Results of the search

Review authors found a total of 187 potential studies in the first search for this systematic review. Of these 187, 57 studies were obtained on 1 January 2006 for further investigation (Figure 2). Of these, 19 studies met the inclusion criteria. The remaining studies were excluded (refer to "Characteristics of excluded studies"). We performed a second search via MEDLINE on 12 September 2006 and identified 26 additional potential studies. Of these 26, two studies qualified for inclusion. We performed another search on 4 June 2007. This search yielded 12 results, of which three qualified for inclusion in our review. We performed another search on 21 May 2008. This time, we identified 15 additional references, of which one study qualified. On 30 June 2009, we performed another search that identified 32 studies. Of the 32 references pulled in the MEDLINE search, 15 were duplicates that had been pulled in the 21 May 2008 search. Of the remaining 17 new references, six qualified as potential studies for inclusion in the meta‐analysis. Of these six studies, two qualified. One of our Russian translators identified two additional studies for inclusion in our review. Later, on 20 July 2011, we performed another search, which yielded 29 results. Of these 29 potential studies, six fit the inclusion criteria. One study (Kahan 2009) was a duplicate from our previous search; therefore, five studies were picked for inclusion. On 12 June 2012, another search was performed to make sure that we included all studies that had been published up to the completion of our meta‐analysis. This search yielded 764 results. Of these, sixteen were pulled for further review. Upon examination, five of these studies were deemed to fit the inclusion criteria and were added to our meta‐analysis. A review of other meta‐analyses performed on chondroitin revealed seven new possible studies for inclusion, of which two fit our inclusion criteria. A final search update was done in November, 2013 that yielded 794 titles/abstracts. An additional 4 studies were included from this search, of which one was a long‐term follow‐up of a previously included study by Wildi 2011A, but with a different outcome. Overall, therefore, 45 studies were included in our review. One study, Brandao 2009, could not be obtained, and another study, Vertkin 2007, was found to have an obvious mismatch in its data, as reported by the translating abstractor. Both studies were therefore excluded, yielding a total of 43 studies to be included in this systematic review and meta‐analysis.

Figure 2

---

---

Figure 2. Study Flow chart

Included studies

See Figure 2 for a flow diagram of the search results. The "Characteristics of included studies table" provides further information about the included studies.

A total of 43 studies with 9,110 participants were included in this review. In all, 4,962 participants were treated with chondroitin (with/without glucosamine, NSAIDs) and 4,148 participants were included in the control group. Trial duration varied 1 month to 3 years. Data were usable from 30 studies in the meta‐analysis. Examples of some studies with non‐usable data were as follows: the treatment regimens were too different from the cohort of included studies (Alekseeva 2008; Cohen 2003; Magrans‐Courtney 2011; Nguyen 2001), or they did not report clinical data (Rovetta 2002; Rovetta 2004).

1. Design

All included studies were randomized trials. Most were also blinded, with very few exceptions (Alekseeva 2005a; Alekseeva 2008; Lila 2005). Most studies were parallel‐arm trials.

2. Sample sizes

Sample sizes ranged from ≤ 30 participants in a few trials (e.g., Magrans‐Courtney 2011; Rovetta 2002) to 1,583 in the GAIT trial (Clegg 2006). Most studies had a sample size larger than 50 participants; in particular, several studies had sample sizes of 100 or more (Alekseeva 1999; Alekseeva 2005a; Alekseeva 2008; Clegg 2006; Gabay 2011; Kahan 2009; L'Hirondel 1992; Mazieres 2001; Michel 2005; Moller 2010; Morreale 1996; Nasonova 2001; Pavelka 1999; Pavelka 2010; Rai 2004; Sawitzke 2008; Sawitzke 2010; Uebelhart 2004; Fardellone 2013).

3. Setting

Many studies were single‐center studies (e.g., Magrans‐Courtney 2011; Nakasone 2011; Rovetta 2002; Uebelhart 2004). Other studies were multicenter studies (e.g., Clegg 2006; Uebelhart 2004; Wildi 2011).

4. Participants

Most studies included participants with knee osteoarthritis, with a few exceptions; a few studies included participants with hand osteoarthritis (Gabay 2011; Rovetta 2002; Rovetta 2004; Verbruggen 2002); one study included participants with hip osteoarthritis (Conrozier 1998).

5. Interventions

One study used a cream that included chondroitin (Cohen 2003); this study was not included in the analyses. Other studies administered oral chondroitin. Most studies gave chondroitin at 800 mg daily dose or higher, with notable exceptions such as Kanzaki 2011, Nakasone 2011; Nguyen 2001; Pavelka 1999; Rai 2004, in which a dose of chondroitin < 800 mg/d was used.

6. Outcomes

All studies had pain and function or quality of life as primary or secondary outcome. Two studies examined only radiographic outcomes (Rovetta 2002; Sawitzke 2008).

A summary of studies is provided below.

Alekseeva 1999 was a parallel arm RCT that included 100 patients, 50 each randomzied to chondroitin sulfate and control group in patients aged 45 and older, with knee osteoarthritis diagnosed according to ACR (1986) criteria and with Kellgren‐Lawrence X‐ray grade II or III. They had moderate to severe pain (> 30 mm on 100‐mm VAS), Lequese’s index 4 to 11 points, and had taken NSAIDs for at least 30 days over previous three months. Patients received either chondroitin sulfate 1000 mg/d (two capsules of 250 mg twice a day) with meals or ibuprofen up to 1200 mg/d (no rules for dose selection given). Paracetamol was allowed as rescue analgesia. The outcomes included VAS Pain (walking/rest), Lequesne's index, Patient global assessment (% improved), Patient global assessment on VAS and concomitant medication use. The study was sponsored by a manufacturer of chondroitin sulfate.

Alekseeva 2005a was a randomized, parallel, independent group study involving 90 participants (45 in the control group, 45 in the active chondroitin sulfate group) that compared 1000 mg and 500 mg of chondroitin sulfate with 100 mg diclofenac daily for six months. Participants were included if they were diagnosed with knee osteoarthritis according to American College of Rheumatology (ACR) 1991 criteria, had a Kellgren‐Lawrence X‐ray grade II or III, moderate to severe pain on motion assessed on a 0 to 100‐mm VAS scale (> 40 mm), and had been taking nonsteroidal anti‐inflammatory drugs for at least 30 days in the previous three months. The active group consisted of 45 participants who were treated with 1000 mg of chondroitin sulfate daily for the first month, then with 500 mg of chondroitin sulfate daily for the next five months. The control group was treated with 100 mg diclofenac daily. Outcomes assessed included (1) Total WOMAC score, (2) Patient global assessment, (3) Physician global assessment, and (4) Adverse events.

Alekseeva 2005b was an open‐label, randomized, multicenter clinical trial involving 375 participants for six months. Participants were included if they were diagnosed with knee osteoarthritis according to ACR criteria, were in the Kellgren‐Lawrence X‐ray grade of II or III, had pain on walking of greater than 40 mm on the 0 to 100‐mm VAS, and had taken nonsteroidal anti‐inflammatory drugs for at least 30 days in the past three months. A total of 203 participants were randomly assigned to the chondroitin sulfate in combination with glucosamine group, and 172 participants to the control group. Participants assigned to the chondroitin sulfate in combination with glucosamine treatment group received 1000 mg of chondroitin sulfate daily plus 1000 mg of glucosamine daily for the first month, then 500 mg chondroitin sulfate and 500 mg glucosamine daily for the next five months. Participants in the active group were also allowed up to 100 mg diclofenac daily with dose reductions and cessation permitted with restrictions. The control group was treated with 100 mg of diclofenac daily for six months. This group was also allowed dose reduction and cessation of the use of diclofenac without restrictions. Outcomes reported were WOMAC Total, pain, physical function, and stiffness.

Alekseeva 2008 was an open‐label, comparative, randomized, independent, parallel‐group experimental design clinical trial that compared the effects of intermittent versus constant therapy with chondroitin sulfate and glucosamine. Participants were included in the study if they were diagnosed with RA according to the ACR (1991) criteria, were in the Kellgren‐Lawrence X‐ray grade II or III, had pain on walking of greater than 40 mm on the 0 to 100‐mm VAS, and had taken NSAIDs for at least 30 days over the previous three months. A total of 50 participants were randomly assigned to the active group, which was treated with chondroitin sulfate combined with glucosamine for nine months daily. In the comparator group, chondroitin sulfate and glucosamine were taken for two three‐month cycles with a three‐month treatment‐free interval between cycles. Participants were also allowed 1200 mg of Ibuprofen taken in 400‐mg capsules three times daily, with dose reductions permitted at the physician's discretion in both groups. The study reported on the following outcomes: WOMAC pain, performance‐based physical function, Patient and Physician global assessments, knee ultrasound, and adverse events.

Artemenko 2005 was a nine‐month, open‐label, comparative, parallel‐design trial that assessed the efficacy of 1000 mg of chondroitin sulfate in combination with 1000 mg of glucosamine for one month, followed by five months of treatment with 500 mg of chondroitin sulfate and 500 mg of glucosamine chloride in comparison with treatment with 50 to 100 mg of diclofenac. A total of 31 participants were randomly assigned to the active arm, and 16 participants were randomly assigned to the control arm. Participants diagnosed with knee osteoarthritis according to ACR criteria, with Kellgren‐Lawrence X‐ray grade II or III, and with pain on walking greater than 40 mm on the 100‐mm VAS were included in the study. Participants with pain on walking less than 40 mm on the 100‐mm VAS, with secondary osteoarthritis, or receiving treatment with other “chondroprotective drugs” (not specified) during the six months before commencement of the trial were excluded from the study. Outcomes of interest in this study were pain, stiffness, Total WOMAC scores, rest and motion pain on VAS, Lequesne's index score, and Patient and Physician global assessments.

Bourgeois 1998 was a three‐month, phase III, randomized, participant‐ and investigator‐blind, double‐dummy, parallel‐group clinical trial that assessed the effects of chondroitin sulfate 4&6 in participants with internal or external, femoral‐tibial knee osteoarthritis. Participants of either sex, aged 45 or older, with femoral tibial, unilateral or bilateral knee osteoarthritis grade I to III diagnosed according to ACR criteria. Participants must have also required stable daily administration of one of the authorized NSAIDs for at least one month before the trial. All participants meeting the inclusion criteria were randomly assigned to one of three groups: chondroitin sulfate 4&6 1200 mg daily (N = 40), CS 4&6 400 mg three times daily (N = 43), or placebo (N = 44). The primary outcome of interest in the study was Lequesne's index. Secondary outcomes included spontaneous pain on 0 to 100‐mm VAS, Patient and Physician global assessments, consumption of NSAIDs, adverse effects, withdrawals, and deaths.

Bucsi 1998 was a randomized, participant‐ and investigator‐blind, placebo‐controlled studying involving two centers and lasting six months. A total of 39 and 46 participants were randomly assigned to the active and control groups, respectively. Participants in the chondroitin sulfate group were treated with 400 mg of chondroitin sulfate twice daily for six months. Participants in the control group were given an identical administration of placebo. Participants were included in the study if they were hospitalized patients or outpatients with idiopathic or secondary clinically symptomatic knee osteoarthritis for longer than six months who showed upon entry a Kellgren‐Lawrence radiologic score of I to III. The primary end point was spontaneous joint pain on a 0 to 100‐mm VAS when pain during daily physical activity was considered. Secondary outcomes were paracetamol consumption, time taken for a 20‐m walk on a flat surface, pain, maximal walking distance, discomfort in daily life movements, Lequesne's index, and Patient and Physician global assessment.

Clegg 2006 was a multicenter, participant‐ and investigator‐blind, placebo‐ and celecoxib‐controlled Glucosamine/chondroitin Arthritis Intervention Trial (GAIT). Participants were included in the trial if they were at least 40 years old, showed clinical and radiographic evidence of osteoarthritis, had a summed pain score of 125 to 400 on the index knee according to WOMAC, and belonged to ARA functional class I, II, or III. Participants were randomly assigned to one of five treatment groups: 1500 mg of glucosamine daily, 1200 mg of chondroitin sulfate daily, both glucosamine and chondroitin sulfate daily, 200 mg of celecoxib daily, or placebo. A total of 313 participants were assigned to the placebo group, 317 to the glucosamine alone group, 318 to the chondroitin group, 317 to the glucosamine and chondroitin sulfate combination group, and 318 to the celecoxib group. The primary outcome for this study was a 20% decrease in the summed score for the WOMAC pain subscale from baseline to week 24. Secondary outcomes for this study were WOMAC stiffness and physical function subscale scores; Patient and Physician global assessment of response to therapy and pain; presence or absence of soft tissue swelling, effusion, or both in the index knee; SF‐36 scores; and analgesic use. Data used for global assessments were Patient and Physician global assessments of "response to therapy scores."

