Portal de Periódicos da CAPES

Video capsule endoscopy to prospectively assess small bowel injury with celecoxib, naproxen plus omeprazole, and placebo

2005; Elsevier BV; Volume: 3; Issue: 2 Linguagem: Inglês

10.1016/s1542-3565(04)00619-6

1542-7714

Jay L. Goldstein, Glenn M. Eisen, Blair S. Lewis, Ian M. Gralnek, Steve Zlotnick, John G. Fort,

Gastroesophageal reflux and treatments

Background & Aims: Data indicate that cyclooxygenase-2–specific inhibitors cause less gastroduodenal mucosal damage than nonspecific NSAIDS, but their effects on the small bowel mucosa are less well recognized. In a multicenter, double-blind, placebo-controlled trial with video capsule endoscopy (VCE) we prospectively evaluated the incidence of small bowel injury in healthy subjects treated with celecoxib compared to naproxen plus omeprazole. Methods: We randomly assigned subjects with normal baseline VCEs to celecoxib 200 mg twice daily (n = 120), naproxen 500 mg twice daily plus omeprazole 20 mg once daily (n = 118), or placebo (n = 118) for 2 weeks. The primary end point was the mean number of small bowel mucosal breaks per subject. Results: Baseline VCE found small bowel lesions in 13.8% (57/413) of screened subjects, who became ineligible for randomization. The mean number of small bowel mucosal breaks per subject and the percentage of subjects with these mucosal breaks were 2.99 ± 0.51, 55% for naproxen/omeprazole compared to 0.32 ± 0.10, 16% for celecoxib and 0.11 ± 0.04, 7% for placebo (P < .001, both comparisons). The magnitude of the difference between celecoxib and placebo was small but statistically significant (P = .04). Conclusions: Among healthy subjects with lesion-free baseline VCEs, celecoxib was associated with significantly fewer small bowel mucosal breaks than naproxen plus omeprazole. This study also showed that the background incidence of small bowel lesions in healthy adults is not insignificant and should be considered in future trials with VCE. Background & Aims: Data indicate that cyclooxygenase-2–specific inhibitors cause less gastroduodenal mucosal damage than nonspecific NSAIDS, but their effects on the small bowel mucosa are less well recognized. In a multicenter, double-blind, placebo-controlled trial with video capsule endoscopy (VCE) we prospectively evaluated the incidence of small bowel injury in healthy subjects treated with celecoxib compared to naproxen plus omeprazole. Methods: We randomly assigned subjects with normal baseline VCEs to celecoxib 200 mg twice daily (n = 120), naproxen 500 mg twice daily plus omeprazole 20 mg once daily (n = 118), or placebo (n = 118) for 2 weeks. The primary end point was the mean number of small bowel mucosal breaks per subject. Results: Baseline VCE found small bowel lesions in 13.8% (57/413) of screened subjects, who became ineligible for randomization. The mean number of small bowel mucosal breaks per subject and the percentage of subjects with these mucosal breaks were 2.99 ± 0.51, 55% for naproxen/omeprazole compared to 0.32 ± 0.10, 16% for celecoxib and 0.11 ± 0.04, 7% for placebo (P < .001, both comparisons). The magnitude of the difference between celecoxib and placebo was small but statistically significant (P = .04). Conclusions: Among healthy subjects with lesion-free baseline VCEs, celecoxib was associated with significantly fewer small bowel mucosal breaks than naproxen plus omeprazole. This study also showed that the background incidence of small bowel lesions in healthy adults is not insignificant and should be considered in future trials with VCE. Data from uncontrolled case series and epidemiologic studies have shown that nonsteroidal anti-inflammatory drugs (NSAIDs) may cause small bowel lesions, including ulcerations.1Lengeling R.W. Mitros F.A. Brennan J.A. et al.Ulcerative ileitis encountered at ileo-colonoscopy likely role of nonsteroidal agents.Clin Gastroenterol Hepatol. 2003; 1: 160-169Abstract Full Text Full Text PDF PubMed Scopus (76) Google Scholar, 2Allison M.C. Howatson A.G. Torrance C.J. et al.Gastrointestinal damage associated with the use of nonsteroidal anitiinflammatory drugs.N Engl J Med. 1992; 327: 751-754Crossref Scopus (969) Google Scholar, 3Morris A.J. Wasson L.A. MacKenzie J.F. Small bowel enteroscopy in undiagnosed gastrointestinal blood loss.Gut. 1992; 33: 887-889Crossref PubMed Scopus (148) Google Scholar, 4Morris A.J. Madhok R. Sturrock R.D. et al.Enteroscopic evidence of small bowel ulceration in patients receiving non-steroidal anti-inflammatory drugs.Lancet. 