• Osteoarthritis;
  • Nonsteroidal antiinflammatory drugs;
  • Cyclooxygenase 2 inhibitors


  1. Top of page
  2. Abstract
  7. Acknowledgements


To examine whether the current widespread use of antiinflammatory drugs may reflect a lack of informed choice (i.e., unawareness of adverse effects or potential treatment alternatives) among older patients with knee osteoarthritis (OA).


Consecutive patients with symptomatic knee OA (n = 100) completed a questionnaire to assess their awareness of drug toxicity. Patients also completed an Adaptive Conjoint Analysis task so that the influence of providing an additional treatment alternative on patient preferences for nonselective nonsteroidal antiinflammatory drugs (NSAIDs) and cyclooxygenase-2 (COX-2) inhibitors could be assessed.


Fifty-four percent of the patients surveyed were unaware of any adverse effects related to NSAIDs and 80% were unaware of any toxicity related to COX-2 inhibitors. When given a choice between NSAIDs and COX-2 inhibitors, 57% of patients preferred COX-2 inhibitors over NSAIDs. When presented with a third less effective, but safer alternative, 100% of patients switched preferences to the safer, albeit less effective, option.


Our findings suggest that the widespread use of NSAIDs may reflect lack of informed choice among older patients with OA.


  1. Top of page
  2. Abstract
  7. Acknowledgements

Nonsteroidal antiinflammatory drugs (NSAIDs) are the most widely used medications for patients with arthritis and other painful conditions, with 70 million to 100 million prescriptions written in the US per year. Cyclooxygenase-2 (COX-2) inhibitors continue to expand the market of antiinflammatory drugs, with total revenues expected to exceed $2.5 billion by 2008 (1). The widespread use of NSAIDs contrasts with clinical studies that have demonstrated that, when fully informed of expected risk and benefits, older arthritis patients are reluctant to accept the risks of drug-related adverse effects (AEs) for modest drug benefits (2–4).

One possible explanation for the discrepancy between patient risk aversion and the widespread use of antiinflammatory drugs is that patients may be unaware of the AEs related to this class of medications. This possibility is supported by previous studies demonstrating that many patients have limited knowledge about their medications (5–7) and that physicians disclose limited risk-related information when prescribing NSAIDs (8) as well as other medications (5).

Another possible explanation is that patients do not explicitly consider the tradeoffs related to available treatment options. This possibility is supported by recent studies that have demonstrated that, despite efforts to promote shared decision making, patients tend to comply with physicians' recommendations rather than participate in interactive discussions about available options (9–11).

The objective of this study was to examine whether the widespread use of NSAIDs may reflect a lack of informed choice (i.e., unawareness of AEs or potential treatment alternatives) among patients with arthritis. To test this hypothesis, we 1) ascertained patient knowledge of arthritis medications and their associated toxicity, and 2) assessed whether asking patients to consider additional options influences their preferences for NSAIDs and COX-2 inhibitors.


  1. Top of page
  2. Abstract
  7. Acknowledgements


Consecutive patients with radiographic knee osteoarthritis (OA), treated in community rheumatology practices affiliated with a university hospital, were contacted by telephone ∼1 week after having received a letter describing the study. Patients having pain in 1 or both knees on most days of the month and not having rheumatoid arthritis, gout, pseudogout, or bilateral knee replacements were asked to complete a questionnaire examining opinions about arthritis treatments. The study was approved by the Human Investigations Committee at our institution.

Data collection.

All data were collected in face-to-face interviews by a research assistant. Patient characteristics were based on self report and the Western Ontario and McMaster Universities Osteoarthritis Index (12). Patient knowledge regarding available treatment options for OA was ascertained using open-ended questions based on the Basic Arthritis Test (13): “What medications are available to patients with your type of arthritis?” Patients were classified as being aware of NSAIDs as a possible treatment option if they listed one or more generic or brand name NSAIDs. Patients were classified as being aware of COX-2 inhibitors if they listed the generic or brand name of either of the 2 coxibs available at the time of the study (rofecoxib and celecoxib). For each of the medications listed, patient awareness of AEs was ascertained using the following question (13): “Name the possible side effects of this medication.” Patients answering “none” or “no side effects” were classified as being unaware of AEs related to the medication(s) listed.

