Dyspepsia tolerability from the patients perspective: a comparison of celecoxib with diclofenac

Aliment Pharmacol Ther 2002; 16: 819 827. Dyspepsia tolerability from the patients perspective: a comparison of celecoxib with diclofenac J. L. GOLDSTEIN*, G. M. EISEN, T. A. BURKEà, B. M. PEÑA, J. LEFKOWITH & G. S. GEIS *Department of Medicine, Section of Digestive and Liver Diseases, College of Medicine, University of Illinois at Chicago, Chicago, IL, USA; Department of Gastroenterology, Vanderbilt University Medical Center, Nashville, TN, USA; àglobal Health Outcomes, Pharmacia Corp, Peapack, NJ, USA; Outcomes Research, Pfizer Inc., New York, NY, USA; Research & Development, Pharmacia Corp, Skokie, IL, USA Accepted for publication 28 November 2001 SUMMARY Aim: To compare celecoxib (800 mg/day, n ¼ 1997) with diclofenac (150 mg/day, n ¼ 1996) on dyspepsiarelated tolerability. Methods: In one of the two protocols comprising the Celecoxib Long-Term Arthritis Safety Study, a randomized double-blind trial, patients completed the Severity of Dyspepsia Assessment Questionnaire at baseline and at weeks 4, 13, 26 and 52 for the following three scales: Pain Intensity, Non-Pain Symptoms and Satisfaction with Dyspepsia-Related Health. Results: For the Pain Intensity scale, patients given diclofenac had significantly higher (worsening dyspepsia) mean changes, defined as follow-up minus baseline, than patients given celecoxib (P < 0.001, at all assessments). The mean changes in the Pain Intensity scale (scale, 2 47; higher score is higher pain intensity) were 0.99 (95% confidence interval (CI): 0.50, 1.48) for celecoxib and 2.76 (95% CI: 2.28, 3.25) for diclofenac at 4 weeks. Satisfaction was superior with celecoxib at all assessments (P < 0.001). At 4 weeks, the mean changes in the Satisfaction scale (scale, 7 35; higher score is higher satisfaction) were 0.02 (95% CI: ) 0.26, 0.29) for celecoxib and ) 0.72 (95% CI: ) 1.00, ) 0.45) for diclofenac. Diclofenac patients had significantly higher Non-Pain Symptoms at 4 weeks (P ¼ 0.005). Conclusions: Celecoxib, at two to four times the recommended dose, demonstrated a superior dyspepsia-related tolerability and satisfaction compared with standard dosages of diclofenac. INTRODUCTION Upper gastrointestinal symptoms are a common sideeffect of non-steroidal anti-inflammatory drug (NSAID) therapy. 1 3 Between 15% and 34% of NSAID users report dyspepsia or abdominal pain adverse events, and approximately 10% of patients withdraw due to these adverse events in clinical trials. 4 9 The association between upper gastrointestinal symptoms and NSAID use has also been evaluated in observational studies. Correspondence to: Dr J. L. Goldstein, Department of Medicine (M/C 787), 840 S. Wood St, Room 718E, Chicago, IL 60612, USA. E-mail: Jlgoldst@uic.edu Singh et al. reported that 12.5% of NSAID users stopped or switched their NSAID due to dyspepsia and 34% used gastrointestinal medications over a 2.5-year follow-up period. 10 Furthermore, gastrointestinal symptoms, particularly dyspepsia and upper abdominal/epigastric pain, are associated with a reduced health-related quality of life in arthritis patients. 11 The tolerability of NSAIDs is particularly important in chronic conditions, such as rheumatoid arthritis and osteoarthritis, which often require long-term NSAID therapy. Recently, a new class of medications, the cyclooxygenase-2 inhibitors, has become available for use in arthritis. The cyclo-oxygenase-2 specific inhibitors have been shown to provide the full therapeutic efficacy Ó 2002 Blackwell Science Ltd 819