1991; 337: 520Abstract PubMed Scopus (155) Google Scholar Clinically relevant complications of so-called small bowel NSAID-associated enteropathy may include the development of webs, strictures/obstructions, acute bleeding, perforations, and occult gastrointestinal bleeding, which may lead to the development of iron-deficiency anemia.1Lengeling R.W. Mitros F.A. Brennan J.A. et al.Ulcerative ileitis encountered at ileo-colonoscopy likely role of nonsteroidal agents.Clin Gastroenterol Hepatol. 2003; 1: 160-169Abstract Full Text Full Text PDF PubMed Scopus (76) Google Scholar, 2Allison M.C. Howatson A.G. Torrance C.J. et al.Gastrointestinal damage associated with the use of nonsteroidal anitiinflammatory drugs.N Engl J Med. 1992; 327: 751-754Crossref Scopus (969) Google Scholar, 3Morris A.J. Wasson L.A. MacKenzie J.F. Small bowel enteroscopy in undiagnosed gastrointestinal blood loss.Gut. 1992; 33: 887-889Crossref PubMed Scopus (148) Google Scholar, 4Morris A.J. Madhok R. Sturrock R.D. et al.Enteroscopic evidence of small bowel ulceration in patients receiving non-steroidal anti-inflammatory drugs.Lancet. 1991; 337: 520Abstract PubMed Scopus (155) Google Scholar, 5Smale S. Tibble J. Sigthorsson G. et al.Epidemiology and differential diagnosis of NSAID-induced injury to the mucosa of the small intestine.Best Pract Res Clin Gastroenterol. 2001; 15: 723-738Abstract Full Text PDF PubMed Scopus (59) Google Scholar, 6Ihse I. Lunderquist A. Akerman M. Chronic bleeding from a primary non-specific small intestinal ulceration localized by angiography.Acta Chir Scand. 1978; 144: 189-192Google Scholar, 7Walker C.O. Sorsdahl O.A. Anemia due to nonspecific ileal ulceration.South Med J. 1968; 61: 127-128Crossref Scopus (1) Google Scholar Although small bowel enteroscopy has been used in clinical practice to examine suspected small bowel pathologies, its utility has been limited by technical difficulties, relative discomfort for the patient, and incomplete evaluation of the small bowel. A noninvasive video capsule endoscopy (VCE) system providing detailed digital imaging of the small bowel has been recently approved by the Food and Drug Administration and is recommended as a first-line diagnostic tool for detecting small bowel pathologies (Given Diagnostic System; Given Imaging Ltd, Yoqneam, Israel).8Lewis B. Goldfarb N. The advent of capsule endoscopy a not-so-futuristic approach to obscure gastrointestinal bleeding.Aliment Pharmacol Ther. 2003; 17: 1085-1096Crossref PubMed Scopus (68) Google Scholar, 9Bowel disorders detection system approved. FDAnews Daily Bulletin July 11, 2003. Available at: www.fdanews.com/dailies/bulletin/1_134/news/15318-1.html. Accessed April 26, 2004.Google Scholar, 10Swain P. Wireless capsule endoscopy.Gut. 2003; 52: iv48-50PubMed Google Scholar The gastroduodenal ulcer complications associated with nonspecific NSAID use have been attributed to cyclooxygenase-1 (COX-1) inhibition.11Crofford L.J. Lipsky P.E. Brooks P. et al.Basic biology and clinical application of specific cyclooxygenase-2 inhibitors.Arthritis Rheum. 2000; 43: 4-13Crossref PubMed Scopus (273) Google Scholar In contrast, COX-2–specific inhibitors spare COX-1 at therapeutic doses, and their use has been associated with lower rates of endoscopic gastroduodenal ulcers and ulcer complications.12Silverstein F.E. Faich G. Goldstein J.L. 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: 1247-1255Crossref PubMed Scopus (3068) Google Scholar, 13Bombardier C. Laine L. Reicin A. et al.Comparison of upper gastrointestinal toxicity of rofecoxib and naproxen in patients with rheumatoid arthritis VIGOR Study Group.N Engl J Med. 2000; 343: 1520-1528Crossref PubMed Scopus (3691) Google Scholar, 14Deeks J.J. Smith L.A. Bradley M.D. Efficacy, tolerability, and upper gastrointestinal safety of celecoxib for treatment of osteoarthritis and rheumatoid arthritis systematic review of randomised controlled trials.BMJ. 2002; 325: 619-626Crossref PubMed Google Scholar, 15Goldstein J.L. Silverstein F.E. Agrawal N.M. et al.Reduced risk of upper gastrointestinal ulcer complications with celecoxib, a novel COX-2 inhibitor.Am J Gastroenterol. 