Aspirin as an Adjunct to Screening for Prevention of Sporadic Colorectal Cancer: A Cost-Effectiveness Analysis

Colorectal cancer is the second leading cause of cancer-related death in the United States, with 129 400 new cases and 56 600 deaths estimated to have occurred in 1999 (1). Most cases of colorectal cancer arise from adenomatous polyps (2). Screening has been shown to decrease colorectal cancer incidence and mortality rates (3, 4), and various decision analyses support its cost-effectiveness (510). Despite demonstrated clinical and economic benefits, however, use of screening tests is low (1113). In a large population-based study, 19.8% of respondents reported having fecal occult blood testing during the preceding year and 30.4% reported having sigmoidoscopyproctoscopy during the preceding 5 years; 9.5% had had both tests (13). In a national strategy to decrease the burden of colorectal cancer, aspirin chemoprevention is a potentially powerful adjunct to screening. Numerous casecontrol (1421) and cohort studies (2226) support the hypothesis that aspirin and other nonsteroidal anti-inflammatory drugs (NSAIDs) decrease the incidence of colorectal adenoma and colorectal cancer and the mortality rate from colorectal cancer by 30% to 65%. Many animal models show that NSAIDs have an antitumor effect in the colon (27). However, the Physicians' Health Study, a randomized study of male physicians that was not designed to examine colorectal cancer specifically, found no decrease in colorectal cancer incidence when 325 mg of aspirin on alternate days was compared with placebo (28). These results may conflict because of the specific dose and duration used and the patients studied. Aspirin is of proven benefit in secondary prevention of cardiovascular morbidity and death (29) and in primary prevention of myocardial infarction in men (30). However, its use for primary cardiovascular prevention remains controversial, in part because mortality prevention is unproven and no randomized studies in women exist (29). Although it is well recognized that aspirin may cause major complications (3035), prophylactic aspirin is prescribed on the basis of cardiovascular considerations. In contrast, aspirin is unlikely to be prescribed for colorectal cancer prevention until better estimates of its benefits and risks become available. Since such data are unlikely to be available for the general U.S. population in the near future, we constructed a decision analytic model to explore the incremental clinical and economic effects of aspirin when it is added to an accepted screening program for sporadic colorectal cancer. Accepted screening programs consisted of flexible sigmoidoscopy every 5 years and yearly fecal occult blood testing (FS/FOBT) (8, 36) or screening colonoscopy every 10 years (COLO), an emerging alternative screening strategy (710, 37). Our primary aim was to clarify whether aspirin chemoprevention might prove to be a cost-effective adjunct to screening. In secondary analyses, we examined the potential role of aspirin chemoprevention in an unscreened population and the effects of screening when added to existing aspirin use. Methods Literature Review We searched MEDLINE from 1980 to 1999 for English-language literature that provided data on colorectal cancer, screening, aspirin chemoprevention, and aspirin-related complications. Model inputs were based on the literature search and on rigorous literature reviews by a multidisciplinary expert panel from the Agency for Health Care Policy and Research (8) and by the Office of Technology Assessment (7) (Table 1). Table 1. Inputs in the Cost-Effectiveness Model Decision Analytic Model We created a Markov model using DATA 3.0 (TreeAge Software Inc., Williamstown, Massachusetts); Excel 2000 (Microsoft Corp., Redmond, Washington) was used for analysis. The Markov model estimated the clinical and economic consequences of six strategies: natural history (no aspirin or screening), FS/FOBT, COLO, aspirin alone (ASA), FS/FOBT and aspirin (FS/FOBT/ASA), and colonoscopy and aspirin (COLO/ASA). Beginning at 50 years of age, patients progressed through the model for 30 1-year cycles; principal states were normal, polyp, cancer (localized, regional, or distant), and dead. The Markov model is discussed in detail in the Appendix. To reflect the effect of aspirin in the general population and not in an individual, we varied the percentage of persons in the population who adhered to screening. Using population-based data (13), we selected a screening adherence rate of 25% to reflect current practice. The general population, men, and women were examined separately. Natural History of Sporadic Colorectal Cancer We constructed a model of the natural history of sporadic colorectal cancer in patients at average risk for the disease. Epidemiologic data on colorectal cancer were derived from the Surveillance, Epidemiology, and End Results (SEER) program (38). According to results of autopsy studies, adenomatous polyp prevalence was assumed to increase