Effect of Oral Glucosamine on Joint Structure in Individuals With Chronic Knee Pain: A Randomized, Placebo-Controlled Clinical Trial


  • ClinicalTrials.gov identifier: NCT00377286.



To determine the short-term efficacy of oral glucosamine supplementation by evaluating structural lesions in the knee joints, as assessed using 3T magnetic resonance imaging (MRI).


This study was designed as a randomized, double-blind, placebo-controlled trial. Recruitment was performed via mass mailings and an arthritis registry in southwestern Pennsylvania. In total, 201 participants with mild-to-moderate pain in one or both knees, as defined by a Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain score ≥25 and ≤100, were enrolled. Of these subjects, 69.2% had a Kellgren/Lawrence grade ≥2 in at least 1 knee. Participants received 24 weeks of treatment with 1,500 mg glucosamine hydrochloride in beverage form or a placebo beverage. The primary outcome was decreased worsening of cartilage damage on 3T MRI of both knees, assessed according to a validated scoring system, the Whole-Organ MRI Score (WORMS). Secondary outcomes included change in bone marrow lesion (BML) scores in all knees and change in excretion of urinary C-terminal crosslinking telopeptide of type II collagen (CTX-II).


The adjusted odds ratio (OR) for the likelihood of decreased cartilage damage over 24 weeks in any WORMS-scored subregion of the knee in the glucosamine treatment group compared to the control group was 0.938 (95% confidence interval [95% CI] 0.528, 1.666). Compared to subjects treated with glucosamine, control subjects showed more improvement in BMLs (adjusted OR 0.537, 95% CI 0.291, 0.990) but no difference in worsening BMLs (adjusted OR 0.691, 95% CI 0.410, 1.166) over 24 weeks. There was no indication that treatment with glucosamine decreased the excretion of urinary CTX-II (β = −0.10, 95% CI −0.21, 0.002).


The results of this short-term study provide no evidence of structural benefits (i.e., improvements in MRI morphologic features or urinary CTX-II excretion) from glucosamine supplementation in individuals with chronic knee pain.

Osteoarthritis (OA), the most common form of arthritis ([1]), is a major public health problem. It is the primary cause of disability in the elderly, and is associated with high health care expenditures ([2]). There are currently no approved drugs to slow the structural progression of OA, and the current pharmacologic therapies for OA are suboptimal due to limited efficacy and concerns regarding their toxicity ([3]). Thus, individuals with OA often look for alternative therapies. Joint pain and arthritis are the third and fourth most common conditions, respectively, targeted for treatment using complementary and alternative therapies, and glucosamine is the second most commonly used nonvitamin, nonmineral natural product ([4]). According to a 2007 Gallup study, >10% of Americans older than age 18 years are current glucosamine users. Global sales of glucosamine supplements increased >60% between 2003 ($1.3 billion) and 2010 (more than $2.1 billion) ([5]).

Glucosamine supplementation is believed to slow the degradation of articular cartilage, one of the hallmark features of the OA disease process, through the inhibition of catabolic enzymes ([6, 7]). However, evidence to support the benefits of glucosamine in improving symptoms and decreasing structural progression in knee OA is very conflicting ([3, 8-15]). There are marked differences in the efficacy of glucosamine between industry-sponsored trials and independently funded trials. Two reviews of the efficacy of glucosamine and an evidence-based review by the Agency for Healthcare Research and Quality concluded that there was strong positive-reporting bias in industry-funded trials, and no evidence for the efficacy of glucosamine from independently funded trials ([16-18]).

Prior studies in which the effects of glucosamine on joint structure have been evaluated utilized plain radiographs to assess the presence and extent of cartilage damage ([13, 17, 19]); however, plain radiographs are an indirect, insensitive measure of structural progression in OA ([20, 21]), and there has been a call for the use of better biomarkers to evaluate the benefits of glucosamine ([22]). Magnetic resonance imaging (MRI) allows for the noninvasive, direct visualization of cartilage damage and other features of knee OA, such as subchondral bone marrow lesions (BMLs) ([23]). Higher levels of urinary C-terminal crosslinking telopeptide of type II collagen (CTX-II), a molecular marker of cartilage tissue degradation, have been shown to be associated with both the prevalence and the progression of radiographic OA of the knee and hip ([24-26]), and with the presence of both cartilage defects and cartilage loss noted on knee MRI ([27]). A decrease in BMLs is associated with a significant reduction in the urinary CTX-II levels over short intervals ([28]). Therefore, biomarkers such as MRI morphologic features and levels of urinary CTX-II excretion may be more sensitive measures for monitoring structural changes of disease progression and assessing the benefits of glucosamine supplementation.