2000; 95: 1681-1690Crossref PubMed Google Scholar, 16Sikes D.H. Agrawal N.M. Zhao W.W. et al.Incidence of gastroduodenal ulcers associated with valdecoxib compared with that of ibuprofen and diclofenac in patients with osteoarthritis.Eur J Gastroenterol Hepatol. 2002; 14: 1101-1111Crossref PubMed Scopus (110) Google Scholar, 17Goldstein J.L. Stenson W. Agrawal N. et al.Valdecoxib is associated with a significantly lower incidence of ulcer complications and symptomatic ulcers in arthritis patients compared to NSAID.Gastroenterology. 2002; 122 (abstr #T1403): A469PubMed Google Scholar, 18Langman M.J. Jensen D.M. Watson D.J. et al.Adverse upper gastrointestinal effects of rofecoxib compared with NSAIDs.JAMA. 1999; 282: 1929-1933Crossref PubMed Scopus (673) Google Scholar Because COX-1 is constitutively expressed in the human small bowel,19Kargman S. Charleson S. Cartwright M. et al.Characterization of prostaglandin G/H synthase 1 and 2 in rat, dog, monkey, and human gastrointestinal tracts.Gastroenterology. 1996; 111: 445-454Abstract Full Text Full Text PDF PubMed Scopus (452) Google Scholar we hypothesized that COX-2–specific inhibitors would also be less injurious than nonspecific NSAIDs in the small bowel. Although there have been no prospectively designed enteroscopic studies directly comparing these 2 drug classes with regard to small bowel mucosal injury, this hypothesis is supported by the greater rate of fecal 51Cr RBC loss reported in healthy subjects treated with the nonspecific NSAID ibuprofen compared to the COX-2–specific inhibitor rofecoxib.20Hunt R.H. Bowen B. Mortensen E.R. et al.A randomized trial measuring fecal blood loss after treatment with rofecoxib, ibuprofen, or placebo in healthy subjects.Am J Med. 2000; 109: 201-206Abstract Full Text Full Text PDF PubMed Scopus (79) Google Scholar Moreover, and of clinical relevance, in the Celecoxib Long-term Arthritis Safety Study, significantly more patients receiving therapeutic doses of ibuprofen or diclofenac had decreases in hematocrit ≥10% and/or decreases in hemoglobin ≥2 g/dL compared to those receiving a supratherapeutic dose of celecoxib (400 mg twice daily).12Silverstein F.E. Faich G. Goldstein J.L. 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: 1247-1255Crossref PubMed Scopus (3068) Google Scholar This difference persisted even when potential upper gastroduodenal ulcer complications were removed from the analysis, suggesting that this anemia might have resulted from occult blood loss from beyond the upper gastrointestinal tract. To better understand and quantify the differential small bowel injury pattern of nonspecific NSAIDs and COX-2–specific inhibitors, we designed a prospective, double-blind, randomized study to evaluate the incidence of small bowel lesions in healthy subjects treated with celecoxib compared to naproxen plus omeprazole or placebo by using VCE. We purposely elected to study the NSAID in combination with a proton pump inhibitor in an attempt to achieve balance among the treatment groups with regard to the upper gastrointestinal tract risk of gastroduodenal mucosal injury,21Chan F.K. Hung L.C. Suen B.Y. et al.Celecoxib versus diclofenac and omeprazole in reducing the risk of recurrent ulcer bleeding in patients with arthritis.N Engl J Med. 2002; 347: 2104-2110Crossref PubMed Scopus (487) Google Scholar, 22Hawkey C.J. Karrasch J.A. Szczepanski L. et al.Omeprazole compared with misoprostol for ulcers associated with nonsteroidal anti-inflammatory drugs OMNIUM study Group.N Engl J Med. 1998; 338: 727-734Crossref PubMed Scopus (859) Google Scholar, 23Yeomans N.D. Tulassay Z. Juhasz L. et al.A comparison of omeprazole with ranitidine for ulcers associated with nonsteroidal anti-inflammatory drugs ASTRONAUT Study Group.N Engl J Med. 1998; 338: 719-726Crossref PubMed Scopus (759) Google Scholar and because the protective effects of proton pump inhibitors would not be expected to occur within the remainder of the small bowel beyond the duodenum. Healthy adults (18–70 years of age) who had no