelká and colleagues77) were included in all the relevant reviews. For chondroitin, only six studies were included in all the relevant reviews. The largest number of primary studies was included in the review by Towheed and colleagues62 for glucosamine and Reichenbach and colleagues61 for chondroitin, and these two reviews also tended to provide the most detail. Therefore, most of the detail presented on the primary studies derives from these two reviews.

TABLE 7 - Distribution of primary studies in the included reviews

	Glucosamine only		Chondroitin only		Both
	Poolsup 200565	Towheed 200573	Bana 200674	Reichenbach 200761	NCCCC 20081	Richy 200375
Glucosamine
Cibere 200478		✓			✓
Crolle 198079					✓
D’Ambrosio 198180		✓			✓
Drovanti 198081					✓
Herrero-Beaumont 200782					✓
Houpt 199983		✓			✓
Hughes 200284					✓	✓
McAlindon 200485		✓			✓
Müller-Fassbender 199486		✓			✓
Noack 199487		✓			✓	✓
Pavelká 200277		✓			✓	✓
Pujalte 198088					✓	✓
Qiu 199889		✓			✓
Reginster 200176		✓			✓	✓
Reichelt 199490					✓
Rindone 200091		✓			✓	✓
Rovati 199792					✓	✓
Usha 200493		✓			✓
Vajaradul 198194					✓
Vaz 198295		✓			✓
Zenk 200296					✓
Chondroitin
Alekseeva 199997				✓	✓
Bourgeois 199898			✓	✓	✓	✓
Bucsi 199899			✓	✓	✓	✓
Conrozier 1992100				✓	✓
Fleisch 1997101				✓	✓
Kahan 2007102				✓	✓
Kerzberg 1987103				✓	✓
L’Hirondel 1992104				✓	✓	✓
Mazieres 1992105			✓	✓	✓	✓
Mazieres 2001106			✓	✓	✓	✓
Mazieres 2007107				✓	✓
Michel 2005108			✓	✓	✓
Morreale 1996109				✓	✓
Nasonova 2001110				✓	✓
Pavelká 1999111				✓	✓	✓
Rovetta 1991112				✓	✓
Soroka 2002113				✓	✓
Uebelhart 1998114			✓	✓	✓	✓
Uebelhart 1999115				✓	✓
Uebelhart 2004116			✓	✓	✓	✓
Glucosamine/chondroitin
Clegg 2006117				✓	✓
Cohen 2003118					✓
Das 2000119					✓
Rai 2004120					✓

The glucosamine trials included between 20 and 1583 participants. With the exception of two trials, the proportion of women included was generally greater than the proportion of men (up to 91% women, with two trials including 48% and 5% women). The mean age range of participants was between 51 and 75 years. Study duration was between 3 weeks and 3 years, with 15 trials having a study duration of no longer than 3 months and two having a duration of at least 1 year. The chondroitin trials included between 17 and 622 participants. As in the glucosamine trials, there were generally more women than men in the trials (33–94%), and mean age was between 50 and 67 years. Trial duration was between 6 weeks and 2 years, with the trials generally having a longer duration than the glucosamine trials (one had a duration of less than 3 months; five had a duration of at least 1 year). One trial, studying a combination of chondroitin and glucosamine, had 1583 participants with a mean age of 58 years and 27% women, and a trial duration of 24 weeks.

Only one of the single therapy and two of the combination therapy trials gave glucosamine hydrochloride rather than glucosamine sulphate. One single therapy trial switched from glucosamine sulphate to glucosamine hydrochloride halfway through the trial owing to supply problems for the glucosamine/placebo product initially used. Towheed and colleagues62 reported data for nine trials using the Rotta preparation of glucosamine and for eight trials using a non-Rotta preparation. None of the studies included Alateris, the only licensed glucosamine product in the UK.

Findings

The findings of the reviews need to be compared with caution; some reviews were more inclusive than others. Of the glucosamine trials, three did not clearly include only knee OA, four gave glucosamine by injection and five had active rather than placebo control groups (three compared with ibuprofen, two compared with piperazine plus chlorbutanol). These glucosamine trials were included in the review by Towheed and colleagues,62 but in none of the others. Of the chondroitin trials, one considered hip rather than knee OA and two gave chondroitin by injection. These chondroitin trials were included in the review by Reichenbach and colleagues,61 but in none of the others. Reichenbach and colleagues,61 also included several studies published in abstract form only, although additional information was obtained where possible. Of the combination therapy trials, three had a duration of less than a year, one studied a topical application of glucosamine and chondroitin, and two examined no structural outcomes. These combination therapy studies were included in the NCCCC OA guideline 2008. 1

Appendix 3 summarises the results of the meta-analyses of the included reviews and Appendix 4 shows the results of any relevant subgroup or sensitivity analyses. Fundamental to our research question, none of the reviews presented data separately for trials of longer duration for all outcomes. For structural changes, most of the trial data have been limited to longer trials, so conclusions from three of the reviews were based solely on trials of more than 1-year follow-up. The other two reviews, and the guideline, based their conclusions about structural changes on a mixture of shorter- and long-term studies

As the NCCCC OA guideline1 focused mainly on two systematic reviews already included here and the only additional RCT eligible for the present review is already described below, the results of that guideline are not shown in further detail. Table 8 summarises the conclusions and recommendations of the reviews. Trials and reviews are hereafter referred to by the name of the first author.

TABLE 8 - Conclusions and recommendations of reviews

BMI, body mass index; CONSORT, Consolidated Standards of Reporting Trials; ICERs, incremental cost-effectiveness ratios; NSAIDs, non-steroidal anti-inflammatory drugs; RCTs, randomised controlled trials; VAS, visual analogue scale.

Study	Authors’ conclusions	Recommendations for research	Additional comments
Glucosamine
Poolsup 200573	Pain/function: The results showed evidence in support of an improvement of symptoms by glucosamine sulphate Structure (based on two studies of > 1-year follow-up): The results showed evidence in support of a delay of structural progression by glucosamine sulphate Adverse effects: The evidence suggests that glucosamine sulphate is safe in long-term use	High quality, placebo-controlled long-term trials of various forms (different salts, different forms of glucosamine sulphate) of glucosamine are needed	An effect on physical function may only be achieved during long-term use
Towheed 200562	Pain/function: Overall, glucosamine showed a moderate clinically significant effect on pain compared with placebo. Pooled results from studies using a non-Rotta preparation of glucosamine or adequate allocation concealment failed to show benefit in pain and WOMAC function, while those studies evaluating the Rotta preparation showed that glucosamine was superior to placebo in the treatment of pain and functional impairment resulting from symptomatic osteoarthritis. WOMAC outcomes of pain, stiffness and function did not show a superiority of glucosamine over placebo for both Rotta and non-Rotta preparations of glucosamine Structure (based on three studies of variable follow-up): Glucosamine may modify the natural radiological progression of osteoarthritis of the knee (delaying progression) Adverse effects: Glucosamine was as safe as placebo	Questions remaining to be answered Are different glucosamine preparations sold by different manufacturers equally effective and safe in the therapy of OA? Is glucosamine sulphate equally effective as glucosamine hydrochloride? Is there any further benefit obtained by using mixed glucosamine preparations that contain additional therapeutic products, such as chondroitin sulphate? Is glucosamine helpful for all stages of OA severity? What is the optimal dose for maximising efficacy and minimising toxicity? What are the patient specific predictors of favourable effects on radiological progression of OA?	There is uncertainty about the effect of glucosamine on glucose metabolism with long-term use – no evidence of an adverse effect was found by two studies
Chondroitin
Bana 200674	Pain/function: The results analysed suggested a true, but modest effect of chondroitin sulphate on relief of pain and improvement of joint function in patients treated. However, the authors concluded that the evaluation could be biased, as only few methodologically sound papers were available Structure (based on three studies > 1 year): Structure (mean and minimum joint space width) appears to be preserved with chondroitin sulphate (versus deterioration in the placebo groups), but in the main study showing a structural advantage, no effect on the WOMAC score was seen Adverse effects: The tolerance of chondroitin sulphate can be considered as excellent	Further large studies lasting at least 2 years are needed to confirm any effects of chondroitin sulphate on preservation of structure/joint space width Studies needed to ascertain for which OA patients these drugs are useful (e.g. patients with mild to moderate disease)	Pain evaluation is biased as in most studies concomitant analgesic/NSAID use is allowed – but often not reported in detail A latency period of about 1 month was observed before an advantage of chondroitin for pain relief/function could be seen The activity of chondroitin appears to last for up to 3 months after the drug is stopped, so cycles of treatment/no treatment might be justified
Reichenbach 200761	Pain/function: There was a high degree of heterogeneity between trials, making interpretation difficult; most studies were poor quality; large-scale, methodologically sound trials indicate that the symptomatic benefit of chondroitin is minimal or non-existent Structure (based on five studies of varying duration): A small effect on joint space width in favour of chondroitin was seen, with uncertain clinical significance Adverse effects: No evidence was found that chondroitin is unsafe Practice recommendation: The use of chondroitin in routine clinical practice should be discouraged. In patients with advanced OA an effect is unlikely; in patients with low-grade OA use of chondroitin should be restricted to RCTs	Future trials should adhere to methodological standards reducing possible biases; reporting should adhere to standards such as the CONSORT criteria A rigorously designed, adequately powered randomised placebo-controlled trial restricted to patients with low-grade OA would be needed to assess if chondroitin is more effective in patients with mild disease than in patients with more severe disease	A significantly larger effect of chondroitin was seen in trials in which use of rescue analgesics was reported to be higher in the control groups Concomitant use of medication (analgesics, NSAIDs) was poorly reported – but small/lack of effect seen was unlikely to be related to concomitant analgesic/NSAID use Recently performed methodologically sound trials with effect sizes near zero tended to have lower proportions of patients with low-grade OA – but the effect of chondroitin may depend on OA stage
Glucosamine/chondroitin
NCCCC guideline 20081	General/pain/function/adverse events: The evidence from the trials is often difficult to compare owing to differences in the products used (and their bioavailability), between-study populations, patient BMI, and the use of analgesia at the time of pain and function assessment in the trials. Overall, trials using glucosamine sulphate as a single dose of 1500 mg rather than hydrochloride 500 mg three times a day showed a small benefit compared with placebo for treatment of knee OA. However, at the time the guideline was prepared, the hydrochloride preparation has been granted a European Medicines Evaluatory Agency licence, while the sulphate has not. The evidence for efficacy of chondroitin was less convincing Evidence to support the efficacy of glucosamine hydrochloride as a symptom modifier is poor. For the non-licensed product (glucosamine sulphate), the evidence is not strong enough to warrant recommending that it should be prescribed on the NHS. Notwithstanding some evidence of benefit and very little evidence of harm in clinical practice, and despite the extra scrutiny these agents have received, the economic cost–consequence table shows that only glucosamine sulphate is potentially cost-effective out of the interventions considered in this section.	Study of short- and long-term effects in the very elderly Benefits of individual and combination OA therapies in people with multiple joint region pain Identifying subsets of people with OA in whom existing treatments are more beneficial and cost-effective
	A wide range of ICERs were reported and the poorest estimates of efficacy would take it beyond the threshold of affordability in the NHS. Because only one glucosamine hydrochloride product is licensed, it would not be cost-effective to prescribe glucosamine on the NHS Many people with OA take over-the-counter nutraceutical products and may benefit from clear, evidence-based information. In particular, the guideline development group felt that it would be beneficial to advise people who wanted to trial over-the-counter glucosamine that the only potential benefits identified in early research are related purely to a reduction of pain (to some people, and to only mild or modest degree) with glucosamine sulphate 1500 mg daily. They could also benefit from advice on how to perform their own trial of therapy, that is, to evaluate their pain before starting glucosamine and ensure they review the benefits of glucosamine after 3 months
	Structure: In assessing the outcomes given in the evidence base, the guideline development group regarded measurement of joint space narrowing as of questionable value in assessing any potential beneficial structural modification, and convincing evidence of improvement in patient-centred outcomes consequent on any structural modification is still lacking. There is therefore no positive recommendation regarding structure modification Overall recommendation: The use of glucosamine or chondroitin products is not recommended for the treatment of OA
Richy 200375	Pain/function: Comparable symptomatic efficacies (Lequesne index, VAS pain and mobility) of chondroitin and glucosamine were shown	Long-term studies are needed to confirm and evaluate the structural efficacy of chondroitin For glucosamine, further studies on the relationship between structural and symptomatic changes, controlling for baseline characteristics including OA stage, are needed, as well as studies on the possible use of glucosamine in prevention	It is important to note that rescue medications were allowed in all trials; however, cumulative doses were low and it might be unlikely that rescue medication affected the pain-relieving effect of glucosamine and chondroitin

Glucosamine

Structure

For glucosamine, only two trials were included in the analysis of structural outcomes in all the three relevant systematic reviews (including the Richy review where only glucosamine trials were included in the analysis of structural outcomes), namely the trials by Reginster76 and Pavelká. 77 Both the Poolsup73 and Towheed62 reviews concluded that glucosamine may delay the structural progression of OA of the knee.

Symptoms/pain

While Poolsup73 concluded that their results showed evidence in support of an improvement of symptoms with glucosamine sulphate, Towheed62 was more cautious. They stated that overall, glucosamine was shown to have a moderate clinically significant effect on pain compared with placebo. With respect to function, results were mixed, with significant results for the Lequesne index, but not for the WOMAC total score or any of the WOMAC subscores; in their subgroup analyses, results were only significant for pain and the WOMAC total score for studies using the Rotta preparation of glucosamine, while they were non-significant for trials using a non-Rotta preparation. Poolsup73 only considered long-term studies (whereas Towheed62 included mainly short-term studies) and commented that an effect on physical function may only be achieved with long-term use of glucosamine.

Adverse events

Both reviews concluded that glucosamine was safe. Towheed62 did, however, mention the concern and uncertainty with respect to a possible effect of glucosamine on glucose metabolism; while not including this particular question in their review, they did quote two studies specifically addressing this question that found no evidence of an adverse metabolic effect of glucosamine.

Chondroitin

Structure

Both the review by Bana74 and the review by Reichenbach61 came to similar conclusions with respect to the structural benefits of chondroitin sulphate. Both suggested a small effect on joint space width in favour of chondroitin, with uncertain clinical significance (and possibly in the absence of a functional benefit). They considered the same two trials of more than one year follow-up (with the addition of one further shorter study included in the Reichenbach review61).

Symptoms/pain

Both Bana74 and Reichenbach61 were cautious in their interpretation of the data with respect to benefits of chondroitin sulphate on joint pain and function. Bana74 concluded that there was a true, but modest, effect of pain relief and improvement of function but also concluded that the results were probably biased owing to the poor quality of the data. They pointed out that pain evaluation was complicated in most studies as concomitant analgesic or NSAID use was allowed, but actual consumption was often not reported in detail. They also observed a latency period of about a month before an advantage of chondroitin with respect to pain relief and function was seen, as well as an effect lasting beyond the cessation of treatment, which they thought might justify cycles of treatment to be used.

Reichenbach61 stressed the high degree of heterogeneity between studies and the poor quality of many studies and concluded that large-scale methodologically sound trials indicated that the symptomatic effect of chondroitin sulphate was minimal or non-existent. They suggested that if there was an effect, chondroitin sulphate was more likely to benefit patients with low-grade OA than patients with advanced OA. They also mentioned the problem of concomitant use of rescue analgesia, but observed that in trials where use of rescue analgesia was reported to be greater in the placebo group, the effect of chondroitin sulphate was actually reported to be larger than in trials reporting equal use in both comparison groups.

Adverse events

Both reviews concluded that chondroitin sulphate was safe.

Glucosamine and/or chondroitin

Analysing trials of glucosamine or chondroitin versus placebo together in one meta-analysis, Richy75 found glucosamine/chondroitin to be superior to placebo for all the outcomes investigated. The authors concluded that glucosamine sulphate or chondroitin sulphate had a comparable benefit with respect to symptomatic outcomes (Lequesne index, VAS pain and mobility). Authors reported that adequate structural data were only available for glucosamine, and a statistically significant benefit in the low to medium range was reported. Exclusion of structural data from certain chondroitin studies was a result of inadequately detailed reporting. The authors mentioned the concomitant use of rescue medication, but concluded that cumulative doses were low and that rescue medication was unlikely to have affected the pain-relieving effect of glucosamine and chondroitin.

The NCCCC OA guideline1 did not recommend the use of glucosamine or chondroitin products for use in OA. They concluded that the evidence from the trials was often difficult to compare owing to differences in the products used (and their bioavailability), between-study populations, patient BMI, and the use of analgesia at the time of pain and function assessment in the trials. They found that, overall, trials using glucosamine sulphate as a single dose of 1500 mg rather than hydrochloride at 500 mg three times a day showed a small benefit compared with placebo for treatment of knee OA. However, at the time the guideline was prepared, the hydrochloride preparation had been granted a European Medicines Evaluatory Agency licence, while the sulphate had not. The evidence for efficacy of chondroitin was less convincing.

Their results suggested that the evidence to support the efficacy of glucosamine hydrochloride as a symptom modifier was poor, and for the non-licensed product (glucosamine sulphate), the evidence was not strong enough to warrant recommending that it should be prescribed on the NHS. Notwithstanding some evidence of benefit and very little evidence of harm in clinical practice, and despite the extra scrutiny these agents had received, their economic cost–consequence table showed that only glucosamine sulphate was potentially cost-effective out of the interventions considered in this section; however, a wide range of incremental cost-effectiveness ratios (ICERs) was reported and the poorest estimates of efficacy would take it beyond the threshold of affordability in the NHS. Because only a glucosamine hydrochloride product was licensed, it would not be cost-effective to prescribe glucosamine on the NHS.

The guideline, however, also stated that many people with OA take over-the-counter nutraceutical products and may benefit from clear, evidence-based information. In particular, the guideline development group felt that it would be beneficial to advise people who wanted to trial over-the-counter glucosamine that the only potential benefits identified in early research were related purely to a reduction of pain (to some people, and to only a mild or modest degree) with glucosamine sulphate 1500 mg daily. They could also benefit from advice on how to perform their own trial of therapy, that is, to evaluate their pain before starting glucosamine and ensure they review the benefits of glucosamine after 3 months.

