Everyone agrees that colorectal polyps are common in the general population, that a small proportion are malignant, and that fecal occult blood tests (FOBTs) can help uncover asymptomatic bowel tumors. The interrelationships among these three points, however, are still unclear. How often do benign neoplastic polyps become cancerous? How long does this take? How well do FOBTs detect polyps and cancer? Should all patients with polyps have ongoing surveillance to help prevent cancer? If so, by what means and how often? Does FOBT screening actually decrease mortality from colorectal cancer? The answers to these and related questions remain uncertain and are often hotly contested.

Two papers in this issue of the Annals explore these questions. Both articles will probably be as controversial as they are stimulating. In one paper, Koretz [1] claims that malignant polyps are often "sheep in wolves' clothing; " that is, despite being cancerous they can behave clinically benign for many years. From published data on the size distribution, proportion of malignancy, and age-prevalence of colonic polyps, Koretz calculates that about 725 000 people in the United States harbor at least one malignant polyp. Because only about 150 000 new cases of colorectal cancer are reported each year, he concludes that already cancerous polyps can remain clinically silent for several years. Moreover, he argues that these lesions are cancerous from the beginning rather than representing malignant transformation of benign polyps.

The latter conclusion is on shaky ground, and readers should carefully weigh Koretz's arguments against traditional evidence supporting the adenoma-carcinoma sequence [2, 3]. His first point is more compelling, although its validity hinges on the accuracy of epidemiologic data extracted from the autopsy, endoscopic, and surgical literature on polyps. Unfortunately the literature is not uniform, and Koretz averages widely disparate estimates to yield his calculations. Relatively minor errors in "true" age-related polyp size, prevalence, and malignancy will have a major multiplicative effect on the final estimate of malignant polyps in the U.S. population; such "stack of cards" calculations should therefore be viewed with caution.

Nevertheless there is good reason to accept Koretz's general point, even if his quantitation proves inaccurate. Recent colonoscopic surveys [4, 5] show neoplastic polyps in about one third of the asymptomatic general population over the age of 50—a remarkably high figure. Clearly, almost all never develop into cancer. In addition, even though the risk for cancer rises with increasing polyp size, only about 8% of unresected polyps larger than 1 cm develop malignancy during a decade of observation [6]. Indeed, people with small (<1 cm) tubular adenomas discovered at sigmoidoscopy have a very low long-term risk for developing colorectal cancer, even in the absence of ongoing surveillance [7]. Further, only about 8% to 15% of malignant polyps with invasion beyond the muscularis mucosae are accompanied by nodal lymphatic spread [8]. Thus it is reasonable to conclude that the risk for cancer from small colonic polyps is often overestimated and that at least some polyps can harbor malignancy for a considerable time while behaving benign biologically. Surveillance strategies need to be evaluated in this light.

The other article in this issue examines the relationship between FOBT screening and mortality from colorectal cancer. In a retrospective case–control study, Selby and colleagues [9] report that patients who died of colorectal cancer were less likely than a matched control group to have had FOBT screening before their tumor was diagnosed. The apparent benefit of FOBTs ranged from 25% to 30% for testing done within a year of the diagnosis, diminished as the interval between test and diagnosis lengthened, and was no longer evident by 3 to 4 years. However, many complexities are associated with retrospective case–control methods, especially bias in case selection and the control of confounding factors. The authors acknowledge this and attempt to take these problems into account in their design and analyses, but important uncertainty about the validity of their results still remains.

For example, only 36% of the FOBTs done within a year of cancer diagnosis were positive, so the overall 25% to 30% benefit implies that a positive test actually decreases tumor mortality by about 70% to 80%. This is implausible, especially because virtually all the study cancers were relatively advanced at the time of diagnosis (98% were Dukes stage B2 or worse). Moreover the odds ratio of benefit varied wildly for FOBTs done 3 to 4 years, 4 to 5 years, and 5 to 6 years before the diagnosis of malignancy—1.00, 0.21, and 2.04, respectively. Selby and colleagues rationalize this as "the instability of estimates based on small numbers," but the strikingly disparate results raise serious unease about the validity of their figures for earlier testing. In addition, sigmoidoscopy is an important confounding variable in this study, and many of the same patients actually participated in the authors' recently published parallel study that examined the effect of screening sigmoidoscopy on cancer mortality [10]. Even though Selby and colleagues incorporate sigmoidoscopy and other variables into their analyses, the fact remains that many factors other than FOBT screening could have substantially influenced the results. Wisely, the authors emphasize the wide confidence intervals around their odds ratios and advise awaiting results of randomized controlled trials of FOBTs before recommending widespread screening.

In this regard, preliminary mortality data are now available from two of the five controlled trials of FOBTs that are underway. Both show a handful of fewer deaths from colorectal cancer in the screening group, but the numbers are very small in relation to the large number of persons studied [11, 12]. More time is needed, but at this juncture it seems unlikely that non-targeted FOBTs will prove a highly effective strategy to decrease mortality from colorectal cancer. This conclusion also seems reasonable a priori because FOBT screening misses a substantial proportion of proven cancers [13, 14] and is poorly sensitive to polyps. Indeed it appears that most polyp "detection" by FOBT screening is merely fortuitous [15, 16]. Other arguments for and against FOBT screening have been reviewed elsewhere [13, 17].

Taken together, the interesting articles by Koretz [1] and by Selby and colleagues [9] raise both hope and caution: hope, that malignant colonic polyps may be biologically less aggressive than generally feared and that FOBT screening may have a modest impact on mortality from colorectal cancer; and caution, that both of these results hinge on uncertain methodologic calculations and may therefore be incorrect. Although these two papers are stimulating and potentially important, there is still much to learn about the relationships among colonic polyps, cancer, and FOBT screening.