Journal Logo

Research Article

Mixed Treatment Comparisons for Nonsurgical Treatment of Knee Osteoarthritis: A Network Meta-analysis

Jevsevar, David S. MD, MBA; Shores, Peter B. MPH; Mullen, Kyle MPH; Schulte, Danielle M. MS; Brown, Gregory A. MD, PhD; Cummins, Deborah S. PhD

Author Information
Journal of the American Academy of Orthopaedic Surgeons: May 1, 2018 - Volume 26 - Issue 9 - p 325-336
doi: 10.5435/JAAOS-D-17-00318
  • Free


Epidemiological research estimates that the lifetime risk of developing symptomatic knee osteoarthritis (KOA) is 45%.1 Because of the shifting demographics with an increasing percentage of the US population older than 65 years, the burden of KOA will continue to increase.2,3 Although a recent randomized controlled trial demonstrated that total knee replacement is more effective than nonsurgical treatment of end-stage KOA,4 effective nonsurgical treatments are required to manage KOA until surgical intervention becomes medically necessary.

Strong evidence of clinical effectiveness supports the use of self-managed strengthening, low-impact aerobic exercises, and nonsteroidal anti-inflammatory drugs (NSAIDs).5 Moderate evidence of clinical effectiveness supported the recommendation for weight loss in patients with a body mass index ≥25. Strong evidence supported the recommendation that viscosupplementation (intra-articular [IA] hyaluronic acid [HA] injections) was not clinically effective. This conclusion that IA HA injections do not provide clinically significant improvement is supported by two additional independent meta-analyses.6,7 Medical evidence for the treatment effectiveness of IA corticosteroid injections and platelet-rich plasma (PRP) injections was inconclusive.5

An evidence-based method is needed to determine the relative effectiveness of NSAIDs, acetaminophen, IA corticosteroid injections, IA PRP injections, and IA HA injections compared with oral placebo and IA placebo. This network meta-analysis (NMA) of high-quality clinical trials attempts to provide that evidence. Although there is strong evidence for the use of NSAIDs, this NMA uses statistical ranking techniques to provide evidence regarding which of the most common NSAIDs are most probable to decrease pain and improve function in patients with KOA and whether this is a comparable effect to other common conservative treatments. The NSAID studies also provide an effective source of overlapping data to generate indirect comparisons needed to fill gaps in evidence for more inconclusive and/or conflicting evidence for treatments such as HA, PRP, and corticosteroids. A previous NMA of KOA treatments included low-quality clinical trials with poor randomization, poor allocation, and/or poor blinding.8 Inclusion of low-quality clinical trials can bias the results of an NMA.9


Data Sources and Searches

We conducted a systematic search using PubMed, EMBASE, and Cochrane Central Register of Controlled Trials (CENTRAL) combined searches for KOA with randomized controlled trials (RCTs) for each intervention of interest. The most recent search for all databases was conducted on October 7, 2015. No date limits were used, results were restricted to the English language, and reference lists from relevant systematic reviews were manually searched. A total of 2956 articles were returned by an initial search of relevant articles; two reviewers assessed the abstracts and retrieved 727 articles for full-text review. Seventy-nine RCTs examining KOA treatments met all inclusion criteria for the NMA. Twenty-six of the included articles were excluded from the NMA because they were not the best available evidence as defined in the Study Selection section.

Study Selection

Articles were included in the NMA and evaluated for quality if the study met all criteria: (1) RCT; (2) written or translated into the English language; (3) human subjects; (4) evaluated treatment(s), outcome(s), and comparison(s) of interest; (5) population that included at least 80% patients with KOA; (6) a minimum of 30 patients per study group10; and (7) participants followed for at least 28 days. Treatments of interest included IA HA, IA corticosteroids, IA PRP, IA placebo, acetaminophen, diclofenac, ibuprofen, naproxen, celecoxib, and oral placebo. Two reviewers evaluated the quality of the studies on the following appraisal criteria: randomization, allocation concealment, blinding, completeness of outcome data, selective reporting, and the presence of other biases such as conflicts of interest or industry funding, confounding factors or treatments, lack of intention-to-treat analysis when applicable, and significant differences in baseline measurements. If potential risks of bias were present in these criteria, articles were downgraded from high quality for those particular design flaws. If the study's methodology was questionably biased but unclear for at least two appraisal criteria, the study would be downgraded to moderate quality; however, the clear presence of a risk of bias in the methodology would result in an immediate downgrade. A full list of quality flaws for each article is available in Supplementary Materials, Supplemental Digital Content 1, We assessed any appraisal disagreements on a case-by-case basis and reached a consensus on inclusion/exclusion.