In this study, we used a community-based recruitment approach and conducted a 24-week, randomized, double-blind, placebo-controlled trial to evaluate the efficacy of oral glucosamine hydrochloride on joint health in subjects with chronic knee pain. The primary outcome was decreased worsening of cartilage damage. Secondary outcome measures included decreased worsening of BMLs and decreased excretion of urinary CTX-II.


Setting and participants

Participants were between the ages of 35 years and 65 years and had symptoms of mild-to-moderate chronic, frequent knee pain typical of knee OA (Western Ontario and McMaster Universities Osteoarthritis Index [WOMAC] score ≥25 and ≤100) ([29]). They were recruited from the community through mass mailings and physician offices, and used the University of Pittsburgh Arthritis Registry with a telephone screening interview and an in-person screening visit at the Arthritis Research Clinic at the University of Pittsburgh. Individuals were excluded if they had a Kellgren/Lawrence (K/L) grade of 4 on radiographs of both knees ([30]), inflammatory arthritis, renal disease, liver disease, diabetes mellitus, cancer, plans for elective surgery in the subsequent 12 months, an inability to undergo MRI of the knee, or an inability to walk without a cane or other assistive device. In addition, individuals who were taking bisphosphonates or who had taken glucosamine or other dietary supplements for knee pain within the prior 6 months were excluded. Those individuals who were unwilling to avoid treatment of knee pain with the use of nonsteroidal antiinflammatory drugs (NSAIDs) or any pain relievers other than acetaminophen for 24 weeks were also excluded. This study was approved by the Institutional Review Board of the University of Pittsburgh.

Randomization and intervention

Eligible participants were randomized to 1 of 2 treatment groups in blocks of 10, using a computer-generated, blinded, randomization scheme. Each subject was randomly assigned to receive either 1,500 mg of oral glucosamine or a placebo beverage as control. The glucosamine was delivered in a 16-ounce bottle of diet lemonade, with the control group receiving a 16-ounce bottle of diet lemonade identical in appearance and taste but without glucosamine. Non-nutritive sweeteners and flavorings (e.g., sucralose and acesulfame K) were used to minimize the provision of additional calories and to ensure similar taste ([31]).

For this study, Regenasure glucosamine hydrochloride (Cargill) was used. The glucosamine content was tested periodically, and was found to remain stable over time. At enrollment, participants were instructed to drink 1 bottle of diet lemonade each morning. To enhance retention, monitor compliance, and record adverse events, all participants were contacted via telephone every 4 weeks from baseline to the 24-week followup visit.

Outcomes and followup


For MRI assessment of both knees, 3T MR images (Siemens Trio) were acquired using an OA Initiative MRI scanner. The MRI pulse sequence protocol was identical to the OA Initiative protocol but without the T2 multiecho spin-echo sequence and the 3-dimensional fast low-angle shot sequence (for more details, see http://www.oai.ucsf.edu) ([32]).

The primary outcome was decreased worsening of cartilage damage in each knee, as assessed at baseline compared to the 24-week followup visit. MR images were not obtained at the 12-week visit. Cartilage damage was scored in 14 articular subregions, using a validated semiquantitative scoring system, the Whole-Organ MRI Score (WORMS) ([33]). All of the readings were performed by an experienced musculoskeletal radiologist (FWR) who was blinded with regard to the clinical data and experimental group assignment, but not to time point. The scale used to assess cartilage damage ranged from 0 to 6, where 0 = normal cartilage morphology.

Change in BMLs was prespecified as a secondary outcome. BMLs were assessed in the same 14 articular subregions. The BML scores ranged from 0 to 3, where 0 = normal. In a modification of the WORMS scoring system that was developed for longitudinal readings, the use of coding of within-grade changes (i.e., definite change that does not cover a full grade increase or decrease in cartilage damage) for cartilage assessment was introduced ([34-36]). Any change, including a within-grade change, was defined as cartilage loss, and any full grade increase or decrease was defined as a change in BMLs.

The interreader reliability (weighted kappa value) for the readings of the different features between the reader in the present study and an experienced reader in another published study ([37]) was 0.62 for BMLs and 0.78 for cartilage morphology.