With respect to structural outcomes, the guideline development group regarded measurement of joint space narrowing as of questionable value in assessing any potential beneficial structural modification, and convincing evidence of improvement in patient-centred outcomes consequent on any structural modification was still lacking. There was therefore no positive recommendation regarding structure modification.

Research recommendations and conclusions from the systematic reviews

Table 8 summarises the researchers’ conclusions and recommendations from the reviews and guideline. The conclusions about clinical effectiveness of glucosamine and chondroitin were not consistent. Authors noted that, at best, a modest effect on pain and function was observed. The effect on joint space was more consistent, but again the effect size was considered to be small and the long-term significance was hampered by the lack of long-term follow-up studies. The variability in quality of some of the trials, in particular for chondroitin, was noted, as was the variability in the preparations used.

For the purposes of our research question about the effectiveness in modifying long-term disease progression, the reviews we identified, with the exception of Poolsup,73 included only a very small proportion of trials of adequate follow-up. While conclusions regarding structural changes were, in three of the six reviews and guidelines, based on studies of at least 1 year, none presented their data for long-term QoL outcomes separately. In addition, the reporting of data extracted from primary studies varied in the reviews with some inconsistencies. We therefore went on to review RCTs of at least 12 months’ duration or follow-up and the results are reported below.

Review of randomised controlled trials

Search results

The range of trials included in the reviews presented above varied widely and it proved difficult to apply the findings to assessing the effectiveness of glucosamine and chondroitin on long-term outcomes such as disease progression. In addition, the reporting of the reviews on primary studies was not always consistent as not all reviews described all the details about the primary studies. We have, therefore, reviewed separately all the primary studies fulfilling our inclusion criteria (both the ones included in the reviews and new ones identified in our searches and not included in the reviews).

Within the reviews included above, there were only two glucosamine RCTs fulfilling the inclusion criteria of our review,76,77 three chondroitin trials108,114,116 (an earlier publication, Conrozier 1998,121 is included in the Reichenbach review61), and one combination therapy trial. 120

Table 9 compares the characteristics of the primary studies included in the systematic reviews against the inclusion criteria of our review. The most frequent reason for not fulfilling the inclusion criteria of the present review was short study duration (18 glucosamine, 15 chondroitin and one combined glucosamine/chondroitin trial).

TABLE 9 - Randomised controlled trials (RCTs) included in reviews and identified in additional searches against inclusion criteria for this review (highlighted trials met all of the inclusion criteria for this review)

Ch, chondroitin; Gl, glucose.

RCT	Knee OA	Duration ≥ 1 year	Oral medication	Placebo control	Structural outcomes included	Full text publication
Glucosamine
From systematic reviews
Cibere 200478	✓	No	✓	✓	No	✓
Clegg 2006117	✓	No	✓	✓	No	✓
Crolle 198079	?	No	No	No	No	✓
D’Ambrosio 198180	Various sites	No	No	No	No	✓
Drovanti 198081	Various sites	No	✓	✓	No	✓
Herrero-Beaumont 200782	✓	No	✓	✓	No	✓
Houpt 199983	✓	No	✓	✓	No	✓
Hughes 200284	✓	No	✓	✓	No	✓
McAlindon 200485	✓	No	✓	✓	No	✓
Müller-Fassbender 199486	✓	No	✓	No	No
Noack 199487	✓	No	✓	✓	No	✓
Pavelká 200277	✓	✓	✓	✓	✓	✓
Pujalte 198088	✓	No	✓	✓	✓	✓
Qiu 199889	✓	No	✓	No	No	✓
Reginster 200176	✓	✓	✓	✓	✓	✓
Reichelt 199490	✓	No	No		No	✓
Rindone 200091	✓	No	✓	✓	No	✓
Rovati 199792	✓	No	✓	✓	No	?
Usha 200493	✓	No	✓	✓	No	✓
Vajaradul 198194	✓	No	No	✓	No	✓
Vaz 198295	✓	No	✓	No	No	✓
Zenk 200296	✓	No	✓	✓	No	✓
Additional searches
Braham 2003125	?	No	✓	✓	No	✓
Frestedt 2008126	✓	No	✓	✓	No	✓
Mehta 2007127	✓	No	✓	No	No	✓
Zhang 2007128	✓	No	✓	No	No	✓
Chondroitin
From systematic reviews
Alekseeva 199997		No	✓	✓	No	✓
Bourgeois 199898	✓	No	✓	✓	No	✓
Bucsi 199899	✓	No	✓	✓	No	✓
Conrozier 1992100	No	No	✓	✓	No	✓
Fleisch 1997101	✓	✓	✓	✓	No	No
Kahan 2007102	✓	✓	✓	✓	✓	No
Kerzberg 1987103	✓	No	No	✓	No	✓
L’Hirondel 1992104	✓	No	✓	✓	No	✓
Mazieres 1992105	✓	No	✓	✓	No	✓
Mazieres 2001106	✓	No	✓	✓	No	✓
Mazieres 2007107	✓	No	✓	✓	No	✓
Michel 2005108	✓	✓	✓	✓	✓	✓
Morreale 1996109	✓	No	✓	✓	No	✓
Nasonova 2001110	✓	No	✓	✓	No	✓
Pavelká 1999111	✓	No	✓	✓	No	✓
Rovetta 1991112	✓	No	No		No	✓
Soroka 2002113	✓	No	✓	✓	No	No
Uebelhart 1998114/Conrozier 1998121	✓	✓	✓	✓	✓	✓
Uebelhart 1999115	✓	No	✓	✓	No	No
Uebelhart 2004116	✓	✓	✓	✓	✓	✓
Glucosamine/chondroitin
From systematic reviews
Clegg 2006117	✓	No	✓	✓	No	✓
Cohen 2003118	✓	No	No		No	✓
Das 2000119	✓	No	✓	✓	✓	✓
Rai 2004120	✓	✓	✓	✓	✓	✓
Additional searches
Messier 2007122	✓	✓	✓	✓	No	✓
Kawasaki 2008123	✓	✓	✓	✓	✓	✓
Debi 2000129	✓	No	No	✓	No	✓
Leffler 1999130	✓	?	?, not only Gl/Ch	✓	No	✓
Nguyen 2001131	No	No	✓	✓	No	✓

In addition, Table 9 includes the details of studies identified in our additional searches. Through examination of the excluded reviews and through supplemental searches, we identified four further RCTs studying glucosamine, and four studying a combination of glucosamine and chondroitin. However, only one trial fulfilled our inclusion criteria,122 reporting only on symptoms, but not on structure. A fifth trial was identified through a PubMed auto-alert in June 2008. Kawasaki123 reported both on symptoms and on structure. For details of reasons for trial exclusion, see Table 9.

One study excluded from our review is worthy of noting here because of its high profile: the GAIT (Glucosamine/chondroitin Arthritis Intervention Trial) study, funded by the National Institute of Health (NIH) and conducted in multiple centres across the USA. 117 The original study was an RCT of 24 weeks’ duration and was thus excluded from our review. 117 Recently (October 2008) a further report was published with 24 months of follow-up. 124 This report is of a subset of participants in the original trial who met certain criteria and who accepted an invitation to participate in the longer follow-up. Inclusion in the follow-up study required that patients had appropriate baseline imaging and continued on treatment and participating in follow-up at least until repeat imaging was done at 12 months. The authors describe the subsequent study as a prospective observational study. 124 Substantial numbers of patients who were eligible for this extension study were not included in the analysis; the most common reason was withdrawal prior to completing 12 months of follow-up. For example, in the glucosamine arm of the study, 134 patients were eligible, but only 77 were analysed, with 33 withdrawing from the study. The withdrawals were not described. The resulting study is at high risk of biases around participation, where patients who dropped out were likely to differ, in a non-random way, from those who continued. The baseline characteristics of patients participating in the extension study differed between groups with regard to sex, symptoms, radiological grade and BMI.

The eight studies fulfilling our inclusion criteria are highlighted in grey in Table 9.

The eight trials included did not allow a detailed comparison of different glucosamine preparations (sulphate versus hydrochloride or Rotta versus non-Rotta preparations). Similarly, none of the trials included a comparison between glucosamine and chondroitin, or a comparison between glucosamine and chondroitin combination therapy with either glucosamine or chondroitin monotherapy. Likewise, none of the trials compared different doses of the treatment.

Table 10 summarises the characteristics of the included trials. The results for glucosamine and chondroitin are presented separately. Appendix 5 provides all of the data in a single summary table for direct comparison between treatments.

TABLE 10 - Characteristics of randomised controlled trials fulfilling the inclusion criteria of this review

ACR, American College of Rheumatology; BMI, body mass index; K–L, Kellgren and Lawrence scores; NSAID(s), non-steroidal anti-inflammatory drug(s); NTX, terminal type I collagen telopeptides; VAS, visual analogue scale.

