- Email: [email protected]
Analysis of air contrast barium enema, computed tomographic colonography, and colonoscopy: prospective comparison
Articles
Analysis of air contrast barium enema, computed tomographic colonography, and colonoscopy: prospective comparison D C Rockey, E Paulson, D Niedzwiecki, W Davis, H B Bosworth, L Sanders, J Yee, J Henderson, P Hatten, S Burdick, A Sanyal, D T Rubin, M Sterling, G Akerkar, M S Bhutani, K Binmoeller, J Garvie, E J Bini, K McQuaid, W L Foster, W M Thompson, A Dachman, R Halvorsen
Summary Background The usefulness of currently available colon imaging tests, including air contrast barium enema (ACBE), computed tomographic colonography (CTC), and colonoscopy, to detect colon polyps and cancers is uncertain. We aimed to assess the sensitivity of these three imaging tests. Methods Patients with faecal occult blood, haematochezia, iron-deficiency anaemia, or a family history of colon cancer underwent three separate colon-imaging studies—ACBE, followed 7–14 days later by CTC and colonoscopy on the same day. The primary outcome was detection of colonic polyps and cancers. Outcomes were assessed by building an aggregate view of the colon, taking into account results of all three tests. Findings 614 patients completed all three imaging tests. When analysed on a per-patient basis, for lesions 10 mm or larger in size (n=63), the sensitivity of ACBE was 48% (95% CI 35–61), CTC 59% (46–71, p=0·1083 for CTC vs ACBE), and colonoscopy 98% (91–100, p<0·0001 for colonoscopy vs CTC). For lesions 6–9 mm in size (n=116), sensitivity was 35% for ACBE (27–45), 51% for CTC (41–60, p=0·0080 for CTC vs ACBE), and 99% for colonoscopy (95–100, p<0·0001 for colonoscopy vs CTC). For lesions of 10 mm or larger in size, the specificity was greater for colonoscopy (0·996) than for either ACBE (0·90) or CTC (0·96) and declined for ACBE and CTC when smaller lesions were considered. Interpretation Colonoscopy was more sensitive than other tests, as currently undertaken, for detection of colonic polyps and cancers. These data have important implications for diagnostic use of colon imaging tests.
Introduction Air contrast barium enema (ACBE) and colonoscopy are imaging tests that are generally used to detect colon polyps and cancers. These tests have potential advantages and disadvantages: many clinicians prefer colonoscopy to ACBE for colonic imaging because they believe it offers the greatest accuracy; however, issues such as cost, safety, availability, and patients’ experience can favour ACBE.1–5 Computed tomographic colonography (CTC; also known as virtual colonoscopy) has been introduced in the past decade as a method of colonic imaging. Findings of studies vary, with some indicating that CTC is highly sensitive for detection of large colonic lesions6–10 whereas others show that sensitivity is limited.11,12 We postulated that ACBE, CTC, and colonoscopy would each detect significant colonic lesions effectively. The primary objective of the study was therefore to establish the comparative accuracy of these three procedures for the detection of large colon polyps and cancers in a cohort of patients at high risk for gastrointestinal colonic neoplasia.
Methods Patients We designed this prospective cohort study to recruit patients with a high likelihood of colon abnormalities www.thelancet.com Vol 365 January 22, 2005
undergoing colon assessment. Recruitment began in December, 2000 (protocol development began in early 2000); new recruitment was halted in October, 2003, and all patients completed testing by February, 2004. To be eligible, we required participants to have one of the following: one or more positive faecal occult blood tests; one or more episodes of bright red blood per rectum within the previous 3 months; iron-deficiency anaemia (defined as haemoglobin <130 g/L for men and <120 g/L for women on at least one laboratory measurement and abnormally low ferritin, iron-binding saturation, or absent bone marrow stores); or history of colon cancer or adenoma in a first-degree relative diagnosed before age 60 years, or any two first-degree relatives with colon cancer or adenoma diagnosed at any age. We excluded patients for the following reasons: active gastrointestinal haemorrhage (reported or witnessed haematemesis, melaenic stools, repeated haematochezia); serious medical illness within the previous 6 weeks; pregnancy, or woman of childbearing age not using birth control; previous colon surgery; normal colonoscopy within the previous 2 years; known inflammatory bowel disease; prisoners; age younger than 18 years; current participation in research involving drugs, medical devices, or biological interventions; need for special precautions in undertaking endoscopic procedures (eg, antibiotic prophylaxis); or weight 136 kg or greater.
Lancet 2005; 365: 305–11 Published online December 17, 2004 http://image.thelancet.com/ extras/04art11289web.pdf See Comment page 275 Durham VA Medical Center (W Davis MD, W L Foster MD), Duke University Medical Center (D C Rockey MD, E Paulson MD, D Niedzwiecki PhD, H B Bosworth PhD, L Sanders MPH, W M Thompson MD), Durham, NC, USA; San Francisco VA Medical Center, University of California at San Francisco, San Francisco, CA, USA (J Yee MD, K McQuaid MD); Troy Internal Medicine, Troy, MI, USA (J Henderson MD); Indian River Radiology, Vero Beach, FL, USA (P Hatten MD); University of Texas at Southwestern, Dallas, TX, USA (S Burdick MD); Virginia Commonwealth University Medical Center, Richmond, VA, USA (A Sanyal MD, R Halvorsen MD); University of Chicago, Chicago, IL, USA (D T Rubin MD, A Dachman MD); University of Medicine and Dentistry of New Jersey, Newark, NJ, USA (M Sterling MD); Seacoast Gastroenterology, Exeter, NH, USA (G Akerkar MD); University of Texas Medical Branch, Galveston, TX, USA (M S Bhutani MD); California Pacific Medical Center, San Francisco, CA, USA (K Binmoeller MD); University of California at San Diego, San Diego, CA, USA (J Garvie MD); and New York University, New York, NY, USA (E J Bini MD) Correspondence to: Dr Don C Rockey, Duke University Medical Center, Box 3083, Sands Building, Rm 334, Research Drive, Durham, NC 27710, USA [email protected]
305
Articles
The study was approved by every local institutional review board for clinical investigations and met all criteria put forth by the Declaration of Helsinki. All participants provided written informed consent before participation in the study. A data safety and monitoring board provided data and safety oversight.
Procedures We undertook ACBE according to standard guidelines.13 Before the procedure, we gave patients a clear liquid diet for 24 h, nothing to eat or drink after 0000 h, bisacodyl tablets (10 mg) in the evening, and bisacodyl suppository (10 mg) by rectum. We gave magnesium citrate (300 mL) the day before the examination. We undertook ACBE with high-density barium (100% w/v). After administration of barium and distension of the colon with room air, we took spot films of all specific colon segments. Additionally, we obtained overhead radiographs in the prone, prone 35º angled, supine, left and right lateral decubitus, and left lateral positions without the rectal tube.13,14 Glucagon was not used routinely. 7–14 days after ACBE, we prepared patients for CTC with a phosphate-based diarrhoeal drug (the standard preparation was a sodium phosphate-based osmotic laxative, 90 mL in total; stool tagging was not done), and undertook the procedure early the next morning. We placed a rectal tube with the patient in the left lateral decubitus position. After the individual was turned supine, the colon was insufflated with either room air or carbon dioxide. After distension of the colon, we obtained a scout radiograph to ensure adequate distension of the colon. If distension was inadequate, we administered additional room air or CO2. We scanned patients in both the supine and prone acquisitions (with identical parameters), using fourslice (n=384) or eight-slice (n=240) multidetector CT scanners. The nominal slice thickness was 2·5 mm with 1 mm reconstruction intervals. Table speed ranged from 7·5 mm/s to 15 mm/s. We scanned patients from the superior aspect of the distended colon to the rectum. Immediately on completion of CTC, we transferred images to a workstation for analysis. We interpreted the images with a primary two-dimensional (2D) reading technique with three-dimensional (3D) problemsolving, using Vitrea version 3.2 (Vital Imaging, Minneapolis, MN, USA; n=467), Advantage Windows version 4.0 or higher (General Electric Medical Systems, Milwaukee, WI, USA; n=85), or another program (n=62). We interpreted CTC immediately before undertaking colonoscopy. We did colonoscopy after CTC in the usual manner. We administered pain and sedative drugs intravenously. Lesions identified at the time of colonoscopy were measured in comparison with open biopsy forceps (7 mm in diameter), photographed, and (if retrieved) sent to pathology departments for analysis. 306
Figure 1: Enrolment details
We used local pathological assessments to determine histology. ACBE and CTC were undertaken by radiology technicians, attending doctors, or radiology residents supervised by an attending doctor. Attending radiologists interpreted all radiographic studies. Attending doctors or gastroenterology fellows supervised by attending gastroenterologists did colonoscopy. We did not provide feedback to readers during the study. The mean duration of experience for ACBE readers was 19 years (SD 7·6). Approximately half the CTC readers had experience of reading more than 50 cases: those who had not read more than 50 cases before beginning the study were required to complete a CTC training module (Joel Fletcher, Mayo Clinic). The mean duration of experience for clinicians doing colonoscopy was 14 years (SD 6·9). To investigate the effect of reader experience, we analysed CTC by comparing readers with experience of 50 cases or more with those who had read fewer than 50 cases. Study participants and investigators undertaking all tests were unaware of the findings detected during the other tests, which we assured by using a specific method of blinding and then reconciling findings. We used a specific technique, which we have termed segmental unblinding, during the colonoscopy procedure. Here, colonoscopy proceeds as usual, and during withdrawal of the colonoscope, the study coordinator recorded the findings of the colonoscopist before declaration of the ACBE and CTC results. This unblinding technique was applied in a segmental fashion corresponding to the major parts of the colon (caecum, ascending, transverse, descending, sigmoid, rectum) as the colonoscope was withdrawn. After withdrawal of the colonoscope from the segment in question, discrepant findings were reconciled by specific re-examination of the specific segment in question. With this methodology, the presence of a true lesion was established after reconciliation of all tests, by developing a consensus view of the colon. In situations where lesions larger than 6 mm were identified by either ACBE or CTC that could not be identified at the time of colonoscopy (including after www.thelancet.com Vol 365 January 22, 2005
Articles
suspected polypoid lesion, the certainty of the lesion was recorded (on a five-point scale, from definitely a lesion to definitely not a lesion). For imaging tests, lesion size was calculated with either a mm-calibrated ruler (ACBE) or an electronic ruler (CTC). For comparative analysis, the colon was divided into the six major colon segments described above. If a lesion identified by one test was within one segment of the lesion identified by another test, and was within 50% of the size of the lesion identified by colonoscopy (at the lower bound) or was any size larger than the lesion seen at colonoscopy, it was regarded as the same lesion. We developed a questionnaire about the patient’s experience with every test. During the study, we administered these questionnaires after every test, and a further separate questionnaire was done 3 days after completion of all tests. Details of this specific analysis will be presented separately.