Biomarker analysis

Urine samples (second void of the morning) were collected and frozen at −20°C. Assays of the urine samples from all participants and at all time points were run at one time and in duplicate. Levels of urinary creatinine were measured using a kinetic method, based on the Jaffe reaction ([38]); the interassay coefficient of variation (CV) was 6.0%. Concentrations of urinary CTX-II were measured using an enzyme-linked immunosorbent assay procedure developed by Immunodiagnostic Systems; the interassay CV was 8.6%. The concentration of CTX-II (in ng/ml) was standardized to that of total urinary creatinine (in mmoles/ml), and the corrected CTX-II concentration was expressed in ng/mmoles ([24]). For the adjusted CTX-II measurement, a lower score constituted an improvement.

Clinical assessment

Self-reported symptoms of knee pain and knee function were measured at baseline, 12 weeks, and 24 weeks. For these assessments, subjects completed the WOMAC 3.1, utilizing a visual analog scale ([29]).


Radiographs of the tibiofemoral joint (standing, semiflexed views; posterior-anterior projection) were obtained from all participants at the clinic screening visit ([39]). One experienced musculoskeletal radiologist (AG) graded all of the baseline knee radiographs for radiographic features of OA according to the K/L scale ([30]) and the Osteoarthritis Research Society International atlas for osteophytes and joint space narrowing (JSN) ([40]). Radiographic tibiofemoral OA was considered present if the K/L grade was ≥2. Scores were assigned for lateral and medial tibiofemoral osteophytes, lateral and medial tibiofemoral JSN, and K/L grade. One knee radiograph was not read because the individual had undergone a total knee replacement. The reader was blinded with regard to each subject's clinical data and experimental group.

Statistical analysis

Sample size estimates were based on the primary outcome, cartilage damage, as assessed by knee MRI. We calculated that 88 subjects would be required in each group to detect an odds ratio (OR) of 2.0 for the likelihood of decreased worsening of cartilage damage in the glucosamine group compared to the control group, with the alpha value set at 0.05, the statistical power set at 0.90, and an estimate of 5% of subregions exhibiting worsening of cartilage damage in the control group. These power calculations also assumed an average correlation of 0.1 between subregions within a knee in terms of cartilage worsening. Thus, after allowing for attrition, ∼100 subjects per group were required. All analyses were done on an intent-to-treat basis, with the last score carried forward in the case of missing data. For those who dropped out, this means that no change was assumed. In all analyses, the dependent variable was coded so that a positive coefficient or an OR >1 indicated a better outcome (either more improvement or less worsening) for the glucosamine group.

For the primary analysis of decreased worsening in cartilage damage, the data were examined by knee and by subregion. First, the number of subregions in a knee with any cartilage worsening was calculated, and the groups were compared according to whether subjects had 0, 1, or ≥2 subregions with cartilage worsening. This was analyzed using ordinal logistic regression, controlling for within-person correlation. We also used logistic regression in a model in which each subregion was considered a separate case, accounting for correlations or clustering within person/knee, to determine the odds of not worsening based on treatment assignment by each subregion. The model was adjusted for age, sex, body mass index (BMI) (normal, overweight, or obese), baseline WOMAC scores, and initial K/L grade, all of which are factors that may potentially influence the worsening of cartilage damage.

To examine the secondary outcome of the odds of change in BMLs, we utilized the same approach as described above, analyzing the data in logistic regression models and keeping each subregion of the knee as a separate case. Improving BMLs and worsening BMLs were modeled separately, with comparisons by subregion. Both models were adjusted and cluster-controlled as in the cartilage model. For the ancillary analysis involving assessment of changes in urinary CTX-II excretion, analysis of covariance (ANCOVA) was used, controlling for the same covariates as above, to compare the control and glucosamine treatment groups. Since the measurements of CTX-II were very skewed, log (natural)–transformed values were used in the analyses. For the ancillary analysis of the self-reported WOMAC scores, we used ANCOVA models that were controlled for age, sex, BMI, and K/L grade, to compare the rates of change in WOMAC scores between the 2 groups.



The recruitment and enrollment of participants occurred from September 2006 to October 2007 (Figure 1). The primary reasons for exclusion after the clinic screening visit included a WOMAC pain score of ≤25, and the use of excluded medications and/or treatments. An additional 129 participants dropped out before they completed their baseline visit. A total of 201 participants were enrolled, with 98 randomized to receive glucosamine hydrochloride and 103 to receive placebo.