When including articles in the meta-analysis, only articles fitting the best available evidence criteria were considered for each comparison. Criteria for the best available evidence were predetermined and defined as the highest-quality studies with consistent durations within each comparison, which is consistent with recommended procedures of previously published articles that used high-quality methodology and data.9,11 First, high-quality studies are included for the target outcomes for each comparison in the meta-analysis (excluding composite outcomes when data are not parsed into pain and function components). Three or more articles for a given comparison were necessary to perform the meta-analysis, and moderate-quality articles were included in analysis only when three or more high-quality articles were not present. If predetermined and reported qualitative risks were proven to display a sufficient and unacceptable level of quantitative statistical heterogeneity, articles with those risks were excluded from analysis for that comparison. Examples of qualitative risks monitored as sources of heterogeneity include potential confounding treatments, inconsistent number of injections, concomitant drug use, >20% loss to follow-up, or inadequate (missing or uninterpretable) reporting; a full list of exclusions and quality scores is available in Supplemental Materials, Supplemental Digital Content 1,

Data Extraction and Quality Assessment

Data from included studies were extracted by two abstractors from study tables and graphs as necessary. Consistent with previously published guidelines, data were extracted from the first follow-up time at or after 4 weeks, with the most common follow-up among studies being 42 days.5 From all studies, reported mean values or mean changes from baseline and SDs were extracted or calculated for the outcomes of interest (pain and function). All validated study outcome metrics assessing pain or function were uniformly converted to either a 0 to 100 visual analog scale for pain or a 0 to 100 Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) scale for function. The final data were then converted by the division of the uniform scores by minimal clinically important difference (MCID) units, which are defined as a reduction of 19.1 units on a 100-mm visual analog pain scale and a reduction of 8.0 units on a 100-mm WOMAC function scale.1214 The weighted mean difference of the MCID conversion results was used to derive the effect sizes of each treatment comparison (Table 1).

Table 1
Table 1:
Direct Comparisons, Homogeneity, and Consistency Values by Pain and Functional Outcomes

Data Synthesis and Analysis

As necessitated by the three assumptions of NMA, the homogeneity, transitivity, and consistency of the data were assessed by a clinical statistician.15,16 Homogeneity is useful in establishing the validity and interpretability of the results and strengthens the assumption of transitivity. Transitivity is the synthesis of indirect comparisons used to compare all treatment options not found in primary research. We used stringent inclusion criteria, quality appraisal, and homogeneity testing to fulfill the transitivity property, which was demonstrated by consistency testing (Table 1). After conversion, the outcomes were analyzed for heterogeneity using STATA software 12.1 to produce forest plots of direct comparisons and funnel plots to assess publication bias (Supplemental Materials, Supplemental Digital Content 1, The data were then entered into GeMTC GUI open source software 14.3 to generate precursory treatment rank comparison data and create code for analysis in OpenBUGS open source software. Bayesian medians were calculated for both pain and function measures using OpenBUGS software. The resulting estimates of mixed (combined direct and indirect) effects were deemed clinically significant at a value of ≥1.0 and as potentially clinically significant if the value was between 0.5 and 1.0.17 Indirect treatment effects were extrapolated from the differences of the mixed treatment effects and the direct treatment effects. Direct and indirect treatment comparisons were evaluated for consistency (Table 1).15,16 Transitivity was assessed through the comparison of treatment effects across trials, and meta-regression was performed to assess the reliability of the quality appraisal criteria (Supplemental Materials, Supplemental Digital Content 1,

The statistician then employed Bayesian posterior distribution to rank the effectiveness probabilities all treatments. The process was done empirically using the Markov Chain Monte Carlo sampling methodology across 200,000 iterations that were monitored for convergence; averages were then taken as estimates and probabilistic estimations were made.16 To decisively rank each treatment on a scale of 1 to 10 (rank 1 = most effective), we used a stepwise process for gathering the cumulative probabilities of each rank, starting from the highest to lowest rank. We first designated the most effective treatment as the treatment with the highest probability of ranking first. For rank 2, we found the treatment with the highest probability of ranking first or second by adding the probability values for rank 2 and rank 1 to get the cumulative rank probability sum (Supplemental Materials, Supplemental Digital Content 1, This process was repeated in a stepwise fashion to create definitive ranks reflecting their highest probabilities of effectiveness. The rank probability sum represents the current rank added to all rank probabilities before it to summarize its potential to place in any of those particular ranks. To determine the improvement of the combination of pain and function collectively, the average of the probability of each individual rank was calculated for each treatment. The stepwise process of cumulative rank probability sums was used as described above. Based on the software available, a noninformative prior was used for the Bayesian distribution.5


We assessed 53 references containing 56 included studies evaluating various comparisons for pain and/or function in the data analysis (Figure 1).1871 Most of the best available articles used in the NMA were of high quality, although some studies contained blinding concerns (1%), improper allocation concealment (3.5%), inadequate description of allocation concealment methods (20%), and inadequate randomization description (34%). In this context, “inadequate” often refers to articles describing processes as properly allocated and/or randomized but lacking details regarding the methods used. The assumptions of homogeneity, transitivity, and consistency were satisfied and supported by maintaining a heterogeneity coefficient (I2) equal to 0% in >90% of the included articles, whereas the remaining articles were still within the predetermined limit of <30% (Table 1). Only three high-quality articles with sufficient reporting were removed for high heterogeneity because of quality risks of inconsistent numbers of injection and concomitant drug use. All individual effect sizes, expressed in MCID units, and consistency values for all direct comparisons were found to be not significantly different compared with indirect comparisons (Table 1). Funnel plots and the random-effects meta-regression model assisted in verifying a lack of publication bias, based on the effect the quality appraisal components had on effect sizes of meta-analyzed treatment comparisons and individual included studies (Supplemental Materials, Supplemental Digital Content 1,

Figure 1
Figure 1:
Schematic illustration showing the number of articles by direct comparison and study quality.

