Flexible Sigmoidoscopy Plus Air Contrast Barium Enema Versus Colonoscopy for Suspected Lower Gastrointestinal Bleeding DOUGLAS K. REX, RICHARD A. WEDDLE, GLEN A. LEHMAN, DAVID C. POUND, KATHERINE W. O’CONNOR, ROBERT H. HAWES, ROBERT S. DITTUS, JOHN C. LAPPAS, and
Divisions of Gastroenterology and General Internal Medicine, Department of Medicine, and Department of Radiology, Indiana University Medical Center; and Richard L. Roudebush Veterans Administration Hospital, Indianapolis, Indiana.
A randomized, controlled trial was performed to compare the diagnostic yields and cost-effectiveness of two strategies for the evaluation of nonemergent lower gastrointestinal bleeding. Three hundred eighty patients aged 240 yr were randomized to undergo initial flexible sigmoidoscopy plus air contrast harium enema or colonoscopy; 332completed the initial studies. Initial colonoscopy detected more cases of polyps t9 mm in size, adenomas, and arteriovenous malformations hut fewer cases of diverticulosis. No significant difference was found between strategies in the number of patients detected with cancers or polyps 29 mm in size. In both strategies, cancers were more common in subjects aged 255 yr (8% overall) than in those aged ~55 yr (1%). Among patients aged t55 yr with suspected lower gastrointestinal bleeding, initial flexible sigmoidoscopy plus air contrast barium enema is a more cost-effective strategy for the detection of colonic neoplasms than initial colonoscopy. However, initial colonoscopy is more cost effective for those aged 255 yr.
olorectal cancer is the second leading cause of cancer death in the United States. In patients aged 240 yr with suspected lower gastrointestinal (GI) bleeding, the entire colon should be imaged for neoplaMany strategies are available for colon examsia (1,Z). ination (3-5). Two commonly used strategies for initial evaluation are diagnostic colonoscopy and the combination of flexible sigmoidoscopy and air contrast barium enema (ACBE]. Flexible sigmoidoscopy is combined with ACBE because the sigmoid colon is difficult to evaluate radiographically (6,7).
While both colonoscopy and flexible sigmoidoscopy plus ACBE are frequently used for the initial evaluation of suspected lower GI bleeding, the relative merits of the strategies are still debated and it is not known which is more cost effective. Therefore, we performed a randomized, controlled clinical trial comparing colonoscopy with flexible sigmoidoscopy plus ACBE as the initial diagnostic procedures for the evaluation of suspected lower GI bleeding. Methods Study Sample Eligible patients were those aged 240 yr who were referred to the Division of Gastroenterology of the Indiana University Medical Center with the clinical suspicion of nonemergent lower GI bleeding. Hemoccult-positive stools were largely (95%) obtained by digital rectal examinations rather than home testing. Outpatients were referred from internal medicine, surgical, and various specialty clinics at the Wishard Memorial Hospital (a county hospital], the Richard L. Roudebush Veterans Administration (V.A.) Hospital, and the Indiana University Hospital [a tertiary care hospital). Inpatients were also referred from each hospital. Patients were excluded if they had prior colorectal neoplasia or vascular malformations, colonoscopy, or barium enema within the previous 18 mo, if they had significant coagulopathy, or if they were unable to give informed consent. Enrolled patients did not represent consecutive patients because of the location of the physicians primarily involved in randomization, the referral of private patients Abbreviation used in this paper: ACBE, air contrast barium enema. 0 1990 by the American Gastroenterological Association 0016-5085/90/$3.00
REX ET AL.
specifically for colonoscopy, and at times insufficient space on the endoscopy schedule to perform potential colonoscopy generated by randomization. However, we compared these nonconsecutive patients to a sample of 100 consecutive patients and found them to be demographically comparable. Between March 1985 and November 1987,191 patients were randomized to undergo flexible sigmoidoscopy plus ACBE first and 189 to undergo colonoscopy first. Of the 380 randomized patients, 332 kept their appointments and completed the initial colon tests: 168 (88%) in the flexible sigmoidoscopy plus ACBE strategy and 164 (87%) in the colonoscopy strategy. Of the 332 study subjects, 205 were recruited from Wishard Hospital, 110 from the V.A. Hospital, and 17 from Indiana University Hospital.