Study	Design	Participants	Interventions	Outcome measures	Quality
Glucosamine
Kawasaki 2008123 Japan	Focus: effects of glucosamine, risedronate or no treatment in patients with mild to moderate knee OA concurrently performing therapeutic exercise Single centre Duration: 18 months Follow-up: no post-intervention follow-up Setting: outpatient Funding: not reported	Total number: 142 Glucosamine: n = 49; 31 completed trial Risedronate: n = 51; 33 completed trial (not considered here) Control: n = 42; 30 completed trial Inclusion criteria: postmenopausal women with untreated OA of medial femorotibial compartment of knee and no other inflammatory diseases; selected according to criteria of ACR; joint space width ≥ 1 mm or and < 6 mm (equivalent to K–L grade II or III) Exclusion criteria: standard exclusion criteria (references given) Age: glucosamine: 68.5 years, SD 7.3; control: 69.5 years, SD 7.1 Gender: all female BMI: glucosamine: 23.9 kg/m2, SD 2.5; control: 24.0 kg/m2, SD 3.0 OA stage: not reported OA duration: not reported Comorbidities: not reported Subgroups: effects of age, BMI, level of joint progression, urine NTX, concomitant use of glucosamine or risedronate with home exercise at baseline (but not for intervention groups separately)	Glucosamine: 1500 mg/day glucosamine hydrochloride Risedronate: 2.5 mg/day risedronate (not considered here) Control: no supplement All groups: instructed by physical therapists to perform home exercise; instruction also given in a brochure; exercises as follows: 1. isometric muscle exercises of the lower limbs, 2. range of motion exercise after the knee was warmed (bath or shower) Co-interventions/rescue analgesia: sodium loxoprofen prescribed in advance for rescue and any use of the medicine recorded Rescue analgesia use: no significant difference between groups Adherence assessment: not reported	Primary: unclear Pain: WOMAC pain subscale, VAS pain scale WOMAC: reported Function: Japan Orthopaedic Association (JOA) score (pain during walking, pain when using stairs, range of motion, swelling, 100 points = full score); WOMAC Joint space loss: tibiofemoral joint space width (determined at the centre point of the medial femoral condyle) Adverse events: no Other: NSAID consumption, urinary NTX (uNTX, bone absorption marker) Assessment: regular assessment every 3 months	Method of randomisation: method not described Allocation concealed? Unclear Participants blinded? No Outcome assessors blinded? No; assessment of joint space width blinded Proportion of participants excluded/lost to follow-up: glucosamine: 35% withdrawals; C: 29% withdrawals; no significant difference in rates or reasons Handling of missing data: last observation carried forward Intention-to-treat analysis performed? Yes Statistical analysis appropriate? Yes Groups comparable at baseline? Yes
Pavelká 200277 Czech Republic	Focus: long-term effects of glucosamine sulphate on progression of joint structure and symptom changes in knee OA Single centre Duration: 3 years Follow-up: no post-intervention follow-up Setting: outpatient Funding: Rotta Research Group (Italy)	Total number: 202 Intervention: n = 101; 66 completed trial Control: n = 101; 55 completed trial Inclusion criteria: 45–70 years, primary knee OA based on clinical and radiological criteria of ACR; Lequesne index at least 4 points, but not > 12 Exclusion criteria: history of other clinically significant articular or rheumatic diseases, inflammatory rheumatic disease, secondary OA, history of trauma or lesions of knee joint, severe articular inflammation, evidence of rapidly progressing OA (obtained before the trial); BMI > 27 kg/m2, clinically significant alterations in haematological variables and renal, hepatic and metabolic functions; systemic or intra-articular corticosteroid therapy in previous 3 months Age: intervention: 61.2 years, SD 7.3; control: 63.5 years, SD 6.9 Gender: intervention: 79% female; control: 76% female BMI: intervention: 25.7 kg/m2, SD 2.1; control: 25.7 kg/m2, SD 1.8 OA stage: intervention: 54% K–L grade 2, 46% grade 3; control: 53% K–L grade 2, 47% grade 3 OA duration: intervention: 10.1 years, SD 8.1; control: 11.0 years, SD 6.8 Comorbidities: not reported Subgroups: none	Intervention: 1500 mg/day glucosamine sulphate (powder for oral solution) (Rotta) Control: placebo packets identical in external appearance Co-interventions/rescue analgesia: paracetamol 500 mg tablets, use recorded in patient diaries; no other pharmacologic treatments for OA allowed; only hydrotherapy, exercise or ultrasound allowed as physical therapies (no significant difference between groups) Rescue analgesia use: no significant difference between groups Adherence assessment: at least 86% of patients reported more than 90% drug intake	Primary: changes of minimum joint space width Pain: WOMAC joint pain subscale WOMAC: reported Function: WOMAC function and stiffness subscales; Lequesne index Joint space loss: minimum joint space width, joint space narrowing > 5 mm, other structural modifications Adverse events: withdrawal rates and reason for drop-out, adverse events, routine laboratory tests Other: paracetamol consumption Assessment: quarterly assessment of symptoms; radiographs and safety laboratory tests at end of each year Note: in patients with bilateral disease, the target knee was the one more severely affected	Method of randomisation: described, adequate Allocation concealed? Yes Participants blinded? Yes Outcome assessors blinded? Yes Proportion of participants excluded/lost to follow-up: intervention: 35% withdrawals; control: 46% withdrawals; no significant difference in rates or reasons Handling of missing data: worst case analysis and random sampling analysis Intention-to-treat analysis performed? Yes Statistical analysis appropriate? Yes Groups comparable at baseline? Yes
Reginster 200176 Belgium	Focus: effect of glucosamine sulphate on long-term progression of OA joint structure changes and symptoms Single centre Duration: 3 years Follow-up: no post-intervention follow-up Setting: outpatient Funding: Rotta Research Group (Italy)	Total number: 212 Intervention: n = 106; 68 completed trial Control: n = 106; 71 completed trial Inclusion criteria: > 50 years, primary knee OA of medial femorotibial compartment, diagnosed according to clinical and radiological criteria of ACR Exclusion criteria: history or active presence of rheumatic diseases that could be cause of secondary knee OA; severe articular inflammation; traumatic knee lesions; BMI > 30 kg/m2; substantial abnormalities in haematological, hepatic, renal or metabolic functions; intra-articular or systematic corticosteroids within 3 months preceding enrolment Age: intervention: 66.0 years, SD 8.1; control: 65.5 years, SD 7.5 Gender: intervention: 75% female; control: 78% female BMI: intervention: 27.3 kg/m2, SD 2.6; control: 27.4 kg/m2, SD 2.7 OA stage: intervention: 71% K–L grade 2, 29% grade 3; control: 70% K–L grade 2, 30% grade 3 OA duration: intervention: 8.0 years, SD 7.5; control: 7.6 years, SD 7.5 Comorbidities: not reported Subgroups: highest vs lowest quartile for joint space width (see Bruyere 2003213)	Intervention: 1500 mg/day glucosamine sulphate (powder for oral solution) (Rotta) Control: placebo (no details) Co-interventions/rescue analgesia: paracetamol (500 mg tablets) or NSAIDs (diclofenac 50 mg tablets, piroxicam 20 mg capsules or proglumetacin 150 mg tablets); washout before symptom assessment (at least five half-lives of selected medication allowed) Rescue analgesia use: no significant difference in consumption of rescue medication between groups Adherence assessment: over 70% drug intake in 81–91% without significant difference between groups	Primary: mean joint space width of medial compartment of tibiofemoral joint, total WOMAC score Pain: WOMAC joint pain subscale WOMAC: reported Function: WOMAC function and stiffness subscales Joint space loss: mean, minimum joint space width Adverse events: adverse events, routine laboratory tests, fasting glucose Other: use of rescue medication Assessment: radiographs taken at baseline, 1 and 3 years	Method of randomisation: described, adequate Allocation concealed? Yes Participants blinded? Yes Outcome assessors blinded? Yes Proportion of participants excluded/lost to follow-up: intervention: 33% withdrawals; control: 36% withdrawals; no significant difference in rates or reasons Handling of missing data: 3 approaches – (i) worst scenario analysis, (ii) last observation carried forward, (iii) random sampling approach Intention-to-treat analysis performed? Yes Statistical analysis appropriate? Yes Groups comparable at baseline? Yes
Chondroitin
Michel 2005108 Switzerland	Focus: effect of chondroitin sulphate on cartilage loss in knee OA Single centre Duration: 2 years Follow-up: no post-intervention follow-up Setting: outpatient clinic Funding: unclear	Total number: 300 Intervention: n = 150; 110 completed trial Control: n = 150; 109 completed trial Inclusion criteria: 40–85 years, clinically symptomatic knee OA (knee pain while standing, walking, and/or motion for at least 25 of the 30 days prior to study entry, no required minimum level of pain on day of entry); diagnosed according to ACR clinical and radiographic criteria for knee OA; patients with K–L grade 1, 2, 3; target knee defined as the most symptomatic knee on study entrance Exclusion criteria: presence of any causes of secondary knee OA, traumatic knee lesions, severe comorbidity (especially severe heart, renal, lung or neurological disease), previous joint surgery, intra-articular medications including corticosteroids in past month, foreseeable prospect of major surgery during 2-year study period; washout period of 3 months required for potentially longer-acting substances, e.g. glucosamine and chondroitin sulphate Age (mean): intervention: 62.5 years, SD 9.1; control: 63.1 years, SD 10.7 Gender: intervention: 51% female; control: 52% female BMI: intervention: 27.7 kg/m2, SD 5.2; control: 28.1 kg/m2, SD 5.5 OA stage/severity: not reported (stated that K–L scores were not significantly different between groups)	Intervention: 800 mg tablet Condrosulf® (IBSA, Hungary) (chondroitins 4- and 6-sulphate) daily Control: identical placebo tablet Co-interventions/rescue analgesia: paracetamol 500 mg tablets at maximum dose of 3 g/day allowed as rescue analgesia; for secondary rescue, NSAIDs were allowed up to a maximum period of 5 consecutive days if primary rescue analgesia with paracetamol was insufficient; physical therapy limited to application of warmth and strengthening exercises when deemed necessary by patient; no other interventions allowed, including steroid injections Rescue analgesia use: similar amounts taken, no details Adherence assessment: pill counts and 12- and 24-month visits; 69% in chondroitin group and 72% in placebo group took more than 70% of tablets during trial (no significant difference between groups	Primary: minimum and mean joint space width of more severely affected compartment of target knee Pain: WOMAC joint pain subscale WOMAC: reported Function: WOMAC function subscale Joint space loss: joint space narrowing in mm; minimum and mean joint space width measured using digitised radiographs and an image analysis system Adverse events: reported Other: none Assessment: follow-up assessments conducted by mail every 3 months over 2-year study period (WOMAC and treatment diaries); radiographs at baseline and 24 months; clinical examination and routine laboratory tests at baseline, 12 and 24 months	Method of randomisation: described, adequate Allocation concealed? Yes Participants blinded? Yes Outcome assessors blinded? Yes Proportion of participants excluded/lost to follow-up: 27% in each study group; no significant difference in reasons for withdrawal Handling of missing data: data at last visit used; radiographic evaluation done at time of drop-out of patients who dropped out (except for 16 patients who dropped out within 1 month of study entry) Intention-to-treat analysis performed? Yes Statistical analysis appropriate? No measures of variability reported Groups comparable at baseline? Yes
		OA duration: not reported Comorbidities: not reported Subgroups: for joint space changes, all patients (n = 300) versus patients with minimum joint space ≥ 1 mm at entry (n = 225)
Uebelhart 1998114 Switzerland	Focus: clinical, radiological, biological efficacy and tolerability of chondroitin 4- and 6-sulphate in knee OA Single centre Duration: 1 year Follow-up: no post-intervention follow-up Setting: in- or outpatient Funding: IBSA (producer)	Total number: 46 Intervention: n =  23; 21 completed trial Control: n =  23; 21 completed trial Inclusion criteria: 35–78 years, mono- or bilateral clinically symptomatic knee OA with a minimum remaining femorotibial joint space of 25% on standard X-ray Exclusion criteria: patients with inflammatory or systemic diseases or other conditions involving joints, lower limbs axial deviation > 5°, steroids or bone-oriented treatments < 3 months before study began Age: intervention: 60 years, SD 13; control: 57 years, SD 11 Gender: intervention: 48% female; control: 56% female Weight: intervention: 72 kg, SD 11; control: 76 kg, SD 14 OA stage: intervention: 43% K–L grade 1, 48% grade 2, 9% grade 3; control: 48% K–L grade 1, 43% grade 2, 9% grade 3 OA duration: not reported Comorbidities: not reported Subgroups: none	Intervention: 800 mg/day chondroitin 4- and 6-sulphate (Condrosulf, two sachets of 400 mg, dissolved in half a glass of water) Control: same daily amount of vehicle alone Co-interventions/rescue analgesia: free access to paracetamol as rescue medication Rescue analgesia use: analgesic consumption not reported	Primary: spontaneous joint pain (Huskisson VAS), overall mobility capacity (VAS) Pain: Huskisson VAS WOMAC: no Function: mobility capacity (VAS) Joint space loss: joint space measurement Adverse events: tolerability Other: bone joint metabolism Assessment: at entry and 3, 6, and 12 months; X-rays at entry and 12 months Note: only the most damaged knee was taken into account for inclusion and follow-up	Method of randomisation: method not described Allocation concealed? Unclear Participants blinded? Yes Outcome assessors blinded? Yes for joint space measurements; other measurements unclear Proportion of participants excluded/lost to follow-up: 8.7% in both comparison groups; chondroitin group: 1 death, 1 left the country; placebo group: 1 left the country, 1 not satisfied Handling of missing data: not described Intention-to-treat analysis performed? Unclear (probably not) Statistical analysis appropriate? Yes Groups comparable at baseline? Yes
Uebelhart 2004116 Switzerland/Belgium/France/Italy	Focus: efficacy of a 3-month, twice a year intermittent treatment with chondroitin sulphate in knee OA Multicentre Duration: 1 year Follow-up: no post-intervention follow-up Setting: not reported Funding: IBSA (producer)	Total number: 110 Intervention: n = 54; 43 completed trial Control: n = 56; 41 completed trial Inclusion criteria: age 40 or over, mono- or bilateral clinically symptomatic idiopathic knee OA according to ACR criteria; patients with K–L score 1–3, with a minimum 25% remaining medial femorotibial joint space upon entry Exclusion criteria: other inflammatory joint diseases or other systemic conditions affecting or involving joints; neoplasias, bone metabolic diseases and/or other metabolic and systemic diseases; treatments such as intra-articular steroids, NSAIDs, symptom-modifying agents or bone-oriented therapies (e.g. fluorides, bisphosphonates, calcitonin, hormonal substitution within 3 months before beginning of study Age: intervention: 63.2 years, SD 9.1; control: 63.7 years, SD 8.1 Gender: intervention: 80% female; control: 82% female Weight: intervention: 76.8 kg, SD 15.8; control: 76.4 kg, SD 13.8 OA stage: intervention: 13% K–L grade 1, 59% grade 2, 28% grade 3; control: 11% K–L grade 1, 59% grade 2, 30% grade 3 OA duration: intervention: 50.1 months; control: 52.4 months Comorbidities: not reported Subgroups: none	Intervention: 800 mg/day chondroitin 4- and 6-sulphate (Condrosulf sachets taken with a glass of water); treatment administered intermittently from entry to month 3, and between months 6 and 9 Control: identical sachets with vehicle only Co-interventions/rescue analgesia: medications as listed in exclusion criteria not allowed; paracetamol up to 4 g/day allowed; paracetamol treatment stopped 24 hours prior to each visit Rescue analgesia use: after first month, paracetamol consumption significantly greater in placebo group (at 12 months 55.5 tablets, SD 68.1 versus 25.8, SD 37.0 in placebo group) Adherence assessment: adherence 93–98%, no significant difference between groups	Primary: Lequesne index Pain: spontaneous joint pain (Huskisson VAS) WOMAC: no Function: 20 m walking time Joint space loss: joint space surface area, mean and minimum joint space width Adverse events: adverse events, clinical laboratory evaluation Other: global patient and physician assessment; paracetamol consumption Note: in patients with bilateral knee OA, the most symptomatic knee was selected as the study target knee Assessment: every 3 months; radiographs at entry and 12 months	Method of randomisation: described, adequate Allocation concealed? Yes Participants blinded? Yes Outcome assessors blinded? Yes Proportion of participants excluded/lost to follow-up: 10 patients (of 120) not included in intention-to-treat population (did not turn up for second visit, did not take any medication); 11 drop-outs in intervention and 15 in placebo group; no significant difference Handling of missing data: last observation carried forward Intention-to-treat analysis performed? Yes Statistical analysis appropriate? Yes Groups comparable at baseline? Yes
Glucosamine/chondroitin
Messier 2007122 USA	Focus: effect of glucosamine hydrochloride plus chondroitin sulphate alone and combined with exercise on physical function, pain, strength, balance and mobility in older adults with knee OA Single centre Duration: 1 year Follow-up: no post-intervention follow-up Setting: outpatient clinical research centre Funding: Rexall Sundown Inc.	Total number: 89 Intervention: n = 45; 37 completed trial Control: n = 44; 35 completed trial Inclusion criteria: ≥ 50 years, radiographic evidence of mild to moderate knee OA (K–L grades 2–3) who met ACR classification criteria Exclusion criteria: currently participating in another trial, dementia, active cancer other than skin cancer, anaemia, severe renal insufficiency, hepatic diseases, excess alcohol use, knee-joint replacement, inability to walk unassisted at least 128 m in 6 minutes, completion of more than 20 minutes of formal exercise per week during past 3 months, planned absence for longer than 3 weeks during required study visits, residing further than 1 hour drive from research facilities, shellfish allergy, exposure to glucosamine/chondroitin in 6 months prior to randomisation, unwilling to discontinue current arthritis medication during 2-week washout and 12-month intervention period (other than rescue medication), less than 80% compliance during run-in period, failure to complete graded exercise test, required human assistance with activities related to knee pain, recent knee surgeries or knee injections, e.g. cortisone and hyaluronic acid Age: intervention: 70.0 years, SE 1.28; control: 74.1 years, SE 1.32; p = 0.03 Gender: intervention: 75.6% female; control: 65.9% female	Intervention: 1500/1200 mg glucosamine hydrochloride plus chondroitin sulphate (GH/CS); choice of 1× or 3× a day regimen; 6 healthy lifestyle classes; after 6-month exercise programme (1 hour twice a week) added for last 6 months (combined facility and home-based exercise programme, two 15-minute walking sessions, separated by 20 minutes of strength training) Control: placebo of identical size, colour, shape, regimen; lifestyle classes and exercise programme after 6 months as in GH/CS group Co-interventions/rescue analgesia: paracetamol maximum dosage 4 g/day Rescue analgesia use: decrease in paracetamol use by 12 months of 37% in intervention and 11% in placebo group (not stated if significant) Run-in/washout period: 2 weeks before start of intervention; discontinuation of all medications; rescue medications/required medication unrelated to OA permitted; unblinded placebo supply (3 pills/day, compliance assessed) Adherence assessment: 91–97%	Primary: WOMAC function Pain: WOMAC pain subscale WOMAC: yes Function: WOMAC function subscale; mobility (distance walked in 6 minutes) Joint space loss: not assessed Adverse events: not systematically reported Other: knee concentric extension and flexion strength (Kin-Com 125E isokinetic dynamometer); balance (AMTI force platform) Assessment: screening visit, baseline, 6 and 12 months	Method of randomisation: not explicitly described, computer allocation implied Allocation concealed? Yes Participants blinded? Yes Outcome assessors blinded? Yes Proportion of participants excluded/lost to follow-up: intervention: 18% withdrawals; control: 20% withdrawals; not reported if significant differences between groups Handling of missing data: last observation carried forward Intention-to-treat analysis performed? Yes Statistical analysis appropriate? Yes Groups comparable at baseline? No, GH/CS were significantly younger, more overweight, and had a higher income than placebo group
		BMI: intervention: 30.7 kg/m2, SE 0.93; control: 27.3 kg/m2, SE 0.71; p = 0.005 OA stage/severity/duration: not reported Comorbidities: listed in detail, no significant difference between groups
Rai 2004120 India	Focus: effect of chondroitin plus glucosamine sulphate on joint structure and symptoms in knee OA Single centre Duration: 1 year Follow-up: no post-intervention follow-up Setting: orthopaedic outpatient centre Funding: unclear	Total number: 100 Intervention: n = 50 Control: n = 50 Inclusion criteria: > 50 years, primary knee OA mainly of medial femorotibial compartment diagnosed according to clinical and radiological criteria of ACR Exclusion criteria: history or active presence of other rheumatic diseases responsible for secondary OA, severe articular inflammation as confirmed by physical examination, K–L grade 4, Lequesne index higher than 12, history of intra-articular injection or systemic steroid intake in 3 months preceding enrolment Age: intervention: 54.7 years; control: 53.9 years Gender: not reported Weight: intervention: 69.1 kg; control: 68.9 kg OA stage/severity: no details, stated that OA was comparable at baseline and mild to moderate OA duration: not reported Comorbidities: not reported Subgroups: none	Intervention: oral combination of glucosamine sulphate and chondroitin sulphate (Kondro®, capsule) (Panacea Biotec Ltd, India) – dosing not stated, presumably one capsule a day; Kondro contains 250 mg glucosamine sulphate and 200 mg chondroitin sulphate Control: matching placebo capsules Co-interventions/rescue analgesia: patients advised to take paracetamol for pain Rescue analgesia use: differences between groups not reported	Primary: joint space width changes in narrowest medial compartment of tibiofemoral joint Pain: no separate assessment WOMAC: no Function: Lequesne index Joint space loss: minimum joint space width Adverse events: not reported Other: none Assessment: every 4 weeks; standardised radiographs at enrolment and after 1 year	Method of randomisation: not reported Allocation concealed? Unclear Participants blinded? Yes, described as double blind Outcome assessors blinded? Yes, described as double blind Proportion of participants excluded/lost to follow-up: not reported Handling of missing data: not reported Intention-to-treat analysis performed? Not reported Statistical analysis appropriate? Yes Groups comparable at baseline? Yes

Findings

Glucosamine

Trial characteristics

Three trials compared glucosamine with placebo. The trials by Pavelká77 and Reginster76 were both randomised, double-blind, placebo-controlled trials with a duration of 3 years. Follow-up data for knee arthroplasty for both trials taken together were available for a mean follow-up of 8 years (5 years after cessation of treatment52). Both were funded by the Rotta Research Group. The third trial, by Kawasaki123 was a Japanese open-label RCT with a duration of 18 months.

A total of 556 people participated in these trials. The mean age of participants was between 61 and 70 years and more women than men were included (between 75% and 100% women). Most had OA of Kellgren–Lawrence grade 2 at baseline (between 53% and 54% for Pavelká and between 70% and 71% for Reginster); the rest had OA of Kellgren–Lawrence grade 3 (no details given in the trial by Kawasaki which just stated that both Kellgren–Lawrence grades 2 and 3 were included). In all trials, baseline characteristics were well balanced in the comparison groups.

Interventions

In the Pavelká and Reginster trials, 1500 mg per day glucosamine sulphate (Rotta) was given as a powder oral solution. Kawasaki used 1500 mg per day glucosamine hydrochloride. All trials reported that there was no significant difference in rescue analgesia use between comparison groups and that medication adherence was good.

Outcome measures and further trial details are summarised in Table 10.

Trial quality

Details for trial quality are shown in Table 10. Trial quality for the trials by Pavelká and Reginster was good. The trial by Kawasaki was of poor quality. Randomisation and allocation concealment were not described and there was no blinding (with the exception of assessment of joint space width).

Results

Structure

Table 11 summarises the results for trials of glucosamine on joint structure.

TABLE 11 - Results for trials of glucosamine on joint structure

C, control; I, intervention; NNT, number needed to treat; NS, not significant.

Study	Outcome	No. of participants	Baseline	Change from baseline/difference between groups	p-value (between groups)
Kawasaki 2008123 (Hydrochloride)	Joint space width	I: n = 49 C: n = 42	I: 2.6 mm, SD 1.2 C: 3.1 mm, SD 1.0	Change from baseline: I: 0.0 mm (95% CI –2.2 to 1.7) C: –0.3 mm (95% CI –1.8 to 1.6) Difference: 0.2 mm	NS
Pavelká 200277 (Sulphate)	Minimum joint space width	I:n = 101 C:n = 101	I: 3.89 mm, SD 1.48 C: 3.63 mm, SD 1.57	Change from baseline: I: +0.04 mm (95% CI –0.06 to 0.14) C: –0.19 mm (95% CI –0.29 to –0.09) Difference: 0.23 mm (95% CI 0.09 to 0.37)	0.001
	Severe joint space narrowing (> 0.5 mm)			End of study: I: n = 5 C: n = 14 NNT = 11	0.05
Reginster 200176 (Sulphate)	Mean joint space width	I:n = 106 C:n = 106	Total joint space width: I: 5.23 mm, SD 1.36 C: 5.39 mm, SD 1.29	Change from baseline: I: –0.06 mm (95% CI –0.22 to 0.09) C: –0.31 mm (95% CI –0.48 to –0.13) Difference: 0.24 mm (95% CI 0.01 to 0.48)	0.043
	Minimum joint space width	I:n = 106 C:n = 106	I: 3.82, SD 1.32 mm C: 3.95, SD 1.24 mm	Change from baseline: I: –0.07 mm (95% CI –0.22 to 0.07) C: –0.40 mm (95% CI –0.56 to –0.24) Difference: 0.33 mm (95% CI 0.12 to 0.54)	0.003

The trials by Pavelká and Reginster showed a statistically significant effect in favour of glucosamine for minimum joint space width, with a weighted mean difference from baseline to end of study of 0.26 mm (95% CI 0.17 to 0.35), indicating less joint space loss with glucosamine (Figure 3). Both trials reported a point estimate in favour of glucosamine sulphate, with minimal evidence of heterogeneity in the size of the effect being observed. Pavelká found significantly more patients with severe joint space narrowing (more than 0.5 mm) in the placebo group than in the intervention group (n = 14 versus n = 5, p < 0.05).

FIGURE 3.

Glucosamine sulphate: a meta-analysis of minimum joint space width – difference from baseline (fixed-effects model produces same point estimate; random-effects model presented as confidence interval better reflects the uncertainty around that point estimate). IV, inverse difference.

Mean joint space width was reported by Reginster only, who observed a significant difference of 0.24 mm (95% CI 0.01 to 0.48) in mean joint space narrowing, indicating a preservation of joint space with glucosamine. Kawasaki reported results for joint space width (determined at the centre point of the medial femoral condyle) and found no significant effect of glucosamine hydrochloride versus control.

A subgroup analysis of the Reginster trial (Table 12) compared patients in the lowest quartile for mean joint space width (joint space width < 4.5 mm) with patients in the highest quartile for mean joint space width (joint space width > 6.2 mm). In the patients in the lowest quartile (n = 29 intervention, n = 23 control), mean joint space width increased by 0.22 mm (SD 0.66 mm) in the intervention group and 0.13 mm (SD 0.81 mm) in the control group (no statistically significant difference). In the patients in the highest quartile (n = 27 intervention, n = 26 control), mean joint space width decreased by 0.45 mm (SD 1.04 mm) in the intervention group and by 1.05 mm (SD 1.28 mm) in the control group (difference not statistically significant, p = 0.1). There was a statistically significant difference in joint space change between the two subgroups (p < 0.01). The authors speculated that the results may indicate that patients with less severe disease may benefit more from structure-modifying drugs, although improvements in the joint space width were also observed in the control group.

