Issues in Pharmacological Management of Osteoarthritis

Jan 31, 2011

Many persons with osteoarthritis (OA) have only mild disease that can be managed successfully with reassurance, acetaminophen, and education in joint protection. However, patients with more severely symptomatic OA may be helped by a comprehensive program of pharmacological and nonpharmacological measures.

ABSTRACT: Many persons with osteoarthritis (OA) have only mild disease that can be managed successfully with reassurance, acetaminophen, and education in joint protection. However, patients with more severely symptomatic OA may be helped by a comprehensive program of pharmacological and nonpharmacological measures. The combination of aspirin, as may be taken for cardiovascular prophylaxis or for pain, and a nonselective NSAID is particularly damaging to the gastric mucosa. Upper GI ulceration and ulcer complications are serious adverse events associated with over-the-counter NSAIDs and prescription NSAIDs. The long-term effectiveness of NSAIDs and acetaminophen in patients with OA is far from satisfactory. For many patients, it is possible to discontinue NSAIDs, decrease the dose, or use NSAIDs only on an as-needed basis. (J Musculoskel Med. 2011;28:45-55)

Although many persons with osteoarthritis (OA) seek medical attention, a large proportion have only mild disease that can be managed successfully with reassurance, the use of acetaminophen, and education in principles of joint protection.1 Indeed, considering the adverse effects associated with the drugs that often are used for OA pain, these patients may do even better by avoiding care. However, many patients with more severely symptomatic OA may be helped by a comprehensive program of pharmacological and nonpharmacological measures, the rationale for which stems from understanding of the abnormal mechanophysiology of the OA joint.

For patients with knee OA, for example, helpful measures might involve instruction of the patient in principles of joint protection and in exercises to strengthen the periarticular muscles; weight reduction, if the patient is obese; avoidance of excessive loading of the arthritic joint with use of a properly fitted cane in the contralateral hand; use of shoes with well-cushioned soles; use of orthoses for patients with varus or valgus deformity; adoption of a toe-out gait and a slower walking speed, if the patient has medial tibiofemoral compartment OA; and patellar taping (for patellofemoral OA).

Implementation of such measures can improve pain and function and enhance quality of life, and the benefits may persist for years. Many such patients can be made better. For patients with significantly symptomatic OA, however, if physicians believe they can “do it all” with only a prescription for the pharmacist, they are likely to fail.

In this 3-part article, I review the pathogenesis, diagnosis, and management of OA. The first part (“Defining osteoarthritis: What it is, and what it is not,” The Journal of Musculoskeletal Medicine, September 2010, page 338) offered a contemporary, operational, evidence-based definition of the disease that has evolved from growing knowledge of its cause.

In the second part (“Osteoarthritis diagnosis: Avoiding the pitfalls,” The Journal of Musculoskeletal Medicine, November 2010, page 445), I discussed the clinical aspects of OA, including diagnosis and diagnostic pitfalls. This third part discusses some current approaches to treatment, focusing selectively on some pragmatic issues that arise in clinical practice relative to the pharmacological management of symptoms.

Concomitant use of NSAIDs and low-dose aspirin
Because OA affects mostly older patients (ie, a population segment at risk for myocardial infarction and stroke, for whom low-dose aspirin prophylaxis for primary and secondary prevention is being used with increasing frequency), physicians should recognize that the combination of aspirin and a nonselective NSAID is particularly damaging to the gastric mucosa. In a study in which the incidence rate ratio of GI bleeding with low-dose aspirin was 2.6, the rate was twice as high in patients who also were taking an NSAID.2 Use of low-dose aspirin is a major risk factor for NSAID-induced gastropathy and is an indication for gastroprotective therapy in patients with OA who take NSAIDs.

Risk with over-the-counter (OTC) NSAIDs
Upper GI ulceration and ulcer complications (eg, bleeding, perforation, and obstruction) are well recognized as serious and, at times, catastrophic adverse events associated with the use of prescription NSAIDs. That they also are associated with the use of OTC products is less well known.3

In one study, 421 patients at a large inner-city hospital were evaluated for upper GI hemorrhage. Peptic ulcer disease was concluded to be the cause in more than 50% of patients. Use of an OTC aspirin or nonaspirin NSAID during the week before admission was reported by 35% and 9% of patients, respectively, and use of a prescription aspirin or nonaspirin NSAID was reported in only 6% and 14%, respectively. Given the high frequency with which OTC agents were used (often for nonmedicinal purposes), short-term NSAID use may be a major cause of ulcer-related hemorrhage. In taking a patient’s history, inquiring about the use of not only prescribed medications but also OTC medications is important. Many patients do not regard OTC preparations as drugs.