We assessed treatment preferences using Adaptive Conjoint Analysis (ACA; version 4.0, Sawtooth Software, Sequim, WA), a well-validated interactive computer program that computes individual patient values (utilities) for specific medication characteristics based on a series of ranking and rating questions. ACA derives individual patient preferences based on the utilities derived from the conjoint questionnaire using least squares regression analyses (14). Because preferences are derived based on tradeoffs between treatment characteristics and not full descriptions of the medications, they are not influenced by product recognition or brand preferences.

We ascertained patient values (i.e., utilities) for the following medication characteristics: time to benefit, response rate, common AEs, risk of ulcer, and monthly copayments (reflective of 2001 pharmaceutical costs in Connecticut). All characteristics were described using lay terminology adapted from patient information sheets published by the Arthritis Foundation. The face and content validity of the characteristic descriptions were assessed by rheumatologists, conjoint experts, and lay persons.

In view of the literature documenting significant variability in both patients' ability to interpret probabilities and preference for the presentation of probabilistic information (15), we used qualitative and quantitative frequency formats to describe the likelihood of AEs. The ranges of probabilities of benefits and AEs were based on randomized controlled data and observational studies (16–24). The characteristics assigned to the treatment options studied are described in Table 1.

Table 1. Description of treatment options studied*
CharacteristicNSAIDsCOX-2 inhibitorsAlternative 1Alternative 2Alternative 3Alternative 4
  • *

    NSAIDs = nonsteroidal antiinflammatory drugs; COX-2 = cyclooxygenase 2; po = by mouth; bid = twice a day; qd = every day; qid = 4 times a day.

Route of administrationpo bidpo qdpo qdpo qidpo qidpo qid
Onset of action1–2 hours1–2 hours1–2 hoursWithin 4 weeksWithin 4 weeksWithin 4 weeks
Percent of patients who benefit50%50%25%25%25%25%
Common adverse effectsNausea, diarrhea, heartburnNausea, diarrhea, heartburnWell toleratedWell toleratedWell toleratedWell tolerated
Annual risk of ulcer2%1%No added risk of ulcerNo added risk of ulcerNo added risk of ulcerNo added risk of ulcer
Out-of-pocket cost per month$17.50$70.00$17.50$17.50$25$50

ACA questionnaire.

Respondents were first asked to rate the importance of the difference between the best and worst estimate of each characteristic on a 4-point scale. The characteristics were presented in random order to eliminate any possible ordering effects (Figure 1). Second, to refine respondents' utilities, respondents evaluated a series of paired comparisons (Figure 1).

thumbnail image

Figure 1. Two examples of Adaptive Conjoint Analysis questions.

Download figure to PowerPoint

A 9-point scale was used for paired comparisons so that finer differences in strength of preference could be derived. The program uses the information obtained from each paired comparison to update the estimates of each respondent's utilities and to select the next pair of options. Final utilities are then generated using regression analysis (14). Additional details regarding this methodology have been published previously (14, 25–27).


Preference data derived from ACA were imported into SAS computer files (SAS software, version 6.12; SAS Institute, Cary, NC) and merged with the patient characteristics data set. We used ACA to determine the percentage of patients preferring specific treatment options. ACA is capable of performing choice simulations that demonstrate the percentage of patients preferring specific treatment options defined by the researcher (14). Individual patient preferences are derived based on each patient's responses to the rating and paired-comparison tasks (27).

We first examined treatment preferences for medications describing NSAIDs and COX-2 inhibitors using ACA's simulation capabilities. We subsequently examined the influence of adding a third, safer but less effective, treatment option on patient preferences.


  1. Top of page
  2. Abstract
  7. Acknowledgements

Older adults with symptomatic knee OA were interviewed (n = 100, participation rate 84%). The mean ± SD age of our sample was 70 ± 7 years (range 59–90 years); 79% were female, 92% were white, 69% had some college education, and 32% reported having an annual household income ≥$60,000.

The mean ± SD duration of knee pain was 11 ± 9 years; mean ± SD knee pain score was 12 ± 4 (on a scale of 0–20), and mean ± SD physical function score was 42 ± 14 (on a scale of 0–68). Thirty-five percent reported previously having dyspepsia due to NSAIDs, 22% had an ulcer, and 5% had been hospitalized for a gastrointestinal bleed. The majority of patients reported current or previous use of NSAIDs (85%) and COX-2 inhibitors (76%) (Table 2).

Table 2. Utilization of NSAIDs and COX-2 inhibitors among the 100 patients surveyed*
 NSAIDsCOX-2 inhibitors
  • *

    NSAIDs = nonsteroidal antiinflammatory drugs; COX-2 = cyclooxygenase 2.

