| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Radiology, Beth Israel Deaconess Medical Center, Harvard Medical
School, 330 Brookline Ave., Boston, MA 02215.
2 Boston Biostatistics Research Foundation, 4 California Ave., Framingham, MA
01701.
Received March 6, 2003;
accepted after revision June 3, 2003.
Address correspondence to J. Sosna
(jsosna{at}bidmc.harvard.edu).
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Articles comparing CT colonography and
conventional colonoscopy were identified, and a standardized form was used to
extract relevant study data. Fisher's exact test and the Mantel-Haenszel test
were used for pooling of data. A 95% confidence interval (CI) was selected to
determine sensitivity and specificity, and the Kruskal-Wallis exact test was
used to identify trends relating to polyp size. Meta-analysis methods were
used to test strength of results. Comparisons were made for the percentage of
polyps detected grouped by size (
10 mm, 6–9 mm, 
5 mm) and the
percentage of patients identified who had polyps of the same size.
RESULTS. Fourteen studies fulfilled all the study inclusion criteria and gave a total of 1,324 patients and 1,411 polyps. The pooled per-patient sensitivity for polyps 10 mm or larger was (sensitivity [95% CI]) 0.88 (0.84–0.93), for polyps 6–9 mm it was 0.84 (0.80–0.89), and for polyps 5 mm or smaller it was 0.65 (0.57–0.73). The pooled per-polyp sensitivity for polyps 10 mm or larger was 0.81 (0.76–0.85), for polyps 6–9 mm it was 0.62 (0.58–0.67), and for polyps 5 mm or smaller it was 0.43 (0.39–0.47). Sensitivity for detection of polyps increased as the polyp size increased (p < 0.00005). The pooled overall specificity for detection of polyps larger than 10 mm was 0.95 (0.94–0.97).
CONCLUSION. The specificity and sensitivity of CT colonography are high for polyps larger than 10 mm.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Study Eligibility
Studies were included only if they satisfied all the following inclusion
criteria: prospective studies, studies involving patients with full colorectal
preparation who underwent CT colonoscopy in which the entire colon was
evaluated, studies in which the reference standard was full conventional
colonoscopy, studies in which reviewers were blinded to the results of
conventional colonoscopy, studies in which CT colonoscopy was performed in
both the prone and supine positions after insufflation of air or
CO2, studies performed using at least a single-detector CT scanner
with slice thickness no greater than 5 mm, studies in which both 2- and 3D
(full or partial) analyses were performed as part of the evaluation, and
studies that reported performance results in absolute numbers as well as in
percentages for polyps or masses of different sizes. In some studies, only a
subgroup of patients fulfilled the inclusion criteria. Of 146 articles on CT
colonography, a total of 14 fulfilled the inclusion criteria
[5–18]
(Table 1).
|
Data Extraction
Two investigators extracted the data from each article separately, and when
disagreement occurred, it was resolved by consensus. The investigators were
aware of the site of origin of publication, journal, and year of publication.
The following data were recorded for each article: author and year of
publication, sample size, CT colonography technique, number of lesions per
study, sensitivity for detection of lesions by size ( 10 mm, 6–9 mm,
5 mm), number of patients per study, sensitivity for detection of polyps
of those sizes, and specificity for detection of polyps greater than 10
mm.
Data Analysis
The suitability for data pooling across studies was assessed using an exact
contingency table test. Fisher's exact test and the Mantel-Haenszel test were
used for pooling of data. Supporting analyses were performed to test the
impact of pooling data across studies using the following three study pooling
algorithms: all individual studies with data combined irrespective of the
sample size, all studies with data combined by weighting individual studies by
the square root of the sample size, and all individual studies with data
combined irrespective of sample size after removing outlier studies one at a
time to meet a pooling criterion of a p value greater than 0.05 until
all remaining studies could be pooled.
A 95% confidence interval (CI) was selected to determine sensitivity and
specificity, and the Kruskal-Wallis exact test was used to identify trends and
to test whether the three categories of polyp size were identically
distributed. Specific comparisons were made between CT colonography and
conventional colonoscopy, including the percentage of polyps detected grouped
by size ( 10 mm, 6–9 mm, 5 mm) and the percentage of patients
who had polyps, also grouped by size. For each data combination algorithm, a
95% CI was constructed for the sensitivity per polyp, for the sensitivity per
patient, and for specificity for polyps greater than 10 mm.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Study characteristics are given in Table 1. Seven of 14 studies had 50 subjects or fewer, four of 14 had between 50 and 100, and three of 14 had more than 180 patients. Twelve (86%) of 14 studies included high-risk patients. All but one study [17] had double interpretations of the CT colonography studies. One study was performed using a MDCT scanner [18].