Vol. 98. No. 4
Cost analyses were performed from the perspective of third-party payers using our local procedure charges as a proxy for cost. The charges for all diagnostic and therapeutic procedures, including the alternative lower GI tests, were included in the total procedural charges for the original strategy to which the patient was assigned. The procedural charges included room charges and professional fees. Procedural charges were flexible sigmoidoscopy, $154; ACBE, $255 (flexible sigmoidoscopy plus ACBE, $409); diagnostic colonoscopy, $543; and polypectomy, $137. One fee was charged per patient for polypectomy regardless of the number of polyps removed. Diagnoses considered potential sources of GI bleeding included hemorrhoids, diverticulosis, colitis, polyps r5 mm, cancer, and arteriovenous malformations.
Study Design The study was a randomized, controlled clinical trial of two diagnostic strategies: [a) initial colonoscopy and (b) initial flexible sigmoidoscopy plus ACBE. Baseline data collected before randomization included the reason for referral; demographic data; previous colorectal diseases; recent colorectal symptoms; results of abdominal, rectal, and stool examinations; and routine laboratory data (hemoglobin, mean corpuscular volume, white blood cell count, blood urea nitrogen, creatinine, albumin). Patients were randomized into one of the two study groups using a randomly varying block design with block sizes of two and four. All patients received standard colon cleansing suitable for either type of examination. In inadequately cleansed patients, the preparation and assigned test(s) were repeated. Colonoscopy was performed by fellows under staff supervision. All polyps found during colonoscopy were removed or destroyed. Flexible sigmoidoscopy was performed with 60-cm flexible sigmoidoscopes by internal medicine or surgical residents under supervision of a staff gastroenterologist. Air [and not CO,) was used for endoscopic procedures. Air contrast barium enema was usually performed the same day as flexible sigmoidoscopy and was interpreted by a staff radiologist. Flexible sigmoidoscopy findings were available to radiologists performing ACBE in all cases. We successfully collected follow-up data on the occurrence of complications in 98% of subjects using telephone follow-up, review of computerized medical records, and review of the patients’ charts. Patients with an incomplete initial lower GI test (ACBE that in the opinion of the radiologists inadequately examined a portion of colon to rule out the presence of neoplasm or colonoscopy that failed to reach the cecal tip] were referred for the corresponding alternative lower GI test (colonoscopy or ACBE). Patients with polyps sized ~5 mm discovered by flexible sigmoidoscopy or ACBE were referred for colonoscopy and polypectomy. Patients with masses on ACBE outside the reach of flexible sigmoidoscopy were referred for diagnostic colonoscopy. Anemic patients without colorectal cancer, arteriovenous malformations, or colitis were referred for upper endoscopy unless this had been done before randomization. Patients with upper GI symptoms alone were referred for upper endoscopy at the discretion of the managing physician.
