Gastrointestinal damage induced by nonsteroidal

Transcription

1 A Randomized Trial Measuring Fecal Blood Loss after Treatment with Rofecoxib, Ibuprofen, or Placebo in Healthy Subjects Richard H. Hunt, MD, Barry Bowen, MScPhm, Eric R. Mortensen, MD, PhD, Thomas J. Simon, MD, Cindy James, RN, Angeline Cagliola, BBA, Hui Quan, PhD, James A. Bolognese, MStat PURPOSE: Gastrointestinal microbleeding, as assessed by the measurement of 51 chromium-labeled red blood cells, is a marker of the mucosal injury associated with the use of nonsteroidal anti-inflammatory drugs. This study tested the hypotheses that cyclooxygenase-2 specific inhibition with rofecoxib would cause less fecal blood loss than a therapeutic dose of ibuprofen and would be equivalent to placebo. SUBJECTS AND METHODS: In this randomized, doubleblind group study, gastrointestinal blood loss was assessed by measurement of fecal 51 chromium radioactivity during a 1-week placebo baseline period and during 4 weeks of treatment with rofecoxib (25 mg or 50 mg once daily), ibuprofen (800 mg three times daily), or placebo in 67 healthy subjects. Gastrointestinal blood loss during treatment weeks 2 to 4 (versus the baseline period) was expressed as the geometric mean ratio of fecal radioactivity in weeks 2 to 4 compared with baseline. RESULTS: Ibuprofen caused significantly (P 0.001) greater gastrointestinal blood loss (geometric mean ratio of 5.2, 95% confidence interval [CI]: 4.2 to 6.3) than the 25-mg dose of rofecoxib (2.6, 95% CI: 2.2 to 3.1), the 50-mg dose of rofecoxib (2.6, 95% CI: 2.2 to 3.0), or placebo (2.1, 95% CI: 1.8 to 2.5). In contrast, gastrointestinal blood loss with both doses of rofecoxib were equivalent to placebo by a predetermined clinical similarity bound. CONCLUSIONS: In healthy subjects, treatment with rofecoxib, at 2 to 4 times the doses that are currently recommended for the treatment of patients with osteoarthritis, produced significantly less fecal blood loss than a therapeutic dose of ibuprofen and was equivalent to placebo. Am J Med. 2000;109: by Excerpta Medica, Inc. Gastrointestinal damage induced by nonsteroidal anti-inflammatory drugs (NSAIDs) is the most common drug-related serious adverse event for patients in industrialized nations (1 3). All conventional NSAIDs inhibit both cyclooxygenase (COX) isoenzymes involved in prostaglandin synthesis: COX-1 and COX-2 (4). COX-1 is constitutively expressed and generates prostaglandins believed to be involved in gastrointestinal mucosal protection (5), whereas COX-2 is primarily induced to generate prostaglandins that are associated with the mediation of inflammation and pain (6). Compounds that provide specific inhibition of COX-2 may provide the anti-inflammatory effects of NSAIDs, which are mediated essentially through inhibition of COX-2 (7), while avoiding the deleterious effects on the gastrointestinal tract that are mediated primarily by inhibition of COX-1 (4,8). Gastrointestinal microbleeding can be assessed by the From McMaster University (RHH, BB, CJ), Hamilton, Ontario, Canada; the University of Toronto (BB), Toronto, Ontario, Canada; and Merck Research Laboratories (ERM, TJS, AC, HQ, JAB), West Point, Pennsylvania. Supported by a grant from Merck Research Laboratories, West Point, Pennsylvania. Requests for reprints should be addressed to Eric Mortensen, MD, PhD, Merck Research Laboratories, PO Box 4, BL 1 4, West Point, Pennsylvania Manuscript submitted December 16, 1999, and accepted in revised form May 3, measurement of 51 chromium-labeled red blood cells in fecal material, which permits a precise, continuous, and quantitative assessment of total gastrointestinal tract blood loss (9). For conventional NSAIDs, the magnitude of drug-induced fecal blood loss is dose dependent (10 13), correlates with the severity of endoscopically observed gastric lesions and mucosal ulcerations (9,14,15), and may produce daily blood loss of clinical importance (16). This 5-week, randomized, double-blind study tested the hypotheses that specific inhibition of COX-2 with rofecoxib (17,18), administered at doses 2 to 4 times the doses currently recommended for the treatment of patients with osteoarthritis (19,20), would cause significantly less fecal blood loss than a therapeutically similar dose of ibuprofen (800 mg three times daily) and would be equivalent to placebo. SUBJECTS AND METHODS Subjects Healthy men (aged 18 to 60 years) who were within 20% of ideal body weight were eligible to enroll if they had a history of regular bowel habits (at least one stool per day) and their fecal red blood cell loss during a pretreatment baseline week was within normal limits. Subjects were excluded if they had a prior history of gastrointestinal or medical disease, or gastrointestinal surgery; a history or 2000 by Excerpta Medica, Inc /00/$ see front matter 201 All rights reserved. PII S (00)