Data Analysis of Per-Patient Sensitivity
A total of 1,324 patients were included in the 14 articles analyzed
(Table 2). One hundred
eighty-seven patients reported in six studies had polyps 10 mm or larger as
seen on conventional colonoscopy, of which 165 were identified on CT
colonography. The pooled per-patient sensitivity for polyps 10 mm or larger
was 0.88 (0.84–0.93) with a corresponding 0.85–0.88 range for the
other pooling algorithms. Results were consistent, although pooling of all
available studies was not justified (p = 0.0162); study 3 and study
12 (as numbered in Table 1)
were outliers. When the results were limited to the poolable studies, the
pooled sensitivity fell to 0.85 (107/126) (0.79–0.91) with an
overlapping 95% CI (Fig.
1).
|
|
Two hundred twenty-four patients reported in five studies had polyps 6–9 mm seen on conventional colonoscopy, of which 189 were identified on CT colonoscopy (Table 3). The pooled per-patient sensitivity for polyps 6–9 mm was 0.84 (0.80–0.89) with a corresponding 0.84–0.89 range for the other pooling algorithms (Table 4). Results were consistent even though study pooling of all available studies was not justified (p = 0.0011); studies 3, 5, and 6 were outliers. When the results were limited to the poolable studies, the pooled sensitivity rose to 0.89 (164/184) (0.85–0.94) with an overlapping 95% CI (Fig. 1).
|
|
One hundred thirty-three patients reported in four studies had polyps less than or equal to 5 mm seen on conventional colonoscopy, of which 87 were identified on CT colonography (Table 3). The pooled per-patient sensitivity for polyps less than or equal to 5 mm was 0.65 (0.57–0.73) with a corresponding 0.65–0.80 range for the other pooling algorithms (Table 4). Results were consistent even though study pooling of all available studies was not justified (p < 0.00005); studies 3 and 5 were outliers. When the results were limited to the poolable studies, the pooled sensitivity rose to 0.80 (70/87) (0.72–0.89) with a nonoverlapping 95% CI (Fig. 1).
Data Analysis of Per-Polyp Sensitivity
A total of 1,411 polyps were evaluated in the 14 articles analyzed
(Table 2). Three hundred sixty
polyps greater than or equal to 10 mm were seen on conventional colonoscopy,
of which 290 were identified on CT colonography. Sensitivity values were
available from all 14 studies. The pooled per-polyp sensitivity for polyps
greater than or equal to 10 mm was 0.81 (0.76–0.85) with a corresponding
0.80–0.81 range for the other pooling algorithms
(Table 4). Results were
consistent even though study pooling of all available studies was not
justified (p = 0.0048); studies 1, 5, and 11 were outliers. When the
results were limited to the poolable studies, the pooled sensitivity was also
0.81 (30/37) (0.68–0.94) with an overlapping 95% CI
(Fig. 2).
|
Four hundred sixty-eight polyps 6–9 mm were seen on conventional colonoscopy, of which 291 were identified on CT colonography and reported in 12 studies (Table 3). The pooled per-polyp sensitivity for polyps 6–9 mm was 0.62 (0.58–0.67) with a corresponding 0.45–0.63 range for the other pooling algorithms (Table 4). Results were consistent even though study pooling of all available studies was not justified (p < 0.00005); studies 2, 5, 7, 9, 13, and 14 were outliers. When the results were limited to the poolable studies, the pooled sensitivity fell to 0.45 (101/224) (0.39–0.52) with a nonoverlapping 95% CI (Fig. 2).
Five hundred seventy-two polyps 5 mm or less were seen on conventional colonoscopy, of which 247 were identified on CT colonography and reported in nine studies (Table 3). The pooled per-polyp sensitivity for polyps 5 mm or less was 0.43 (0.39–0.47) with a corresponding 0.17–0.50 range for the other pooling algorithms (Table 4). Results were consistent, although study pooling of all available studies was not justified (p < 0.00005); studies 1, 2, 3, 5, 11, and 14 were outliers. When the results were limited to the poolable studies, the pooled sensitivity fell to 0.17 (36/207) (0.12–0.23) with a nonoverlapping 95% CI (Fig. 2).