Displayed as Bayesian medians, 25 mixed treatment comparisons of pain and function outcomes were found to have statistical significance, eight comparisons had large enough effect sizes to have the potential of clinical significance, and one comparison, favoring naproxen over oral placebo (1.18 [0.95 to 1.43]), had clinical significance (Figure 2). All active oral treatments (NSAIDs and acetaminophen) were found to be significant over oral placebo when evaluating pain or function. All active IA treatments were found to have statistical significance over oral placebo, whereas only corticosteroids showed significance over IA placebo when evaluating pain alone. When evaluating function, no active IA treatments showed any statistical significance over IA placebo or oral placebo. Because of their large effect size, IA corticosteroids (0.55 [0.27 to 0.86]) and IA PRP (0.52 [0.2 to 0.87]) were found to be potentially clinically significant for pain. Celecoxib (0.82 [0.65 to 1]), diclofenac (0.88 [0.56 to 1.2]), and ibuprofen (0.66 [0.17 to 1.15]) were found to be potentially clinically significant for function compared with oral placebo. Because of the large effect size of naproxen, it was found to be potentially clinically significant over acetaminophen, IA HA, and IA placebo (0.78 [0.38 to 1.17], 0.7 [0.04 to 1.38], and 0.79 [0.14 to 1.46]), respectively.

Figure 2
Figure 2:
Tables showing the mixed effects and clinical significance of all treatment comparisons for pain and function. Mixed effects = combination of direct and indirect effects generated from network meta-analysis. Bold indicates a statistically significant difference between treatments. *≥0.5 indicates a statistically significant difference between treatments with potential clinical significance for an appreciable number of patients. **≥1 indicates a clinically significant difference between treatments.

The five treatments with the highest probability of ranking 1 to 5 (1 being most effective) when evaluating pain reduction are IA corticosteroids, ibuprofen, IA PRP, naproxen, and celecoxib (Figure 3). The five treatments with the highest probability of ranking 1 to 5 when evaluating functional improvement are naproxen, diclofenac, celecoxib, ibuprofen, and IA PRP. When evaluating both pain reduction and functional improvement, the top 1 to 5 ranking treatments are naproxen, IA corticosteroids, IA PRP, ibuprofen, and celecoxib. On the basis of cumulative probability, all treatments were ranked 1 to 10 for outcomes of pain (rank 1 = IA corticosteroids), function (rank 1 = naproxen), and the combination of pain and function (rank 1 = naproxen) (Figure 3; Supplemental Materials, Supplemental Digital Content 1,

Figure 3
Figure 3:
Graphs showing the probability rankings of the most effective treatments. Cumulative ranking, as described in the Methods section, was used to more accurately reflect the true probability of one treatment being more effective than another. For example, corticosteriods were most probable to rank first; ibuprofen had the highest combined probability value for ranking first and second; platelet-rich plasma had the highest combined probability value for ranking first, second, third, and so on.


This NMA of the best available evidence demonstrated that IA placebo was found to have a statistically greater effect on decreasing pain compared with oral placebo; however, there was no statistical difference between the effect of IA placebo and oral placebo on function. Intra-articular corticosteroids were found to have a statistically greater effect on reducing pain than IA placebo, whereas all other IA treatments showed no statistical difference in effect. Intra-articular corticosteroids and IA-PRP demonstrate the “potential to be clinically significant” at decreasing pain because of a large statistically different magnitude of effect (0.55 [0.27 to 0.86] and 0.52 [0.2 to 0.87], respectively) compared with oral placebo.16 No IA treatments were found to be statistically significant over IA placebo in improving function. An RCT published after the completion of our secondary systematic search reported that an IA corticosteroid was favored for improving pain and function compared with placebo; however, this result did not reach statistical significance.72 This finding was consistent with the individual results of the three RCTs included in this NMA which, when combined, showed that IA corticosteroids were statistically better than placebo because of the increased statistical power. Therefore, inclusion of this recent study would not have affected the overall result.

Celecoxib, diclofenac, and ibuprofen demonstrate statistical significance and the “potential to be clinically significant” at increasing function because of a large statistically different magnitude of effect (0.82 [0.65 to 1], 0.88 [0.56 to 1.2], and 0.66 [0.17 to 1.15], respectively) compared with oral placebo. Naproxen is clinically significant at increasing function 1.18 (0.95 to 1.43) compared with oral placebo and demonstrates the “potential to be clinically significant” because of a large statistically different magnitude of effect over three other comparators (acetaminophen, IA-HA, and IA-placebo). Overall, naproxen was found to be the most effective single treatment for decreasing pain and improving function.

