Nonsteroidal anti-inflammatory drugs

Effects on mortality after colorectal cancer diagnosis

Abstract

BACKGROUND:

Nonsteroidal anti-inflammatory drug (NSAID) use has been associated with a decreased colorectal cancer (CRC) risk. However, to the best of the authors' knowledge, the effects of NSAID on clinical outcomes after CRC diagnosis are not well defined. The authors investigated the association between prediagnosis NSAID use and mortality after CRC diagnosis among women in the California Teachers Study cohort.

METHODS:

Women aged <85 years participating in the California Teachers Study, without a prior CRC diagnosis at baseline (1995-1996), and who were diagnosed with CRC during follow-up through December 2005, were eligible for analysis of the association between prediagnosis NSAID use and mortality. NSAID use (including aspirin and ibuprofen) was collected through a self-administered questionnaire. Cancer occurrence was identified through California Cancer Registry linkage. Multivariate Cox proportional hazards regression models were used to estimate hazards ratios (HR) for death and 95% confidence intervals (95% CIs).

RESULTS:

Among 621 CRC patients who were identified, 64% reported no prediagnosis regular NSAID use, 17% reported use of 1 to 6 days/week, and 20% reported daily use. A duration of NSAID use <5 years was reported by 17% of patients and a use of ≥5 years was reported by 18%. Regular prediagnosis NSAID use (1-3 days/week, 4-6 days/week, and daily) versus none was associated with improved overall survival (OS) (HR, 0.71; 95% CI, 0.53-0.95) and CRC-specific survival (HR, 0.58; 95% CI 0.40-0.84) after adjustment for clinically relevant factors. Prediagnosis NSAID use ≥5 years (vs none) was found to be associated with improved OS (HR, 0.55; 95% CI, 0.37-0.84) and CRC-specific survival (HR, 0.40; 95% CI, 0.23-0.71) in adjusted analyses.

CONCLUSIONS:

When used regularly or over a prolonged duration before CRC diagnosis, NSAIDs are associated with decreased mortality among female CRC patients. Cancer 2009. © 2009 American Cancer Society.

During the year 2007, an estimated 112,340 new cases of colon cancer and 41,420 cases of rectal cancer were diagnosed in the United States, making colorectal cancer (CRC) the third most common cancer among US men and women.1 A wide body of evidence links use of nonsteroidal anti-inflammatory drugs (NSAIDs) to a reduced risk of CRC. First, rodent models have demonstrated that NSAIDs inhibit intestinal cancer development.2-8 Consistent with experimental data, clinical trial results involving patients with familial adenomatous polyposis have demonstrated that the NSAIDs sulindac,9-11 celecoxib,12 and rofecoxib13 prevent colorectal adenomas. Aspirin has been implicated in the regression and prevention of premalignant colorectal adenomas in clinical trials,14, 15 and various NSAIDs have been associated with a reduced risk of CRC in the vast majority of cohort and case-control studies.16-23 In addition to these studies demonstrating preventive effects on CRC risk, 2 important studies suggest that NSAID use after a diagnosis of CRC may improve survival.24, 25 In the male Health Professionals Study, for example, aspirin was associated with a decreased risk of CRC or adenoma, and also with a decreased risk of advanced stage CRC and fatal CRC.24 However, to our knowledge, these reports are limited to a cohort study of males, and an unpublished, secondary analysis of clinical trial data restricted to stage III colon cancer patients. Survival outcomes from key therapeutic trials involving NSAIDs among CRC patients have not to our knowledge been reported to date. Thus, a clear determination of potential survival benefits associated with NSAID use after CRC diagnosis has not been established. By using data from women in the California Teachers Study (CTS) cohort, we investigated the correlation between NSAIDs and mortality among CRC patients.

MATERIALS AND METHODS

Study Population

The CTS is a prospective cohort of current, recent, and retired female public school teachers and administrators who were members of the California State Teachers Retirement System at the time of study inception in 1995.26 Cohort members completed a detailed self-administered baseline questionnaire. A total of 133,479 women comprise the CTS cohort. The CTS has been approved by the institutional review boards of the State of California, the University of Southern California, the University of California at Irvine, the Northern California Cancer Center, and the City of Hope National Medical Center.