Cohen 2003 was a single‐center, randomized, participant‐ and investigator‐blind, placebo‐controlled trial lasting eight weeks. The study assessed the effects of chondroitin sulfate preparation cream. Participants diagnosed with osteoarthritis of the knee according to ACR criteria had knee pain due to osteoarthritis greater than 4 cm on a 0 to 10‐cm VAS in one or both knees for longer than four weeks. A total of 30 participants were randomly assigned to the active cream group, and 29 were randomly assigned to the inactive cream group. The treatment cream consisted of 30 mg/g of glucosamine sulfate, 50 mg/g of chondroitin sulfate, and 140 mg/g of shark cartilage, of which 10% to 30% is chondroitin sulfate, 32 mg/g is camphor, and 9 mg/g is peppermint oil scent. The inactive cream was a simple conventional cosmetic cream of identical scent and appearance. The primary outcome of this study was pain on a 0 to 100‐mm VAS. Secondary outcomes were pain, stiffness, and physical function on WOMAC and the SF‐36 questionnaire.

Conrozier 1992 was a participant‐ and investigator‐blind, placebo‐controlled study that took place in France for six months. Participants were included in the study if they were diagnosed with hip joint arthrosis with narrowing but intra‐articular space (degree I, II, III) and pain of the hip joint requiring regular analgesics or NSAIDs. A total of 56 participants were randomly assigned to each of the active and treatment groups Participants in the treatment group received three capsules of 400 mg of chondroitin sulfate, and those in the placebo group received an identical administration of placebo capsules. The primary outcomes of the study were Lequesne's index, pain relief on Huskisson's VAS, consumption of analgesics or nonsteroidal antirheumatics (NSARs), and participant evaluation. Secondary outcomes included morning stiffness, maximum walking distance, frequency of waking at night, and intermalleolar distance.

Das 2000 was a randomized, placebo‐controlled, participant‐ and investigator‐blind study. Participants were included if they were in grade II, III, or IV of the Kellgren‐Lawrence scale, were between the ages of 45 and 75, were able to walk, were symptomatic for longer than six months, and had unilateral or bilateral osteoarthritis. Included participants must have been willing to comply with the protocol. A total of 46 participants were assigned to the chondroitin sulfate and glucosamine group and received two capsules of 500 mg of glucosamine and 400 mg of chondroitin sulfate plus 76 mg of manganese. Those in the placebo group received placebo tablets. The primary outcomes in this study were Lequesne's index, WOMAC score, and Patient global assessment. Secondary outcomes were consumed amounts of rescue pain medications and adverse events.

Debi 2000 was a randomized, participant‐ and investigator‐blind study that compared glucosamine sulfate and chondroitin sulfate versus placebo and lasted one month. The original language of this study was Hebrew; thus we acquired a translator and translated the article. All study participants had osteoarthritis; 36 participants were randomly assigned to the active chondroitin sulfate group and 20 to the inactive group. All patients "suffering from osteoarthritis" were included in the study. Outcomes of the study were Patient global assessment, performance‐based physical function, and radiographic changes on X‐ray.

Fardellone 2013 was a multicenter, prospective, randomized, double‐blind, double‐placebo, active‐controlled, parallel group study performed in patients with symptomatic knee OA that compared two chondroitin preparations, Structum® 500 mg twice daily or Chondrosulf® 400 mg three times daily over a 24‐week duration. Inclusion criteria were as follows: age 50 to 80 years, medial and/or lateral femoro‐tibial OA of the knee according to American College of Rheumatology (ACR) criteria, symptomatic for more than 6 months, with a baseline level of symptoms as follows, global pain score on a Visual Analog Scale (VAS 0 to 100) of at least 40 millimeters (mm), a Lequesne’s algofunctional index (LFI 0 to 24) score greater than or equal to 7, radiographic OA as defined by a Kellgren‐Lawrence grade II or III on an antero‐posterior weight‐bearing view of both knees taken. Primary outcome was global pain on 0 to 100 mm VAS. Secondary outcomes included patient’s and investigator’s global assessment scores (VAS), Osteoarthritis of the knee or hip Quality of Life dimension score (OAKHQOL), Short‐form 12 (SF‐12) scores and the use of analgesic medications (paracetamol and/or NSAIDs).

Gabay 2011 was an investigator‐initiated, single‐center, randomized, participant‐ and investigator‐blind, placebo‐controlled clinical trial that examined the efficacy of chondroitin in participants with hand osteoarthritis. Participants were included in the trial if they were at least 40 years old and fulfilled the ACR criteria for the classification of hand osteoarthritis. In addition, radiographic features of hand osteoarthritis affecting at least two joints of the target hand on standard plain films obtained within six months of enrolment and at least two painful flares of osteoarthritis in the finger joints during the previous 12 months were required. The target hand was defined as the participant's most symptomatic hand or, when both hands were equally painful, the participant's dominant hand. To be eligible for the study, participants had to have presented with symptomatic osteoarthritis. Minimum symptoms included joint pain of at least 40 mm on a 0 to 100‐mm VAS and a Functional Index for Hand Osteoarthritis (FIHOA) score of at least 6 in the target hand (0 to 30 scale). The study lasted for six months. A total of 80 participants were randomly assigned to receive 800 mg of chondroitin sulfate daily. In all, 82 participants were randomly assigned to receive placebo sachets. The primary outcomes of the study were change in the Patient’s assessment of global spontaneous hand pain on VAS and change in Patient’s assessment of global spontaneous hand function on the FIHOA score from baseline to month six. Secondary outcomes were grip strength and duration of morning stiffness.

Kahan 2009 is an international, randomized, participant‐ and investigator‐blind, placebo‐controlled trial that lasted two years. Participants were included in the trial if they were outpatients between 40 and 80 years old who had been diagnosed with primary knee osteoarthritis of the medial tibiofemoral compartment according to ACR criteria. A total of 309 participants were randomly assigned to the active chondroitin group, which received 800 mg of chondroitin sulfate daily for two years. In all, 313 participants were randomly assigned to the placebo group and received placebo sachets of identical appearance and regimen. The primary outcome of the study was pain modification in the minimum joint space width of the medial compartment of the target tibiofemoral joint. Secondary outcomes included assessment of pain using a 0 to 100‐mm VAS, WOMAC Total and subscale scores, Patient and Physician global assessments, cumulative consumption of acetaminophen, and cumulative consumption of NSAIDs. Results and data for joint space width are taken from Kahan's supplementary report, published in 2006, in which results for joint space narrowing were presented for the placebo and chondroitin sulfate groups.

Kanzaki 2011 was a prospective, randomized, placebo‐controlled, parallel‐group comparative study that was designed to assess the efficacy and safety of a GCQ supplement (that contained chondroitin) for a duration of six months. The study involved two clinical service organization centers that were under the control of two medical investigators in Japan. Male and female Japanese participants between 40 and 85 years of age were diagnosed with knee osteoarthritis, and the presence of knee pain was confirmed by assessment of scores on the "walking" subscale of the Japanese Orthopedic Association (JOA). A total of 20 participants were randomly assigned to receive 1200 mg glucosamine hydrochloride, 300 mg shark cartilage extract (60 mg as chondroitin sulfate), and 45 mg quercetin glycosides in a daily dose of six tablets in the active group. In all, 20 participants were randomly assigned to the placebo group and received indistinguishable placebo tablets, which were administered identically with six tablets a day. The primary outcomes of the study were Japan Orthopaedic Association subscale scores. Subscales documented walking ability, ascending and descending ability and pain on stairs, range of motion, and degree of joint swelling. The study’s secondary outcomes were pain at rest on VAS, pain on walking on VAS, and pain on ascending and descending of stairs on VAS.

L'Hirondel 1992 was a participant‐ and investigator‐blind, parallel‐design, randomized study in which orally administered chondroitin sulfate was compared with placebo in participants with tibiofemoral gonarthrosis for six months; the study was conducted in France. Participants were included if they had painful tibiofemoral gonarthrosis with an intra‐articular space but without dislocation of the main axis and with or without meniscus calcification. A total of 63 participants were randomly assigned to the active chondroitin sulfate group, and 62 participants were randomly assigned to the placebo group. The chondroitin sulfate group received three sachets of 400 mg chondroitin sulfate daily, and the placebo group received three identical sachets of placebo. Participants were allowed to continue their ongoing treatments through the first two months of the study and were then allowed 500 mg of paracetamol from months 2 to 6, according to their pain intensity. The primary outcomes of this study were Lequesne's index, adverse effects, pain relief on Huskisson's VAS, consumption of analgesics and nonsteroidal antirheumatics, and total efficacy as evaluated by the investigator. Secondary outcomes included extent of flexion and extension of the joint, extent of joint circumference, and verification of a static incident, an abnormal movement, or an intra‐articular effusion.

Lila 2005 was a randomized, open efficacy trial that included participants 45 to 75 years of age who were diagnosed with osteoarthritis according to ACR 1987 criteria and were in Kellgren‐Lawrence X‐ray grade I to III. They had a pain score greater than 40 mm on a VAS (0 to 100 mm) for pain on walking and had daily morning stiffness for at least 30 minutes. A total of 30 participants who fulfilled these criteria were randomly assigned to the active chondroitin sulfate in combination with glucosamine and placebo groups. Those assigned to the active group received 800 mg of chondroitin sulfate in combination with 1000 mg of glucosamine daily for the first month, then 400 mg of chondroitin sulfate plus 500 mg of glucosamine daily for the next two months. The control group received diclofenac 75 mg daily. Outcomes of in the study included WOMAC subscales, rest or motion pain on a 0 to 100‐mm VAS, morning stiffness, Patient and Physician global assessments, and adverse events.

Magrans‐Courtney 2011 was a randomized, participant‐ and investigator‐blind, placebo‐controlled, parallel clinical trial that lasted 14 weeks. Women with physician‐diagnosed osteoarthritis between the ages of 18 and 70 years who had a body mass index (BMI) greater than 27 kg/m² and no recent participation in a diet or exercise program were included in the study. A total of 16 participants were assigned to the active treatment arm, in which participants received a total of 1500 mg of glucosamine (from d‐glucosamine HCL), 1200 mg/d of chondroitin sulfate (from chondroitin sulfate sodium), 120 mg/d of niacin, 120 mg/d of sodium, 45 mg/d of zinc, 900 mg/d of MSM, 300 mg/d of Boswellia serrata extract, 180 mg/d of white willow bark extract, and 15 mg/d of rutin powder. A total of 14 participants were randomly assigned to receive placebo of identical appearance and dosage. All randomly assigned participants also participated in a 14‐week circuit‐style workout consisting of 14 exercises (e.g., elbow flexion/extension, knee flexion/extension, shoulder press/lat pull, hip abductor/adductor, chest press/seated row, horizontal leg press, squat, abdominal crunch/back extension, pec deck, oblique, shoulder shrug/dip, hip extension, side bends, stepping three times a week). Participants in each treatment group were then randomly assigned to begin a high‐protein or high‐carbohydrate diet program. Both programs were composed of three phases. During the first phase, which lasted one week, participants were allowed to consume a total of 1200 kcal daily. For the next nine weeks, participants were given 1600 kcal/d. The final four weeks of the diet (Phase III) served as a weight maintenance period. The high‐protein (HP) diet regimen was composed of 7% carbohydrate, 63% protein, and 30% fat during Phase I, and 15% carbohydrate, 55% protein, and 30% fat during Phase II. Upon completion of the high‐carbohydrate diet program, participants were instructed to consume a diet consisting of 55% carbohydrate, 15% protein, and 30% fat. In the final phase of the diet program, all participants were allowed 55% carbohydrate, 15% protein, and 30% fat. The primary outcomes were measures of pain, including pain on VAS, stiffness, physical function, and Total score on the WOMAC scale. Secondary outcomes included weight loss, body composition, measures of quality of life, changes in energy intake, anthropometrics, body composition, resting energy expenditures, cardiovascular and muscular fitness, balance and functional capacity, serum and whole blood clinical markers, hormonal profiles, pain indices, psychosocial parameters, and knee range of motion and circumference.

Mazieres 1992 was a multicenter, randomized, controlled, parallel, independent, participant‐ and investigator‐blind study that lasted three months; participants were 50 years of age or older; had a pain score of at least 40 mm on a 0 to 100‐mm VAS and a score greater than 4 on Lequesne's Index; and were diagnosed with gonarthroses of the femorotibial and internal or external compartment or coxarthrosis. A total of 58 participants were randomly assigned to the active group and received 2000 mg of chondroitin sulfate daily for three months. In all, 56 participants were randomly assigned to the placebo group. Outcomes of interest in the study were rest or motion pain on VAS, Lequesne's index, Patient and Physician global assessments, and NSAID use.