Current use, no.3433
Past use, no.5143
Never used, no.1218
Don't know, no.36

Eighty percent of patients listed NSAIDs and 81% listed COX-2 inhibitors as possible therapeutic options (Table 3). Of these, 54% were unaware of any AEs related to NSAIDs and 80% were unaware of any toxicity related to COX-2 inhibitors. Patients reporting previous gastrointestinal AEs due to antiinflammatory medications and those with longer duration of knee OA symptoms were more likely to be aware of at least 1 AE associated with NSAIDs compared with those without prior AEs and shorter duration of symptoms (Table 4). In a logistic regression model containing all 5 variables, previous gastrointestinal AEs remained a clinically significant predictor of awareness of AEs related to NSAIDs. The adjusted odds ratio [OR] for gastrointestinal AEs was 4.5, 95% confidence interval [95% CI] 1.7–12.1; and for duration of knee OA 1.1, 95% CI 1.0–1.1. The number of patients aware of AEs related to COX-2 inhibitors was very small, making it difficult to examine associations between patient characteristics and awareness of drug toxicity (Table 4).

Table 3. Percentage of patients unaware of the specific adverse effects associated with antiinflammatory drugs*
Adverse effectNSAIDs (n = 80) %COX-2 inhibitors (n = 81) %
  • *

    NSAIDs = nonsteroidal antiinflammatory drugs; COX-2 = cyclooxygenase 2; GI = gastrointestinal.

Stomach upset/heartburn5480
Ulcer/GI bleed9596
Renal toxicity9699
Table 4. Association between patient characteristics and knowledge of adverse effects*
CharacteristicsNSAIDsCOX-2 inhibitors
Aware of AEs (n = 37)Unaware of AEs (n = 43)PAware of AEs (n = 16)Unaware of AEs (n = 65)P
  • *

    NSAIDs = nonsteroidal antiinflammatory drugs; COX-2 = cyclooxygenase:2; AEs = adverse effects; OA = osteoarthritis.

Age, mean ± SD years69.1 ± 6.070.3 ± 7.20.468.2 ± 6.270.0 ± 6.80.3
College education, %51420.431490.2
Annual household income ≥$60,000, %35330.831320.9
Previous gastrointestinal AEs, %65330.00450501.0
Duration of OA, mean ± SD years13.4 ± 10.89.3 ± ± 6.411.0 ± 9.80.3

Fifty-seven percent of patients preferred COX-2 inhibitors over NSAIDs (Table 5). Adding a third option that is 50% less effective but does not cause dyspepsia or serious gastrointestinal events resulted in a dramatic change in preferences, with 100% of patients preferring the safer, but less effective option (Table 5). If the third alternative is described as having a delayed onset of action (4 weeks) and must be taken 4 times daily, the number preferring this option drops to 69%. Fifty percent of the patients surveyed continued to prefer the less effective, less convenient, but safer option, even when it costs more than double the cost of the option describing traditional NSAIDs.

Table 5. Treatment preferences for NSAIDs, COX-2 inhibitors, and a third treatment option*
Description of the additional treatment optionPreferring NSAIDs mean ± SD, %Preferring COX-2 inhibitors mean ± SD, %Preferring third option mean ± SD, %
  • *

    NSAIDs = nonsteroidal antiinflammatory drugs; COX-2 cyclooxygenase 2; AEs = adverse effects; qid = 4 times a day.

43.0 ± 5.057.0 ± 5.0
Alternative 1: No AEs, 50% less effective, $17.50/month.00100
Alternative 2: No AEs, 50% less effective, delayed onset of action, taken qid, $17.50 per month.11.0 ± 3.120.0 ± 4.069.0 ± 4.6
Alternative 3: No AEs, 50% less effective, delayed onset of action, taken qid, $25.00 per month.14.0 ± 3.522.0 ± 4.164.0 ± 4.8
Alternative 4: No AEs, 50% less effective, delayed onset of action, taken qid, $50.00 per month.23.0 ± 4.227.0 ± 4.450.0 ± 5.0

We found no relationship between the percentage of patients preferring the additional option and age, income, education, pain, functional status, or previous gastrointestinal side effects (data not shown).