TABLE 12 - Subgroup analysis: summary from Reginster 2001 trial76

C, control; I, intervention.

Study	Subgroup	Outcome	Change from baseline/difference between groups	p-value
Reginster 200176	Mean joint space width lowest quartile (< 4.5 mm, n = 29 intervention, n = 23 control) vs Highest quartile (> 6.2 mm, n = 27 intervention, n = 26 control)	Change in mean joint space width	Joint space width < 4.5 mm: I: +6.2%, SD 17.5; +0.22 mm, SD 0.66 C: +3.8%, SD 23.8; +0.13 mm, SD 0.81 Joint space width > 6.2 mm: I: –6.0%, SD 15.1; –0.45 mm, SD 1.04 C: –14.9%, SD 17.9; –1.05 mm, SD 1.28	< 0.01 for joint space change between the two quartiles

Symptoms/pain

For function and pain assessment, the trials did not report data consistently (for example, Pavelká and Kawasaki reported WOMAC scores as points, whereas Reginster used per cent change on a VAS), only Pavelká used the Lequesne index, and Kawasaki used the Japan Orthopaedic Association (JOA) score and a visual analogue pain scale, so data could not be summarised in a meta-analysis (Table 13).

TABLE 13 - Summary of glucosamine trial results for pain, function and composite health state scores

C, control; I, intervention, NS, not significant.

Study	No. of participants	Outcome	Baseline	Change from baseline/difference between groups	p-value
Kawasaki 2008123	I: n = 49 C: n = 42	JOA total score (range 0–100)	I: 74.2, SD 13.3 C: 69.6, SD 14.4	I: +13.7 (95% CI –18.5 to 43) C: +16.3 (95% CI –4.8 to 39.8)	NS
		JOA gait (range 0–30)	I: 23.3, SD 5.5 C: 21.5 SD 6.6	I: +4.4 (95% CI 2.8 to 11) C: +6.3 (95% CI –3.6 to 20)	NS
		JOA stairs (range 0–25)	I: 14.7, SD 5.6 C: 11.9, SD 6.1	I: +5.5 (95% CI –8.2 to 20.3) C: +6.3 (95% CI –1.1 to 17.5)	NS
		JOA range of motion (range 0–35)	I: 29.4, SD 5.0 C: 29.5, SD 4.8	I: +1.4 (95% CI –5.5 to 5.5) C: +1.9 (95% CI –6.1 to 10)	NS
		JOA swelling (range 0–10)	I: 6.8, SD 2.7 C: 6.7, SD 3.3	I: +2.6 (95% CI –0.3 to 5) C: +1.9 (95% CI –5.0 to 10)	NS
		VAS pain (range 0–10)	I: 5.0, SD 2.0 C: 4.9, SD 1.8	I: –1.6 (95% CI –5.0 to 2.2) C: –1.7 (95% CI –5.5 to 2.5)	NS
		WOMAC total (range 0–96)	I: 31.5, SD 15.6 C: 28.7, SD 16.0	I: –14.9 (95% CI –45.5 to 6) C: –11.5 (95% CI –38.8 to 7.5)	NS
		WOMAC pain (range 0–20)	I: 7.2, SD 3.9 C: 6.5, SD 3.4	I: –3.6 (95% CI –14.1 to 3) C: –2.9 (95% CI –9.5 to 0.5)	NS
		WOMAC stiffness (range 0–8)	I: 3.4, SD 1.6 C: 2.7, SD 1.6	I: –1.5 (95% CI –5.1 to 1.0) C: –0.7 (95% CI –3 to 1.5)	NS
		WOMAC function (range 0–68)	I: 20.8, SD 11.0 C: 19.6, SD 11.5	I: –9.8 (95% CI –31.2 to 5.2) C: –7.5 (95% CI –27.2 to 7)	NS
Pavelká 200277	I: n = 101 C: n = 101	Lequesne index (points)	I: 8.95, SD 2.30 C: 8.94, SD 2.27	Change from baseline: I: –1.7 (95% CI –2.2 to –1.2) C: –0.82 (95% CI –1.1 to –0.51) Difference: 0.91 (95% CI 0.34 to 1.5)	0.002
		WOMAC total (points)	I: 30.70, SD 14.40 C: 30.48, SD 14.43	Change from baseline: I: –8.0 (95% CI –9.8 to –6.3) C: –4.9 (95% CI –6.5 to 3.2) Difference: 3.1 (95% CI 0.77 to 5.5)	0.01
		WOMAC pain (points)	I: 6.61, SD 3.45 C: 6.33, SD 3.13	Change from baseline: I: –2.0 (95% CI –2.4 to –1.5) C: –1.3 (95% CI –1.7 to 0.88) Difference: 0.7 (95% CI 0.06 to 1.3)	0.03
		WOMAC function (points)	I: 21.84, SD 10.67 C: 22.00, SD 11.03	Change from baseline: I: –5.8 (95% CI –7.1 to –4.4) C: –3.7 (95% CI –4.9 to –2.5) Difference: 2.1 (95% CI 0.28 to 3.9)	0.02
		WOMAC stiffness (points)	I: 2.25, SD 1.47 C: 2.15, SD 1.44	Change from baseline: I: –0.31 (95% CI –0.55 to 0.07) C: +0.11 (95% CI –0.12 to 0.34) Difference: 0.42 (95% CI 0.09 to 0.75)	0.01
Reginster 200176	I: n = 106 C: n = 106	WOMAC total (out of 2400mm)	I: 1030.2 mm, SD 473.8 C: 939.7 mm, SD 484.8	Change from baseline: I: –11.7% (95% CI –20.3 to –3.2) C: +9.8% (95% CI –6.2 to 25.8) Difference: 21.6% (95% CI 3.5 to 39.6)	0.020
		WOMAC pain	I: 194.1 mm, SD 101.9 C: 172.2 mm, SD 104.5	Change from baseline: I: ~–36.7 mm, SE 8.3 C: ~–7.5 mm, SE 10.6 (estimated from graph)	0.047
		WOMAC function	I: 740.1 mm, SD 364.2 C: 670.8 mm, SD 367.8	Change from baseline: I: ~–160 mm, SE 24.2 C: ~–65.5 mm, SE 31.5 (estimated from graph)	0.020
		WOMAC stiffness	I: 96.0 mm, SD 54.8 C: 96.7 mm, SD 54.6		NS

Pain

Pavelká found a statistically significant effect in favour of glucosamine compared with placebo at the end of the 3-year trial for the WOMAC pain subscale [difference between groups 0.7 points (95% CI 0.06 to 1.3)]. Similarly, Reginster found a statistically significant effect in favour of glucosamine for the WOMAC pain subscale [change from baseline approximately –36.7, SE 8.3 mm (VAS) for glucosamine and –7.5 SE, 10.6 mm for the control group (estimated from graph)]. Kawasaki did not find any statistically significant difference in the WOMAC pain subscale or the VAS pain scale between the glucosamine hydrochloride and control groups. Meta-analysis (Figure 4) illustrates the small, but statistically significant favourable effect for glucosamine.

FIGURE 4.

Glucosamine sulphate and hydrochloride: a meta-analysis of WOMAC pain scores at end of study (fixed-effects model produces same point estimate; random-effects model presented as confidence interval better reflects the uncertainty around that point estimate). IV, inverse difference.

Little detail was reported in the three trials about the use of rescue analgesia with the exception of stating in all three trials that there was no statistically significant difference in the volume of rescue analgesia used.

Function

Pavelká found a statistically significant effect in favour of glucosamine compared with placebo for both the WOMAC function and stiffness subscales (difference between groups 2.1 points (95% CI 0.28 to 3.9) for function and 0.42 points (95% CI 0.09 to 0.75) for stiffness). Reginster found statistically significant effects in favour of glucosamine for the WOMAC function subscale [change from baseline about –160 SE, 24.2 mm (VAS) for glucosamine and –65.5, SE 31.5 mm for the control group (estimated from graph)], but not for the stiffness subscale. Kawasaki did not find any statistically significant difference in the JOA total score or the gait, stairs, range of motion or swelling subscores, or in the WOMAC stiffness and function subscores between glucosamine hydrochloride and the control group. Meta-analysis (Figure 5) illustrates the small, but statistically significant favourable effect for glucosamine. While the CIs overlap, visual inspection identifies substantial heterogeneity, with the glucosamine hydrochloride product demonstrating little evidence of an effect.

FIGURE 5.

Glucosamine sulphate and hydrochloride: a meta-analysis of WOMAC function scores at end of study (fixed-effects model produces same point estimate; random-effects model presented as confidence interval better reflects the uncertainty around that point estimate). IV, inverse difference.

Function/pain composite scores

Pavelká found a statistically significant effect in favour of glucosamine compared with placebo for both the Lequesne index and the WOMAC total score (difference between groups 0.91 points (95% CI 0.34 to 1.5) for the Lequesne index and 3.1 (95% CI 0.77 to 5.5) for the WOMAC total score). Reginster found statistically significant effects in favour of glucosamine for the WOMAC total score [per cent change from baseline, difference between groups 21.6% (95% CI 3.5 to 39.6)]. Kawasaki did not find any significant difference in the WOMAC total score between glucosamine hydrochloride and the control group.

Knee arthroplasty

In a follow-up analysis52 of the trials by Reginster and Pavelká taken together, patients with at least 12 months of treatment were followed up for their joint replacement status. Of 340 eligible patients, 275 (81%) could be followed up. Of these, 144 had been in the glucosamine groups of the trials and 131 had been in the placebo groups. No significant baseline differences were found between the patients followed up and those lost to follow-up. After a mean post-intervention follow-up of 5 years (8 years from trial baseline), significantly fewer patients in the former glucosamine groups had received knee arthroplasty than patients in the former placebo groups (9/144 in the glucosamine groups and 19/131 in the placebo groups, RR 0.43 (95% CI 0.20 to 0.92, p = 0.024).

Adverse events

Neither Pavelká nor Reginster found significant differences in reported adverse effects between the glucosamine groups and the placebo groups. No significant differences in routine laboratory test results were seen, and Reginster did not note any change in glucose homeostasis. Adverse events were not reported systematically in the study by Kawasaki. One patient in the glucosamine group and two in the control group withdrew from the trial because of adverse events.

Chondroitin

Trial characteristics

Of the three chondroitin trials, all were described as double-blind, placebo-controlled RCTs. The trial by Michel108 had a duration of 2 years, whereas the two trials by Uebelhart114,116 had a duration of 1 year. The two trials by Uebelhart114,116 were industry funded. No funding information was available for the trial by Michel. 108

A total of 456 people participated in the three trials. The trials included participants of a mean age of between 57 and 64 years, with between 48% and 82% being female. Osteoarthritis severity was not reported in the trial by Michel; in the trials by Uebelhart, between 13% and 48% had OA of Kellgren–Lawrence grade 1 at baseline, between 43% and 59% had grade 2, and between 9% and 30% had grade 3 [with more patients in the Uebelhart (2004) study116 having more severe disease]. In all trials, baseline characteristics were well balanced in the comparison groups.

Interventions

In the trial by Michel, 800 mg/day chondroitins 4- and 6-sulphate [Condrosulf^® (IBSA, Hungary)] were given as a tablet. In the trials by Uebelhart (1998)114 and Uebelhart (2004),116 800 mg/day chondroitins 4- and 6-sulphate (Condrosulf) were given in powder form. In the Michel and Uebelhart (1998) trials the treatment was given continuously, whereas in the trial by Uebelhart (2004) treatment was administered intermittently for 3-month periods (from study entry to the third month, and from the sixth to the ninth months). Michel reported that analgesia use was similar between the comparison groups; no information was given in the trial by Uebelhart (1998), whereas in the trial by Uebelhart (2004) paracetamol consumption was significantly greater in the placebo group at 12 months than in the intervention group.

Outcome measures and further details of the trials are summarised in Table 10.

Trial quality

The Michel and Uebelhart (2004) trials were of good quality. The trial by Uebelhart (1998) was of poor quality, reporting little detail of the randomisation procedure and allocation concealment.

Results

Structure

Table 14 summarises the results for chondroitin on structural outcomes.

TABLE 14 - Summary of the effect of chondroitin on joint structure

C, control; I, intervention; NS, not significant.

Study	Outcome	Baseline	End of study	Change from baseline/difference between groups	p-value (between groups)
Michel 2005108	Minimum joint space width	I: 2.41 mm, SD 0.14 C: 2.35 mm, SD 0.14		Change from baseline: I: +0.045 mm, SD 0.48 (95% CI –0.03 to 0.12) C: –0.07 mm, SD 0.56 (95% CI –0.16 to 0.02) Difference: 0.12 mm, SD 0.52 (95% CI 0.00 to 0.24)	0.05
	Mean joint space width	I: 3.04 mm, SD 0.14 C: 3.00 mm, SD 0.15		Change from baseline: I: 0.0 mm, SD 0.53 (95% CI –0.08 to 0.09) C: –0.14 mm, SD 0.61 (95% CI –0.24 to –0.04) Difference: 0.14 mm, SD 0.57 (95% CI 0.01 to 0.27)	0.04
Uebelhart 1998114	Medial femorotibial joint minimum width	I: (n = 14) 0.34 cm, SD 0.10 C: (n = 12) 0.40 cm, SD 0.10	I: (n = 14) 0.35 cm, SD 0.10 C: (n = 12) 0.36 cm, SD 0.13		NS
	Medial femorotibial joint mean width	I: (n = 14) 0.44 cm, SD 0.11 C: (n = 12) 0.51 cm, SD 0.10	I: (n = 14) 0.44 cm, SD 0.10 C: (n = 12) 0.46 cm, SD 0.14		< 0.05
	Medial femorotibial joint surface area	I: (n = 14) 1.08 cm2, SD 0.32 C: (n = 12) 1.29 cm2, SD 0.28	I: (n = 14) 1.09 cm2, SD 0.33 C: (n = 12) 1.17 cm2, SD 0.36		NS
	Bone metabolism			Significant effect on bone metabolism found: osteocalcin, pyridinoline/creatinine and deoxypyridinoline/creatinine significantly higher, and keratin sulphate significantly lower than in placebo group, p = 0.001
Uebelhart 2004116	Joint surface area	I: 68.0 mm2, SD 27.2 C: 63.3 mm2, SD 24.4	I: 67.8 mm2, SD 26.9 C: 58.7 mm2, SD 20.9	Change from baseline: I: –0.19 mm2 (95% CI –3.56 to 3.17) C: –4.55 mm2 (95% CI –8.61 to –0.49) Difference: 4.36 mm2 (95% CI –0.19 to 8.91)	NS
	Minimum joint space width	I: 3.65 mm, SD 1.46 C: 3.54 mm, SD 1.39	I: 3.61 mm, SD 1.51 C: 3.23 mm, SD 1.27	Change from baseline: I: –0.04 mm (95% CI –0.23 to 0.14) C: –0.32 mm (95% CI –0.57 to –0.07) Difference: 0.27 mm (95% CI 0.004 to 0.55)	0.047
	Mean joint space width	I: 4.20 mm, SD 1.51 C: 4.03 mm, SD 1.47	I: 4.20 mm, SD 1.58 C: 3.74 mm, SD 1.28	Change from baseline: I: –0.006 mm (95% CI –0.20 to 0.18) C: –0.29 mm (95% CI –0.53 to –0.04) Difference: 0.28 mm (95% CI 0.01 to 0.55)	0.039

The three chondroitin trials showed a significant effect in favour of chondroitin for minimum joint space width, with the weighted mean difference changing from 0.02 mm (95% CI –0.20 to 0.24) at baseline to 0.18 mm (95% CI 0.07 to 0.30) at the end of the trials (p = 0.001, random-effects model), no statistically significant heterogeneity, but the estimate was driven by the larger trial by Michel (Figure 6). Similarly, a significant effect in favour of chondroitin was shown for mean joint space width, with the weighted mean difference changing from –0.01 mm (95% CI –0.30 to 0.28) at baseline to 0.20 mm (95% CI 0.07 to 0.32) at the end of the trials (p = 0.003, random-effects model) (Figure 7). These results indicate less joint space loss with chondroitin sulphate. Uebelhart (1998) also found a significant positive effect of chondroitin sulphate compared with placebo on some indicators of bone metabolism.

FIGURE 6.

Chondroitin sulphate: a meta analysis of minimum joint space width at end of study (fixed-effects model produces same point estimate; random-effects model presented as confidence interval better reflects the uncertainty around that point estimate). IV, inverse difference

FIGURE 7.

Chondroitin sulphate: a meta-analysis of minimum joint space width at end of study (fixed-effects model produces same point estimate; random-effects model presented as confidence interval better reflects the uncertainty around that point estimate). IV, inverse difference

Michel did a subgroup analysis of the patients with a minimum joint space width of 1 or more millimetres at entry (225/300 patients) and found no significant difference to their main results (Table 15). They also state that mean joint space width at baseline had no influence on radiographic progression.

TABLE 15 - Subgroup analysis: summary from Michel 2005108

C, control; I, intervention.

Study	Subgroup	Outcome	Change from baseline/difference between groups	p-value
Michel 2005108	Patients with minimum joint space width ≥ 1 mm at entry (I: n = 114, C: n = 111)	Minimum joint space width	Change from baseline: I: +0.05 mm, SD 0.53 (95% CI –0.05 to 0.14) C: –0.14 mm, SD 0.57 (95% CI –0.25 to –0.03) Difference: 0.19 mm, SD 0.55 (95% CI 0.04 to 0.33)	0.01
		Mean joint space width	Change from baseline: I: +0.01 mm, SD 0.54 (95% CI –0.09 to 0.11) C: –0.20 mm, SD 0.58 (95% CI –0.31 to –0.09) Difference: 0.21 mm, SD 0.56 (95% CI 0.06 to 0.36)	0.006

Symptoms/pain

Results for function and pain were very mixed and significant changes in joint space were not always mirrored by significant changes in pain/function (Table 16).

TABLE 16 - Summary of the effect of chondroitin on pain, function and composite measures of health status

C, control; I, intervention; NS, not significant; VAS, visual analogue scale.

