The cyclooxygenase-2-selective inhibitors rofecoxib and celecoxib prevent colorectal neoplasia occurrence and recurrence

GASTROENTEROLOGY 2003;125:404 – 412

The Cyclooxygenase-2–Selective Inhibitors Rofecoxib and Celecoxib Prevent Colorectal Neoplasia Occurrence and Recurrence ELHAM RAHME,*,‡ ALAN N. BARKUN,§,㛳 YOUSSEF TOUBOUTI,‡ and MARC BARDOU§,㛳,¶ *Department of Medicine, McGill University, Montreal, Canada; ‡Research Institute, McGill University Health Center, Montreal, Canada; §Department of Medicine, Division of Gastroenterology, McGill University, Montreal, Canada; 㛳McGill University Health Center, Montreal, Canada; and ¶Department of Medicine, Clinical Pharmacology Unit, Faculty of Medicine, Dijon, France

Background & Aims: Colorectal cancer is one of the leading causes of cancer death. Most colorectal cancers are believed to develop from colorectal adenomas. We examined the effect of the selective cyclooxygenase-2 inhibitors rofecoxib and celecoxib, nonselective nonsteroidal anti-inflammatory drugs, aspirin, and acetaminophen on colorectal neoplasia (colorectal cancer, colorectal adenoma, or both). Methods: This was a nested case-control study, which used data from a government insurance database on patients 65 years and older who underwent a diagnostic test or procedure for colorectal neoplasia between January and June 2001. Logistic regression models were used to determine the effect of exposure to the drugs of interest for at least 3 months on the occurrence or recurrence of colorectal neoplasia. Results: The control group included 2568 patients found to be free of colorectal neoplasia; 730 patients were diagnosed with colorectal adenoma, and 179 were diagnosed with colorectal cancer. Patients more likely to have colorectal adenoma (odds ratio, 95% confidence interval) were those diagnosed with colorectal adenoma (4.12, 3.27–5.18) or colorectal cancer (3.74, 2.32–6.03) in the previous 1–3 years and those with hemorrhage of the rectum or unspecified anemia in the prior month (3.19, 2.46 – 4.12). Exposures to rofecoxib (0.67, 0.46 – 0.98) and nonselective nonsteroidal anti-inflammatory drugs (0.41, 0.21– 0.83) reduced the risk of colorectal adenoma. Rofecoxib, celecoxib, and nonselective nonsteroidal anti-inflammatory drugs were all protective against both neoplasias (0.64, 0.45– 0.91; 0.73, 0.54 – 0.99; and 0.47, 0.26 – 0.86, respectively). Conclusions: Rofecoxib, celecoxib, and nonselective nonsteroidal antiinflammatory drugs seem to protect against the development of colorectal neoplasia.

olorectal cancer (CRC) is the third most prevalent cancer in industrialized countries1 and accounts for 3.7% of all deaths.1–3 Its incidence is age related and approximates 3 per 1000 persons in those older than 65 years.4,5 CRC develops in a stepwise fashion from normal mucosa to adenomatous polyps to carcinoma, a process

C

that takes several years.6 It is widely accepted that detection and removal of adenomatous polyps at early stages prevents the progression to cancer.7,8 Family history increases the risk of CRC9 –12; however, most CRCs, which are asymptomatic, occur in patients with no identifiable risk factors other than age. A number of pharmacological agents have been studied for the prevention of CRC, including nonselective nonsteroidal anti-inflammatory drugs (NSAIDs) and cyclooxygenase (COX)-2–selective inhibitors or coxibs.9,13–16 COX-2 expression is low or undetectable in normal gastrointestinal (GI) mucosa but is found in approximately half of adenomas and 90% of CRCs.17,18 The effectiveness of COX-2 inhibitors has been reported in patients with a family history of adenomatous polyposis9 and has been suggested for other conditions, such as hereditary nonpolyposis CRC and sporadic adenomas.19 There is direct genetic evidence from a mouse model that the COX-2 gene contributes to the development of adenomas, where the specific COX-2 inhibitor rofecoxib resulted in a dosedependent reduction in the number and size of intestinal and colonic polyps in the Apcdelta716 mouse, a knockout model for the COX-2 gene.20 With the same Apc Min mouse model, another COX-2 inhibitor, celecoxib, was also found to exert prevention and regression effects against colonic polyps. Furthermore, compared with the nonselective NSAID piroxicam, celecoxib caused no GI side effects at plasma drug levels similar to those obtained in early clinical trials in humans.16 Celecoxib was also shown to reduce the number of colorectal adenomas (CRA) in patients with familial adenomatous polyposis.9 Long-term aspirin and nonaspirin nonselective NSAIDs have also been Abbreviations used in this paper: COX, cyclooxygenase; CRA, colorectal adenoma; CRC, colorectal cancer; GI, gastrointestinal; RAMQ, Re ´gie de l’Assurance Maladie du Que ´bec. © 2003 by the American Gastroenterological Association 0016-5085/03/$30.00 doi:10.1016/S0016-5085(03)00880-1

August 2003

associated with a reduced incidence of CRC.21–24 However, information on any possible dose-related effect of these agents has been controversial. At least 1 study suggested that low doses of NSAIDs were as effective as high doses for the prevention of colon cancer. In addition, none of the nonselective NSAIDs has been found to be more effective than another. At least 2 studies did not find any association between aspirin use and CRC reduction.25,26 However, more recently, 2 randomized, controlled trials found that aspirin has a moderate chemopreventive effect on CRA in the large bowel in patients with a recent history of histologically documented adenomas and a significant reduction effect in the incidence of CRA in patients with previous CRA or CRC, respectively.27,28 The aim of this study was to determine the presence and magnitude of a possible protective effect of the nonselective NSAIDs and the coxibs for the development of CRA in a general population of patients aged 65 years and older.

