Is Colonoscopy Indicated for Small Adenomas Found by Screening Flexible Sigmoidoscopy?

Methods

Patients

Data were collected prospectively on flexible sigmoidoscopic examinations performed in 4490 patients as part of a colon cancer screening program at a large urban Veterans Affairs medical center and a health maintenance organization from 1992 to 1996. Baseline variables were collected on a standardized data form and included family history of colorectal cancer in first- and second-degree relatives, age at diagnosis, history of blood in the stool, and previous endoscopy. Flexible sigmoidoscopy was offered to patients referred by a primary care provider for colorectal cancer screening who were asymptomatic, were older than 50 years of age, did not have a first-degree relative 55 years of age or younger with colon cancer, had a negative result on a fecal occult blood test, and had not had previous barium enema or colonoscopy.

Patients were prepared for screening by administration of two phosphosoda enemas on the morning of the examination. Examinations were performed by a trained physician assistant, nurse practitioner, or gastroenterology fellow under the supervision of a staff gastroenterologist. An endoscope was inserted to a depth of 60 cm or until solid stool was encountered. All polyps smaller than 10 mm identified by flexible sigmoidoscopy were biopsied and examined histologically. Patients with polyps larger than 10 mm on flexible sigmoidoscopy either underwent biopsy or were referred directly for colonoscopy (during which the polyp was removed). Polyps were characterized as non-neoplastic (normal mucosa, hyperplastic, or inflammatory) or neoplastic (tubular adenoma, tubulovillous adenoma, villous adenoma, high-grade dysplasia, or adenocarcinoma) by a gastrointestinal pathologist. The size of the polyp was estimated to the nearest millimeter by using an open biopsy forceps placed in direct contact with the polyp before polyp removal.

All patients with neoplastic polyps were offered colonoscopy. Patients with single small adenomas (1 to 5 mm) were advised of the controversy over the implications of a single small distal adenoma found on flexible sigmoidoscopy. Colonoscopy was performed within 6 months of flexible sigmoidoscopy. Patients were prepared for colonoscopy by administration of polyethylene glycol-electrolyte gastrointestinal lavage or oral phosphosoda. All colonoscopies were performed by staff gastroenterologists or gastroenterology fellows under the supervision of a staff gastroenterologist. If a polyp was encountered, its location was estimated according to the distance from the anal verge as measured with markings on the endoscope. Polyp size was estimated to the nearest millimeter according to the methods described above. All visualized polyps were removed and characterized histologically. Polyps were considered advanced if they were larger than 10 mm or were characterized as tubulovillous, villous, high-grade dysplasia, or adenocarcinoma.

Patients were prospectively subgrouped according to the number (1 or >1), size (1 to 5 mm, 6 to 10 mm, or >10 mm), and histologic characteristics (nonadvanced or advanced) of distal neoplastic polyps. Proximal neoplastic polyps were defined as those that were identified by colonoscopy (but not flexible sigmoidoscopy) and were proximal to the sigmoid-descending colon junction.

Statistical Analysis

The prevalence of advanced proximal polyps was determined, and subgroup analysis was performed on the basis of the size and number of neoplastic polyps found in the distal colon by flexible sigmoidoscopy. Confidence intervals for the prevalence rates were calculated by using exact methods for binomial data [18]. A chi-square test for linear trend was used to determine whether the three categories of distal polyps (single 1 to 5 mm, multiple or 6 to 10 mm, or advanced) showed an increasing association with advanced proximal polyps. Logistic regression analysis was performed by using S-Plus software (Statistical Sciences, Inc., Seattle, Washington) to evaluate for confounding by age and sex. Differences between patients with a distal neoplastic polyp who did or did not undergo colonoscopy were compared by using the Student t-test for continuous variables and the z-test for categorical variables. All tests were two-sided.

The costs and consequences of the screening evaluation were determined by creating a decision model in DATA Software (TreeAge Inc., Williamstown, Massachusetts). Cost data for colonoscopy, biopsy, polypectomy, and flexible sigmoidoscopy were obtained from the Agency for Health Care Policy and Research publication on colorectal cancer screening [6]. The likelihood of having a polyp was taken directly from the observed data in the study. An estimate of the prevalence of proximal polyps among patients with no distal adenomas was obtained from the study by Rex and colleagues [19]. To include the observed data and 95% CIs, sensitivity analysis was performed on the prevalence of advanced proximal polyps among patients with tubular adenomas 1 to 5 mm in diameter on flexible sigmoidoscopy.