Mazieres 2001 was a participant‐ and investigator‐blind, randomized, parallel‐group study that lasted six months. The study included outpatients older than 50 years of age who met the following criteria: (1) clinically and radiographically confirmed osteoarthritis of one or both knees according to the criteria of the ACR; (2) an algofunctional index of Lequesne (AFI) ≥ 4; (3) pain with activity ≥ 30 mm on a 100‐mm VAS; (4) regular consumption of NSAIDs for three months; and (5) radiographic grade II to III on the Kellgren‐Lawrence scale. In participants with osteoarthritis of both knees, the most painful side was assessed. A total of 67 participants were randomly assigned to the active chondroitin sulfate group and received 500 mg of chondroitin sulfate twice daily for three months. In all, 63 participants were randomly assigned to the placebo treatment group and received placebo sachets. The primary outcome of the study was Lequesne's AFI. Secondary outcomes were pain on physical activity as measured on a 0 to 100‐mm VAS, pain on rest as measured on a 0 to 100‐mm VAS, self‐assessed effect of osteoarthritis on daily activities as measured on a 0 to 100‐mm VAS, participant's overall assessment of change as measured on a five‐point categorical scale (much better, better, unchanged, somewhat worse, much worse), and daily NSAID and analgesics consumption.

Messier 2007 was a participant‐ and investigator‐blind, placebo‐controlled, randomized clinical trial that lasted 12 months, of which six months was applicable to the current review. For the remaining six months, investigators assessed the effect of exercise combined with chondroitin sulfate and glucosamine, and thus the data could not be used. Participants were included if they were 50 years of age or older, had mild to moderate knee osteoarthritis on a Kellgren‐Lawrence scale (II or III) according to ACR criteria, and were not currently participating in another study. A total of 45 participants were randomly assigned to the chondroitin sulfate and glucosamine hydrochloride group and received 1200 mg of chondroitin sulfate plus 1500 mg of glucosamine hydrochloride once or three times daily. In all, 44 participants were randomly assigned to the placebo group and received placebo sachets of identical appearance and odor in a blinded fashion. The primary outcome of the study was physical function as measured on the WOMAC. Secondary outcomes were pain on WOMAC, distance walked in six minutes, mental status as measured on the Mini Mental State Examination (MMSE), balance, and strength of concentric extension and flexion in the most affected knee.

Michel 2005 was a randomized, participant‐ and investigator‐blind, placebo‐controlled trial that lasted two years and was conducted in Zurich, Switzerland. Participants were included in the study if they were between the ages of 40 and 85 years and had clinically symptomatic knee osteoarthritis (knee pain while standing, walking, and/or on motion for at least 25 or 30 days before study entry, with no required minimum level of pain on the day of entry). Partcipants had to be diagnosed with knee osteoarthritis according to ACR clinical and radiographic criteria for osteoarthritis of the knee. Participants with osteoarthritis of grade I, II or III according to Kellgren‐Lawrence (KL) were eligible for study entry. Participants with osteoarthritis of KL grade 4, indicating a greatly narrowed joint space with sclerosis of subchondral bone, were excluded. The target knee was defined as the most symptomatic knee at study entry. A total of 150 participants were randomly assigned to the active chondroitin sulfate group and received 800 mg of chondroitin sulfate once daily for two years. In all, 150 participants were randomly assigned to the placebo group and received placebo sachets. The primary outcome of interest was the minimum and mean joint space width of the more severely affected compartment of the target knee. Secondary outcomes were stiffness and physical function on WOMAC, total consumption of rescue drugs during the trial, average number of tablets of study drug taken per day, and adverse events.

Moller 2010 was a randomized, participant‐ and investigator‐blind, placebo‐controlled multicenter study that lasted three months. Participants eligible for the study were men and women 40 years of age or older who had osteoarthritis of the knee as defined by the criteria of the ACR, pain in the affected knee scored as 30 on a continuous 0 to 100‐mm Huskisson’s VAS, and a confirmatory knee X‐ray diagnosis (Kellgren‐Lawrence grade I to III) associated with cutaneous plaque‐type psoriasis with a Psoriasis Area and Severity Index (PASI) score of 5. A total of 60 participants were randomly assigned to receive 800 mg of chondroitin sulfate daily. In all, 56 participants were randomly assigned to receive matched placebo capsules. Primary outcomes were a decrease in pain intensity as assessed by VAS and clinical improvement of psoriasis as determined by PASI score at the end of treatment as compared with baseline. Secondary outcomes were pain relief and function improvement in the knee according to the Lequesne AFI, acetaminophen consumption, histopathologic data, changes in psoriatic lesions according to physician global assessment (PGA), assessment of efficacy by participants and investigators, and quality of life as measured by SF‐36.

Morreale 1996 was a six‐month, randomized, participant‐ and investigator‐blind, parallel‐group study that was conducted at two centers. Participants were between 40 and 75 years of age and had grade I or II monolateral or bilateral knee osteoarthritis; they were not taking NSAIDs or chondroprotective treatment for 15 to 30 days before study initiation. A total of 74 participants were randomly assigned to the chondroitin sulfate group and received 400 mg of chondroitin sulfate three times daily plus three tablets of placebo for the first month, only 400 mg of chondroitin sulfate for the second month, and only placebo tablets from months 3 to 6. A total of 72 participants were randomly assigned to the control group; they received 50 mg of diclofenac sodium three times daily plus three tablets of placebo in the first month of the trial; only three placebo sachets in the second month of the trial, and placebo from months 3 to 6. Only the first month's results were applicable to the study because the treatment setup created dependence over the months of the study and did not allow for stratification of the data into three separate calculable data sets. The primary outcomes of the study were Lequesne's AFI score, spontaneous pain on a 0 to 100‐mm VAS, and pain on load on a 4‐point scale of absent, light, moderate, and intense. The secondary outcome of interest in the study was paracetamol consumption.

Nakasone 2011 was a randomized, participant‐ and investigator‐blind, placebo‐controlled study designed to assess the efficacy and safety of the test supplement in adult participants with symptomatic knee osteoarthritis; the study lasted 16 weeks and was conducted at two clinical service organizations in Yokohama, Japan. All male and female Japanese participants were 40 to 83 years of age and presented with clinical and radiographic evidence of mild knee osteoarthritis, defined by a score of 30 to 75 on a 100‐mm VAS and by radiologic severity of affected knee joints mainly graded 1 to 2 on the Kellgren‐Lawrence (KL) scale. Participants with diagnosed bilateral knee osteoarthritis were asked to specify the worse affected knee at baseline, and this knee was evaluated throughout the study period. A total of 16 participants were randomly assigned to the treatment group, which received 1200 mg of orally administered glucosamine hydrochloride, 200 mg of shark cartilage (60 mg of chondroitin sulfate), 300 mg of MSM, 105 mg of guava leaf extract, 5.6 μg of vitamin D, and 7.35 mg of vitamin B1 seven times daily; 16 participants were assigned to the inactive placebo group. The primary efficacy outcomes of this study were scores on the Japanese Knee Osteoarthritis Measure subscales of pain/stiffness, conditions of daily life, general conditions, and health conditions. Other reported outcomes were pain on VAS, pain on rest on VAS, pain on walking on VAS, pain on ascending or descending stairs on VAS, synovial inflammation, and cartilage metabolism.

Nguyen 2001 is a three‐month, randomized, participant‐ and investigator‐blind pilot study that included participants with pain in one or both temporomandibular joints (TMJs) and moderate or severe pain on lateral or dorsal palpation of the TMJ. A total of 24 participants were randomly assigned to the chondroitin sulfate group and received two tablets of 250 mg of glucosamine hydrochloride and 200 mg of chondroitin daily. Outcomes of the study were pain on VAS (0 to 100 mm), on the McGill Pain Questionnaire's pain rating scales, and on the mood and functioning questionnaire; tenderness on TMJ palpation; jaw range of motion; daily pain rating on the VAS; analgesic use; and daily change in pain intensity.

Pavelka 1999 was a phase II, randomized, participant‐ and investigator‐blind, dose‐effect study. Participants were older than 30 years of age and had femorotibial osteoarthritis of the knee, according to ACR criteria, with clinical symptoms that persisted for at least three months; Lequesne's Index greater than or equal to 8 points; pain on Huskisson's VAS greater than or equal to 40 mm during physical activity; and persistence of some articular joint space as documented on radiography. Participants were randomly assigned to one of four groups. In all, 35 participants were randomly assigned to the 200‐mg chondroitin sulfate group and received one sachet of 200 mg chondroitin sulfate; 35 participants to 800 mg of chondroitin sulfate and given one sachet of 800 mg chondroitin sulfate; 35 participants to the 1200‐mg chondroitin sulfate group and given one sachet of 1200 mg chondroitin sulfate; and 35 participants to the placebo group and given one placebo sachet. The primary outcome of the study was Lequesne's Index and Spontaneous Pain on Huskisson's VAS on a 0 to 100‐mm scale, Patient and Physician Global Efficacy Evaluations, and Paracetamol Consumption from Day 15 to Day 90.

Pavelka 2010 was a six‐month, controlled, randomized, multinational, multicenter, participant‐ and investigator‐blind, double‐dummy, parallel‐group study that was carried out at five centers in the Czech Republic, three in the Slovak Republic, five in Hungary, seven in Poland, and six in Romania. Participants had to be “aged 45 years or above and [to have] femoro‐tibial osteoarthritis of the knees longer than six months with pain and functional discomfort over one month during the past three months, were complying with the clinical and radiologic criteria of the American College of Rheumatology of knee osteoarthritis, had a Lequesne's index between 5 and 13 and a radiologic score of grade I, II or III of the modified Kellgren‐Lawrence scale on a frontal image of extended knee, on both knees, the image being not older than six months, had pain on movement and/or pain at rest in the last 48 hours [of] at least 40 mm evaluated on a VAS, and/or at least 40 mm evaluated on at least two items among the five items n the A‐section of the WOMAC Index, with no intake of analgesics for 48 hours and NSAID for 5 days.” A total of 142 participants were randomly assigned to receive 300 mg of avocado soybean unsaponifiable daily. In all, 121 participants in the comparator group were randomly assigned to receive 400 mg three times daily. The primary outcome of the study was the change in the WOMAC index from the beginning of the study to the end of treatment. Secondary outcomes were the Lequesne index, pain on active movement and at rest, and global assessment of efficacy.

Rai 2004 was a randomized, participant‐ and investigator‐blind, placebo‐controlled trial with participants older than 50 years of age who had primary knee osteoarthritis, mainly on the medial femorotibial compartment, according to the clinical and radiographic criteria of ACR and a Lequesne's index score of 4 or greater. The treatment group (n = 50) received a drug called kondro and was compared with a placebo group (n = 50). Kondro is a combination of chondroitin and glucosamine, although the exact amounts of either were not specified on the pharmaceutical company's website. Outcomes of interest in the study were Lequesne's index scores and joint space narrowing as shown on X‐rays.

Railhac 2012 was a randomized, double‐blind, placebo‐controlled, parallel group, multicenter study conducted in 20 French rheumatology practice centers comparing Structum® 500 mg (chondroitin) or matching placebo, twice daily by oral route as from baseline to week 48. Primary outcomes were pain on VAS (0 to 100) and Lequesne index scores. Secondary outcomes were clinical improvement according to the patient and the investigator and the use of rescue medication (paracetamol and/or NSAIDs). The first author was an employee of the maker of chondroitin product discussed.

Raynauld 2013 was a long‐term follow‐up study of Wildi 2011, as described above. Four‐year outcome of total knee replacement was presented along with its predictors, where one group was treated with chondroitin or placebo once daily for the first 6 months (double‐blind phase) followed by 6 months of treatment with 800 mg chondroitin sulfate once daily for both groups (open‐label phase). Thirty‐five patients were randomized each to chondroitin and placebo, of whom 57 patients were followed for this study. Thirteen patients underwent total knee replacement at 4‐years, 9/34 in the placebo and 4/35 in the chondroitin group. Details of inclusion/exclusion criteria are provided in Wildi 2011.

Rovetta 2002 was a 24‐month, randomized trial involving 24 "consecutive" participants of both sexes (2 men and 22 women) with a mean age of 53 years who were suffering from osteoarthritis and were showing central erosions of the distal interphalangeal (DIP) and/or proximal interphalangeal (PIP) joints. Participants were randomly assigned to one of two groups: naproxen 500 mg/daily only (Group A) or chondroitin sulfate 800 mg daily plus naproxen 500 mg daily (Group B). The only outcome of interest in the study was the joint count for erosions; results were taken at baseline and at 12 and 24 months.

Rovetta 2004 is the same trial as Rovetta 2002; however, the authors present results on several different outcomes. The outcomes of this study are radiographic joint counts for Heberden and Bouchard nodes; the Dreiser index was used to assess pain and function and Patient and Physician global assessments on a 0 to 10‐cm VAS.

