Simon LS, Weaver AL, Graham DY, Kivitz AJ, Lipsky PE, Hubbard RC, Isakson PC, Verburg KM, Yu SS, Zhao WW, Geis GS. Anti-inflammatory and Upper Gastrointestinal Effects of Celecoxib in Rheumatoid ArthritisA Randomized Controlled Trial. JAMA. 1999;282(20):1921–1928. doi:10.1001/jama.282.20.1921
Author Affiliations: Department of Medicine, Division of Rheumatology and Metabolic Bone Disease, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, Mass (Dr Simon); the Arthritis Center of Nebraska, Lincoln (Dr Weaver); Department of Medicine, Veterans Affairs Medical Center/Baylor College of Medicine, Houston, Tex (Dr Graham); Altoona Center for Clinical Research, Altoona, Pa (Dr Kivitz); Rheumatic Diseases Division, University of Texas Southwestern Medical Center, Dallas (Dr Lipsky); and Departments of Clinical Research (Drs Hubbard, Verburg, Yu, Zhao, and Geis) and COX-2 Technology (Dr Isakson), Searle Research and Development, Skokie, Ill.
Context In vitro studies have shown that celecoxib inhibits cyclooxygenase 2
(COX-2) but not COX-1, suggesting that this drug may have anti-inflammatory
and analgesic activity without adverse upper gastrointestinal (GI) tract effects
that result from COX-1 inhibition.
Objective To test whether celecoxib has efficacy as an anti-inflammatory and analgesic
with reduced GI tract mucosal damage compared with conventional nonsteroidal
anti-inflammatory drugs in patients with rheumatoid arthritis.
Design Randomized, multicenter, placebo-controlled, double-blind trial lasting
12 weeks, with follow-up at weeks 2, 6, and 12, from September 1996 thorugh
Setting Seventy-nine clinical sites in the United States and Canada.
Patients A total of 1149 patients aged 18 years or older with symptomatic rheumatoid
arthritis who met inclusion criteria were randomized; 688 (60%) of these completed
Interventions Patients were randomized to receive celecoxib, 100 mg, 200 mg, or 400
mg twice per day (n = 240, 235, and 218, respectively); naproxen, 500 mg twice
per day (n = 225); or placebo (n = 231).
Main Outcome Measures Improvement in signs and symptoms of rheumatoid arthritis as assessed
using standard measures of efficacy and GI tract safety as assessed by upper
GI tract endoscopy before and after treatment, compared among treatment groups.
Results All dosages of celecoxib and naproxen significantly improved the signs
and symptoms of arthritis compared with placebo. Maximal anti-inflammatory
and analgesic activity was evident within 2 weeks of initiating treatment
and was sustained throughout the 12 weeks. The incidence of endoscopically
determined gastroduodenal ulcers in placebo-treated patients was 4 (4%) of
99, and the incidences across all dosages of celecoxib were not significantly
different (P>.40): 9 (6%) of 148 with 100 mg twice
per day, 6 (4%) of 145 with 200 mg twice per day, and 8 (6%) of 130 with 400
mg twice per day. In contrast, the incidence with naproxen was 36 (26%) of
137, significantly greater than either placebo or celecoxib (P<.001). The overall incidences of GI tract adverse effects were
19% for placebo; 28%, 25%, and 26% for celecoxib 100 mg, 200 mg, and 400 mg
twice per day, respectively; and 31% for naproxen.
