Realistically, it is impossible to prove or disprove the theory that such adenomas are premalignant, that is, that a particular benign lesion will later become malignant. One cannot remove an adenoma to prove that it is totally benign (histologically) and then put it back in to see if it becomes malignant (clinically). The closest clinical experiences we have to this are the reports of polypoid growths that have been followed without excision. Malignancy has been found in such lesions known to have been present for over a decade [9, 10]. At first glance, this would only seem to be explicable by postulating a transformation from benignancy to malignancy. After all, how long could a "malignant growth" remain in the colon before it causes overt symptoms due to increasing bulk or distal spread?

Clearly there must be some "latent phase," that is, a time between the first appearance of a malignant process and the onset of local or distant symptoms. For any given lesion, this is impossible to calculate because such growths are removed when they are discovered. However, if we knew how many asymptomatic malignant lesions were in the population at a given time and how many symptomatic colon cancers subsequently occurred every year, we could calculate the average latent phase, that is, the average span of time during which these malignant lesions remain silent before causing symptoms.

Methods

Top Methods Results Discussion Author & Article Info References

Definitions

For purposes of this discussion, I define the following terms. When unmodified, the word "polyp" refers to any mucosal protrusion (including non-neoplastic processes) into the lumen of the colon. An "adenoma" (or "adenomatous polyp") is such a growth that has been examined histologically and found to be a tumor composed of benign gland-like structures (that is, a benign neoplasm); the term encompasses both tubular and villous lesions. "Malignant polyp" refers to a polyp containing histologic evidence of malignant cells extending beyond the muscularis mucosae; this term does not include adenomas in which there is only atypia, dysplasia, or carcinoma in situ limited to the mucosal layer. A "neoplasm" or "neoplastic polyp" refers to a mucosal protrusion that is either an adenoma or a malignant polyp.

Prevalence of Neoplasms

In order to estimate the number of malignant polyps present in the population, I first estimated the number of persons with neoplastic polyps. Ideally, such prevalence data should be obtained from "in-vivo" screening colonoscopic examinations of representative samples of asymptomatic individuals. Unfortunately, data from colonoscopic screening of truly asymptomatic populations are only available in three reports, two of which are reports from Veterans Affairs hospitals [11-13]. (Patients found to have hemoccult-positive stools are not "asymptomatic".) Sigmoidoscopic screening data are not useful because only a portion of the colon is examined. Hence, I used autopsy data from the United States and Western Europe [14-22] to estimate neoplasm prevalences in the general population. The data from the colonoscopic screening studies were used to validate the model. I made systematic efforts to identify all available autopsy studies.

Most of these autopsy studies were published many years ago; if neoplastic growths are increasing in numbers in recent decades [23], these data will underestimate the current prevalence. On the other hand, small lesions might be found in autopsy material but missed by endoscopic examination, overestimating the magnitude of the problem (because colonoscopy would be the only available technique to find and excise polyps in live patients).

In each autopsy study, the prevalences of neoplasms in men and women were identified separately for each age decade. The estimated prevalence for each sex and decade was calculated by averaging the respective prevalences from each study (deleting the highest and lowest to avoid "outliers"). Using these estimated prevalences and the total population for each group [24], one can estimate the number of people in the United States who have neoplasms.

A certain percentage of neoplasms are malignant. To calculate this figure, other information is needed. First, how often is malignancy found in lesions of different sizes? (Larger lesions are more likely to have a malignant component.) Only a limited amount of this information was available from autopsy studies, from the Western world [21] or elsewhere [25]. Hence, data from two other sources were obtained: sigmoidoscopic screening studies of asymptomatic individuals [26-28] and surgical polypectomies [29-36]. Lesions were divided into four sizes: less than 5 mm, 5 to 10 mm, 10 to 20 mm, and greater than 20 mm. Again, the estimated prevalences of malignancy in each size group were calculated by averaging the data from each study.

Next, how often does each size neoplasm occur (that is, what percentage of the total number of neoplastic polyps is < 5 mm, what percentage is 5 to 10 mm, and so forth)? I estimated these figures using autopsy and screening sigmoidoscopy data [16, 18, 21, 22, 27, 28, 37]. I did not use surgical data because many of those lesions were removed specifically for size considerations. Although two studies found a tendency for older patients to have larger neoplasms [18, 22], two others did not [20, 21]. For purposes of this calculation, I will assume that the percentages of the various sized neoplasms are the same for each age group.