Celecoxib and low-dose
aspirin: the CLASS study

Two major, large-size GI safety studies were published, the Celecoxib Long-term Arthritis Safety Study (CLASS)4 and the Vioxx Gastrointestinal Outcomes Research (VIGOR) trial.5 These studies, conducted by the manufacturers of celecoxib and rofecoxib, respectively, were designed to ascertain whether treatment with a coxib results in a lower incidence of clinically important NSAID-associated ulcers and ulcer complications than that seen with nonselective NSAIDs. Rofecoxib subsequently was withdrawn from the market because data indicated an increased risk of cardiovascular (CV) adverse events. Celecoxib currently is the only coxib on the market in the United States.

In CLASS, the difference between the annualized incidence rates of upper GI ulcer complications (the primary outcome measure) with celecoxib and with the comparator nonselective NSAID (diclofenac or ibuprofen) was not statistically significant. However, when ulcer complications and symptomatic ulcers were considered together, the difference reached significance and, among nonaspirin users, the results over the first 6 months showed a significant reduction in ulcer complications with celecoxib, mimicking the results obtained with rofecoxib in the VIGOR trial.

However, the use of low-dose aspirin mitigated the 6-month gastroprotective effect of celecoxib in CLASS. Among aspirin users, the incidence of upper GI complications was no lower in those receiving celecoxib than in those receiving diclofenac or ibuprofen, regardless of the presence of ulcer symptoms. It also should be noted that only 22% of subjects in CLASS were taking low-dose aspirin, because in clinical practice as many as 60% of patients older than 60 years may do so.

Also, even though the results were available for as long as 12 to 15 months, the major publication describing the results of CLASS4 presented actual data for only the first 6 months of treatment and annualized those results to provide 12-month estimates. However, examination of all the data indicated that nearly all the complications that occurred after the first 6 months were in patients taking celecoxib.6 As has been pointed out,6,7 even among nonaspirin users no difference between treatment groups existed with respect to the incidence of ulcer complications among patients who were treated for 12 to 15 months; ie, the difference between treatment groups that favored celecoxib after the first 6 months of treatment was not sustained.

Poor correlation between upper GI bleeding
and dyspepsia in NSAID users

Although dyspepsia is common with NSAID use, it correlates poorly with clinical episodes of GI bleeding and with the presence of endoscopically visualized lesions. Most patients who incur a serious GI complication with NSAID use have not had previous GI symptoms.8 Nonspecific GI adverse events (eg, dyspepsia, abdominal pain, bloating) are common in NSAID users, however, and even if they are not life-threatening they are important. That is because they affect adherence to prescribed NSAID dosing, result in use of additional therapies for management of the symptoms, and lead to expensive radiographic or endoscopic investigations to rule out the presence of ulcer or malignancy, increasing the cost of managing the disease and, at times, producing additional adverse effects. The incidence of nonspecific GI adverse effects with a coxib may not differ appreciably from that with a nonselective NSAID.9,10

In addition to age and dose, risk factors associated with NSAID-induced gastroenteropathy include CV disease, a history of ulcer, GI hemorrhage, use of corticosteroids, use of anticoagulants, poor general health (comorbidities), concomitant use of more than 1 NSAID (including aspirin), alcohol use, and cigarette smoking. In considering prescription of an NSAID, the physician should consider each of these factors. If an NSAID is recommended, it should be with caution in patients who have CV risk factors. In all patients, NSAIDs should be used in the lowest possible dose and for the shortest period necessary.

Adverse effects of NSAIDs are clearly dose-dependent, although this often is not the case with respect to the efficacy of NSAIDs within the therapeutic dose range.11 Differences between patients with respect to the degree of improvement in pain and function with NSAID treatment may relate more to differences in pharmacodynamics and pharmacokinetics than to dose.12

Long-term efficacy of NSAIDs in OA
Most clinical trials of NSAID efficacy in OA have been only 1 to 3 months in duration, although many patients with OA take these drugs for years. In a 2-year trial in which Dieppe and associates13 randomized 89 patients who had been taking NSAIDs long-term for knee OA to diclofenac, 100 mg/d, or to placebo and patients were permitted to take acetaminophen, up to 4 g/d, as rescue medication, only 57% completed the study; 27% of the placebo group and 7% of the diclofenac group withdrew because of lack of efficacy, primarily within the first 3 months.