Data Analysis of Trends
The Kruskal-Wallis exact test for trends showed increasing sensitivity for
detecting polyps as their size increased (p < 0.00005) for
per-patient sensitivity (across all studies) and per-polyp sensitivity (across
all studies as well as for poolable studies).
Specificity
CT colonography confirmed the absence of polyps 10 mm or larger in 732 of
768 patients described in nine studies
(Table 4). The pooled
specificity for polyps 10 mm or larger was 0.95 (0.94–0.97) with a
corresponding 0.95–0.96 range for the other pooling algorithms
(Table 4). Pooling of all
studies was justified (p = 0.9996).
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
We analyzed only articles published in the English language from American as well as European journals because these could be reviewed and evaluated without language obstacles. Only a relatively small number of studies published in the literature addressed the comparison between CT colonography and conventional colonoscopy directly. The other publications were either reviews or letters or addressed different technical aspects of the usage of CT colonography. Of the studies that compared CT colonography and conventional colonoscopy, 70% fulfilled all our inclusion criteria and had in common the usage of an acceptable CT colonography technique: dual positioning; at most 5-mm slice thickness, 2- and 3D reviewing, and at least a single-detector helical CT, if not multidetector, scanning capability.
The studies analyzed differ regarding technical factors such as pitch and reconstruction interval (Table 1). To the best of our knowledge, these factors have not been proven to alter the performance of CT colonography in clinical settings [27]. Tube current was also variable (50–260 mA), but even lower tube currents were recently shown to have good diagnostic yield [28]. The effect of using different oral bowel-cleansing solution preparations has never been proven to affect diagnostic yield directly [13]. Although the results on CT colonography are compared with the traditional gold standard of conventional colonoscopy, the latter is not perfect. Six percent of polyps were missed in a back-to-back conventional colonoscopy study [29]. Therefore, the comparison of CT colonography is not with another perfect technique but rather with one that is approximately 95% sensitive for polyps 10 mm or larger.
An issue that needs to be addressed is whether the results in the studies included can be generalized to represent the current performance of CT colonography. Had we loosened our inclusion criteria, would the results have been different (e.g., inclusion of studies with slice thickness larger than 5 mm or a single scanning position)? Inclusion of such studies with suboptimal scanning techniques would have resulted in irrelevancy of this analysis for the future. We decided to restrict the review to studies of high quality so as to make our findings applicable to currently acceptable performance of CT colonography. We believe that our systematic quantitative review using the combination of the end points for statistical analysis was justified and that our results represent the current performance of CT colonography despite variances among studies and methods [30].
One of the steps in the systematic review of diagnostic tests is the pooling of results. We thus performed a statistical analysis to assess whether all the available studies could be pooled. If pooling was not possible, the poolable studies were analyzed and the outlier studies were excluded. Although the outliers might have had a one-sided distribution, they did not when the size of the studies was taken into account. The outliers did not appear to be systematically distributed about the results of the larger studies, which might have introduced bias.
We tested the strength of our results by performing analyses using three study-pooling algorithms: all individual studies with data combined irrespective of the sample size, all studies with data combined by weighting individual studies by the square root of the sample size, and all individual studies with data combined irrespective of sample size after removing outlier studies one at a time.
These multiple analyses were used because we found that sensitivity varied among studies, and it would have been misleading to try to estimate the true common sensitivity using only the first method of combining individual data irrespective of sample size. Fifty percent of the studies had small sample sizes and were overrepresented. Although some smaller studies were eliminated as outliers, larger studies were also eliminated as outliers, which then increased the weight given to the smaller studies even more. Therefore, we weighted the sample size using its square root [31].
Finally, simply discarding outliers as in the third method would have been an inadequate measure for treating the statistically significant variation in the outcome. Indeed, little variation in results appeared in most categories among the three pooling algorithms used, which further emphasizes the soundness of this metaanalysis. Although inherent heterogeneity exists among the studies, it did not seem to significantly affect the results among the three pooling algorithms. Upcoming technical advances will probably further improve the performance of CT colonography. Although a metaanalysis does not replace large prospective clinical trials, it has been shown in therapeutic trials not to differ significantly from those obtained in large trials [32].