Knowing both the rate of malignancy in each size lesion and the frequency with which each size occurs in the neoplasm population, the average risk that any given neoplasm is malignant can be calculated. However, the question being asked is what fraction of people have malignant polyps. People who have multiple colonic lesions are at higher risk for having at least one malignant lesion. Limited data regarding multiplicity of neoplasms are available [14-16, 20-22, 25, 38, 39]. If an individual has n neoplasms, the probability that at least one of them is malignant is (1 –[probability that all are benign]), or (1 -[probability that one is benign]ⁿ), assuming that the probability of one being malignant is independent of the presence or absence of other neoplasms in the same colon.

The specific details about these calculations are available from the author on request. Only the final estimated figures appear in the tables.

Incidence of Colon Cancer

Approximately 150 000 cases of colon cancer are reported every year [40]. No information is currently available to determine how many of these were found as the result of screening asymptomatic individuals.

Estimated Latent Phase

According to classic epidemiology, in a steady-state situation, prevalence equals incidence times duration. Using this formula, the average latent phase (equivalent to duration) can be calculated by dividing the prevalence of malignant polyps by the annual incidence of symptomatic colon cancer arising from malignant polyps. This latter number is no larger than 150 000, the annual incidence of all colon cancers. If some symptomatic colon cancers arise de novo from the mucosa (that is, not from malignant polyps) and if some of those 150 000 cases of cancer represent asymptomatic patients found during screening procedures, the appropriate denominator is less than 150 000.

Results

Top Methods Results Discussion Author & Article Info References

Number of Neoplastic Polyps in the General Population

The estimated prevalence rates of neoplasms in each decade for both men and women are summarized in Table 1. The prevalence of neoplasms in men tends to be somewhat higher than that in women and increases with age in both sexes. The three colonoscopic screening surveys [11-13] found age-specific prevalences that generally agree with the calculated estimates (Table 2).

Using these prevalence data, I calculated the absolute number of people in the United States expected to have neoplastic polyps (see Table 1); slightly under 27 million people currently have neoplastic polyps in their colons. About 25 million of them are 40 or more years of age; the prevalence of neoplasms in this population is 25 120/90 166 or 28%. The prevalences of neoplasms in these various age groups are remarkably similar to those reported when hemoccult-positive asymptomatic individuals are evaluated [41, 42], suggesting that these adenomas may not have been responsible for the positive test.

Number of Malignant Polyps in the General Population

Estimates of both the frequencies at which malignancy is found in neoplastic polyps of various sizes and the distribution of neoplasms by size are summarized in Table 3. It can be calculated that 1.4% of all neoplastic polyps are malignant. For comparison, this percentage ranged from 0.2% to 5.4% (mean, 1.6%) in the autopsy studies [14-16, 19-22, 25, 43, 44]. As validation, five malignant polyps were found in 285 neoplasms that were identified in the colonoscopic screening series, a frequency of 1.8% [11-13]. The 1.4% figure is used in the following calculations.

This does not mean that only 1.4% of all people with neoplastic polyps have malignant ones. About half of the people with neoplasms in their colons have at least two; this number ranges between 40% and 60% in most autopsy studies [14-16, 21, 22, 25, 38, 39] and was 39% and 70% in the two colonoscopic series where this information was available [11, 12].

The two autopsy studies that gave more specific information about how many patients had two; how many, three; and so forth [16, 20] are summarized in Table 4. From this limited information, I estimated a sample distribution that matches the data. It assumes that half of the patients with neoplasms have only one; half of the remainder, two; half of that remainder, three; and so on. This distribution is also shown in Table 4. Although three studies have found that multiplicity increases with age [17, 25, 41], others have not shown this phenomenon [15, 18, 19, 22] except perhaps at the extremes of age (persons over the age of 70 [19] or 80 [15] years). Because the prevalence of neoplasms increases with age, it might seem reasonable to also expect the prevalence of multiple neoplasms to rise; however, in the absence of more data, I assumed that this is not the case and that the hypothetical distribution is true for all age groups. Multiplicity appears to be the same in men and women [19, 21].