Some 30% of the noncompleters were withdrawn because of adverse effects and 15% because of poor adherence, but the 2 treatment groups did not differ significantly in this respect. Among patients who were still in the trial at the end of 2 years, 52% of those randomized to diclofenac reported they were better; 16%, the same; and 32%, worse. The respective distributions among those in the placebo group were similar. The average daily amount of rescue acetaminophen consumed by patients who received placebo was 2 g/d, compared with 1.7 g/d by those in the diclofenac arm.

In another 2-year study, Williams and colleagues14 randomized 178 patients with knee OA who had not been receiving long-term NSAID therapy to naproxen, 750 mg/d, or acetaminophen, 2600 mg/d. The completion rate was only 35%-even lower than that in the Dieppe study-31% of those randomized to acetaminophen and 39% in the naproxen group. Withdrawals because of adverse effects were slightly more common with naproxen than with acetaminophen (23% vs 18%); withdrawals because of lack of efficacy and other reasons were less common in the naproxen group (16% vs 22%). For those patients who remained in the trial for 2 years, efficacy of both drugs was only modest and few differences between the treatment groups were apparent, although naproxen appeared to be slightly more effective.

These trials demonstrate that the long-term effectiveness of NSAIDs and acetaminophen in patients with OA is far from satisfactory. They also suggest that although NSAIDs are superior to placebo, for many patients they are not clearly superior to acetaminophen, a finding supported by some shorter-term studies.11 Indeed, data from the Dieppe study, for example, suggest that a substantial proportion of patients with OA who receive long-term NSAID therapy may do as well with withdrawal of their NSAID and use of acetaminophen only as needed. It is not surprising that only 15% to 20% of patients with OA for whom an NSAID was prescribed were still taking the same NSAID 12 months later.15

Are NSAIDs chondroprotective?

Claims that NSAIDs have structure-modifying activity in OA (ie, that they are disease-modifying OA drugs) are based almost exclusively on in vitro evidence that they may modify proteoglycan or collagen metabolism or inhibit cytokine-mediated matrix degeneration, the release or activity of cartilage matrix metalloproteinases, or the actions of toxic oxygen metabolites.16 The number of in vivo studies of NSAIDs in experimental models of OA is remarkably limited, largely because all the animal species that are frequently used as models (dog, mouse, rabbit, guinea pig) are markedly sensitive to the GI adverse effects of NSAIDs and GI hemorrhage or perforation develops before the joint pathology is well established.

Although the potential implications of the in vitro data are obvious, prediction of the in vivo effects of an NSAID based on its in vitro effects is naive. In vitro studies cannot predict the relative importance of the effect of the NSAID on synovial inflammation, its direct effect on chondrocyte metabolism, and its analgesic action (perhaps resulting in overload of the damaged joint).

In humans, in vivo evidence that NSAIDs exhibit structure-modifying OA drug (SMOAD) activity is no more adequate than in animal models. In the prospective double-blind study in which naproxen was compared with acetaminophen,14 the very high dropout rate precluded meaningful conclusions regarding the question of whether NSAIDs affect progression of OA in humans. In the 2-year placebo-controlled trial of diclofenac in patients with knee OA,13 no evidence of radiographic progression of OA was seen in the experimental or control group, although the critical importance of rigid standardization of radioanatomical positioning of the joint in a study that assesses the progression of structural damage was not appreciated at the time. However, more recent studies of putative SMOADs, using highly standardized radiographic protocols, have not shown unambiguous SMOAD activity in humans.