Importantly, all included direct comparisons in this NMA for pain and function were found to be quantitatively and qualitatively homogenous. All direct and indirect comparisons were found to be consistent with no statistical differences. Transitivity was satisfied by the careful selection of the inclusion criteria, the proven consistency between direct and indirect comparisons, and the qualitative and quantitative homogeneity of the included studies. Meta-regression analysis evaluated sources of heterogeneity among quality components, and funnel plots also showed no evidence of publication bias for either pain or function, further validating the inclusion criteria and higher quality of included studies (Supplemental Materials, Supplemental Digital Content 1,

Our NMA differs in many ways from that previously reported by Bannuru et al.8 This NMA included only the best available evidence, significantly reducing the potential for bias and improving homogeneity, resulting in observed effect sizes with more reliability and proper estimation of effect. Bannuru et al8 reported 20% blinding risk, 50% inadequate allocation concealment, and 45% inadequate or unclear randomization compared with 1%, 3.5%, and 34%, respectively, in this NMA. Bannuru et al8 included 52 articles, many containing confounded and questionable data, evaluating the comparison of HA with IA placebo. This resulted in the core argument that HA is the most effective treatment of KOA to be based on data confounded by unclear, unpublished, and/or insufficient studies. Our NMA found 12 articles to be high-quality evidence with low heterogeneity, resulting in the more valid observation that HA is not significantly different than IA placebo for the treatment of KOA. We also selected follow-up at the first follow-up at or after 4 weeks (most commonly, 42 days) versus the wider range of 2 to 6 months used by Bannuru et al,8 although the methods stated a 3-month interval. Each NMA used different tests of clinical importance, with this study using an anchor-based MCID methodology. Our NMA was also confined to the highest quality evidence published in English compared with unpublished data and foreign language studies used in the previous NMA. Finally, this NMA included PRP as an IA treatment of KOA.

The use of NSAIDs for arthritic conditions such as KOA has been shown to have a wide range of potential risks, including an increased risk of adverse cardiovascular effects. Indeed, the US Federal Drug Administration strengthened “an existing warning in prescription drug labels and over-the-counter drug fact labels to indicate that NSAIDs can increase the chance of a heart attack or stroke, either of which can lead to death.” Although the routine use of NSAIDs in chronic arthritic conditions should be cautioned, evidence indicates that naproxen has less potential for adverse cardiovascular events.73,74

This NMA supports naproxen as the conservative treatment of choice, as it is most probable to improve pain and function in patients with symptomatic KOA, followed by an IA corticosteroid, IA PRP, celecoxib, and ibuprofen. Probability ranking indicated that IA corticosteroids are the most likely to reduce pain, with ibuprofen, IA PRP, naproxen, and celecoxib being the top treatment choices. For function, NSAIDs demonstrate a higher probability of improved function, with naproxen having the highest rank, followed by diclofenac, celecoxib, ibuprofen, and IA PRP. The ability of NSAIDS to improve function more effectively than pain may be attributed to the fact that knee joint effusion is a contributor to limited knee joint function, and impaired function in patients with KOA is often secondary to the inflammatory aspect of the disease, which leads to joint effusion. Acetaminophen ranks near the bottom of the probabilistic rankings for pain, function, and combined pain and function. Acetaminophen shows superiority over oral placebo only for combined pain and function. Acetaminophen, with a relatively low level of effect at improving pain and function in patients with KOA and potential for hepatic toxicity, must be considered a lower-level treatment recommendation. Intra-articular HA does not achieve a rank in the top five in pain, function, or combined pain and function. Intra-articular HA's highest rank is in the pain analysis, where it ranks sixth. Although IA HA achieves superiority over acetaminophen, and both placebos, it does not demonstrate a better treatment effect than any of the other active treatments. Because of the demonstrated lack of superior efficacy over other active treatments, its relatively high cost, and reported adverse effects, the routine use of IA HA must be questioned.

This study had several potential limitations. As discussed previously in the results, moderate-quality studies were forcibly included in several instances because of the lack of direct treatment comparisons, specifically between IA and oral treatments. Furthermore, the link between IA and oral treatment comparisons contained no more than 1 direct comparison, if any. Indirect comparisons were evaluated where direct comparison data were lacking, and the results were proven consistent. Heterogeneity was eliminated between included studies for direct comparisons whenever possible, but in the few cases when it was not possible, a heterogeneity limit of <30% was still maintained. Durations were not always consistent among studies, so data at the timepoints closest to the most common duration used by all studies were used; the most common follow-up was 42 days (6 weeks). This NMA lacked the inclusion of exercise, physical therapy, and/or lifestyle changes as conservative comparators. It was determined that including these treatment options was not necessary because of the existence of strong evidence regarding these treatments, the complexities and barriers of comparing these techniques to other conservative treatments, and the ability to use these options in collaboration with the reviewed conservative treatments. This NMA also excluded any articles that were not written or translated to English, but the included references of past systematic reviews were evaluated and suggested that very few, if any, untranslated articles provide sufficient or effective evidence to this topic. Most of the direct comparisons were made using comparable dosing and drug preparations, but on a few occasions, there was variation in the compared doses. This was not a common occurrence, and indirect comparison results were highly consistent with direct comparisons, suggesting that this limitation had a little effect on the ultimate effectiveness rating.