  1. Top of page
  2. Abstract
  7. Acknowledgements

Many older patients with knee OA in this well-educated cohort were unaware of the toxicity associated with NSAIDs, and almost all were unaware of the toxicity associated with COX-2 inhibitors. Awareness of NSAID-related toxicity was associated with duration of knee OA and previous experience of gastrointestinal AEs. These results are in agreement with a survey in which 55% of Americans taking NSAIDs were unaware of NSAID-related gastrointestinal complications (7), and are supported by the study by Katz et al (8) that demonstrated that rheumatologists disclose relatively minimal information when prescribing NSAIDs. To the best of our knowledge, patient awareness regarding the toxicity associated with COX-2 inhibitors has not been previously studied. Taken together, these findings suggest that, while most arthritis patients are currently or have previously used NSAIDs or COX-2 inhibitors, they may not be fully informed of the potential risks associated with these medications.

We also found that, when given a choice, older adults almost always opt for safer (albeit less effective) medications over treatment options with known (even if small) risks of toxicity. We would expect that even fewer patients would choose antiinflammatory drugs had we included other drug-related risks, such as renal toxicity. These results suggest that many older patients might want to try relatively safer treatment options, such as capsaicin, before agreeing to accept the gastrointestinal toxicity associated with antiinflammatory drugs.

Our results must be interpreted in view of the limitations of the study. ACA derives individual patient preferences based on how each patient values tradeoffs between specific medication characteristics. Therefore, while providing preference data that is unbiased by brand name recognition and providers' preferences, this method is limited by the spectrum of characteristics included in the ACA survey. In addition, we measured preferences at only one point in time; one would expect that patient preferences would change depending on response to treatment and development of AEs.

The relatively uniform and small population studied limits the generalizability of our results. Although NSAID use and AEs experienced in this cohort are within the ranges of those reported in the general arthritis population (7), it is possible that younger patients or those with more severe disease would be more willing to accept the risks associated with NSAIDs and COX-2 inhibitors. Last, the number of patients preferring antiinflammatory drugs might have decreased simply as a result of adding a third alternative. However, the dramatic switch in preferences observed would be unlikely to occur solely on the basis of expanding the choice set.

Our findings have important ethical and clinical implications. First, this study adds to the literature suggesting that older patients may not be adequately informed of the risks associated with antiinflammatory medications. This finding is worrisome because older patients are among those at the highest risk for serious gastrointestinal complications, and are the most frequent chronic users of antiinflammatory medications. Second, we found that the widespread use of these drugs (as described in this study) is not consistent with patient preferences, and that when given a choice, many patients opt for safer, albeit less effective, alternatives. This finding is particularly important because NSAIDs do not affect the natural history of the disease.

In summary, this study suggests that the widespread use of antiinflammatory drugs may, in part, reflect lack of informed choice among older patients with OA. Health care providers should encourage patient participation in decision-making to ensure informed choice among older adults with arthritis.


  1. Top of page
  2. Abstract
  7. Acknowledgements

We thank Drs. R. T. Schoen, J. Kenney, C. DiSabatino, A. Liebling, J. Evans, D. Trock, M. Schwartz, R. Maclean, and C. Arnold for their help in recruiting patients. Most importantly, we thank all the patients who participated in this study for their time and effort.