from 15% at 50 years of age to 47% at 75 years of age (16% to 54% in men; 15% to 40% in women) (3941). We assumed that 90% of tumors develop from polyps and that cancer progresses from localized to regional (2 years in each state) to disseminated, with symptomatic presentation by stage as reflected in SEER data. Age-specific transition probabilities were calculated between normal, polyp, and cancer to yield polyp and cancer rates derived from the literature. Age-specific noncolorectal cancer mortality rates were derived from U.S. life tables (46). Effect of Colorectal Cancer Screening We selected FS/FOBT as an accepted strategy endorsed by published guidelines (8, 36). In the model, any positive test result was followed by colonoscopy and, if necessary, polypectomy. If the results of colonoscopy were normal, screening resumed in 10 years. After an adenoma was detected, surveillance colonoscopy was performed every 5 years. Although specific patients with high-risk or low-risk adenomas may undergo surveillance at different intervals, we chose 5 years to reflect the average surveillance interval for the population. Colonoscopy was performed at cancer diagnosis and 3 years and every 5 years thereafter (8). As long as results of screening tests remained negative, FS/FOBT was continued. Emerging data from clinical studies and decision analyses suggest that colonoscopy is a cost-effective screening tool (710, 37). Therefore, COLO was examined as an alternative screening strategy. In the model, colonoscopy was performed every 10 years as long as no adenoma or cancer was detected. Surveillance after adenoma detection was the same as that described for FS/FOBT. Effect of Aspirin Aspirin, 325 mg/d, was added to the screening and natural history strategies. In the base case, aspirin was estimated to reduce colorectal cancer risk by 30% through equal reductions in adenoma incidence and progression to cancer (1426). Aspirin may increase occult gastrointestinal blood loss (4753), which could affect the FS/FOBT strategy. In the base case, we assumed that aspirin did not affect fecal occult blood testing because increases in blood loss have been reported only with high doses (4853) and low doses have been reported to have no effect on fecal occult blood testing (5456). The protective effect of NSAIDs is not explained by earlier detection of adenomas or cancer (19, 26); however, increased sensitivity and decreased specificity of FS/FOBT in patients receiving aspirin were assessed in the sensitivity analysis. In the model, aspirin-related complications were followed by aspirin withdrawal and return to baseline risk for colorectal cancer. Using data from trials of low-dose aspirin, we calculated rates of excess serious complications in patients taking aspirin. The rates were 1.4 per 10 000 person-years in patients younger than 65 years of age and 28 to 40 per 10 000 person-years in older patients with cerebrovascular disease (30, 3234). The 1-year prevalence of serious gastrointestinal NSAID-related complications is estimated to increase eightfold after 65 years of age (31). Thus, we selected a base-case rate of 2 per 10 000 person-years before 65 years of age and 16 per 10 000 person-years thereafter for serious aspirin-related complications. We modeled mortality rates of 2% to 8% after a serious aspirin-related complication (4244). We did not include changes in cardiovascular outcomes in the base case for several reasons. First, the role of aspirin in primary cardiovascular prevention is controversial, largely because prevention of death is unproven (29). A recent decision analysis suggested that aspirin may be harmful in cohorts at low risk for cardiovascular events and may be only minimally beneficial as risk increases (57). Second, aspirin may increase the incidence of hemorrhagic stroke (30), and the balance of risks and benefits is unknown, particularly in women (29). Third, in patients with cardiovascular indications for aspirin, a potential effect on colon cancer is unlikely to affect prescribing decisions. Thus, we chose to examine the incremental benefit of colorectal cancer screening in a population already receiving aspirin. In the sensitivity analysis, we examined the effect of aspirin by assuming that it reduced cardiovascular mortality. Cardiovascular death accounts for 30% to 42% of deaths, depending on age (46). Costs Procedure costs were derived from Medicare fee schedules and include professional fees and median procedure reimbursement. We used the wholesale cost of aspirin at the University of Michigan pharmacy, Ann Arbor, Michigan. Complication costs were derived from relevant diagnostic related groups (45). Costs for care of stage-specific colon cancer were taken from reports to the National Cancer Institute (7). All costs are in 1998 U.S. dollars (Table 1). Outcomes Colorectal cancer cases by stage, deaths by cause, average life-years after age 50 years, total costs, and itemized costs were determined for each strategy. Costs and life-years were discounted