evidence of either a mucosal break or blood in the small bowel at the baseline VCE and who had no clinically significant laboratory or physical examination abnormalities were eligible for randomization. Important criteria for exclusion were active gastrointestinal disease or history of gastrointestinal ulcers or bleeding; known or suspected complete or partial stenosis of the small intestine; prior gastric or intestinal surgery (resection); established delayed gastric emptying or diabetic gastroparesis; hemoglobin <10 g/dL (women) or 12 g/dL (men); positive fecal occult blood test results at screening; frequent (more than 3 times per week) use of aspirin or other NSAIDs within 2 weeks before screening; use of any known or suspected ulcerogenic medication (including aspirin or other NSAIDs) within 2 weeks before the baseline VCE (run-in period); and use of antiulcer medications (ie, sucralfate, antacids, histamine-2-receptor antagonists, proton pump inhibitors, and misoprostol) within 4 weeks before the baseline VCE, which corresponds to at least 2 weeks before the initial screening visit and entry into the run-in period. This was a prospective, randomized, double-blind, triple-dummy, placebo-controlled, multicenter study. After the initial screening visit, subjects underwent a 2-week run-in period, during which time alcohol consumption and use of any known or suspected ulcerogenic medications or antiulcer agents were prohibited. At the baseline visit, eligible subjects underwent a baseline VCE. If any mucosal break or blood was observed in the small bowel, or if small bowel visualization time was less than 2 hours, the subject was not eligible for randomization. Eligible subjects were subsequently randomly assigned to receive celecoxib (Celebrex; Pfizer, Inc, New York, NY) 200 mg twice daily, naproxen (Naprosyn; Roche Laboratories, Inc, Nutley, NJ) 500 mg twice daily plus omeprazole (Prilosec; AstraZeneca, Wilmington, DE) 20 mg once daily, or placebo for 2 weeks. After this 2-week treatment period, each subject underwent VCE for determination of the primary and secondary end points of the study. A computer-generated randomization technique was used to assign subjects in a 1:1:1 ratio. During the study, NSAIDs (including topical agents and aspirin), known or suspected ulcerogenic agents, analgesics (including opioids), and antiulcer medications were prohibited. Acetaminophen up to 4 g daily up to twice per week was permitted for general pain relief. The Institutional Review Boards at each of the 9 centers approved the protocol, and all subjects provided written informed consent at screening. VCE was performed by using the Given M2A video capsule system (Given Imaging Ltd).8Lewis B. Goldfarb N. The advent of capsule endoscopy a not-so-futuristic approach to obscure gastrointestinal bleeding.Aliment Pharmacol Ther. 2003; 17: 1085-1096Crossref PubMed Scopus (68) Google Scholar This system has been described in detail elsewhere.24Lewis B. Gostout C. The performance of capsule endoscopy.in: Halpern E. Jacob H. Atlas of capsule endoscopy. Rahash Printing Ltd, Haifa, Israel2002: 15-20Google Scholar Smoking was prohibited for 24 hours before the procedure was initiated, and only liquids were to be ingested beginning at lunchtime of the day before the procedure. No purges were given, and no prokinetic medications were administered. Subjects fasted 12 hours before swallowing the capsule. A light snack was permitted 4 hours after capsule ingestion. Data were collected for up to 8 hours after capsule ingestion. Repeat VCEs as a result of system failures (image loss or low signal) were permitted only if capsule excretion was confirmed. Each local investigator as well as each member of the Small Bowel Events Committee were required to attend a standardized training session on the use of the Given Diagnostic System and to pass a certification examination to ensure proficiency with the use of the Given Diagnostic System and the evaluation of the M2A video images. The investigator at each center reviewed and interpreted the baseline (pretreatment) video capsule images for determination of eligibility into the trial. If no mucosal break or blood was observed, the subject was randomized to the 