Patients (n=614) Mean age (SD, years) Women Ethnic origin White Black Other Indication for colon assessment (%)* Haematochezia Faecal occult blood Family history of colon cancer Iron-deficiency anaemia
57 (10·0) 186 (30%) 430 (70%) 145 (24%) 39 (6%) 259 (42%) 234 (38%) 194 (32%) 49 (8%)
Date are number of patients (%), unless otherwise stated. *Some patients had more than one indication for colon evaluation.
Table 1: Patients’ characteristics
reconciliation), an independent committee (blindly) reviewed ACBE or CTC; if there was agreement about the presence of a colonic lesion at the time of re-review of ACBE or CTC, we recommended that all three tests be repeated. Assessment of colon preparation was carefully undertaken for all studies. We divided preparations into high quality or moderate quality. Quality was scored for every study on the basis of consistency and colour of the last bowel movement, presence of residual stool in the lumen and on the mucosa, the degree of colonic distension, and the reader’s estimation of their overall ability to use the study to assess for lesions. We developed scoring systems for all tests and applied these to the six major colon segments described above. We used a quantitative scale from 0 to 4 (0 represented no stool on the mucosa, no stool in the lumen, and complete ability to assess for lesions; 4 represented extensive stool coating the lumen, extensive stool in the lumen, and inability to assess for lesions). Assessments were undertaken for CTC in the prone and supine position. To analyse the effect of the quality of the preparation, we divided preparations into high quality and moderate quality. Any patient without a single assessment in any region of the colon scored as 3 or 4 (vs 0–2) based on the ability to assess parameters, was assigned to a high quality preparation and others to the moderate quality group. Additionally, for every
Statistical analysis The primary aim of the trial was to estimate sensitivity for every imaging modality within specified precision with 95% confidence. We planned two formal tests of hypothesis: a test of non-inferiority between CTC and colonoscopy for the detection of significant lesions, defined as cancers or lesions 1 cm in diameter or larger; and a test of superiority between CTC and ACBE for detection of significant lesions. Interim analyses were also done. Additionally, a futility analysis was incorporated, allowing for early stopping if the sensitivity of CTC was reported to be significantly inferior to that of colonoscopy. We calculated exact binomial CI estimates, adjusted for interim analyses, for prevalence, sensitivity, and specificity. At interim analyses, we used the Lan-DeMets analogue of the O’Brien-Fleming boundaries truncated at 3·5 (one-sided, =0·025). A CI approach was used to establish the non-inferiority of CTC versus colonoscopy.15,16 The variance estimate was calculated assuming unequal sensitivity of CTC and colonoscopy. We compared the sensitivities of ACBE and CTC by McNemar’s test (=0·025), and the sensitivity and specificity between modalities by McNemar’s test for per
Lesion size 10 mm Total lesions Caecum Ascending Transverse Descending Sigmoid Rectum Total
9 14 12 8 24 9 76
Lesion size 6–9 mm
Adenoma
Cancer
*Other
Total lesions
Adenoma
*Other
6 9 9 4 13 5 46
2 0 1 1 2 3 9
1 5 2 3 9 1 21
15 38 34 17 41 13 158
6 28 25 12 18 8 97
9 10 9 5 23 5 61
*Includes hyperplastic lesions, lipoma, normal mucosa, and five patients in whom histological information could not be retrieved (one in the 10 mm group).
Table 2: Location of colon lesions
www.thelancet.com Vol 365 January 22, 2005
307
Articles
Results
Figure 2: Colon lesions identified by ACBE, CTC, and colonoscopy Panels A–D show imaging studies for a patient with a 7 mm adenomatous polyp identified by ACBE (A), 2D CTC (B), 3D CTC (C), and colonoscopy (D). The lesion shown is in the descending colon. Panels E–H show imaging studies for a patient with a large rectal adenocarcinoma identified by ACBE (E), 2D CTC (F), 3D CTC (G), and colonoscopy (H). In (A) and (B), the arrow identifies the polypoid lesion. In (E) and (F), the arrow identifies the cancerous lesion.
patient, per lesion, and per pathology (adenoma or cancer) analyses within subgroups (5 mm; 6–9 mm; 6 mm; 10 mm). At the second interim analysis, the results of the futility analysis comparing the sensitivity of CTC and colonoscopy led the data safety and monitoring board to recommend that the trial be halted.
Role of the funding source The sponsor of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. Representatives from the National Cancer Institute (NCI) participated as ex-officio members of the data safety and monitoring board. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.
Sensitivity (95% CI) Per patient
p
Lesion size 10 mm Number 63 ACBE 0·48 (0·35–0·61) 0·1083 CTC 0·59 (0·45–0·71) <0·0001 Colonoscopy 0·984 (0·91–1·00) Lesion size 6–9 mm Number 116 ACBE 0·35 (0·27–0·45) 0·0080 CTC 0·51 (0·41–0·60) <0·0001 Colonoscopy 0·99 (0·95–1·00) Lesion size 6 mm Number 155 ACBE 0·41 (0·33–0·49) 0·0033 CTC 0·55 (0·47–0·63) <0·0001 Colonoscopy 0·987 (0·95–1·00)
Per lesion
p
Per histology
p
76 0·45 (0·33–0·57) 0·53 (0·41–0·64) 0·987 (0·93–1·00)
0·2008 <0·0001
55 0·55 (0·40–0·68) 0·64 (0·49–0·77) 0·982 (0·90–1·00)
0·2253 <0·0001
158 0·30 (0·23–0·38) 0·47 (0·39–0·56) 0·987 (0·96–1·00)
0·0005 <0·0001
97 0·31 (0·22–0·41) 0·60 (0·49–0·70) 1·00 (0·96–1·00)
<0·0001 <0·0001
234 0·35 (0·29–0·41) 0·49 (0·43–0·56) 0·987 (0·96–1·00)
0·0002 <0·0001
152 0·39 (0·32–0·48) 0·61 (0·53–0·69) 0·993 (0·96–1·00)
<0·0001 <0·0001
Sensitivity is based on lesion size and method of analysis—ie, whether per patient, per lesion, or per histology (adenoma or cancer). p values associated with ACBE are for ACBE vs CTC. p values associated with CTC are for CTC vs colonoscopy. p values are not shown for ACBE vs colonoscopy, but all were p<0·0001.