To identify incident invasive colorectal cancers that occurred during follow-up of the CTS, we first delineated a cohort of eligible women. We excluded in a hierarchical manner women who 1) moved out of the state or whose history of invasive cancer during the follow-up period was unknown (n = 10,443), 2) were aged >85 years at baseline (n = 1674), 3) had developed a prior malignancy (n = 169); 4) lacked data regarding family history of CRC (n = 3844), and 5) lacked data concerning the frequency of NSAID use (n = 3106). The resulting cohort was comprised of 114,243 women.

Case Ascertainment and Follow-Up

Incident diagnoses of invasive carcinoma of the colon and rectum (International Classification of Diseases for Oncology [ICD-O] topography codes C18.0-18.9, C19.9, and C20.9) were identified through annual linkages with the California Cancer Registry. The California Cancer Registry also provided details regarding anatomic subsite within the colorectum and extent of disease. As part of a state mandate for cancer reporting, the California Cancer Registry receives reports on >99% of all cancer diagnoses occurring in California residents.27 Overall, 621 CTS participants considered eligible for analysis were diagnosed with invasive colorectal carcinoma during follow-up: 427 with colon cancer, 173 with rectal cancer, and 21 with large bowel cancer of unspecified site.

Follow-up was calculated from the date of diagnosis of colorectal cancer to death, or the end of the follow-up period on December 31, 2005. Treatment during the first course of therapy was ascertained using available California Cancer Registry data describing receipt of surgical resection and/or radiotherapy. Cause of death was recorded according to the ICD-O criteria in effect at the time of death.28

Assessment of Medication Use

The self-administered questionnaire asked “Have you taken any of the following medications regularly (at least once a week)?” Available options included 1) aspirin (Anacin, Bufferin, or Excedrin) or 2) ibuprofen (Advil, Motrin, or Nuprin). For positive responses, participants indicated the average number of days per week (frequency) (1-3, 4-6, or 7) and the total duration of use in number of years (<1, 1, 2, 3-4, 5-9, or ≥10). For the primary analyses, frequency of medication use was categorized into 2 categories: no use or regular use (1-3 days per week, 4-6 days per week, or daily use). Duration was categorized as no use, duration <5 years, or duration ≥5 years. The reconstituted NSAID variables in the CTS were used for all analyses, because these accounted for combinations of aspirin and ibuprofen use.

Assessment of CRC Risk Factors

Family history of CRC was defined as having a first-degree relative diagnosed with CRC (parent, sibling, or child). Information regarding other potential CRC risk factors was collected in the baseline questionnaire, including race/ethnicity; personal history or family history of colorectal polyps; body mass index (BMI); recreational physical activity; lifetime tobacco exposure; oral contraceptive use; hormone replacement use; years since last mammogram or Papanicolaou smear; and total daily intake of calcium, folate, fiber, iron, and energy (in kilocalories).29

Statistical Analyses

We used multivariate Cox proportional hazards regression analyses to estimate the association between NSAID or aspirin-only use and overall mortality or CRC-specific mortality. All models included the categoric variable for NSAID or aspirin-only use. Adjustment was made for variables known to predict survival in CRC: age, stage of disease at the time of presentation, disease site within the colorectum (ie, colon or rectum), and family history of CRC. Other potential confounders were not included in models, because their inclusion did not change risk estimates by >5%. Local, regional, and advanced (ie, remote) stages at presentation were categorized according to the Surveillance, Epidemiology, and End Results summary stage indicator when available.30P values were not adjusted for multiple comparisons. All statistical analyses were performed using SAS statistical software (version 9.2; SAS Institute, Inc., Cary, NC).

