Accuracy of Capsule Colonoscopy to Detect Colorectal Polyps
Accuracy of Capsule Colonoscopy to Detect Colorectal Polyps
This report describes a capsule endoscopy study for colorectal polyp detection in asymptomatic persons.
Several aspects of the study design and the results of this trial are important to consider because they indicate that a complete understanding of capsule performance relative to colonoscopy is difficult to achieve. First, polyps were matched by colon segment whereas in previous trials of the second-generation capsule polyps were matched by the entire colon. When we performed a secondary analysis of our results with matching by the entire colon, the results for capsule sensitivity were numerically better and comparable with those of the initial smaller studies. Both colonoscopy and capsule are inferior for localization compared with CT colonography (CTC). Inaccurate localization by one or both tests in this study could have reduced the sensitivity of the capsule. Because a capsule finding of a polyp 6 mm or larger could be expected to generate a follow-up complete colonoscopy in clinical practice, whole-colon matching of polyp findings between the capsule and colonoscopy in clinical trials seems clinically relevant. Second, matching polyps by size, even with prespecified rules for polyp measurement and seemingly liberal rules for matching, is subject to error. In a preliminary study of the accuracy of colonoscopy at measuring polyps, we found that colonoscopists presented with the same photographs of polyps with aligned forceps reached variable conclusions about polyp size. Anecdotally, analysis of video recordings from this study suggested a variety of other errors occur in size measurement during colonoscopy, such as failure to expose the longest polyp dimension for photography, failure to push the forceps against the polyp, failure to orient the forceps along the longest axis of the polyp, and so forth. In addition, it might be expected that the capsule and colonoscopy would photograph the same polyp from different perspectives, with different degrees of luminal distention, and with different magnifications (capsule photographs are through water and colonoscopy photographs are through gas). Furthermore, the software used to measure polyp size during the capsule studies had a 40% plus-minus error range when tested on balls of known size in a model (Given Imaging, Inc, unpublished data), but there are no published studies of the accuracy of this software. With regard to these issues in size matching, we used a size matching rule in this study (polyps matched if the 50% plus range of the smaller size measurement was within the 50% minus range of the larger size measurement, and either the colonoscopy size or the capsule size measurement could serve as the larger measure) that was more liberal (allowed more frequent polyp matching) than that used in previous studies. Thus, in this study we used a more liberal size matching rule and a stricter location matching rule compared with previous studies. Any set of matching rules for polyps detected by the capsule and colonoscopy might operate to increase or decrease the calculated sensitivity of capsule incorrectly. Third, colonoscopy is known to be an imperfect gold standard. In this study, the average performance of colonoscopy was excellent (overall ADR, 39%), but 3 colonoscopists had ADRs that were below the recommended thresholds, and in some instances a polyp clearly visible by capsule was not verified by colonoscopy. In these cases a polyp that should be a true positive for the capsule was counted as false. Finally, any imaging test measured against colonoscopy will have its apparent performance affected by the quality of the colonoscopy. For example, the best results ever reported for the performance of CTC came from a center where the combined ADR of the colonoscopists during the study was only 16.5%. In that case, the sensitivity of CTC likely was improved by the relatively poor performance of the gold standard colonoscopies. In previous studies of the capsule, the average ADR of the colonoscopies was not reported. The average ADR in this study was among the highest ever reported for a mixed-gender screening population. We suggest that ADRs be reported in future studies of noncolonoscopic imaging because this information allows assessment of the colonoscopy quality as the gold standard in the study, and may allow improved understanding of how varying results for performance of the test modality occur in different studies.
Lesions in the serrated class were not detected well by the capsule in this study compared with conventional adenomas. In previous studies of the capsule, detection of serrated lesions was not reported. Sessile serrated polyps are commonly flat or sessile, pale in color, and subtle in appearance when viewed by colonoscopy. Initial studies suggest colonoscopists have more variability in the detection of sessile serrated polyps compared with conventional adenomas. CTC also has difficulty detecting these lesions, and a recent trial found that fecal DNA testing had some sensitivity for serrated lesions whereas fecal immunochemical testing had no sensitivity. Our results indicate that additional work is needed to understand the appearance of serrated lesions at capsule and to improve their detection.