Figure 1.

Flow diagram of enrollment and outcomes, according to the Consolidated Standards of Reporting Trials, in subjects with chronic knee pain randomized to receive either oral glucosamine or a placebo control beverage.

Table 1 shows the baseline characteristics of the study participants by treatment group. The glucosamine and control groups were similar with regard to age, BMI, sex, race, baseline WOMAC pain score, baseline WOMAC function score, and baseline WOMAC total score. There were no statistically significant differences by group in the percentage of participants taking narcotics, steroids, or NSAIDs over the course of the study. Similarly, there were no statistically significant differences in the use of any medications during the 1-week washout period before clinic assessments, or in the use of acetaminophen as rescue medication. The number of doses was also examined, and no statistically significant differences between the 2 groups were found. The groups were also similar in terms of the presence of radiographic OA in at least 1 knee, the presence of definite osteophytes, and the K/L grade on the baseline radiograph. The control group did, however, have more JSN, particularly medial JSN. There were no statistically significant differences in the baseline MRI findings between the groups with regard to either the number of knee subregions with cartilage damage or the severity of BMLs (Table 1).

Table 1. Characteristics of the study participants at baseline*
 Glucosamine groupPlacebo control group
  1. Except where indicated otherwise, values are the number (%) of subjects (n = 98 in the glucosamine group and n = 103 in the control group) for demographic data, Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) scores, use of pain medication, and radiographic features, and the number (%) of knees for baseline Kellgren/Lawrence (K/L) stage (n = 195 in the glucosamine group and n = 206 in the control group) and magnetic resonance imaging (MRI) features (n = 192 in the glucosamine group and n = 196 in the control group). BMI = body mass index; OA = osteoarthritis; JSN = joint space narrowing; BMLs = bone marrow lesions.
  2. aIncludes narcotics, steroids, cyclobenzaprine hydrochloride, and nonsteroidal antiinflammatory drugs.
  3. bIn the glucosamine group, 195 knees were assessed for K/L grade, due to 1 subject having undergone a total knee replacement.
  4. cEight individuals in the control group underwent MRI of 1 knee only.
Age, mean ± SD years52.17 ± 6.0552.29 ± 6.72
BMI, mean ± SD kg/m228.81 ± 4.1528.99 ± 4.44
Female51 (52.04)47 (45.63)
White88 (89.8)96 (93.2)
WOMAC score (scale 0–100), mean ± SD  
Pain subscale45.20 ± 14.0347.72 ± 16.65
Function subscale42.76 ± 16.8947.34 ± 19.60
Total43.98 ± 15.4348.16 ± 18.07
Use of any pain medicationa16 (17.2)20 (23.26)
Use of medications during 1-week washout19 (20.43)18 (20.93)
Use of acetaminophen as rescue medication55 (59.14)51 (59.3)
Radiographic features  
Radiographic OA, either knee68 (69.39)71 (68.93)
Any osteophytes (≥2 subregions)34 (17.44)42 (20.39)
Any JSN (≥2 subregions)24 (12.31)45 (21.84)
Any medial JSN (≥2 subregions)19 (9.74)40 (19.42)
Baseline K/L stageb  
Grade 066 (33.85)59 (28.64)
Grade 113 (6.67)26 (12.62)
Grade 226 (13.33)16 (7.77)
Grade 381 (41.54)90 (43.69)
Grade 49 (4.62)15 (7.28)
MRI featuresc  
Cartilage damage  
0–1 affected subregions60 (31.25)59 (30.1)
2–3 affected subregions64 (33.33)43 (21.94)
4–5 affected subregions29 (15.1)49 (25)
≥6 affected subregions39 (20.31)45 (22.96)
Worst (highest graded) BMLs, either knee  
Grade 052 (27.08)53 (27.04)
Grade 154 (28.13)64 (32.65)
Grade 258 (30.21)51 (26.02)
Grade 328 (14.58)28 (14.29)

Twenty-two of the participants who enrolled did not complete the study and had no followup MRI. All but 3 of these participants dropped out before the first followup interview at 12 weeks, of whom 5 were in the glucosamine group and 17 were in the control group. Thus, 16.5% of the control group did not complete the study, compared to only 5.1% of those receiving glucosamine (P = 0.0096 by chi-square test). The differences in the baseline characteristics between completers and noncompleters are summarized in Table 2.