Materials and Methods Data Extraction In Quebec, the Government’s health insurance agency, the Re´gie de l’Assurance Maladie du Que´bec (RAMQ), covers all in- and outpatient medical services for all persons, as well as the cost of prescription drugs for all patients 65 years or older, those receiving social assistance, and, since 1997, those who do not have private drug insurance. Drugs dispensed to patients during stays in hospitals or public nursing homes are not covered by RAMQ. The RAMQ database has been used in other studies.29,30 Aspirin, nonaspirin NSAIDs, celecoxib and rofecoxib (coxibs), and acetaminophen are listed on the Quebec formulary with no specific restrictions to their use. Aspirin, acetaminophen, and ibuprofen, but none of the coxibs, are also available over the counter. From the RAMQ database we obtained in- and outpatient physician claims and prescription drug and demographic data on a random cohort of patients consisting of 25% of all seniors 65 years and older (approximately 900,000 persons) who were prescribed nonselective nonaspirin NSAIDs, aspirin, coxibs, or acetaminophen between January 1997 and September 2001.

Patient Selection We conducted a nested case-control study with both cases and controls selected from the cohort described previously. Patients were included if they were at least 66 years old (they were 65 years old at entry into the cohort and had been covered by the drug plan for at least 1 year at the time of analysis) and had undergone a diagnostic test or a procedure for CRA or CRC, colonoscopy, barium enema, or surgery (International Classification of Diseases-9 and RAMQ codes available on request) between January and June 2001. The date of the first diagnostic test or procedure was termed the index date. Hereafter, unless otherwise stated, “prior” will refer to before the index date. Prescription drug information and in- and outpatient data for the prior year

COXIBS PREVENT COLORECTAL NEOPLASIA

405

and 4 years, respectively, were available for all patients. We excluded patients if a diagnostic test or procedure for CRA or CRC, a diagnosis of cancer (any site), or more than 1 day of hospitalization had occurred within the prior year. Patients hospitalized for more than 1 day were excluded because drugs given at the hospital are not covered by RAMQ and it is not clear whether, once hospitalized, patients will interrupt their treatment.

Outcomes All outcomes were assessed within the 3 months after and including the index date. The main outcome was a diagnosis of CRA. Secondary outcomes were a diagnosis of CRC or any neoplasia (CRA/CRC). International Classification of Diseases-9 codes and RAMQ procedure codes that specifically indicated the presence of a CRA or CRC were used to assess the outcomes. Three groups of patients were formed according to diagnosis: (1) the healthy group consisted of those who were not diagnosed with either CRA or CRC; (2) the CRA group consisted of those who were diagnosed with a CRA but not CRC; and (3) the CRC group consisted of those who were diagnosed with CRC. To study the CRA, CRC, or CRA/CRC occurrence or recurrence, cases were selected to be all patients in the CRA, CRC, or either CRA or CRC groups, respectively, and controls were those in the healthy group.

Patient Data Collected at Baseline Patient characteristics at baseline were defined on the basis of medications received in the prior year and health services and physician encounters in the prior 4 years. The following variables were measured at the index date: age, sex, comorbidity indicators (number of drugs, number of physician encounters, and all-cause hospitalizations in the prior year), and risk factors for CRA or CRC—a history of CRA, defined as a diagnosis of CRA but not CRC in the past 1– 4 years; a history of CRC, defined as a diagnosis of CRC in the past 1– 4 years; a diagnostic procedure for CRA or CRC with “negative results” (no diagnosis of either of these conditions) in the past 1– 4 years; and GI symptoms likely to be associated with CRA or CRC (Appendix 1) in the prior month (GI symptoms were further separated into 2 categories: “hemorrhage of rectum or unspecified anemia” and “other symptoms”).

Exposure Patients who filled a prescription for any nonselective NSAID, aspirin, rofecoxib, celecoxib, or acetaminophen were classified as exposed to these respective drugs. The number of days’ supply for each filled prescription of each study drug was recorded. For each patient, the exposure duration to each study drug was the total number of days’ supply in the prior year. The daily dose of each drug was also recorded and dichotomized to high or low daily doses according to the maintenance dose recommended in the Compendium of Pharmaceuticals and Specialties.31,32 For each patient, the total number of days on a high dose was calculated for each drug.

406

RAHME ET AL.

Statistical Analyses Descriptive analyses (means ⫾ SD and proportions) were used to compare patient characteristics at the index date among the 3 groups. Main analysis. In this analysis, cases were those patients in the CRA group, and controls were those in the healthy group. Logistic regression models were used to determine the effect on CRA of at least 3 months of exposure to rofecoxib, celecoxib, nonselective NSAIDs, aspirin, or acetaminophen as compared with no exposure to any of these drugs. The model adjusted for age, sex, and patient baseline characteristics. A backward selection was performed where exposure to the drugs of interest was forced in the model, and other patient characteristics (potential effect modifiers) were retained in the model if they were significant at P ⬍ 0.1. Additional analyses. Several additional analyses were performed. First, because patients could have been exposed to several study drugs during the past year, the main analysis was recalculated 5 times to study the effect of each drug separately. For example, to study the sole effect of rofecoxib, patients exposed to nonselective NSAIDs or celecoxib were excluded. Second, we further dichotomized patients with at least 3 months of exposure to each drug into 2 groups: those who had at least 3 months of exposure to high doses of the drug and those who did not. Third, to study the secondary outcome of CRC, we used logistic regression models with the same variables as in the main analysis to compare patients in the CRC group with controls (patients in the healthy group). Fourth, we also used logistic regression models to study all neoplasia (CRA/ CRC). Fifth, we repeated the main analysis in patients with a history of neoplasia. Finally, because rofecoxib was listed on the Quebec drug formulary in April 2000 and therefore was not reimbursed in the first 4 months of 2000, whereas all other study drugs were, we repeated all analyses, redefining patient characteristics and prior exposure to be within the prior 8 months.