Conclusion In this study, all dosages of celecoxib were efficacious in the treatment
of rheumatoid arthritis and did not affect COX-1 activity in the GI tract
mucosa as evidenced by less frequent incidence of endoscopic ulcers compared
Prostanoic acids are synthesized in response to physiologic stimuli
that modulate and maintain homeostasis. Prostanoic acids are also produced
during acute and chronic inflammatory processes, and it is generally accepted
that they mediate many of the symptoms of inflammation such as edema and pain.1,2 The 2 isoforms of cyclooxygenase (COX),
COX-1 and COX-2, catalyze the committed step in the synthesis of prostanoic
acids from arachidonic acid.3 Recent pharmacological
evidence reinforces the likelihood that these isoenzymes mediate different
biological functions.4,5 COX-1
is constitutively expressed in many tissues and produces prostanoic acids
that predominantly regulate normal cellular processes.6- 8
In contrast, COX-2 activity is typically undetectable in most tissues; however,
COX-2 expression can be rapidly induced by proinflammatory cytokines or by
growth factors.8- 11
As a result of the research that characterized the role of COX-2 in
prostanoic acid production,4,5,9,12- 14
a class of anti-inflammatory and analgesic agents that primarily inhibit COX-2
while sparing COX-1 at therapeutic dosages has been developed.15- 18
The clinical rationale for this effort is that by sparing COX-1 activity,
COX-2–specific inhibitors are not expected to interfere with homeostatic
prostanoid-dependent processes such as upper gastrointestinal (GI) tract mucosal
protection and platelet aggregation. The potential clinical benefit of this
strategy is important given that patients who take nonsteroidal anti-inflammatory
drugs (NSAIDs), which inhibit both COX-1 and COX-2,15
incur a 3- to 10-fold higher risk of gastroduodenal injury and death than
those who do not.19- 22
Endoscopic studies have shown that the prevalence of gastroduodenal ulcers
is 15% to 30% among users of conventional (ie, nonspecific) NSAIDs.23- 26 Large,
randomized trials have suggested that endoscopic ulcers are surrogate markers
for NSAID-induced complications such as bleeding, perforation, and obstruction.27,28 Celecoxib has been shown to inhibit
COX-2 and spare COX-1 activity in vitro while possessing effective anti-inflammatory
and analgesic properties when studied in vivo.8,18,29- 31
It is recommended for the treatment of osteoarthritis at 100 mg 2 times a
day or 200 mg once daily, and for the treatment of rheumatoid arthritis (RA)
at 100 to 200 mg twice per day. This randomized, placebo-controlled, double-blind,
12-week trial was conducted to test the hypothesis that celecoxib has efficacy
as an anti-inflammatory and analgesic drug through COX-2 inhibition but has
little effect on COX-1 activity at efficacious doses as evidenced by reduced
GI tract mucosal damage defined by endoscopy. The efficacy and upper GI tract
safety of celecoxib in treating RA was assessed and compared with the effects
of naproxen and placebo.
Men and women outpatients aged 18 years or older were eligible to participate
in the study if they fulfilled the American College of Rheumatology (ACR-20)
criteria for a diagnosis of RA evident for 3 months or longer32
and were in a functional class of I, II, or III.33
Additional selection criteria were based on disease activity.
Patients were eligible to participate if the dosages of any glucocorticoids,
disease-modifying antirheumatic drugs, or methotrexate had been stable and
were expected to remain constant throughout the study.
Patients were excluded from the study if they had active GI tract, renal,
hepatic, or coagulation disorders; history of malignancy (unless removed surgically
with no recurrence within 5 years); esophageal or gastroduodenal ulceration
within the previous 30 days; or a history of gastric or duodenal surgery other
than an oversew. In addition, patients were excluded if the upper GI tract
endoscopy performed at baseline disclosed an esophageal, gastric, or duodenal
ulcer or more than 10 erosions in the stomach or duodenum. Patients were not
excluded for a history of peptic ulcer disease.
This prospective, randomized, double-blind trial was conducted at 79
clinical sites in the United States and Canada in accordance with the principles
of Good Clinical Practice and the Declaration of Helsinki. The protocol was
approved by the institutional review board at each clinical site, and all
patients were required to provide written informed consent. Quality control
measures included site visits, verification of case report forms against source
medical records, and site audits by sponsor personnel.
Prior to enrollment, patients completed a physical examination and clinical
laboratory testing. A baseline serological antibody test for Helicobacter pylori (FlexSure, Beckman-Coulter, Palo Alto, Calif) was
included. Screening or baseline clinical assessments of arthritis included
patients' and physicians' global assessment of arthritis, scored on a scale
of 1 (very good) to 5 (very poor); the patients' assessment of arthritis pain
marked on a visual analog scale (VAS) from 0 mm (no pain) to 100 mm (severe
pain); a complete count of tender/painful joints; a complete count of swollen
joints (hip joints were not assessed); duration of morning stiffness; the
health assessment questionnaire Functional Disability Index; and plasma levels
of C-reactive protein.34- 38
Following a 2- to 7-day washout period of NSAIDs or any analgesic medication,
symptomatic RA (flare) was confirmed at a baseline visit according to the
following definition: physicians' and patients' global assessments of "fair,"
"poor," or "very poor" and the first 2 plus either the third or the fourth
of the following: (1) the presence of at least 6 tender or painful joints
with an increase of 20% or at least 2 joints; (2) a minimum of 3 swollen joints
with an increase of 20% or at least 2 joints; (3) a minimum of 45 minutes
of morning stiffness and increase of at least 15 minutes; or (4) patients'
assessment of pain of at least 40 mm on the VAS and an increase of at least
20% or 10 mm.
An upper GI tract endoscopic evaluation was performed within 7 days
prior to the first dose of study medication. The mucosae of the stomach and
duodenum were evaluated separately for the presence of petechiae, erosions,
and ulcers. An ulcer was defined as any break in the mucosa at least 3 mm
in diameter with unequivocal depth.