Sawitzke 2008 was the follow‐up of Clegg 2006 study and presented the mean change in JSW in the medial compartment of the knee over 2 years.

Sawitzke 2010 was a 24‐month, participant‐ and investigator‐blind, placebo‐controlled study that was conducted at nine sites in the United States ancillary to the Glucosamine/chondroitin Arthritis Intervention Trial (GAIT), a follow‐up report of Clegg 2006; investigators enrolled 662 participants with knee osteoarthritis who satisfied radiographic criteria (Kellgren‐Lawrence (KL) grade II or III changes and baseline joint space width of at least 2 mm). Participants were included in the study if they were at least 40 years old, had been diagnosed with osteoarthritis for at least six months, and showed radiographic evidence of osteoarthritis by KL grade II or III. A total of 134 participants were assigned to receive 500 mg of glucosamine three times daily. In all, 126 participants were assigned to receive 400 mg of chondroitin sulfate three times daily; 129 to receive a combination of the glucosamine and chondroitin sulfate dosages; 142 to receive 200 mg of celecoxib daily; and 131 to receive a placebo. The study’s primary outcome was a 20% reduction in WOMAC pain over 24 months. Secondary outcomes were pain reduction attributable to each treatment WOMAC function subscale and the likelihood of achieving an OMERACT/OARSI response over 24 months.

Uebelhart 1998 was a one‐year, randomized, participant‐ and investigator‐blind, controlled pilot study with 46 participants of both sexes between the ages of 35 and 79 who had symptomatic osteoarthritis. The chondroitin active group received 400 mg of chondroitin sulfate twice daily, while the placebo group received placebo sachets. A total of 23 participants were randomly assigned to chondroitin sulfate, and 23 participants were randomly assigned to the placebo group. Primary outcomes of the study were the degree of spontaneous joint pain as measured on VAS (0 to 100 mm) and overall mobility capacity on VAS. Secondary outcomes included the actual joint space measurement taken by X‐ray and the levels of biochemical markers of bone and joint metabolism. In our meta‐analysis, we subtracted the rating for overall mobility capacity from 100 because the scale for this outcome took 0 as the worst score and 100 as the best score. Because all of the other studies that presented data on this outcome did so on an opposite scale, which took 0 as the best and 100 as the worst, we inverted the scale for this study's outcome of physical function to standardize and incorporate its data.

Uebelhart 2004 was a multicenter, participant‐ and investigator‐blind, placebo‐controlled, one‐year trial that compared chondroitin sulfate with placebo. Participants 40 years of age or older had stiffness of less than 30, bilateral or monolateral idiopathic knee osteoarthritis according to the ACR, and a Kellgren‐Lawrence score of I to III, with a minimum of 25% remaining medial femorotibial joint space. A total of 54 participants were randomly assigned to the chondroitin sulfate group and received 800 mg daily of chondroitin sulfate for two periods of three months (1 to 90 days and 181 to 270 days) over a two‐year period. The primary outcome of the study was the score on Lequesne's index. Secondary outcomes were spontaneous pain on a 100‐mm VAS, time to walk 20 meters, Patient global assessment of efficacy on a 0 to 4 Scale, consumption of analgesics, joint space narrowing on X‐ray, and adverse events.

Verbruggen 2002 was a randomized, participant‐ and investigator‐blind, placebo‐controlled trial that assessed the effects of chondroitin sulfate as a disease modifying osteoarthritis drug (DMOAD) at Ghent University Hospital. Participants were included if they were between 40 and 70 years of age and had symptoms producing osteoarthritis of the finger joint that were confirmed according to the presence of osteophytes and/or joint space narrowing, with or without subchondral sclerosis on hand X‐rays. The study reported that all participants were Caucasian. A total of 44 participants were randomly assigned to the active chondroitin sulfate group and received 400 mg of chondroitin sulfate three times daily for three years. In all, 48 participants were randomly assigned to the placebo group and received placebo sachets three times daily for three years.

Wildi 2011 was a multicenter, randomized, participant‐ and investigator‐blind, controlled trial that compared chondroitin sulfate with placebo in participants with primary knee osteoarthritis lasting six months. Participants were included if they were diagnosed with primary osteoarthritis of the knee in accordance with the clinical and radiologic criteria of the ACR and if they had clinical signs of synovitis (warmth, swelling, or effusion), a disease severity grade 2 or 3 based on the Kellgren–Lawrence radiographic system, a minimal medial joint space width (JSW) of 2 mm on standing knee X‐ray, and a VAS pain index of at least 40 mm while walking. Participants were required to have no significant laboratory abnormalities. In all, 35 participants were randomly assigned to receive 800 mg of chondroitin sulfate as two capsules of 400 mg once daily. A total of 34 participants were randomly assigned to receive placebo once daily. The primary outcome of the study was synovial membrane thickness. Secondary outcomes were cartilage volume; bone marrow lesions; WOMAC pain, function, stiffness, and total score; pain on VAS; and quality of life as assessed by SF‐36.

Zegels 2012 was a multicenter, randomized, double blind, double‐dummy study with an allocation ratio of 1:1:1, in 10 centres in Belgium, three in France and two in Switzerland that compared oral sachet of chondroitin sulfate 1200 mg/day, one oral capsule of chondroitin sulfate 400 mg three times a day and placebo.  Primary outcome was Lequesne's index and secondary outcomes were Pain VAS, treatment compliance and adverse effects.  

Excluded studies

After review of the full texts, 58 studies were excluded, all of which are included in the reference list (Excluded studies). Of these studies, 11 were excluded because they were reviews and 13 because they were not randomized. Six studies were excluded because they did not provide data on any clinical outcomes specified. Two studies were excluded because they were duplicates, 16 because they were abstracts, one because it could not be obtained and the author did not respond to requests to provide the study for review, four because they were commentaries, another four because investigators did not study the efficacy of chondroitin or did not study osteoarthritis, and one because an obvious mismatch was noted in the data provided. Details are given in the excluded studies section (Characteristics of excluded studies).

Allocation

Risk of bias assessment was performed by two assessors independently (SN, PF/JAS). All disagreements were settled by consensus.

Of the 43 studies included in this meta‐analysis, 30 did not report methods of randomization adequately (Alekseeva 1999; Alekseeva 2005a; Alekseeva 2005b; Alekseeva 2008; Artemenko 2005; Bourgeois 1998; Bucsi 1998; Cohen 2003; Conrozier 1992; Conrozier 1998; Das 2000; Debi 2000; Kanzaki 2011; L'Hirondel 1992; Lila 2005; Magrans‐Courtney 2011; Mazieres 1992; Mazieres 2001; Morreale 1996; Nakasone 2011; Nasonova 2001; Pavelka 1999; Rai 2004; Rovetta 2002; Rovetta 2004; Sawitzke 2010; Uebelhart 1998; Verbruggen 2002; Wildi 2011; Raynauld 2013). There were 13 studies that reported adequate methods of randomization ( Clegg 2006; Gabay 2011; Kahan 2009; Messier 2007; Michel 2005; Moller 2010; Nguyen 2001; Pavelka 2010; Sawitzke 2008; Uebelhart 2004; Fardellone 2013; Railhac 2012; Zegels 2012).

In all, 34 studies did not report the methods of allocation concealment used or had unclear methods of allocation concealment (Alekseeva 1999; Alekseeva 2005a; Alekseeva 2005b; Alekseeva 2008; Artemenko 2005; Bourgeois 1998; Bucsi 1998; Cohen 2003; Conrozier 1992; Conrozier 1998; Debi 2000; Gabay 2011; Kanzaki 2011; L'Hirondel 1992; Lila 2005; Magrans‐Courtney 2011; Mazieres 1992; Mazieres 2001; Michel 2005; Morreale 1996; Nakasone 2011; Nasonova 2001; Pavelka 1999; Pavelka 2010; Rai 2004; Rovetta 2002; Rovetta 2004; Sawitzke 2008; Sawitzke 2010; Uebelhart 1998; Verbruggen 2002; Fardellone 2013; Railhac 2012; Zegels 2012). Nine studies reported adequate methods of allocation concealment (Clegg 2006; Das 2000; Kahan 2009; Messier 2007; Moller 2010; Nguyen 2001; Uebelhart 2004; Wildi 2011; Raynauld 2013). For a more detailed description of allocation concealment in the individual studies, refer to the risk of bias tables included under the characteristics of included studies and to Figure 1 and Figure 2.

Blinding

Of the 43 studies included in this meta‐analysis, 14 used unclear methods of blinding (Bucsi 1998; Cohen 2003; Conrozier 1998; Debi 2000;Gabay 2011;L'Hirondel 1992; Mazieres 1992;Nakasone 2011;Rai 2004;Sawitzke 2008;Sawitzke 2010;Verbruggen 2002; Wildi 2011;Raynauld 2013). Nine studies did not have participant and investigator blinding (Alekseeva 1999, Alekseeva 2005a, Alekseeva 2005b, Alekseeva 2008, Artemenko 2005, Lila 2005, Nasonova 2001, Rovetta 2004, Rovetta 2002); 20 studies (Bourgeois 1998, Clegg 2006, Conrozier 1992, Das 2000, Kahan 2009, Kanzaki 2011, Magrans‐Courtney 2011, Mazieres 2001, Messier 2007, Michel 2005, Moller 2010, Morreale 1996, Nguyen 2001, Pavelka 1999, Pavelka 2010, Uebelhart 1998, Uebelhart 2004; Fardellone 2013; Railhac 2012; Zegels 2012) used investigator and participant blinding and described adequately the methods used.

Incomplete outcome data

Of the 43 studies included in this meta‐analysis, 10 used unclear methods in handling missing outcomes (Alekseeva 2005a; Alekseeva 2005b; Alekseeva 2008; Bucsi 1998; Debi 2000; Kanzaki 2011; L'Hirondel 1992; Nakasone 2011; Sawitzke 2010; Uebelhart 1998). Of the remaining 33 studies, 25 were assessed to have adequately handled missing outcome data by accounting for and providing a clear explanation of all dropouts or performing the appropriate statistical methods for handling missing data (Bourgeois 1998; Clegg 2006; Cohen 2003; Conrozier 1992; Conrozier 1998; Das 2000; Gabay 2011; Kahan 2009; Magrans‐Courtney 2011; Mazieres 1992; Mazieres 2001; Messier 2007; Michel 2005; Morreale 1996; Pavelka 1999; Pavelka 2010; Rovetta 2002; Rovetta 2004; Sawitzke 2008; Uebelhart 2004; Verbruggen 2002; Fardellone 2013; Railhac 2012; Raynauld 2013; Zegels 2012). The remaining eight studies failed to provide an adequate account of their withdrawals and missing outcome data or to perform the appropriate statistical methods.

Selective reporting

Only one study (Clegg 2006) reported on outcome measures as recommended by the Outcome Measures in Rheumatology (OMERACT) group. Therefore, it was difficult to find commonly reported outcomes across the studies. Furthermore, because review authors could not gain access to the study protocols, it was difficult to decipher whether the studies reported on all outcomes specified in the study protocol.

Other potential sources of bias

Although no other sources of bias were detected, most of the included studies were funded by a manufacturer of chondroitin, which may qualify as a confounding source. This issue is discussed further in the Discussion section of this review, as the source of funding for trials is not a factor that directly influences the methods used in the analysis. For more information on the source of funding for each study, please refer to the "Characteristics of included studies tables."

CHONDROITIN SULFATE VERSUS PLACEBO

(See Table 1 under Data and analyses.)

1. Pain: Twelve studies provided data (Bourgeois 1998; Bucsi 1998; Clegg 2006; Mazieres 2001; Morreale 1996; Pavelka 1999; Sawitzke 2010; Uebelhart 1998; Uebelhart 2004; Wildi 2011; Railhac 2012; Zegels 2012) for short‐term and/or long‐term pain. The difference was statistically significant in the eight short‐term studies (Bourgeois 1998; Mazieres 1992; Morreale 1996; Pavelka 1999; Sawitzke 2010; Uebelhart 1998; Uebelhart 2004; Railhac 2012), with a mean difference of ‐10.14 units (95% CI, ‐14.58 to ‐5.71) between chondroitin and placebo arms on a 0 to 100 pain scale. This translated into an absolute risk difference of 10% lower (95% CI, 15% to 6% lower) and a NNT=5 (95% CI 3 to 8) (n=8 trials; level of evidence, low). In six long‐term studies (Bucsi 1998; Clegg 2006; Uebelhart 1998; Uebelhart 2004; Wildi 2011; Zegels 2012), chondroitin was associated with a mean difference of ‐9.00 (95% CI ‐17.7 to ‐0.34) on a 0‐100 scale, which corresponds to absolute risk difference of 9% lower (95% CI, 18% to 0% lower; p=0.05). Heterogeneity was moderate in both long‐term and short‐term studies, with T² of 0.07 and I² of 69% in the short‐term and T² of 0.12 and I² of 83% in the long‐term pain studies. Because 12 studies were included for short‐ or long‐term pain outcomes, we created a funnel plot to detect bias. Figure 5 shows the presence of bias, as the plot does not take the shape of a funnel.