Data Analysis Baseline and outcome data were compared between the two groups, flexible sigmoidoscopy plus ACBE vs. colonoscopy. Differences between the means of continuous variables were tested for significance using the Student’s t-test (8). Differences between dichotomous variables were tested using the x2 test (8). p Values ~0.05 were considered significant. The study sample size provided a power of 0.8 at an alpha of 0.05 to detect a 10% difference in cancer prevalence between the two study groups. To evaluate cost effectiveness, the total procedural charges incurred using each strategy were summed for two different outcomes: [a] total charges per patient with a potential lower GI bleeding source found and (b] total charges per patient with at least one significant neoplasm [polyp 2 5 mm or cancer). To explore the impact on the relative cost effectiveness of the two diagnostic strategies, sensitivity analyses were performed. The charge for flexible sigmoidoscopy plus ACBE was held constant at $409, while the charge for diagnostic colonoscopy [and thus the colonoscopy-flexible sigmoidoscopy plus ACBE charge ratio] was allowed to vary. The charge for polypectomy was $137 (our current actual charge] when the charge for diagnostic colonoscopy was $543 (our current actual charge) and increased or decreased in proportion to changes in the diagnostic colonoscopy charge. This simulates clinical practice because increases in the charge for diagnostic colonoscopy are likely to be accompanied by proportional increases in the charge for polypectomy. The formula for total charges for each study group in the sensitivity analysis included all procedural charges and the probability [based on the data from this study) that the alternative procedure was indicated for visualization of unexamined or inadequately examined colon, the probability that colonoscopy was needed for polypectomy or biopsy, the probability of polypectomy, the probability that colonoscopy failed to pass the splenic flexure (in which case the charge was that for flexible sigmoidoscopy and not for colonoscopy], and the probability of false-positive ACBE (in which case there was no charge for polypectomy when colonoscopy was performed). For the purposes of sensitivity analysis, it was assumed that the 6 patients in the flexible sigmoidoscopy plus ACBE group with polyps 25 mm who
FLEX SIG PLUS ACBE VS. COLONOSCOPY
have not yet undergone colonoscopy and polypectomy [see Results] will eventually undergo it, and these charges were included in the cost of flexible sigmoidoscopy plus ACBE. Similarly, for the 4 patients in the colonoscopy group recommended to undergo ACBE who have not done so, the charges for ACBE were included. These charges have been included because these patients have not complied with the follow-up recommended and because we believe compliance is likely to be higher in a private practice population. Results
Baseline information was collected on all patients. There were no significant demographic or historical differences between the two study groups (Table 1). The mean age for all subjects was 62 yr (range, 40-88); 50% were black, 58% were male, and 43% were inpatients. The reasons for referral were hemoccult-positive stools (93%), hematochezia (8%), and melena with a negative upper GI evaluation (4%). Anemia was present in 19%. There were also no significant differences in physical examination or laboratory variables between the two study groups. Flexible sigmoidoscopy was successful (insertion to at least 30 cm) in 161 patients (96%). The mean depth
Table 1. Comparison of Demographic and Historical Variables” Flexible sigmoidoscopy plus ACBE [n = 188)
Age ( ~4 Race (% white)
62.2 45.8 56.0 45.5
Sex [% male) Inpatient status (%) Reason for referral Heme positive stool (%) 94.0 Melena with no upper GI 5.4 source (%) Hematochezia (%) 6.7 Personal history of diverticulosis 1%) 4.2 1.2 Family history of polyps [%) Family history of colon cancer (%) 4.3 Recent [within 6 months] history of: Hematochezia (%) 35.5 8.4 >2 wk of diarrhea (%) Lower abdominal pain (%I 19.9 Any change in bowel habits (%I 25.9 Weight loss without dieting (%I 28.3 21.5 Mean weight lost (lb][n = 89) “All values are means. NS. difference p > 0.05.