Table 2. Baseline characteristics of the study completers compared to the study noncompleters
 Completers (n = 179)Noncompleters (n = 22)P
  1. aRadiographic osteoarthritis (OA) was defined as a Kellgren/Lawrence grade of ≥2.
  2. bNormal body mass index (BMI) was defined as 18.5 kg/m2 to <25 kg/m2.
  3. cWOMAC = Western Ontario and McMaster Universities Osteoarthritis Index.
Female, no. (%)82 (45.81)16 (72.73)0.0171
White, no. (%)162 (90.5)22 (100)0.1308
Radiographic OA in either knee, no. (%)a128 (71.5)11 (50.0)0.0393
Normal BMI, no. (%)b34 (18.99)9 (40.91)0.0497
Self-reported WOMAC score (scale 0–100), mean ± SDc   
Pain subscale46.03 ± 15.4650.25 ± 15.140.2264
Function subscale44.62 ± 18.0049.03 ± 21.640.2907
Total45.64 ± 16.6150.03 ± 19.230.2524

Primary outcome measure

In assessing the primary outcome measure of decreased cartilage worsening, we observed that in both groups at baseline, 88% of knees had prevalent cartilage damage in at least 1 subregion. Overall, across the 24 weeks of followup, the cartilage status declined (i.e., deterioration worsened) in 79 (1.41%) of the 5,614 subregions assessed. These results based on the primary outcome measure of decreased cartilage worsening are summarized in Table 3. There was no statistically significant difference in cartilage worsening between the glucosamine group and the control group (OR 0.846, 95% CI 0.465, 1.538). When all 5,614 subregions were included in an adjusted logistic regression model, controlling for clustering by subregions within knees and between knees within an individual, there was no association with less cartilage deterioration in the glucosamine group as compared to the control group (adjusted OR 0.938, 95% CI 0.528, 1.666).

Table 3. Magnetic resonance imaging outcomes by treatment group
 Overall, no. (%)Glucosamine group, no. (%)Placebo control group, no. (%)Odds ratio (95% confidence intervals)
  1. aAdjusted odds ratios >1 indicate improvement (either less worsening in cartilage and bone marrow lesions [BMLs] or improvement in BMLs) in the glucosamine group as compared to the control group. Models are adjusted for age, sex, body mass index, Western Ontario and McMaster Universities Osteoarthritis Index scores, and initial radiographic Kellgren/Lawrence grade, and are controlled for clustering by person/knee.
  2. bCompared to all other subregions as referent group (e.g., worsening BMLs compared to subregions with no change and subregions with improving BMLs).
  3. cSignificant at P < 0.05.
Primary outcome, decreased cartilage damage    
By knee (n = 401 knees)    
Worsening in at least 1 subregion37 (9.3)15 (7.7)22 (10.7)0.846 (0.465, 1.538)
Worsening in ≥2 subregions19 (4.8)10 (5.1)9 (4.4) 
By subregion (n = 5,614 subregions), worsening cartilagea79 (1.4)37 (1.4)42 (1.5)0.938 (0.528, 1.666)
Secondary outcome, change in BMLs by subregion (n = 5,614 subregions)    
Worsening BMLsa, b105 (1.9)58 (2.12)47 (1.63)0.691 (0.410, 1.166)
Improvement in BMLsa, b55 (1.0)19 (0.7)36 (1.3)0.537 (0.291, 0.990)c

Secondary outcome measures

Change in BMLs

Overall, over the 24-week followup, 70% of knees showed no change in BMLs, 18% of knees showed worsening of BMLs (in 105 subregions [1.87%]), and 10% of knees showed improvement in BMLs (55 subregions [0.98%]). Only 8 knees showed worsening in some subregions but improvement in others.

Improvement in BMLs and worsening in BMLs were modeled separately, and the models were controlled for clustering by knee and adjusted for demographic and clinical characteristics. In all analyses reported, an adjusted OR >1 indicated a better outcome (e.g., either more improvement in BMLs or less worsening in BMLs) for the glucosamine group as compared to the control group. Less worsening in BMLs was not associated with group assignment. However, more improvement in BMLs was associated with being in the control group (Table 3). The results were similar when changes in BMLs were assessed using a within-grade approach (results not shown).