Table 3: Colon imaging test sensitivity
308
Figure 1 shows enrolment characteristics of the study cohort. We approached a total of 1409 potentially eligible subjects, of whom 634 were excluded. Reasons for exclusion included the following: refusal to participate or to complete all studies; inability to complete ACBE or CTC; and intercurrent medical illness. A substantial number of patients withdrew after ACBE, but before other testing. Reasons that patients did not complete all three studies included the following: 72 withdrew consent before study testing; 60 were lost to follow-up (including 28 after completion of ACBE); nine had inadequate preparations and therefore ACBE or CTC were aborted; five had adverse events unrelated to study procedures while they were enrolled and withdrew; technical difficulties precluded completion of CTC in five patients; the caecum was not reached during colonoscopy in four; and other reasons in six. 31 individuals received a waiver for enrolment for several reasons: some patients’ weight was greater than 136 kg but the scanner at that site could accommodate this weight; several people reporting a family history of colorectal cancer could not recall specific details; and some individuals who had undergone colonoscopy recently had de novo symptoms. The maximum number of patients examined at any one site was 165 and the minimum was two; the median number enrolled per site was 22·5 (IQR 16–66). Table 1 shows the demographics of the study population. As expected, because two Veterans’ Affairs hospitals participated, more men than women were included. The mean total time to complete ACBE was 47·9 min, 38·5 min for the procedure and 9·4 min for the doctor to interpret the test results. For CTC, the total time was 34·3 min, 17·4 min for the procedure and 16·9 min for interpretation. For colonoscopy, the total time in the endoscopy suite was 31·0 min, the time to reach the caecum was 10·3 min, and the average time spent on reconciliation was 4·1 min. No serious procedurerelated adverse events happened for any participant. 63 patients had 76 lesions of 10 mm or larger in size, of which 55 were either adenomas or cancers (table 2). 116 patients had 158 lesions of 6–9 mm in size, of which 97 were adenomas; none had evidence of adenocarcinoma. The likelihood of having adenomatous or adenocarcinoma histology was 72% in patients with lesions of 10 mm or larger in size and 61% in those with lesions of 6–9 mm in size. 366 individuals had a total of 1023 lesions of 5 mm or smaller in size, of which 400 had adenomatous histology and none had adenocarcinoma. Figure 2 shows examples of concordant lesions identified by the imaging tests. The sensitivity of colon imaging tests was calculated on a per-patient, per-lesion, and per-histology (adenoma or cancer) basis (table 3). Colonoscopy was the most sensitive test, irrespective of the outcome measured. For lesions of 10 mm or larger in size, both ACBE and CTC www.thelancet.com Vol 365 January 22, 2005
Articles
Specificity (95% CI)
p
Lesion size 10 mm (n=551) ACBE 0·90 (0·87–0·92) CTC 0·96 (0·94–0·98) Colonoscopy 0·996 (0·99–1·00) Lesion size 6 mm (n=459) ACBE 0·82 (0·78–0·85) CTC 0·89 (0·86–0·92) Colonoscopy 0·996 (0·98–1·00)
<0·0001 <0·0001
0·0007 <0·0001
Specificity is based on lesion size on a per-patient basis only. If specificity is calculated taking into account histology only—eg, colonoscopy identified non-neoplastic, normal, or hyperplastic lesions—the specificity of colonoscopy was 0·972 with lesion size 10 mm; for lesion size 6 mm, the point estimate is 0·90. p values associated with ACBE are for ACBE vs CTC. p values associated with CTC are for CTC vs colonoscopy.
Table 4: Colon imaging test specificity
did not detect 28 of 76 lesions when compared with the consensus colon result, ACBE did not detect 14 lesions that were detected by CTC, ACBE detected eight lesions that CTC did not detect, and both tests detected 26 of 76 lesions. One lesion of 10 mm or larger in size missed by colonoscopy was detected by both CTC and ACBE. For lesions of 6–9 mm in size, both ACBE and CTC failed to detect 65 of 158 lesions, ACBE did not detect 46 lesions that were detected by CTC, ACBE detected 18 lesions that CTC did not detect, and both tests detected 29 of 158 lesions. For lesions of 5 mm or smaller in size, per-patient sensitivities were 0·32 (120 of 375 lesions) for ACBE, 0·45 (167/375) for CTC, and 0·99 (370/375) for colonoscopy. The sensitivity of detection of adenocarcinoma was 0·89 (8/9) for ACBE, 0·78 (7/9) for CTC, and 1·00 (9/9) for colonoscopy. Four of these adenocarcinomas were Duke’s stage A, four were B, and one was C. Reconciliation was needed for 103 lesions identified by ACBE and for 83 lesions identified by CTC. In three patients, ACBE and CTC identified lesions in the same segment of bowel that were larger than 6 mm and could not be reconciled at the time of colonoscopy. Committee review of the data confirmed their concordance and a high suspicion that they represented genuine lesions; therefore, patients underwent a repeat of all three procedures. In these three individuals, the initially
ACBE 10 mm 6–9 mm CTC 10 mm 6–9 mm
High preparation quality
Moderate preparation quality
Total
p
21/45 (0·47) 33/90 (0·37)
9/18 (0·50) 7/25 (0·28)
30/63 (0·48) 40/115 (0·35)
0·4838
18/30 (0·60) 33/54 (0·61)
19/33 (0·58) 25/57 (0·44)
37/63 (0·59) 58/111 (0·52)
0·0877
Data are number of lesions seen/total lesions (sensitivity). Patients who did not have complete preparation data—ie, fields filled for all six segments, for all three tests—were excluded from the analysis.
Table 5: Colon imaging test sensitivity by preparation quality
www.thelancet.com Vol 365 January 22, 2005
identified lesion could not be seen by any of the three repeated procedures. The specificity for all tests was best when large lesions were considered (table 4). Although the specificity of ACBE and CTC for lesions of 10 mm or larger was high, for colonoscopy it was greater than that of either of the other tests, even taking into account the possibility that normal colonic or hyperplastic pathology could be a false-positive lesion. We assessed the role of the quality of the preparation on sensitivity of imaging procedures (table 5). Preparation quality did not clearly affect detection of large lesions for any test. Since sensitivity for colonoscopy was so high, differences in detection rates were not apparent on the basis of preparation quality (data not shown). In our additional analysis of CTC on the basis of reader experience, those with less case experience detected more lesions of 10 mm or larger than did those with the most experience (23/33 [70%] vs 14/30 [47%]; p=0·0775), and a similar finding was recorded for lesions of 6–9 mm (33/57 [58%] vs 26/59 [44%]; p=0·1435). CTC identified at least one extracolonic abnormality in 348 patients. Most of these lesions were not clinically significant. Abdominal aortic aneurysms were seen in 12 patients and malignant mass lesions in four other individuals.
Discussion We have shown that colonoscopy is more sensitive than either ACBE or CTC to detect polypoid lesions in the colon. Colonoscopy also has the greatest specificity. The strength of our study was that all patients received all three imaging tests, enabling comparison of all three methods. ACBE is an attractive imaging technique to examine the colon because it is widely available, inexpensive, and safe.4,17 For this reason, ACBE has been popular for many years (emphasised by the mean reader experience of 19 years). Despite these important advantages, we identified limitations for ACBE. While considering such limitations, we should highlight that the current study overcame several flaws noted in previous studies of ACBE and colonoscopy. First, we investigated a much larger number of patients than has been reported on previously. Second, our study design incorporated methods to overcome verification bias, ascertainment bias, and reference bias, all of which have been important issues in previous studies.4 Our reference standard was the combined dataset for all three colon studies rather than for colonoscopy alone—an important issue, since use of colonoscopy alone as a reference standard has been questioned because it can miss lesions, especially those behind folds.18 Although we implemented a rigorous study design, with the intent to optimise ACBE performance and to study ACBE as it is currently done in standard practice, the procedure was insensitive for detection of polypoid 309
Articles