Role of Funding Source

The Harvard Pilgrim Health Care Foundation, the source of funding for this study, did not participate in data collection, analysis, or publication of this study.

Results

A neoplastic polyp was detected in 401 of 4490 asymptomatic patients who underwent flexible sigmoidoscopy, did not have first-degree relatives younger than 55 years of age with colorectal cancer, and had a negative result on a fecal occult blood test. Data on the size, location, and histologic characteristics of all polyps was complete for 392 patients (98%). Nine patients were excluded because of incomplete data (polyp was resected but not recovered for pathologic analysis [n = 6]) or polyp size was not reported [n = 3]). The mean age (±SD) of patients with neoplasia was 62.5 ± 8.1 years; 71% were male, and 29% were female. The histologic characteristics and size of the most advanced neoplastic polyps detected by flexible sigmoidoscopy among patients who had subsequent complete colonic visualization is shown in Table 1.

All 392 patients with distal neoplastic polyps were offered colonoscopy, and 301 (77%) agreed to undergo the procedure. All of the 91 patients who did not undergo colonoscopy had tubular adenomas. Eighty-six patients had a single adenoma 1 to 5 mm in diameter, 2 had multiple adenomas 1 to 5 mm in diameter, 2 had a single adenoma 6 to 10 mm in diameter, and 1 had multiple adenomas 6 to 10 mm in diameter. No patient with advanced distal adenomas refused colonoscopy.

To look for potential bias between patients who did or did not undergo colonoscopy for a tubular adenoma 1 to 5 mm in diameter found by flexible sigmoidoscopy, we compared the mean age, presence of a family history of colorectal cancer other than in a first-degree relative younger than 55 years of age, and the site at which flexible sigmoidoscopy was performed (Veterans Affairs medical center or health maintenance organization). Patients who did and patients who did not undergo colonoscopy for a tubular adenoma 1 to 5 mm in diameter did not differ for age or family history of colorectal cancer (data not shown). Acceptance of a recommendation for colonoscopy differed significantly by the site at which flexible sigmoidoscopy was done. Among patients with distal adenomas, the Veterans Affairs medical center completed colonoscopy in 56 of 58 patients (96.6%) and the health maintenance organization completed colonoscopy in 245 of 334 patients (83.3%) (P < 0.001).

Of the 301 patients in whom colonoscopy was attempted, 295 (98%) had colonoscopy of the cecum and 6 had incomplete colonoscopy but subsequently had a complete barium enema. Sixty-three of the 301 patients had advanced neoplastic polyps in the distal colon detected at the time of flexible sigmoidoscopy, and 238 had nonadvanced (tubular) neoplastic polyps. We further stratified these patients according to the histologic characteristics (tubular, tubulovillous, villous, or cancerous) of the most advanced distal neoplastic polyp, the size of the largest distal neoplastic polyp, and the number of distal neoplastic polyps.

Ninety patients had a single tubular adenoma 1 to 5 mm in diameter, 16 had multiple tubular adenomas 1 to 5 mm in diameter, 100 had a single tubular adenoma 6 to 10 mm in diameter, and 32 had multiple tubular adenomas 6 to 10 mm in diameter detected by flexible sigmoidoscopy. At the time of colonoscopy, all polyps in the proximal colon were removed and were characterized histologically. The prevalence of neoplastic polyps in the proximal colon in each group is shown in Table 2.

Most of the neoplastic polyps detected in the proximal colon were nonadvanced with respect to histology and size. The prevalence of advanced proximal neoplastic polyps is shown in Table 3. Among the 90 patients with a single tubular adenoma 1 to 5 mm in diameter, 0% (95% CI, 0.0% to 4.0%) had an advanced proximal polyp. In contrast, among patients who had multiple tubular adenomas 1 to 5 mm in diameter or any number of distal polyps 6 to 10 mm in diameter, 8 of 148 (5.4% [CI, 2.4% to 10.4%]) had an advanced proximal polyp (2 of 16 patients with multiple adenomas 1 to 5 mm in diameter, 3 of 100 patients with single adenomas 6 to 10 mm in diameter, and 3 of 32 patients with multiple adenomas 6 to 10 mm in diameter. Among patients with advanced distal polyps, 5 of 63 [7.9% (CI, 2.6% to 17.6%)] had an advanced proximal polyp. Therefore, the prevalence of advanced proximal polyps increased as the size and number of distal polyps increased (P = 0.013 for trend).