Figure 5

---

---

Funnel plot of comparison: 1 Chondroitin versus placebo, outcome: 1.1 Pain on a 0 to 100 scale.

Sensitivity analysis showed that the results were robust with respect to blinding and ITT, but not robust with respect to allocation concealment, study size, study sponsorship, or publication year (Analysis 6.1; Analysis 10.1; Analysis 14.1; Analysis 18.1; Analysis 22.1; Analysis 31.1). Studies that were adequately blinded had a similar result to the original estimate, and most studies included in the analyses had adequate blinding. Studies using appropriate ITT analyses had an effect size of ‐0.46 for pain which was significant , similar to the overall effect, while those with unclear ITT had a larger effect size of ‐1.00, while 1 study with high risk of bias for ITT analyses had an effect size 0.24, which was insignificant. Most included studies used appropriate ITT analyses.

Among studies that used adequate methods of allocation concealment, no significant difference was reported between placebo and chondroitin while studies that used unclear methods of allocation concealment, participants treated with chondroitin had statistically significantly better scores than those in the placebo group.A difference was seen between the results of studies with large and small sample sizes. Studies with large sample sizes reported no statistically significant difference in pain scores between chondroitin and placebo while studies with small sample sizes reported that pain scores for participants in the chondroitin group were significantly better than those in the placebo group.

Studies funded by chondroitin sulfate manufacturers showed that participants in the chondroitin group had statistically significantly better pain scores than those in the placebo group, and the effect size was ‐0.52. Studies that did not specify their source of funding or with no pharmaceutical sponsorship did not report a statistically significant difference between participants in the chondroitin group and those in the placebo group, but the effect size was greater at ‐0.73. Thus, the effect of study sponsorship on findings was mixed.

Studies published before 2000 showed that participants treated with chondroitin had significantly better pain scores than those in the placebo group while studies published after 2000 reported no statistically significant difference between chondroitin and placebo.

2. MCII WOMAC: Two studies provided data (Clegg 2006; Kahan 2009). A significantly greater number of participants in the chondroitin sulfate group experienced an MCII in pain on WOMAC (Risk ratio of 1.12, 95% CI 1.01 to 1.24; P = 0.04) with an absolute difference of 6% (95% CI 1% to 11%) and an NNT=16 (95% CI 9 to 136). Sensitivity analyses showed the results were robust to study sponsorship.

3. Physical function: Three studies provided data (Clegg 2006; Sawitzke 2010; Uebelhart 1998) in the long‐term analysis, and two studies provided data in the short‐term analysis (Sawitzke 2010; Uebelhart 1998). Uebelhart 1998 presented data on 0 = best and 100 = worst scale, and the mean score was transformed by subtracting the scores from 100 to make the data compatible with those of the other studies, which presented data on a 0 = best and 100 = worst scale. No statistically significant difference was noted between chondroitin and placebo in the short‐term studies (SMD 0.11, 95% CI ‐0.47 to 0.68; P = 0.72; T² = 0.13; moderate level of heterogeneity with an I²of 70%) or long‐term studies (SMD ‐0.20, 95% CI ‐0.53 to 0.14; P = 0.25; T² = 0.06; moderate level of heterogeneity, with an I²of 77%). This equated a mean difference of 2.39 (95% CI, ‐9.81 to 14.58) in short‐term and ‐4.31 (95% CI, ‐11.33 to 2.70) in long‐term studies, on a 0‐100 scale. This equated to an absolute risk difference of 8% lower with chondroitin versus placebo (24% lower to 7% higher; n = 3 trials; level of evidence, moderate) for long‐term physical function studies.

Sensitivity analyses demonstrated the results were robust to blinding but not to allocation concealment, sample size, pharmaceutical sponsorship, publication year or ITT analyses (Analysis 6.3; Analysis 10.3; Analysis 14.3; Analysis 18.3; Analysis 18.3; Analysis 22.3; Analysis 31.3). Studies with an unclear allocation, small sample size, with pharmaceutical sponsorship, with publication prior to 1999, and with unclear impact of missing data showed that participants in the chondroitin sulfate group performed statistically significantly better on physical function scales than those in the placebo arm.

4. WOMAC stiffness: One study provided data (Clegg 2006). No significant difference between chondroitin and placebo was noted (P = 0.21).

5. Patient global assessment of good to very good: Three studies provided data (Bourgeois 1998; Bucsi 1998; Pavelka 1999). A total of 68% of participants in the chondroitin sulfate group and 33% of participants in the placebo group reported a good to very good assessment of their status. Participants in the chondroitin sulfate group were twice as likely as those in the placebo group to report good to very good overall assessment with a significant risk ratio (RR) of 2.11 (95% CI 1.49 to 2.99; P < 0.0001) in short‐term and RR of 2.04 (95% CI 1.28 to 3.27; P = 0.003) in long‐term studies.

6. Patient global assessment on VAS: Three studies provided data (Clegg 2006; Gabay 2011; Kahan 2009). No significant difference was reported between chondroitin and placebo in short‐ or long‐term studies, with SMD of ‐0.03 (95% CI ‐0.48 to 0.21; P = 0.45) and 0.01 (95% CI ‐0.21 to 0.23; P = 0.92), respectively.

7. Physician global assessment of good to very good: Three studies provided data (Bourgeois 1998; Bucsi 1998; Pavelka 1999). A total of 71% of physician global assessments in the chondroitin sulfate group and 36% of those in the placebo group were good to very good assessments, leading to a significant risk in short‐ or long‐term studies, with RR of 1.95 (95% CI 1.42 to 2.69; P < 0.00001) and 2.12 (95% CI 1.33 to 3.38; P = 0.002), respectively.

8. Physician global assessment on VAS: Two studies provided data (Clegg 2006; Kahan 2009). No significant difference was noted between chondroitin and placebo (SMD 0.09, 95% CI ‐0.05 to 0.23, P = 0.19).

9. Total Knee arthroplasty during follow‐up: One study provided data (Raynauld 2013), which was a follow‐up of Wildi 2011. No significant difference was noted between chondroitin and placebo, RR was 0.43 (95% CI 0.15 to 1.27; P =0.13).

10. Grip strength (kg/cm²): One study provided data (Gabay 2011). No significant difference was noted between chondroitin and placebo, mean difference was 0.90 (95% CI ‐2.29 to 4.09; see Table 1).

11. Morning stiffness (minutes): One study provided data (Gabay 2011). No significant difference was reported between chondroitin and placebo ‐0.60 (95% CI ‐5.16 to 3.96; P = 0.80; see Table 1).

12. Cartilage volume loss (global): One study provided data (Wildi 2011). Participants in the chondroitin group had significantly less global cartilage volume loss than those in the placebo group, mean difference was 1.80 (95% CI 0.23 to 3.37; P = 0.02; see Table 1).

13. Cartilage volume loss (lateral compartment): One study provided data (Wildi 2011). Participants in the chondroitin group had significantly less lateral compartment cartilage volume loss than those in the placebo group, mean difference was 2.19 (95% CI 0.31 to 4.07; P = 0.02; see Table 1).

14. Cartilage volume loss (medial compartment): One study provided data (Wildi 2011). No significant difference in cartilage volume loss was noted between chondroitin and placebo, mean difference was 1.47 (95% CI ‐0.88, 3.82; P = 0.22).

15. Cartilage volume loss (condyles): One study provided data (Wildi 2011). No significant difference in cartilage volume loss was reported between chondroitin and placebo, mean difference was 0.64 (95% CI ‐1.59, 2.87;P = 0.57; see Table 1).

16. Cartilage volume loss (lateral condyles): One study provided data (Wildi 2011). No significant difference in cartilage volume loss was noted between chondroitin and placebo, mean difference was 1.84 (95% CI ‐0.44 to 4.12; P = 0.11; see Table 1).

17. Cartilage volume loss (medial condyles): One study provided data (Wildi 2011). No significant difference in cartilage volume loss was reported between chondroitin and placebo, mean difference was ‐0.65 (95% CI ‐4.14 to 2.84; P = 0.71; see Table 1).

18. Cartilage volume loss (tibial plateaus): One study provided data (Wildi 2011). Participants in the chondroitin group had significantly less tibial plateau cartilage volume loss than those in the placebo group, mean difference was 2.98 (95% CI 1.15 to 4.81; P = 0.001; see Table 1).

19. Cartilage volume loss (lateral tibial plateaus): One study provided data (Wildi 2011). No significant difference in cartilage volume loss was reported between chondroitin and placebo, mean difference was 2.31 (95% CI ‐0.09 to 4.71; P = 0.06; see Table 1).

20. Cartilage volume loss (medial tibial plateaus): One study provided data (Wildi 2011). Participants in the chondroitin group had significantly less cartilage volume loss in their medial tibial plateaus than those in the placebo group, mean difference was 4.47 (95% CI 1.57 to 7.37; P = 0.003; see Table 1).

21. Cartilage volume loss (trochlea): One study provided data (Wildi 2011). No significant difference in cartilage volume loss was reported between chondroitin and placebo, mean difference was 0.65 (95% CI ‐1.79 to 3.09; P = 0.60; see Table 1).

22. Change in bone marrow lesion scores (global): One study provided data (Wildi 2011). No significant difference between chondroitin and placebo was noted across all bone marrow lesion regions, mean difference was ‐0.07 (95% CI ‐0.65 to 0.51; Table 1).

23. Change in bone marrow lesion scores (lateral compartment): One study provided data (Wildi 2011). No significant difference between chondroitin and placebo was noted across all bone marrow lesion regions, mean difference was ‐0.10 (95% CI ‐0.30 to 0.10; P = 0.32; see Table 1).

24. Change in bone marrow lesion scores (medial compartment): One study provided data (Wildi 2011). No significant difference between chondroitin and placebo was noted across all bone marrow lesion regions, mean difference was 0.02 (95% CI ‐0.52 to 0.56; P = 0.94; see Table 1).

25. Change in bone marrow lesion scores (condyles): One study provided data (Wildi 2011). No significant difference between chondroitin and placebo was noted across all bone marrow lesion regions, mean difference was 0.02 (95% CI ‐0.35 to 0.39; P = 0.92; see Table 1).

26. Change in bone marrow lesion scores (lateral condyles): One study provided data (Wildi 2011). No significant difference between chondroitin and placebo was noted across all bone marrow lesion regions, mean difference was ‐0.04 (95% CI ‐0.14 to 0.06; P = 0.45; see Table 1).

27. Change in bone marrow lesion scores (medial condyles): One study provided data (Wildi 2011). No significant difference between chondroitin and placebo was noted across all bone marrow lesion regions, mean difference was 0.13 (95% CI ‐0.22 to 0.48; P = 0.47; see Table 1).

28. Change in bone marrow lesion scores (tibial plateaus): One study provided data (Wildi 2011). No significant difference between chondroitin and placebo was noted across all bone marrow lesion regions, mean difference was ‐0.10 (95% CI ‐0.46 to 0.26; P = 0.58; see Table 1).

29. Change in bone marrow lesion scores (lateral tibial plateaus): One study provided data (Wildi 2011). No significant difference between chondroitin and placebo was noted across all bone marrow lesion regions, mean difference was ‐0.01 (95% CI ‐0.18 to 0.16; P = 0.91; see Table 1).

30. Change in bone marrow lesion scores (medial tibial plateaus): One study provided data (Wildi 2011). No significant difference between chondroitin and placebo was noted across all bone marrow lesion regions, mean difference was ‐0.10 (95% CI ‐0.40 to 0.20; P = 0.22; see Table 1).

31. Change in bone marrow lesion scores (trochlea): One study provided data (Wildi 2011). No significant difference between chondroitin and placebo was noted across all bone marrow lesion regions, mean difference was ‐0.16 (95% CI ‐0.10 to 0.42; P = 0.52; see Table 1).

32. OMERACT‐OARSI responder: One study provided data (Clegg 2006). The proportion of participants achieving responder status in the chondroitin sulfate group was not significantly different from that in the placebo group, mean difference was (95% CI ; P = 0.09; see Table 1).

33. Lequesne's index score: Nine studies provided data (Bourgeois 1998; Bucsi 1998; Mazieres 2001; Morreale 1996; Moller 2010; Pavelka 1999;Uebelhart 2004; Railhac 2012; Zegels 2012). Lequesne's Index score was significantly lower (better) in chondroitin‐treated participants than in placebo‐treated participants with (mean difference (MD) ‐1.98, 95% CI ‐2.79 to ‐1.17; P < 0.00001) in short‐term studies. This translates into an absolute risk difference of 8.3% lower in chondroitin versus placebo (11.6% lower to 4.9% lower; T² = 0.78; I² = 67%; n = 7 trials; level of evidence, moderate). For long‐term studies, SMD was ‐1.60, (95% CI ‐3.49 to 0.29; P =0.10; T² = 1.88; I² = 68%;), not statistically significant.