Once the qualitative and quantitative variability was assessed for direct comparisons, incorporation into the Bayesian hierarchical random effects model allowed review of the mixed effects of all treatment comparisons. The probabilities of treatment effect were then displayed through the posterior distribution over 200,000 iterations, where the higher number of iterations would validate the true value of the results with increasing statistical accuracy. Bayesian medians and credibility intervals are presented, as they can be interpreted probabilistically as opposed to mean values and confidence intervals, which cannot be interpreted probabilistically.

Heretofore, physicians have struggled with approaches in conservative treatment for KOA. Concerns about efficacy, side effects, and cost have blurred the available treatment alternatives. This NMA, using the best available evidence for conservative, pharmacologic treatment of KOA, proposes a methodologic rank-order, which physicians can use to establish their treatment hierarchy. This study supports the use of naproxen sodium, which is both low cost and has a favorable cardiac risk profile, and demonstrates that naproxen is the most effective sole treatment for improving pain and function compared with other conservative KOA treatments. The combined greatest effect based on this NMA would be a combination of an IA corticosteroid and naproxen. All included NSAIDs demonstrated greater effect than oral and IA placebos. Acetaminophen was not different from oral placebo for pain or function. Intra-articular HA was no different from IA placebo. This NMA probability ranking, along with patient pain, function, and comorbidities, should be used to optimize treatment for patients with symptomatic KOA. Future KOA research should use naproxen and IA corticosteroids as active comparators to evaluate new or existing treatments.


References printed in bold type are those published within the past 5 years.