Study	Outcome	Baseline	End of study	Change from baseline/difference between groups	p-value
Michel 2005108	WOMAC total	I: 2.3, SD 1.6 C: 2.6, SD 1.7		Change from baseline: I: –3.9% C: +2.1%	NS
	WOMAC pain	I: 2.5, SD 1.6 C: 2.7, SD 1.8		Change from baseline: I: –11.0% C: –6.2%	NS
	WOMAC function	I: 2.1, SD 1.6 C: 2.5, SD 1.8		Change from baseline: I: –7.8% C: –4.6%	NS
	WOMAC stiffness	I: 3.0, SD 2.3 C: 3.5, SD 2.5		Change from baseline: I: –0.8% C: +5.9%	NS
Uebelhart 1998114	Huskisson VAS (0–10 cm)	I: 5.76 cm, SD 1.61 C: 6.44 cm, SD 1.10	I: 2.1 cm, SD 2.1 C: 4.8 cm, SD 2.5		Between treatments p < 0.001, between times p < 0.05
	Overall mobility capacity (VAS)	I: 5.1 cm, SD 1.5 C: 5.7 cm, SD 1.6	I: 8.6 cm, SD 1.4 C: 6.8 cm, SD 2.3		Between treatments p = NS, between times p < 0.001
Uebelhart 2004116	Lequesne index	I: 9.0, SD 2.8 C: 9.1, SD 3.2	I: 5.8, SD 3.6 C: 7.0, SD 3.9	I: –36% C: –23%	< 0.01
	Huskisson VAS (0–100 mm)	I: 58.8 mm, SD 15.5 C: 61.1 mm, SD 19.0	I: 34.3 mm, SD 27.4 C: 45.8 mm, SD 27.6	I: –42% C: –25%	< 0.05
	Walking time	I: 24.5 seconds, SD 22.7 C: 22.8 seconds, SD 7.5	I: 20.1 seconds, SD 6.8 C: 22.7 seconds, SD 7.7	I: –18% C: –0.5%	< 0.05

Pain

Michel did not find any significant differences between chondroitin sulphate and placebo for the WOMAC pain subscore. The trials by Uebelhart (1998 and 2004), on the other hand, found a statistically significant effect in favour of chondroitin sulphate for the Huskisson visual analogue pain scale [a decrease of 3.66 cm in the intervention group and 1.64 cm in the placebo group (p < 0.001) in Uebelhart (1998) and a decrease of 24.5 mm in the intervention group and 15.3 mm in the placebo group (p < 0.05) in Uebelhart (2004).

Uebelhart (1998) did not report on the use of rescue analgesia but Uebelhart (2004) noted statistically significantly less use of paracetamol in the chondroitin group compared with the placebo group at 12 months (25.8 tablets versus 55 tablets per month; p < 0.05). Michel reported no statistically significant difference.

Function

Michel did not find any significant differences between chondroitin sulphate and placebo for the WOMAC function or stiffness subscores. The trial by Uebelhart (1998) found no significant effect for overall mobility capacity, but the trial by Uebelhart (2004) found statistically significant results in favour of chondroitin sulphate for walking time [a decrease of 4.4 seconds in the intervention group and 0.1 seconds in the placebo group (p < 0.05)].

Function/pain composite scores

Michel did not find any significant differences between chondroitin sulphate and placebo for the WOMAC total score. However, the trial by Uebelhart (2004) found a statistically significant result in favour of chondroitin sulphate for the Lequesne index [a decrease of 3.2 points in the intervention group and 2.1 points in the placebo group (p < 0.01)].

Adverse events

None of the chondroitin trials found significant differences in reported adverse effects between the chondroitin groups and the placebo groups. No significant differences in routine laboratory test results (where reported) were seen.

Combination therapy with glucosamine and chondroitin

Trial characteristics

Both trials of glucosamine and chondroitin combination therapy were double-blind, placebo-controlled single centre trials with a duration of 1 year. The trial by Messier122 was industry funded; no funding information was available for Rai. 120 A total of 189 people participated in these trials. The mean age was between 70 and 74 years for Messier and around 54 years for Rai. The trial by Messier included more women than men (between 66% and 76% women); no gender information was given by Rai. Neither of the trials gave details of OA severity or duration. Rai reported baseline characteristics to be well balanced in the comparison groups, whereas in the trial by Messier, participants in the intervention group were significantly younger, more overweight, and had a higher income than those in the placebo group.

Interventions

In the trial by Rai, it was only stated that Kondro, an oral combination of glucosamine sulphate and chondroitin sulphate, was given. No details on dosing were given, but the Kondro capsules contain 250 mg glucosamine sulphate and 200 mg chondroitin sulphate, and presumably one capsule a day was given. In the trial by Messier, a combination of 1500 mg glucosamine hydrochloride plus 1200 mg chondroitin sulphate was given each day, either in one dose or distributed over three doses a day. During the first 6 months, this was supplemented in both comparison groups by six healthy lifestyle classes, and in the second 6 months, an exercise programme was added that lasted for 6 months. The exercise programme was home and facility based and consisted of two 15-minute walking sessions, separated by 20 minutes of strength training. Exercise sessions were held for an hour each twice a week. In the trial by Messier, paracetamol was allowed as a rescue medication. In the trial by Rai, patients were advised to take paracetamol for pain. Messier did not report whether the difference in analgesic consumption between the groups was significant (a decrease of 37% by 12 months in the intervention and 11% in the placebo group), and Rai did not report analgesic consumption.

Outcome measures and further trial details are summarised in Table 10.

Trial quality

The trial by Messier was of moderate quality, whereas the trial by Rai was of poor quality. Messier did not explicitly describe randomisation, but their description implies that randomisation was done centrally by the sponsor. Rai did not report the method of randomisation, it was unclear whether allocation was concealed and the study was described as double blind, but no further detail was given, withdrawals or drop-outs were not reported, and it was not reported whether analysis was by intention to treat.

Results

Structure

Rai found a statistically significant result for the preservation of minimum joint space with a combination of glucosamine sulphate and chondroitin sulphate compared with placebo [change from baseline –0.04 mm in the intervention group and –0.13 mm in the placebo group (p ≤ 0.01)]. Table 17 summarises the trial findings for the effects on structure.

TABLE 17 - Summary of the effect of combined glucosamine and chondroitin versus placebo on structure

C, control; I, intervention; NS, not significant.

Study	Outcome	Baseline	End of study	Change from baseline/difference between groups	p-value
Rai 2004120	Minimum joint space width	I: 3.66 mm C: 3.65 mm	I: 3.62 mm, p = NS vs baseline C: 3.52 mm, p ≤ 0.01 vs baseline		≤ 0.01

Symptoms/pain

Table 18 summarises the findings for symptoms and pain.

TABLE 18 - Summary of the effect of combined glucosamine and chondroitin versus placebo on measures of pain, function and composite health status scores

C, control; I, intervention; N, Newton; NS, not significant.

Study	Outcome	Baseline	End of study	p-value
Messier 2007122	WOMAC pain	I: 7.1, SE 0.5 C: 5.9, SE 0.5	6 months: I: 6.2, SE0.4 C: 6.2, SE0.4 12 months: I: 6.0, SE0.5 C: 5.18, SE0.5	NS
	WOMAC function	I: 25.9, SE 1.7 C: 21.1, SE 1.5	6 months: I: 21.9, SE 1.1 C: 22.9, SE 1.1 12 months: I: 19.4, SE 1.2 C: 20.6, SE 1.2	NS
	6-minute walk	I: 384.7 m, SE17.6 C: 398.7 m, SE17.3	6 months: I: 393.6 m, SE 8.0 C: 396.5 m, SE 7.9 12 months: I: 409.2 m, SE 8.7 C: 410.5 m, SE 8.6	NS
	Knee concentric extension strength	I: 209.4 N, SE 31.2 C: 163.9 N, SE 20.6	6 months: I: 176.9 N, SE 16.3 C: 202.7 N, SE 17.5 12 months: I: 207.6 N, SE 14.6 C: 209.7 N, SE 15.0	NS
	Knee concentric flexion strength	I: 106.0 N, SE 16.1 C: 83.0 N, SE 10.9	6 months: I: 106.1 N, SE 7.3 C: 106.7 N, SE 7.8 12 months: I: 102.9 N, SE 7.7 C: 124.8 N, SE 8.3	NS at 6 months, p = 0.05 at 12 months
	Balance (foot length)	I: 0.52, SE 0.04 C: 0.53, SE 0.03	6 months: I: 0.523, SE 0.014 C: 0.583, SE 0.017 12 months: I: 0.538, SE 0.017 C: 0.591, SE 0.020	p = 0.01 at 6 months, p = 0.05 at 12 months
Rai 2004120	Lequesne index	I: 4.6 C: 4.9	I: 3.7 C: 11.48	≤ 0.01

Pain

Messier did not report any statistically significant difference in the pain component of WOMAC at 6 or 12 months. Rescue analgesia use was decreased in both the treatment and placebo groups, but there was no statistically significant difference between the groups. Rai did not report on pain as a separate score nor did they report on analgesia use.

Function

Messier did not find any statistically significant differences either at 6 months (i.e. before the exercise intervention) or at 12 months for a combination of glucosamine hydrochloride plus chondroitin sulphate plus exercise (versus placebo plus exercise) for any of the following outcomes: WOMAC function subscale, the distance walked in 6 minutes or knee concentric extension strength. The intervention group had a statistically significant benefit for knee concentric flexion strength and balance.

Pain/function composite scores

Messier did not find any statistically significant differences either at 6 months or at 12 months for a combination of glucosamine hydrochloride plus chondroitin sulphate plus exercise versus placebo plus exercise for results of the WOMAC total score. In the trial by Rai, a statistically significant benefit of the combination treatment compared with placebo was seen on the Lequesne index. However, compared with other studies reporting results of the Lequesne index, the deterioration seen in this trial in the placebo group seems rather extreme (4.6. to 3.7 in the intervention group versus 4.9 to 11.48 in the placebo group).

Adverse events

Neither of the trials of glucosamine/chondroitin combination therapy reported systematically on adverse effects.

Adverse effects of glucosamine or chondroitin – additional data

No long-term observational studies were identified that investigated any adverse events association with glucosamine (any form) or chondroitin sulphate or a combination of the two.

Anderson and colleagues64 conducted a review of adverse effects of glucosamine which did not have the characteristics of a systematic review, but seemed to aim at some thoroughness (literature search described, obviously some formal extraction of data). For their safety assessments in humans, they summarised data from 33 studies in 3063 patients. In 16 studies of a total of 854 participants followed for a weighted average of 37 weeks, no significant changes in blood glucose values were seen with glucosamine administration at therapeutic doses. In 13 studies including over 800 participants, no laboratory or cardiovascular abnormalities were seen with glucosamine administration at therapeutic doses for a weighted average of around 40 weeks. Studies reporting common adverse events included 988 participants followed for a weighted average of 37 weeks. Adverse events were reported less frequently with glucosamine than with placebo [rate ratio 0.76 (95% CI 0.61 to 0.92, p < 0.05)]. Five studies reported adverse events with glucosamine in comparison with adverse events with ibuprofen, and the prevalence of adverse events was 10.0% with glucosamine and 32.5% with ibuprofen.

Stumpf and colleagues132 reviewed data on the effect of glucosamine on glucose control. The review was not a systematic review, but a search strategy was described. Of 16 clinical studies of oral glucosamine, only five included plasma glucose concentrations that could be pooled. After 3–12 weeks of treatment, glucose values in the glucosamine groups were not significantly different from those in the placebo groups. Two 3-year studies (Pavelká and colleagues77 and Reginster and colleagues76) also found no significant effect on glucose levels. However, the authors concluded that data were limited and longer-term effects were uncertain.

One RCT by Scroggie and colleagues133 studied 34 patients with type 2 diabetes and a mean age of 69–71 years. Patients in the intervention group (n = 22) were given a combination of 1500 mg glucosamine hydrochloride plus 1200 mg chondroitin sulphate per day. Twelve patients were in the placebo group. After 90 days of treatment, glycaemic control (HbA1c levels) were not significantly different between the comparison groups.

Knudsen and Sokol134 presented a case report of a potential interaction of glucosamine/chondroitin with warfarin in a single patient, resulting in an increased international normalised ratio (INR) (i.e. a higher risk of bleeding). The authors did a search of the US FDA MedWatch database and the WHO Adverse Drug Reactions database, as well as a literature search of MEDLINE, EMBASE and the International Pharmaceutical Abstracts database. They identified 43 unique cases of an increased INR in people using warfarin and glucosamine or a glucosamine/chondroitin sulphate combination product (20 cases from MedWatch, 21 cases from the WHO database, and one published case report). The reports showed a temporal relationship between glucosamine or combination product use and warfarin, with a number of the reports stating that patients had a stable INR when they started taking glucosamine, and that the event resolved again in most cases when glucosamine was stopped. The Therapeutic Goods Administration of the Australian Government has received 12 reports (nine included in the WHO data) of interactions between warfarin and glucosamine. In 10 of these case reports, clinically significant rises in INR were observed 4–20 days after initiation of glucosamine. 135

A case report in the UK136 reported the death of a man who developed acute liver failure. The otherwise healthy 64-year-old man had recently started treatment with glucosamine. A fatal accident enquiry concluded that there was insufficient evidence to attribute the death to glucosamine.

Summary

We summarise below the evidence of clinical effectiveness of glucosamine and chondroitin, highlighting where data were suitable for inclusion in economic modelling and where there were important gaps or uncertainty in the evidence.

Introduction

The review of clinical effectiveness shows that of the therapies that were reviewed, only glucosamine sulphate was demonstrated to have a statistically significant effect on QoL and/or disease progression. Thus, cost-effectiveness modelling was limited to a comparison of glucosamine sulphate versus current care. The model draws upon several data sources, including the paper by Bruyere and colleagues,52 which uses pooled data from two previous RCTs of the clinical effectiveness of glucosamine sulphate. 76,77 However, there was a lack of data on costs and QoL from the pooled study. Thus, the model also draws upon other studies that report this information. Further detail is outlined in the Methods section and Appendix 7. One important issue to note relates to the definition of current care. In the model presented below, estimates of the costs of current care are based on two UK sources,32,137 while estimates of comparative effectiveness are based on the definition of current care used by Bruyere and colleagues,52 that is: ‘…any care received from health-care providers, with or without physical therapies and investigations (hydrotherapy, exercise, ultrasound, radiological procedures), and treatment with prescribed medicines (includes antacids, hypnotics and anxiolytics, antidepressants, analgesics, corticosteroids, drugs for nutritional, blood, musculo-skeletal and joint diseases, and local anaesthetics)’.

Methods

The costs and effectiveness of glucosamine sulphate were modelled using cohort simulation (Figure 8). Cohort simulation involves the placement of a number of individuals with pre-defined baseline clinical characteristics into particular states of health, with associated costs and QoL, who may then move or not move to other states of health over time. Owing to a lack of well-defined discrete states in knee OA, the approach taken here involved the placement of the cohort into an initial baseline level of health status, followed by change in health status over time. Thus, rather than using discrete states, health was modelled along a continuum. In addition to the continuous health status measure, two additional states were used. First, to reflect the greater costs and QoL implications associated with knee replacement arthroplasty, a separate state was included for this outcome. Knee arthroplasty can include a number of different surgical procedures. For the purpose of the model, total knee replacement (TKR) surgery was assumed. Further, although therapy had no effect on survival, as the time horizon was remaining lifetime, the state of death was also included. An outline schematic of model structure is shown in Figure 8.

FIGURE 8.

Outline schematic of model structure. TKR, total knee replacement.

The cohort was pre-defined according to the baseline clinical characteristics reported within Pavelká and colleagues. 77 We used this source as this was the only published randomised placebo-controlled trial of therapy that reported baseline and follow-up QoL data, using an instrument (WOMAC) that permitted conversion to a preference-based utility scale (HUI3). The conversion from Likert WOMAC score to HUI3 was performed using Grootendorst and colleagues. 138 Grootendorst and colleagues138 conducted a multicentre randomised trial of hylan G-F 20 against appropriate care among 255 Canadian patients with symptomatic knee OA of mild to moderate severity, who had previously been treated with analgesics, in which patients completed WOMAC and HUI3 questionnaires at 2-monthly intervals over 12 months. It is important to use preference-based scales such as the HUI3, in order to permit comparisons of cost-effectiveness with other therapies using quality-adjusted life-year (QALY) measures. A recent paper,139 published after model development, has produced EQ-5D estimates derived from WOMAC for knee pain patients. However, as the model does not account for disease duration, and was not derived from a population of knee OA patients, we chose to retain the Grootendorst and colleagues’ mapping algorithm. 138

Based on Pavelká and colleagues77 and Grootendorst and colleagues,138 the cohort had an initial QoL score of 0.69 (on a 0–1 scale, where 0 = dead and 1 = perfect health). It is useful to note that this value is identical to the EQ-5D based value from Barton and colleagues,139 used by the NCCCC,1 which modelled the cost-effectiveness of oral NSAID or COX inhibitor therapy for OA. The value of 0.69 was computed using the best performing model (model 3); this model had the lowest mean absolute error (MAE) and the lowest root mean square error statistics (RMSE) (MAE = 0.1628, RMSE = 0.2065), relative to the closest alternative model that included Kellgren radiographic grade (MAE = 0.1638, RMSE = 0.2083). Model 3 is shown as follows:

The authors show with the aid of a worked example that the model predicts that a 56-year-old woman with WOMAC pain, stiffness, and function scores of 10, 5, and 24, respectively, who has had OA for 2.5 years, would have a HUI3 score of 0.68. For our cohort, we obtained a similar value (0.69) to this with different parameters. We used less disabled WOMAC pain, stiffness and function scores of 6, 2 and 22 (the higher the score, the worse the disability), and higher values for OA duration and age, reported from the mean intention-to-treat values in the placebo group in Pavelká and colleagues. 77 We were only able to use the Pavelká and colleagues77 estimate as Reginster and colleagues76 reported VAS WOMAC scores rather than the Likert version. Further, for ease of illustration we approximated the Pavelká and colleagues77 demographic estimates by using OA duration of 10 years and a starting age of 60 years (the mean values in the placebo group were 11 years and 63 years respectively).