The sensitivity of CTC for patients with adenomas 6 mm or larger in the National CT Colonography trial was 78%, which is comparable with the 88% sensitivity found in the current trial. Given that both CTC and the capsule are intended for detection of lesions 6 mm or larger and not diminutive lesions, and that CTC is recommended at 5-year intervals when negative, our results suggest that the capsule also might be performed at 5-year intervals when used for screening.
The specificities achieved in the current study were higher than 2 previous studies of the second-generation colon capsule. This finding may reflect the more liberal size matching rule used in the current study, the training of the central readers, or the unblinding after colonoscopy, which converted some cases from false to true positives.
The most significant limitation of this trial was that we excluded 77 patients for the combination of short transit time and poor preparation and this exclusion was not planned before the study. The short transit times likely were related to use of oral sulfate solution as a boost. Outside the United States, sodium phosphate frequently is used as the boost, but because of the occurrence of rare cases of acute irreversible nephropathy after sodium phosphate, the continued use of this agent was considered inappropriate in the United States. Exclusion of these patients is consistent with the anticipated labeling of the capsule by the US Food and Drug Administration, which indicate that patients with a combination of poor preparation and a colonic transit time of fewer than 40 minutes had a technically inadequate study for polyp detection. We did describe the capsule performance in the 77 excluded patients.
In summary, we found that the capsule performed well for detecting asymptomatic patients with any polyp and with any conventional adenoma 6 mm or larger in size. In considering the capsule as a colorectal cancer screening test, it is appropriate to consider that the bowel preparation for the capsule is more extensive than that required for colonoscopy, that technical failures (short transit time plus poor preparation) occurred in 9% of patients, and that the logistics of performing same-day colonoscopy in patients with a positive capsule study are difficult. Given these considerations and the results of this study, we conclude that colonoscopy remains the gold standard for the detection of colorectal polyps. The capsule is a good test for the detection of patients with conventional adenomas 6 mm or larger in size and appears to be an appropriate imaging choice for patients who cannot undergo colonoscopy or had incomplete colonoscopy. Additional study of the capsule, including efforts to improve the detection of serrated lesions, appears warranted.
Discussion
This report describes a capsule endoscopy study for colorectal polyp detection in asymptomatic persons.
Several aspects of the study design and the results of this trial are important to consider because they indicate that a complete understanding of capsule performance relative to colonoscopy is difficult to achieve. First, polyps were matched by colon segment whereas in previous trials of the second-generation capsule polyps were matched by the entire colon. When we performed a secondary analysis of our results with matching by the entire colon, the results for capsule sensitivity were numerically better and comparable with those of the initial smaller studies. Both colonoscopy and capsule are inferior for localization compared with CT colonography (CTC). Inaccurate localization by one or both tests in this study could have reduced the sensitivity of the capsule. Because a capsule finding of a polyp 6 mm or larger could be expected to generate a follow-up complete colonoscopy in clinical practice, whole-colon matching of polyp findings between the capsule and colonoscopy in clinical trials seems clinically relevant. Second, matching polyps by size, even with prespecified rules for polyp measurement and seemingly liberal rules for matching, is subject to error. In a preliminary study of the accuracy of colonoscopy at measuring polyps, we found that colonoscopists presented with the same photographs of polyps with aligned forceps reached variable conclusions about polyp size. Anecdotally, analysis of video recordings from this study suggested a variety of other errors occur in size measurement during colonoscopy, such as failure to expose the longest polyp dimension for photography, failure to push the forceps against the polyp, failure to orient the forceps along the longest axis of the polyp, and so forth. In addition, it might be expected that the capsule and colonoscopy would photograph the same polyp from different perspectives, with different degrees of luminal distention, and with different magnifications (capsule photographs are through water and colonoscopy photographs are through gas). Furthermore, the software used to measure polyp size during the capsule studies had a 40% plus-minus error range when tested on balls of known size in a model (Given Imaging, Inc, unpublished data), but there are no published studies of the accuracy of this software. With regard to these issues in size matching, we used a size matching rule in this study (polyps matched if the 50% plus range of the smaller size measurement was within the 50% minus range of the larger size measurement, and either