The publications analyzed in our study included mostly high-risk patients. This higher prevalence affects and increases the positive predictive value. In a screening population the prevalence of disease is lower. The lessons learned from mammography will have to be implimented if and when CT colonography is used for screening [33]. The extrapolation of our results to screening populations needs to await larger-scale studies because the number of patients enrolled in each study evaluated was too small for us to reach conclusions regarding the general population. The importance of risk-stratification profiles and cost-benefit models should be studied in the future.
The per-polyp and per-patient sensitivity should be determined separately. Per-patient sensitivity is important primarily for clinical decision-making in triaging patients to conventional colonoscopy for a more complete evaluation and biopsy. Per-polyp sensitivity establishes our diagnostic skills because many patients have only one polyp and accurate diagnosis is needed for setting up criteria for accreditation and performance. The per-patient weighted sensitivities for polyps 6–9 mm and for polyps 10 mm or larger were similar (0.87 and 0.88, respectively), which is encouraging for metaanalyses involving the same studies. On a per-polyp analysis, however, this difference was more pronounced—0.63 and 0.81, respectively— and is statistically significant. The performance of CT colonography for small polyps is poor, but from a clinical perspective these are the least important ones.
Currently, one of the major drawbacks of mammography is its high level of false-positive results (i.e., low specificity) [33]. On the basis of our results, CT colonography seems to have an excellent specificity record (false-positive results of 5%) with pooled specificity for patients with polyps 10 mm or larger being 95%. We did not address specificity per polyp because of uncertainty regarding the denominator for this group.
Limitations to our study may include an inherited publication bias because studies that show a positive effect tend to be published more often than those that do not [34]. Several weakly positive studies may seem to add up to a strong positive result. A verification bias may have also occurred because in most studies, high-risk patient results were evaluated and verified using conventional colonoscopy [35]. Another inherent limitation is using size criteria for polyps that might not have been precisely measured, thus causing bias.
In conclusion, the collective evidence of published studies shows CT colonography to be an accurate tool for detecting clinically important colorectal polyps. The specificity and sensitivity of CT colonography are especially good for detecting polyps 10 mm or larger. The metaanalysis of the literature we conducted justifies large-scale studies to validate the results obtained.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  E. P. Whitlock, J. S. Lin, E. Liles, T. L. Beil, and R. Fu Screening for Colorectal Cancer: A Targeted, Updated Systematic Review for the U.S. Preventive Services Task Force Ann Intern Med, October 6, 2008; (2008) 0000605-200811040-00245. [Abstract] [Full Text]
|
|
|
|
 |
|
 J. Sosna, T. Sella, O. Sy, P. T. Lavin, R. Eliahou, S. Fraifeld, and E. Libson Critical Analysis of the Performance of Double-Contrast Barium Enema for Detecting Colorectal Polyps >= 6 mm in the Era of CT Colonography Am. J. Roentgenol., February 1, 2008; 190(2): 374 - 385. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. Burling, J. E East, and S. A Taylor Investigating rectal bleeding BMJ, December 15, 2007; 335(7632): 1260 - 1262. [Full Text] [PDF]
|
|
|
|
 |
|
 M. E. Baker, L. Bogoni, N. A. Obuchowski, C. Dass, R. M. Kendzierski, E. M. Remer, D. M. Einstein, P. Cathier, A. Jerebko, S. Lakare, et al. Computer-aided Detection of Colorectal Polyps: Can It Improve Sensitivity of Less-Experienced Readers? Preliminary Findings Radiology, October 1, 2007; 245(1): 140 - 149. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. Copel, J. Sosna, J. B. Kruskal, V. Raptopoulos, R. J. Farrell, and M. M. Morrin CT Colonography in 546 Patients with Incomplete Colonoscopy Radiology, August 1, 2007; 244(2): 471 - 478. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Halligan and D. G. Altman Evidence-based Practice in Radiology: Steps 3 and 4--Appraise and Apply Systematic Reviews and Meta-Analyses Radiology, April 1, 2007; 243(1): 13 - 27. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