The long-standing inability to pharmacologically affect the underlying processes in OA in humans has been suggested to be the result of a chondrocentric view of OA and persistent overemphasis on articular cartilage as the major focus of OA damage and the progression of joint failure. Excessive emphasis also has been placed on the inflammatory changes in OA, which, although they may be relevant to the patient’s symptoms (and are therefore important), are secondary to breakdown products of the articular cartilage and subchondral bone. That the synovial joint is an organ and that OA is failure of that organ that can be initiated by abnormalities arising in any of its tissues is insufficiently appreciated. Failure to focus on the reduction of the contact area of the joint and on aberrant joint loading has, in my view, misdirected therapeutic efforts in OA.17

In Europe, some rheumatologists and orthopedic surgeons popularized the concept of an “indocin hip,” arguing that the NSAID indomethacin, a potent cyclooxygenase (COX)-2 inhibitor, may accelerate joint damage in patients with knee OA rather than exhibit a beneficial structure-modifying effect. Some surgeons have claimed that they can tell at the time that hip surgery is performed whether the patient was treated with indomethacin because the degree of degeneration is much greater in such patients than in those who have not been treated with indomethacin.

Rashad and colleagues18 reported a greater increase in the rate of joint-space narrowing-implying greater loss of articular cartilage-in radiographs of patients with hip OA who were treated with indomethacin than in those who received the weak COX inhibitor azapropazone. Patients in the indomethacin group came to arthroplasty sooner than those in the comparison group. However, that study had several limitations.

Similarly, Huskisson and associates19 also concluded that indomethacin accelerates joint breakdown in patients with knee OA. In a double-blind parallel study, nearly 400 patients with knee OA completed at least 1 year of treatment with indomethacin, 75 mg/d; tiaprofenic acid, 600 mg/d; or placebo. More than twice as many patients in the indomethacin group as in the placebo group showed joint-space narrowing. However, several concerns have been expressed about the study design.20 Although the conclusion that use of indomethacin in patients with OA may lead to acceleration of joint breakdown may be correct, the supporting evidence remains questionable.

Hurwitz and colleagues21 reported that in patients with medial tibiofemoral OA, the adductor moment at the knee (the force acting on the medial tibiofemoral compartment) was greater while they were receiving an NSAID than after withdrawal from the drug, when their knee pain was more severe-ie, pharmacological amelioration of joint pain resulted in an increase in mechanical loading of the already damaged compartment of the joint. However, long-term follow-up that might have permitted a conclusion about whether the increased mechanical loading was accompanied by an increase in structural damage (“analgesic arthropathy”) was not performed.

A phase 3 clinical trial of tanezumab (a monoclonal antibody that targets nerve growth factor) in patients with chronically painful hip OA was halted recently because a few reports indicated that patients receiving infusions of tanezumab experienced worsening disease and needed joint replacement.22 In clinical studies, tanezumab infusions had appeared remarkably effective in decreasing OA pain over a median treatment duration of 198 days.23 Noted as adverse effects were hypesthesia and paresthesia but not dysesthesia or allodynia. Worsening of arthritis has not been recorded in patients who were receiving tanezumab for other chronically painful conditions but did not have OA.

Withdrawal of NSAIDs from patients with OA
Both patients and physicians generally have a low level of satisfaction with NSAIDs. Patients with OA often are switched from one NSAID to another because of adverse effects or lack of effectiveness, but barring development of a serious adverse event, treatment with NSAIDs is not often discontinued. Because physicians tend to maintain NSAID therapy in virtual perpetuity for patients with OA, it is important to recognize that for many patients it is possible to discontinue NSAIDs, decrease the dose, or use NSAIDs only on an as-needed basis. Because the risk of NSAID-associated adverse effects (but not necessarily the effectiveness of NSAIDs) is dose-dependent, such measures are highly desirable.

The results of an NSAID-withdrawal study in 91 older patients who were admitted acutely to geriatric hospitals for indications unrelated to OA or its treatment and were taking a prescription NSAID at the time of admission are illuminating.24 (Although in younger persons NSAIDs may be prescribed for a multitude of reasons, such as migraine, dysmenorrhea, sprains, and strains, among older persons prescription of an NSAID is tantamount to a diagnosis of OA.) The study protocol involved withdrawal of the NSAID and replacement, as needed, with a simple analgesic, hot or cold packs, massage, or a muscle relaxant, after which the patient’s pain was monitored and, if the above measures were ineffective, treatment with the patient’s customary NSAID was reinstituted. More than 50% of the patients did not require reinstitution of NSAID treatment over the ensuing 6-month follow-up period. The point: Many patients with OA improve symptomatically-they get better-and decreasing the dose or discontinuing the NSAID is possible.