34. HAQ Disability score: One study provided data (Clegg 2006). No significant difference was reported between chondroitin and placebo, mean difference was (P = 0.26).

35. Radiographic—minimum joint space width (mm): Two studies provided data (Uebelhart 1998; Uebelhart 2004). No significant difference was noted between chondroitin and placebo (MD 0.16, 95% CI ‐0.16 to 0.48; P = 0.32).

36. Radiographic—reduction in minimum joint space width (mm): Two studies provided data (Kahan 2009; Michel 2005). Participants in the chondroitin group experienced significantly less reduction in minimum joint space width as compared with those in the placebo group (MD 0.18, 95% CI 0.06 to 0.30; P < 0.0001). This translated into significantly less loss of minimum joint space width in the chondroitin group than in the placebo group, with a relative risk difference of 4.72% less (95% CI 1.58% to 7.87% less; n = 2 trials; level of evidence, high).

37. Radiographic—mean joint space width (mm): One study provided data (Uebelhart 1998). No significant difference was reported between chondroitin and placebo (MD 0.17, 95% CI ‐0.47 to 0.81; P = 0.60).

38. Radiographic—change in mean joint space width (mm): Three studies provided data (Kahan 2009; Michel 2005; Sawitzke 2010). Participants in the chondroitin group experienced significantly less change in their mean joint space width as compared with those in the placebo group (MD 0.13, 95% CI 0.07 to 0.20; P <0.0001).

39. SF‐36 (Physical component score): One study provided data (Moller 2010). No significant difference was reported between chondroitin and placebo (P = 0.81).

40. SF‐36 (Mental component score): One study provided data (Moller 2010). No significant difference was noted between chondroitin and placebo (P = 0.96).

41. All withdrawals: Fifteen studies provided data (Bourgeois 1998; Clegg 2006; Conrozier 1992; Gabay 2011; Kahan 2009; Mazieres 1992; Mazieres 2001; Michel 2005; Moller 2010; Morreale 1996; Pavelka 1999; Uebelhart 1998; Uebelhart 2004; Verbruggen 2002; Wildi 2011). The number of participants who withdrew from the study for any reason did not differ significantly between chondroitin and placebo groups (RR 0.80, 95% CI 0.63 to 1.02; P = 0.07).

42. Withdrawals due to adverse events: Ten studies provided data (Bourgeois 1998; Clegg 2006; Gabay 2011; Kahan 2009; Mazieres 2001; Michel 2005; Moller 2010; Morreale 1996; Uebelhart 2004; Verbruggen 2002), four short‐term and 6 long‐term studies. No statistically significant difference was reported between placebo and chondroitin sulfate (RR 0.95, 95% CI 0.31 to 2.89; P = 0.70) for short‐term or long‐term studies, RR of 1.09 [95% CI, 0.73 to 1.63]. The overall RR combining short‐ and long‐term studies was 1.08 (95% CI, 0.74 to 1.57).

43. Withdrawals due to inefficacy: Ten studies provided data (Bourgeois 1998; Clegg 2006; Kahan 2009; Mazieres 2001; Michel 2005; Morreale 1996; Uebelhart 1998; Uebelhart 2004; Verbruggen 2002; Wildi 2011). The number of participants who withdrew from the study because of inefficacy did not differ significantly between chondroitin and placebo groups (RR 1.22, 95% CI 0.88 to 1.70; P = 0.23).

44. Number of adverse events: Eight studies provided data (Bourgeois 1998; Gabay 2011; Kahan 2009; Mazieres 2001; Moller 2010; Morreale 1996; Pavelka 1999; Wildi 2011). The number of adverse events did not differ significantly between chondroitin and placebo groups (RR 0.96, 95% CI 0.78 to 1.18; P = 0.69).

45. Number of serious adverse events: Six studies provided data (Bourgeois 1998; Gabay 2011; Morreale 1996; Sawitzke 2010; Wildi 2011; Zegels 2012). The risk of serious adverse events was lower with chondroitin compared to placebo in the long‐term with an RR of 0.38 (95% CI 0.17 to 0.84; P = 0.01] , but not in the short‐term studies, RR of 0.50 [95% CI, 0.09 to 2.78]. The overall risk, combining both short‐ and long‐term studies, was 0.40 (95% CI 0.19 to 0.82].

46. Gastrointestinal adverse events: Nine studies provided data (Bourgeois 1998; Gabay 2011; Kahan 2009; Mazieres 1992; Michel 2005; Moller 2010; Uebelhart 2004; Verbruggen 2002; Wildi 2011). The number of gastrointestinal adverse events did not differ significantly between chondroitin and placebo groups (RR 0.68, 95% CI 0.45 to 1.04; P = 0.07).

47. Other adverse events: These were defined as adverse events other than gastrointestinal, cardiac, or hematologic. Six studies provided data (Bourgeois 1998; Gabay 2011; Michel 2005; Moller 2010; Uebelhart 2004; Wildi 2011). The number of other adverse events was not significantly different between chondroitin and placebo groups (RR 0.95, 95% CI 0.79 to 1.15; P = 0.66).

48. Death: Seven studies provided data (Bourgeois 1998; Clegg 2006; Gabay 2011; Moller 2010; Sawitzke 2010; Uebelhart 1998; Wildi 2011). Two deaths were reported in the placebo group and none in the chondroitin group. No significant difference was noted between the two groups (RR 0.34, 95% CI 0.04 to 3.20; P = 0.35).

CHONDROITIN SULFATE VERSUS CONTROL

(See Table 2 under Data and Analyses.)

Control treatment was avocado soybean unsaponifiable (ASU) in Pavelka 2010, “regular treatment” of osteoarthritis according to physician’s preferences in Nasonova 2001, and ibuprofen 1200 mg daily in Alekseeva 1999.

1. Pain (0‐100 scale): One study (Pavelka 2010) provided both short‐ and long‐term data. No significant difference was noted between chondroitin and control; SMD was ‐0.06 [95% CI, ‐0.26 to 0.15] for short‐term and ‐0.10 [95% CI, ‐0.31 to 0.11] for long‐term studies. This translates to a mean differences of ‐1.14 [95% CI, ‐5.32 to 3.04] and ‐2.02 [95% CI, ‐6.12 to 2.08], respectively.

2. WOMAC stiffness: One study (Pavelka 2010) provided both short‐ and long‐term data. No significant difference was reported between chondroitin and control; mean differences in short‐ and long‐terms studies were ‐1.00 [95% CI, ‐5.77 to 3.77] and ‐0.60 [95% CI, ‐5.25 to 4.05], respectively.

3. WOMAC physical function: One study (Pavelka 2010) provided both short‐ and long‐term data. No significant difference was noted between chondroitin and control; mean differences in short‐ and long‐terms studies were ‐2.10 [95% CI, ‐6.44 to 2.24] and ‐1.56 [95% CI, ‐5.70 to 2.58], respectively.

4. WOMAC Total: One study (Pavelka 2010) provided both short‐ and long‐term data. No significant difference was reported between chondroitin and control; mean differences in short‐ and long‐terms studies were ‐1.80 [95% CI, ‐6.06 to 2.46] and ‐1.60 [95% CI, ‐5.69 to 2.49], respectively.

5. Lequesne's index: Two studies provided data (Nasonova 2001; Pavelka 2010). Both were long‐term studies. No statistically significant difference was noted between chondroitin and control; mean difference was ‐1.36 [95% CI, ‐3.60 to 0.89].

6. Patient global assessment (% with improvement): Two studies provided data (Nasonova 2001; Pavelka 2010). Both were long‐term studies. Investigators found that participants treated with chondroitin sulfate scored better than participants in the control group, with an odds ratio (OR) of 5.04 (95% CI 1.95 to 13.05; P = 0.0008).

7. Physician global assessment (% with improvement): Two studies provided data (Nasonova 2001; Pavelka 2010). Both were long‐term studies. Investigators found that participants treated with chondroitin sulfate scored better than participants in the control group (OR 7.68, 95% CI 4.48 to 13.15; P < 0.00001).

8. NSAID consumption: One study (Pavelka 2010) provided both short‐ and long‐term data. No significant difference was reported between chondroitin and control; mean differences in short‐ and long‐terms studies were ‐0.02 [95% CI, ‐0.23 to 0.19] and 0.00 [95% CI, ‐0.20 to 0.20], respectively.

9. All withdrawals: Three studies provided data (Alekseeva 1999; Nasonova 2001;Pavelka 2010 ). No significant difference was noted between chondroitin and control (OR 0.39, 95% CI 0.09 to 1.74; P = 0.22).

10. Withdrawals due to adverse events: Two studies provided data (Alekseeva 1999; Pavelka 2010). No significant difference was reported between chondroitin and control (RR 0.27, 95% CI 0.03 to 2.65; P = 0.26).

11. Number of adverse events: Two studies provided data (Alekseeva 1999; Pavelka 2010). No significant difference was noted between chondroitin and control (RR 0.08, 95% CI 0.00 to 4.82; P = 0.22).

12. Number of GI adverse events: Two studies provided data (Alekseeva 1999; Pavelka 2010). No significant difference was reported between chondroitin and control RR 0.51, 95% CI 0.24 to 1.07; P = 0.08).

CHONDROITIN SULFATE PLUS GLUCOSAMINE VERSUS PLACEBO

(See Table 3 under Data and Analyses.)

1. Pain: Five studies provided data (Clegg 2006; Kanzaki 2011; Messier 2007; Nakasone 2011; Sawitzke 2010). Three were short‐term studies (Kanzaki 2011; Nakasone 2011; Sawitzke 2010), and three were long‐term studies (Clegg 2006; Messier 2007; Sawitzke 2010). No statistically significant difference was noted between the chondroitin plus glucosamine group and the placebo group for either the short‐term studies (SMD ‐0.06, 95% CI ‐0.33 to 0.20; P = 0.63) or the long‐term studies (SMD ‐0.09, 95% CI ‐0.21 to 0.04; P = 0.17).

Sensitvity analyses were robust to allocation concealment, blinding, study size, study sponsorship, publication year, and impact of true ITT analyses.

2. MCII WOMAC pain: One study provided data (Clegg 2006). No significant difference between active treatment and placebo was reported (P = 0.09).

3. Physical function: Four studies provided data (Clegg 2006; Kanzaki 2011; Messier 2007; Sawitzke 2010). Sawitzke 2010 and Kanzaki 2011 were in the short‐term category. The rest of the studies, including Sawitzke 2010 at 24 months, were in the long‐term subgroup. No significant difference was noted between active treatment and placebo in the short‐ (SMD 0.11, 95% CI ‐0.31 to 0.54; P = 0.60) or long‐term subgroups (SMD ‐0.11, 95% CI ‐0.24 to 0.01; P = 0.08).

Sensitivity analyses were robust to allocation concealment, blinding, study size, study sponsorship, publication year, and impact of true ITT analyses.

4. Six‐minute walk, distance in meters: One study provided data (Messier 2007). No significant difference was reported between active treatment and placebo (P = 0.8); mean difference was ‐2.90 [95% CI, ‐24.94 to 19.14].

5. WOMAC stiffness: One study provided data (Clegg 2006). No significant difference was reported between active treatment and placebo (P = 0.25); mean difference was ‐2.20 [95% CI, ‐5.97 to 1.57].

6. WOMAC Total: One study provided data (Clegg 2006). Participants treated with chondroitin sulfate in combination with glucosamine had significantly better scores on WOMAC Total (MD ‐2.60, 95% CI ‐5.29 to 0.72; P = 0.13). The results for WOMAC Total as reported by Clegg 2006 were presented as normalized WOMAC scores on a 0 to 300 scale, which we then normalized to a 0 to 100 scale.

7. Patient global assessment on VAS: Two studies provided data (Clegg 2006; Das 2000). No significant difference was noted between active treatment and placebo (P = 0.39); mean difference was ‐1.60 [95% CI, ‐5.29 to 2.09].

8 Physician global assessment on VAS: One study provided data (Clegg 2006). No significant difference between active treatment and placebo was reported (P = 0.43); mean difference was ‐1.40 [95% CI, ‐4.87 to 2.07].

9. OMERACT‐OARSI responder: One study provided data (Clegg 2006). The proportion of participants achieving responder status in the chondroitin sulfate in combination with glucosamine group was significantly greater in the active treatment group than in the placebo group (P = 0.03); odds ratio was 1.15 [95% CI, 1.02 to 1.31].

