It is well established that screening for colorectal cancer with the fecal occult-blood test significantly lowers the rate of death from the disease. (1) The report by Mandel et al. (2) in this issue of the Journal suggests that it also prevents cancer, reaffirming prior research about the benefits of detecting adenomatous polyps. (3) The mounting evidence that early detection saves lives has sharpened the consensus among professional groups that screening for colorectal cancer should begin routinely at the age of 50. (4)
The current debate centers on which tests to use, at what ages, and how often. Each of the tests -- fecal occult-blood screening, sigmoidoscopy, barium enema, and colonoscopy -- has advocates who consider it superior. Recent attention has focused on colonoscopy, which has been favored by professional groups, (5) editorialists, (6) and television personalities. (7) But determining which test is the best is a value-laden choice. Each test finds cancers, some more effectively than others, but with added benefits come the potential for increased harms and costs. Which trade-off is preferable is a matter of perspective.
For some people, scientific certainty that screening improves outcomes is what matters most. Fecal occult-blood screening, the only test shown in randomized trials to lower mortality from colorectal cancer, (1) ranks highest in this regard. Trials of screening with sigmoidoscopy are under way, (8,9) but case-control studies currently provide the only evidence that sigmoidoscopy lowers mortality. (10) Some observers find retrospective evidence unconvincing; patients who undergo sigmoidoscopy may have fewer complications from colorectal cancer for other reasons (e.g., a healthier lifestyle). Others are persuaded by these studies, because mortality was lower exclusively for cancers that could be reached by the sigmoidoscope. (10)
No direct evidence proves that whole-bowel screening (colonoscopy or barium enema) reduces mortality, though studies of this topic are in process. Those who require such evidence consider the endorsement of whole-bowel screening to be premature. Those who are willing to bridge the gap with logic argue that if sigmoidoscopy -- which evaluates only the distal bowel -- reduces mortality, then complete bowel examination must save more lives. (6) Colonoscopy certainly detects more proximal disease than does sigmoidoscopy. (11,12) Those who accept this implicit evidence of benefit question the ethics of waiting for mortality data.
If one selects tests on the basis of their accuracy, colonoscopy wins. It enables inspection of the entire colon, detection of almost all important neoplasms, and immediate polypectomy. Barium enema also enables whole-bowel examination but is less accurate. (4,13) Flexible sigmoidoscopy is as accurate as colonoscopy, but only for the distal bowel. (4) Fecal occult-blood screening, a test for peroxidase, has limited sensitivity for neoplasms (37 percent to 69 percent (4)) and must be repeated every one to two years. Approximately 85 percent to 90 percent of positive screening tests are false positives. Performing fecal occult-blood screening together with sigmoidoscopy improves sensitivity for lesions that elude sigmoidoscopy alone but not for nonbleeding lesions, particularly polyps.
Some contend that accuracy matters less than the magnitude of benefit. By one estimate, 1173 persons must undergo fecal occult-blood screening to prevent one death in 10 years (a 0.09 percent probability of preventing death for the individual patient). (1) Some would say that this is immaterial if screening saves lives and that only those concerned about costs would think otherwise. But the millions who undergo screening for no apparent gain -- the denominator -- are subject to harms (such as the follow-up for false positive results) that could cumulatively outweigh the benefits to the smaller group in the numerator. Do one person's benefits justify the potential harms to which 1173 persons are exposed? The answer depends, of course, on the seriousness of the harms.
Concern about harms might seem unwarranted, given the safety of colonoscopy and sigmoidoscopy. Bleeding or perforation occurs in only 10 to 30 persons per 10,000 examinations, and death occurs in 1 per 10,000 colonoscopies. (4,12) But if the probability of benefit is also small (e.g., the rate of death from colorectal cancer among persons 50 to 54 years of age is 1.8 per 10,000 (14)), the number of persons harmed by screening could offset the number who benefit.