RESULTS

NSAID Frequency Before CRC Diagnosis

Among the 621 CRC cases identified after formation of the cohort, the observed median duration between baseline NSAID assessment and CRC diagnosis was 5.2 years (range, 0-10.1 years; mean duration, 5.2 years), and the median follow-up duration after CRC diagnosis was 2.8 years (range, 0-9.8 years; mean duration, 3.4 years). Table 1 presents the baseline characteristics of CRC cases in the CTS stratified by frequency of NSAID use (no use vs regular use [defined as daily use, use 4-6 times per week, or use 1-3 times per week]). As can be seen in Table 1, these groups were similar across most of the variables considered. Regular NSAID users had a higher BMI than nonusers, and were more likely to have undergone a recent mammogram. NSAID frequency usage and duration of NSAID use among CRC cases are presented in Table 2. The majority (64%) of cases did not use NSAIDs regularly, whereas 20% were daily NSAID users. Among NSAID users, 39% (ie, 87 of 224 individuals) reported use for ≥10 years, and 50% (ie, 111 of 224 individuals) reported use for at least 5 years.

Table 1. Selected Baseline Characteristics in Relation to NSAID Use Among 621 Colorectal Cancer Cases in the California Teachers Study (1995-2005)

	Any Regular NSAID Use Reported at Baseline,* n=224	No Regular NSAID Use Reported at Baseline, n=397	P
NSAID indicates nonsteroidal anti-inflammatory drug; SD, standard deviation; CRC, colorectal cancer; BMI, body mass index; Pap, Papanicolaou. *, † Variable based on response to question regarding ever having used NSAIDs daily, 4 to 6 times/week, and 1 to 3 times/week versus never having used them at least once weekly.
Age at diagnosis, y
Mean (±SD)	65.0±11.6	63.2±12.6	.08
Race
Caucasian	196 (88%)	353 (89%)	.68
African American	8 (4%)	19 (5%)
Others	20 (8%)	25 (6%)
Personal history of polyps	15 (7%)	35 (9%)	.35
Family history of polyps	18 (8%)	42 (11%)	.30
Family history of CRC in first-degree relative	30 (13%)	41 (10%)	.25
BMI, kg/m^2 (±SD)	27.0±6.0	25.2±5.1	.0004
Total kcal per day, mean (±SD)	1518±564	1503±519	.78
Total daily calcium intake, mean mg (±SD)	743±416	698±430	.23
Total daily iron intake, mean mg (±SD)	10.9±4.2	10.9±4.1	.96
Total daily fiber intake, mean g (±SD)	14.9±6.4	14.8±6.6	.94
Total daily folate intake, mean μg (±SD)	300±117	303±122	.78
Strenuous and moderate lifetime physical activity, mean h/wk (±SD)	4.0±4.2	4.0±4.1	.93
Active and passive lifetime tobacco exposure
None	27 (13%)	60 (16%)	.72
Lifetime tobacco exposure in pack-years
Quartile 1 (0.05-10.04)	25 (12%)	52 (14%)
Quartile 2 (10.05-17.04)	34 (16%)	59 (15%)
Quartile 3 (17.05-31.54)	57 (27%)	97 (25%)
Quartile 4 (≥31.55)	72 (33%)	113 (30%)
Years since last mammogram
Never	6 (3%)	35 (9%)	.009
<1 y	127 (57%)	222 (57%)
1-2 y	71 (32%)	92 (24%)
≥3 y	17 (8%)	41 (11%)
Had mammogram previously, but at unspecified time	1 (0.5%)	1 (0.3%)
Years since last Pap smear
Never	1 (0.5%)	5 (1%)	.57
<1 y	116 (53%)	200 (54%)
1-2 y	66 (30%)	97 (26%)
≥3 y	33 (15%)	69 (19%)
Had previously, but at unspecified time	1 (0.5%)	2 (0.5%)
Tumor site
Colon	156 (72%)	271 (72%)	.87
Rectum	62 (28%)	111 (28%)
Histology
Adenocarcinoma	181 (82%)	315 (80%)	.58
Carcinoma	14 (6%)	18 (5%)
Mucinous carcinoma	24 (11%)	54 (14%)
Not otherwise specified	3 (1%)	7 (2%)
Stage of disease at presentation
Local	102 (47%)	163 (43%)	.48
Regional	76 (35%)	138 (36%)
Remote	39 (18%)	82 (21%)
Tumor grade
1	14 (7%)	35 (10%)	.38
2	132 (69%)	223 (65%)
3	42 (22%)	83 (24%)
4	4 (2%)	3 (1%)
Primary treatment involved surgery	207 (92%)	361 (91%)	.53
Primary treatment involved radiotherapy	17 (8%)	43 (11%)	.19
Primary treatment involved chemotherapy	19 (8%)	26 (7%)	.37