10. HAQ Disability score: One study provided data (Clegg 2006). No significant difference between active treatment and placebo was noted (P = 0.24); mean difference was ‐0.04 [95% CI, ‐0.11 to 0.03].

11. Objective joint function (flexion): Only one study provided data on this outcome (Messier 2007). No significant difference between active treatment and placebo groups was reported (P = 0.96); mean difference was ‐0.60 [95% CI, ‐21.54 to 20.34].

12. Objective joint function (extension): Only one study provided data on this outcome (Messier 2007). No significant difference between active treatment and placebo groups was reported (P = 0.29); mean difference was ‐25.80 [95% CI, ‐72.67 to 21.07].

13. Objective joint function (balance): Only one study provided data on this outcome (Messier 2007). Participants in the active treatment group had significantly lower (better) scores on balance in objective joint function (P = 0.008); mean difference was ‐0.06 [95% CI, ‐0.10 to ‐0.02].

14. All withdrawals: Three studies provided data on this outcome (Clegg 2006; Das 2000; Nguyen 2001). No significant difference between active treatment and placebo groups was noted (Risk Ratio [RR] 1.24, 95% CI 0.40 to 3.85; P = 0.70).

15. Withdrawals due to adverse events: Four studies provided data (Clegg 2006; Das 2000; Nakasone 2011;Nguyen 2001). The numbers of participants who withdrew from the study because of adverse events did not differ significantly between active and placebo groups (RR 1.21, 95% CI 0.57 to 2.55; P = 0.62).

16. Withdrawals due to inefficacy: Three studies provided data (Clegg 2006; Das 2000; Nguyen 2001). No significant difference was noted between chondroitin and placebo in the number of withdrawals due to inefficacy (RR 0.76, 95% CI 0.41 to 1.41; P = 0.39).

17. Number of adverse events: Three studies provided data (Das 2000; Nakasone 2011;Nguyen 2001). No significant difference was noted in the number of adverse events between chondroitin and placebo groups (RR 1.11, 95% CI 0.67 to 1.84; P = 0.68).

18. Number of serious adverse events: Three studies provided data (Das 2000; Nguyen 2001; Sawitzke 2010). No significant difference between placebo and chondroitin groups was reported (RR 1.45, 95% CI 0.77 to 2.75; P = 0.25).

19. Gastrointestinal adverse events: Two studies provided data (Das 2000; Nguyen 2001). No difference between chondroitin and placebo groups was reported (RR 0.68, 95% CI 0.34 to 1.37; P = 0.28).

20. Hematologic adverse events: Two studies provided data (Das 2000; Nguyen 2001). No difference between chondroitin and placebo groups was noted (RR 0.34, 95% CI 0.01 to 8.15; P = 0.51).

21. Other adverse events: These were defined as adverse events other than gastrointestinal, cardiac, or hematologic. Two studies provided data (Das 2000; Nguyen 2001). The number of other adverse events was not significantly different between chondroitin and placebo groups (RR 1.35, 95% CI 0.43 to 4.19; P = 0.61).

22. Death: Four studies provided data (Clegg 2006; Das 2000; Nguyen 2001; Sawitzke 2010). No significant difference between the two groups was reported (RR 0.34, 95% CI 0.01 to 8.23; P = 0.51).

CHONDROITIN SULFATE PLUS GLUCOSAMINE VERSUS NSAIDs CONTROL

(See Table 4 under Data and Anayses.)

1. Pain: Four studies provided data (Artemenko 2005; Clegg 2006; Lila 2005; Sawitzke 2010). Participants in the chondroitin sulfate and glucosamine group had significantly lower pain scores than those in the NSAIDs control group (SMD ‐1.41, 95% CI ‐2.18 to ‐0.63; P = 0.0004). Two studies provided short‐term data (Lila 2005; Sawitzke 2010), and four studies provided long‐term data (Artemenko 2005; Clegg 2006; Lila 2005;Sawitzke 2010). Short‐term studies showed no significant difference between NSAIDs control and chondroitin sulfate in combination with glucosamine (SMD ‐1.41, 95% CI ‐4.41 to 1.58; P = 0.35). Heterogeneity was high among these studies (T² = 4.60). Long‐term studies showed that participants in the chondroitin and glucosamine combination group experienced significantly better pain scores than those in the NSAIDs control group (SMD ‐1.48, 95% CI ‐2.51 to ‐0.44; P = 0.005). Heterogeneity was high (T² = 1.02).

Sensitivity analyses were robust to study sponsorship and publication year. Results were not robust to allocation concealment, blinding, study size, and impact of true ITT analyses.

In groups with both adequate and unclear methods of allocation concealment the estimate was similar to the original estimate. However, among studies with inadequate methods of allocation concealment and blinding, small studies (N<100), unclear pharmaceutical sponsorship, and high risk of bias due to unclear or non‐use of ITT analyses, participants treated with chondroitin sulfate had statistically better pain scores than those in the NSAID group.

2. MCII WOMAC: One study provided data (Clegg 2006). No significant difference was reported between chondroitin sulfate in combination with glucosamine and the NSAIDs control (P = 0.33); 0.85 [0.61, 1.19].

3. Physical function: Two studies provided data (Clegg 2006; Sawitzke 2010). No significant difference was noted between chondroitin sulfate in combination with glucosamine and the NSAIDs control with SMD in short‐ and long‐terms studies of 0.04 [95% CI, ‐0.20 to 0.28; p=0.35] and 0.02 [95% CI, ‐0.11 to 0.15; p=0.005], respectively. This translates to mean differences of 0.80 [95% CI, ‐4.09 to 5.69] and 0.34 [95% CI, ‐2.46 to 3.14], respectively.

4. WOMAC stiffness: Three studies provided data (Artemenko 2005; Clegg 2006; Lila 2005). Participants in the chondroitin sulfate in combination with glucosamine group achieved significantly better scores in WOMAC Stiffness than those in the NSAID group with mean differences in short‐ and long‐terms studies of ‐2.50 [95% CI, ‐4.27 to ‐.073; p=0.006] and ‐7.72 [95% CI, ‐15.36 to ‐0.08; p=0.05], respectively.

5. WOMAC Total: Three studies provided data (Artemenko 2005; Clegg 2006; Lila 2005). No significant difference between chondroitin sulfate in combination with glucosamine and NSAIDs control  was noted (MD ‐4.77, 95% CI ‐9.75 to 0.20, P = 0.06). WOMAC Total for Clegg 2006 was presented as a normalized WOMAC score on a 0 to 300 scale, which we normalized further to a 0 to 100 scale.

6. Patient global assessment of good to very good: Three studies provided data (Alekseeva 2005a; Alekseeva 2005b; Lila 2005). Significantly more participants in the chondroitin sulfate in combination with glucosamine group were able to achieve a good to very good global assessment as compared with the NSAIDs control (SMD 1.43, 95% CI 1.29 to 1.58; P < 0.00001).

7. Physician global assessment of good to very good: Three studies provided data (Alekseeva 2005a; Alekseeva 2005b; Lila 2005). Significantly more participants in the chondroitin sulfate in combination with glucosamine group said they were able to achieve a good to very good global assessment as compared with those in the NSAIDs control group (SMD 1.51, 95% CI 1.35 to 1.68; P < 0.00001).

8. OMERACT‐OARSI responder: One study provided data (Clegg 2006). The proportion of participants achieving responder status in the chondroitin sulfate group was not significantly different from the proportion in the active treatment group (P = 0.65); RR was 0.93 [95% CI 0.67 to 1.29].

9. HAQ Disability score: One study provided data (Clegg 2006). No significant difference was reported between chondroitin sulfate in combination with glucosamine and NSAIDs control (P = 0.77); mean difference was 0.01 [95% CI ‐0.06 to 0.08].

10. Radiographic outcome: change in mean JSW: One study provided data (Sawitzke 2010). No difference was noted between chondroitin sulfate in combination with glucosamine and NSAIDs control (P = 0.63); mean difference was 0.08 [95% CI, ‐0.26 to 0.42].

11. All withdrawals: Five studies provided data on this outcome (Alekseeva 2005a; Alekseeva 2005b; Artemenko 2005; Clegg 2006;Lila 2005). No significant difference between chondroitin sulfate in combination with glucosamine and active treatment was reported (RR 0.31, 95% CI 0.08 to 1.18; P = 0.09).

12. Withdrawals due to adverse events: One study provided data (Clegg 2006). The number of participants who withdrew from the study because of adverse events did not differ significantly between chondroitin sulfate in combination with glucosamine and NSAIDs control groups (RR 1.72, 95% CI 0.69 to 4.31; P = 0.25).

13. Withdrawals due to inefficacy: One study provided data (Clegg 2006). The number of participants who withdrew from the study because of inefficacy did not differ significantly between chondroitin sulfate in combination with glucosamine and NSAIDs control groups (RR 1.55, 95% CI 0.74 to 3.26; P = 0.25).

14. Adverse events: Four studies provided data (Alekseeva 2005a; Alekseeva 2005b; Lila 2005; Sawitzke 2010). No significant difference was reported in adverse events between chondroitin sulfate in combination with glucosamine and NSAIDs control groups (SMD 0.45, 95% CI 0.17 to 1.21; P = 0.11).

15. Serious adverse events: Two studies provided data (Alekseeva 2005a; Lila 2005). No significant difference was noted between the chondroitin sulfate in combination with glucosamine and NSAIDs control groups (RR 3.00, 95% CI 0.13 to 70.83; P = 0.50).

16. Gastrointestinal adverse events: Three studies provided data (Alekseeva 2005a; Clegg 2006; Lila 2005). No difference was reported between the chondroitin sulfate in combination with glucosamine and NSAIDs control groups (RR 0.54, 95% CI 0.29 to 1.01; P = 0.06).

17. Other adverse events: These were defined as adverse events other than gastrointestinal, cardiac, or hematologic. Two studies provided data (Alekseeva 2005a; Lila 2005). The number of other adverse events was not significantly different between chondroitin sulfate in combination with glucosamine and NSAIDs control groups (RR 0.42, 95% CI 0.16 to 1.13; P = 0.08).

18. Death: One study provided data (Clegg 2006). No deaths were reported in either the chondroitin sulfate in combination with glucosamine group or the NSAIDs control group (not estimable).

CHONDROITIN SULFATE ALONE OR WITH GLUCOSAMINE OR WITH OTHER SUPPLEMENT VERSUS PLACEBO OR CONTROL

(See Table 5 under Data and Anayses.)

In this group, we combined all studies in all comparison arms to assess the overall effect of chondroitin (with or without glucosamine) compared with placebo or control resulting in a greater number of studies for the outcomes of pain and physical function. This decision was clinically relevant because many patients take glucosamine and chondroitin together.

1. Pain (All studies): Before sensitivity analysis was performed, when all studies were combined into one large overall group, 17 studies were identified (Artemenko 2005; Bourgeois 1998; Bucsi 1998; Clegg 2006; Kanzaki 2011; Lila 2005; Mazieres 2001; Messier 2007: Morreale 1996; Nakasone 2011; Pavelka 1999; Pavelka 2010; Uebelhart 1998; Uebelhart 2004; Wildi 2011; Railhac 2012; Zegels 2012). Participants treated with chondroitin alone and in combination with other supplements such as glucosamine showed significantly better pain scores than those treated with a control or placebo in short‐ and long‐terms studies with an SMD ‐0.65, 95% CI ‐0.95 to ‐0.35; P < 0.0001). This translates into an absolute risk difference of which translated into an absolute risk difference of ‐9.6 (95% CI ‐14 to ‐5.2) on a 0 to 100 scale (absolute percent change 10% lower (14% lower to 5% lower ;T² = 0.33; n = 17 trials; level of evidence, low). Because were more than 10 studies were included, we created a funnel plot to detect bias. As Figure 6 shows, the possibility of publication bias is high. Removing one study with a small standard deviation (Lila 2005), reduced the SMD for pain to ‐0.47 (95% CI ‐0.70 to ‐0.23) and I² to 84%. 

Figure 6

---

---

Funnel plot of comparison: 5 chondroitin sulfate/CSGH versus Placebo/Control, outcome: 5.1 Pain on a 0 to 100 scale (short‐ and long‐term results).

Sensitivity analyses showed that the results were robust to blinding but not robust to allocation concealment, study size, study sponsorship, publication year, and impact of true ITT analyses (Analysis 9.1; Analysis 13.1; Analysis 17.1; Analysis 21.1; Analysis 25.1; Analysis 34.1). Among studies with adequate allocation concealment, sample size greater than 100, no pharmaceutical sponsorship, publication year >2010, and high risk of bias due to the use of non‐ITT analyses, there was no statistically significant difference between treatment and control groups.