the colonoscopy size or the capsule size measurement could serve as the larger measure) that was more liberal (allowed more frequent polyp matching) than that used in previous studies. Thus, in this study we used a more liberal size matching rule and a stricter location matching rule compared with previous studies. Any set of matching rules for polyps detected by the capsule and colonoscopy might operate to increase or decrease the calculated sensitivity of capsule incorrectly. Third, colonoscopy is known to be an imperfect gold standard. In this study, the average performance of colonoscopy was excellent (overall ADR, 39%), but 3 colonoscopists had ADRs that were below the recommended thresholds, and in some instances a polyp clearly visible by capsule was not verified by colonoscopy. In these cases a polyp that should be a true positive for the capsule was counted as false. Finally, any imaging test measured against colonoscopy will have its apparent performance affected by the quality of the colonoscopy. For example, the best results ever reported for the performance of CTC came from a center where the combined ADR of the colonoscopists during the study was only 16.5%. In that case, the sensitivity of CTC likely was improved by the relatively poor performance of the gold standard colonoscopies. In previous studies of the capsule, the average ADR of the colonoscopies was not reported. The average ADR in this study was among the highest ever reported for a mixed-gender screening population. We suggest that ADRs be reported in future studies of noncolonoscopic imaging because this information allows assessment of the colonoscopy quality as the gold standard in the study, and may allow improved understanding of how varying results for performance of the test modality occur in different studies.
Lesions in the serrated class were not detected well by the capsule in this study compared with conventional adenomas. In previous studies of the capsule, detection of serrated lesions was not reported. Sessile serrated polyps are commonly flat or sessile, pale in color, and subtle in appearance when viewed by colonoscopy. Initial studies suggest colonoscopists have more variability in the detection of sessile serrated polyps compared with conventional adenomas. CTC also has difficulty detecting these lesions, and a recent trial found that fecal DNA testing had some sensitivity for serrated lesions whereas fecal immunochemical testing had no sensitivity. Our results indicate that additional work is needed to understand the appearance of serrated lesions at capsule and to improve their detection.
The sensitivity of CTC for patients with adenomas 6 mm or larger in the National CT Colonography trial was 78%, which is comparable with the 88% sensitivity found in the current trial. Given that both CTC and the capsule are intended for detection of lesions 6 mm or larger and not diminutive lesions, and that CTC is recommended at 5-year intervals when negative, our results suggest that the capsule also might be performed at 5-year intervals when used for screening.
The specificities achieved in the current study were higher than 2 previous studies of the second-generation colon capsule. This finding may reflect the more liberal size matching rule used in the current study, the training of the central readers, or the unblinding after colonoscopy, which converted some cases from false to true positives.
The most significant limitation of this trial was that we excluded 77 patients for the combination of short transit time and poor preparation and this exclusion was not planned before the study. The short transit times likely were related to use of oral sulfate solution as a boost. Outside the United States, sodium phosphate frequently is used as the boost, but because of the occurrence of rare cases of acute irreversible nephropathy after sodium phosphate, the continued use of this agent was considered inappropriate in the United States. Exclusion of these patients is consistent with the anticipated labeling of the capsule by the US Food and Drug Administration, which indicate that patients with a combination of poor preparation and a colonic transit time of fewer than 40 minutes had a technically inadequate study for polyp detection. We did describe the capsule performance in the 77 excluded patients.
In summary, we found that the capsule performed well for detecting asymptomatic patients with any polyp and with any conventional adenoma 6 mm or larger in size. In considering the capsule as a colorectal cancer screening test, it is appropriate to consider that the bowel preparation for the capsule is more extensive than that required for colonoscopy, that technical failures (short transit time plus poor preparation) occurred in 9% of patients, and that the logistics of performing same-day colonoscopy in patients with a positive capsule study are difficult. Given these considerations and the results of this study, we conclude that colonoscopy remains the gold standard for the detection of colorectal polyps. The capsule is a good test for the detection of patients with conventional adenomas 6 mm or larger in size and appears to be an appropriate imaging choice for patients who cannot undergo colonoscopy or had incomplete colonoscopy. Additional study of the capsule, including efforts to improve the detection of serrated lesions, appears warranted.
Source...