Table 2. Patterns of NSAID Frequency and Duration of Use Among 621 CRC Cases in the CTS

NSAID indicates nonsteroidal anti-inflammatory drug; CRC, colorectal cancer; CTS, California Teachers Study.
Detailed NSAID Frequency
No regular NSAID use	397 (64%)
1-3 d/wk	75 (12%)
4-6 d/wk	28 (5%)
Daily	121 (20%)
Duration of NSAID use
No regular NSAID use	397 (64%)
<1 y	32 (5%)
1 y	18 (3%)
2 y	22 (4%)
3-4 y	33 (5%)
5-9 y	24 (4%)
≥10 y	87 (14%)
Unknown duration	8 (1%)

Association Between NSAID Use and CRC Survival

Cause of death analysis revealed that 222 (36%) of the 621 identified CRC cases died during the study period. Of these 222 deaths, 145 (65%) were attributed to CRC. Fifteen of the 222 deaths (6.8% of all deaths) were because of heart disease.

The primary results are presented for analysis of regular NSAID use compared with no regular NSAID use. In the Cox models, after adjustment for age at diagnosis, family history of CRC in a first-degree relative, disease site (colon or rectum), stage of disease at diagnosis, time from NSAID assessment to CRC diagnosis, and treatment with surgery (yes/no), regular NSAID use was found to be associated with a statistically significant decreased risk of death compared with no regular NSAID use (Table 3). For CRC-specific survival, the hazards ratio (HR) was 0.58 (95% confidence interval [95% CI], 0.40-0.84); for overall survival, the HR was 0.71 (95% CI, 0.53-0.95). Restricting analysis to colon cancer cases, the multivariate relative hazards for regular NSAID use versus no use revealed significant differences in overall survival (HR, 0.68; 95% CI, 0.48-0.97) and CRC-specific survival (HR, 0.57; 95% CI, 0.36-0.90). No significant differences were observed when the analysis was limited to patients with rectal cancer.

Table 3. Univariate and Multivariate Adjusted* Overall Survival Analysis and CRC-Specific Survival Analysis for CRC Cases by Self-Reported Frequency of NSAID and Aspirin Use†

	Reported NSAID Frequency		Reported Aspirin frequency
	No Regular NSAID Use	Regular NSAID Use	No Regular Aspirin Use	Regular Aspirin Use‡
CRC indicates colorectal cancer; NSAID, nonsteroidal anti-inflammatory drug; HR, hazards ratio; 95% CI, 95% confidence interval. * Includes adjustment for age (in years), family history of CRC in a first-degree relative, tumor site (colon or rectum), stage of disease at diagnosis, time from NSAID assessment to CRC diagnosis, and treatment with surgery. † Includes NSAID use daily, 4 to 6 times/week, and 1 to 3 times/week. ‡ Data regarding aspirin frequency were missing for 1 case.
Overall mortality
No. of events	150	72	167	54
No. at risk	397	224	456	164
Unadjusted HR (95% CI)	1 (reference)	0.83 (0.63-1.10)	1 (reference)	0.89 (0.65-1.21)
Adjusted HR (95% CI)*	1 (reference)	0.71 (0.53-0.95)	1 (reference)	0.74 (0.54-1.01)
CRC-specific mortality
No. of events	103	42	113	32
No. at risk	397	224	456	164
Unadjusted HR (95% CI)	1 (reference)	0.71 (0.49-1.01)	1 (reference)	0.77 (0.52-1.14)
Adjusted HR (95% CI)*	1 (reference)	0.58 (0.40-0.84)	1 (reference)	0.62 (0.41-0.94)