Sample Size We calculated that a sample size of 724 patients (362 in each group) was needed for a 2-group continuitycorrected ␹2 test with a 0.05 2-sided significance level to have 80% power to detect the difference between a 20% CRA proportion in those not exposed and a 12% CRA proportion in those exposed for at least 3 months of rofecoxib (odds ratio, 0.55). Assuming that 25% of study patients are exposed to at least 3 months of rofecoxib or not exposed to any of the study drugs, we estimated that 2896 patients would be needed for this study. Nonexposed patients were assumed to have the baseline CRA risk of 20%.33–36 Rofecoxib and celecoxib were assumed to reduce the risk of CRA occurrence or recurrence by 40%, for the purposes of the sample size calculation.

GASTROENTEROLOGY Vol. 125, No. 2

Results Cohort Selection A total of 8033 patients had a colorectal procedure between January and June 2001. Of these, 2196 had a colorectal procedure or a diagnosis of CRA or CRC in the prior year and were excluded. Also excluded were an additional 1013 patients who had a cancer diagnosis (any site) and an additional 1347 patients who were hospitalized for more than 1 day or were in a public nursing home or long-term care facility in the prior year. This left 3477 patients in the study population, including 2568 patients (73.9%) in the healthy group, 730 (21.0%) in the CRA group, and 179 (5.1%) in the CRC group. Cohort Characteristics Patient characteristics at the index date are shown in Table 1. Key patient characteristics are presented in controls vs. patients with CRA, CRC, or all neoplasias. Patients undergoing a procedure for CRA or CRC were mostly women (62.4%), of whom 18.4% were diagnosed with CRA and 4.3% with CRC; 25.3% of men were diagnosed with CRA and 6.6% with CRC. Approximately 50.2% of patients experienced GI symptoms more or less likely to be associated with CRA or CRC (Appendix 1) in the prior month, of whom 17.6% had a diagnosis of CRA and 5.3% one of CRC, compared with 24.4% CRA and 5.0% CRC for patients without symptoms. Among those with hemorrhage of the rectum or anemia in the prior month (11.2% of total), 40.2% were diagnosed with CRA and 11.8% with CRC. Patients in the CRA group were younger and more often had a prior CRA compared with patients in either the healthy or CRC groups. A total of 38 patients died in the 3 months after the index procedure. Of these, 19 were in the healthy group (0.7% of the group), 7 in the CRA group (1% of the group), and 12 in the CRC group (6.7% of the group). Effect of NSAIDs or Coxibs on Colorectal Adenoma The results of the main analysis are shown in Table 2. Logistic regression models were used to determine the effect of exposure to at least 3 months of the study drugs on CRA occurrence and recurrence while adjusting for baseline factors. Patients at higher risk for CRA (odds ratio, 95% confidence interval) were those with a history of CRA (4.12, 3.27–5.18) and those with a history of CRC (3.74, 2.32– 6.03) compared with those with no prior neoplasia. Patients with “hemorrhage of rectum or unspecified anemia” in the prior month were also at higher risk of CRA (3.19, 2.46 – 4.12) compared with patients without symptoms. Female patients (0.73,

August 2003

COXIBS PREVENT COLORECTAL NEOPLASIA

407

Table 1. Patient Characteristics at the Index Date Variable No. patients Demographics (%) Female Age (yr) % 66–74 75–84 ⱖ85 Comorbidity indicators All-cause hospitalization (1 day) (%) No. drugs in prior year (mean ⫾ SD) Number of physician encounters in prior year (mean ⫾ SD) Cancer any site in prior 1–3 years (%) Colonoscopy at index date GI symptoms in prior month (%) Hemorrhage of rectum or unspecified anemia Other symptoms History of neoplasia (%) Colorectal procedure CRA CRC Exposure (%) Rofecoxib ⬎90 days ⬎90 days of high dose Celecoxib ⬎90 days ⬎90 days of high dose NSAIDs ⬎90 days Acetaminophen ⬎90 days Aspirin ⬎90 days Not exposed Exposure duration, mean ⴞ SD (median, interquartile range) Rofecoxib ⬎90 days Celecoxib ⬎90 days NSAIDs ⬎90 days Acetaminophen ⬎90 days Aspirin ⬎90 days Death in the 3 mo after the index procedure

Healthy

CRA

CRC

2568

730

179

65.3

54.7

52.0

59.8 34.4 5.8

64.8 31.1 4.1

45.8 43.0 11.2

18.5 8.3 ⫾ 5.4

16.4 7.8 ⫾ 5.0

17.3 6.9 ⫾ 5.0

14.5 ⫾ 13.7 14.5 57.4 52.4

13.2 ⫾ 10.2 18.6 91.4 42.1

11.1 ⫾ 10.2 19.0 79.3 51.4

7.3 45.1

21.5 20.6

25.7 25.7

29.3 9.0 1.7

45.1 27.8 4.9

26.8 8.4 10.6

26.8 7.7 0.5 27.5 10.0 3.2 13.4 3.2 26.3 7.8 37.5 31.3 21.8

25.2 5.6 0.3 24.0 8.9 2.7 12.1 1.4 25.5 6.6 38.9 31.8 24.7

15.6 4.5 0 16.8 2.8 0.6 10.6 2.2 23.5 6.1 39.7 35.8 27.9

79 ⫾ 85 (42, 96) 192 ⫾ 77 (173, 51) 102 ⫾ 104 (60, 126) 218 ⫾ 86 (210, 165) 69 ⫾ 80 (30, 75) 190 ⫾ 76 (163, 110) 79 ⫾ 91 (37, 108) 199 ⫾ 77 (183, 115) 263 ⫾ 116 (330, 180) 303 ⫾ 77 (330, 90)