lesions. Our data accord with those of two small studies that compared ACBE with endoscopy3,19 but contrast with most work in this area, in which a sensitivity for ACBE of 70–100% has been reported for detection of polyps.20–23 By contrast, we noted ACBE to be highly sensitive for detection of large cancerous lesions, in accordance with previous investigations.24,25 Thus, our findings suggest that ACBE is highly accurate for detection of large cancerous lesions, but if the desired result of ACBE is to detect (precancerous) polypoid lesions, then it might not be sensitive enough to warrant its use for this reason. CTC, as undertaken in our study, was not as sensitive as colonoscopy for detection of lesions of any size. However, our data must be interpreted in the greater context of published work on CTC. Several points are noteworthy. First, in our study, advanced (four or eight slice) multidetector CT scanners were used in all patients. Second, we used primary 2D reads with 3D problem-solving. This method differs from that in a previous report, in which CTC using primary 3D reading (in addition to stool tagging) was superior to colonoscopy for detection of colonic polypoid lesions of 10 mm or larger.9 Nonetheless, we currently do not know whether performance of CTC using 3D reading is superior to 2D reading, as it was in our study, and this specific issue needs further investigation. We should emphasise that most clinicians currently undertaking CTC use methods (including preparation and hardware and software) very similar to those used by us. Thus, in view of the controversy about the current sensitivity of CTC,6–9,11,12,26 comparisons of specific preparative and software techniques will be essential. What are the mechanisms underlying the inability to identify lesions? Technical or perception errors are both potentially important. We used consistent preparation techniques and quantified preparation quality so that we could begin to understand these mechanisms. The sensitivity of ACBE and CTC did not vary greatly by quality of preparations, suggesting that preparation quality throughout our study was adequate. These data indicate that in a setting of optimum reader interpretation, lesions are nonetheless missed. This finding raises the possibility that inherent technical issues such as inability to adequately distend the colon or clear it of fluid might limit the sensitivity of ACBE or CTC. Reader experience could have a role in perceptive errors, although for CTC, lesser-experienced readers did better than more experienced readers—a possible result of training. As for colonoscopy, lesions behind folds are the types most commonly missed.18 Thus, further study is needed to investigate the reasons underlying falsenegative and false-positive readings and to understand technical and perceptive attributes of colon imaging tests. We are unsure whether the results of this study can be generalised to colorectal-cancer screening. On one hand, we only included patients with an increased likelihood of having colorectal abnormalities (notably, the prevalence 310
of 10·3% [63/614 patients] was lower than predicted). On the other hand, we specifically designed the study to address sensitivity of the various colon imaging tests, as commonly used in practice, including in colorectalcancer screening. From the standpoint that a key objective of colorectal cancer screening is to detect earlystage lesions such as adenomas,27 our results are important and have substantial implications. However, the higher prevalence of underlying lesions than in a screening cohort could introduce bias because the increased likelihood of multiple lesions might lead to overestimation of per-patient sensitivity.4,17 Our results have important implications for clinicians considering use of colon imaging tests, and potentially for colon cancer screening programmes. As currently undertaken, the sensitivity of colonoscopy is superior to ACBE and CTC. Although the methods used to undertake ACBE are mature, those for CTC are rapidly evolving, and this test will almost certainly undergo significant further advances and improvements.28 As such, clinicians using this test must stay abreast of performance and patient preference issues. Contributors D C Rockey, E Paulson, J Yee, D Rubin, W L Foster, A Dachman, and R Halvorsen had the idea for and designed the study, analysed data, and wrote the report. D Niedzwiecki did the statistical analysis and wrote the report. W Davis had the idea for and designed the study. H B Bosworth had the idea for and designed the study, did the statistical analysis, and wrote the report. L Sanders did the statistical analysis. J Henderson, P Hatten, S Burdick, A Sanyal, M Sterling, G Akerkar, M S Bhutani, K Binmoeller, J Garvie, E J Bini, and B Thompson analysed data and wrote the report. K McQuaid had the idea for and designed the study and wrote the report. Study investigators L Ho, V Low, D Provenzale, D Tendler, N Agrawal, B O’Connor, J Wilson, M Byrne, S Brazer, K McGrath, B Broomer, L Fulford, M McCown, A Chaoui, C Murray, R Wille, J Weber, S Bologna, J Norconk, P Skaggs, J Wernicki, J Pelaez, W McCormack, R McGovern, C Brewington, T Forte, M Ulissey, L Bilhartz, J Sreenarasimhaiah, M Turner, L Carucci, D Switz, K Cho, M Simmons, A Raslavicus, M Marrero, R Ernst, A Oto, M Schreiber, S Faruqi, S Jafri, O S Ozkan, P J Pasricha, G S Raju, J M Shabot, N Snyder, E M Walser, E Baker, G Casola, G Ghahremani. Conflict of interest statement AD is a consultant for GE Health Care and E-Z-EM, has received research funding from E-Z-EM, and is co-patent holder for CAD software licensed to R2 Technologies. All other authors declare that they have no conflict of interest. Acknowledgments This work was supported by a grant from the NCI (grant R01 CA82344). This study was coordinated by the Duke Clinical Research Institute, Durham, NC, USA. We thank all study coordinators, nurses, and technical staff who cared for the patients involved in this study. References 1 Hixson LJ, Fennerty MB, Sampliner RE, Garewal HS. Prospective blinded trial of the colonoscopic miss-rate of large colorectal polyps. Gastrointest Endosc 1991; 37: 125–27. 2 Ott DJ. Barium enema: colorectal polyps and carcinoma. Semin Roentgenol 1996; 31: 125–41. 3 Winawer SJ, Stewart ET, Zauber AG, et al. A comparison of colonoscopy and double-contrast barium enema for surveillance after polypectomy. N Engl J Med 2000; 342: 1766–72.
www.thelancet.com Vol 365 January 22, 2005
Articles
4
5
6
7
8
9
10
11
12
13
14 15 16
Glick S. Double-contrast barium enema for colorectal cancer screening: a review of the issues and a comparison with other screening alternatives. AJR Am J Roentgenol 2000; 174: 1529–37. Steine S. Which hurts the most? A comparison of pain rating during double-contrast barium enema examination and colonoscopy. Radiology 1994; 191: 99–101. Fenlon HM, Nunes DP, Schroy PC, Barish MA, Clarke PD, Ferrucci JT. A comparison of virtual and conventional colonoscopy for the detection of colorectal polyps. N Engl J Med 1999; 341: 1496–503. Yee J, Akerkar GA, Hung RK, Steinauer-Gebauer AM, Wall SD, McQuaid KR. Colorectal neoplasia: performance characteristics of CT colonography for detection in 300 patients. Radiology 2001; 219: 685–92. McFarland EG, Pilgram TK, Brink JA, et al. CT colonography: multiobserver diagnostic performance. Radiology 2002; 225: 380–90. Pickhardt PJ, Choi JR, Hwang I, et al. Computed tomographic virtual colonoscopy to screen for colorectal neoplasia in asymptomatic adults. N Engl J Med 2003; 349: 2191–200. Johnson CD, Toledano AY, Herman BA, et al. Computerized tomographic colonography: performance evaluation in a retrospective multicenter setting. Gastroenterology 2003; 125: 688–95. Pineau BC, Paskett ED, Chen GJ, et al. Virtual colonoscopy using oral contrast compared with colonoscopy for the detection of patients with colorectal polyps. Gastroenterology 2003; 125: 304–10. Cotton PB, Durkalski VL, Pineau BC, et al. Computed tomographic colonography (virtual colonoscopy): a multicenter comparison with standard colonoscopy for detection of colorectal neoplasia. JAMA 2004; 291: 1713–19. Laufer I. Barium studies. In: Gore RM, Levine MS, ed. Textbook of gastrointestinal radiology. Philadelphia: W B Saunders, 1994: 1028–40. Simpkins KC. Double-contrast examination, part III: colon. Clin Gastroenterol 1984; 13: 99–121. Durrleman S, Simon R. Planning and monitoring of equivalence studies. Biometrics 1990; 46: 329–36. Amato DA. The design and analysis of equivalence trials. In: Finkelstein DM, Schoenfeld DA, eds. AIDS clinical trials. New York: Wiley-Liss, 1995: 155–76.
www.thelancet.com Vol 365 January 22, 2005
17
18
19
20
21
22
23
24
25
26
27
28
de Zwart IM, Griffioen G, Shaw MP, Lamers CB, de Roos A. Barium enema and endoscopy for the detection of colorectal neoplasia: sensitivity, specificity, complications and its determinants. Clin Radiol 2001; 56: 401–09. Pickhardt PJ, Nugent PA, Mysliwiec PA, Choi JR, Schindler WR. Location of adenomas missed by optical colonoscopy. Ann Intern Med 2004; 141: 352–59. Brewster NT, Grieve DC, Saunders JH. Double-contrast barium enema and flexible sigmoidoscopy for routine colonic investigation. Br J Surg 1994; 81: 445–47. Fork FT. Reliability of routine double contrast examination of the large bowel: a prospective study of 2590 patients. Gut 1983; 24: 672–77. Durdey P, Weston PM, Williams NS. Colonoscopy or barium enema as initial investigation of colonic disease. Lancet 1987; 2: 549–51. Ott DJ, Scharling ES, Chen YM, Wu WC, Gelfand DW. Barium enema examination: sensitivity in detecting colonic polyps and carcinomas. South Med J 1989; 82: 197–200. Jensen J, Kewenter J, Asztely M, Lycke G, Wojciechowski J. Double contrast barium enema and flexible rectosigmoidoscopy: a reliable diagnostic combination for detection of colorectal neoplasm. Br J Surg 1990; 77: 270–72. Brady AP, Stevenson GW, Stevenson I. Colorectal cancer overlooked at barium enema examination and colonoscopy: a continuing perceptual problem. Radiology 1994; 192: 373–78. Strom E, Larsen JL. Colon cancer at barium enema examination and colonoscopy: a study from the county of Hordaland, Norway. Radiology 1999; 211: 211–14. Johnson CD, Harmsen WS, Wilson LA, et al. Prospective blinded evaluation of computed tomographic colonography for screen detection of colorectal polyps. Gastroenterology 2003; 125: 311–19. Rex DK, Johnson DA, Lieberman DA, Burt RW, Sonnenberg A. Colorectal cancer prevention 2000: screening recommendations of the American College of Gastroenterology. Am J Gastroenterol 2000; 95: 868–77. Van Dam J, Cotton P, Johnson CD, et al. AGA future trends report: CT colonography. Gastroenterology 2004; 127: 970–84.