Once the cohort was placed in the starting level of health, thereafter individuals had a particular probability of entering the TKR state, or avoiding TKR and progressing to a different state of health, or dying. Within the glucosamine group, individuals were assumed to remain on therapy until death. Probability of entering the state of TKR was derived from Bruyere and colleagues;52 the TKR risk was annualised using a standard formula, p* = 1 – (1 – p)^1/t, where p = event probability over study duration and t = time in years. Probability of death was estimated from age-specific all-cause life tables. 140 As explained above, the QoL associated with health change over time among those neither dying nor entering the TKR state was taken from Grootendorst and colleagues,138 who estimated regression models that permit conversion of HUI3 scores using WOMAC subscale scores, age, and duration of OA in the study knee. Owing to a lack of data reporting QoL prior to TKR within Bruyere and colleagues,52 QoL prior to TKR was estimated from another source, using baseline (48 hours prior to surgery) WOMAC subscale scores published by Nunez and colleagues,141 giving a value of 0.51. Within the model, individuals who entered the ‘prior to TKR’ health state did so at the beginning of each annual cycle and remained in that state for the complete cycle length, i.e. 12 months, then rejoined the larger non-progressive cohort in future cycles. Added to this, for both progressors and non-progressors, account was taken of the natural history of knee OA, using data from Johnson and colleagues,22 who reported 7-year WOMAC subscale score data for 43 Canadian patients (mean age at baseline 69 years, 77% female) with moderate to severe hip or knee OA. Converting to HUI3 using the Grootendorst and colleagues model,138 the difference in QoL after 7 years of 0.04 gives an annual QoL decrement value of 0.006.

The effectiveness of therapy on health states was obtained from two main sources. First, we used the estimate from Bruyere and colleagues52 of the effectiveness of therapy in reducing the risk of undergoing TKR surgery (RR = 0.43, 95% CI 0.20 to 0.92). The Bruyere estimate was based on pooled data from 275 patients who had participated in two previous randomised placebo-controlled trials of glucosamine sulphate,76,77 with follow-up data for a mean period of 5 years after the end of both 3-year trial periods.

Second, the effectiveness of therapy in improving QoL for those who did not progress to TKR was estimated using Pavelká and colleagues,77 who demonstrated a statistically significant change in WOMAC total scale and subscale scores. To generate the annual utility improvement with therapy, the WOMAC subscales scores at baseline and follow-up, with and without therapy, were converted to HUI3 using the regression model of Grootendorst and colleagues. 138 The calculation is shown in Table 20.

Health-care costs associated with knee OA were estimated from a UK study conducted by Lord and colleagues. 32 This study was an RCT of primary care-based education for knee OA. In both control and intervention groups, costs were related to the resources consumed for GP visits, medications, outpatient visits, inpatient care, professions allied to medicine consultations, complementary therapists and X-ray procedures for a population of patients with radiographic evidence of knee OA. Data were collected from case notes, supplemented by patient interview for data on professions allied to medicine consultations. We used the reported NHS direct cost element, uprated to 2007–8 prices, for the control group who did not receive the education programme (£216 per patient per year). This estimate was very similar to the values reported in a more recent study by Hurley and colleagues,137 who calculated 6-month health-care and social care costs of £103 for usual care among chronic knee pain patients in the UK at 2003–4 prices. The cost of TKR surgery (£6126) was calculated by uprating to 2007–8 prices the 2005–6 NHS Reference Costs estimate for elective primary knee replacement (code HRG H04, £5747). 142 We used the lower and upper interquartile range estimates, uprated to 2007–8 prices, in probabilistic analysis, applying a uniform distribution. No UK published prices were available for glucosamine sulphate, as the product has no UK licence. UK prices were, however, available for the licensed hydrochloride preparation. Listed prices for glucosamine sulphate in the USA were US$34.95 for 30 days’ supply,143 giving a total annual cost of US$429.40 with international shipping, or £242.60 at an exchange rate of £1 = US$1.77. Purchasing power parity (PPP) conversion from US$ gives a cost of £280.40. The price to the NHS is likely to be cheaper, however, given that the NHS is a monopoly buyer, and glucosamine sulphate would be likely to be priced at a similar level to competitor products, such as the hydrochloride formulation. For the purposes of assessing the cost-effectiveness of a likely guide price in the future, we therefore used current UK published market prices of glucosamine hydrochloride, giving an annual cost of £221 per patient per year. In addition however, based on the PPP conversion price of £280.40, we used upper and lower bounds of 27% in deterministic sensitivity analysis to assess the impact of alternative prices on cost-effectiveness. Both costs and health outcomes were discounted at 3.5%. Assumptions for model parameters are outlined in Table 21.

Results

Over a lifetime horizon (mean life expectancy of 22.61 years) the mean discounted costs per patient associated with current care were calculated as £4634. Care costs with the addition of glucosamine sulphate were estimated to be £7039 over the same time period. Thus, the additional discounted costs associated with glucosamine were £2405. Mean discounted QALYs were 8.17 for current care and 8.28 for glucosamine, giving a QALY gain of 0.11. Avoiding rounding error, the incremental cost per QALY gain was £21,335.

Deterministic sensitivity analysis revealed that estimates were reasonably robust to variations in assumptions (Figure 9). Estimates were most affected by changes in the QoL gain associated with therapy. Changes to the costs of current care or TKR costs had negligible effects.

FIGURE 9.

Deterministic sensitivity analysis. QALY(s), quality-adjusted life-year(s); QoL, quality of life; RRR, relative risk reduction; TKR, total knee replacement.

The results of the probabilistic sensitivity analysis are presented in Figures 10 and 11. Figure 10 shows the scatterplot of differences in costs and QALYs from the 5000 replications. It was apparent that therapy was unlikely to be cost saving, as there were only 41 occurrences of costs within the glucosamine group being lower than those in the current care group. However, there was a wide scatter across the x-axis (depicting QALY differences). Figure 11 shows the cost and QALY differences between therapy and current care within the cost-effectiveness acceptability curve framework. This depicts the change in the probability for differing values of willingness to pay for a QALY (or ‘ceiling ratio’). This shows that glucosamine was not expected to be cost-effective if we were only prepared to pay up to £10,000 per QALY (probability of cost-effectiveness = 0.08), but for increasing amounts of willingness to pay per QALY, the probability of cost-effectiveness rises (e.g. if we were prepared to pay up to £20,000 per QALY, the probability of cost-effectiveness = 0.43). At a willingness-to-pay ceiling ratio of above approximately £22,000, the probability of cost-effectiveness becomes greater than 0.5.

FIGURE 10.

Scatterplot of incremental cost and quality-adjusted life-year (QALY) gained.

FIGURE 11.

Cost-effectiveness acceptability curves.

The EVPI provides an upper bound of the value of undertaking further research to reduce uncertainty over whether therapy is more cost-effective than current care. Figure 12 shows the relationship between these values and different ceiling ratios of willingness to pay per QALY. This reveals that the value of further research was high for a range of commonly applied ceiling ratios (e.g. £10,000–£50,000), and far exceeded the likely costs of research. The EVPI became lower as the ceiling ratio approached £100,000, as uncertainty surrounding cost-effectiveness became less for higher levels of willingness to pay per QALY.

FIGURE 12.

Expected value of perfect information (EVPI).

Figures 13 and 14 show the analysis of covariance results. Figure 13 reveals that the greatest uncertainty in costs was related to variation in TKR costs, followed by variation in the probability of TKR surgery with therapy, followed by variation in costs of current care. Figure 14 shows that the most important parameter that contributed to variation in QALYs was the QoL change arising from the addition of glucosamine to current care. Quality of life prior to TKR surgery, and the probability of progression to TKR surgery, make smaller, although similar, contributions in comparison. Figure 15 shows the partial EVPI for parameters. This reveals quite clearly that the most important parameter is the QoL gain associated with therapy. This is followed by the TKR probability with therapy. Current care costs, the quality of life associated with TKR and TKR costs have much smaller values in comparison.

FIGURE 13.

Individual parameter contribution to variation in costs. TKR, total knee replacement.

FIGURE 14.

Individual parameter contribution to variation in quality-adjusted life-years. QoL, quality of life; TKR, total knee replacement.

FIGURE 15.

Expected value of perfect information (EVPI) for parameters. QoL, quality of life; TKR, total knee replacement.

Discussion

The addition of glucosamine sulphate therapy to current care for knee OA was estimated to lead to additional costs and gains in health-related QoL. With a cost per QALY estimate of approximately £21,335, the magnitude of potential gains appears moderate in relation to the additional costs. Deterministic sensitivity analysis suggested that the cost-effectiveness of glucosamine sulphate therapy was particularly dependent on the magnitude of the QoL gains produced by therapy. Probabilistic sensitivity analysis showed that these estimates of cost-effectiveness were imprecise, suggesting some degree of decision uncertainty.

Our estimate of cost-effectiveness was higher in comparison with other available studies; the NCCCC1 produced two cost per QALY estimates for glucosamine sulphate. Using only 3-year data from Pavelká and colleagues,77 the mean cost per QALY was £10,880 (range £3581–£21,219). Using data only from Reginster and colleagues,76 the mean cost per QALY was £2427 (range £799–£6519). Without further information on the sources of data or assumptions used, it is difficult to explain the differences. One possible explanation is the difference in the timeframe adopted, although this seems unlikely, as the longer follow-up period adopted in the model presented here allows the incorporation of cost savings and QoL gains associated with the avoidance of knee replacement surgery, which would lead to lower, rather than higher, cost per QALY estimates. A more plausible explanation is the use of alternative estimates of QoL. As the same cost difference is used to generate the two estimates of cost per QALY reported by the NCCCC,1 the cost per QALY difference is driven by differences in the QALY gain. Such differences may be due to the use of alternative ways to value health changes. For example, the lower cost per QALY gain figures of Reginster and colleagues76 is generated from use of the VAS version of the WOMAC rather than the Likert scale version.

Value of information analysis indicated that further research to reduce decision uncertainty would be beneficial. It would appear that a high priority should be given to obtaining further evidence of the QoL gains produced by glucosamine sulphate relative to placebo. The current uncertainty surrounding the magnitude of such gains may be related to sample size and the requirement to map between health status instruments to obtain suitable QoL values in order to calculate QALYs. Thus, any future trial should aim to collect data using a generic preference-based QoL instrument, such as the HUI3, SF-6D or EQ-5D.

It is important to emphasise that the effectiveness data within economic model are built mainly around data from one study that pooled data from two trials,51 while the QoL gain with therapy is estimated from one trial. 58 Further evidence is also required on the relationship between QoL and costs. Owing to lack of data, other than through progression to TKR surgery, it was not possible to build into the model estimates of cost changes that arise from changes in QoL. This may be an important limitation of the current model. We would anticipate that incorporation of these effects would have made current cost estimates less precise, as there was large variation around the magnitude of QoL improvements, but would have made little change to mean cost per QALY estimates. This suggests that a future trial should also collect resource use and cost data, to allow estimation of the resource impact of changes in QoL.

Estimates were not particularly sensitive to QoL prior to TKR surgery, nor to costs of surgery. Further, cost-effectiveness was moderately related to the volume of TKR surgery in deterministic analysis. To assess the importance of this parameter, we were only able to apply arbitrary 50% changes in volume in deterministic sensitivity analysis, resulting in small changes of between 1% and 2% in either direction in the likelihood of requiring surgery. A nationally representative cohort survey to determine the current level of TKR surgery would help to ensure that the modelled estimates presented here are generalisable to the UK health-care system.

Glucosamine, chondroitin and hyaline cartilage

As discussed in Chapter 1, glucosamine and chondroitin are contained within hyaline cartilage. The non-collagenous 50% of hyaline cartilage is mainly made up of proteoglycan molecules. These molecules (also called aggrecans) are composed of long-chain glycosaminoglycans linked to a core protein. Loss of aggrecan from the extracellular matrix in cartilage leads to a change in the biomechanical properties of the tissue, and in so doing accelerates the loss of articular cartilage.

The glycosaminoglycans consist of a hexuronic acid and a hexosamine (amino sugar).

Glucosamine is the hexosamine constituent of keratan sulphate, which is found in hyaline cartilage, alongside the glycosaminoglycans chondroitin 4-sulphate and chondroitin 6-sulphate (which are therefore much larger molecules than glucosamine). 63 Chondroitin is a large gel-forming molecule which forms part of cartilage and confers resistance to compression. Despite being a large molecule, it is partially absorbed from the diet or supplements.

Anabolic state

Many authors have studied the role of glucosamine on both chondrocytes themselves and also the chondroprogenitor cells, such as the human mesenchymal stem cells (hMSCs). Dodge and Jimenez166 studied the effects of glucosamine sulphate on freshly isolated chondrocytes obtained from patients with knee OA. They observed that 40% of chondrocytes from OA patients failed to respond to glucosamine sulphate. Of the remaining 60%, however, it was observed that the presence of glucosamine sulphate upregulated messenger ribonucleic acid (mRNA) levels of the aggrecan gene in a dose-dependent manner, leading to a corresponding increase in observed levels of aggrecan. They also observed that the presence of glucosamine sulphate decreased matrix metalloproteinase (MMP-3) protein levels as well as enzymatic activity, and increased glycosaminoglycan content. 166

Noyszewski and colleagues167 have also shown that the addition of radiolabelled glucosamine sulphate to cultured chondrocytes facilitates the production of proteoglycan components with direct uptake of the added radiolabelled glucosamine sulphate.

In another in vitro study, Derfoul and colleagues168 found that addition of 100 μM glucosamine hydrochloride led to enhanced expression of collagen II and aggregan, and an increased content of sulphated glycosaminoglycan. It was also observed that the addition of glucosamine partially blocked the effects of interleukin (IL)-1β on downregulating collagen II and aggrecan expression. Following addition of glucosamine, the expression of the matrix-degrading enzyme, MMP-13, was also downregulated in both chondrocyte and hMSC cultures. 168

The effects of glucosamine have also been attributed to the upregulation of chondrocyte proliferation, matrix synthesis and gene expression. Varghese and colleagues169 found that the addition of glucosamine hydrochloride upregulated matrix production in 2D and 3D bovine chondrocyte layers. Up to a dose of 2 mM, this effect was seen principally on aggrecan and type II collagen. This was not seen at higher doses. It was also noted that transforming growth factor (TGF)-β1 was upregulated in a dose-dependent manner in the presence of glucosamine hydrochloride. 169

Nishomoto and colleagues170 observed upregulation of the mRNA of collagen II and an increase in aggrecan mRNA expression in porcine chondrocytes after the addition of chondroitin sulphate. However, the effect was not seen with all subtypes of chondroitin sulphate. 170

Mroz and colleagues171 studied whether supplementation of glucosamine to cultured human chondrocytes leads to increased production of chondroitin sulphate. Figure 16 shows possible pathways.

FIGURE 16.

Biosynthetic scheme for incorporation of galactosamine into chondroitin from glucose or glucosamine precursors. 171 (Reproduced from: Mroz PJ, Silbert JE. Use of 3H-glucosamine and 35S-sulfate with cultured human chondrocytes to determine the effect of glucosamine concentration on formation of chondroitin sulfate. Arthritis Rheum 2004;50(11):3574–9. Copyright 2004 American College of Rheumatology. Reprinted with permission of John Wiley & Sons, Inc.)

Copyright 2004 American College of Rheumatology. Reprinted with permission of John Wiley & Sons, Inc.

They concluded that the addition of exogenous H-glucosamine did not stimulate an increased production of chondroitin sulphate. 171

Further work by Kim and colleagues172 explored the anabolic effects of physiological doses of glucosamine sulphate on human osteoblast-like MG-63 cells. The addition of glucosamine sulphate increased alkaline phosphatase (ALP) activity, collagen synthesis, osteocalcin secretion and mineralization of the cultured osteoblastic cells in vitro in a dose-dependent manner from 10–1000 μg/ml, generally increasing anabolic activity.

Not all studies support an increase in anabolic activity with glucosamine sulphate, however. Using pre-cultured human osteoarthritic explants, Uitterlinden and colleagues173 studied the effects of 0.5 mM or 5 mM addition of either glucosamine hydrochloride or glucosamine sulphate. A 5 mM concentration of both glucosamine subtypes was seen to significantly downregulate aggrecan and collagen type II gene expression suggesting inhibited anabolic activity. At the same doses, however, there was also significant anti-catabolic activity with downregulation of MMP-3, and aggrecanase-1. The addition of 0.5 mM glucosamine was not seen to alter either anabolic or catabolic activity at a significant level; however, similar trends were noted when compared with the 5 mM group. 173

Anticatabolic effects

Piperno and colleagues174 studied the effects of glucosamine sulphate on proteoglycan synthesis, caseinase, collagenase, and phospholipase A2 (PLA2), and on production of nitric oxide (NO), cyclic adenosine monophosphate (cAMP), and protein kinase C, in human osteoarthritic chondrocyte cell cultures. Glucosamine sulphate was found to inhibit PLA2 and collagenase activity in a dose-dependent manner. A dose-dependent increase in protein synthesis (minimum concentration 50 μM) and protein kinase C production was seen.

Monfort and colleagues175 studied the in vitro effects of chondroitin sulphate on MMP-3 (stromelysin-1) synthesis in human osteoarthritic chondrocytes. MMP-3 is a cartilage proteolytic enzyme which induces cartilage breakdown and also acts as an inflammatory mediator. Chondrocytes were obtained from human osteoarthritic cartilage from patients of both sexes, over the age of 40 years, who were undergoing arthroplasty. The study noted a marked inhibitory effect of chondroitin sulphate on MMP-3 synthesis at all concentrations tested. No significant differences were found between the various concentrations. 175

The majority of the studies analysing the in vitro effects of glucosamine and chondroitin sulphate have focused purely on chondrocyte cultures. The onset of OA is increasingly understood to involve changes within the subchondral bone, including subchondral bone resorption. The balance of bone remodelling factors osteoprotegerin (OPG), and receptor activator of nuclear factor κB ligand (RANKL), are important in this process. Tat and colleagues176 studied the effects of both glucosamine sulphate and chondroitin sulphate and found that the chondroitin sulphate reduced production of OPG and RANKL, whereas glucosamine sulphate significantly reduced resorptive activity. Both chondroitin sulphate and glucosamine sulphate may, therefore, play an important role in reducing the catabolic effects of subchondral bone in OA.