Results for CRC cases based on the duration of NSAID use are presented in Table 4. Colorectal cancer patients who reported using NSAIDs for at least 5 years before diagnosis were found to have a statistically significant decreased risk of overall mortality after adjustment for clinically relevant factors compared with those reporting no NSAID use (HR, 0.55; 95% CI, 0.37-0.84). Similarly, a statistically significant decreased risk of CRC-specific mortality was observed in the multivariate analysis for CRC patients reporting at least 5 years of NSAID use before diagnosis compared with patients reporting no NSAID use (HR, 0.40; 95% CI, 0.23-0.71). No statistically significant associations with overall or CRC-specific mortality were observed for cases reporting <5 years duration of NSAID use compared with those reporting no NSAID use. Among colon cancer patients, at least 5 years duration of NSAID use before diagnosis was associated with a significant reduction in CRC-specific mortality (HR, 0.46; 95% CI, 0.23-0.94), but not overall mortality (HR, 0.65; 95% CI, 0.40-1.05). No significant differences were observed for subset analysis of NSAID duration among rectal cancer patients. Time from NSAID assessment (at baseline) to CRC diagnosis was not found to be significantly associated with mortality (data not shown). Results shown in Tables 3 and 4 based on aspirin use rather than all NSAIDs are consistent with the results for all NSAID use.

Table 4. Univariate and Multivariate Adjusted* Overall Survival Analysis and CRC-Specific Survival Analysis for All Colorectal Cancer Cases by Self-Reported Duration of NSAID and Aspirin Use†

	Reported NSAID Duration			Reported Aspirin Duration
	None	<5 Years	≥5 Years	None	<5 Years	≥5 Years
CRC indicates colorectal cancer; NSAID, nonsteroidal anti-inflammatory drug; HR, hazards ratio; 95% CI, 95% confidence interval. * Includes adjustment for age (in years), family history of CRC in a first-degree relative, tumor site (colon or rectum), stage of disease at diagnosis, time from NSAID assessment to CRC diagnosis, and treatment with surgery. † Data were missing regarding NSAID and aspirin duration for 8 cases.
Overall mortality
No. of events	150	30	42	167	26	24
No. at risk	397	105	111	456	65	92
Unadjusted HR (95% CI)	1 (reference)	0.94 (0.66-1.34)	0.67 (0.45-0.99)	1 (reference)	1.06 (0.70-1.60)	0.71 (0.46-1.08)
Adjusted HR (95% CI)†	1 (reference)	0.83 (0.58-1.19)	0.55 (0.37-0.84)	1 (reference)	1.02 (0.67-1.55)	0.53 (0.34-0.83)
CRC-specific mortality
No. of events	103	24	24	113	18	11
No. at risk	397	105	111	456	65	92
Unadjusted HR (95% CI)	1 (reference)	0.87 (0.56-1.35)	0.50 (0.29-0.87)	1 (reference)	1.08 (0.66-1.78)	0.48 (0.26-0.88)
Adjusted HR (95% CI)*	1 (reference)	0.73 (0.46-1.15)	0.40 (0.23-0.71)	1 (reference)	1.03 (0.62-1.72)	0.33 (0.18-0.63)

DISCUSSION

In this analysis of women diagnosed with a first primary invasive CRC after enrolling in the large CTS cohort, we found that prediagnosis regular NSAID use and prolonged duration of NSAID use were independently associated with decreased overall mortality and CRC-specific mortality. Prolonged NSAID duration (>5 years) reported among women with subsequent diagnosis of CRC was associated with a 50% (unadjusted) and 60% (adjusted) lower risk of CRC-specific mortality. Aspirin-only survival effects were consistent with the effects observed for NSAIDs. If confirmed in other studies, these findings may have important implications for tertiary cancer prevention.