2 evidence of nasal, oral, or rectal bleeding; a positive fecal occult blood test at the prestudy visit; or hemorrhoids on anorectal examination. Subjects were also excluded if they used any analgesic or sedative-hypnotic medications, or if they had used antacids, histamine 2 -receptor antagonists, proton pump inhibitors, or misoprostol more than twice during the 4 weeks before the study, or any NSAIDs in the previous 2 weeks. Additional exclusions included excessive alcohol use, cigarette smoking, daily consumption of more than four 8-oz cups of caffeinnated beverages, or allergy or intolerance to NSAIDs. The study was approved by the institutional review board of Chedoke-McMaster Hospitals; all subjects gave written informed consent before enrollment. Study Design and Treatment At the prestudy visit (up to 14 days before radiolabeling), all subjects underwent physical examination, including anoscopy, laboratory evaluation, electrocardiography, and fecal occult blood testing. One day before admission to the clinical investigation unit (study day 8), subjects received an injection of an aliquot of their own red blood cells labeled with 300 Ci of 51 chromium. They then entered a 1-week, single-blind placebo treatment period to establish baseline fecal blood loss. Subjects were not enrolled if fecal blood loss was 2 ml on any single day or if their average fecal blood loss during the first 6 days of the week was 1.5 ml per day. A computerized randomization scheme randomly assigned eligible subjects to treatment for 28 days with 25 mg of rofecoxib (Vioxx; Merck & Co., Inc., West Point, Pennsylvania) once daily, 50 mg of rofecoxib once daily, 800 mg of ibuprofen three times daily, or placebo. Fewer patients were assigned to ibuprofen because of the larger effects anticipated with that drug. Treatment began on day 0. Blinding was maintained by using a matching placebo for each study medication; all study personnel remained blinded to the subject s treatment until after study completion. The study staff monitored daily administration of study medication to ensure protocol compliance throughout the trial (35 days). All subjects lived in a dormitory environment, where medication administration, diet, laboratory values, daily activity, and stool collection were closely monitored. Procedures scheduled for the last day of therapy (study day 28) were performed at the time of withdrawal for any subject who discontinued the study early. Adverse events (intensity, duration, seriousness, relation to study drug, and outcome) were collected throughout the study and at the follow-up visit. Fecal Blood Loss Measurement Mean daily fecal blood loss was measured with the 51 chromium-labeled red blood cell method (9,21). Twenty milliliters of venous blood were drawn from an antecubital vein of each subject and transferred into a sterile vial containing 3 ml of acid-citrate-dextrose solution (USP); 300 Ci of sodium chromate 51 chromium (Merck Frosst Laboratories, Montreal, Canada) was added, and the solution was incubated at room temperature for 30 minutes. The blood was gently centrifuged, and the red blood cells were then resuspended in 4 ml sodium chloride 0.9% injection USP. Each syringe was weighed, and the amount of radioactivity was measured in an ionization chamber. This suspension was then slowly injected into each subject. Administration of 300 Ci is estimated to result in a radiation dose of 0.8 msv, equivalent to approximately 14% of the annual limit recommended for radiation workers by the International Commission on Radiation Protection. Samples and Calculations All stool samples