The probability of finding at least one malignant polyp in a colon in which there are "n" neoplasms equals (1 -[0.986]ⁿ) because 0.986 is the likelihood that any particular neoplasm is benign. If we consider 100 000 patients with neoplasms, the number who have at least one malignancy would be 2745 (Table 5); in other words, 2.7% of patients with neoplastic polyps can be expected to have invasive cancer in at least one of those growths. Five of the 147 (3.4%) patients with neoplasms in the three colonoscopic screening trials had malignant polyps [11-13].

Because this estimate was assumed to be true across all age groups, 2.7% of the almost 27 million people with neoplasms will have malignant polyps. Thus, I estimate that approximately 725 000 people in the United States have at least one malignant polyp in their colon. Recall that 150 000 new cases of colon cancer are seen every year. If all colon cancers arise from adenomas, and if all 150 000 of those new cases represent patients with symptomatic cancer, it will take about 5 years (725 000 per population/150 000 per population per year = 4.8 years), on the average, for these malignant polyps to become clinically apparent cancer.

If some of these cancers arise de novo and if many of these 150 000 cancers are diagnosed in asymptomatic persons (as almost incidental findings of screening), this latent phase is even longer. For example, if half of the colon cancers arise de novo and if only 125 000 of the new cancers are found because of symptoms, the average latent phase is more than 11 years (725 000/62 500 = 11.6).

Discussion

Top Methods Results Discussion Author & Article Info References

A widely held belief is that adenomas of the colon are premalignant and will degenerate into cancer [45]. As a result of heightened awareness of colon polyps and cancer, large numbers of patients who have polyps are being identified; they are being entered into prospective programs of colonoscopic excisions and follow-ups. A large national cooperative trial is currently assessing the role of yearly examinations versus examinations every 3 years [5]. The pursuit of polyps and the patients who harbor them has become a large component of the day-to-day clinical activity of gastrointestinal endoscopists.

As noted, the entire body of evidence supporting the hypothesis that adenomas are premalignant is circumstantial. Many studies have reported a statistical association between adenomas and cancer, but association can never infer causation. An equally tenable explanation is that the colonic mucosa of these patients is, in some way, predisposed to growing both benign and malignant neoplasms. The normal-appearing portion of mucosa of individuals who have neoplasms contains higher levels of tyrosine kinase and ornithine decarboxylase than does the normal mucosa of those who do not have such growths [46-48], findings consistent with the presence of pre-existing mucosal defects. If the problem is "bad soil," removing all benign growths will not prevent a malignant one from appearing.

As alluded to in the definition of a malignant polyp, a number of reports have described malignant foci in otherwise benign-appearing adenomatous polyps [3-5]. The relevance of this observation to the question of benign becoming malignant relates to the ability of the histologist both to determine past and future behavior based on a single sample in time and to define "benign" and "malignant" in a clinical sense. As we have seen, "malignant polyps" may not always behave in a very malignant fashion.

If adenomas and cancers had unique markers, one might use such an observation to infer that the two lesions had some commonality. Vogelstein and colleagues described genetic mutations and chromosomal abnormalities that were not seen very often in small benign adenomas but were present more frequently as they examined larger lesions and malignant growths [7]. The authors assumed that the small-medium-large adenomas were different phases of a progression to cancer, and used their data to hypothesize that these various genetic changes were the causative links along the way. However, the data only demonstrate an association between the size of the neoplasm and the presence and frequency of genetic abnormalities. One does not even know which came first.

One of the most frequently cited arguments that adenomas become malignant is the polypectomy experience of Gilbertsen at the University of Minnesota Cancer Detection Center [8]. In 1974, he claimed that the "risk of death from cancer of the lower bowel—the proctosigmoidoscopic area—can be virtually eliminated" [8]. He performed annual proctosigmoidoscopy (with routine polypectomies) on more than 18 000 patients followed for as long as 25 years (although the average follow-up period was about 5 years). During this time, he only found 11 cancers, although, based on the age and sex distribution, he calculated that 75 to 80 should have been observed. He attributed this 85% reduction to the identification and removal of the "premalignant" polyp.