820 J. L. GOLDSTEIN et al. of traditional NSAIDs, but with a significantly reduced risk of endoscopic ulcers and symptomatic gastroduodenal ulcers/ulcer complications. In two large-scale gastrointestinal outcomes trials of somewhat similar design, the risks of symptomatic gastroduodenal ulcers/ ulcer complications were significantly reduced for both celecoxib, 800 mg/day (relative risk ¼ 0.59; 95% confidence interval (CI), 0.38 0.94), and rofecoxib, 50 mg/ day (relative risk ¼ 0.5; 95% CI, 0.3 0.6). 7, 8 In one of the two protocols comprising the Celecoxib Long-Term Arthritis Safety Study (CLASS) trial, upper gastrointestinal tolerability was evaluated using the patientreported outcome measure, the Severity of Dyspepsia Assessment (SODA). The purpose of this paper is to compare the dyspepsia experience of celecoxib-treated patients with that of patients treated with diclofenac using SODA to assess patients experience of dyspepsia. MATERIALS AND METHODS Data source CLASS was a prospective, randomized, double-blind trial comparing celecoxib vs. diclofenac and ibuprofen in patients with osteoarthritis and rheumatoid arthritis. CLASS consisted of two companion protocols. In the first protocol, patients were randomized to receive ibuprofen (2400 mg/day) or celecoxib (800 mg/day). In the second protocol, patients were randomized to receive diclofenac (150 mg/day) or celecoxib (800 mg/day). Celecoxib, 800 mg/day, is twice the therapeutic dose for rheumatoid arthritis and four times that for osteoarthritis. The study end-points of CLASS were upper gastrointestinal ulcer complications (perforation, obstruction, bleeding) and symptomatic ulcers. The safety results of CLASS have been published elsewhere. 8 In the diclofenac protocol, dyspepsia-related health was assessed by SODA questionnaire. These data provide the basis for this report. Patients Study patients were men and women, aged 18 years or over, diagnosed with osteoarthritis or rheumatoid arthritis, evident for 3 months or longer. Patients were expected to require continuous NSAID treatment for the duration of the trial. The minimum expected participation for an individual patient was 6 months, with a maximum study participation of up to 12 months. Assessments At the screening visit, all patients were questioned about their medical histories and specifically asked to note any history of upper gastrointestinal bleeding, gastroduodenal ulcer or gastrointestinal-related NSAID intolerance. A history of gastrointestinal-related NSAID intolerance was defined as any prior gastroduodenal ulcer, erosive gastritis or NSAID-related upper gastrointestinal symptoms of sufficient severity to discontinue NSAIDs. Patients visited the study sites for safety and efficacy assessments at baseline, at weeks 4, 13, 26, 39 and 52, and at early termination. Safety assessments consisted of laboratory testing (haematology and biochemistry) and assessments of any signs and symptoms experienced by the patient. At the study visits, the investigators or their assigned study personnel were instructed to query patients for signs and symptoms by asking: Since your last visit, have you experienced or do you currently have any symptoms that are not associated with your arthritis?. Signs or symptoms associated with arthritis were specifically excluded because they were captured within the arthritis efficacy assessments. The investigators summarized signs and symptoms in the case report form. The investigators descriptions of the signs and symptoms were mapped to the World Health Organization Adverse Reaction Terminology (WHOART) list (information can be obtained at: http://www.who-umc.org/ umc.html). SODA questionnaire SODA was originally developed as a multidimensional measure of dyspepsia-related health to serve as the primary outcome measure in trials involving patients with uninvestigated dyspepsia. SODA consists of three scales, each of which measures a different dimension of dyspepsia-related health that has been identified through previous empirical research. 12, 13 In a previous psychometric report based on this trial by Rabeneck et al., the SODA scales exceeded the recommended reliability criterion for group-level comparison (alpha > 0.70), demonstrated fair to good levels of reproducibility (intraclass correlation coefficient between 0.40 and 0.75), and were sensitive to changes in dyspepsia-related health. 14 Rabeneck et al. used dyspepsia withdrawal status (yes or no) and dyspepsia severity (none, mild, moderate, severe) as external criteria. 14