Decreased excretion of urinary CTX-II

Evaluation of the urinary CTX-II (log-transformed) change scores was carried out using ANCOVA, with adjustment for demographic and clinical characteristics. In assessing the difference in urinary CTX-II excretion from baseline to 24 weeks, we observed a marginal association with a relative decline in CTX-II in the control group (β = −0.10, 95% CI −0.21, 0.002). However, neither the change in CTX-II excretion from baseline to 12 weeks (β = −0.04, 95% CI −0.15, 0.07) nor the change in CTX-II excretion from 12 weeks to 24 weeks (β = −0.06, 95% CI −0.17, 0.05) showed any association with group membership.

Self-reported pain and function

There were no significant differences in the WOMAC pain or function subscale scores or in the total WOMAC score between the glucosamine group and the control group from the baseline to 12-week assessment, the 12 -week to 24-week assessment, or the baseline to 24-week assessment. These results are summarized in Table 4.

Table 4. Changes in WOMAC scores in the glucosamine group compared to the placebo control group*
 Glucosamine group, mean ± SD (n = 98)Placebo control group, mean ± SD (n = 103)Adjusted β coefficientP
  1. Changes in Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) scores were assessed in an ordinary least squares model, controlled for age, sex, body mass index, and Kellgren/Lawrence radiographic grade. Positive beta indicates that the WOMAC score in the glucosamine group numerically declined more than in the control group.
Change, baseline to 12 weeks    
Pain−20.0714 ± 17.31−20.0893 ± 21.33−0.05650.9838
Difficulty−18.0873 ± 17.85−18.718 ± 22.30−0.59240.8381
Total−18.5597 ± 17.29−19.1691 ± 21.64−0.589740.8338
Change, baseline to 24 weeks   
Pain−17.402 ± 20.99−20.8893 ± 21.34−3.778090.212
Difficulty−15.4138 ± 21.34−19.404 ± 21.21−4.048070.1835
Total−15.7157 ± 20.48−19.9213 ± 20.85−4.350910.1409
Change, 12 weeks to 24 weeks    
Pain2.669388 ± 19.33−1.39223 ± 17.53−4.251990.1081
Difficulty2.673514 ± 18.48−1.50103 ± 15.25−4.196570.0828
Total2.843995 ± 18.14−1.49729 ± 15.50−4.433440.0664

Adverse events

There was no difference in reported side effects between the 2 groups. At the 24-week followup visit, 12.9% of the glucosamine group and 15.1% of the control group reported side effects (P = 0.6738 by Fisher's exact test). Seven subjects (4 in the control group and 3 in the glucosamine group) dropped out of the study because of side effects (e.g., nausea, weight gain, migraines).

Success of the blinding

A total of 63.89% of those in the glucosamine group thought they were receiving glucosamine, compared to 65.63% of those in the control group (P = 0.8591 by Fisher's exact test).

Adherence to the protocol

There was no significant difference in protocol compliance by group assignment. At the final assessment, there were no significant differences between the glucosamine group and control group with regard to missing at least 1 day of drinking the study beverage (62.6% versus 74.4%; P = 0.107 by Fisher's exact test) or with regard to the number of missed bottles (mean ± SD 6.43 ± 9.89 versus 8.84 ± 11.91; t = −1.47, P = 0.142 by t-test). Participants were instructed to bring back any bottles not consumed, and there was also no difference between the glucosamine group and the placebo group with regard to the number of bottles returned over the course of the study (mean ± SD 4.42 ± 16.69 versus 6.65 ± 16.32; t = −0.86, P = 0.3936 by t-test).


This is the first study to evaluate the benefits of glucosamine hydrochloride on joint health using 2 different MRI parameters to measure outcomes. In this 24-week study, we did not find any evidence that glucosamine is more effective than placebo in improving joint health, when assessed according to the outcomes of decreased cartilage deterioration on MRI, improvement of BMLs on MRI, decreased excretion of urinary CTX-II, and decreased pain or improved function. We used full and within-grade assessment of cartilage damage as a more sensitive measure of cartilage deterioration. When we used only full-grade assessment of cartilage damage, there was still no significant evidence that glucosamine was more effective than placebo. Our secondary outcome of change in BMLs suggested that there was less worsening of BMLs and more improvement in BMLs in the control group as compared to the glucosamine group. We evaluated the benefit of glucosamine over a study period of 24 weeks, since the results of the report from the Cochrane Collaboration indicated that glucosamine improves pain and function, as assessed using WOMAC scores, after 24 weeks ([14]).