Colonoscopy [n = 164)
62.0 53.7 59.8 40.1
NS NS NS NS
40.1 8.6 24.7
NS NS NS
not statistically significant,
of insertion was 50 cm. Flexible sigmoidoscopy was normal in 31 of 168 patients (18%). Abnormalities discovered by flexible sigmoidoscopy included hemorrhoids, 98 (58%); diverticulosis, 30 (19%); any polyps, 37 (23%); cancer, 7 (4%); and proctitis, 4 (2%). Air contrast barium enema was successful (examination adequate to rule out major pathology) in 157 patients (93%). Causes of unsuccessful ACBE were inability to distend or fill the right colon adequately (n = 51, repeated1 y inadequate preparation to rule out mass lesions (n = 41, and inability to retain the enema adequately (n = 2). Air contrast barium enema was normal in 48 of 168 patients (29%). Abnormalities diagnosed by ACBE included hemorrhoids, 1 (1%); diverticulosis, 82 (48%]; any polyp, 43 (26%); stricture, 3 (2%); and cancer, 7 (4%). Colonoscopy was successful (insertion to the cecum) in 151 patients (92%). Causes of unsuccessful colonoscopy were obstructing cancers (n = 6) and technical factors (n = 7). Colonoscopy was normal in 18 of 162 patients (11%). The diagnostic yields (number of patients with abnormalities] of the two study strategies are summarized in Table 2. Of 168 patients in the flexible sigmoidoscopy plus ACBE group, 64 (38%) had at least 1 colorectal polyp demonstrated by the initial diagnostic tests. These 64 patients had 101 polyps demonstrated by flexible sigmoidoscopy plus ACBE, with the following size distribution: ~4 mm, n = 45; 5-8 mm, n = 29; r9 mm, n = 27. These 64 patients included 4 patients with 7 polyps who also had colorectal cancer (polyp size distribution: ~4 mm, n = 2; 5-8 mm, n = 2; r9 mm, n = 3). Patients with polyps r5 mm were referred for colonoscopy. Despite repeated passes through the area in question in 4 of the 38 patients (10.5%) with polyps ~5 mm seen with ACBE, no polyp could be found at colonoscopy. These patients were considered to have had false-positive ACBE results. Among remaining patients referred for polypectomy, all (28 patients, including the 4 who also had cancer] had at least one adenoma. Thus, 33 patients (20%) in the flexible sigmoidoscopy plus ACBE group had either cancer or adenoma documented by the initial tests or subsequent colonoscopy. Among patients referred for polypectomy, colonoscopy detected an additional 25 polyps not visualized by the initial flexible sigmoidoscopy plus ACBE (size distribution: ~4 mm, n = 18; 5-8 mm, n = 5; 29 mm, n = 2). Despite contacts by telephone and letter, 6 patients with polyps r5 mm have yet to submit to colonoscopy and polypectomy. Nine patients in the flexible sigmoidoscopy plus ACBE group had cancer: 3 had Dukes’ B tumors (all with serosal involvement], 1 had a C lesion, and 4 had Dukes’ D tumors. One patient with Dukes’ B cancer had a negative ACBE. Four weeks later he developed
858 REX ET AL.
Table 2. Comparison
All patients Flexible sigmoidoscopy plus ACBE [n = 168) Internal or external hemorrhoids Diverticulosis Any colorectal polyp Any colorectal polyp 25 mm Any colorectal polyp 29 mm Colonic stricture Colon cancer Colitis or proctitis Arteriovenous malformation
Colonoscopy (N = 164)
Age 255 yr Flexible sigmoidoscopy plus ACBE (n = 127) p Value n (%I
Vol. 98. No. 4
Colonoscopy [n = 123)” nI%l
Age t55 yr Flexible sigmoidoscopy plus ACBE (n = 41) n(%) p Value
Colonoscopy (n = 41) n (%)
99(59) 85(51) 64(38)
97(59) 56(34) 86(52)
NS 0.002 0.009
76(60) 67[53) 50(39)
NS NS NS
“Total number of patients. NS, difference not statistically significant, p > 0.05.