311
Analysis of air contrast barium enema, computed tomographic colonography, and colonoscopy: prospective comparison D C Rockey, E Paulson, D Niedzwiecki, W Davis, H B Bosworth, L Sanders, J Yee, J Henderson, P Hatten, S Burdick, A Sanyal, D T Rubin, M Sterling, G Akerkar, M S Bhutani, K Binmoeller, J Garvie, E J Bini, K McQuaid, W L Foster, W M Thompson, A Dachman, R Halvorsen
Summary Background The usefulness of currently available colon imaging tests, including air contrast barium enema (ACBE), computed tomographic colonography (CTC), and colonoscopy, to detect colon polyps and cancers is uncertain. We aimed to assess the sensitivity of these three imaging tests. Methods Patients with faecal occult blood, haematochezia, iron-deficiency anaemia, or a family history of colon cancer underwent three separate colon-imaging studies—ACBE, followed 7–14 days later by CTC and colonoscopy on the same day. The primary outcome was detection of colonic polyps and cancers. Outcomes were assessed by building an aggregate view of the colon, taking into account results of all three tests. Findings 614 patients completed all three imaging tests. When analysed on a per-patient basis, for lesions 10 mm or larger in size (n=63), the sensitivity of ACBE was 48% (95% CI 35–61), CTC 59% (46–71, p=0·1083 for CTC vs ACBE), and colonoscopy 98% (91–100, p<0·0001 for colonoscopy vs CTC). For lesions 6–9 mm in size (n=116), sensitivity was 35% for ACBE (27–45), 51% for CTC (41–60, p=0·0080 for CTC vs ACBE), and 99% for colonoscopy (95–100, p<0·0001 for colonoscopy vs CTC). For lesions of 10 mm or larger in size, the specificity was greater for colonoscopy (0·996) than for either ACBE (0·90) or CTC (0·96) and declined for ACBE and CTC when smaller lesions were considered. Interpretation Colonoscopy was more sensitive than other tests, as currently undertaken, for detection of colonic polyps and cancers. These data have important implications for diagnostic use of colon imaging tests.
Introduction Air contrast barium enema (ACBE) and colonoscopy are imaging tests that are generally used to detect colon polyps and cancers. These tests have potential advantages and disadvantages: many clinicians prefer colonoscopy to ACBE for colonic imaging because they believe it offers the greatest accuracy; however, issues such as cost, safety, availability, and patients’ experience can favour ACBE.1–5 Computed tomographic colonography (CTC; also known as virtual colonoscopy) has been introduced in the past decade as a method of colonic imaging. Findings of studies vary, with some indicating that CTC is highly sensitive for detection of large colonic lesions6–10 whereas others show that sensitivity is limited.11,12 We postulated that ACBE, CTC, and colonoscopy would each detect significant colonic lesions effectively. The primary objective of the study was therefore to establish the comparative accuracy of these three procedures for the detection of large colon polyps and cancers in a cohort of patients at high risk for gastrointestinal colonic neoplasia.
Methods Patients We designed this prospective cohort study to recruit patients with a high likelihood of colon abnormalities www.thelancet.com Vol 365 January 22, 2005
undergoing colon assessment. Recruitment began in December, 2000 (protocol development began in early 2000); new recruitment was halted in October, 2003, and all patients completed testing by February, 2004. To be eligible, we required participants to have one of the following: one or more positive faecal occult blood tests; one or more episodes of bright red blood per rectum within the previous 3 months; iron-deficiency anaemia (defined as haemoglobin <130 g/L for men and <120 g/L for women on at least one laboratory measurement and abnormally low ferritin, iron-binding saturation, or absent bone marrow stores); or history of colon cancer or adenoma in a first-degree relative diagnosed before age 60 years, or any two first-degree relatives with colon cancer or adenoma diagnosed at any age. We excluded patients for the following reasons: active gastrointestinal haemorrhage (reported or witnessed haematemesis, melaenic stools, repeated haematochezia); serious medical illness within the previous 6 weeks; pregnancy, or woman of childbearing age not using birth control; previous colon surgery; normal colonoscopy within the previous 2 years; known inflammatory bowel disease; prisoners; age younger than 18 years; current participation in research involving drugs, medical devices, or biological interventions; need for special precautions in undertaking endoscopic procedures (eg, antibiotic prophylaxis); or weight 136 kg or greater.
Lancet 2005; 365: 305–11 Published online December 17, 2004 http://image.thelancet.com/ extras/04art11289web.pdf See Comment page 275 Durham VA Medical Center (W Davis MD, W L Foster MD), Duke University Medical Center (D C Rockey MD, E Paulson MD, D Niedzwiecki PhD, H B Bosworth PhD, L Sanders MPH, W M Thompson MD), Durham, NC, USA; San Francisco VA Medical Center, University of California at San Francisco, San Francisco, CA, USA (J Yee MD, K McQuaid MD); Troy Internal Medicine, Troy, MI, USA (J Henderson MD); Indian River Radiology, Vero Beach, FL, USA (P Hatten MD); University of Texas at Southwestern, Dallas, TX, USA (S Burdick MD); Virginia Commonwealth University Medical Center, Richmond, VA, USA (A Sanyal MD, R Halvorsen MD); University of Chicago, Chicago, IL, USA (D T Rubin MD, A Dachman MD); University of Medicine and Dentistry of New Jersey, Newark, NJ, USA (M Sterling MD); Seacoast Gastroenterology, Exeter, NH, USA (G Akerkar MD); University of Texas Medical Branch, Galveston, TX, USA (M S Bhutani MD); California Pacific Medical Center, San Francisco, CA, USA (K Binmoeller MD); University of California at San Diego, San Diego, CA, USA (J Garvie MD); and New York University, New York, NY, USA (E J Bini MD) Correspondence to: Dr Don C Rockey, Duke University Medical Center, Box 3083, Sands Building, Rm 334, Research Drive, Durham, NC 27710, USA [email protected]
305
Articles
The study was approved by every local institutional review board for clinical investigations and met all criteria put forth by the Declaration of Helsinki. All participants provided written informed consent before participation in the study. A data safety and monitoring board provided data and safety oversight.
Procedures We undertook ACBE according to standard guidelines.13 Before the procedure, we gave patients a clear liquid diet for 24 h, nothing to eat or drink after 0000 h, bisacodyl tablets (10 mg) in the evening, and bisacodyl suppository (10 mg) by rectum. We gave magnesium citrate (300 mL) the day before the examination. We undertook ACBE with high-density barium (100% w/v). After administration of barium and distension of the colon with room air, we took spot films of all specific colon segments. Additionally, we obtained overhead radiographs in the prone, prone 35º angled, supine, left and right lateral decubitus, and left lateral positions without the rectal tube.13,14 Glucagon was not used routinely. 7–14 days after ACBE, we prepared patients for CTC with a phosphate-based diarrhoeal drug (the standard preparation was a sodium phosphate-based osmotic laxative, 90 mL in total; stool tagging was not done), and undertook the procedure early the next morning. We placed a rectal tube with the patient in the left lateral decubitus position. After the individual was turned supine, the colon was insufflated with either room air or carbon dioxide. After distension of the colon, we obtained a scout radiograph to ensure adequate distension of the colon. If distension was inadequate, we administered additional room air or CO2. We scanned patients in both the supine and prone acquisitions (with identical parameters), using fourslice (n=384) or eight-slice (n=240) multidetector CT scanners. The nominal slice thickness was 2·5 mm with 1 mm reconstruction intervals. Table speed ranged from 7·5 mm/s to 15 mm/s. We scanned patients from the superior aspect of the distended colon to the rectum. Immediately on completion of CTC, we transferred images to a workstation for analysis. We interpreted the images with a primary two-dimensional (2D) reading technique with three-dimensional (3D) problemsolving, using Vitrea version 3.2 (Vital Imaging, Minneapolis, MN, USA; n=467), Advantage Windows version 4.0 or higher (General Electric Medical Systems, Milwaukee, WI, USA; n=85), or another program (n=62). We interpreted CTC immediately before undertaking colonoscopy. We did colonoscopy after CTC in the usual manner. We administered pain and sedative drugs intravenously. Lesions identified at the time of colonoscopy were measured in comparison with open biopsy forceps (7 mm in diameter), photographed, and (if retrieved) sent to pathology departments for analysis. 306
Figure 1: Enrolment details
We used local pathological assessments to determine histology. ACBE and CTC were undertaken by radiology technicians, attending doctors, or radiology residents supervised by an attending doctor. Attending radiologists interpreted all radiographic studies. Attending doctors or gastroenterology fellows supervised by attending gastroenterologists did colonoscopy. We did not provide feedback to readers during the study. The mean duration of experience for ACBE readers was 19 years (SD 7·6). Approximately half the CTC readers had experience of reading more than 50 cases: those who had not read more than 50 cases before beginning the study were required to complete a CTC training module (Joel Fletcher, Mayo Clinic). The mean duration of experience for clinicians doing colonoscopy was 14 years (SD 6·9). To investigate the effect of reader experience, we analysed CTC by comparing readers with experience of 50 cases or more with those who had read fewer than 50 cases. Study participants and investigators undertaking all tests were unaware of the findings detected during the other tests, which we assured by using a specific method of blinding and then reconciling findings. We used a specific technique, which we have termed segmental unblinding, during the colonoscopy procedure. Here, colonoscopy proceeds as usual, and during withdrawal of the colonoscope, the study coordinator recorded the findings of the colonoscopist before declaration of the ACBE and CTC results. This unblinding technique was applied in a segmental fashion corresponding to the major parts of the colon (caecum, ascending, transverse, descending, sigmoid, rectum) as the colonoscope was withdrawn. After withdrawal of the colonoscope from the segment in question, discrepant findings were reconciled by specific re-examination of the specific segment in question. With this methodology, the presence of a true lesion was established after reconciliation of all tests, by developing a consensus view of the colon. In situations where lesions larger than 6 mm were identified by either ACBE or CTC that could not be identified at the time of colonoscopy (including after www.thelancet.com Vol 365 January 22, 2005
Articles
suspected polypoid lesion, the certainty of the lesion was recorded (on a five-point scale, from definitely a lesion to definitely not a lesion). For imaging tests, lesion size was calculated with either a mm-calibrated ruler (ACBE) or an electronic ruler (CTC). For comparative analysis, the colon was divided into the six major colon segments described above. If a lesion identified by one test was within one segment of the lesion identified by another test, and was within 50% of the size of the lesion identified by colonoscopy (at the lower bound) or was any size larger than the lesion seen at colonoscopy, it was regarded as the same lesion. We developed a questionnaire about the patient’s experience with every test. During the study, we administered these questionnaires after every test, and a further separate questionnaire was done 3 days after completion of all tests. Details of this specific analysis will be presented separately.