67 ⫾ 79 (30, 75) 193 ⫾ 73 (187, 114) 101 ⫾ 98 (60, 160) 213 ⫾ 72 (210, 120) 54 ⫾ 83 (24, 48) 259 ⫾ 97 (267, 155) 71 ⫾ 84 (32, 84) 195 ⫾ 72 (195, 114) 255 ⫾ 120 (300, 210) 300 ⫾ 77 (330, 105)

87 ⫾ 103 (30, 130) 233 ⫾ 76 (232, 154) 60 ⫾ 79 (30, 45) 197 ⫾ 122 (120, 220) 64 ⫾ 90 (20, 64) 219 ⫾ 77 (183, 83) 72 ⫾ 89 (30, 93) 195 ⫾ 84 (192, 147) 286 ⫾ 95 (330, 120) 309 ⫾ 65 (330, 90)

0.7

1.0

6.7

0.60 – 0.87) and patients with “other symptoms” (symptoms other than “hemorrhage of rectum or unspecified anemia”) in the prior month (0.41, 0.33– 0.51) were less likely to have a diagnosis of CRA. Exposure to either rofecoxib (0.67, 0.46 – 0.98) or nonselective NSAIDs (0.41, 0.21– 0.83) was found to reduce the risk of CRA. Exposure to celecoxib was not associated with significant protection (0.87, 0.63–1.19). The results of the main analysis did not change when we studied the effect of

each study drug separately by excluding patients who had taken the other drugs in the prior year (Table 3). The results of rofecoxib and celecoxib differed slightly from those of the main analysis when we separated the patients exposed to at least 3 months according to whether or not they had been exposed to high or low doses. The odds ratios for rofecoxib were 0.37 (0.07– 1.88) at high doses and 0.70 (0.47–1.03) at low doses, whereas those for high- and low-dose celecoxib were 0.76

408

RAHME ET AL.

GASTROENTEROLOGY Vol. 125, No. 2

Table 2. Effect of Exposure to Coxibs, NSAIDs, Aspirin, or Acetaminophen on Colorectal Adenoma Adjusted for Patient Characteristics at the Index Date (Main Analysis) Variable

Odds ratio (95% CI)

Female sex Age 66–74 yr Age 75–85 yr Age ⱖ85 yr Symptoms in the prior month No symptoms Hemorrhage of rectum or unspecified anemia Other symptoms No CRA or CRC history or previous CRA/CRC procedure CRA history CRC history Previous CRA/CRC procedure with negative result Not exposed Exposeda to rofecoxib Exposed to celecoxib Exposed to NSAIDs Exposed to aspirin Exposed to acetaminophen

0.73 (0.60–0.87) 1 (REF) 0.80 (0.66–0.97) 0.62 (0.40–0.97) 1 (REF) 3.19 (2.46–4.12) 0.41 (0.33–0.51) 1 (REF) 4.12 (3.27–5.18) 3.74 (2.32–6.03) 0.95 (0.74–1.22) 1 (REF) 0.67 (0.46–0.98) 0.87 (0.63–1.19) 0.41 (0.21–0.83) 0.84 (0.67–1.04) 0.80 (0.55–1.14)

NOTE. The model also adjusted for number of drugs, number of physician encounters, and all-cause hospitalization in prior year. Exposure (⬎90 days) to high vs. low doses of either rofecoxib or celecoxib was not found to be different. REF, reference value; CI, confidence interval. aExposed in this table refers to exposure for ⬎90 days.

(0.44 –1.31) and 0.92 (0.63–1.33), respectively. Highdose aspirin was not studied in this analysis because in Quebec, most aspirin is prescribed at low doses for antithrombotic prophylaxis.30 An insufficient number of patients were exposed to at least 3 months of high-dose

nonselective NSAIDs, and this precluded a similar analysis in this group. Effect of NSAIDs or Coxibs on Colorectal Cancer Only celecoxib showed a statistically significant protective effect against CRC occurrence or recurrence (Table 4). Compared with nonexposed patients, the odds ratios were 0.67 (0.24 –1.92) for NSAIDs, 0.53 (0.25– 1.14) for rofecoxib, and 0.23 (0.09 – 0.59) for celecoxib. Effect of NSAIDs or Coxibs on Neoplasia When assessing all neoplasia (CRC or CRA), nonselective NSAIDs, rofecoxib, and celecoxib were associated with a significant protective effect. The results were 0.47 (0.26 – 0.86), 0.64 (0.45– 0.91), and 0.73 (0.54 – 0.99) in those exposed to nonselective NSAIDs, rofecoxib, and celecoxib, respectively, as compared with nonexposed patients (Table 4). Effect of NSAIDs or Coxibs on Neoplasia in Patients With a History of Neoplasia When we restricted the analyses to patients with a history of neoplasia, the effect of rofecoxib on neoplasia was greater (0.43, 0.19 – 0.95) than that observed in the main analysis. In contrast, celecoxib did not display any significant effect (0.97, 0.49 –1.90). NSAIDs (0.54, 0.13–2.17) were also protective; however, their effect did not reach statistical significance. The results did not change when we repeated all analyses, considering exposure and patients’ characteristics only in the prior 8 months (since the listing of rofecoxib; data not shown).