right lower quadrant abscess which was surgically drained. He then underwent colonoscopy, which showed a cecal cancer, and he subsequently had surgical resection. One patient with a transverse colon cancer demonstrated by ACBE has continued to refuse surgery. One hundred sixty-four patients were randomized to colonoscopy, and 86 had at least one polyp (52%). The 86 patients had 194 polyps with the following size distribution: 14 mm, n = 108; 5-8 mm, n = 56; r9 mm, n = 29. Nine of the 86 patients with polyps also had colorectal cancer. These 9 patients had 16 polyps (~4 mm, n = 9; 5-9 mm, n = 5; r9 mm, n = 2). Among the 86 patients with polyps, 72 had at least one histologically documented adenoma [including all 9 patients with both cancer and polyps], 6 had only hyperplastic polyps (all ~4 mm), 1 had a carcinoid, 1 had an inflammatory polyp, 1 had tissue interpreted histologically to represent “lymphoid aggregate,” and 5 had no specimen obtained. The latter patients had diminutive polyps that were destroyed without obtaining a specimen, Thus, 76 (46%) of the 164 patients in the colonoscopy group had colonic adenoma or carcinoma. Thirteen patients in the colonoscopy group had cancer, 2 had Dukes’ A tumors (confined to pedunculated polyps; these patients did not require laparotomy), 8 had Dukes’ B tumors (3 with only submucosa involvement, 1 with involvement extending to muscle but not serosa, 4 with serosal involvement], 2 had Dukes’ D lesions, and 1 with a transverse colon cancer has continued to refuse surgery. Among 168 patients randomized to flexible sigmoidoscopy plus ACBE, 53 (32%) had subsequent colonosa
copy recommended, 11 because of an inadequate study, 38 for polypectomy, and 4 for biopsies on lesions outside the reach of flexible sigmoidoscopy. In the 164 patients randomized to diagnostic colonoscopy, 13 (8%) had subsequent ACBE recommended because of inability to advance the colonoscope to the cecum. There was a fourfold difference in the frequency of recommended alternative lower GI procedure (32% vs. 8%, p 5 0.0001). Examination of diagnostic yields with respect to age indicated diversion in polyp and cancer yield for patients aged 255 yr (Table 2). Within each age strata, there were no significant differences in demographic, patient history, or laboratory variables between the study groups. The superior detection of polyps in the strategy of colonoscopy first was entirely accounted for by the finding of polyps t9 mm in size in patients aged ~55 yr [Table 2). The overall yield of cancers in patients aged t55 yr was very low (1%) compared with 8% in those aged 255 yr. More patients with polyps r9 mm were found by the strategy of flexible sigmoidoscopy plus ACBE first than by the strategy of colonoscopy first in patients aged t55 yr (p = 0.021). Complication surveillance detected both possible procedural complications and a variety of postprocedural symptoms that did not appear to be related to the procedures (Table 3). Transient respiratory distress was noted in 6 patients. None required oxygen or mechanical ventilation. Phlebitis occurred twice as often in the colonoscopy-first strategy, but this difference was not statistically significant. However, our study did not have sufficient power to detect a true difference in the incidence of phlebitis of this magni-
FLEX SIG PLUS ACBE VS. COLONOSCOPY
Comparison of Complication Surveillance Data
Flexible sigmoidoscopy plus ACBE
Complications Respiratory distress Phlebitis Other postprocedural symptoms Rectal bleeding Other complications
NS. not significant, p > 0.05.
tude. A number of patients reported symptoms within 2 wk of the initial procedures that were considered in all cases to represent continuation of their preprocedure symptoms including abdominal pain, chest pain, and dizziness in 44 patients (13%). There were no transfusions, hospitalizations, prolongation of hospital stay, or deaths attributable to a diagnostic procedure in this study. The mean total charge per patient using our procedural charges to complete lower GI evaluation and treatment was $609.64 for the group undergoing flexible sigmoidoscopy plus ACBE first and $603.11 for the group undergoing colonoscopy first. To generalize the comparisons of total charges for the two study groups so that readers with different local procedural charges can apply these data to their settings, we performed sensitivity analyses [Figure 1).
The vertical axis in Figure 1 is the ratio of total procedural charges incurred in the flexible sigmoidoscopy plus ACBE group to total procedural charges incurred in the colonoscopy group. The horizontal axis is the ratio of diagnostic colonoscopy charge to flexible sigmoidoscopy plus ACBE charge. The diagnostic colonoscopy-flexible sigmoidoscopy plus ACBE charge ratio at which total procedural charges for the two study groups were equal was 1.34. In Table 4 this charge ratio is listed for two outcomes and for 2 cost-effectiveness endpoints. For each particular outcome or endpoint, a diagnostic colonoscopy-flexible sigmoidoscopy plus ACBE charge ratio greater than that shown in Table 4 would favor flexible sigmoidoscopy plus ACBE as the more cost effective initial strategy for that outcome or endpoint. Likewise, a charge ratio lower than that listed would favor colonoscopy as the more cost-effective initial strategy. Table 4 includes the charges that would have been incurred if colonoscopy and polypectomy had been performed for patients in the flexible sigmoidoscopy plus ACBE group with only polyps