were collected each day, weighed, and frozen. Blood and stool samples were placed in plastic containers (Bayer Tissue-tek; Bayer, Etobicoke, Ontario, Canada) for counting. Estimates of red blood cell survival and blood-specific activity were made from serial measurements of 51 chromium radioactivity in aliquots of blood samples obtained at intervals after the intravenous injection of 51 chromium-labeled red blood cells on days 7, 4, 1, 4, 7, 11, 14, 18, 21, 25, and 28. The samples were counted by placing each sample on the top of a shielded 5-inch sodium iodide crystal doped with thallium. The signal was amplified and then analyzed; the integrated counts for the background and the sample were recorded. Because of the need to ensure that the system was stable, a 137 cesium standard was analyzed at each counting session to calibrate the instrument and to determine the amount of drift of the spectrum. The calculation of fecal blood loss depended on the red blood cell specific activity. Daily 51 chromium blood-specific activity was interpolated by calculation of each subject s 51 chromium red blood cell survival curve. Statistical Analysis Treatment with ibuprofen and several other NSAIDs has been associated with a more than twofold increase in fecal blood loss (9,22,23). The predetermined primary period for analysis in this study was the second to the fourth week of drug treatment to exclude the possibility that acute mucosal injury with ibuprofen during the first week would increase the apparent rate of blood loss in this treatment group. The primary study hypothesis was that compared with ibuprofen (800 mg three times daily), 51 chromium fecal red blood cell loss during a 4-week period of treatment would be lower with rofecoxib (25 mg once daily). Gastrointestinal blood loss was expressed as the geometric mean ratio of fecal radioactivity during weeks 2 to 4 of treatment compared with the pretreatment baseline. For the comparison of ibuprofen with rofecoxib, the sample size provided 89% power to detect a ratio (ibuprofen to rofecoxib) of 2 or more, based on a 202 August 15, 2000 THE AMERICAN JOURNAL OF MEDICINE Volume 109

3 Table 1. Baseline Characteristics of the Subjects, by Treatment Group* Placebo (n 17) Rofecoxib 25 mg (n 18) Rofecoxib 50 mg (n 19) Ibuprofen 2400 mg (n 13) Number (Percent) or Mean SD (Range) Age (years) 22 2 (19 25) 24 4 (20 37) 22 2 (19 24) 23 2 (21 27) Race Asian 2 (12) 1 (6) 0 1 (8) Black 1 (6) 0 2 (11) 0 White 14 (82) 17 (94) 17 (90) 12 (92) * All subjects were male. Two subjects (1 assigned to 25 mg of rofecoxib and 1 assigned to 50 mg of rofecoxib) are included in this table even though they were removed from the study because of fecal blood loss 2mLon a single day during the baseline pre-treatment period. log-scaled SD of 0.5 ml per day, estimated from prior data. For comparisons with placebo, a predetermined similarity bound was established, based on a previous aspirincontrolled study that demonstrated no difference between the lowest labeled dose of nabumetone and placebo (24). Data from that study showed a geometric mean ratio for nabumetone to placebo of 1.2, with an upper limit of the 90% confidence interval of approximately 1.9. We adopted a more restrictive value as the predetermined similarity bound. Rofecoxib would be considered equivalent to placebo if the one-sided upper limit of the 95% confidence interval (equivalent to a two-sided 90% confidence interval) for the rofecoxib to placebo geometric mean ratio for daily fecal blood loss was 1.7. (Because a protective effect of rofecoxib was not hypothesized, a one-sided confidence interval was used.) Both upper and lower (one-sided 95% or two-sided 90%) confidence intervals for the active treatment to placebo ratios are presented. We had 90% power to demonstrate that the upper limit of the one-sided 95% confidence interval for the rofecoxib to placebo ratio would be 1.7, if the true ratio was 1 (at an alpha of 0.05). Daily fecal blood loss data followed a normal distribution more closely in the log scale (P 0.11) than in the original scale (P 0.001). Therefore, an analysis of covariance model with terms for baseline and treatment was used to analyze change from baseline in a log scale for daily fecal