Table 3. Results of 5 Separate Analyses, Illustrating the Effect on Colorectal Adenoma of Each Study Drug Separately Variable Effect of acetaminophen, excluding patients who took NSAIDs, rofecoxib, celecoxib, or aspirin in the prior year ⬎90 days of acetaminophen ⱕ90 days of acetaminophen Effect of aspirin, excluding patients who took NSAIDs, rofecoxib, or celecoxib in the prior year ⬎90 days of aspirin ⱕ90 days of aspirin Effect of celecoxib, excluding patients who took NSAIDs, rofecoxib, or aspirin in the prior year ⬎90 days of celecoxib ⱕ90 days of celecoxib Effect of rofecoxib, excluding patients who took NSAIDs, celecoxib, or aspirin in the prior year ⬎90 days of rofecoxib ⱕ90 days of rofecoxib Effect of NSAIDs, excluding patients who took rofecoxib or celecoxib in the prior yeara ⬎90 days of NSAIDs ⱕ90 days of NSAIDs

Odds ratio (95% CI) 1.12 (0.54–2.30) 0.81 (0.50–1.30) 0.85 (0.65–1.11) 0.73 (0.42–1.26) 0.78 (0.48–1.26) 0.98 (0.66–1.44) 0.74 (0.41–1.34) 0.91 (0.63–1.32) 0.46 (0.19–1.13) 1.04 (0.66–1.62)

NOTE. All models were adjusted for patient characteristics at the index date. CI, confidence interval. aWe did not exclude patients previously exposed to aspirin because the resulting number of patients exposed to NSAIDs would have been too small to be studied in a model.

August 2003

COXIBS PREVENT COLORECTAL NEOPLASIA

409

Table 4. Effect of Exposure to Rofecoxib, Celecoxib, NSAIDs, Aspirin, or Acetaminophen on Colorectal Cancer or Neoplasia of Any Kind (CRA/CRC) Adjusted for Patient Characteristics at the Index Date Variable Female sex Age 66–74 yr Age 75–84 yr Age ⱖ 85 yr Symptoms in the prior month No symptoms Hemorrhage of rectum or unspecified anemia Other symptoms No CRA or CRC diagnosis or procedure in prior 4 years CRA history CRC history Previous CRA/CRC procedure with negative result Not exposed Exposeda to rofecoxib Exposed to celecoxib Exposed to NSAIDs Exposed to aspirin Exposed to acetaminophen

CRC, odds ratio (95% CI)

CRA/CRC, odds ratio (95% CI)

0.59 (0.43–0.81) 1 (REF) 1.64 (1.17–2.29) 2.57 (1.48–4.46)

0.69 (0.58–0.81) 1 (REF) 0.92 (0.77–1.10) 0.95 (0.66–1.37)

1 (REF) 3.57 (2.37–5.38) 0.61 (0.42–0.89) 1 (REF) 1.00 (0.57–1.76) 6.05 (3.31–11.08) 0.52 (0.31–0.87) 1 (REF) 0.53 (0.25–1.14) 0.23 (0.09–0.59) 0.67 (0.24–1.92) 0.87 (0.59–1.28) 0.71 (0.36–1.39)

1 (REF) 3.23 (2.55–4.10) 0.45 (0.37–0.54) 1 (REF) 3.36 (2.69–4.19) 4.25 (2.78–6.51) 0.84 (0.67–1.06) 1 (REF) 0.64 (0.45–0.91) 0.73 (0.54–0.99) 0.47 (0.26–0.86) 0.85 (0.70–1.04) 0.78 (0.55–1.09)

NOTE. The model also adjusted for number of drugs, number of physician encounters, and all-cause hospitalization (for 1 day) in the prior year. REF, reference category; CI, confidence interval. aExposed in this table refers to exposure for ⬎90 days.

Discussion This analysis is a population-based epidemiological study that examines the effect of exposure to COX-2 inhibitors on CRA occurrence and recurrence. It was conducted by using a database that includes the entire elderly population of Quebec (approximately 900,000 persons). We showed that exposure to at least 3 months of rofecoxib or nonselective NSAIDs (all doses) had a significant protective effect. Exposure to at least 3 months of celecoxib or aspirin also showed a trend to a protective effect, which did not reach statistical significance. Exposure to acetaminophen did not have any effect. Exposure to at least 3 months of high-dose rofecoxib or celecoxib seemed to have a stronger effect than exposure to low doses, but the results did not reach statistical significance, perhaps because of the small number of patients in each of these groups. Exposure to at least 3 months of high-dose NSAIDs could not be studied because almost no one in this study met that criterion. This finding is not surprising, because prolonged use of high-dose nonselective NSAIDs in the elderly is considered a significant risk factor for untoward GI events.37,38 We also did not study the effect of highdose aspirin because our previous study30 showed that aspirin is most often prescribed chronically at low doses in the elderly for its antithrombotic effect. Patients diagnosed with CRA had much higher rates of prior recent “hemorrhage of rectum or unspecified anemia” (in the month before colorectal procedure) and

fewer “other symptoms” than patients not diagnosed with either CRA or CRC. Given that the large majority of polyps are silent, this might be explained by the fact that asymptomatic patients were referred because of an increased risk for neoplasia (mostly family or personal history of either CRA or CRC), whereas symptomatic patients were more likely to have other nonneoplastic GI conditions, such as irritable bowel syndrome. It was recently reported that patients with irritable bowel syndrome have no increased risk for CRC compared with the general population.39 This finding emphasizes the importance of screening in patients with a positive family history. Female patients had more colorectal procedures than male patients but less CRA. This is in agreement with a recent epidemiological study in which it was reported that men were at a greater risk of polyps (odds ratio, 1.5) and tumors (odds ratio, 1.4) than women.40 We also examined the occurrence of CRC to ensure that the reduction in CRA was not due to an increase in CRC. Exposure to at least 3 months of celecoxib (all doses) was found to have a protective effect, whereas exposure to at least 3 months of rofecoxib was also protective but failed to reach statistical significance. Patients in the CRC group had also much higher rates of prior recent “hemorrhage of rectum or unspecified anemia” and fewer “other symptoms” than patients in the healthy group, and female patients had fewer CRC diagnoses than male patients.

410

RAHME ET AL.