A controlled study of the effects of an education intervention provides additional evidence of the feasibility of NSAID withdrawal.25 In this study of older nursing home residents who were regular NSAID users, nursing home staff and the responsible physician were informed that such persons are at high risk for NSAID-associated adverse effects, and pharmacological and nonpharmacological alternatives that could permit discontinuation of the NSAID or reduction in dose were provided. In lieu of the patient's NSAID, acetaminophen was prescribed, in a dose of 650 mg 3 times a day and at bedtime as needed. To ensure that they did not experience an increase in pain as a result of the intervention, the nurse reevaluated the patients throughout the study. If that occurred, low-dose ibuprofen was added as needed. If that was unsuccessful, a higher dose of ibuprofen was prescribed on a regular basis. If the latter was not successful, treatment with the original NSAID was reinstituted.

The patients had a considerable level of chronic pain at baseline. The intervention resulted in reduction of NSAID use in the experimental group, from an average of 7 to about 2 days per week. For many patients, discontinuing the NSAID was possible. Little increase in acetaminophen use was noted in the control nursing homes, but acetaminophen use in the intervention nursing homes increased from 2 days to 5 days per week. Opioid analgesic use decreased significantly in the intervention homes but increased in the control nursing homes. In both treatment groups, pain scores showed little change from baseline-the intervention did not exacerbate pain or disability. In sum, in a controlled nursing home setting, with the cooperation of nursing home staff and physicians, the intervention resulted in a significant reduction in NSAID dose with no adverse effects on pain or function.

Notably, when a similar intervention was attempted by targeting primary care physicians in the community, rather than nursing home staff, only 10% of patients in the intervention group were given a trial of acetaminophen (vs 1% of patients in the practices of physicians in the control group).26 No difference was noted between the groups with respect to use of other pain medications or gastroprotective drugs or general health status.

Why the education intervention was so successful in reducing NSAID use among nursing home patients but did not alter prescribing behaviors of community physicians is unclear. I suggest the following: Older patients with OA pain who visit the office of a community physician often have a number of other problems related to serious comorbidities (eg, diabetes mellitus, heart disease, neurological problems) that must be addressed in the course of a relatively brief follow-up visit and are given precedence over management of the patient’s joint pain.

A rational approach to
managing OA pain with NSAIDs

In patients who are at low risk for a serious NSAID-related adverse event, ie, those who are not taking low-dose aspirin and have no history of ulcer disease or other risk factors for upper GI complications from NSAIDs, it is unknown whether celecoxib is preferable to a low dose of a nonselective NSAID. However, a cost-effectiveness analysis in Canada found that except in older patients, coxibs are not cost-effective in patients at average risk.28

In considering use of a systemic NSAID for patients with OA pain, note that other options are available, such as acetaminophen, topical NSAIDs, opioids, and nonpharmacological measures. In a patient who is at low risk for a GI complication from NSAIDs (not taking low-dose aspirin and with no history of ulcer disease or the risk factors mentioned above), prescribing a nonselective NSAID in the lowest effective dose and for the shortest time necessary is reasonable.

Coxibs and CV disease
In a patient at increased risk for GI problems, it is preferable to use a COX-2–selective agent (although it is no more effective in relieving OA pain than a nonselective NSAID) or a nonselective NSAID with coprescription of a proton pump inhibitor (PPI) or misoprostol. However, misoprostol is expensive and its effectiveness in this setting is incomplete. In addition, diarrhea is somewhat common with misoprostol and the drug often does not relieve NSAID-induced dyspepsia. Therefore, the daily quality of life of a patient taking misoprostol and an NSAID may be poorer than that of a patient taking an NSAID alone. Also, as shown in the CLASS study,4 if the patient is taking low-dose aspirin for CV prophylaxis, the gastroprotection associated with the use of celecoxib is lost.