Because this was an analysis of prediagnosis NSAID use, it is possible that the improvement in survival associated with NSAID use may be a consequence of prediagnosis NSAID use leading to the development of more indolent, less aggressive cancers. Alternatively, prediagnosis NSAID use may simply reflect postdiagnosis NSAID use, and potentially beneficial effects of NSAID use on survival after CRC diagnosis. This is supported by data regarding our surviving CRC patients, which indicated that 82% who reported regular NSAID use in the prediagnosis questionnaire also reported regular NSAID use in a later questionnaire that was completed in 2007 (data not shown). Despite numerous studies reporting NSAID effects on CRC risk, to the best of our knowledge only limited data are available that describe NSAID effects on survival among individuals with established colon or rectal cancer. In the male Health Professionals Study, aspirin was associated with decreased risk of advanced stage CRC and fatal CRC.24 In a prospective analysis of 803 stage III colon cancer patients who received adjuvant chemotherapy (leucovorin-modulated 5-fluorouracil with or without irinotecan), consistent aspirin use among 74 patients was associated with statistically significant decreases in HRs for disease recurrence and overall mortality.25 Cyclooxygenase (COX)-2 overexpression is important in the progression of the colon adenoma-carcinoma sequence.31 Both the COX-2 selective and nonselective NSAIDs have been shown to induce apoptosis in colon cancer model systems.3, 4, 32 Aspirin's preventive effect on CRC carcinogenesis is limited to lowering the risk of COX-2–overexpressing tumors.33 A phase 1 clinical trial using celecoxib in combination with leucovorin-modulated 5-fluorouracil as treatment for advanced colorectal adenocarcinoma suggests that this regimen is safe and worthy of further investigation.34 Various NSAIDs have been used in phase 1, 2, and 3 clinical trials in combination with either radiotherapy or chemotherapy for regional and metastatic colorectal cancer.34-36 However, the majority of these trials are currently in progress, and survival outcomes are not yet reported. COX-2 inhibitors have been described as “cotherapeutic agents” in the treatment of colorectal cancer,36 because of their primary effect of inhibiting tumor growth rather than causing tumor regression.35, 37 Therefore, incorporating COX-2 inhibitors or nonselective NSAIDs into combination therapy with chemotherapy, radiotherapy, other investigational chemopreventive agents, or targeted therapies such as vascular endothelial growth factor inhibitors (eg, bevacizumab) and epidermal growth factor inhibitors (eg, cetuximab, panitumumab) may prove beneficial.

NSAIDs exert their antiproliferative effects on colonic cells by inhibiting prostaglandin synthesis through reversible binding to COX and through other mechanisms. Aspirin has been noted to reduce the recurrence of adenomatous polyps in patients at moderate risk and with acceptable toxicity, but such is not the case for COX-2–selective inhibitors. Several randomized trials have demonstrated that COX-2–selective NSAIDs result in a substantial reduction in adenoma recurrence, but at the cost of significantly increased cardiovascular events.38, 39 Thus, COX-2–selective NSAIDs cannot be recommended routinely for CRC prevention. Whether similar adverse results will be found with less selective NSAIDs awaits the results of ongoing trials. In light of these data, the US Preventive Services Task Force has recommended that, despite evidence of efficacy, NSAIDs are not warranted for CRC prevention among average-risk individuals because of cardiovascular and gastrointestinal toxicities.40 Of note, only 6.8% of deaths in the current study were attributed to heart disease.

Family history itself was not found to be associated with a decreased risk of overall mortality or CRC-specific survival in the current study. Despite convincing evidence that a family history of CRC is a risk factor for colorectal polyps and cancer,41-44 conflicting reports have emerged regarding the association of family history with survival after CRC diagnosis. Consistent with our findings, a family history of CRC was not found to be associated with survival after CRC diagnosis in the Melbourne CRC cohort study45 or in a study from the Utah Population Database.46 However, there is not a consensus on this issue, because new studies have emerged demonstrating survival benefits for colon cancer patients with a family history of CRC compared with those without.47-50 It is possible that unique cohort characteristics (ie, stage, sex, ethnicity) and statistical power account for the different observations concerning the effects of CRC family history on survival after diagnosis of CRC.