Among the advanced lesions found in the proximal colon, 10 were large tubular adenomas 35 mm (1 polyp), 30 mm (2 polyps), 20 mm (1 polyp), 18 mm (1 polyp), 15 mm (2 polyps), and 12 mm in diameter (3 polyps). Two polyps had high-grade dysplasia, and 1 20-mm polyp had invasive carcinoma (Duke stage C). Polyps with high-grade dysplasia or cancer were seen only in patients with single or multiple polyps 6 to 10 mm in diameter. No proximal villous polyps, polyps with high-grade dysplasia, or carcinomas were detected among the 106 patients with polyps 1 to 5 mm in diameter. No evidence of confounding by age or sex was found by logistic regression modeling.

We estimated the cost savings and consequences of performing colonoscopy only on patients in whom sigmoidoscopy revealed multiple adenomas, adenomas larger than 5 mm, or adenomas with advanced histologic characteristics (strategy 1) compared with performing colonoscopy on all patients with adenomas (strategy 2). Using strategy 1 resulted in an overall decrease in the cost of screening (flexible sigmoidoscopy plus biopsy plus colonoscopy with polypectomy) from $126 per patient to $114 per patient, a 10% or $53 880 reduction in the cost to screen all 4490 patients. In addition, we performed sensitivity analysis on the estimate of advanced proximal polyps among patients with a single distal adenoma 1 to 5 mm in diameter. If the prevalence was 0% (as we observed), no advanced proximal polyps would have been missed, whereas 4 patients with advanced proximal polyps of the original cohort of 4490 patients would have been missed if the prevalence was 4.0%.

Discussion

We found that asymptomatic patients with negative results on fecal occult blood tests and a single tubular adenoma 1 to 5 mm in diameter on screening flexible sigmoidoscopy had a low prevalence of advanced proximal polyps. In contrast, patients with multiple polyps, tubular adenomas 6 to 10 mm in diameter, or an advanced polyp in the distal colon had a much higher prevalence of advanced proximal polyps. Our findings agree with those of studies that included only patients with tubular adenomas. In the study by Zarchy and Ershoff [14], 0.8% of patients with single or multiple tubular adenomas 10 mm or smaller in the distal colon had advanced proximal polyps. Grossman and associates [20] found a 3% prevalence of advanced proximal polyps among patients with tubular adenomas 10 mm or smaller. Tripp and colleagues [9] evaluated the relation of proximal and distal polyps among patients undergoing screening colonoscopy. They also noted a very low rate (0 of 32 patients) of advanced proximal polyps among the groups with tubular adenomas 5 mm or smaller in the distal colon but a higher rate (16 of 103 patients) of advanced proximal polyps among patients with distal polyps larger than 5 mm.

In contrast to these investigators, Read and associates [17] reported a much higher rate (6%) of advanced proximal polyps among patients with distal tubular or villous adenomas 5 mm or smaller. However, they included patients with advanced (tubulovillous or villous) distal polyps, which have been shown in our series and others [14, 17] to be associated with a higher prevalence of advanced proximal polyps. In our study, villous polyps of any size were considered advanced and were therefore not grouped with tubular lesions of 1 to 5 mm or 6 to 10 mm. This may account for our lower estimate of advanced proximal polyps among patients with distal tubular adenomas 1 to 5 mm in diameter. We did not note any distal villous polyps 1 to 5 mm in diameter.

Lemmel and colleagues [21] and others [22] have argued that because many patients with proximal colon cancer do not have any adenomatous polyps in the distal colon, flexible sigmoidoscopy would miss a high percentage of patients with colon cancer. Although these studies raise concern, they did not address what percentage of such patients had another accepted indication for colonoscopy (such as a family history of colorectal cancer, a positive result on a fecal occult blood test, or iron deficiency anemia) and thus would not have been considered for flexible sigmoidoscopy screening.