Patients were assigned by a computer-generated randomization schedule
to 1 of 5 treatment groups: placebo, celecoxib 100 mg twice per day, celecoxib
200 mg twice per day, celecoxib 400 mg twice per day, or naproxen 500 mg twice
per day (Figure 1).
Randomization was stratified by center using a block size of 10 treatments.
All treatment regimens were fully masked so that all patients took the same
number of capsules, and all regimens were identical in appearance and frequency.
Stable doses of aspirin no more than 325 mg/d were allowed, and acetaminophen
up to 2 g/d for no longer than 3 consecutive days was also allowed except
within 48 hours prior to arthritis assessments, during which no analgesics
were allowed. NSAIDs, injectable corticosteroids, and anticoagulants were
prohibited. Stable doses of oral glucocorticoids (up to 10 mg of prednisone
per day) or disease-modifying antirheumatic drugs (DMARDs) were allowed and
antiulcer drugs were prohibited.
Clinical efficacy and safety assessments were performed at weeks 2,
6, and 12. Efficacy assessments were identical to those performed at the screening
and baseline visits. Safety was evaluated according to the incidence and type
of adverse reactions and clinical laboratory abnormalities. At the final treatment
(or early termination) visit, each patient underwent a second upper GI tract
endoscopy, and a CLO test for H pylori was performed
on a tissue sample taken from the greater curvature of the stomach. In all
cases, the endoscopist was blinded to the treatment a patient was receiving.
Patient demographic and baseline characteristics are shown in Table 1.
Homogeneity of the treatment groups at baseline was analyzed using the χ2 (for sex and race), 2-way analysis of variance (for continuous demographic
variables and baseline disease activity), and the Cochran-Mantel-Haenszel
test (for patients' and physicians' global assessments). Differences among
the treatment groups in concurrent medication use were analyzed with the Fisher
Efficacy analyses were based on the intent-to-treat cohort, defined
as all patients who took at least 1 dose of study medication. In all efficacy
measures, including the composite ACR-20 analysis, missing values for any
assessment time were imputed by carrying forward the last observed value for
any patient who discontinued the study for any reason (including treatment
failure) before completing 12 weeks. Continuous efficacy variables were compared
among treatment groups using analysis of covariance with treatment and center
as factors and the corresponding baseline value as a covariate. Hochberg's
step-up procedure was used to control for type-1 error associated with multiple-treatment
comparisons at each time point within each efficacy variable.39
For categorical efficacy variables (Patient's and Physician's Global
Assessments and the ACR-20 responder criteria40),
the Cochran-Mantel-Haenszel test, stratified by center, was used to compare
results among treatment groups. Incidence of withdrawal due to treatment failure
was analyzed with the Fisher exact test.
The gastroduodenal ulcer incidences at week 12 were analyzed with Cochran-Mantel-Haenszel
tests stratified by baseline status; 95% confidence intervals (CIs) for the
ulcer incidences were also calculated. The overall effects of H pylori status and H pylori status by treatment
interaction were examined using both analysis of covariance and Cochran-Mantel-Haenszel
test. The effects of concurrent aspirin or corticosteroid use, history of
gastroduodenal ulcers, and history of GI tract bleeding were analyzed in a
The planned sample size was based on the expectation that 35% of patients
receiving active treatment would show improvement compared with 20% of placebo-treated
patients. A sample size of 200 patients per treatment group was sufficient
to detect this difference with 80% power at an α level of .05 adjusted
for 3 celecoxib doses vs placebo by the Bonferroni method. This sample size
was also sufficient to detect an anticipated difference in endoscopic gastroduodenal
ulcer rate of 3% (celecoxib 400 mg twice per day) vs 11% (naproxen 500 mg
twice per day) at the same α level and power.
A total of 1149 patients were enrolled. No significant differences among
the treatment groups at entry were detected with respect to baseline characteristics
(Table 1). The study was completed
by 688 patients (60%). Figure 1
shows reasons for early discontinuation from the study and includes the numbers
of patients withdrawing during each interval, indicating the extent of data
extrapolation at each assessment time.
Baseline endoscopic scores were not significantly different among treatment
groups. More than 50% of the patients had normal gastric and duodenal mucosae,
and no patients had an ulcer. The incidence of H pylori positive serology results at baseline was also not statistically significantly
different across the treatment groups, ranging from 23% to 34% of patients.