The protective effect of coxibs on CRA occurrence or recurrence may be explained by COX-2 overexpression in colonic adenomas.41 Our study confirms and extends the results found in very high-risk populations of patients, such as those with a family history of adenomatous polyposis (which is thought to account for approximately 0.5% of all CRC), in whom sulindac was able to reduce the number or the size of CRAs42– 45 and celecoxib was effective in reducing the number of CRAs.9 A more recent and controversial study found that sulindac treatment (75 or 150 mg daily) of patients who were genotypically affected with familial adenomatous polyposis but were phenotypically unaffected did not prevent the development of CRA.46 The literature reports a beneficial effect of nonselective nonaspirin NSAIDs and aspirin against CRC,23,47,48 but no prior study has clearly shown that coxibs protect against CRA or CRC occurrence in a general population. The beneficial effect of aspirin did not reach statistical significance in this study. This is likely to be explained, at least partly, by the time of exposure that was considered, by the fact that aspirin was mostly used at daily doses of 75 to 325 mg, or by patients’ noncompliance with the medication prescribed. Indeed, epidemiological studies have shown that either the dose of aspirin49 or the duration of treatment23 might interfere with the protective effect against CRC. Moreover, 1 study in an Apc Min mice showed that, contrary to indomethacin, aspirin did not reduce the tumor load.50 However, 2 recent randomized, controlled trials27,28 found that aspirin had a significant chemopreventive effect against CRA in patients with previous CRA or CRC, respectively. The risk reduction was significant with a daily dose of 81 mg but not with 325 mg for patients with a history of histologically proven CRA. Our study period was restricted because of the relatively recent introduction of the coxibs to the market. Because cancer latency extends from 5 to 10 years, our study period was too short to address any definitive effect on cancer incidence. Our study design was therefore best suited to assess the occurrence of CRA, and a longer study period with more prolonged coxib exposure is needed to confidently assess CRC occurrence or recurrence. However, despite this limitation, a strong protective effect against CRC was observed with a relatively short exposure to either coxibs or nonselective NSAIDs. This analysis was conducted by using a large, population-based, well-validated medical database.29,30 The coxibs, NSAIDs, acetaminophen, and aspirin are available to all patients regardless of their risk factors. Therefore, the use of the RAMQ database provides the advan-

GASTROENTEROLOGY Vol. 125, No. 2

tage of large sample size, generalizability, and the broad inclusion of patients with multiple risk factors who are typically excluded from clinical trials. The RAMQ prescription claims database has been found to be both accurate and reliable when patients’ clinical records were validated with their RAMQ records.51 Retrospective claims database analyses have 2 common limitations: inaccuracies from the use of a payment database to infer clinical events, and selection bias. The occurrence of CRA or CRC was inferred from patterns of resource utilization and payment diagnosis codes, rather than from the review of chart records. However, any miscoding would drive the odds ratio toward no effect, as a result of nondifferential misclassification bias.52 Furthermore, because nonselective NSAIDs and coxibs are believed to reduce CRC occurrence, patients with a family history of CRC would likely have received more NSAIDs or coxibs; a selection bias would therefore have resulted in an effect that was the reverse of the one noted. Because of delays in physician encounters and the possible difficulty in ascertaining the presence of CRA vs. CRC at the first colorectal procedure, we assessed the diagnoses in the 3 months after the first procedure. Other potential limitations of the study include uncertainty with respect to actual doses taken and unknown but concurrent use of over-the-counter drugs (especially aspirin, ibuprofen, and acetaminophen). A survey conducted by Sante´ Que´bec, a government of Quebec public health agency, showed that in 1998, 17.0%, 2.2%, and 46.3% of the elderly who consumed, respectively, NSAIDs, aspirin, or acetaminophen acquired them over the counter. Quebec has maintained the same drug insurance policy since 1997. The high number of patients who acquire acetaminophen or NSAIDs over the counter suggests that some patients may have been exposed for longer than 3 months but were misclassified into the less-than-3-months group. However, this misclassification would not have affected the results in those classified as exposed for more than 3 months. In summary, the coxibs seem very promising as protective drugs against CRA and CRC. Prospective interventional studies are now warranted to confirm these results and assess their risks and benefits.

Appendix 1 International Classification of Diseases-9 Codes Used to Identify Symptoms That May Be Associated With Colorectal Adenoma or Colorectal Cancer 1. Hemorrhage of rectum or unspecified anemia: 5693, 2859, 2819

August 2003

2. Other symptoms: 7890, 7893, 7894, 7896, 5589, 5799, 5640,5641, 5645, 5649,5694,5569, 0091, 7876, 5609, 7873, 5621, 2859, 2819, 5649, 5650, 5651, 7879, 5550–5559