European Society of Gastrointestinal and Abdominal Effect of Directed Training on Reader Performance for CT Colonography: Multicenter Study Radiology, January 1, 2007; 242(1): 152 - 161. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A SLATER Alteration in bowel habit Imaging, December 1, 2006; 18(4): 208 - 217. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. J Solomon, S. J Lord, and S. Walleser Review: computed tomographic colonography is accurate for medium and large colorectal polyps and cancer Evid. Based Med., October 1, 2006; 11(5): 153 - 153. [Full Text] [PDF]
|
|
|
|
|
|
R. L. MacCarty, C. D. Johnson, J. G. Fletcher, and L. A. Wilson Occult Colorectal Polyps on CT Colonography: Implications for Surveillance. Am. J. Roentgenol., May 1, 2006; 186(5): 1380 - 1383. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Sosna, A. Blachar, M. Amitai, E. Barmeir, N. Peled, S. N. Goldberg, and J. Bar-Ziv Colonic Perforation at CT Colonography: Assessment of Risk in a Multicenter Large Cohort Radiology, May 1, 2006; 239(2): 457 - 463. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S A Taylor, S Halligan, A Slater, D Burling, M Marshall, and C I Bartram Comparison of Radiologists' confidence in excluding significant colorectal neoplasia with multidetector-row CT colonography compared with double contrast barium enema. Br. J. Radiol., March 1, 2006; 79(939): 208 - 214. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. A. Taylor, S. Halligan, D. Burling, M. E. Roddie, L. Honeyfield, J. McQuillan, H. Amin, and J. Dehmeshki Computer-Assisted Reader Software Versus Expert Reviewers for Polyp Detection on CT Colonography. Am. J. Roentgenol., March 1, 2006; 186(3): 696 - 702. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P Veit, C Kuhle, T Beyer, H Kuehl, C U Herborn, G Borsch, H Stergar, J Barkhausen, A Bockisch, and G Antoch Whole body positron emission tomography/computed tomography (PET/CT) tumour staging with integrated PET/CT colonography: technical feasibility and first experiences in patients with colorectal cancer Gut, January 1, 2006; 55(1): 68 - 73. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Halligan, D. G. Altman, S. A. Taylor, S. Mallett, J. J. Deeks, C. I. Bartram, and W. Atkin CT Colonography in the Detection of Colorectal Polyps and Cancer: Systematic Review, Meta-Analysis, and Proposed Minimum Data Set for Study Level Reporting Radiology, December 1, 2005; 237(3): 893 - 904. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. G. Mang, C. Schaefer-Prokop, A. Maier, E. Schober, G. Lechner, and M. Prokop Detectability of Small and Flat Polyps in MDCT Colonography Using 2D and 3D Imaging Tools: Results from a Phantom Study Am. J. Roentgenol., December 1, 2005; 185(6): 1582 - 1589. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. P. Mulhall, G. R. Veerappan, and J. L. Jackson Meta-Analysis: Computed Tomographic Colonography Ann Intern Med, April 19, 2005; 142(8): 635 - 650. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L Bogoni, P Cathier, M Dundar, A Jerebko, S Lakare, J Liang, S Periaswamy, M E Baker, and M Macari Computer-aided detection (CAD) for CT colonography: a tool to address a growing need Br. J. Radiol., January 1, 2005; 78(suppl_1): S57 - S62. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. H. Dachman, P. Schumm, B. Heckel, H. Yoshida, and P. LaRiviere The Effect of Reconstruction Algorithm on Conspicuity of Polyps in CT Colonography Am. J. Roentgenol., November 1, 2004; 183(5): 1349 - 1353. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Cohnen, C. Vogt, A. Beck, K. Andersen, W. Heinen, S. vom Dahl, V. Aurich, D. Haeussinger, and U. Moedder Feasibility of MDCT Colonography in Ultra-Low-Dose Technique in the Detection of Colorectal Lesions: Comparison with High-Resolution Video Colonoscopy Am. J. Roentgenol., November 1, 2004; 183(5): 1355 - 1359. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. E. van Gelder, E. Birnie, J. Florie, M. P. Schutter, J. F. Bartelsman, P. Snel, J. S. Lameris, G. J. Bonsel, and J. Stoker CT Colonography and Colonoscopy: Assessment of Patient Preference in a 5-week Follow-up Study Radiology, November 1, 2004; 233(2): 328 - 337. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R Lewis, A Flynn, M E Dean, A Melville, A Eastwood, and A Booth Management of colorectal cancers Qual. Saf. Health Care, October 1, 2004; 13(5): 400 - 404. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. H. Dachman and M. E. Zalis Quality and Consistency in CT Colonography and Research Reporting Radiology, February 1, 2004; 230(2): 319 - 323. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||