blood loss. Betweentreatment comparisons were based on the t test from the analysis of covariance model. An intention-to-treat approach was the primary method of analysis for all statistical comparisons. A perprotocol analysis was also performed for the primary endpoint of the ratio of weeks 2 to 4 compared with baseline. Five subjects were classified as protocol violators during the study and were excluded from the per-protocol analysis, the results of which were similar to the intention-to-treat analysis. The geometric mean daily fecal blood loss, with a 84% confidence interval, was plotted. This confidence interval was chosen because nonoverlapping 84% confidence intervals for two treatment groups indicate a between-treatment difference of about P 0.05 with a two-sample t test) (25). The geometric mean (the antilog of the arithmetic mean of the log of the data) was used because this is the most appropriate measure of central tendency for data with a log-normal (skewed) distribution, in which the arithmetic mean is substantially affected by a small number of outlying values. RESULTS Subjects A total of 67 subjects were enrolled, and 65 completed the study. Two subjects, 1 each in the rofecoxib 25-mg and 50-mg treatment groups, were removed from the study because of a protocol violation that occurred after they had taken one dose of blinded test medication: each subject had a fecal blood loss result 2 ml on a single day during the pretreatment baseline week. The treatment groups were similar demographically (Table 1). Fecal Blood Loss Fecal blood loss associated with ibuprofen was significantly (P 0.001) greater than that with placebo throughout the 4 weeks of treatment (Figure 1). In contrast, a comparison of the fecal blood loss associated with both doses of rofecoxib, 25 mg or 50 mg, and with placebo treatment showed no significant differences. Fecal blood loss with ibuprofen was significantly (P 0.001) greater than with rofecoxib. No differences were observed between the rofecoxib 25-mg and 50-mg treatment groups. 51 Chromium activity in the stool increased continuously throughout the study in the placebo treatment group and with both doses of rofecoxib. Compared with baseline, fecal blood loss in the ibuprofen group was significantly greater (P 0.001) than in the other three treatment groups (Table 2, Figure 2). For example, fecal blood loss was 2.4-fold greater in the ibuprofen treatment group than in the placebo group. These differences were seen during the entire 4-week treatment period (Figure 1). When compared with placebo, the one-sided upper August 15, 2000 THE AMERICAN JOURNAL OF MEDICINE Volume

4 Figure 1. Geometric mean daily fecal blood loss with 84% confidence intervals. Nonoverlapping intervals approximately correspond to P 0.05 using a pairwise t test. limit of the 95% confidence interval (the upper limit of the two-sided 90% confidence interval) of 1.5 for the geometric mean ratio for daily fecal blood loss in the rofecoxib treatment groups was less than the predetermined clinical similarity bound of 1.7 (Table 2). In contrast, even the lower limit of the 90% confidence interval for the ratio of ibuprofen to placebo values of 1.9 was significantly greater than the clinical comparability bound. A comparison of rofecoxib versus placebo also showed no significant differences between these groups during any week within the treatment period. Figure 2. Weeks 2 to 4/baseline least-squares geometric mean ratio for daily fecal blood loss, with 95% confidence intervals. Adverse Events No subject was withdrawn from the study due to an adverse event, and there were no serious clinical or laboratory adverse events. Nine of the 17 subjects in the placebo group, 6 of the 18 subjects in the rofecoxib 25-mg group, 8 of the 19 subjects in the rofecoxib 50-mg group, and 5 of the 13 subjects in the ibuprofen group had adverse events during the study. The most frequent adverse event was headache, reported by a total of 6 subjects: 1 in the placebo group, 3 in the rofecoxib 50-mg group, and 2 in the ibuprofen group. There were no clinically relevant differences among treatment groups in the frequency of adverse events. Tabl 2. Fecal Blood Loss (ml per Day) by Treatment Group (Intention-to-Treat