Unlike prior studies of the structural benefit of glucosamine, radiographic knee OA was not a criterion for study entry. Since glucosamine is marketed as a nutraceutical in the United States and does not require a physician's prescription for its use, it is not likely that most individuals who will consume glucosamine will have documented radiographic knee OA. A limitation of prior studies of the structural benefit of glucosamine has been the focus on OA progression being defined as change in radiographic JSN. This required large sample sizes and long followup to assess structural benefit, whereas we chose MRI to assess disease progression, as it is the only imaging method that is able to directly visualize articular cartilage and is considered by many to have sensitivity superior to that of radiography in detecting progressive cartilage damage ([20, 41]). Two studies of glucosamine sulfate demonstrated a decrease in progression of radiographic JSN after 3 years of followup, but these were both industry-funded studies ([17, 19]). The Glucosamine/Chondroitin Arthritis Intervention Trial (GAIT), an NIH-funded study, did not show that glucosamine hydrochloride decreased progression of radiographic JSN ([13]).

Furthermore, we were unable to demonstrate a benefit of glucosamine over placebo with regard to improvement in BMLs. We chose BMLs as a secondary outcome because they have been shown to be associated with subregional cartilage loss ([35]). No prior studies have examined the benefit of glucosamine on changes in BMLs. Three recent studies, however, have demonstrated significant changes in BMLs over a 24-week period. Wildi and colleagues reported that the size of BMLs increased, as assessed using 1.5T MRI, over 6, 12, and 24 months of treatment in a randomized, controlled clinical trial ([42]). In a study by Kubota et al, enlargement of BMLs was observed over a 6-month followup period, with greater enlargement among those with a K/L grade of 3 compared to those with a K/L grade of 1 or 2 at baseline ([43]). The results of a study by Laslett et al indicated that treatment with zoledronic acid decreased the total BML area over 6 months ([44]). Garnero and colleagues observed a change in BMLs in ∼30% of subjects over a 3-month period, with reduction in the extent of bone marrow abnormalities being associated with a decrease in cartilage degradation as measured by CTX-II excretion ([45]).

Similarly, we were not able to demonstrate a benefit of glucosamine over placebo with regard to decreased excretion of urinary CTX-II. We chose urinary CTX-II as a secondary outcome since higher urinary levels of CTX-II have been shown to be associated with radiographic progression and cartilage loss ([27, 46]). Our results indicated that the excretion of urinary CTX-II was decreased in the control group compared to the glucosamine group. Our results also indicated that there was no benefit of glucosamine treatment in terms of WOMAC measures of pain at 12 weeks or 24 weeks.

There are a number of important potential limitations to our study. Although the dropout rate was only 10%, there was differential dropout by treatment group, with a greater proportion of the dropouts occurring in the control group. However, a completer analysis (results not shown), which was performed to check that the imputation involved in an intent-to-treat analysis was not determining the results, did not show any skewing toward the glucosamine group.

Our followup period was only 24 weeks. Nevertheless, recent studies have shown that therapeutic benefit, with evidence of improvement on MRI (e.g., improvement in BMLs or quantitative cartilage volume), can be detected over this time period ([44, 47]). There was only a small amount of worsening of cartilage damage in the control group during this time. Only 79 (1.41%) of the 5,614 possible subregions in both knees showed incident or worsening cartilage damage at the 24-week followup. This may have limited, in particular, our ability to detect the potential short-term and long-term benefits of glucosamine over placebo with regard to structural progression. For context, with the assumption that there was a 1.4% incidence of worsening in subregions of the knees in the placebo group, as was actually observed in the present study, rather than the 5% that we had estimated in planning the trial, the post hoc power to detect an effect size (OR) of 2.0 was 0.4312. Even though the study was underpowered, given the actual incidence of worsening, it is important to emphasize that the point estimate of the OR for all treatment effects was close to 1.0, and that the 95% confidence bounds did not include a sizeable treatment effect (e.g., OR of 2.0). There was limited worsening of outcomes in the control group, but our power calculations suggested that there was adequate power to detect a clinically significant association with cartilage damage as assessed by knee MRI, as well as urinary CTX-II excretion.

Furthermore, the size of the glucosamine and control groups in our study was comparable to that in 2 recent studies of glucosamine, the GAIT study and the GUIDE (Glucosamine Unum in Die Efficacy) study ([8, 9]), and we had similar entry criteria. Since radiographic knee OA was not a requirement for study entry, it is possible that the knee pain was due to causes other than knee OA, but the definition of frequent knee pain that we used is the one that is commonly used in epidemiologic studies of knee OA ([48]). It is also possible that a study population with more severe manifestations of OA (i.e., more radiographic damage) may have yielded a different result.