2-week, blinded treatment phase and, after that, underwent the end-of-study VCE. The investigator at each center (who was blinded to patient treatment) then reviewed and evaluated the end-of-study VCEs and marked all suspected small bowel abnormalities. These investigators were instructed to mark any significant lesion, regardless of whether they believed the image met the criteria of a mucosal break. The members of the independent Small Bowel Events Committee (J.L.G., G.M.E., I.M.G., and B.L.), who were blinded to subject, site, and treatment, then reviewed and adjudicated to consensus these blinded, marked videos, according to the prespecified Small Bowel Lesion Assessment criteria (Table 1) for determination of the primary and secondary end points. Figure 1 describes the flow of the patients during the duration of the trial.Table 1Small Bowel Lesion Assessment TableCategoryDescription1Petechiae2Mucosal break without hemorrhage3Mucosal break with hemorrhage4Web/stricture without mucosal break and without hemorrhage5Web/stricture without mucosal break and with hemorrhage6Web/stricture with mucosal break and without hemorrhage7Web/stricture with mucosal break and hemorrhage8Presence of blood (hemorrhage) without visualized lesions9Other lesionsNOTE. There were 9 categories of small bowel lesions. Each lesion was evaluated and assigned a category. Mucosal breaks included any lesion that appeared as an erosion or ulcer, regardless of perceived size. Hemorrhage was defined as visible blood. Open table in a new tab NOTE. There were 9 categories of small bowel lesions. Each lesion was evaluated and assigned a category. Mucosal breaks included any lesion that appeared as an erosion or ulcer, regardless of perceived size. Hemorrhage was defined as visible blood. The primary end point was the mean number of small bowel mucosal breaks (defined as any break in the mucosa, as depicted by categories 2, 3, 6, and 7 in Table 1) per subject. There was no attempt to differentiate between ulcers and erosions, because video capsule technology cannot accurately assess either lesion size or depth.24Lewis B. Gostout C. The performance of capsule endoscopy.in: Halpern E. Jacob H. Atlas of capsule endoscopy. Rahash Printing Ltd, Haifa, Israel2002: 15-20Google Scholar There were 3 secondary end points: the percentage of subjects with 1 or more small bowel mucosal breaks; the total number of small bowel lesions with or without hemorrhage (categories 2, 3, 4, 5, 6, 7, and 9; Table 1); and the percentage of subjects with visible blood in the small bowel but without visualized lesions in small bowel (category 8, Table 1). We also performed 2 post hoc analyses. One was an analysis of the mean number of mucosal breaks among those subjects with at least 1 mucosal break. The second was an analysis of the distribution of small bowel mucosal breaks across tertiles, which divided the area between the duodenum and cecum into 3 equal parts, on the basis of each subject's specific small bowel transit time. Any video in which the cecum could not be identified was excluded from this specific analysis. Overall safety was assessed by physical examination, laboratory tests, and observed or reported adverse events. Descriptive statistics were used for subject demographics and clinical characteristics and for the outcome variables. The primary end point (number of small bowel mucosal breaks) and the secondary end point (number of small bowel lesions with or without hemorrhage) were not normally distributed, and therefore their rank scores were analyzed with Cochran-Mantel-Haenszel test, stratified by site. The nonparametric test was a Kruskal-Wallis test for the overall comparison and a Wilcoxon rank sum test for the pairwise comparison. The remaining secondary end points (percentage of subjects with mucosal breaks and percentage of subjects with visible blood without visualized lesions) also used Cochran-Mantel-Haenszel test, stratified by site, for overall and pairwise comparisons. A post hoc analysis with a nonparametric test (Friedman chi-square) was performed to assess whether within each treatment group a difference existed in the distribution of small bowel mucosal breaks across tertiles, and pairwise comparisons