Cibere and colleagues177 studied the effects of glucosamine sulphate on type II collagen breakdown in a randomised double-blind placebo-controlled study. One hundred and thirty-seven patients with proven knee OA who had gained at least moderate benefit from glucosamine therapy were randomised to either placebo (discontinuation) or glucosamine (continuation) treatment over a 24-week period. The expectation was that the placebo group would show an increase in type II collagen markers after discontinuation of glucosamine sulphate. However, this study failed to demonstrate any significant difference between the two groups when measuring glucosamine sulphate degradation products in serum and urine. 177

Aggrecanase activity

The aggrecan degradation simulated by IL-1, tumour necrosis factor (TNF)-α, or retinoic acid is thought largely to be due to the action of aggrecanase. Ilic and colleagues178 studied the effects of long-term glucosamine on aggrecan degradation in cultures of bovine cartilage. They concluded that a 5 mM concentration of glucosamine indirectly inhibited degradation of aggrecan under simulated catabolic activity. When glucosamine was then removed from the cultures, aggrecanase activity increased. Uitterlinden and colleagues173 also demonstrated that the addition of 5 mM either glucosamine sulphate or glucosamine hydrochloride to cultured human osteoarthritic explants demonstrated a significant downregulation of aggrecanase-1. Suppression of aggrecanase-2 was also detected in the glucosamine sulphate group.

Furthermore, work by Sandy and colleagues179,180 concluded that aggrecanase activity is significantly decreased in the presence of glucosamine and that this may be via suppression of glycosylphosphatidylinositol-linked proteins.

Sulphate deficiency

A further plausible hypothesis is based on the role of the sulphated amino sugars found in glucosamine hydrochloride and chondroitin sulphate. Cartilage contains tissue-specific glycosaminoglycans that require a source of inorganic sulphate for its synthesis. Although the recommended dietary allowance has previously been derived in young healthy individuals, this does not take into account the increased turnover of glycosaminoglycans during the osteoarthritic process. Furthermore, many patients with OA take paracetamol as a simple analgesic to manage their symptoms. Paracetamol is largely metabolised by sulphation and indeed 40% of paracetamol is excreted in urine conjugated with sulphate, which may contribute to sulphate deficiency. It has been previously shown that sulphate depletion inhibits glycosaminoglycans synthesis in vitro in rat and human articular cartilage. Indeed, human cartilage is thought to be particularly sensitive to sulphate deficiency. 192,193

Hoffer and colleagues194 have also investigated the role of sulphate as the proposed mechanism of action of glucosamine sulphate. They first studied whether oral glucosamine sulphate increases serum sulphate concentrations. They then investigated whether this increase was also found within synovial fluid. Finally, the group studied whether the use of paracetamol affected the serum sulphate response to glucosamine sulphate. Seven healthy adults had venous blood measured for sulphate levels before, and then 3 hours following, ingestion of 1 g glucosamine sulphate (1.65 mmol sulphate). The study was then repeated in a similar fashion with the addition of 1 g paracetamol along with the glucosamine sulphate. In a separate group, 15 patients with known OA, who were to undergo diagnostic needle aspiration in the outpatient setting, consented to the measurement of both synovial and serum sulphate. This was to investigate the relationship between serum and synovial sulphate levels which were found to be virtually identical.

Of the seven healthy patients, the mean baseline serum sulphate level was 331 μmol/l which increased to 375 μmol/l 3 hours following ingestion of glucosamine sulphate. In the glucosamine sulphate and paracetamol group, the baseline level decreased from 325 μmol/l to 290 μmol/l at 3 hours. The study therefore concluded that glucosamine sulphate increases serum inorganic sulphate concentration in normal individuals and that this increase is reversed by coingestion of paracetamol. 194

Cordoba and Nimni195 investigated the relationship between dietary levels of protein and urinary excretion of sulphate. This study assessed whether consumption of chondroitin sulphate and glucosamine sulphate increased urinary excretion of sulphate. Healthy human volunteers were instructed to record their dietary intake. Patients were separated into five groups depending on their routine dietary protein intake. In the two lowest groups for dietary protein intake, consuming 15 mM or less of sulphur over 24 hours, supplementation with chondroitin sulphate did not lead to the expected increase in urinary secretion, suggesting retention of sulphate, presumably due to dietary insufficiency. In the groups consuming 25 mM or more of sulphur, supplementation of either glucosamine sulphate or chondroitin sulphate led to an increase in urinary excretion of sulphate over the measured daily dietary consumption. The effect of glucosamine sulphate for patients with low dietary sulphur was not described in the paper. It is plausible, therefore, that chondroitin sulphate and glucosamine sulphate supplementation may be most effective in patient populations with low dietary sulphur, such as vegetarians or those who consume little red meat.

Summary

While much of the evidence presented above is derived from in vitro experiments which may not reflect the osteoarthritic human knee, there are a number of plausible mechanisms by which glucosamine, chondroitin or the sulphate component of the treatments may act:

• Bioavailability: there was significant evidence that oral supplements of both chondroitin and glucosamine led to an increase in both serum and synovial fluid concentrations.

• Anabolic effect: in vitro studies supported some anabolic function for glucosamine.

• Catabolic effect: there was a stronger evidence base for inhibition of catabolism.

• Sulphate effect: there was evidence to support the intriguing hypothesis that it may be the sulphate rather than the glucosamine which has the effect.

The exact mechanism(s) of either drug is as yet unknown. Furthermore, many but not all of the in vitro studies utilised concentrations in excess of those found from the bioavailability studies, and therefore may not be clinically relevant.

Summary of findings

In this review we set out to address a series of questions with regard to the clinical effectiveness and cost-effectiveness of glucosamine and chondroitin.

• Does glucosamine prevent or slow progression of OA of the knee?

  1. – There was evidence that the Rotta powdered oral preparation of glucosamine sulphate reduced joint space loss and may even improve joint space width after 3 years of therapy when compared with placebo. The clinical significance of joint space narrowing and changes in width over time were less certain. Absolute joint space width is a poor predictor of outcomes in terms of either QoL or need for knee arthroplasty. There is, however, some evidence that changes in joint space width do relate to clinical outcomes such as the need for future knee arthroplasty.

  2. – There was evidence that glucosamine had a beneficial impact on long-term outcome of knee arthroplasty (6.3% in the treated group versus 14.5%); an effect seen despite small numbers and the opportunity for substantial treatment group switching during follow-up which would, if anything, lessen the observed difference between the two study groups.

  3. – A small statistically significant improvement was observed in QoL measured using WOMAC in the treated group versus the placebo group.

• Does chondroitin prevent or slow progression of OA of the knee?

  1. – We found evidence to support the conclusion that chondroitin had a small effect on structural changes, but this time there was no corroborating evidence of a translation of this finding into long-term clinically important outcomes. The impact on QoL appeared to be influenced by the tool used and was inconsistent between studies.

  2. – As there was currently no way to determine what, if any, significance joint space width has on QoL, health utility or long-term outcomes, it was not possible to draw any conclusions about the clinical effectiveness of chondroitin.

• Does the combination of both offer any greater effect?

  1. – While a few studies did use combination therapy, the comparison was placebo not single therapy. It was not possible, therefore, to address whether there was a relatively greater gain with such a combination.

• If they are useful, what dosages should be used?

  1. – With only two trials for the one preparation for which we had concluded that there was evidence of effectiveness, and with both trials using identical dosing regimens, it was not possible to address the question of what was the optimal dose. It is worth noting that both trials used the Rotta glucosamine product and were sponsored by the pharmaceutical company.

• What are the costs, in terms of both cost of the products and any side effects?

  1. – The reported adverse event profile for glucosamine sulphate was unremarkable. The only potential issue identified from data beyond clinical trials was an interaction with warfarin. The one death associated with acute liver failure, however, highlighted the need for long-term monitoring of the use of glucosamine and that, to date, experience from formal study of the treatment draws on a relatively small patient pool. There is no licensed preparation of glucosamine sulphate available in the UK. Over-the-counter preparations vary substantially in cost and content. For modelling, the cost of the licensed glucosamine hydrochloride product was taken as a marker for the likely price to the NHS of a sulphate licensed product should one become available (BNF price of £18.40).

• Are there any savings to the NHS from their use, for example by reduced consumption of prescribed non-steroidal anti-inflammatory drugs, or from avoidance of knee arthroplasty?

  1. – There was little evidence from the trials that glucosamine sulphate reduced the need for oral analgesia, but as noted above, the long-term follow-up data did support a reduction in the need for knee arthroplasty.

• Is there evidence of cost-effectiveness from well constructed economic evaluations?

  1. – We did not identify any economic evaluations focusing on the long-term cost-effectiveness of glucosamine sulphate. Our own analysis estimated the cost per QALY to be £21,335. Gains in cost to the health service were accompanied by modest health-related gains. However, sensitivity analysis suggested that the cost-effectiveness of glucosamine sulphate was largely dependent on the magnitude of QoL gains from therapy. The estimates of cost-effectiveness were imprecise and suggest some degree of decision uncertainty.

  2. From our review we conclude that the evidence is inconclusive, but suggestive enough of benefit to justify further research for glucosamine sulphate. Our conclusions are in keeping with those of the recent NCCCC guidelines1 that included an assessment of glucosamine and chondroitin. While they did not focus on long-term outcomes and used a different methodology for determining utility scores in the economics analysis, they did report that a small effect was observed for glucosamine sulphate on QoL measures and joint space. Like us, they concluded from their economic analysis that evidence of cost-effectiveness was not conclusively established.

Similarly authors of the systematic reviews included in our clinical effectiveness review reported evidence of some clinical effect associated, in particular, with preparations of glucosamine sulphate, and to some extent, chondroitin.

• Are the alleged benefits of glucosamine on knee cartilage biologically plausible?

  1. – Bioavailability of the glucosamine in the synovial fluid has been demonstrated. While a specific mechanism of action has not been established, evidence for a number of potential actions on the joint has been observed for glucosamine, chondroitin and sulphate (in vitro, in animals and in healthy volunteers).

• If, as suggested by the commissioning brief, the evidence suggests that there is no gain in short-term symptom relief, but only in long-term preservation of cartilage (which does imply longer-term reductions in symptoms, advanced OA and perhaps the need for knee arthroplasty), what should the comparator in trials be?

  1. – Because OA is now recognised to be a condition that is not necessarily relentlessly progressive and, in fact, the symptoms experienced by individuals can range widely, a comparator group is essential for the interpretation of trial data. NCCCC identified, in their review of treatments for OA of the knee, that none of the existing therapies have been shown to alter the clinical course of the condition. The comparator, therefore, has to be against placebo.

• If the gain is in long-term outcomes, are there short-term indicators of benefit which could be used, rather than waiting 20 years for advanced OA to become manifest? Joint space narrowing on plain film has been used as one outcome. Could more complex forms of imaging, such as MRI, positron emission tomography, computerised tomography, other radionuclide techniques or ultrasound, provide more detailed assessment of the quality and thickness of remaining cartilage, and hence give earlier confirmation of benefit?

  1. – The two most significant challenges around imaging the knee using X-ray have been the difficulty in relating changes in the image to long-term outcomes and the need to use quantitative measures to assess change.

  2. – There have been growing numbers of reports in the literature about the use of other imaging techniques in assessing joint space changes in OA. While there is evidence of a change in joint space being associated with long-term outcomes such as knee arthroplasty, regardless of the technique used to image the joint, the difficulties remain about the reliability of such measurements. MRI and other techniques afford the additional benefit of providing more information about structural changes with in the joint itself. An evidence-driven consensus conference in 2006 lead by OMERACT and OARSI attempted to summarise opinion about the most potentially useful features derivable from MRI. They concluded that articular cartilage loss, osteophytes, bone marrow abnormalities, synovitis, meniscal abnormalities and synovial effusions held the greatest promise. Despite enthusiasm for these new measures, the need for long-term cohorts and follow-up data was recognised as a priority to determine the clinical significance.

• Which variables or assumptions (e.g. costs, QoL, mortality, extrapolation of effectiveness over time) are most important in generating uncertainty in the cost-effectiveness estimates, and which would provide the greatest return in terms of reducing uncertainty at reasonable cost? These variables and assumptions will be identified using value of information analysis, and will form the main basis in recommendations for further research.

  1. – Value of information analysis indicated that further research to reduce uncertainty would be beneficial. Greatest uncertainty around costs is generated by the uncertainty around the probability and cost of knee arthroplasty. The most important driver in QALY variability is uncertainty around the QoL gain associated with treatment. Overall, variability in cost-effectiveness is most heavily determined by uncertainty regarding the magnitude and duration of QoL gains following treatment.

For chondroitin, biologically plausible mechanisms have been reported, but the clinical effectiveness evidence was less convincing with greater heterogeneity in conclusions. Trials of chondroitin, however, have tended to be of poorer quality, so uncertainties about the clinical effects of chondroitin remain.

Challenges for clinical effectiveness

Despite numerous reviews of these two treatments, most conclusions have previously been driven by short-term findings, particularly for measures of symptoms and function. The failure to undertake subgroup analysis by duration of follow-up made it impossible to determine the effectiveness for long-term outcomes. Studies including structural outcomes tended to be longer term; therefore, the conclusions of previous reviews on this outcome are of relevance. Poolsup and colleagues73 focused on long-term outcomes and identified two of the three trials we included. Their conclusions were similar to ours, supporting evidence of clinical effectiveness for both symptoms and structural outcomes. Towheed and colleagues62 included two of the three trials we included in their assessment of structural outcomes, plus an additional shorter-duration study that we excluded. They were cautious in their interpretation of the significance of the surrogate end point of joint space change and at that time did not have access to the longer follow-up information about knee arthroplasty.

Surrogate end points in trials are widely used for many conditions and often lead to concerns regarding their translation into long-term effects. As a surrogate end point, joint space narrowing has been approved as an acceptable trial outcome measure for OA by OMERACT and regulatory bodies41 but its insensitivity in diagnosis, measuring disease activity and assessing progression has been noted.

To address the challenges of surrogate end points in technology assessment, the HTA are currently establishing a working group to explore the challenges of surrogate markers in clinical trials with a view to preparing guidance on their use in health technology assessment. Critical to our interpretation of the potential effectiveness of glucosamine sulphate has been the evidence of a translation from the surrogate end point of joint space narrowing to the clinical end point of knee arthroplasty.

Where evidence of effectiveness was observed (for glucosamine sulphate) the conclusions are based on only three trials. While two of the three trials were well designed, they followed similar treatment protocols and therefore did not allow any assessment of the relative merits of different doses, products or regimens.

Other challenges included interpretation of QoL measures. A number of different tools were used to assess pain, function and disability. Authors rarely stated a priori expectations of what might represent a clinically important change. There was little information about the long-term natural history of such measures or how such measures relate to other outcomes, including knee arthroplasty or other service utilisation.

Finally, none of the studies were undertaken in the UK and, given the importance of access to services and complexity of decision-making around the timing of surgical intervention, generalisability to the UK health-care setting is difficult.

Challenges for modelling

Three main challenges for modelling are worth noting. First, the lack of robust epidemiological and service utilisation data for the UK meant that there was substantial uncertainty about the generalisability of trial data and the cost of ‘current care’. Second, there was no direct measure of a generic preference-based QoL measure (such as HUI3, SF-6D, EQ-5D) and as a result mapping formulae had to be used to map WOMAC to a utility measure. Finally, the economic model relied on the results of two studies for effect of treatment on WOMAC and knee arthroplasty. The number of participants was relatively small, and only one treatment protocol and glucosamine preparation was used. There remain, therefore, a number of decision uncertainties that could not be addressed in the economic modelling.

Research needs

1.  – Based on sensitivity analysis and value of information analysis, three research priorities were identified:

   • Quality of life – further clarification of the potential QoL gains from treatment with glucosamine sulphate versus placebo over long-term treatment would reduce uncertainty in the cost-effectiveness estimates. There was consistent evidence of some effect on various composite scores, but no study utilised a generic preference-based QoL measure (such as HUI3, SF-6D, EQ-5D) that can readily be used to estimate utility. Any future trial should also inform our understanding of the relationship between QoL and costs collecting resource use and cost data to allow estimation of the resource impact of any changes in QoL.

   • Structural outcomes – further long-term trial data are required to clarify the impact on the ultimate need for knee arthroplasty, including the ability to delay the need for surgery. As joint replacements have a finite lifespan, optimising the timing of surgery is a critical aspect of decision-making for surgeons, and treatments that delay the need for arthroplasty may also reduce the need for subsequent revision surgery. As yet, surrogate marks continue to be proposed, but in the absence of long-term follow-up to surgery, the implications of change in surrogate end points remain uncertain.

   • Knee arthroplasty – a nationally representative cohort study is required to understand what proportion of patients with OA (diagnosed in primary care and referred to secondary care) require knee arthroplasty.

Trials of interventions should focus on glucosamine sulphate and the Rotta product is the only product to date that has demonstrated effectiveness. While uncertainty about other preparations remains, there was insufficient evidence of effectiveness and it was not possible to develop an economic case for further study at this time. Any trial should:

1. include collection of information about co-prescribing, the use of other interventions and adverse events

2. recruit obese and overweight participants and people across stages of OA severity

3. use the opportunity to gather a number of measures of joint structure and damage

4. be of at least 3 years’ follow-up with a mechanism to follow the cohort long term (e.g. through record linkage to hospital data).

In preparing the background and undertaking the clinical effectiveness review and economics modelling, it was apparent that there was a lack of information about the epidemiology, management and natural history of OA of the knee in the UK. Greater understanding of the incidence and prevalence in the community, as presenting to primary care and as referred to specialist services, would benefit any future technology assessment in this clinical area. Alongside this, information about current care should be obtained. Some of this information could be obtained using routine primary care databases, but would require supplementation to estimate the impact in the community and in secondary care.

A number of studies noted that expectations, coping and tolerance of symptoms varied between individuals and had substantial impacts on their utilisation of services. Further research to understand the impact of such patient factors would help in the planning of services and target need (versus demand).