In very-high-risk patients with a history of a GI bleed, the use of celecoxib was found to be associated with a 9% risk of recurrent GI bleeding, although this risk was shown to be abolished by administration of celecoxib and a PPI.29 This result is supported by those of a population-based retrospective cohort study in which a 30% reduction in hospital admissions for GI bleeding was noted among persons older than 75 years for whom a PPI was coprescribed, compared with treatment with celecoxib alone, without an additional gastroprotective agent (GPA).30

NSAID users’ adherence to
gastroprotective therapy

At the end of 2003, the FDA approved a kit containing tablets of naproxen, 500 mg, and delayed-release capsules of the PPI lansoprazole, 15 mg (Prevacid NapraPAC). In April 2010, the FDA approved a fixed-dose formulation of delayed-release enteric-coated naproxen (500 mg) and an immediate-release formulation of the PPI esomeprazole (20 mg).

In 2010, the combination product, Vimovo (formerly known as PN 400), was approved for management of the signs and symptoms of OA, rheumatoid arthritis, and ankylosing spondylitis in patients who are at risk for an NSAID-associated gastric ulcer. In contrast to Prevacid NapraPAC, Vimovo is formulated as a single pill. As with other NSAIDs, dose adjustment of the naproxen component of both products may be required in older patients, and there is no reason to expect that the PPI will confer any protection against other adverse events associated with NSAID use, such as renal insufficiency, edema, and lower GI tract complications.

In part, the motivation for developing agents such as Prevacid NapraPAC and Vimovo stems from evidence that adherence to prescribed GPAs by NSAID users is suboptimal. In a population-based, nested, case-control study that examined the association between adherence to PPIs or H2-receptor antagonists and the risk of NSAID-related upper GI ulcers or hemorrhage in high-risk patients, 15% of nonselective NSAID users received GPAs. For every 10% decrease in adherence to prescribed GPA dosing, the risk of an NSAID-related upper GI complication increased 16%. Compared with patients in whom more than 80% of NSAID treatment days were covered by a GPA, the risk of an NSAID-related upper GI complication increased 2.5-fold and 4.0-fold, respectively, in those with 20% to 80% of NSAID days covered and those with less than 20% of NSAID days covered.31


References1. Dieppe P, Brandt KD. What is important in treating osteoarthritis? Whom shall we treat and how should we treat them? Rheum Dis Clin North Am . 2003;29:687-716.

2. Sørensen HT, Mellemkjaer L, Blot WJ, et al. Risk of upper gastrointestinal bleeding associated with use of low-dose aspirin. Am J Gastroenterol. 2000;95:2218-2224.

3. Wilcox CM, Shalek KA, Cotsonis G. Striking prevalence of over-the counter nonsteroidal anti-inflammatory drug use in patients with upper gastrointestinal hemorrhage. Arch Intern Med. 1994;154:42-46.

4. Silverstein FE, Faich G, Goldstein JL, et al. Gastrointestinal toxicity with celecoxib vs nonsteroidal anti-inflammatory drugs for osteoarthritis and rheumatoid arthritis: the CLASS study: a randomized controlled trial. Celecoxib Long-term Arthritis Safety Study. JAMA. 2000;284:1247-1255.

5. Bombardier C, Laine L, Reicin A, et al; VIGOR Study Group. Comparison of upper gastrointestinal toxicity of rofecoxib and naproxen in patients with rheumatoid arthritis. VIGOR Study Group. N Engl J Med. 2000;343:1520-1528.

6. Jüni P, Rujes AW, Dieppe PA. Are selective COX-2 inhibitors superior to traditional non steroidal anti-inflammatory drugs? [published correction appears in BMJ. 2002;324:1538]. BMJ. 2002;324:1287-1288.

7. Metcalfe S, Dougherty S, McNee W. Systematic review of celecoxib for osteoarthritis and rheumatoid arthritis: celecoxib’s relative gastrointestinal safety is overstated. BMJ. 2003;326:334.

8. Singh G, Ramey DR, Morfeld D, et al. Gastrointestinal tract complications of nonsteroidal anti-inflammatory drug treatment in rheumatoid arthritis: a prospective observational cohort study. Arch Intern Med. 1996;156:1530-1536.

9. Langman MJ, Jensen DM, Watson DJ, et al. Adverse upper gastrointestinal effects of rofecoxib compared with NSAIDs. JAMA. 1999;282:1929-1933.

10. Feldman M, McMahon AT. Do cyclooxygenase-2 inhibitors provide benefits similar to those of traditional nonsteroidal anti-inflammatory drugs, with less gastrointestinal toxicity? [published correction appears in Ann Intern Med. 2000;132:1011]. Ann Intern Med. 2000;132:134-143.