ASSESSING DYSPEPSIA: CELECOXIB AND DICLOFENAC 821 SODA consists of three scales that measure the concepts of Pain Intensity, Non-Pain Symptoms and Satisfaction with Dyspepsia-Related Health. The Pain Intensity scale consists of six items that ask patients to rate their abdominal discomfort (or stomach ache) with respect to average pain intensity or their worst pain intensity. The recall period is 4 weeks. A Pain Intensity scale score is determined using a scoring algorithm, which was established to convert the scale score as interval-level scaling using Rasch analysis. The number of responses for the Pain Intensity items varies, ranging from four to 10. Hence, the Pain Intensity scores fall in the range 2 47, with higher scores indicating more severe pain. The Non-Pain Symptoms scale consists of seven items, with each item assessing a common upper gastrointestinal symptom. These seven symptoms are burping/ belching, heartburn, bloating, passing gas, sour taste, nausea and bad breath. For each symptom, respondents are asked: During the past 7 days, on average, to what extent did each of the following present a problem for you? Response options range from 1 (no problem) to 5 (very severe problem, markedly influences your daily activities and/or requires rest). Non-Pain Symptoms item scores are summed to create a total scale score. The total score falls in the range 7 35, with higher scores indicating increased non-pain symptoms. The Satisfaction with Dyspepsia-Related Health scale consists of four items that assess the degree to which patients are satisfied with their level of abdominal discomfort. Items measure the following concepts: (i) satisfied/dissatisfied with abdominal discomfort/health; (ii) happy/unhappy with present level of abdominal discomfort; (iii) pleased/displeased with abdominal discomfort control; (iv) pleased/displeased with current abdominal discomfort. A scoring algorithm is used to determine the final score (as in the Pain Intensity scale). The number of responses for the Satisfaction items ranges from four to 10; Satisfaction scores fall in the range 2 23, with higher scores indicating greater satisfaction. SODA was administered to patients at baseline, at weeks 4, 13, 26, 39 and 52, and at withdrawal, if this occurred. At each visit, patients completed SODA before any other assessment. Concomitant gastrointestinal medication The short-term use of antacids (up to 7 days of more than one dose per day each month) and the daily use of calcium-containing antacids as a calcium supplement were also permitted. Prescription or non-prescription anti-ulcer drugs (including histamine-2 antagonists, proton pump inhibitors, sucralfate and misoprostol) were prohibited, as were antibiotics for the treatment of Helicobacter pylori infection. Population for analysis All randomized patients with complete SODA responses at baseline were included in the analysis. Of the 3993 patients who were randomized to receive treatment, 386 had a missing response to at least one SODA item and therefore were excluded. A total of 3607 patients (90.3%) qualified for analysis. In general, patients excluded from the analysis tended to be slightly older (mean, 63.6 vs. 60.2 years; P < 0.001), had a longer disease duration (mean osteoarthritis duration, 11.82 vs. 10.79 years; P ¼ 0.039), and had a more severe arthritis condition, as assessed by the patient global assessment of arthritis (P ¼ 0.024). For the purposes of SODA, the last observation carried forward was used to impute missing scale scores. SODA was developed to measure dyspepsia-related health in patients with uninvestigated dyspepsia. Among arthritis patients using traditional NSAIDs, approximately 15 34% report dyspepsia, while 66 85% do not. Therefore, we tested SODA s ability to detect treatment differences within the population for which it was originally intended (i.e. those with dyspepsia). This comparison was one of dyspepsia severity within dyspepsia patients, as the analysis controls for differences in the overall rate of dyspepsia reporting. For this evaluation, a subset population was formed, which consisted of any patient who self-reported an upper gastrointestinal