Analysis) Treatment N Baseline Geometric Mean Treatment Geometric Mean Treatment/Baseline Ratio (95% Confidence Interval)* Comparison of Ratio for Active Drug with Ratio for Placebo (90% Confidence Interval) Weeks 2 4 Placebo ( ) Rofecoxib 25 mg ( ) 1.2 ( ) Rofecoxib 50 mg ( ) 1.2 ( ) Ibuprofen 2,400 mg ( ) 2.4 ( ) Weeks 1 4 Placebo ( ) Rofecoxib 25 mg ( ) 1.2 ( ) Rofecoxib 50 mg ( ) 1.2 ( ) Ibuprofen 2,400 mg ( ) 2.4 ( ) * The mean was calculated for the change from baseline in a log scale, then transformed back to obtain the ratio of the geometric means in the original scale. Ratio of the treatment/baseline geometric mean ratios in the adjacent column. 204 August 15, 2000 THE AMERICAN JOURNAL OF MEDICINE Volume 109

5 DISCUSSION Our study in healthy subjects confirmed the hypothesis that 4 weeks of treatment with rofecoxib, a specific inhibitor of COX-2, was associated with less fecal red blood cell loss than a therapeutic dose of ibuprofen (19,20). Treatment with 25 mg or 50 mg of rofecoxib once daily produced fecal blood losses that were similar to placebo treatment. The amount of 51 chromium-labeled fecal red blood cell loss that is associated with the use of NSAIDs is related to the incidence of gastroduodenal ulcers and erosions induced by these agents (13 16,26). Prolonged drug-induced blood loss may contribute to negative iron balance and anemia in patients who require chronic treatment (11). The hypothesis that gastrointestinal microbleeding is related to the decrease in gastrointestinal mucosal COX-1 activity caused by nonspecific cyclooxygenase inhibitors is supported by this study s demonstration that rofecoxib, a specific inhibitor of COX-2, does not increase fecal blood loss. The hypothesis is also supported by the observation that aspirin-induced fecal blood loss in patients with arthritis was decreased by the oral administration of prostaglandin analogs (27). Ibuprofen was chosen as the conventional NSAID in this study because it is commonly used for the treatment of osteoarthritis and was considered to be one of the safest NSAIDs in a large meta-analysis (28). Although aspirin (29,30) or indomethacin (22) are frequently used in gastrointestinal microbleeding studies, previous studies of 51 chromium-labeled fecal blood loss have shown that gastrointestinal microbleeding associated with a variety of NSAIDs is less than that associated with therapeutic doses of these two drugs (9,22,24,29). Ibuprofen was therefore also chosen to provide a more conservative estimate of the amount of fecal blood loss to be anticipated with treatment using conventional NSAIDs. The doses of rofecoxib given in this study were 2 to 4 times greater than the doses currently recommended for the treatment of patients with osteoarthritis, and were similar in therapeutic efficacy to ibuprofen (800 mg three times daily) (19,20). Fecal blood losses in the rofecoxib treatment groups were virtually identical. There was no evidence of a doserelated increase, as has been reported for conventional NSAIDs (10 12). Although treatment with rofecoxib was associated with a small increase in fecal blood loss compared with placebo, the predetermined criteria for equivalence to placebo were met for both rofecoxib doses. Furthermore, the apparent increase in labeled blood loss with rofecoxib of approximately 0.1 ml per day was within the range of variability of fecal blood loss (30). Fecal blood loss increased with time in the placebo group. This observation is consistent with previous findings (24). The reason for this increase is not known, but may be related to biliary excretion of small quantities of radiochromium from senescent 51 chromium-labeled erythrocytes (31,32). Our findings are consistent with other studies that have suggested that rofecoxib may have better gastrointestinal tolerability compared with nonselective NSAIDs (10,33). Rofecoxib inhibits COX-2 activity with greater than 800-fold selectivity compared with COX-1 in in vitro assays, and did not significantly inhibit COX-1 activity in ex vivo human whole blood assays in subjects treated with up to 1,000 mg of the drug (18). In a 1-week endoscopic study of gastroduodenal injury, 250 mg of rofecoxib daily (10 to 20 times the dose for treatment