References 1. Micheli A, Mugno E, Krogh V, Quinn MJ, Coleman M, Hakulinen T, Gatta G, Berrino F, Capocaccia R. Cancer prevalence in European registry areas. Ann Oncol 2002;13:840 – 865. 2. Gwyn K, Sinicrope FA. Chemoprevention of colorectal cancer. Am J Gastroenterol 2002;97:13–21. 3. NINC. Canadian Cancer Statistics 2000. Available at: http:// www.cancer.ca. 4. Bakhle YS. COX-2 and cancer: a new approach to an old problem. Br J Pharmacol 2001;134:1137–1150. 5. Gaudette LA, Altmayer CA, Wysocki M, Gao RN. Cancer incidence and mortality across Canada. Health Rep 1998;10:51– 66. 6. Hill MJ, Morson BC, Bussey HJ. Aetiology of adenoma– carcinoma sequence in large bowel. Lancet 1978;1:245–247. 7. Pignone M, Rich M, Teutsch SM, Berg AO, Lohr KN. Screening for colorectal cancer in adults at average risk: a summary of the evidence for the U.S. Preventive Services Task Force. Ann Intern Med 2002;137:132–141. 8. Wan J, Zhang ZQ, Zhu C, Wang MW, Zhao DH, Fu YH, Zhang JP, Wang YH, Wu BY. Colonoscopic screening and follow-up for colorectal cancer in the elderly. World J Gastroenterol 2002;8:267– 269. 9. Steinbach G, Lynch PM, Phillips RKS. The effect of celecoxib, a cyclo oxygenase-2 inhibitor, in familial adenomatous polyposis. N Engl J Med 2000;342:1946 –1952. 10. Samowitz WS, Curtin K, Lin HH, Robertson MA, Schaffer D, Nichols M, Gruenthal K, Leppert MF, Slattery ML. The colon cancer burden of genetically defined hereditary nonpolyposis colon cancer. Gastroenterology 2001;121:830 – 838. 11. Mukherjee D, Nissen SE, Topol EJ. Risk of cardiovascular events associated with selective COX-2 inhibitors. JAMA 2001;286: 954 –959. 12. Hampel H, Peltomaki P. Hereditary colorectal cancer: risk assessment and management. Clin Genet 2000;58:89 –97. 13. Bus PJ, Verspaget HW, Lamers CB, Griffioen G. Chemoprevention of colorectal cancer by non-steroidal anti-inflammatory drugs. Scand J Gastroenterol Suppl 2000;232:101–104. 14. Ricciardiello L, Roda E, Bazzoli F. Chemoprevention in colorectal neoplasias: what is practical and feasible? Dig Dis 2002;20:70 – 72. 15. Vaisman N, Arber N. The role of nutrition and chemoprevention in colorectal cancer: from observations to expectations. Best Pract Res Clin Gastroenterol 2002;16:201–217. 16. Jacoby RF, Seibert K, Cole CE, Kelloff G, Lubet RA. The cyclooxygenase-2 inhibitor celecoxib is a potent preventive and therapeutic agent in the min mouse model of adenomatous polyposis. Cancer Res 2000;60:5040 –5044. 17. Chapple KS, Cartwright EJ, Hawcroft G, Tisbury A, Bonifer C, Scott N, Windsor AC, Guillou PJ, Markham AF, Coletta PL, Hull MA. Localization of cyclooxygenase-2 in human sporadic colorectal adenomas. Am J Pathol 2000;156:545–555. 18. Sheehan KM, Sheahan K, O’Donoghue DP, MacSweeney F, Conroy RM, Fitzgerald DJ, Murray FE. The relationship between cyclooxygenase-2 expression and colorectal cancer. JAMA 1999;282: 1254 –1257. 19. Lynch PM. COX-2 inhibition in clinical cancer prevention. Oncology (Huntingt) 2001;15:21–26. 20. Oshima M, Murai N, Kargman S, Arguello M, Luk P, Kwong E, Taketo MM, Evans JF. Chemoprevention of intestinal polyposis in the Apcdelta716 mouse by rofecoxib, a specific cyclooxygenase-2 inhibitor. Cancer Res 2001;61:1733–1740.

COXIBS PREVENT COLORECTAL NEOPLASIA

411

21. Marnett JL. Aspirin and the potential role of prostaglandins in colon cancer. Cancer Res 1992;52:5575–5589. 22. Giovannucci E, Egan KM, Hunter DJ, Stampfer MJ, Colditz GA, Willett WC, Speizer FE. Aspirin and the risk of colorectal cancer in women. N Engl J Med 1995;333:609 – 614. 23. Collet JP, Sharpe C, Belzile E, Boivin J-F, Hanley J, Abenhaim L. Colorectal cancer prevention by non-steroidal anti-inflammatory drugs: effects of dosage and timing. Br J Cancer 1999;81:62– 68. 24. Smalley W, Ray WA, Daugherty J, Griffin MR. Use of nonsteroidal anti-inflammatory drugs and incidence of colorectal cancer: a population-based study. Arch Intern Med 1999;159:161–166. 25. Gann PH, Manson JE, Glynn RJ, Buring JE, Hennekens CH. Lowdose aspirin and incidence of colorectal tumors in a randomized trial. J Natl Cancer Inst 1993;85:1220 –1224. 26. Paganinihill A, Chao A, Ross RK, Henderson BE. Aspirin use and chronic diseases—a cohort study of the elderly. BMJ 1989;299: 1247–1250. 27. Sandler RS, Halabi S, Baron JA, Budinger S, Paskett E, Keresztes R, Petrelli N, Pipas JM, Karp DD, Loprinzi CL, Steinbach G, Schilsky R. A randomized trial of aspirin to prevent colorectal adenomas in patients with previous colorectal cancer. N Engl J Med 2003;348:883– 890. 28. Baron JA, Cole BF, Sandler RS, Haile RW, Ahnen D, Bresalier R, McKeown-Eyssen G, Summers RW, Rothstein R, Burke CA, Snover DC, Church TR, Allen JI, Beach M, Beck GJ, Bond JH, Byers T, Greenberg ER, Mandel JS, Marcon N, Mott LA, Pearson L, Saibil F, van Stolk RU. A randomized trial of aspirin to prevent colorectal adenomas. N Engl J Med 2003;348:891– 899. 29. Rahme E, Pilote L, LeLorier J. Association between naproxen use and protection against acute myocardial infarction. Arch Intern Med 2002;162:1111–1115. 30. Rahme E, Joseph L, Kong SX, Watson DJ, LeLorier J. Cost of prescribed NSAID-related gastrointestinal adverse events in elderly patients. Br J Clin Pharmacol 2001;52:185–192. 31. Compendium of pharmaceuticals and specialties (CPS). The Canadian drug reference for health professionals. Ottawa: Canadian Pharmacists Association, 2002. 32. Rahme E, Pettit D, LeLorier J. Determinants and sequelae associated with acetaminophen utilization versus traditional NSAIDs in an elderly population. Arthritis Rheum 2002;46:3046 –3054. 33. Imperiale TF, Wagner DR, Lin CY, Larkin GN, Rogge JD, Ransohoff DF. Results of screening colonoscopy among persons 40 to 49 years of age. N Engl J Med 2002;346:1781–1785. 34. Lieberman DA, Weiss DG, Bond JH, Ahnen DJ, Garewal H, Chejfec G. Use of colonoscopy to screen asymptomatic adults for colorectal cancer. Veterans Affairs Cooperative Study Group 380. N Engl J Med 2000;343:162–168. 35. Paspatis GA, Papanikolaou N, Zois E, Michalodimitrakis E. Prevalence of polyps and diverticulosis of the large bowel in the Cretan population. An autopsy study. Int J Colorectal Dis 2001; 16:257–261. 36. Saitoh Y, Waxman I, West AB, Popnikolov NK, Gatalica Z, Watari J, Obara T, Kohgo Y, Pasricha PJ. Prevalence and distinctive biologic features of flat colorectal adenomas in a North American population. Gastroenterology 2001;120:1657–1665. 37. Hochberg MC, Altman RD, Brandt KD, Clark BM, Dieppe PA, Griffin MR, Moskowitz RW, Schnitzer TJ. Guidelines for the medical management of osteoarthritis. Part II. Osteoarthritis of the knee. American College of Rheumatology. Arthritis Rheum 1995; 38:1541–1546. 38. Hochberg MC, Altman RD, Brandt KD, Clark BM, Dieppe PA, Griffin MR, Moskowitz RW, Schnitzer TJ. Guidelines for the medical management of osteoarthritis. Part I. Osteoarthritis of the hip. American College of Rheumatology. Arthritis Rheum 1995; 38:1535–1540. 39. Garcia Rodriguez LA, Ruigomez A, Wallander MA, Johansson S,

