Adenomatous polyps are precursors of most colorectal cancers, and their prevalence increases with age. The chance of detecting adenomatous polyps at colonoscopy is generally independent of the indication for the procedure [5]. The practice of removing polyps at colonoscopy is based on the assumption that their removal prevents progression to cancer. This concept, often called the adenoma-carcinoma sequence, has never been directly proved [6]. The most compelling evidence substantiating the theory has come from the National Polyp Study, a cohort study in which the incidence of colorectal cancer in more than 1418 patients who had a complete colonoscopy and from whom one or more adenomatous polyps was removed was compared with the incidence in three reference groups [7]. Colonoscopic polypectomy resulted in an incidence of colorectal cancer that was 76% to 90% lower than expected. The results of this landmark study not only confirmed the adenoma-carcinoma sequence hypothesis, but also served to justify the widely held practice of endoscopic screening for colorectal cancer, with its end point of colonoscopic polypectomy.

Although endoscopic screening for colorectal cancer has been practiced for decades by many physicians, the current era of health care reform has created an environment in which medical practices, no matter how well reasoned or well intentioned, must be subjected to vigorous study to demonstrate their efficacy. However, a prospective study in which patients with colorectal polyps are randomly assigned either to receive polypectomy or simply surveillance would be neither ethical nor feasible [7]. Thus, indirect evidence of the protective effect of polypectomy must suffice.

In this issue, Muller and Sonnenberg [8] used the well-established case–control method to determine the ability of flexible sigmoidoscopy, colonoscopy, and polypectomy to prevent colorectal cancer in a population of more than 32 000 veterans. Of 16 351 persons in the case population, 1051 had endoscopic procedures involving the large intestine before colorectal cancer was diagnosed. Of the 16 351 persons in the control population, 1166 had endoscopic procedures, a difference of only 115 procedures. Despite the relatively small difference in the number of endoscopic procedures done in the two large study groups, the results showed that endoscopic procedures of the large intestine in conjunction with polyp removal reduced the risk for developing colorectal cancer by 50%.

The study had the practical limitations of the case–control design. In addition, as noted by the authors, the patient population consisted entirely of veterans served by the Department of Veterans Affairs medical system, a population that may not adequately reflect the health behavior and medical care of other Americans. Nevertheless, the study contributes significantly to our current understanding and management of colorectal cancer. The power and importance of the study is its large patient population followed within a relatively closed system. As the authors have noted, the data obtained by such outcomes research test the effectiveness of medical service that average physicians provide to average patients. The results of the study add substantially to the already strong evidence that flexible endoscopy and polypectomy reduce the incidence of and mortality from colorectal cancer.

The study by Muller and Sonnenberg [8] must be evaluated in conjunction with the case–control study reported by Selby and colleagues [3], in which 261 patients with fatal cancer of the rectum and distal colon were compared with 868 age-matched controls. The results of that study showed that screening sigmoidoscopy using the rigid sigmoidoscope reduced mortality from cancer in the region of the intestine examined by the rigid instrument by 60%. The studies by Muller and Sonnenberg [8] and by Selby and colleagues [3] differ in many respects. First, the instruments used involve a different depth of insertion; the rigid sigmoidoscope is limited to evaluating the rectum and distal sigmoid colon. Second, Muller and Sonnenberg [8] chose as their end point the diagnosis of colorectal cancer, whereas Selby and colleagues chose mortality from colorectal cancer. It may be assumed that not all patients with a diagnosis of colorectal cancer in the study by Muller and Sonnenberg [8] died of their disease. Indeed, the reduction in mortality resulting from the screening of asymptomatic persons that has been observed in other studies has been attributed to finding potentially curable Dukes A cancers as well as to removal of malignant precursors. Thus, a reduction in mortality from colorectal cancer may be expected in the population studied by Muller and Sonnenberg [2].

The ideal screening test for colorectal cancer should be sensitive (few false-negative results) and specific (few false-positive results), simple to perform (for both patient and physician), and cost-effective. Although colonoscopy is a powerful diagnostic tool for detecting adenomatous polyps when present as grossly elevated mucosal lesions, recent reports have identified a few "flat or depressed" adenomas and cancers that often require adjunctive techniques for detection; this suggests that colonoscopy as currently performed is not infallible [9-11]. Colonoscopy is also expensive, as measured by direct and indirect costs, and its cost-effectiveness and public acceptance as a screening tool have been challenged.

In the future, screening for colorectal cancer may rely on methods far more sensitive and specific than flexible endoscopy performed in average-risk persons or fecal occult blood testing, in which a positive result indicates gastrointestinal bleeding and not necessarily the presence of neoplasia [12]. The success of research aimed at determining the molecular basis for colorectal cancer suggests that new screening strategies may be based on molecular markers for disease. Study of hereditary nonpolyposis colon cancer, the most common form of inherited colon cancer, has led to the identification of a series of defective DNA repair genes (hMSH2, hMLH1, hPM1, and hPM2) involved in maintaining DNA constancy during replication, which is sometimes referred to as "proofreading" [13, 14]. The defective genes appear to be carried in a high percentage of patients with hereditary nonpolyposis colon cancer and may be found in as many as 15% of patients with sporadic colorectal cancer. An example of screening targeted to such high-risk persons was provided recently by Jarvinen and colleagues [15], who compared the mortality and incidence of colorectal cancer in asymptomatic at-risk members of 22 families with hereditary nonpolyposis colon cancer [15]. The results showed that screening at 3-year intervals by colonoscopy (or barium enema with flexible sigmoidoscopy) reduced the incidence of colorectal cancer in these high-risk patients by 62% (4.5% of screened patients compared with 11.9% of controls). Therefore, endoscopy and polypectomy confers the same protective effect against colorectal cancer in patients with hereditary nonpolyposis colon cancer as it does in persons without a hereditary predisposition. Other genetic or biochemical alterations, such as ras mutations, have been detected in the stool of patients with large adenomatous polyps and colon cancers and someday may be used to supplement conventional screening strategies [16]. Thus, in the future, endoscopic screening may be targeted to the segment of the population genetically predisposed to be at risk for the disease. Until then, fecal occult blood testing, flexible sigmoidoscopy, colonoscopy, and polypectomy remain the best methods for screening and preventing colorectal cancer.