across treatments were analyzed by a Wilcoxon rank sum test. All analyses were performed on the modified intention-to-treat cohort (defined as all randomized subjects who completed 2 weeks of double-blind therapy and who had valid end-of-study VCEs). Because of current interest in the gastrointestinal mucosal effects of COX-2–specific inhibitors compared to nonspecific NSAIDs, we also performed a post hoc analysis of the mean number of mucosal breaks among the subgroup of subjects who had at least 1 mucosal break. The nonparametric test was a Kruskal-Wallis test for the overall comparison and a Wilcoxon two-sample test for the pairwise comparisons. All hypothesis tests were conducted by using a Type I error rate of 5%, and all tests were 2-tailed, with two-sided 95% confidence intervals provided. SAS Version 8.2 (SAS Institute, Inc, Cary, NC) was used for all analyses. Between September 2002 and March 2003, we screened 462 healthy adult subjects from 9 centers (8 centers in the United States and 1 center in Israel). A total of 40 subjects did not meet general inclusion and/or exclusion criteria, and in 9 other subjects the baseline videos were technical failures. Of the remaining 413 subjects, 57 (13.8%) were found to have lesions at the baseline VCE and were not eligible for randomization. The remaining 356 subjects underwent randomization: 120 to celecoxib, 118 to naproxen plus omeprazole, and 118 to placebo (Figure 1). Adverse events resulted in the withdrawal of 4 celecoxib subjects and 1 naproxen/omeprazole subject. A total of 351 subjects completed the study and underwent a final video capsule examination. Technical failures of the video occurred in 12 subjects. The remaining 339 subjects constituted the modified intention-to-treat cohort. These subjects (as well as all randomized subjects) had similar baseline characteristics (Table 2).Table 2Baseline Demographic Characteristics of the SubjectsCharacteristicAll randomized subjectsModified intention-to-treat cohortCelecoxib (n = 120)Naproxen/omeprazole (n = 118)Placebo (n = 118)Celecoxib (n = 115)Naproxen/omeprazole (n = 111)Placebo (n = 113)Age (y)33.9 ± 10.632.8 ± 10.133.6 ± 11.033.9 ± 10.833.1 ± 10.333.6 ± 11.0Gender, no. of subjects (%)Male51 (42)47 (40)42 (36)50 (43)46 (41)39 (35)Female69 (58)71 (60)76 (64)65 (57)65 (59)74 (65)Race, no. of subjects (%)White93 (78)92 (78)85 (72)88 (77)87 (78)84 (74)Black9 (8)12 (10)12 (10)9 (8)12 (11)11 (10)Asian10 (8)8 (7)16 (14)10 (9)7 (6)15 (13)OtheraOther includes other racial groups and where the race was not allowed to be asked.8 (7)6 (5)5 (4)8 (7)5 (5)3 (3)NOTE. Plus-minus values are means ± standard deviations.a Other includes other racial groups and where the race was not allowed to be asked. Open table in a new tab NOTE. Plus-minus values are means ± standard deviations. As shown in Table 3, the mean number of small bowel mucosal breaks per subject was significantly higher in the naproxen/omeprazole group (2.99) compared to the celecoxib (0.32) and placebo groups (0.11) (P < .001 for both comparisons). The difference between celecoxib and placebo was also significant (P = .04). A representative lesion meeting the criteria for category 2 (mucosal break without hemorrhage) is shown in Figure 2.Table 3Small Bowel Lesions After Two Weeks of TreatmentEnd pointCelecoxib (n = 115)Naproxen/omeprazole (n = 111)Placebo (n = 113)P valueaP value based on Wilcoxon rank sum method for the end points relating to number of lesions and on Cochran-Mantel-Haenszel for end points relating to percentage of subjects with small bowel lesions or blood.Primary end pointNumber of small bowel mucosal breaks per subject0.32 ± 0.102.99 ± 0.510.11 ± 0.04<.001bCelecoxib vs naproxen/omeprazole.cNaproxen/omeprazole vs placebo..04dCelecoxib vs placebo.Secondary end pointsSubjects with small bowel mucosal breaks, no. of subjects (%)18 (16)61 (55)8 (7)<.001bCelecoxib vs naproxen/omeprazole.cNaproxen/omeprazole vs placebo..04dCelecoxib vs placebo.Number of small bowel lesions with or without hemorrhage per subject0.50 ± 0.194.18 ± 0.720.14 ± 0.05<.001bCelecoxib vs naproxen/omeprazole.cNaproxen/omeprazole vs placebo..09dCelecoxib vs placebo.Subjects with blood in the small bowel, without visualized lesions, no. of