1.1 Pain ‐ restricted to chondroitin dose ≧800 mg/day (therapeutic dose): 14 studies were included in this analysis (Artemenko 2005; Bourgeois 1998; Bucsi 1998; Clegg 2006; Lila 2005; Mazieres 2001; Messier 2007: Morreale 1996; Pavelka 2010; Uebelhart 1998; Uebelhart 2004; Wildi 2011; Railhac 2012; Zegels 2012) .Participants treated with chondroitin alone at a therapeutic dose of ≧800 mg/day and in combination with other supplements such as glucosamine showed significantly better pain scores than those treated with a control or placebo in short‐ and long‐terms studies with an SMD ‐0.67, 95% CI ‐0.99 to ‐0.34; P < 0.0001); absolute percent change 10% lower (14.6% lower to 5% lower) (Figure 7).

Figure 7

---

---

Forest plot of comparison: 5 Chondroitin sulfate or Chondroitin sulfate + Glucosamine versus Placebo or Control, outcome: 5.5 Pain on 0 to 100 scale (short‐ or long‐term) for CS dose >= 800 mg/day.

2. Physical function (All studies—physical function): Overall, before sensitivity analysis was performed, five studies provided data on physical function (Clegg 2006; Kanzaki 2011; Messier 2007; Pavelka 2010; Uebelhart 1998). No significant difference was reported between chondroitin or a combination of chondroitin with other supplements and placebo or control (SMD ‐0.07, 95% CI ‐0.31 to 0.17) which equates to an absolute risk difference of ‐1% lower (‐6% lower to 3% higher; level of evidence, moderate).

Sensitivity analyses demonstrated that results were robust with respect to allocation concealment, blinding, study size, study sponsor, and impact of use of non‐ITT analyses (Analysis 9.2; Analysis 13.2; Analysis 17.2; Analysis 21.2; Analysis 25.2; Analysis 34.2). Results were not robust to publication year, with one study published 1999 showing a statistically significant result for improvement in function in favour of the treatment group.

3. Lequesne's index: 10 studies provided data. Chondroitin with or without glucosamine was associated with statistically significantly better score on Lequesne's index compared to control or placebo, with a mean difference of 2.14 lower (95% CI 1.39 lower to 2.88 lower; p<0.00001) and SMD of ‐0.48 (95% CI ‐0.72 to ‐0.24).

SUMMARY OF SENSITIVITY ANALYSIS BY RISK OF BIAS AND OTHER CRITERIA

(See Tables 5‐25 and Tables 31‐34 under Data and Anayses.)

In summary, the beneficial effects of chondroitin on pain severity persisted when evidence was limited to studies with adequate blinding or studies using ITT analyses. On the other hand, pain reduction effects were smaller or were nonexistent when we limited data to studies with appropriate allocation concealment or a large study sample (> 200) or to studies without pharmaceutical funding.

SENSITIVITY ANALYSIS INCLUDING STUDIES WITH AN ESTIMATED STANDARD DEVIATION

(See Tables 26‐30 under Data and Anayses.)

In this analysis, along with studies that provided standard deviations in their results section, we included studies for which the standard deviation was not provided and was, therefore, estimated using the estimation method described in the Methods section. This analysis revealed significant differences in the results on the efficacy of chondroitin in several outcomes under each comparison group.

In the chondroitin sulfate versus placebo group, five studies were added when we included estimated standard deviation studies (Conrozier 1992; Conrozier 1998; L'Hirondel 1992; Alekseeva 1999; Alekseeva 2005b) for pain outcome and six studies for Lequesne's index (Conrozier 1992; Conrozier 1998; L'Hirondel 1992; Alekseeva 1999; Das 2000; Rai 2004) . The addition of these studies did not affect the results for pain and Lequesne's index.

In the comparison group of chondroitin sulfate versus control, one study was added (Alekseeva 1999). The addition of this study resulted in notable differences in the outcomes of pain and Lequesne's index. For pain, the main analysis revealed no significant difference between chondroitin sulfate and control (SMD ‐0.08, 95% CI ‐0.23 to 0.07; P = 0.29); this held true after Alekseeva 1999 was added with an estimated standard deviation (SMD 0.34, 95% CI ‐0.85 to 0.18; P = 0.20). In the main analysis, Lequesne's index showed no significant difference between control and active treatment with a mean difference of ‐1.36 (95% CI ‐3.60 to 0.89; P = 0.24). After Alekseeva 1999 was added to the meta‐analysis, however, these results switched with a mean difference of ‐1.99 (95% CI ‐3.27 to ‐0.70; P = 0.002).

The sensitivity analysis revealed no noteworthy difference between active treatment and control in the glucosamine and chondroitin sulfate versus placebo comparison group after studies with estimated standard deviations were added (Das 2000).

In the comparison group of chondroitin sulfate and glucosamine versus NSAIDs, the sensitivity analysis revealed notable changes in physical function, WOMAC stiffness, and WOMAC Total. For physical function, a moderate difference between the results of the main analysis and the results of the sensitivity analysis was observed, with the SMD increasing to ‐0.17 (95% CI ‐0.57 to 0.23; P = 0.41). In the outcome of stiffness on WOMAC, a slight difference was noted between the main analysis and the sensitivity analysis, with the mean difference decreasing to ‐8.49 (95% CI ‐14.52 to ‐2.46; P = 0.006). In WOMAC Total, a significant difference was observed, with the results switching from non significant to significant in favor of chondroitin (MD ‐6.94, 95% CI ‐12.86 to ‐1.01; P = 0.02).

In the comparison arm of chondroitin with/without glucosamine versus placebo/control, a moderate difference in the results was noted after the addition of studies with estimated SDs (SMD ‐0.73, 95% CI ‐1.00 to ‐0.46; P < 0.00001 for pain). Removing one study with a small standard deviation (Lila 2005), the SMD was ‐0.52 (95% CI, ‐0.71 to 0.33; I² = 82%) for pain. 

Studies Not Included in Data Analyses

Nguyen 2001 studies the efficacy of chondroitin sulfate and glucosamine hydrochloride in participants diagnosed with capsulitis, disk displacement, disk dislocation, or painful osteoarthritis of the temporomandibular joint (TMJ). This study found no statistically significant decrease in the VAS pain ratings of participants treated with chondroitin sulfate‐glucosamine hydrochloride, although it did find that participants in the placebo group experienced a statistically significant decrease in their pain ratings. In the outcome of the McGill Pain Questionnaire (MPQ), again no statistically significant change was observed in the sensory, affective, miscellaneous, total, and number of words pain rating indices in the chondroitin sulfate‐glucosamine hydrochloride group. Only in the evaluative Pain Rating index was a significant decrease seen in the pain rating, which changed from 3.4 to 2.4. It is interesting to note that participants in the placebo group again underwent statistically significant changes in sensory, evaluative, miscellaneous, and number of words Pain Rating indices. In the placebo group, the evaluative pain rating also decreased from 3.1 to 2. In summary, this study showed that participants treated with chondroitin sulfate‐glucosamine hydrochloride showed improvement in one scale of the MPQ, and participants treated with placebo experienced statistically significant improvement on four scales of the MPQ. In measures of mood and functioning, the results were again interesting, as participants in the chondroitin sulfate‐glucosamine hydrochloride group showed slight worsening of their mood and functioning score, although this decrease was not statistically significant. Participants in the placebo group, on the other hand, showed improvement in their mood and functioning, although this was not significant either. In the outcome of TMJ palpation score, participants in the chondroitin sulfate‐glucosamine hydrochloride group showed a statistically significant decrease, and those in the placebo group did not. In the outcome of myofascial pain, no statistically significant improvement was observed in the placebo group or the active group. In the outcome of jaw range of motion, no statistically significant improvement was observed. However, this study did show that chondroitin sulfate‐glucosamine hydrochloride participants took statistically significantly less medication per day than those in the placebo group (1.42 in the placebo group vs 0.72 in the active group).

Rovetta 2004 studies the efficacy of chondroitin at 800 mg daily plus naproxen versus naproxen alone in participants with erosive osteoarthritis of the hands. This study found that participants in the active and placebo groups showed worsening in their scores on nearly all outcomes as compared with baseline. In the outcomes of degree of erosion, Heberden and Bouchard nodes, and Dreiser index scores, the treated group showed significant worsening, and the Patient and Physician global assessment showed no significant change. The untreated group also showed significant worsening in degree of erosion, Heberden and Bourchard nodes, Dreiser index, and Physician and Patient global assessment scores. Using the Mann‐Whitney test, the authors concluded at the end of the study that the results show less worsening in study outcomes in the treated group than in the placebo group.

Rovetta 2002 is a study conducted to evaluate the effect of 800 mg of chondroitin sulfate in combination with naproxen per day on the joint count for erosions in participants with erosive osteoarthritis of the hands as compared with naproxen alone. In both the active group and the placebo group, the general tendency for the joint count was for erosion to increase over time. Progression of erosion over 24 months, however, was significantly less for participants in the chondroitin sulfate–treated group than for those treated with naproxen alone. After the first year, although 11 of the 12 participants in the chondroitin sulfate group had remained at the same joint erosion count, only six in the placebo group had done so. Although only one participant in the chondroitin sulfate group increased by one erosive joint count, five participants in the naproxen alone group experienced a one‐count increase in the number of their erosive joints. The authors concluded that the difference between participants with erosive joint count increases and those without in group B versus group A is statistically significant.

Alekseeva 2008 studied the efficacy of chondroitin sulfate plus glucosamine administered daily for nine months ("constant therapy") versus the efficacy of the two supplements administered for three‐month cycles with a treatment‐free cycle between the two treated three‐month cycles. The study revealed no significant difference between constant and intermittent treatment groups in WOMAC Total, physical function, pain and stiffness scales, or in the time it took to complete a 10‐feet week at the end of the 12‐month trial.

Magrans‐Courtney 2011 is a study undertaken to evaluate the efficacy of glucosamine and chondroitin sulfate in combination with a 14‐week exercise program and one of two different diet programs. All participants participated in a 14‐week exercise plan. Simultaneously, participants were randomly assigned to receive a high‐protein or a high‐carbohydrate diet program. Then, they were randomly assigned to receive a placebo or 1500 mg of glucosamine, 1200 mg of chondroitin, and 900 mg of methylsulfonylmethane (MSM) daily. The study found that participants in all groups experienced improvement in their physical functioning test, as well as on WOMAC, VAS, and quality of life measures. However, the authors do not note any significant correlation between the use of chondroitin sulfate, glucosamine, and MSM and improvement on these scales, stating, "no significant differences [were] observed among diet and supplement groups," which would indicate such a correlation.

Debi 2000 studied the efficacy of glucosamine sulfate and chondroitin sulfate in participants suffering from osteoarthritis. The authors reported on the outcome of Patient global assessment, finding that participants treated with glucosamine and chondroitin sulfate had significantly better global assessments than those treated with placebo.

Cohen 2003 studied the efficacy of a chondroitin cream versus a cream containing placebo. Participants in the chondroitin sulfate in combination with glucosamine group had significantly lower (better) pain scores on VAS (0 to 100 mm) after eight weeks. Those treated with chondroitin sulfate also had improved scores on pain on the WOMAC and SF‐36 scales.

Fardellone 2013 compared two chondroitin preparations, Structum® 500 mg twice daily or Chondrosulf® 400 mg three times daily over a 24‐week duration. There was no control group or placebo group.

Raynauld 2013 was a long‐term follow‐up study of Wildi 2011, that assessed 4‐year outcome of total knee replacement. Two groups of patients were treated with chondroitin or placebo once daily for the first 6 months (double‐blind phase) followed by 6 months of treatment with 800 mg chondroitin sulfate once daily for both groups (open‐label phase). Thirteen patients underwent total knee replacement at 4‐years, 9/34 in the placebo and 4/35 in the chondroitin group.  However since both patient groups were exposed to chondroitin, data from this study could not be used. However, the rates of arthroplasty seemed to be trending towards a difference between groups.

FDA and EMEA on Chondroitin

The U.S. FDA focused its recent review on the efficacy of chondroitin in reducing the risk of osteoarthritis and “....tentatively concluded that a relationship between glucosamine and chondroitin sulfate and a reduced risk of osteoarthritis is not established" (FDA 2004b).

In 2004, the FDA reviewed health claims from Weider Nutrition International, Inc., regarding glucosamine and chondroitin sulfate, and osteoarthritis, joint degeneration, cartilage deterioration, and osteoarthritis‐related joint pain, tenderness, and swelling (FDA 2004a). The FDA rejected these claims. 

The FDA did not evaluate the safety of chondroitin, citing “...the Agency did not perform a full safety review and make its own determination on this issue. It was not necessary for FDA to do so because the Agency is denying the proposed claims for lack of credible evidence, as discussed in section II below.”  We could not find any other safety data on the FDA website. A similar search on EMEA website revealed a waiver to Bioiberica S.A., Spain, but no information related to adverse events was provided.