complaint. A combined set of WHOART preferred terms, similar to the concepts measured by SODA, was used to define an upper gastrointestinal complaint. The WHO- ART preferred terms were abdominal pain, dyspepsia, nausea, flatulence, abdominal fullness, eructation, halitosis and taste perversion. Statistical analysis Two-way analysis of variance (ANOVA), with treatment and centre as factors in the models, was used to compare treatment groups for continuous variables. Chi-squared tests were used to compare treatment groups for categorical variables. Separate analysis of covariance (ANCOVA) models were used to estimate the

822 J. L. GOLDSTEIN et al. adjusted mean change for each of the three SODA scales, where centre and treatment were factors and the patient s baseline score was the covariate. 15 This analysis was performed twice, once in the overall population and once in the subgroup with an upper gastrointestinal complaint. Upper gastrointestinal adverse event incidences were reported and compared using Fisher s exact test. SAS version 6.12 (SAS Institute, Cary, NC, USA) was used to complete the analyses. In all analyses, P values were two-sided. RESULTS Study population were comparable with respect to all the study variables reported in Table 1. The mean age of the study population was 60.5 years (s.d. ¼ 11.4 years) in the celecoxib treatment group and 59.9 years (s.d. ¼ 12.2 years) in the diclofenac group. Most patients were female (diclofenac, 67.2%; celecoxib, 68.3%), Caucasian (diclofenac, 89.6%; celecoxib, 90.3%) and had a diagnosis of osteoarthritis (celecoxib, 72.8%; diclofenac, 73.2%). In addition, 17.1% of patients in the celecoxib group and 17.4% in the diclofenac group used low-dose aspirin ( 325 mg/day) for more than 10% of days during the study follow-up period. SODA results The clinical and demographic characteristics of the study subjects are shown in Table 1. The treatment groups Table 1. Clinical and demographic characteristics of the study population Treatment At baseline, SODA scores were distributed on the lower end for the Pain Intensity and Non-Pain Symptoms Celecoxib 400 mg b.d. (n ¼ 1791) Diclofenac 75 mg b.d. (n ¼ 1816) P value Mean age (years) (s.d.) 60.5 (11.4) 59.9 (12.2) 0.065* Caucasian ethnicity (%) 90.3 89.6 0.84 Female gender (%) 68.3 67.2 0.50 Type of arthritis (%) Osteoarthritis 72.8 73.2 0.82 Rheumatoid arthritis 27.2 26.8 Duration of arthritis (years) Osteoarthritis, mean (s.d.) 10.3 (9.7) 10.2 (10.2) 0.91* Rheumatoid arthritis, mean (s.d.) 11.2 (10.1) 10.4 (9.4) 0.21* History of gastrointestinal bleed (%) 2.0 1.5 0.35 History of gastroduodenal ulcer (%) 8.9 8.6 0.76 History of GI-related NSAID intolerance (%) 10.3 10.1 0.80 Helicobacter pylori positive (%) 38.0 37.2 0.60 Aspirin use (%) 0.77 None 78.2 77.4 One dose to < 10% study days 4.7 5.2 > 10% study days 17.1 17.4 Corticosteroid use (%) 0.402 None 69.7 71.7 One dose to < 10% study days 10.0 9.6 > 10% study days 20.3 18.7 Patient s global assessment of arthritis at baseline (%) 0.82à Very good 2.6 2.0 Good 19.7 21.0 Fair 59.0 59.3 Poor 15.9 15.1 Very poor 2.7 2.3 GI, gastrointestinal; NSAID, non-steroidal anti-inflammatory drug. *P value is based on analysis of variance (ANOVA) after controlling for treatment and centre as factors. P value is based on Pearson chi-squared test. àp value from Cochran Mantel Haenszel test stratified by centre.

ASSESSING DYSPEPSIA: CELECOXIB AND DICLOFENAC 823 scales (indicating a low level of dyspepsia) and the higher end for the Satisfaction scales (indicating a high level of satisfaction). Treatment groups were comparable at baseline for each scale (P ¼ 0.20 0.49) (see Table 2). Pain Intensity changes (follow-up minus baseline) were statistically significantly higher (indicating a worsening status) for diclofenac compared to celecoxib at each follow-up assessment (P < 0.001). The unadjusted mean changes were 2.81 (s.d. ¼ 10.5) for the diclofenac arm and 1.15 (s.d. ¼ 10.4) for the celecoxib arm at the 4-week assessment. For all SODA scales, the means adjusted for the baseline score and study centre were similar to