subjects (%)8 (7)9 (8)1 (1).75bCelecoxib vs naproxen/omeprazole..007cNaproxen/omeprazole vs placebo..01dCelecoxib vs placebo.NOTE. Modified intention-to-treat cohort. Plus-minus values are unadjusted means ± standard errors.a P value based on Wilcoxon rank sum method for the end points relating to number of lesions and on Cochran-Mantel-Haenszel for end points relating to percentage of subjects with small bowel lesions or blood.b Celecoxib vs naproxen/omeprazole.c Naproxen/omeprazole vs placebo.d Celecoxib vs placebo. Open table in a new tab NOTE. Modified intention-to-treat cohort. Plus-minus values are unadjusted means ± standard errors. A significantly higher percentage of subjects in the naproxen/omeprazole group (55%) had mucosal breaks compared to those in the celecoxib (16%) and placebo (7%) groups (P < .001 for both comparisons); the difference between celecoxib and placebo was also significant (P = .04). The number of small bowel lesions with or without hemorrhage was significantly higher in the naproxen/omeprazole group (4.18) than in the celecoxib (0.50) and placebo groups (0.14) (P < .001 for both comparisons); there was no difference between celecoxib and placebo (P = .09) (Table 3). The percentage of subjects with blood (without visualized lesions) in the small bowel (category 8, Table 1) was similar for celecoxib (7%) and naproxen/omeprazole (8%), and each of these treatments was significantly different from placebo (1%) (P ≤ .01 for both comparisons) (Table 3). When including all subjects with hemorrhage (categories 3,5,7, and 8; Table 1), the percentages remained the same for celecoxib and placebo and increased to 14% for naproxen/omeprazole. Only the difference between the active treatment groups and placebo, however, was significant (P ≤ .01 for both comparisons). In a post hoc analysis, we calculated the mean number of mucosal breaks among those subjects who developed at least 1 mucosal break; there was a significant difference between celecoxib and naproxen/omeprazole. The mean number (± standard error) of mucosal breaks for the celecoxib group was 2.06 ± 0.48, which was significantly lower (P = .004) than the mean observed in the naproxen/omeprazole group (5.44 ± 0.79), but not significantly different from that observed in the placebo group (1.50 ± 0.33) (P = .58 for celecoxib vs placebo). The difference between naproxen/omeprazole and placebo was also significant (P = .01). The distribution of subjects with 1 or more mucosal breaks is shown in Figure 3. An additional post hoc analysis found that there was no statistically significant difference in the distribution of mucosal breaks throughout tertiles of the small bowel within each treatment group (Table 4). Furthermore, the pattern of the treatment difference for the comparison of naproxen/omeprazole to celecoxib and placebo was consistent within each tertile and similar to the overall results (which did not control for tertile location). However, in each tertile, the difference between celecoxib and placebo was no longer statistically significant, possibly because of fewer small bowel mucosal breaks in each tertile.Table 4Number of Small Bowel Mucosal Breaks per Tertile of the Small Bowel LengthTertileCelecoxib (n = 115)Naproxen/omeprazole (n = 111)Placebo (n = 113)P valueaP value based on Wilcoxon rank sum test.First0.13 ± 0.041.28 ± 0.280.07 ± 0.04<.001bCelecoxib vs naproxen/omeprazole.cNaproxen/omeprazole vs placebo..10dCelecoxib vs placebo.Second0.06 ± 0.031.02 ± 0.230.01 ± 0.01<.001bCelecoxib vs naproxen/omeprazole.cNaproxen/omeprazole vs placebo..14dCelecoxib vs placebo.Third0.13 ± 0.080.68 ± 0.170.02 ± 0.01<.001bCelecoxib vs naproxen/omeprazole.cNaproxen/omeprazole vs placebo.0.17dCelecoxib vs placebo.PeP value based on Friedman χ2 test..08.23.37NOTE. Modified intention-to-treat cohort. Plus-minus values are unadjusted means ± standard errors.a P value based on Wilcoxon rank sum test.b Celecoxib vs naproxen/omeprazole.c Naproxen/omeprazole vs placebo.d Celecoxib vs placebo.e P value based on Friedman χ2 test. Open table in a new tab NOTE. Modified intention-to-treat cohort. Plus-minus values are unadjusted means ± standard errors. All treatments were well tolerated, and there were no significant