  1. Top of page
  2. Abstract
  7. Acknowledgements
  • 1
    IMS Health Web site. Accessed January 16, 2004. URL:
  • 2
    Fraenkel L, Bogardus S, Concato J, Felson D. Does decreasing the likelihood of toxicity make patients more likely to use medications? J Rheumatol 2003; 30: 4438.
  • 3
    O'Brien BJ, Elswood J, Calin A. Willingness to accept risk in the treatment of rheumatic disease. J Epidemiol Community Health 1990; 44: 24952.
  • 4
    Pullar T, Wright V, Feely M. What do patients and rheumatologists regard as an acceptable risk in the treatment of rheumatic disease? Br J Rheumatol 1990; 29: 2158.
  • 5
    Ascione FJ, Kirscht JP, Shimp LA. An assessment of different components of patient medication knowledge. Med Care 1986; 24: 101828.
  • 6
    Ribeiro V, Blakeley J, Laryea M. Women's knowledge and practices regarding the prevention and treatment of osteoporosis. Health Care Women Int 2000; 21: 34753.
  • 7
    Singh G, Triadafilopoulos G. Epidemiology of NSAID induced gastrointestinal complications. J Rheumatol 1999; Suppl 26: 1824.
  • 8
    Katz JN, Daltroy LH, Brennan TA, Liang MH. Informed consent and the prescription of nonsteroidal antiinflammatory drugs. Arthritis Rheum 1992; 35: 125763.
  • 9
    O'Cathain A, Walters SJ, Nicholl JP, Thomas KJ, MKirkham M. Use of evidence based leaflets to promote informed choice in maternity care: randomised controlled trial in everyday practice. BMJ 2002; 324: 6436.
  • 10
    Oliver S, Rajan L, Turner H, Oakley A, Entwistle V, Watt I, et al. Informed choice for users of health services: views on ultrasonography leaflets of women in early pregnancy, midwives, and ultrasonographers. BMJ 1996; 313: 12513.
  • 11
    Stapleton H, Kirkham M, Thomas G. Qualitative study of evidence based leaflets in maternity care. BMJ 2002; 324: 63945.
  • 12
    Bellamy N, Buchanan WW, Goldsmith CH, Campbell J, Stitt LW. Validation study of WOMAC: a health status instrument for measuring clinically important patient relevant outcomes to anti-rheumatic drug therapy in patients with osteoarthritis of the hip or knee. J Rheumatol 1988; 15: 183340.
  • 13
    Branch VK, Lipsky K, Nieman T, Lipsky PE. Positive impact of an intervention by arthritis patient educators on knowledge and satisfaction of patients in a rheumatology practice. Arthritis Care Res 1999; 12: 3705.
  • 14
    Johnson RM. Adaptive conjoint analysis: Sawtooth Software conference proceedings. Sequim (WA): Sawtooth Software; 1987. p. 25365.
  • 15
    Mazur DJ, Hickam DH. Patients' interpretations of probability terms. J Gen Intern Med 1991; 6: 23740.
  • 16
    Altman RD, Moskowitz R. A randomized clinical trial of intra-articular sodium hyaluronate in patients with osteoarthritis of the knee: a summary. Am J Orthop 1999; 28(11 Suppl ): 34.
  • 17
    Cannon GW, Caldwell JR, Holt P, McLean B, Seidenberg B, Bolognese J, et al. Rofecoxib, a specific inhibitor of cyclooxygenase 2, with clinical efficacy comparable with that of diclofenac sodium: results of a one-year, randomized, clinical trial in patients with osteoarthritis of the knee and hip. Arthritis Rheum 2000; 43: 97887.
  • 18
    Deal CL, Schnitzer TJ, Lipstein E, Seibold JR, Stevens RM, Levy MD, et al. Treatment of arthritis with topical capsaicin: a double-blind trial. Clin Ther 1991; 13: 38395.
  • 19
    Ehrich EW, Schnitzer TJ, McIlwain H, Levy R, Wolfe F, Weisman M, et al, Rofecoxib Osteoarthritis Pilot Study Group. Effect of specific COX-2 inhibition in osteoarthritis of the knee: a 6 week double blind, placebo controlled pilot study of rofecoxib. J Rheumatol 1999; 26: 243847.
  • 20
    Houpt JB, McMillan R, Wein C, Paget-Dellio SD. Effect of glucosamine hydrochloride in the treatment of pain of osteoarthritis of the knee. J Rheumatol 1999; 26: 242330.
  • 21
    Langman MJ, Jensen DM, Watson DJ, Harper SE, Zhao PL, Quan H, et al. Adverse upper gastrointestinal effects of rofecoxib compared with NSAIDs. JAMA 1999; 282: 192933.
  • 22
    Peloso PM, Bellamy N, Bensen W, Thomson GT, Harsanyi Z, Babul N, et al. Double blind randomized placebo control trial of controlled release codeine in the treatment of osteoarthritis of the hip or knee. J Rheumatol 2000; 27: 76471.
  • 23
    Roth SH, Fleischmann RM, Burch FX, Dietz F, Bockow B, Rapoport RJ, et al. Around-the-clock, controlled-release oxycodone therapy for osteoarthritis-related pain: placebo-controlled trial and long-term evaluation. Arch Intern Med 2000; 160: 85360.
  • 24
    Towheed TE, Hochberg MC. A systematic review of randomized controlled trials of pharmacological therapy in osteoarthritis of the knee, with emphasis on trial methodology. Semin Arthritis Rheum 1997; 26: 75570.
  • 25
    Fraenkel L, Bogardus ST, Wittink DR. Understanding patient preferences for the treatment of lupus nephritis using adaptive conjoint analysis. Med Care 2001; 39: 120316.
  • 26
    Green PE, Srinivasan V. Conjoint analysis in marketing: new developments with implications for research and practice. J Marketing 1990; 54: 317.
  • 27
    Wittink DR, Bergenstuen T. Forecasting with conjoint analysis. In: Armstrong JS, editor. Principles of forecasting: a handbook for researchers and practitioners. Norwell (MA): Kluwer Academic Publishers; 2001.