the unadjusted means. The adjusted means are reported in Table 2. Changes (follow-up minus baseline) for the SODA Non- Pain Symptoms score were much smaller than those reported for the Pain Intensity scale. Celecoxib was statistically better than diclofenac at 4 weeks (P ¼ 0.005), but not at weeks 13, 26 and 52 (P ¼ 0.12 and P > 0.20). The unadjusted mean changes were 0.22 (s.d. ¼ 3.46) for diclofenac and unchanged (< 0.01) (s.d. ¼ 3.53) for celecoxib at the 4-week assessment. Celecoxib was statistically superior to diclofenac at each follow-up assessment (P < 0.001) for the Satisfaction with Dyspepsia-Related Health scale. Unadjusted mean changes were ) 0.68 (s.d. ¼ 6.03) for diclofenac and ) 0.11 (s.d. ¼ 5.94) for celecoxib at the 4-week assessment. SODA results: dyspeptic population The analysis (reported in Table 2) was repeated in the subset with a self-reported upper gastrointestinal complaint during the first 26 weeks of the trial. This analysis was a comparison of upper gastrointestinal symptom severity among those with an upper gastrointestinal complaint, as the subsetting controls for any differences in dyspepsia between the two treatment groups. The 32% of celecoxib-treated patients (576 of 1791) and 40.8% of diclofenac-treated patients (741 of 1816) who reported a dyspepsia event within 26 weeks were included in this analysis. This specific analysis has clinical relevance as dyspepsia is the most commonly reported upper gastrointestinal adverse event and was often the cause of patient withdrawal from the trial. Table 2. Mean scores on the Severity of Dyspepsia Assessment Questionnaire Scale score Treatment Celecoxib 400 mg b.d. (n ¼ 1791) Diclofenac 75 mg b.d. (n ¼ 1816) P value Pain Intensity (range 2 47) Baseline 11.18 ± 9.53 11.34 ± 9.32 > 0.20 Change at week 4 0.99 (0.50, 1.48) 2.76 (2.28, 3.25) < 0.001 Change at week 13 1.25 (0.73, 1.77) 3.03 (2.51, 3.55) < 0.001 Change at week 26 1.29 (0.75, 1.83) 2.87 (2.34, 3.40) < 0.001 Change at week 52 1.32 (0.78, 1.86) 2.89 (2.35, 3.43) < 0.001 Non-Pain Symptoms (range 7 35) Baseline 12.00 ± 3.73 12.16 ± 3.66 > 0.20 Change at week 4 ) 0.09 () 0.26, 0.08) 0.21 (0.04, 0.37) 0.005 Change at week 13 0.05 () 0.13, 0.24) 0.21 (0.03, 0.39) 0.17 Change at week 26 0.11 () 0.08, 0.29) 0.18 (0.00, 0.37) > 0.20 Change at week 52 0.03 () 0.16, 0.22) 0.18 ()0.01, 0.36) > 0.20 Satisfaction (range 2 23) Baseline 18.07 ± 5.43 17.83 ± 5.40 > 0.20 Change at week 4 0.02 () 0.26, 0.29) ) 0.72 () 1.00, ) 0.45) < 0.001 Change at week 13 ) 0.29 () 0.58, 0.00) ) 0.97 () 1.26, ) 0.69) < 0.001 Change at week 26 ) 0.26 () 0.55, 0.04) ) 0.95 () 1.24, ) 0.66) < 0.001 Change at week 52 ) 0.30 () 0.59, 0.00) ) 0.93 () 1.22, ) 0.63) < 0.001 Means are adjusted by baseline score and centre. Change is defined as follow-up score minus baseline score. Baseline values are means ± s.d.; all other values are mean change from baseline (95% confidence interval). For Pain Intensity and Non-Pain Symptoms, a higher score indicates higher symptoms/symptom severity. For Satisfaction, a higher score indicates higher satisfaction.

824 J. L. GOLDSTEIN et al. For the Pain Intensity and Satisfaction scales, the mean changes were significantly higher for diclofenac at each time assessment (P < 0.001 for Pain Intensity and P ¼ 0.005 0.048 for Satisfaction). Differences between treatment groups in the mean Pain Intensity and Satisfaction changes were larger than the differences observed in the overall population analysis. For example, the difference in mean change between treatment groups was 1.77 for the overall population and 2.33 in the dyspeptic population. For the Non-Pain Symptoms scale, the mean changes were comparable (P ¼ 0.13 0.46). Adjusted means are reported in Table 3. Gastrointestinal adverse events Celecoxib-treated patients had a lower overall incidence of any gastrointestinal adverse event (40.4% vs. 48.1%; P < 0.001) (see Table 4). Furthermore, patients in the celecoxib treatment group had lower rates for each gastrointestinal adverse event. Overall, 12.8% of diclofenac-treated patients withdrew due to a gastrointestinal