of osteoarthritis) produced significantly less mucosal injury than 800 mg of ibuprofen three times daily or 650 mg of aspirin four times daily (33). This study found no significant difference between the incidence of gastroduodenal mucosal injury induced by placebo and by high-dose rofecoxib. In addition, a crossover study to evaluate intestinal permeability also showed that rofecoxib (25 or 50 mg daily) was equivalent to placebo and superior to indomethacin (150 mg) (34). The assessment of gastrointestinal microbleeding using 51 chromium-labeled fecal red blood cell loss offers some advantages for investigating gastrointestinal damage. For example, endoscopic techniques cannot assess the condition of the entire length of the gastrointestinal tract (9), and measurements of intestinal permeability do not assess gastric or colonic mucosal injury (35). The assessment of fecal blood loss through the use of semiquantitative tests, including colorimetric and immunologic methods, is effective for the detection of bleeding in the lower gastrointestinal tract (36,37), but these methods are of limited use in the evaluation of upper gastrointestinal bleeding (24,38). In summary, in healthy subjects, treatment for 28 days with rofecoxib (25 mg or 50 mg daily), at doses 2 to 4 times those that are currently recommended for patients with osteoarthritis, was associated with significantly less fecal blood loss than ibuprofen (800 mg three times daily) and was equivalent to placebo. These results are consistent with the hypothesis that treatment with rofecoxib, a specific inhibitor of COX-2, is not associated with the gastrointestinal mucosal injury associated with treatment using nonselective (conventional) nonsteroidal anti-inflammatory drugs. ACKNOWLEDGMENT We thank Seymour Fein, MD, for help with protocol design and Christopher Lines, PhD, for assistance in preparing the manuscript. August 15, 2000 THE AMERICAN JOURNAL OF MEDICINE Volume

6 REFERENCES 1. Fries JF. NSAID gastropathy: the second most deadly rheumatic disease? Epidemiology and risk appraisal. J Rheumatol. 1991; 18(suppl 28): Fries JF, Miller SR, Spitz PW, et al. Toward an epidemiology of gastropathy associated with nonsteroidal antiinflammatory drug use. Gastroenterology. 1989;96: Roth SH. Nonsteroidal anti-inflammatory drug gastropathy: we started it can we stop it? Arch Intern Med. 1986;146: Cryer B. Nonsteroidal anti-inflammatory drugs and gastrointestinal disease. In: Feldman M, Scharschmidt BF, Sleisenger MH, eds. Sleisinger and Fordtran s Gastrointestinal and Liver Disease. 6th ed. Philadelphia: Saunders; 1998: Meade EA, Smith WL, DeWitt DL. Differential inhibition of prostaglandin endoperoxide synthase (cyclooxygenase) isozymes by aspirin and other non-steroidal anti-inflammatory drugs. J Biol Chem. 1993;268: Fosslien E. Adverse effects of nonsteroidal anti-inflammatory drugs on the gastrointestinal system. Ann Clin Lab Sci. 1998;28: Mitchell JA, Akarasereenont P, Thiemermann C, et al. Selectivity of nonsteroidal antiinflammatory drugs as inhibitors of constitutive and inducible cyclooxygenase. Proc Natl Acad Sci USA. 1994;90: Hawkey, CJ. COX-2 inhibitors. Lancet. 1999;353: Lussier A, Arsenault A, Varady J, et al. The use of a 51 Cr technique to detect gastrointestinal microbleeding associated with nonsteroidal antiinflammatory drugs. Semin Arthritis Rheum. 1988;17(suppl 2): Chan C-C, Boyce S, Brideau C, et al. Pharmacology of a selective cyclooxygenase-2 inhibitor L-745,337: a novel nonsteroidal antiinflammatory agent with an ulcerogenic sparing effect in rat and nonhuman primate stomach. J Pharmacol Exp Ther. 1995;274: Pierson RN, Holt PR, Watson RM, Keating RP. Aspirin and gastrointestinal bleeding. Chromate 51 blood loss studies. Am J Med. 1961; 31: Menasse-Gdynia R, Krupp P. Quantitative measurement of gastrointestinal bleeding in rats: the effect of nonsteroidal anti-inflammatory drugs. Toxicol Appl Pharmacol. 1974;29: Aabakken L, Dybdahl JH, Larsen S, et al. A double-blind comparison of gastrointestinal effects of ibuprofen standard and ibuprofen sustained release assessed by means of endoscopy and 51 Cr-labelled erythrocytes. Scand J Rheumatol. 