The current study was based on a prospective study design such that NSAID reporting predated CRC diagnosis, with cases accrued during a median cohort follow-up of 10.1 years. Self-reported NSAID use was obtained from the baseline questionnaire, whereas survival effects may be more closely related to NSAID use after CRC diagnosis. Among our surviving CRC patients, we have additional data regarding NSAID use obtained in 2007; although 82% of those who reported using NSAIDs at baseline continued their use, 38% of the nonusers from the prediagnosis questionnaire reported regular NSAID use in the later questionnaire. These subsequent NSAID assessments were collected in the final year of follow-up for mortality, meaning that deceased cases could not provide relevant data regarding NSAID use. Thus, we cannot determine how changing patterns of NSAID use over time influenced CRC mortality. We also recognize that our findings could reflect effects of lead time bias if NSAID use were associated with earlier diagnosis because of early symptomatology (eg, rectal bleeding). Our analyses of rectal cancer were limited in sample size and therefore in the number of deaths occurring in this group, resulting in low statistical power to detect effects of NSAID use on survival. The study has several other important limitations, including a lack of information regarding screening practices such as fecal occult blood testing or the use of endoscopy (colonoscopy or flexible sigmoidoscopy). It is recognized that screening may be associated with NSAID use, and thus differences in screening could theoretically explain the observed associations. In the current analysis, there were observed differences in the screening rates for mammograms but not Papanicolaou smears between CRC patients reporting regular NSAID use and those reporting no regular NSAID use, but stage of CRC at diagnosis did not appear to vary by NSAID use (Table 1). Detailed NSAID history was not available, including information regarding the specific type of NSAID and dosage, information regarding NSAIDs other than aspirin or ibuprofen, and information concerning current or former use. In addition, our study was underpowered to investigate risk differences based on tumor subsite location within the colon or rectum.

The findings of the current study cannot support a general recommendation for NSAID use as tertiary prevention of colorectal cancer among CRC patients. However, they do justify future validation studies to better understand prediagnostic and postdiagnostic NSAID use among CRC cases and underscore the importance of performing clinical trials of NSAID tertiary prevention in CRC patients, such as those being planned in the cooperative group setting. These findings clearly suggest that prior NSAID use patterns are an important variable to consider when planning and interpreting future trials, and that completed therapeutic trials demonstrating benefit may need to revisit whether prior NSAID use patterns were adequately addressed in the study design and analysis.

Acknowledgements

We thank the California Teachers Study Steering Committee members who are responsible for the formation and maintenance of the cohort within which this study was conducted but are not included as authors of the current article: Gretchen Agha, Rosemary Cress, Pamela Horn-Ross, Rich Pinder, Peggy Reynolds, and Dee W. West.

CONFLICT OF INTEREST DISCLOSURES

Supported by Grants R01 CA77398 and R25 CA85771 from the National Cancer Institute, Grant P30 ES 07,048 from the National Institute of Environmental Health Sciences, and Contract 97-10,500 from the California Breast Cancer Research Fund.

The collection of the cancer incidence data used in this study was supported by the California Department of Health Services as part of the statewide cancer reporting program mandated by California Health and Safety Code Section 103885; the National Cancer Institute's Surveillance, Epidemiology, and End Results Program under Contract N01-PC-35,136 awarded to the Northern California Cancer Center, Contract N01-PC-35,139 awarded to the University of Southern California, and Contract N02-PC-15,105 awarded to the Public Health Institute; and the Centers for Disease Control and Prevention's National Program of Cancer Registries, under Agreement #U55/CCR921930-02 awarded to the Public Health Institute. The ideas and opinions expressed herein are those of the authors, and endorsement by the State of California, Department of Health Services, the National Cancer Institute, the Centers for Disease Control and Prevention, and/or the Genetic Epidemiology Research Institute of the University of California, Irvine is not intended nor should be inferred.