Celecoxib produced significant improvement in the signs and symptoms
of RA for all efficacy measures. As shown by the reduced number of tender/painful
and of swollen joints among those treated (Figure 2), celecoxib produced statistically significant and maximal
effects by week 2, which were sustained through 12 weeks. All celecoxib doses
generally demonstrated similar efficacy, and all were comparable to naproxen
500 mg twice per day.
The percentages of patients who responded (improved) by ACR-20 criteria
at weeks 2, 6, and 12 are shown in Figure
3. The results show significant and comparable treatment effects
among patients in all the dose groups of celecoxib and naproxen, with maximal
effect achieved by week 2.
For other efficacy measures, week 12 results are presented in Table 2. In the patients' and physicians'
global assessments, celecoxib was associated with statistically significant
treatment effects compared with placebo. For patients' global assessment,
all celecoxib dose groups had significantly better scores than the placebo
group. However, for Physicians' Global Assessment, only those in the 200-mg
and 400-mg, twice-per-day celecoxib-dose groups had significantly better scores
than those in the placebo group. Naproxen was not significantly different
from placebo at week 12 in either measure of efficacy. In patients' assessment
of arthritis pain and duration of morning stiffness, all active treatments
showed significant improvement and were statistically distinct from placebo.
Improvements in the health assessment questionnaire functional disability
scores were significant for those taking celecoxib 200 mg and 400 mg twice
per day (P<.001 for both) and those taking naproxen
compared with those taking placebo (P = .008). Neither
celecoxib nor naproxen was associated with demonstrable effects on C-reactive
Withdrawals from the study due to treatment failure were significantly
lower for all active treatment groups (P<.001
for all) than for the placebo group: 104 (45%) of placebo patients compared
with 67 (28%) of patients receiving 100 mg, 50 (21%) receiving 200 mg, and
59 (27%) receiving 400 mg of celecoxib twice per day, and 65 (29%) of patients
receiving naproxen (Figure 1).
Figure 4 shows the incidences
of ulceration over the 12-week course of the trial in patients who completed
the study and underwent final endoscopic evaluation. Any endoscopic finding
other than ulcer was categorized as unknown if the data were obtained before
the 12-week visit; only patients with endoscopy results categorized as known,
including all patients found to have an ulcer at any time, are included in
the analysis. In 99 patients receiving placebo, gastroduodenal ulcers developed
in 4 (4% [95% CI, 0.1%-7.9%]); in 148 receiving 100-mg celecoxib, ulcers developed
in 9 (6% [95% CI, 2.2%-10.0%]); in 145 receiving 200-mg celecoxib, ulcers
developed in 6 (4% [95% CI, 0.9%-7.3%]); and in 130 receiving 400-mg celecoxib
twice daily, ulcers developed in 8 (6% [95% CI, 2.1%-10.4%]). In comparison,
of 137 patients receiving naproxen, 36 developed gastroduodenal ulcers (26%
[95% CI, 18.9%-33.7%]). There were no statistically significant differences
in the incidence of gastroduodenal ulcers between the placebo group and any
of the celecoxib groups (P>.40) and no evidence of
a dose response, whereas the incidence of ulceration in the naproxen group
was significantly greater than in each of the other treatment groups (P<.001). A comparison of the effects of H pylori status, concurrent aspirin or corticosteroid use, history
of GI tract bleeding, or history of GI tract ulcers on the incidence of gastroduodenal
ulcers within treatment groups showed that none of these factors was associated
with an effect on ulceration.
All doses of celecoxib were well tolerated in this study. The incidences
of adverse events among the celecoxib treatment groups were generally higher
than in the placebo group but did not suggest a dose response. The adverse
events with the highest incidence were headache, upper respiratory tract infection,
dyspepsia, diarrhea, and abdominal pain (Table 3).
The incidences of the most frequently reported GI tract adverse events
(dyspepsia, diarrhea, abdominal pain, nausea, and flatulence) combined were
19% for placebo; 28% for 100 mg, 25% for 200 mg, and 26% for 400 mg of celecoxib
twice per day; and 31% for naproxen.
No adverse renal effects of celecoxib were detected. The incidences
of peripheral edema and hypertension were low (0%-2%) and were similar among
all treatment groups (Table 3).
As representative measures, mean blood pressures and creatinine values decreased
slightly over the 12 weeks in all treatment groups (Table 4).
Serious adverse events (representing hospitalizations or malignancies
detected during study participation) were reported for 5 patients (2%) receiving
placebo; 4 patients (2%) receiving 100 mg, 5 (2%) receiving 200 mg, and 4
(2%) receiving 400 mg of celecoxib twice per day; and 4 patients (2%) receiving
naproxen. None of these events was considered to be related to study medication.