We used an intermediate outcome, the detection of an advanced proximal polyp, to predict the likelihood that patients would or would not benefit from a complete colonoscopy. There is strong evidence that advanced polyps have greater malignant potential than nonadvanced polyps. Cross-sectional studies have shown that cancer is rare in patients with polyps smaller than 6 mm [13]. A prospective surveillance study of small polyps left in situ showed that the rate of malignant conversion and even growth to larger polyps is low [11]. In two retrospective studies analyzing patients who had small polyps on sigmoidoscopy before the availability of flexible colonoscopy, Atkin [15] and Spencer [16] and their coworkers showed that the risk for subsequent cancer in the proximal colon was lower than predicted. Although these studies were retrospective, they demonstrate strongly that colonoscopy may not be beneficial for low-risk patients.

Our study has several limitations. Ninety-one of 392 patients with nonadvanced distal adenomas did not undergo colonoscopy. Eighty-six of these patients had a distal single tubular adenoma 1 to 5 mm in diameter. It is possible that some of these patients had advanced proximal polyps that were missed, and we may have underestimated the prevalence of advanced proximal polyps in this group. However, patients who did not undergo colonoscopy were similar to patients who did undergo colonoscopy in terms of age, family history, lack of symptoms, and any other risk factor for colorectal cancer. Thus, the lack of any proximal advanced polyps in the patients with a single small distal tubular adenoma is likely to be a characteristic of all study patients with a single small distal tubular adenoma. In addition, 9 patients who had both sigmoidoscopy and colonoscopy had incomplete data with which to classify their status. Because of other polyps found in all of these patients, no patient could have had a single distal adenoma 1 to 5 mm in diameter and an advanced proximal polyp. Thus, including these patients would not have changed our conclusion about single distal adenomas 1 to 5 mm in diameter.

The accuracy of estimating polyp size is imperfect, and polyp size is consistently underestimated [23]. As in most other published studies, we used an open biopsy forceps placed adjacent to the polyp, a method available at all endoscopy centers that perform biopsy. Other methods, such as direct measurement of polyp size at the bedside immediately after removal or after fixation by the pathologist, are also inaccurate [24] and are more difficult to compare across studies. Although more accurate endoscopic measuring tools have been used in research studies [11], they are not widely available.

An additional limitation of our study is the broad CIs found in even the lowest-risk group. Although 0 of 90 patients in this group had an advanced proximal polyp, the CI extends from 0% to 4%. Given the possibility of a 4% prevalence of advanced proximal polyps, it would be premature to make firm recommendations that colonoscopy need not be performed for single small distal adenomas. When the prevalence of a disease is estimated in the population in which the observed number in the sample is 0, the 95% CI will always include a low but finite prevalence of that disease. The decision to perform colonoscopy on a patient who has undergone flexible sigmoidoscopy must be made with the potential benefits and the risk and costs of the procedure in mind.

Several authors have argued that patients with no adenomatous polyps or hyperplastic polyps are at low risk for proximal adenomas and are therefore unlikely to benefit from colonoscopy [4, 19]. In our study, the prevalence of any proximal neoplasia among patients with single adenomas 1 to 5 mm in diameter was 12%, a rate similar to that seen in patients with no distal adenomatous polyps (15%) or hyperplastic polyps (18%) in other large prospective studies [19].

We conclude that the prevalence of tubular adenomas larger than 10 mm or villous adenomas or cancer in the proximal colon is very low among asymptomatic patients with a single adenoma 1 to 5 mm in diameter found by flexible sigmoidoscopy who have no other risk factors for colon cancer. These patients account for almost one half of patients referred for colonoscopy because of a polyp found by flexible sigmoidoscopy. Deferral of colonoscopy in this low-risk group would result in substantial cost savings and is unlikely to reduce the effectiveness of colon cancer prevention. Further study is necessary to refine the estimation of this prevalence with sufficient confidence to make firm policy recommendations. Our findings support the idea that patients undergoing flexible sigmoidoscopy should have all visualized polyps sent for biopsy and that those with multiple adenomatous polyps of any size, tubular adenomas larger than 5 mm, or advanced polyps in the distal colon should undergo colonoscopy.

Dr. Kemp: Harvard Vanguard Medical Association, 185 Dartmouth Street, Boston, MA 02116.

Dr. Trnka: Harvard Vanguard Medical Association, 133 Brookline Avenue, Boston, MA 02115.

Dr. Donovan: Brockton/West Roxbury Veterans Affairs Medical Center, 1400 VFW Parkway, West Roxbury, MA 02132.

Dr. Farraye: Harvard Vanguard Medical Associates, 291 Independence Drive, West Roxbury, MA 02167.