The biological mechanism of glucosamine sulphate and chondroitin remains uncertain. In particular, the proposal that the active substance may be sulphate should be explored further. Other sulphate-rich supplements might provide a useful comparator, e.g. glutathione and l-methionine. Paracetamol, metabolised by sulphation may potentially lead to sulphate deficiency, and therefore a better understanding of this potential mechanism has important clinical implications. Most work has focused on the effect of compounds on the chondrocyte and cartilage. Further work is needed to explore the impact on subchondral bone, which is increasingly seen as the primary disease initiator.

Conclusion

There was evidence that glucosamine sulphate shows some clinical effectiveness in the treatment of OA of the knee. Cost-effectiveness was not conclusively demonstrated, but substantial uncertainty was observed for some key determinants in the model. There was evidence from biological studies to support the potential clinical impact of glucosamine sulphate. For other preparations, the evidence base was less consistent (chondroitin) or absent (glucosamine hydrochloride).

Further research to reduce uncertainty and confirm the impact on surgical outcomes in a UK clinical setting would be beneficial.

Contribution of authors

Corri Black (Senior Clinical Lecturer) led the review and drafted the final report. Christine Clar (Systematic Reviewer) led the review of clinical effectiveness and drafted this chapter. Rob Henderson (Consultant in Public Health Medicine) drafted sections of the background. Campbell MacEachern (Specialist Registrar) drafted sections of the background and the chapter on biological plausibility. Paul McNamee (Senior Research Fellow) and Zahidul Quayyum (Research Fellow) undertook the cost-effectiveness analysis and drafted this section. Pam Royle (Senior Research Fellow) developed the search strategy and undertook all searching. Sian Thomas (Systematic Reviewer) data extracted and contributed to the draft clinical effectiveness chapter. All authors read and commented on the final draft.

Disclaimers

The views expressed in this publication are those of the authors and not necessarily those of the HTA programme or the Department of Health.

Health Technology Assessment reports published to date

1. Home parenteral nutrition: a systematic review.

   By Richards DM, Deeks JJ, Sheldon TA, Shaffer JL.

2. Diagnosis, management and screening of early localised prostate cancer.

   A review by Selley S, Donovan J, Faulkner A, Coast J, Gillatt D.

3. The diagnosis, management, treatment and costs of prostate cancer in England and Wales.

   A review by Chamberlain J, Melia J, Moss S, Brown J.

4. Screening for fragile X syndrome.

   A review by Murray J, Cuckle H, Taylor G, Hewison J.

5. A review of near patient testing in primary care.

   By Hobbs FDR, Delaney BC, Fitzmaurice DA, Wilson S, Hyde CJ, Thorpe GH, et al.

6. Systematic review of outpatient services for chronic pain control.

   By McQuay HJ, Moore RA, Eccleston C, Morley S, de C Williams AC.

7. Neonatal screening for inborn errors of metabolism: cost, yield and outcome.

   A review by Pollitt RJ, Green A, McCabe CJ, Booth A, Cooper NJ, Leonard JV, et al.

8. Preschool vision screening.

   A review by Snowdon SK, Stewart-Brown SL.

9. Implications of socio-cultural contexts for the ethics of clinical trials.

   A review by Ashcroft RE, Chadwick DW, Clark SRL, Edwards RHT, Frith L, Hutton JL.

10. A critical review of the role of neonatal hearing screening in the detection of congenital hearing impairment.

    By Davis A, Bamford J, Wilson I, Ramkalawan T, Forshaw M, Wright S.

11. Newborn screening for inborn errors of metabolism: a systematic review.

    By Seymour CA, Thomason MJ, Chalmers RA, Addison GM, Bain MD, Cockburn F, et al.

12. Routine preoperative testing: a systematic review of the evidence.

    By Munro J, Booth A, Nicholl J.

13. Systematic review of the effectiveness of laxatives in the elderly.

    By Petticrew M, Watt I, Sheldon T.

14. When and how to assess fast-changing technologies: a comparative study of medical applications of four generic technologies.

    A review by Mowatt G, Bower DJ, Brebner JA, Cairns JA, Grant AM, McKee L.

1. Antenatal screening for Down’s syndrome.

   A review by Wald NJ, Kennard A, Hackshaw A, McGuire A.

2. Screening for ovarian cancer: a systematic review.

   By Bell R, Petticrew M, Luengo S, Sheldon TA.

3. Consensus development methods, and their use in clinical guideline development.

   A review by Murphy MK, Black NA, Lamping DL, McKee CM, Sanderson CFB, Askham J, et al.

4. A cost–utility analysis of interferon beta for multiple sclerosis.

   By Parkin D, McNamee P, Jacoby A, Miller P, Thomas S, Bates D.

5. Effectiveness and efficiency of methods of dialysis therapy for end-stage renal disease: systematic reviews.

   By MacLeod A, Grant A, Donaldson C, Khan I, Campbell M, Daly C, et al.

6. Effectiveness of hip prostheses in primary total hip replacement: a critical review of evidence and an economic model.

   By Faulkner A, Kennedy LG, Baxter K, Donovan J, Wilkinson M, Bevan G.

7. Antimicrobial prophylaxis in colorectal surgery: a systematic review of randomised controlled trials.

   By Song F, Glenny AM.

8. Bone marrow and peripheral blood stem cell transplantation for malignancy.

   A review by Johnson PWM, Simnett SJ, Sweetenham JW, Morgan GJ, Stewart LA.

9. Screening for speech and language delay: a systematic review of the literature.

   By Law J, Boyle J, Harris F, Harkness A, Nye C.

10. Resource allocation for chronic stable angina: a systematic review of effectiveness, costs and cost-effectiveness of alternative interventions.

    By Sculpher MJ, Petticrew M, Kelland JL, Elliott RA, Holdright DR, Buxton MJ.

11. Detection, adherence and control of hypertension for the prevention of stroke: a systematic review.

    By Ebrahim S.

12. Postoperative analgesia and vomiting, with special reference to day-case surgery: a systematic review.

    By McQuay HJ, Moore RA.

13. Choosing between randomised and nonrandomised studies: a systematic review.

    By Britton A, McKee M, Black N, McPherson K, Sanderson C, Bain C.

14. Evaluating patient-based outcome measures for use in clinical trials.

    A review by Fitzpatrick R, Davey C, Buxton MJ, Jones DR.

15. Ethical issues in the design and conduct of randomised controlled trials.

    A review by Edwards SJL, Lilford RJ, Braunholtz DA, Jackson JC, Hewison J, Thornton J.

16. Qualitative research methods in health technology assessment: a review of the literature.

    By Murphy E, Dingwall R, Greatbatch D, Parker S, Watson P.

17. The costs and benefits of paramedic skills in pre-hospital trauma care.

    By Nicholl J, Hughes S, Dixon S, Turner J, Yates D.

18. Systematic review of endoscopic ultrasound in gastro-oesophageal cancer.

    By Harris KM, Kelly S, Berry E, Hutton J, Roderick P, Cullingworth J, et al.

19. Systematic reviews of trials and other studies.

    By Sutton AJ, Abrams KR, Jones DR, Sheldon TA, Song F.

20. Primary total hip replacement surgery: a systematic review of outcomes and modelling of cost-effectiveness associated with different prostheses.

    A review by Fitzpatrick R, Shortall E, Sculpher M, Murray D, Morris R, Lodge M, et al.

1. Informed decision making: an annotated bibliography and systematic review.

   By Bekker H, Thornton JG, Airey CM, Connelly JB, Hewison J, Robinson MB, et al.

2. Handling uncertainty when performing economic evaluation of healthcare interventions.

   A review by Briggs AH, Gray AM.

3. The role of expectancies in the placebo effect and their use in the delivery of health care: a systematic review.

   By Crow R, Gage H, Hampson S, Hart J, Kimber A, Thomas H.

4. A randomised controlled trial of different approaches to universal antenatal HIV testing: uptake and acceptability. Annex: Antenatal HIV testing – assessment of a routine voluntary approach.

   By Simpson WM, Johnstone FD, Boyd FM, Goldberg DJ, Hart GJ, Gormley SM, et al.

5. Methods for evaluating area-wide and organisation-based interventions in health and health care: a systematic review.

   By Ukoumunne OC, Gulliford MC, Chinn S, Sterne JAC, Burney PGJ.

6. Assessing the costs of healthcare technologies in clinical trials.

   A review by Johnston K, Buxton MJ, Jones DR, Fitzpatrick R.

7. Cooperatives and their primary care emergency centres: organisation and impact.

   By Hallam L, Henthorne K.

8. Screening for cystic fibrosis.

   A review by Murray J, Cuckle H, Taylor G, Littlewood J, Hewison J.

9. A review of the use of health status measures in economic evaluation.

   By Brazier J, Deverill M, Green C, Harper R, Booth A.

10. Methods for the analysis of quality-of-life and survival data in health technology assessment.

    A review by Billingham LJ, Abrams KR, Jones DR.

11. Antenatal and neonatal haemoglobinopathy screening in the UK: review and economic analysis.

    By Zeuner D, Ades AE, Karnon J, Brown J, Dezateux C, Anionwu EN.

12. Assessing the quality of reports of randomised trials: implications for the conduct of meta-analyses.

    A review by Moher D, Cook DJ, Jadad AR, Tugwell P, Moher M, Jones A, et al.

13. ‘Early warning systems’ for identifying new healthcare technologies.

    By Robert G, Stevens A, Gabbay J.

14. A systematic review of the role of human papillomavirus testing within a cervical screening programme.

    By Cuzick J, Sasieni P, Davies P, Adams J, Normand C, Frater A, et al.

15. Near patient testing in diabetes clinics: appraising the costs and outcomes.

    By Grieve R, Beech R, Vincent J, Mazurkiewicz J.

16. Positron emission tomography: establishing priorities for health technology assessment.

    A review by Robert G, Milne R.

17. The debridement of chronic wounds: a systematic review.

    By Bradley M, Cullum N, Sheldon T.

18. Systematic reviews of wound care management: (2) Dressings and topical agents used in the healing of chronic wounds.

    By Bradley M, Cullum N, Nelson EA, Petticrew M, Sheldon T, Torgerson D.

19. A systematic literature review of spiral and electron beam computed tomography: with particular reference to clinical applications in hepatic lesions, pulmonary embolus and coronary artery disease.

    By Berry E, Kelly S, Hutton J, Harris KM, Roderick P, Boyce JC, et al.

20. What role for statins? A review and economic model.

    By Ebrahim S, Davey Smith G, McCabe C, Payne N, Pickin M, Sheldon TA, et al.

21. Factors that limit the quality, number and progress of randomised controlled trials.

    A review by Prescott RJ, Counsell CE, Gillespie WJ, Grant AM, Russell IT, Kiauka S, et al.

22. Antimicrobial prophylaxis in total hip replacement: a systematic review.

    By Glenny AM, Song F.

23. Health promoting schools and health promotion in schools: two systematic reviews.

    By Lister-Sharp D, Chapman S, Stewart-Brown S, Sowden A.

24. Economic evaluation of a primary care-based education programme for patients with osteoarthritis of the knee.

    A review by Lord J, Victor C, Littlejohns P, Ross FM, Axford JS.

1. The estimation of marginal time preference in a UK-wide sample (TEMPUS) project.

   A review by Cairns JA, van der Pol MM.

2. Geriatric rehabilitation following fractures in older people: a systematic review.

   By Cameron I, Crotty M, Currie C, Finnegan T, Gillespie L, Gillespie W, et al.

3. Screening for sickle cell disease and thalassaemia: a systematic review with supplementary research.

   By Davies SC, Cronin E, Gill M, Greengross P, Hickman M, Normand C.

4. Community provision of hearing aids and related audiology services.

   A review by Reeves DJ, Alborz A, Hickson FS, Bamford JM.

5. False-negative results in screening programmes: systematic review of impact and implications.

   By Petticrew MP, Sowden AJ, Lister-Sharp D, Wright K.

6. Costs and benefits of community postnatal support workers: a randomised controlled trial.

   By Morrell CJ, Spiby H, Stewart P, Walters S, Morgan A.

7. Implantable contraceptives (subdermal implants and hormonally impregnated intrauterine systems) versus other forms of reversible contraceptives: two systematic reviews to assess relative effectiveness, acceptability, tolerability and cost-effectiveness.

   By French RS, Cowan FM, Mansour DJA, Morris S, Procter T, Hughes D, et al.

8. An introduction to statistical methods for health technology assessment.

   A review by White SJ, Ashby D, Brown PJ.

9. Disease-modifying drugs for multiple sclerosis: a rapid and systematic review.

   By Clegg A, Bryant J, Milne R.

10. Publication and related biases.

    A review by Song F, Eastwood AJ, Gilbody S, Duley L, Sutton AJ.

11. Cost and outcome implications of the organisation of vascular services.

    By Michaels J, Brazier J, Palfreyman S, Shackley P, Slack R.

12. Monitoring blood glucose control in diabetes mellitus: a systematic review.

    By Coster S, Gulliford MC, Seed PT, Powrie JK, Swaminathan R.

13. The effectiveness of domiciliary health visiting: a systematic review of international studies and a selective review of the British literature.

    By Elkan R, Kendrick D, Hewitt M, Robinson JJA, Tolley K, Blair M, et al.

14. The determinants of screening uptake and interventions for increasing uptake: a systematic review.

    By Jepson R, Clegg A, Forbes C, Lewis R, Sowden A, Kleijnen J.

15. The effectiveness and cost-effectiveness of prophylactic removal of wisdom teeth.

    A rapid review by Song F, O’Meara S, Wilson P, Golder S, Kleijnen J.

16. Ultrasound screening in pregnancy: a systematic review of the clinical effectiveness, cost-effectiveness and women’s views.

    By Bricker L, Garcia J, Henderson J, Mugford M, Neilson J, Roberts T, et al.

17. A rapid and systematic review of the effectiveness and cost-effectiveness of the taxanes used in the treatment of advanced breast and ovarian cancer.

    By Lister-Sharp D, McDonagh MS, Khan KS, Kleijnen J.

18. Liquid-based cytology in cervical screening: a rapid and systematic review.

    By Payne N, Chilcott J, McGoogan E.

19. Randomised controlled trial of non-directive counselling, cognitive–behaviour therapy and usual general practitioner care in the management of depression as well as mixed anxiety and depression in primary care.

    By King M, Sibbald B, Ward E, Bower P, Lloyd M, Gabbay M, et al.

20. Routine referral for radiography of patients presenting with low back pain: is patients’ outcome influenced by GPs’ referral for plain radiography?

    By Kerry S, Hilton S, Patel S, Dundas D, Rink E, Lord J.

21. Systematic reviews of wound care management: (3) antimicrobial agents for chronic wounds; (4) diabetic foot ulceration.

    By O’Meara S, Cullum N, Majid M, Sheldon T.

22. Using routine data to complement and enhance the results of randomised controlled trials.

    By Lewsey JD, Leyland AH, Murray GD, Boddy FA.

23. Coronary artery stents in the treatment of ischaemic heart disease: a rapid and systematic review.

    By Meads C, Cummins C, Jolly K, Stevens A, Burls A, Hyde C.

24. Outcome measures for adult critical care: a systematic review.

    By Hayes JA, Black NA, Jenkinson C, Young JD, Rowan KM, Daly K, et al.

25. A systematic review to evaluate the effectiveness of interventions to promote the initiation of breastfeeding.

    By Fairbank L, O’Meara S, Renfrew MJ, Woolridge M, Sowden AJ, Lister-Sharp D.

26. Implantable cardioverter defibrillators: arrhythmias. A rapid and systematic review.

    By Parkes J, Bryant J, Milne R.

27. Treatments for fatigue in multiple sclerosis: a rapid and systematic review.

    By Brañas P, Jordan R, Fry-Smith A, Burls A, Hyde C.

28. Early asthma prophylaxis, natural history, skeletal development and economy (EASE): a pilot randomised controlled trial.

    By Baxter-Jones ADG, Helms PJ, Russell G, Grant A, Ross S, Cairns JA, et al.

29. Screening for hypercholesterolaemia versus case finding for familial hypercholesterolaemia: a systematic review and cost-effectiveness analysis.

    By Marks D, Wonderling D, Thorogood M, Lambert H, Humphries SE, Neil HAW.

30. A rapid and systematic review of the clinical effectiveness and cost-effectiveness of glycoprotein IIb/IIIa antagonists in the medical management of unstable angina.

    By McDonagh MS, Bachmann LM, Golder S, Kleijnen J, ter Riet G.

31. A randomised controlled trial of prehospital intravenous fluid replacement therapy in serious trauma.

    By Turner J, Nicholl J, Webber L, Cox H, Dixon S, Yates D.

32. Intrathecal pumps for giving opioids in chronic pain: a systematic review.

    By Williams JE, Louw G, Towlerton G.

33. Combination therapy (interferon alfa and ribavirin) in the treatment of chronic hepatitis C: a rapid and systematic review.

    By Shepherd J, Waugh N, Hewitson P.

34. A systematic review of comparisons of effect sizes derived from randomised and non-randomised studies.

    By MacLehose RR, Reeves BC, Harvey IM, Sheldon TA, Russell IT, Black AMS.

35. Intravascular ultrasound-guided interventions in coronary artery disease: a systematic literature review, with decision-analytic modelling, of outcomes and cost-effectiveness.

    By Berry E, Kelly S, Hutton J, Lindsay HSJ, Blaxill JM, Evans JA, et al.

36. A randomised controlled trial to evaluate the effectiveness and cost-effectiveness of counselling patients with chronic depression.

    By Simpson S, Corney R, Fitzgerald P, Beecham J.

37. Systematic review of treatments for atopic eczema.

    By Hoare C, Li Wan Po A, Williams H.

38. Bayesian methods in health technology assessment: a review.

    By Spiegelhalter DJ, Myles JP, Jones DR, Abrams KR.

39. The management of dyspepsia: a systematic review.

    By Delaney B, Moayyedi P, Deeks J, Innes M, Soo S, Barton P, et al.

40. A systematic review of treatments for severe psoriasis.

    By Griffiths CEM, Clark CM, Chalmers RJG, Li Wan Po A, Williams HC.