Patients (n=614) Mean age (SD, years) Women Ethnic origin White Black Other Indication for colon assessment (%)* Haematochezia Faecal occult blood Family history of colon cancer Iron-deficiency anaemia
57 (10·0) 186 (30%) 430 (70%) 145 (24%) 39 (6%) 259 (42%) 234 (38%) 194 (32%) 49 (8%)
Date are number of patients (%), unless otherwise stated. *Some patients had more than one indication for colon evaluation.
Table 1: Patients’ characteristics
reconciliation), an independent committee (blindly) reviewed ACBE or CTC; if there was agreement about the presence of a colonic lesion at the time of re-review of ACBE or CTC, we recommended that all three tests be repeated. Assessment of colon preparation was carefully undertaken for all studies. We divided preparations into high quality or moderate quality. Quality was scored for every study on the basis of consistency and colour of the last bowel movement, presence of residual stool in the lumen and on the mucosa, the degree of colonic distension, and the reader’s estimation of their overall ability to use the study to assess for lesions. We developed scoring systems for all tests and applied these to the six major colon segments described above. We used a quantitative scale from 0 to 4 (0 represented no stool on the mucosa, no stool in the lumen, and complete ability to assess for lesions; 4 represented extensive stool coating the lumen, extensive stool in the lumen, and inability to assess for lesions). Assessments were undertaken for CTC in the prone and supine position. To analyse the effect of the quality of the preparation, we divided preparations into high quality and moderate quality. Any patient without a single assessment in any region of the colon scored as 3 or 4 (vs 0–2) based on the ability to assess parameters, was assigned to a high quality preparation and others to the moderate quality group. Additionally, for every
Statistical analysis The primary aim of the trial was to estimate sensitivity for every imaging modality within specified precision with 95% confidence. We planned two formal tests of hypothesis: a test of non-inferiority between CTC and colonoscopy for the detection of significant lesions, defined as cancers or lesions 1 cm in diameter or larger; and a test of superiority between CTC and ACBE for detection of significant lesions. Interim analyses were also done. Additionally, a futility analysis was incorporated, allowing for early stopping if the sensitivity of CTC was reported to be significantly inferior to that of colonoscopy. We calculated exact binomial CI estimates, adjusted for interim analyses, for prevalence, sensitivity, and specificity. At interim analyses, we used the Lan-DeMets analogue of the O’Brien-Fleming boundaries truncated at 3·5 (one-sided, =0·025). A CI approach was used to establish the non-inferiority of CTC versus colonoscopy.15,16 The variance estimate was calculated assuming unequal sensitivity of CTC and colonoscopy. We compared the sensitivities of ACBE and CTC by McNemar’s test (=0·025), and the sensitivity and specificity between modalities by McNemar’s test for per
Lesion size 10 mm Total lesions Caecum Ascending Transverse Descending Sigmoid Rectum Total
9 14 12 8 24 9 76
Lesion size 6–9 mm
Adenoma
Cancer
*Other
Total lesions
Adenoma
*Other
6 9 9 4 13 5 46
2 0 1 1 2 3 9
1 5 2 3 9 1 21
15 38 34 17 41 13 158
6 28 25 12 18 8 97
9 10 9 5 23 5 61
*Includes hyperplastic lesions, lipoma, normal mucosa, and five patients in whom histological information could not be retrieved (one in the 10 mm group).
Table 2: Location of colon lesions
www.thelancet.com Vol 365 January 22, 2005
307
Articles
Results
Figure 2: Colon lesions identified by ACBE, CTC, and colonoscopy Panels A–D show imaging studies for a patient with a 7 mm adenomatous polyp identified by ACBE (A), 2D CTC (B), 3D CTC (C), and colonoscopy (D). The lesion shown is in the descending colon. Panels E–H show imaging studies for a patient with a large rectal adenocarcinoma identified by ACBE (E), 2D CTC (F), 3D CTC (G), and colonoscopy (H). In (A) and (B), the arrow identifies the polypoid lesion. In (E) and (F), the arrow identifies the cancerous lesion.
patient, per lesion, and per pathology (adenoma or cancer) analyses within subgroups (5 mm; 6–9 mm; 6 mm; 10 mm). At the second interim analysis, the results of the futility analysis comparing the sensitivity of CTC and colonoscopy led the data safety and monitoring board to recommend that the trial be halted.
Role of the funding source The sponsor of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. Representatives from the National Cancer Institute (NCI) participated as ex-officio members of the data safety and monitoring board. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.
Sensitivity (95% CI) Per patient
p
Lesion size 10 mm Number 63 ACBE 0·48 (0·35–0·61) 0·1083 CTC 0·59 (0·45–0·71) <0·0001 Colonoscopy 0·984 (0·91–1·00) Lesion size 6–9 mm Number 116 ACBE 0·35 (0·27–0·45) 0·0080 CTC 0·51 (0·41–0·60) <0·0001 Colonoscopy 0·99 (0·95–1·00) Lesion size 6 mm Number 155 ACBE 0·41 (0·33–0·49) 0·0033 CTC 0·55 (0·47–0·63) <0·0001 Colonoscopy 0·987 (0·95–1·00)
Per lesion
p
Per histology
p
76 0·45 (0·33–0·57) 0·53 (0·41–0·64) 0·987 (0·93–1·00)
0·2008 <0·0001
55 0·55 (0·40–0·68) 0·64 (0·49–0·77) 0·982 (0·90–1·00)
0·2253 <0·0001
158 0·30 (0·23–0·38) 0·47 (0·39–0·56) 0·987 (0·96–1·00)
0·0005 <0·0001
97 0·31 (0·22–0·41) 0·60 (0·49–0·70) 1·00 (0·96–1·00)
<0·0001 <0·0001
234 0·35 (0·29–0·41) 0·49 (0·43–0·56) 0·987 (0·96–1·00)
0·0002 <0·0001
152 0·39 (0·32–0·48) 0·61 (0·53–0·69) 0·993 (0·96–1·00)
<0·0001 <0·0001
Sensitivity is based on lesion size and method of analysis—ie, whether per patient, per lesion, or per histology (adenoma or cancer). p values associated with ACBE are for ACBE vs CTC. p values associated with CTC are for CTC vs colonoscopy. p values are not shown for ACBE vs colonoscopy, but all were p<0·0001.