1. Murphy L, Schwartz TA, Helmick CG, et al.: Lifetime risk of symptomatic knee osteoarthritis. Arthritis Rheum 2008;59:1207-1213.
2. Projected state-specific increases in self-reported doctor-diagnosed arthritis and arthritis-attributable activity limitations: United States, 2005–2030. MMWR Morb Mortal Wkly Rep 2007;56:423-425.
3. Prevalence of doctor-diagnosed arthritis and arthritis-attributable activity limitation: United States, 2010–2012. MMWR Morb Mortal Wkly Rep 2013;62:869-873.
4. Skou ST, Roos EM, Laursen MB, et al.: A randomized, controlled trial of total knee replacement. N Engl J Med 2015;373:1597-1606.
5. Jevsevar DS, Brown GA, Jones DL, et al.: The American Academy of Orthopaedic Surgeons evidence-based guideline on: Treatment of osteoarthritis of the knee, 2nd edition. J Bone Joint Surg Am 2013;95:1885-1886.
6. Rutjes AW, Jüni P, da Costa BR, Trelle S, Nüesch E, Reichenbach S: Viscosupplementation for osteoarthritis of the knee: A systematic review and meta-analysis. Ann Intern Med 2012;157:180-191.
7. McAlindon TE, Bannuru RR, Sullivan MC, et al.: OARSI guidelines for the non-surgical management of knee osteoarthritis. Osteoarthritis Cartilage 2014;22:363-388.
8. Bannuru RR, Schmid CH, Kent DM, Vaysbrot EE, Wong JB, McAlindon TE: Comparative effectiveness of pharmacologic interventions for knee osteoarthritis: A systematic review and network meta-analysis. Ann Intern Med 2015;162:46-54.
9. Jevsevar D, Donnelly P, Brown G, Cummins D: Viscosupplementation for osteoarthritis of the knee. J Bone Joint Surg Am 2015;97:2047-2060.
10. Columbia University: Statistical sampling and regression: t-distribution. Accessed July 22, 2016.
11. Slavin RE: Best-evidence synthesis: An alternative to meta-analytic and traditional reviews. Educ Res 1986;15:5-11.
12. Angst F, Aeschlimann A, Michel BA, Stucki G: Minimal clinically important rehabilitation effects in patients with osteoarthritis of the lower extremities. J Rheum 2002;29:131-138.
13. Tubach F, Ravaud P, Baron G, et al.: Evaluation of clinically relevant changes in patient reported outcomes in knee and hip osteoarthritis: The minimal clinically important improvement. Ann Rheum Dis 2005;64:29-33.
14. Thorlund K, Walter SD, Johnston BC, Furukawa TA, Guyatt GH: Pooling health-related quality of life outcomes in meta-analysis: A tutorial and review of methods for enhancing interpretability. Res Syn Meth 2011;2:188-203.
15. Dias S, Welton NJ, Caldwell DM, Ades AE: Checking consistency in mixed treatment comparison meta-analysis. Stat Med 2010;29:932-944.
16. Dias S, Welton NJ, Sutton AJ, Ades AE: NICE DSU technical support document 2: A generalised linear modelling framework for pairwise and network meta-analysis of randomised controlled trials. 2011. Accessed April 2014.
17. Johnston BC, Thorlund K, Schünemann HJ, et al.: Improving the interpretation of quality of life evidence in meta-analyses: The application of minimal important difference units. Health Qual Life Out 2010;8:116.
18. Filardo G, DiMatteo B, DiMartino A, et al.: Platelet-rich plasma intra-articular knee injections show no superiority versus viscosupplementation: A randomized controlled trial. Am J Sports Med 2015;43:1575-1582.
19. Van der Weegen W, Wullems JA, Bos E, Noten H, van Drumpt RA: No difference between intra-articular injection of hyaluronic acid and placebo for mild to moderate knee osteoarthritis: A randomized, controlled, double-blind trial. J Arthroplasty 2015;30:754-757.
20. Prior MJ, Harrison DD, Frustaci ME: A randomized, double-blind, placebo-controlled 12 week trial of acetaminophen extended release for the treatment of signs and symptoms of osteoarthritis. Curr Med Res Opin 2014;30:2377-2387.
21. Gibofsky A, Hochberg MC, Jaros MJ, Young CL: Efficacy and safety of low-dose submicron diclofenac for the treatment of osteoarthritis pain: A 12 week, phase 3 study. Curr Med Res Opin 2014;30:1883-1893.
22. Housman L, Arden N, Schnitzer TJ, et al.: Intra-articular hylastan versus steroid for knee osteoarthritis. Knee Surg Sports Traumatol Arthrosc 2014;22:1684-1692.
23. Arden NK, Akermark C, Andersson M, Todman MG, Altman RD: A randomized saline-controlled trial of NASHA hyaluronic acid for knee osteoarthritis. Curr Med Res Opin 2014;30:279-286.
24. Essex MN, Beha R, O'Connell MA, Brown PB: Efficacy and tolerability of celecoxib and naproxen versus placebo in Hispanic patients with knee osteoarthritis. Int J Gen Med 2014;7:227-235.
25. Leighton R, Akermark C, Therrien R, et al.: NASHA hyaluronic acid vs. methylprednisolone for knee osteoarthritis: A prospective, multi-centre, randomized, non-inferiority trial. Osteoarthritis Cartilage 2014;22:17-25.
26. Conaghan PG, Dickson J, Bolten W, Cevc G, Rother M: A multicentre, randomized, placebo- and active-controlled trial comparing the efficacy and safety of topical ketoprofen in transfersome gel (IDEA-033) with ketoprofen-free vehicle (TDT 064) and oral celecoxib for knee pain associated with osteoarthritis. Rheumatology (Oxford) 2013;52:1303-1312.
27. Filardo G, Kon E, DiMartino A, et al.: Platelet-rich plasma vs hyaluronic acid to treat knee degenerative pathology: Study design and preliminary results of a randomized controlled trial. BMC Musculoskelet Disord 2012;13:229.
28. Essex MN, Bhadra P, Sands GH: Efficacy and tolerability of celecoxib versus naproxen in patients with osteoarthritis of the knee: A randomized, double-blind, double-dummy trial. J Int Med Res 2012;40:1357-1370.