adverse event, compared to 8.9% of celecoxib-treated patients (P < 0.001). DISCUSSION Our results indicated that patients treated with celecoxib, at two to four times the maximum recommended dose as used in this study, reported superior dyspepsia tolerability compared to those treated with diclofenac at standard dosages. Celecoxib was superior in all of the three dimensions of dyspepsia (Pain Intensity, Non-Pain Symptoms and Satisfaction with Dyspepsia-Related Health) at the initial assessment. Changes in dyspepsia-related health occurred primarily within the first 4 weeks of the trial and were sustained throughout the observation period. Adverse event data corroborated the SODA results reported herein, as celecoxib patients were significantly less likely to report or withdraw due to gastrointestinal adverse events. These data support the conclusion that celecoxib (800 mg/day) is associated with a superior dyspepsia-related health and dyspepsia tolerability than is diclofenac (150 mg/day). The results reported here are consistent with previous literature on the upper gastrointestinal tolerability of celecoxib relative to non-selective NSAIDs. Previous reports have been based on adverse events measured by the response of the patients to the question: Did you Table 3. Mean scores on the Severity of Dyspepsia Assessment Questionnaire: patients with a self-reported upper gastrointestinal complaint Treatment Scale score Celecoxib 400 mg b.d. (n ¼ 576) Diclofenac 75 mg b.d. (n ¼ 741) P value Pain Intensity (range 2 47) Baseline 12.68 ± 9.88 12.03 ± 9.47 > 0.20 Change at week 4 4.01 (3.03, 4.99) 6.33 (5.43, 7.22) < 0.001 Change at week 13 4.96 (3.91, 6.00) 7.26 (6.32, 8.21) < 0.001 Change at week 26 4.89 (3.81, 5.97) 7.21 (6.22, 8.19) < 0.001 Non-Pain Symptoms (range 7 35) Baseline 12.59 ± 3.73 12.70 ± 3.65 > 0.20 Change at week 4 0.99 (0.66, 1.33) 1.29 (0.98, 1.59) 0.13 Change at week 13 1.31 (0.95, 1.67) 1.46 (1.14, 1.79) > 0.20 Change at week 26 1.28 (0.92, 1.64) 1.55 (1.22, 1.87) > 0.20 Satisfaction (range 2 23) Baseline 17.24 ± 5.59 17.42 ± 4.41 > 0.20 Change at week 4 ) 1.65 () 2.21, ) 1.09) ) 2.44 () 2.95, ) 1.93) 0.005 Change at week 13 ) 2.10 () 2.67, ) 1.53) ) 3.04 () 3.56, ) 2.52) 0.015 Change at week 26 ) 2.23 () 2.81, ) 1.64) ) 3.05 () 3.58, ) 2.52) 0.048 Means are adjusted by baseline score and centre. Change is defined as follow-up score minus baseline score. Baseline values are means ± s.d.; all other values are mean change from baseline (95% confidence interval). For Pain Intensity and Non-Pain Symptoms, a higher score indicates higher symptoms/symptom severity. For Satisfaction, a higher score indicates higher satisfaction.

ASSESSING DYSPEPSIA: CELECOXIB AND DICLOFENAC 825 Table 4. Incidence of gastrointestinal (GI) adverse events Treatment Celecoxib 400 mg b.d. (n ¼ 1791) Diclofenac 75 mg b.d. (n ¼ 1816) P value GI adverse events (% of patients) Any GI adverse event 40.4 48.1 < 0.001 Dyspepsia-like* adverse event 35.5 43.3 < 0.001 Specific GI adverse events Dyspepsia 17.1 19.6 0.053 Abdominal pain 12.2 18.9 < 0.001 Diarrhoea 11.4 15.1 0.001 Nausea 8.4 12.3 < 0.001 Constipation 2.4 6.4 < 0.001 Flatulence 9.8 11.5 0.094 GI adverse event causing withdrawal (%) Any GI adverse event 8.9 12.8 < 0.001 Abdominal pain 3.9 6.6 < 0.001 Dyspepsia 3.5 4.3 N.S. Diarrhoea 1.8 2.8 0.057 Nausea 1.8 2.9 0.036 GI adverse events are those with incidence 5% in either treatment group. Withdrawal due to GI adverse event are those with incidence 2% in either treatment group. * Dyspepsia-like adverse events include: abdominal pain, dyspepsia, nausea, flatulence, abdominal fullness, eructation, halitosis and taste perversion. notice any problems with your arthritis study medications?. In two studies using therapeutic dosages, celecoxib (200 mg/day in osteoarthritis and 400 mg/ day in rheumatoid arthritis) was compared to diclofenac (150 mg/day). In these trials, the rate of various upper gastrointestinal adverse events (e.g. abdominal pain, dyspepsia) consistently favoured celecoxib. Furthermore, the results of the aforementioned trials at therapeutic dosages (200 400 mg/day) were similar to the results reported in this trial with celecoxib, 800 mg/day. 6, 9 The results of this trial are also consistent with literature characterizing NSAID-related dyspepsia. Straus et al. found that abdominal discomfort symptoms are most strongly associated with NSAID use. In their meta-analysis, they found that the relationship of NSAIDs to dyspepsia was strongest when the definition was restricted to abdominal discomfort/pain symptoms (relative risk ¼ 1.46; 95% CI, 1.16 1.85). 