One clinically significant upper GI tract ulcer complication occurred
during the study. An 80-year-old woman who received naproxen 500 mg twice
per day developed an ulcer on the superior wall of the duodenal bulb and a
large postbulbar ulcer on the anterosuperior wall of the duodenum, creating
a partial gastric outlet obstruction after 22 days of treatment.
We tested the hypothesis that an agent that inhibits COX-2 while sparing
COX-1 will be as effective as conventional NSAIDs (that inhibit both COX-1
and COX-2) but at therapeutic doses will not interfere with other prostaglandin-dependent
homeostatic processes such as upper GI tract mucosal integrity.7,8,29,30,41,42
The results of our study provide evidence supporting the hypothesis. Celecoxib
demonstrated anti-inflammatory and analgesic efficacy comparable with naproxen,
with a significantly lower incidence of gastroduodenal ulceration than naproxen,
and not significantly different from placebo.
All doses of celecoxib were associated with anti-inflammatory and analgesic
efficacy. This efficacy was reflected by improvements in all efficacy measures
beginning at week 2 and sustained over 12 weeks. Total daily celecoxib doses
of 200 mg to 400 mg were maximally efficacious, with no further benefit observed
with the 400 mg twice per day regimen (800 mg/d). The efficacy of celecoxib
was comparable with naproxen, and the improvement in patients treated with
naproxen was similar to previously reported results from RA efficacy trials
investigating the efficacy of naproxen and other NSAIDs.43,44
Approximately 20% to 30% of patients who take conventional NSAIDs develop
persistent adverse effects, and more than 10% are estimated to discontinue
treatment as a result.45 In this study, the
GI tract tolerability of celecoxib was found to be intermediate between that
for placebo and that for naproxen, as shown by incidences of GI tract adverse
events and withdrawals due to GI tract adverse events. (Because crude incidences
are not normalized for differing lengths of exposure, the ability to interpret
these data is limited.) Overall, celecoxib was well tolerated.
It is well established that conventional NSAID therapy can lead to gastroduodenal
ulceration and associated serious complications of perforation, hemorrhage,
and gastric outlet obstruction.19- 26
There is evidence to suggest that NSAID-induced ulcers and their resulting
complications are largely caused by NSAID-mediated inhibition of mucosal prostaglandin
production, primarily mediated by COX-1 activity.45,46
Prostaglandins have been shown to modulate gastroduodenal mucosal protection
by several interrelated mechanisms.47,48
In animal models, NSAID-induced GI tract toxicity has been isolated to inhibition
of COX-1 activity.17,30
The results of this study provide clinical evidence for the association
of celecoxib with improved endoscopic upper GI tract safety compared with
naproxen. Moreover, the incidence rates of gastroduodenal ulcers associated
with celecoxib, even at 4 times the recommended dose, were not significantly
different from that observed with placebo. However, it should be noted that
the study was not powered to show equivalence between celecoxib and placebo.
The incidence of gastroduodenal ulcers among patients receiving placebo
in our study is similar to that observed in previous studies in which the
point prevalence of gastroduodenal ulcers in normal asymptomatic volunteers
was examined by upper GI tract endoscopy.49,50
The prevalence of gastroduodenal ulcers in untreated patients in these previous
studies ranged from 1.7% to 4.3%. The incidence of gastroduodenal ulcers among
patients who received naproxen in our study was 26%, which is similarly consistent
with previous endoscopic studies of upper GI tract damage induced by naproxen
or other NSAIDs.23- 26
The precise cause of the ulcers that develop in the patients treated
with either placebo or celecoxib in our study is uncertain. Neither H pylori–positive status nor low-dose aspirin use
(≤325 mg/d), both known ulcerogenic factors,51,52
was shown to be a factor contributing to gastroduodenal ulcer formation.
The observed differences in upper GI tract ulceration between celecoxib
and naproxen are important since ulcers are generally thought to be precursors
for potentially fatal ulcer complications (perforation, bleeding, or obstruction).27,28 If true, these data would indicate
that drugs that inhibit COX-2 while sparing COX-1 may result in a decreased
rate of ulcer complications compared with conventional NSAIDs.
Overall, these results provide evidence of the clinical benefits of
celecoxib in the treatment of RA. Current therapeutic strategies for RA usually
consist of combination therapy including NSAIDs together with corticosteroids
and disease-modifying agents.53 In this 12-week
study, celecoxib produced improvement in the signs and symptoms of RA comparable
with the effects of naproxen but with a significantly reduced incidence of
endoscopically identified gastroduodenal ulcers. Thus, one of the major impediments
that can limit the effective use of conventional NSAIDs, upper GI tract toxic
effects, may potentially be obviated by the use of celecoxib.