29. Huang TL, Chang CC, Lee CH, Chen SC, Lai CH, Tsai CL: Intra-articular injections of sodium hyaluronate (Hyalgan(R)) in osteoarthritis of the knee: A randomized, controlled, double-blind, multicenter trial in the Asian population. BMC Musculoskelet Disord 2011;12:221.
30. Schnitzer TJ, Hochberg MC, Marrero CE, Duquesroix B, Frayssinet H, Beekman M: Efficacy and safety of naproxcinod in patients with osteoarthritis of the knee: A 53-week prospective randomized multicenter study. Semin Arthritis Rheum 2011;40:285-297.
31. Schnitzer TJ, Kivitz A, Frayssinet H, Duquesroix B: Efficacy and safety of naproxcinod in the treatment of patients with osteoarthritis of the knee: A 13-week prospective, randomized, multicenter study. Osteoarthritis Cartilage 2010;18:629-639.
32. Chao J, Wu C, Sun B, et al.: Inflammatory characteristics on ultrasound predict poorer longterm response to intraarticular corticosteroid injections in knee osteoarthritis. J Rheumatol 2010;37:650-655.
33. Chevalier X, Jerosch J, Goupille P, et al.: Single, intra-articular treatment with 6 ml hylan G-F 20 in patients with symptomatic primary osteoarthritis of the knee: A randomised, multicentre, double-blind, placebo controlled trial. Ann Rheum Dis 2010;69:113-119.
34. Altman RD, Rosen JE, Bloch DA, Hatoum HT, Korner P: A double-blind, randomized, saline-controlled study of the efficacy and safety of EUFLEXXA for treatment of painful osteoarthritis of the knee, with an open-label safety extension (the FLEXX trial). Semin Arthritis Rheum 2009;39:1-9.
35. Simon LS, Grierson LM, Naseer Z, Bookman AA, Zev Shainhouse J: Efficacy and safety of topical diclofenac containing dimethyl sulfoxide (DMSO) compared with those of topical placebo, DMSO vehicle and oral diclofenac for knee osteoarthritis. Pain 2009;143:238-245.
36. Baltzer AW, Moser C, Jansen SA, Krauspe R: Autologous conditioned serum (orthokine) is an effective treatment for knee osteoarthritis. Osteoarthritis Cartilage 2009;17:152-160.
37. Lundsgaard C, Dufour N, Fallentin E, Winkel P, Gluud C: Intra-articular sodium hyaluronate 2 mL versus physiological saline 20 mL versus physiological saline 2 mL for painful knee osteoarthritis: A randomized clinical trial. Scand J Rheumatol 2008;37:142-150.
38. Puopolo A, Boice JA, Fidelholtz JL, et al.: A randomized placebo-controlled trial comparing the efficacy of etoricoxib 30 mg and ibuprofen 2400 mg for the treatment of patients with osteoarthritis. Osteoarthritis Cartilage 2007;15:1348-1356.
39. Rother M, Lavins BJ, Kneer W, Lehnhardt K, Seidel EJ, Mazgareanu S: Efficacy and safety of epicutaneous ketoprofen in transfersome (IDEA-033) versus oral celecoxib and placebo in osteoarthritis of the knee: Multicentre randomised controlled trial. Ann Rheum Dis 2007;66:1178-1183.
40. Altman RD, Zinsenheim JR, Temple AR, Schweinle JE: Three-month efficacy and safety of acetaminophen extended-release for osteoarthritis pain of the hip or knee: A randomized, double-blind, placebo-controlled study. Osteoarthritis Cartilage 2007;15:454-461.
41. Herrero-Beaumont G, Ivorra JA, Del Carmen Trabado M, et al.: Glucosamine sulfate in the treatment of knee osteoarthritis symptoms: A randomized, double-blind, placebo-controlled study using acetaminophen as a side comparator. Arthritis Rheum 2007;56:555-567.
42. Svensson O, Malmenas M, Fajutrao L, Roos EM, Lohmander LS: Greater reduction of knee than hip pain in osteoarthritis treated with naproxen, as evaluated by WOMAC and SF-36. Ann Rheum Dis 2006;65:781-784.
43. Clegg DO, Reda DJ, Harris CL, et al.: Glucosamine, chondroitin sulfate, and the two in combination for painful knee osteoarthritis. N Engl J Med 2006;354:795-808.
44. Birbara C, Ruoff G, Sheldon E, et al.: Efficacy and safety of rofecoxib 12.5 mg and celecoxib 200 mg in two similarly designed osteoarthritis studies. Curr Med Res Opin 2006;22:199-210.
45. Neustadt D, Caldwell J, Bell M, Wade J, Gimbel J: Clinical effects of intraarticular injection of high molecular weight hyaluronan (orthovisc) in osteoarthritis of the knee: A randomized, controlled, multicenter trial. J Rheumatol 2005;32:1928-1936.
46. Schnitzer TJ, Weaver AL, Polis AB, Petruschke RA, Geba GP. Efficacy of rofecoxib, celecoxib, and acetaminophen in patients with osteoarthritis of the knee: A combined analysis of the VACT studies. J Rheumatol 2005;32:1093-1105.
47. Lehmann R, Brzosko M, Kopsa P, et al.: Efficacy and tolerability of lumiracoxib 100 mg once daily in knee osteoarthritis: A 13-week, randomized, double-blind study vs. placebo and celecoxib. Curr Med Res Opin 2005;21:517-526.
48. Sheldon E, Beaulieu A, Paster Z, Dutta D, Yu S, Sloan VS: Efficacy and tolerability of lumiracoxib in the treatment of osteoarthritis of the knee: A 13-week, randomized, double-blind comparison with celecoxib and placebo. Clin Ther 2005;27:64-77.
49. Pham T, Le Henanff A, Ravaud P, Dieppe P, Paolozzi L, Dougados M: Evaluation of the symptomatic and structural efficacy of a new hyaluronic acid compound, NRD101, in comparison with diacerein and placebo in a 1 year randomised controlled study in symptomatic knee osteoarthritis. Ann Rheum Dis 2004;63:1611-1617.
50. Tannenbaum H, Berenbaum F, Reginster JY, et al.: Lumiracoxib is effective in the treatment of osteoarthritis of the knee: A 13 week, randomised, double blind study versus placebo and celecoxib. Ann Rheum Dis 2004;63:1419-1426.
51. Pincus T, Koch G, Lei H, et al.: Patient preference for placebo, acetaminophen (paracetamol) or celecoxib efficacy studies (PACES): Two randomised, double blind, placebo controlled, crossover clinical trials in patients with knee or hip osteoarthritis. Ann Rheum Dis 2004;63:931-939.