Table 3: Colon imaging test sensitivity
308
Figure 1 shows enrolment characteristics of the study cohort. We approached a total of 1409 potentially eligible subjects, of whom 634 were excluded. Reasons for exclusion included the following: refusal to participate or to complete all studies; inability to complete ACBE or CTC; and intercurrent medical illness. A substantial number of patients withdrew after ACBE, but before other testing. Reasons that patients did not complete all three studies included the following: 72 withdrew consent before study testing; 60 were lost to follow-up (including 28 after completion of ACBE); nine had inadequate preparations and therefore ACBE or CTC were aborted; five had adverse events unrelated to study procedures while they were enrolled and withdrew; technical difficulties precluded completion of CTC in five patients; the caecum was not reached during colonoscopy in four; and other reasons in six. 31 individuals received a waiver for enrolment for several reasons: some patients’ weight was greater than 136 kg but the scanner at that site could accommodate this weight; several people reporting a family history of colorectal cancer could not recall specific details; and some individuals who had undergone colonoscopy recently had de novo symptoms. The maximum number of patients examined at any one site was 165 and the minimum was two; the median number enrolled per site was 22·5 (IQR 16–66). Table 1 shows the demographics of the study population. As expected, because two Veterans’ Affairs hospitals participated, more men than women were included. The mean total time to complete ACBE was 47·9 min, 38·5 min for the procedure and 9·4 min for the doctor to interpret the test results. For CTC, the total time was 34·3 min, 17·4 min for the procedure and 16·9 min for interpretation. For colonoscopy, the total time in the endoscopy suite was 31·0 min, the time to reach the caecum was 10·3 min, and the average time spent on reconciliation was 4·1 min. No serious procedurerelated adverse events happened for any participant. 63 patients had 76 lesions of 10 mm or larger in size, of which 55 were either adenomas or cancers (table 2). 116 patients had 158 lesions of 6–9 mm in size, of which 97 were adenomas; none had evidence of adenocarcinoma. The likelihood of having adenomatous or adenocarcinoma histology was 72% in patients with lesions of 10 mm or larger in size and 61% in those with lesions of 6–9 mm in size. 366 individuals had a total of 1023 lesions of 5 mm or smaller in size, of which 400 had adenomatous histology and none had adenocarcinoma. Figure 2 shows examples of concordant lesions identified by the imaging tests. The sensitivity of colon imaging tests was calculated on a per-patient, per-lesion, and per-histology (adenoma or cancer) basis (table 3). Colonoscopy was the most sensitive test, irrespective of the outcome measured. For lesions of 10 mm or larger in size, both ACBE and CTC www.thelancet.com Vol 365 January 22, 2005
Articles
Specificity (95% CI)
p
Lesion size 10 mm (n=551) ACBE 0·90 (0·87–0·92) CTC 0·96 (0·94–0·98) Colonoscopy 0·996 (0·99–1·00) Lesion size 6 mm (n=459) ACBE 0·82 (0·78–0·85) CTC 0·89 (0·86–0·92) Colonoscopy 0·996 (0·98–1·00)
<0·0001 <0·0001
0·0007 <0·0001
Specificity is based on lesion size on a per-patient basis only. If specificity is calculated taking into account histology only—eg, colonoscopy identified non-neoplastic, normal, or hyperplastic lesions—the specificity of colonoscopy was 0·972 with lesion size 10 mm; for lesion size 6 mm, the point estimate is 0·90. p values associated with ACBE are for ACBE vs CTC. p values associated with CTC are for CTC vs colonoscopy.
Table 4: Colon imaging test specificity
did not detect 28 of 76 lesions when compared with the consensus colon result, ACBE did not detect 14 lesions that were detected by CTC, ACBE detected eight lesions that CTC did not detect, and both tests detected 26 of 76 lesions. One lesion of 10 mm or larger in size missed by colonoscopy was detected by both CTC and ACBE. For lesions of 6–9 mm in size, both ACBE and CTC failed to detect 65 of 158 lesions, ACBE did not detect 46 lesions that were detected by CTC, ACBE detected 18 lesions that CTC did not detect, and both tests detected 29 of 158 lesions. For lesions of 5 mm or smaller in size, per-patient sensitivities were 0·32 (120 of 375 lesions) for ACBE, 0·45 (167/375) for CTC, and 0·99 (370/375) for colonoscopy. The sensitivity of detection of adenocarcinoma was 0·89 (8/9) for ACBE, 0·78 (7/9) for CTC, and 1·00 (9/9) for colonoscopy. Four of these adenocarcinomas were Duke’s stage A, four were B, and one was C. Reconciliation was needed for 103 lesions identified by ACBE and for 83 lesions identified by CTC. In three patients, ACBE and CTC identified lesions in the same segment of bowel that were larger than 6 mm and could not be reconciled at the time of colonoscopy. Committee review of the data confirmed their concordance and a high suspicion that they represented genuine lesions; therefore, patients underwent a repeat of all three procedures. In these three individuals, the initially
ACBE 10 mm 6–9 mm CTC 10 mm 6–9 mm
High preparation quality
Moderate preparation quality
Total
p
21/45 (0·47) 33/90 (0·37)
9/18 (0·50) 7/25 (0·28)
30/63 (0·48) 40/115 (0·35)
0·4838
18/30 (0·60) 33/54 (0·61)
19/33 (0·58) 25/57 (0·44)
37/63 (0·59) 58/111 (0·52)
0·0877
Data are number of lesions seen/total lesions (sensitivity). Patients who did not have complete preparation data—ie, fields filled for all six segments, for all three tests—were excluded from the analysis.
Table 5: Colon imaging test sensitivity by preparation quality
www.thelancet.com Vol 365 January 22, 2005
identified lesion could not be seen by any of the three repeated procedures. The specificity for all tests was best when large lesions were considered (table 4). Although the specificity of ACBE and CTC for lesions of 10 mm or larger was high, for colonoscopy it was greater than that of either of the other tests, even taking into account the possibility that normal colonic or hyperplastic pathology could be a false-positive lesion. We assessed the role of the quality of the preparation on sensitivity of imaging procedures (table 5). Preparation quality did not clearly affect detection of large lesions for any test. Since sensitivity for colonoscopy was so high, differences in detection rates were not apparent on the basis of preparation quality (data not shown). In our additional analysis of CTC on the basis of reader experience, those with less case experience detected more lesions of 10 mm or larger than did those with the most experience (23/33 [70%] vs 14/30 [47%]; p=0·0775), and a similar finding was recorded for lesions of 6–9 mm (33/57 [58%] vs 26/59 [44%]; p=0·1435). CTC identified at least one extracolonic abnormality in 348 patients. Most of these lesions were not clinically significant. Abdominal aortic aneurysms were seen in 12 patients and malignant mass lesions in four other individuals.
Discussion We have shown that colonoscopy is more sensitive than either ACBE or CTC to detect polypoid lesions in the colon. Colonoscopy also has the greatest specificity. The strength of our study was that all patients received all three imaging tests, enabling comparison of all three methods. ACBE is an attractive imaging technique to examine the colon because it is widely available, inexpensive, and safe.4,17 For this reason, ACBE has been popular for many years (emphasised by the mean reader experience of 19 years). Despite these important advantages, we identified limitations for ACBE. While considering such limitations, we should highlight that the current study overcame several flaws noted in previous studies of ACBE and colonoscopy. First, we investigated a much larger number of patients than has been reported on previously. Second, our study design incorporated methods to overcome verification bias, ascertainment bias, and reference bias, all of which have been important issues in previous studies.4 Our reference standard was the combined dataset for all three colon studies rather than for colonoscopy alone—an important issue, since use of colonoscopy alone as a reference standard has been questioned because it can miss lesions, especially those behind folds.18 Although we implemented a rigorous study design, with the intent to optimise ACBE performance and to study ACBE as it is currently done in standard practice, the procedure was insensitive for detection of polypoid 309
Articles