This downside explains the reluctance of some to embrace colonoscopy as the "preferred" test. The question is not whether it detects more neoplasms but whether the incremental benefits, as compared with those of alternative tests, justify the harms. Sigmoidoscopy (followed by colonoscopy in the case of suspicious findings) can detect 80 percent of neoplasms (12); additional lesions that it misses (e.g., proximal neoplasms without distal disease) can sometimes be identified by combining sigmoidoscopy with fecal occult-blood screening. What these tests still fail to detect constitutes the incremental benefit of routine colonoscopy. According to one model, the discovery of these lesions saves 2.5 life-years for every major complication that is caused by colonoscopy. (4)
Is this trade-off worthwhile? Unfortunately, there is no mathematical answer. First, reliable data on complications are lacking; published rates are outdated or misrepresent actual rates in the community. (15) Second, the benefits (e.g., prevention of cancer) and harms (e.g., bleeding) are dissimilar; they must be converted to a common unit (e.g., quality-adjusted life-years) so that outcomes are weighted according to their importance. Models have attempted this, (4) but the utilities are based on assumptions. Until quantitative methods improve, the judgment of whether benefits outweigh harms remains subjective. Answers are as varied as peoples' perceptions of the probability and importance of benefits and harms.
For most patients, adverse effects that are more subtle than bleeding and perforation, but far more common, influence test preferences. These include the discomfort, embarrassment, and inconvenience associated with bowel preparation and the examination itself; anxiety and other negative consequences from having received positive results; and harms resulting from false positive results. How often these occur depends on the test and clinical setting, but how much they matter depends on the individual patient. Physicians presume that patients' preferences about which test is best match their own, but they are actually disparate. (16) Although some patients prefer colonoscopy because it need not be done every year and sedation is used, others dislike the fact that time is needed to recover from the test. Which test patients prefer also depends on their access to specialists, their insurance coverage (or lack thereof), potential difficulties in performing the test, their experience with the test, and changes in risk. A patient might reject the benefit-to-harm ratio for colonoscopy at the age of 50 but feel differently at the age of 65, when proximal lesions are more likely to have developed. (12)
For health insurers, costs dictate test preferences, but affordability depends on charges, which vary locally according to volume, market pressures, and contracts. The costs to society, but less so those to payers, are offset by the economic benefits of averting cancer. Numerous studies, (4) including several published recently, (17,18,19) underscore the cost effectiveness of every colorectal screening test (at an estimated cost of <$20,000 per year of life saved), including colonoscopy. Though it is the most expensive test, colonoscopy may ultimately cost less than alternatives because it prevents more cancers and does not need to be performed frequently. (18) However, this argument is unconvincing to many insurers, whose members may have changed plans by the time benefits accrue. Insurers without altruistic priorities prefer tests with lower up-front costs.
Finally, a supposedly better test can lose its superiority in the hands of an incompetent examiner or one without the means of ensuring follow-up of abnormal results and timely rescreening. On the national and local levels, the number of qualified endoscopists may be inadequate to provide colonoscopy (or even sigmoidoscopy) to the people for whom screening is now recommended. Without proper coordination, patients at risk may not receive priority. If unqualified examiners absorb the overflow, increased inaccuracy and complications could undo the incremental benefit that the test offers.
Those who advocate a specific screening test for colorectal cancer have taken a position -- reflecting their priorities with regard to scientific certainty, accuracy, the magnitude of benefit, safety, costs, and feasibility -- that they presume is universal. This approach is justified only when it can be safely assumed that most people, given the same facts, would make the same choice. This is a safe bet when the trade-offs are clear but not, as in this case, when subjective and situational factors determine how the scales tip. Although the average patient has the most to gain from certain screening protocols (e.g., fecal occult-blood screening every 1 to 2 years plus sigmoidoscopy every 5 years or colonoscopy alone every 10 years), the best choices are made when patients, physicians, and insurers weigh the trade-offs from their own perspectives.
Sixty percent of eligible people in the United States have never been screened for colorectal cancer. Allowing patients to select the tests they prefer may do more good -- as long as they choose something -- than whatever is gained with a "preferred" test. The difficulties of shared decision making notwithstanding, patients have the right to make their own choices. Someday, noninvasive forms of screening technology may render today's choices obsolete. Until then, redefining the best test as the one the patient wants may save the most lives.
Steven H. Woolf, M.D., M.P.H.
Virginia Commonwealth University
Fairfax, VA 22033