1989;18: Ridolfo AS, Crabtree RE, Johnson DW, Rockhold FW. Gastrointestinal microbleeding: comparisons between benoxaprofen and other nonsteroidal antiinflammatory agents. J Rheumatol. 1980;7(suppl 6): Loebl DH, Craig RM, Culic DD, et al. Gastrointestinal blood loss. Effect of aspirin, fenoprofen and acetaminophen in rheumatoid arthritis as determined by sequential gastroscopy and radioactive fecal markers. JAMA. 1977;237: Lanza FL, Arnold JD. Etodolac, a new nonsteroidal antiinflammatory drug: gastrointestinal microbleeding and endoscopic studies. Clin Rheumatol. 1989;8(suppl 1): Chan C, Boyce S, Brideau C, et al. Rofecoxib [Vioxx, MK-0966; 4-(4 -methylsulfonylphenyl)-3-phenyl-2-(5h)-furanone]: a potent and orally active cyclooxygenase-2 inhibitor. Pharmacological and biochemical profiles. J Pharmacol Exp Ther. 1999;290: Elrich EW, Dallob A, De Lepeleire I, et al. Characterization of rofecoxib as a cyclooxygenase-2 isoform inhibitor and demonstration of analgesia in the dental pain model. Clin Pharmacol Ther. 1999; 65: Saag K, Fisher C, McKay J, et al. MK-0966, a specific COX-2 inhibitor has clinical efficacy comparable to ibuprofen in the treatment of knee, and hip osteoarthritis (OA) in a 6-week controlled clinical trial. Arth Rheum. 1998;41:S Day R, Morrison B, Luza A, et al. A randomized trial of the efficacy and safety of the COX-2 specific inhibitor, rofecoxib, and ibuprofen in patients with osteoarthritis. Arch Intern Med. 2000;160: Mollison PL, Veall N. The use of the isotope 51 Cr as a label for red cells. Br J Haematol. 1955;1: Salom IL, Jacob G, Jallard N, et al. Gastrointestinal microbleeding associated with the use of etodolac, ibuprofen, indomethacin, and naproxen in normal males. J Clin Pharmacol. 1984;24: Patoia L, Clausi G, Farroni F, et al. Comparison of faecal blood loss, upper gastrointestinal mucosal integrity and symptoms after piroxicam beta-cyclodextrin, prioxicam and placebo administration. Eur J Clin Pharmacol. 1989;36: Lussier A, LeBel E. Radiochromium (Chromium-51) evaluation of gastrointestinal blood loss associated with placebo, aspirin, and nabumetone. Am J Med. 1987;83(suppl 4B): Nelson, L. S. Evaluating overlapping confidence intervals. J Qual Technol. 1989;21: Chernish SM, Rosenak BD, Brunelle RL, Crabtree R. Comparison of gastrointestinal effects of aspirin and fenoprofen. A double blind crossover study. Arth Rheum. 1979;22: Ryan JR, Vargas JR, Clay GA, McMahon FG. Role of misoprostol in reducing aspirin-induced gastrointestinal blood loss in arthritic patients. Am J Med. 1987;83(suppl 1A): Henry D, Lim LL-Y, Rodriguez LAG, et al. Variability in risk of gastrointestinal complications with individual non-steroidal antiinflammatory drugs: results of a collaborative meta-analysis. BMJ. 1996;312: Uthgenannt H, Timm H. The effect of Voltaren on gastrointestinal bleeding. [Translation.] Muenchen Med Wochenschr. 1975;117: Lussier A, Tetreault L, LeBel E. Comparative study of gastrointestinal microbleeding caused by aspirin, fenbufen and placebo. Am J Med. 1983: Rainsford KD. The biochemical pathology of aspirin-induced gastric damage. Agents Act. 1975;5: DeMedicis R, LeBel E, Rioux A, et al. Biliary excretion of radiochromium. Am J Med. 1988;85: Lanza F, Simon T, Quan H, et al. Selective inhibition of cyclooxygenase-2 (COX-2) with MK-0966 (250 mg Q.D.) is associated with less gastroduodenal damage than aspirin (ASA) 650 mg Q.I.D or ibuprofen (IBU) 800 mg T.I.D. Aliment Pharmacol Ther. 1999;13: Bjarnason I, Sigthorsson G, Crane R, et al. COX-2 specific inhibition with MK or 50 mg q.d. does not increase intestinal permeability. A controlled study with placebo (pbo) and indomethacin 50 mg t.i.d.(indo). Am J Gastroenterol. 1998;93: Bjarnason, I, Zanelli G, Smith T, et al. The pathogenesis of nonsteroidal anti-inflammatory drug-induced small intestinal inflammation. Scand J Rheumatol. 1987;64(suppl): Hardcastle JD, Armitage NC. Early diagnosis of colorectal cancer: a review. J R Soc Med. 1984;77: Hoogewerf PE, Hislop TG, Morrison BJ, et al. Patient compliance with screening for fecal occult blood in family practice. Can Med Assoc J. 1987;1327: Rockey DC, Auslander A, Greenberg PD. Detection of upper gastrointestinal blood with fecal occult blood tests. Am J Gastroenterol. 1999;94: August 15, 2000 THE AMERICAN JOURNAL OF MEDICINE Volume 109