lesions. Our data accord with those of two small studies that compared ACBE with endoscopy3,19 but contrast with most work in this area, in which a sensitivity for ACBE of 70–100% has been reported for detection of polyps.20–23 By contrast, we noted ACBE to be highly sensitive for detection of large cancerous lesions, in accordance with previous investigations.24,25 Thus, our findings suggest that ACBE is highly accurate for detection of large cancerous lesions, but if the desired result of ACBE is to detect (precancerous) polypoid lesions, then it might not be sensitive enough to warrant its use for this reason. CTC, as undertaken in our study, was not as sensitive as colonoscopy for detection of lesions of any size. However, our data must be interpreted in the greater context of published work on CTC. Several points are noteworthy. First, in our study, advanced (four or eight slice) multidetector CT scanners were used in all patients. Second, we used primary 2D reads with 3D problem-solving. This method differs from that in a previous report, in which CTC using primary 3D reading (in addition to stool tagging) was superior to colonoscopy for detection of colonic polypoid lesions of 10 mm or larger.9 Nonetheless, we currently do not know whether performance of CTC using 3D reading is superior to 2D reading, as it was in our study, and this specific issue needs further investigation. We should emphasise that most clinicians currently undertaking CTC use methods (including preparation and hardware and software) very similar to those used by us. Thus, in view of the controversy about the current sensitivity of CTC,6–9,11,12,26 comparisons of specific preparative and software techniques will be essential. What are the mechanisms underlying the inability to identify lesions? Technical or perception errors are both potentially important. We used consistent preparation techniques and quantified preparation quality so that we could begin to understand these mechanisms. The sensitivity of ACBE and CTC did not vary greatly by quality of preparations, suggesting that preparation quality throughout our study was adequate. These data indicate that in a setting of optimum reader interpretation, lesions are nonetheless missed. This finding raises the possibility that inherent technical issues such as inability to adequately distend the colon or clear it of fluid might limit the sensitivity of ACBE or CTC. Reader experience could have a role in perceptive errors, although for CTC, lesser-experienced readers did better than more experienced readers—a possible result of training. As for colonoscopy, lesions behind folds are the types most commonly missed.18 Thus, further study is needed to investigate the reasons underlying falsenegative and false-positive readings and to understand technical and perceptive attributes of colon imaging tests. We are unsure whether the results of this study can be generalised to colorectal-cancer screening. On one hand, we only included patients with an increased likelihood of having colorectal abnormalities (notably, the prevalence 310
of 10·3% [63/614 patients] was lower than predicted). On the other hand, we specifically designed the study to address sensitivity of the various colon imaging tests, as commonly used in practice, including in colorectalcancer screening. From the standpoint that a key objective of colorectal cancer screening is to detect earlystage lesions such as adenomas,27 our results are important and have substantial implications. However, the higher prevalence of underlying lesions than in a screening cohort could introduce bias because the increased likelihood of multiple lesions might lead to overestimation of per-patient sensitivity.4,17 Our results have important implications for clinicians considering use of colon imaging tests, and potentially for colon cancer screening programmes. As currently undertaken, the sensitivity of colonoscopy is superior to ACBE and CTC. Although the methods used to undertake ACBE are mature, those for CTC are rapidly evolving, and this test will almost certainly undergo significant further advances and improvements.28 As such, clinicians using this test must stay abreast of performance and patient preference issues. Contributors D C Rockey, E Paulson, J Yee, D Rubin, W L Foster, A Dachman, and R Halvorsen had the idea for and designed the study, analysed data, and wrote the report. D Niedzwiecki did the statistical analysis and wrote the report. W Davis had the idea for and designed the study. H B Bosworth had the idea for and designed the study, did the statistical analysis, and wrote the report. L Sanders did the statistical analysis. J Henderson, P Hatten, S Burdick, A Sanyal, M Sterling, G Akerkar, M S Bhutani, K Binmoeller, J Garvie, E J Bini, and B Thompson analysed data and wrote the report. K McQuaid had the idea for and designed the study and wrote the report. Study investigators L Ho, V Low, D Provenzale, D Tendler, N Agrawal, B O’Connor, J Wilson, M Byrne, S Brazer, K McGrath, B Broomer, L Fulford, M McCown, A Chaoui, C Murray, R Wille, J Weber, S Bologna, J Norconk, P Skaggs, J Wernicki, J Pelaez, W McCormack, R McGovern, C Brewington, T Forte, M Ulissey, L Bilhartz, J Sreenarasimhaiah, M Turner, L Carucci, D Switz, K Cho, M Simmons, A Raslavicus, M Marrero, R Ernst, A Oto, M Schreiber, S Faruqi, S Jafri, O S Ozkan, P J Pasricha, G S Raju, J M Shabot, N Snyder, E M Walser, E Baker, G Casola, G Ghahremani. Conflict of interest statement AD is a consultant for GE Health Care and E-Z-EM, has received research funding from E-Z-EM, and is co-patent holder for CAD software licensed to R2 Technologies. All other authors declare that they have no conflict of interest. Acknowledgments This work was supported by a grant from the NCI (grant R01 CA82344). This study was coordinated by the Duke Clinical Research Institute, Durham, NC, USA. We thank all study coordinators, nurses, and technical staff who cared for the patients involved in this study. References 1 Hixson LJ, Fennerty MB, Sampliner RE, Garewal HS. Prospective blinded trial of the colonoscopic miss-rate of large colorectal polyps. Gastrointest Endosc 1991; 37: 125–27. 2 Ott DJ. Barium enema: colorectal polyps and carcinoma. Semin Roentgenol 1996; 31: 125–41. 3 Winawer SJ, Stewart ET, Zauber AG, et al. A comparison of colonoscopy and double-contrast barium enema for surveillance after polypectomy. N Engl J Med 2000; 342: 1766–72.
www.thelancet.com Vol 365 January 22, 2005
Articles
4
5
6
7
8
9
10
11
12
13
14 15 16
Glick S. Double-contrast barium enema for colorectal cancer screening: a review of the issues and a comparison with other screening alternatives. AJR Am J Roentgenol 2000; 174: 1529–37. Steine S. Which hurts the most? A comparison of pain rating during double-contrast barium enema examination and colonoscopy. Radiology 1994; 191: 99–101. Fenlon HM, Nunes DP, Schroy PC, Barish MA, Clarke PD, Ferrucci JT. A comparison of virtual and conventional colonoscopy for the detection of colorectal polyps. N Engl J Med 1999; 341: 1496–503. Yee J, Akerkar GA, Hung RK, Steinauer-Gebauer AM, Wall SD, McQuaid KR. Colorectal neoplasia: performance characteristics of CT colonography for detection in 300 patients. Radiology 2001; 219: 685–92. McFarland EG, Pilgram TK, Brink JA, et al. CT colonography: multiobserver diagnostic performance. Radiology 2002; 225: 380–90. Pickhardt PJ, Choi JR, Hwang I, et al. Computed tomographic virtual colonoscopy to screen for colorectal neoplasia in asymptomatic adults. N Engl J Med 2003; 349: 2191–200. Johnson CD, Toledano AY, Herman BA, et al. Computerized tomographic colonography: performance evaluation in a retrospective multicenter setting. Gastroenterology 2003; 125: 688–95. Pineau BC, Paskett ED, Chen GJ, et al. Virtual colonoscopy using oral contrast compared with colonoscopy for the detection of patients with colorectal polyps. Gastroenterology 2003; 125: 304–10. Cotton PB, Durkalski VL, Pineau BC, et al. Computed tomographic colonography (virtual colonoscopy): a multicenter comparison with standard colonoscopy for detection of colorectal neoplasia. JAMA 2004; 291: 1713–19. Laufer I. Barium studies. In: Gore RM, Levine MS, ed. Textbook of gastrointestinal radiology. Philadelphia: W B Saunders, 1994: 1028–40. Simpkins KC. Double-contrast examination, part III: colon. Clin Gastroenterol 1984; 13: 99–121. Durrleman S, Simon R. Planning and monitoring of equivalence studies. Biometrics 1990; 46: 329–36. Amato DA. The design and analysis of equivalence trials. In: Finkelstein DM, Schoenfeld DA, eds. AIDS clinical trials. New York: Wiley-Liss, 1995: 155–76.
www.thelancet.com Vol 365 January 22, 2005
17
18
19
20
21
22
23
24
25
26
27
28
de Zwart IM, Griffioen G, Shaw MP, Lamers CB, de Roos A. Barium enema and endoscopy for the detection of colorectal neoplasia: sensitivity, specificity, complications and its determinants. Clin Radiol 2001; 56: 401–09. Pickhardt PJ, Nugent PA, Mysliwiec PA, Choi JR, Schindler WR. Location of adenomas missed by optical colonoscopy. Ann Intern Med 2004; 141: 352–59. Brewster NT, Grieve DC, Saunders JH. Double-contrast barium enema and flexible sigmoidoscopy for routine colonic investigation. Br J Surg 1994; 81: 445–47. Fork FT. Reliability of routine double contrast examination of the large bowel: a prospective study of 2590 patients. Gut 1983; 24: 672–77. Durdey P, Weston PM, Williams NS. Colonoscopy or barium enema as initial investigation of colonic disease. Lancet 1987; 2: 549–51. Ott DJ, Scharling ES, Chen YM, Wu WC, Gelfand DW. Barium enema examination: sensitivity in detecting colonic polyps and carcinomas. South Med J 1989; 82: 197–200. Jensen J, Kewenter J, Asztely M, Lycke G, Wojciechowski J. Double contrast barium enema and flexible rectosigmoidoscopy: a reliable diagnostic combination for detection of colorectal neoplasm. Br J Surg 1990; 77: 270–72. Brady AP, Stevenson GW, Stevenson I. Colorectal cancer overlooked at barium enema examination and colonoscopy: a continuing perceptual problem. Radiology 1994; 192: 373–78. Strom E, Larsen JL. Colon cancer at barium enema examination and colonoscopy: a study from the county of Hordaland, Norway. Radiology 1999; 211: 211–14. Johnson CD, Harmsen WS, Wilson LA, et al. Prospective blinded evaluation of computed tomographic colonography for screen detection of colorectal polyps. Gastroenterology 2003; 125: 311–19. Rex DK, Johnson DA, Lieberman DA, Burt RW, Sonnenberg A. Colorectal cancer prevention 2000: screening recommendations of the American College of Gastroenterology. Am J Gastroenterol 2000; 95: 868–77. Van Dam J, Cotton P, Johnson CD, et al. AGA future trends report: CT colonography. Gastroenterology 2004; 127: 970–84.
311