16 The association was considerably weakened when other dyspepsia-like symptoms (e.g. bloating, vomiting, heartburn) were included in the definition (relative risk ¼ 1.16; 95% CI, 0.99 1.37). This observation is important for interpreting the SODA results: one would expect celecoxib to be more likely to be different from diclofenac in terms of NSAID-related dyspepsia, which was characterized as abdominal discomfort, and less likely to be different from diclofenac in terms of nonpain dyspepsia symptoms (e.g. bloating, sour taste, heartburn). Indeed, this expectation is consistent with the results reported here: the statistical significance of differences on the Pain Intensity scale was greater than that reported for the Non-Pain Symptoms scale. The data here, however, go beyond the characterization of the NSAID-related dyspepsia and the symptoms themselves. Patients perceive the pain intensity, and the result is a higher patient satisfaction with dyspepsiarelated health with celecoxib. An important finding was that pain intensity was significantly higher and satisfaction significantly lower in diclofenac patients with an upper gastrointestinal complaint compared to celecoxib patients with an upper gastrointestinal complaint. These findings are important in terms of the patient s and clinician s response to dyspepsia. An upper gastrointestinal complaint in an NSAID user may, due to its increased severity, require a greater clinical response than an upper gastrointestinal complaint in a celecoxib user. A greater response would be one that more often requires the addition of a gastrointestinal medication

826 J. L. GOLDSTEIN et al. and the selection of a more potent gastrointestinal medication if one is used (e.g. proton pump inhibitor rather than antacid or histamine-2 antagonist). The results also have implications in terms of drug selection in arthritis patients. Clinicians treating NSAID users who present with upper gastrointestinal tolerability problems should consider using celecoxib because celecoxib is associated with a lower risk of gastrointestinal tolerance than NSAIDs, and because patients with a history of intolerance to NSAIDs are more likely to have subsequent intolerance to NSAIDs. 17 Several limitations of the study are also worth noting. One limitation was the use of a single non-selective NSAID, diclofenac, upon which to base the comparison with celecoxib. The degree to which the dyspepsia experiences of patients treated with diclofenac may be generalized to the experiences of patients treated with other NSAIDs is undetermined. Second, a placebo arm was not included in the design of this trial, and therefore we were unable to compare the effect on dyspepsia in the active treatment groups with that in a nontreatment group. However, in a previous report of pooled clinical trials, celecoxib was comparable to placebo in terms of upper gastrointestinal tolerability at dosages ranging from 100 mg/day to 800 mg/ day. 17 Third, further work is needed to identify changes in SODA, which are clinically meaningful in an arthritis population both to individual patients and at the group level. Rabeneck et al. have initiated research in this area. 14 Another potential limitation was patient withdrawal in the trial (13.9% and 18.8% of patients by 4 and 13 weeks, respectively). Early withdrawal is important because a large percentage of the SODA responses represent carried forward scores, as opposed to actual patient responses. Early withdrawal, however, increases the importance of the earliest treatment comparison (i.e. the 4-week