This uncontrolled experience evaluated patients who may not have been representative of the general population. More importantly, a decade earlier Gilbertsen reported having already seen 25 patients with rectal adenocarcinoma and 31 with colon carcinoma [49]. Many of these lesions were seen on the initial examination, and at least these patients were apparently not counted among the 11 described in 1974 [8]. If these people had not been seen for screening, however, and if they were to have subsequently developed symptomatic cancer, they would have contributed some of the 75 to 80 expected cases.

Because some patients developed cancer despite annual examinations and polypectomies, some cancer must not pass through a recognizable benign polyp phase. Indeed, small flat cancers, often with metastatic spread already present, have been observed [50-52]. Making a conservative estimate from the Minnesota data, at least 15% of cancers arise de novo [8, 53]. Others have also seen cancer despite making frequent observations and removal of polyps [54, 55].

Perhaps the most compelling argument that adenomas become malignant are the reports in which polyps have been recognized but not excised, only to have malignancies found at the same place years later [9, 10, 54, 56]. This argument assumes that malignant polyps cannot remain quiescent for long periods. However, it appears that the latent phase for such malignant polyps is usually measurable in terms of years (perhaps decades), and it is entirely possible that the "malignancy" discovered after a number of years was actually present all along.

The calculation that 725 000 people in the United States have malignant polyps is critical to the determination of the latent phase. This calculation was based on a model created from mostly autopsy data. However, three screening colonoscopy studies [11-13] have described similar findings and thus serve as an independent validation of the model.

Ikegami postulated that malignant colon lesions could be divided into two types, flat ones that invade the bowel wall early and polypoid-type tumors that grow only slowly into the lumen and usually do not invade the deeper layers of the colon [52]. He suggested that 70% to 90% of the clinically important colon cancers arose from these flat lesions. (If 80% of colon cancer arises de novo, and 125 000 cases are found because of symptoms, the average latent phase is 29 years.) Under this concept, a "malignant" polyp (by histologic criteria) would not have to behave in a malignant fashion (that is, cause symptoms from metastases or local growth). Of course, this concept also challenges the value of removing such lesions.

Whenever a physician finds and removes a malignant polyp, he or she and the patient rejoice. After all, has the patient not just been saved from a terrible death from colon cancer? Given the large number of malignant polyps that must be present in the general population, this may not be the case. Unless these lesions go away with time (in which case they also need not be removed), many people must be dying with them but not because of them. For those individuals, the malignant polyp is clinically irrelevant, no matter how histologically ominous it is (a sheep in wolf's clothing).

One long-term (10-year follow-up) prospective, randomized controlled trial could not show that hemoccult screening improved colon cancer mortality [57], even though such screening does detect more early-stage malignancies [58-60]. These observations are compatible with the concept that some malignant lesions have long latent phases.

Malignant polyps that spontaneously disappear have an infinite latent phase because the denominator is zero. (A subpopulation of such growths will lengthen the average latent phase.) In this regard, one might speculate that redifferentiations of malignant polyps into histologically benign lesions account for the observations of "adenomas with malignant foci".

It should be remembered that the latent-phase calculation produces a number reflecting the average span of time during which these lesions may produce problems. If the variation around the average is large enough, some of these malignant polyps will cause symptoms. The conundrum we face is in identifying those patients who have these more aggressive processes. As we have seen, histologic examination alone appears inadequate.

Using autopsy (or even screening colonoscopy) data to calculate "growth rates" of cancers will result in an under-representation of more aggressive tumors ("length bias"). However, if one is going to advocate screening programs and polypectomy prophylaxis, it is these same slow-growing lesions that will largely be eradicated [61].

As I noted at the outset, it is impossible to prove that benign adenomas transform into malignancies. Although this concept appears to be widely accepted and is the rationale for removing asymptomatic polyps, the evidence supporting it is circumstantial and arguable. It is certainly likely that not all colorectal cancers begin as benign adenomas, although it is possible that some do. However, answering the question of whether or not adenomas are premalignant is actually not critical to clinical practice. What we need to know is whether the search for neoplasms (and the removal of polyps along the way) ultimately results in an improvement in the morbidity and mortality from colon cancer. The answer is unknown.

"It is what we think we know already that often prevents us from learning".

—Claude Bernard