11. Bradley JD, Brandt KD, Katz BP, et al. Comparison of an anti-inflammatory 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.

12. Bradley JD, Rudy AC, Katz BP, et al. Correlation of serum concentrations of ibuprofen stereoisomers with clinical response in the treatment of hip and knee osteoarthritis. J Rheumatol. 1992;19:130-134.

13. Dieppe P, Cushnaghan J, Jasani MK, et al. A two-year, placebo-controlled trial of non-steroidal anti-inflammatory therapy in osteoarthritis of the knee joint. Br J Rheumatol. 1993;32:595-600.

14. 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.

15. Scholes D, Stergachis A, Penna PM, et al. Nonsteroidal antiinflammatory drug discontinuation in patients with osteoarthritis. J Rheumatol. 1995;22:708-712.

16. Doherty M. ‘Chondroprotection’ by non-steroidal anti-inflammatory drugs. Ann Rheum Dis. 1989;48:619-621.

17. Brandt KD, Dieppe P, Radin EL. Commentary: is it useful to subset “primary” osteoarthritis? A critique based on evidence regarding the etiopathogenesis of osteoarthritis. Semin Arthritis Rheum. 2009;39:81-95.

18. Rashad S, Revell P, Hemingway A, et al. Effect of non-steroidal anti-inflammatory drugs on the course of osteoarthritis. Lancet. 1989;2:519-522.

19. Huskisson EC, Berry H, Gishen P, et al. Effects of antiinflammatory drugs on the progression of osteoarthritis of the knee. LINK Study Group. Longitudinal Investigation of Nonsteroidal Antiinflammatory Drugs in Knee Osteoarthritis. J Rheumatol. 1995;22:1941-1946.

20. Doherty M, Jones A. Indomethacin hastens large joint osteoarthritis in humans-how strong is the evidence? J Rheumatol. 1995;22:2013-2016.

21. Hurwitz DE, Sharma L, Andriacchi TP. Effect of knee pain on joint loading in osteoarthritis. Curr Opin Rheumatol. 1999;11:422-426.

22. Trials halted as Pfizer's tanezumab shown to worsen osteoarthritis. Genetic Engineering & Biotechnology News. GEN news highlights: Jun 24 2010. Accessed January 10, 2011.

23. Lane NE, Schnitzer TJ, Birbara CA, et al. Long-term tanezumab use for treatment of moderate to severe osteoarthritic knee pain: an open-label extension study [abstract]. Arthritis Rheum. 2009;60(suppl 10):S312.

24. Black D, Tuppen J, Heller A. NSAID withdrawal in elderly patients. J Am Geriatr Soc. 1991;39:A26.

25. Stein CM, Griffin MR, Taylor JA, et al. Educational program for nursing home physicians and staff to reduce use of non-steroidal anti-inflammatory drugs among nursing home residents: a randomized controlled trial. Med Care. 2001;39:436-445.

26. Ray WA, Stein CM, Byrd V, et al. Educational program for physicians to reduce use of non-steroidal anti-inflammatory drugs among community-dwelling elderly persons: a randomized controlled trial. Med Care. 2001;39:425-435.

27. Tai-Seale M, McGuire TG, Zhang W. Time allocation in primary care office visits. Health Serv Res. 2007;42:1871-1894.

28. Maetzel A, Krahn M, Naglie G. The cost-effectiveness of celecoxib and rofecoxib in patients with osteoarthritis or rheumatoid arthritis. Ottawa: Canadian Coordinating Office for Health Technology Assessment (CCOHTA); 2001. Technology report No. 23.

29. Chan FK, Wong VW, Suen BY, et al. Combination of a cyclo-oxygenase-2 inhibitor and a proton-pump inhibitor for prevention of recurrent ulcer bleeding in patients at very high risk: a double-blind, randomised trial. Lancet. 2007;369:1621-1626.

30. Rahme E, Barkun AN, Toubouti Y, et al. Do proton-pump inhibitors confer additional gastrointestinal protection in patients given celecoxib? Arthritis Rheum. 2007;57:748-755.

31. van Soest EM, Sturkenboom MC, Dieleman JP, et al. Adherence to gastroprotection and the risk of NSAID-related upper gastrointestinal ulcers and haemorrhage. Aliment Pharmacol Ther. 2007;26:265-275.