comparison), where celecoxib was statistically superior to diclofenac in all three dimensions. These results demonstrated that, in a population of arthritis NSAID users that contained both osteoarthritis (73%) and rheumatoid arthritis (27%) patients and users of low-dose aspirin (22%), the dyspepsia tolerability of celecoxib was superior to that of diclofenac at standard dosages. Celecoxib, at two to four times the manufacturer s recommended dose, was superior in all of the three dimensions of dyspepsia (Pain Intensity, Non-Pain Symptoms and Satisfaction with Dyspepsia- Related Health) at the initial assessment. These changes were generally sustained throughout the observation period. ACKNOWLEDGEMENTS Thanks are due to John Xu for statistical programming. Pharmacia is the manufacturer of celecoxib (Celebrex), which is comarketed by Pharmacia and Pfizer. Jay L. Goldstein and Glenn M. Eisen serve as consultants for Pharmacia and Pfizer; James Lefkowith, G. Steven Geis and Thomas A. Burke are employees of Pharmacia Corporation; Beatriz M. Peña is an employee of Pfizer Inc. REFERENCES 1 Larkai EN, Smith JL, Kidsky MD, et al. Dyspepsia in NSAID users: the size of the problem. J Clin Gastroenterol 1989; 11: 158 62. 2 Griffin MR. Epidemiology of nonsteroidal anti-inflammatory drug-associated gastrointestinal injury. Am J Med 1998; 104(Suppl. 3A): 23S 29S. 3 Spencer-Green G, Spencer-Green E. Nonsteroidal therapy of rheumatoid arthritis and osteoarthritis: how physicians manage treatment failures. J Rheumatol 1998; 25: 2088 93. 4 Silverstein FE, Graham DY, Senior JR, et al. Misoprostol reduces serious gastrointestinal complications in patients with rheumatoid arthritis receiving nonsteroidal anti-inflammatory drugs. A randomized, double-blind, placebo-controlled trial. Ann Intern Med 1995; 123: 241 9. 5 Bensen WG, Fiechtner JJ, McMillen JI, et al. Treatment of osteoarthritis with celecoxib, a cyclooxygenase-2 inhibitor: a randomized controlled trial. Mayo Clin Proc 1999; 74(11): 1095 105. 6 Emery P, Zeidler H, Kvien TK, et al. Celecoxib versus diclofenac in long-term management of rheumatoid arthritis: randomised double-blind comparison. Lancet 1999; 354: 2106 11. 7 Bombardier C, Laine L, Reicin A, et al. Comparison of upper gastrointestinal toxicity of rofecoxib and naproxen in patients with rheumatoid arthritis. N Engl J Med 2000; 343: 1520 8. 8 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. J Am Med Assoc 2000; 284: 1247 55. 9 McKenna F, Borenstein D, Wendt H, et al. Celecoxib versus diclofenac in the management of osteoarthritis of the knee. Scand J Rheumatol 2001; 30: 11 8. 10 Singh G, Ramey DR, Morfeld D, et al. Gastrointestinal tract complications of nonsteroidal anti-inflammatory drug treatment in rheumatoid arthritis. Arch Intern Med 1996; 156: 1530 6. 11 Wolfe F, Kong SX, Watson DJ. Gastrointestinal symptoms and health related quality of life in patients with arthritis. J Rheumatol 2000; 27: 1373 8.

ASSESSING DYSPEPSIA: CELECOXIB AND DICLOFENAC 827 12 Cook KF, Rabenack L, Campbell CJM, et al. Evaluation of a multidimensional measure of dyspepsia-related health for use in a randomized clinical trial. J Clin Epidemiol 1999; 52: 381 92. 13 Rabeneck L, Cook KF, Wristers K, et al. SODA (severity of dyspepsia assessment): a new effective outcome measure for dyspepsia-related health. J Clin Epidemiol 2001; 54: 755 65. 14 Rabeneck L, Wristers K, Goldstein JL, et al. Reliability, validity, and responsiveness of Severity of Dyspepsia Assessment (SODA) in a randomized clinical trial of COX-2 specific inhibitor and traditional NSAID therapy. Am J Gastroenterol 2002; 97: 32 9. 15 Kleinbaum D, Kupper L, Muller K, et al. Applied Regression Analysis and Multivariable Methods, 3rd edn. London: Duxbury Press, 1998. 16 Straus WL, Ofman JJ, MacLean C, Morton S. Do NSAIDs cause dyspepsia? A meta-analysis evaluating alternative definitions. Gastroenterology 1999; 116: A97(Abstract). 17 Bensen WG, Zhao SZ, Burke TA, et al. Upper gastrointestinal tolerability of celecoxib, a COX-2 specific inhibitor, compared to naproxen and placebo. J Rheumatol 2000; 27: 1876 83.