52. Altman RD, Akermark C, Beaulieu AD, Schnitzer T: Efficacy and safety of a single intra-articular injection of non-animal stabilized hyaluronic acid (NASHA) in patients with osteoarthritis of the knee. Osteoarthritis Cartilage 2004;12:642-649.
53. Gibofsky A, Williams GW, McKenna F, Fort JG: Comparing the efficacy of cyclooxygenase 2-specific inhibitors in treating osteoarthritis: Appropriate trial design considerations and results of a randomized, placebo-controlled trial. Arthritis Rheum 2003;48:3102-3111.
54. Sangdee C, Teekachunhatean S, Sananpanich K, et al.: Electroacupuncture versus diclofenac in symptomatic treatment of osteoarthritis of the knee: A randomized controlled trial. BMC Complement Altern Med 2002;2:3.
55. Geba GP, Weaver AL, Polis AB, Dixon ME, Schnitzer TJ: Efficacy of rofecoxib, celecoxib, and acetaminophen in osteoarthritis of the knee: A randomized trial. JAMA 2002;287:64-71.
56. McKenna F, Borenstein D, Wendt H, Wallemark C, Lefkowith JB, Geis GS: Celecoxib versus diclofenac in the management of osteoarthritis of the knee. Scand J Rheumatol 2001;30:11-18.
57. McKenna F, Weaver A, Fiechtner JJ, Bello AE, Fort JG: COX-2 specific inhibitors in the management of osteoarthritis of the knee: A placebo-controlled, randomized, double-blind study. J Clin Rheumatol 2001;7:151-159.
58. Davies GM, Watson DJ, Bellamy N: Comparison of the responsiveness and relative effect size of the Western Ontario and McMaster Universities Osteoarthritis Index and the Short-Form Medical Outcomes Study Survey in a randomized, clinical trial of osteoarthritis patients. Arthritis Care Res 1999;12:172-179.
59. Sandelin J, Harilainen A, Crone H, Hamberg P, Forsskahl B, Tamelander G: Local NSAID gel (eltenac) in the treatment of osteoarthritis of the knee: A double blind study comparing eltenac with oral diclofenac and placebo gel. Scand J Rheumatol 1997;26:287-292.
60. Henderson EB, Smith EC, Pegley F, Blake DR: Intra-articular injections of 750 kD hyaluronan in the treatment of osteoarthritis: A randomised single centre double-blind placebo-controlled trial of 91 patients demonstrating lack of efficacy. Ann Rheum Dis 1994;53:529-534.
61. Williams HJ, Ward JR, Egger MJ, et al.: Comparison of naproxen and acetaminophen in a two-year study of treatment of osteoarthritis of the knee. Arthritis Rheum 1993;36:1196-1206.
62. Bradley JD, Brandt KD, Katz BP, Kalasinski LA, Ryan SI: Comparison of an antiinflammatory dose of ibuprofen, an analgesic dose of ibuprofen, and acetaminophen in the treatment of patients with osteoarthritis of the knee. N Engl J Med 1991;325:87-91.
63. Asmus MJ, Essex MN, Brown PB, Mallen SR: Efficacy and tolerability of celecoxib in osteoarthritis patients who previously failed naproxen and ibuprofen: Results from two trials. Int J Clin Rheumatol 2014;9:551-558.
64. Raynauld JP, Buckland-Wright C, Ward R, et al.: Safety and efficacy of long-term intraarticular steroid injections in osteoarthritis of the knee: A randomized, double-blind, placebo-controlled trial. Arthritis Rheum 2003;48:370-377.
65. Gaffney K, Ledingham J, Perry JD: Intra-articular triamcinolone hexacetonide in knee osteoarthritis: Factors influencing the clinical response. Ann Rheum Dis 1995;54:379-381.
66. Gormeli G, Gormeli CA, Ataoglu B, Colak C, Aslanturk O, Ertem K: Multiple PRP injections are more effective than single injections and hyaluronic acid in knees with early osteoarthritis: A randomized, double-blind, placebo-controlled trial. Knee Surg Sports Traumatol Arthrosc 2017;25:958-965.
67. Verkleij SP, Luijsterburg PA, Willemsen SP, Koes BW, Bohnen AM, Bierma-Zeinstra SM: Effectiveness of diclofenac versus paracetamol in knee osteoarthritis: A randomised controlled trial in primary care. Br J Gen Pract 2015;65:e530-e537.
68. Hochberg MC, Fort JG, Svensson O, Hwang C, Sostek M: Fixed-dose combination of enteric-coated naproxen and immediate-release esomeprazole has comparable efficacy to celecoxib for knee osteoarthritis: Two randomized trials. Curr Med Res Opin 2011;27:1243-1253.
69. Dickson DJ, Hosie G, English JR: A double-blind, placebo-controlled comparison of hylan G-F 20 against diclofenac in knee osteoarthritis. J Clin Res 2001;4:41-52.
70. Altman RD, Moskowitz R: Intraarticular sodium hyaluronate (hyalgan) in the treatment of patients with osteoarthritis of the knee: A randomized clinical trial. Hyalgan Study Group. J Rheumatol 1998;25:2203-2212.
71. Henriksen M, Christensen R, Klokker L, et al.: Evaluation of the benefit of corticosteroid injection before exercise therapy in patients with osteoarthritis of the knee: A randomized clinical trial. JAMA Intern Med 2015;175:923-930.
72. McAlindon T, LaValle MP, Harvey WF, et al.: Effect of intra-articular triamcinilone vs saline on knee cartilage volume and pain in patient with knee osteoarthritis: A randomized clinical trial. JAMA 2017;317:1967-1975.
73. Coxib and Traditional NSAID Trialists' (CNT) Collaboration: Vascular and upper gastrointestinal effects of non-steroidal anti-inflammatory drugs: Meta-analyses of individual participant data from randomised trials. Lancet 2013;382:769-779.
74. Ray WA, Varas-Lorenzo C, Chung CP, et al.: Cardiovascular risks of nonsteroidal antiinflammatory drugs in patients after hospitalization for serious coronary heart disease. Circ Cardiovasc Qual Outcomes 2009;2:155-163.

Supplemental Digital Content

Copyright 2018 by the American Academy of Orthopaedic Surgeons.