412

40.

41.

42.

43.

44.

45.

46.

RAHME ET AL.

Olbe L. Detection of colorectal tumor and inflammatory bowel disease during follow-up of patients with initial diagnosis of irritable bowel syndrome. Scand J Gastroenterol 2000;35:306 – 311. McCashland TM, Brand R, Lyden E, de Garmo P. Gender differences in colorectal polyps and tumors. Am J Gastroenterol 2001; 96:882– 886. Eberhart CE, Coffey RJ, Radhika A, Giardiello FM, Ferrenbach S, Dubois RN. Up-regulation of cyclooxygenase 2 gene expression in human colorectal adenomas and adenocarcinomas. Gastroenterology 1994;107:1183–1188. Giardiello FM, Hamilton SR, Krush AJ, Piantadosi S, Hylind LM, Celano P, Booker SV, Robinson CR, Offerhaus GJ. Treatment of colonic and rectal adenomas with sulindac in familial adenomatous polyposis. N Engl J Med 1993;328:1313–1316. Labayle D, Fischer D, Vielh P, Drouhin F, Pariente A, Bories C, Duhamel O, Trousset M, Attali P. Sulindac causes regression of rectal polyps in familial adenomatous polyposis. Gastroenterology 1991;101:635– 639. Rigau J, Pique JM, Rubio E, Planas R, Tarrech JM, Bordas JM. Effects of long-term sulindac therapy on colonic polyposis. Ann Intern Med 1991;115:952–954. Spagnesi MT, Tonelli F, Dolara P, Caderni G, Valanzano R, Anastasi A, Bianchini F. Rectal proliferation and polyp occurrence in patients with familial adenomatous polyposis after sulindac treatment. Gastroenterology 1994;106:362–366. Giardiello FM, Yang VW, Hylind LM, Krush AJ, Petersen GM, Trimbath JD, Piantadosi S, Garrett E, Geiman DE, Hubbard W, Offerhaus GJ, Hamilton SR. Primary chemoprevention of familial adenomatous polyposis with sulindac. N Engl J Med 2002;346: 1054 –1059.

GASTROENTEROLOGY Vol. 125, No. 2

47. Peleg II, Maibach HT, Brown SH, Wilcox CM. Aspirin and nonsteroidal anti-inflammatory drug use and the risk of subsequent colorectal cancer. Arch Intern Med 1994;154:394 –399. 48. Reeves MJ, Newcomb PA, Trentham-Dietz A, Storer BE, Remington PL. Nonsteroidal anti-inflammatory drug use and protection against colorectal cancer in women. Cancer Epidemiol Biomarkers Prev 1996;5:955–960. 49. Garcia Rodriguez LA, Hernandez-Diaz S. Risk of upper gastrointestinal complications among users of acetaminophen and nonsteroidal anti-inflammatory drugs. Epidemiology 2001;12:570 – 576. 50. Chiu C, McEntee M, Whelan J. Discordant effect of aspirin and indomethacin on intestinal tumor burden in Apc(Min/⫹)mice. Prostaglandins Leukot Essent Fatty Acids 2000;62:269 –275. 51. Tamblyn R, Lavoie G, Petrella L, Monette J. The use of prescription claims databases in pharmacoepidemiological research: the accuracy and comprehensiveness of the prescription claims database in Quebec. J Clin Epidemiol 1995;48:999 –1009. 52. Rothman KJ, Greenland S. Modern epidemiology. 2nd ed. Philadelphia: Lippincott-Raven, 1998.

Received December 18, 2002. Accepted May 8, 2003. Address requests for reprints to: Elham Rahme, Ph.D., Division of Clinical Epidemiology, Montreal General Hospital, 1650 Cedar Avenue, Montreal, Quebec H3G 1A4, Canada. e-mail: [email protected]; fax: (514) 934-8293. Supported in part by the Arthritis Society. Dr. Barkun is a research scholar funded by the FRSQ (Fonds de la Recherche en Sante ´ du Que ´bec). Dr. Rahme is funded by the Arthritis Society. Dr. Rahme has served as consultant for Merck & Co. Inc. and Pfizer Inc., the companies that manufacture rofecoxib and celecoxib.