Although both incident cartilage damage and worsening cartilage damage over 24 weeks were observed in only a minority of participants, we believe that semiquantitative assessment is suited to detect such short-term changes in a sensitive manner. Focal cartilage defects that are commonly seen in knees without OA or in those with mild radiographic OA cannot be detected by other assessment methods. Only assessment by visual evaluation using scoring systems such as the WORMS or other semiquantitative methods can identify these lesions. Quantitative approaches based on segmentation would miss these minute changes ([49]).

In addition, quantitative approaches do not seem sensitive for detecting short-term change, as shown in a short-term study of patients with symptomatic knee OA in which only small, inconsistent compartment changes were observed over a 3–6-month period ([50]). To date, only a few studies have evaluated cartilage changes on MRI over periods as short as 6 months, and direct head-to-head comparisons between semiquantitative and quantitative approaches are missing from those studies. A recent study suggested that increased sensitivity to change could be achieved with volumetric assessment as compared to semiquantitative assessment over a 24-month period ([51]). It is unclear, however, whether volumetric assessment is superior to semiquantitative assessment over shorter periods such as 6 months. A definite shortcoming of semiquantitative assessment is inferiority in the detection of subtle, nonfocal cartilage loss over larger areas, which may not be obvious visually regardless of reader experience ([34]). Whether MRI methods such as compositional techniques are able to monitor change over short periods remains to be shown.

We studied glucosamine hydrochloride delivered in beverage form, whereas prior studies that have demonstrated a benefit of glucosamine have utilized glucosamine sulfate in tablet form ([14]). One report suggested that glucosamine hydrochloride may have different effects compared to glucosamine sulfate ([52]), but another review indicated that the pharmacokinetics of oral glucosamine salts are similar ([53]). Furthermore, a recent study showed that glucosamine hydrochloride inhibited bone resorption and bone remodeling in a murine model of OA ([54]).

Despite these limitations, the current study has several notable strengths. It is the first study to examine the structural benefits of glucosamine with the use of MRI to assess prevention of worsening of cartilage damage or worsening of BMLs. It assessed structural outcomes in both knees, in addition to a potential biomarker of OA progression, urinary CTX-II. Another strength is the consistency of the negative findings among more sensitive outcome measures of joint health, measures that encompass structural changes on MRI (i.e., cartilage damage and BMLs) and cartilage turnover (i.e., urinary excretion of CTX-II) and that parallel the lack of improvement in clinical symptoms (i.e., WOMAC knee pain and function).

The high retention rate of 89% and high adherence to the intervention is a further strength of our study. Moreover, by using a community-based recruitment strategy, the study sample is representative of the types of individuals with chronic knee pain who would be likely to take glucosamine.

In conclusion, the results of our study suggest that administration of glucosamine hydrochloride in a beverage for 24 weeks is not associated with less deterioration of cartilage damage in the knees, less worsening of BMLs, improvement of BMLs in the knees, decreased urinary excretion of CTX-II, and/or decreased pain and functioning in individuals with chronic knee pain from the community.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Kwoh had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Kwoh, Moore, Jakicic, Guermazi, Green.

Acquisition of data. Kwoh, Roemer, Hannon, Guermazi, Green.

Analysis and interpretation of data. Kwoh, Roemer, Hannon, Guermazi, Evans, Boudreau.


Funding for this study was provided by the Beverage Institute for Health & Wellness, The Coca-Cola Company. All statistical analyses were performed at the University of Pittsburgh and were partially supported by the Multidisciplinary Clinical Research Center for Musculoskeletal and Skin Diseases. The Coca-Cola Company had no role in the performance of the statistical analyses. Dr. Moore, formerly Principal Scientist at the Beverage Institute for Health & Wellness, The Coca-Cola Company, was involved with monitoring all aspects of study progress for the funding agency and coordinated the production of the glucosamine beverage, as well as the quality assurance checks on the stability of the beverage contents. Since 2008, Dr. Moore has been a faculty member at Texas Woman's University and was not associated with the Beverage Institute when the study radiographs and MRIs were analyzed. Dr. Moore participated in writing and gave final approval of the manuscript.