Accuracy of Capsule Colonoscopy to Detect Colorectal Polyps
Accuracy of Capsule Colonoscopy to Detect Colorectal Polyps
Figure 1 diagrams the initial enrollment cohort to the final analysis cohort. Of 884 asymptomatic subjects enrolled, 695 were included in the analysis of all polyps. Six patients did not have a pathology report, therefore 689 were included in the adenoma analysis. The mean age of the 689 subjects was 57 years, and 56% were female.
(Enlarge Image)
Figure 1.
Flow chart diagramming the initial enrollment cohort to the final analysis cohort.
There were 77 subjects (9%) excluded for both inadequate cleansing and whole-colon transit time fewer than 40 minutes. Forty-five subjects dropped out before the capsule procedure, 15 subjects dropped out after the capsule procedure and before the colonoscopy, 15 subjects were dropped because a site was terminated early for repeated protocol noncompliance, in 12 subjects the capsule did not reach the colon by 12 hours after ingestion, 10 subjects had protocol violations during the colonoscopy, in 8 subjects the capsule did not leave the cecum, 5 subjects had moderate adverse events before the capsule (related to bowel preparation) and did not ingest the capsule, and 2 subjects could not ingest the capsule.
Bowel preparation was adequate (excellent or good) for the entire colon in 80% (95% CI, 76%–83%) of capsule procedures. The cecal cleansing level was adequate in 67% (95% CI, 63%–70%), ascending colon cleansing level was adequate in 74% (95% CI, 71%–78%), transverse colon cleansing level was adequate in 77% (95% CI, 73%–80%), descending-sigmoid colon cleansing level was adequate in 82% (95% CI, 79%–85%), and rectum cleansing level was adequate in 65% (95% CI, 61%–69%). Among 689 subjects evaluated for conventional adenoma performance, 522 (76%; 95% CI, 72%–79%) excreted the capsule within 6 hours after ingestion, 607 (88%; 95% CI, 85%–90%) excreted the capsule within 8 hours, 626 (91%; 95% CI, 88%–923%) within 10 hours, and 636 (92%; 95% CI, 90%–94%) excreted the capsule within 12 hours. The median transit time of the capsule through the colon before and after all exclusions was 1 hour 16 minutes and 1 hour 39 minutes, respectively (Supplementary Table 1). The total transit time through the colon was shorter than transit times in previous trials of capsule colonoscopy (Supplementary Table 1).
Colonoscopy was complete to the cecum in 99.6% (95% CI, 99%–100%) of subjects. Bowel preparation was adequate (excellent or good) in 93% of subjects (95% CI, 91%–95%). The cecal cleansing level was adequate in 88% (95% CI, 85%–90%), ascending colon cleansing level was adequate in 91% (95% CI, 89%–93%), transverse colon cleansing level was adequate in 92% (95% CI, 90%–94%), descending-sigmoid colon cleansing level was adequate in 92% (95% CI, 90%–94%), and rectum cleansing level was adequate in 93% (95% CI, 91%–95%). The mean withdrawal time in subjects with no lesions detected was 10.6 minutes (SD, 4.2 min; range, 5–32 min). The collective adenoma detection rate (ADR) of colonoscopists for the study was 39%. Among 17 colonoscopists who performed 10 or more colonoscopies, the range of ADRs was 8%–63%, and 3 colonoscopists had ADRs below the recommended 20% threshold. The size of all polyps and the histology (when available) detected at colonoscopy is shown in Table 1. Of 340 lesions detected and retrieved for histology by colonoscopy that were 6 mm or larger in size (excluding the 4 cancers), 211 (62%) were conventional adenomas and 129 (38%) were in the serrated class.
Capsule identified 392 lesions 6 mm or larger and 111 lesions 10 mm or larger in size. Colonoscopy detected 368 lesions 6 mm or larger and 114 lesions 10 mm or larger (Table 2).
Capsule detection of patients with polyps and with conventional adenomas 6 mm or larger and 10 mm or larger is shown in Table 3. The capsule sensitivities for detecting subjects with any polyp 6 mm or larger and 10 mm or larger were 81% (95% CI, 77%–84%) and 80% (95% CI, 74%–86%), respectively (per-segment matching), and the specificities were 93% (95% CI, 91%–95%) and 97% (95% CI, 96%–98%), respectively.
For the analysis of capsule detection of patients with conventional adenomas, lesions in the serrated class were ignored, except that a lesion detected by capsule that proved to not be a conventional adenoma was considered a false-positive capsule detection. Capsule sensitivities for detecting subjects with conventional adenomas 6 mm or larger and 10 mm or larger were 88% (95% CI, 82%–93%) and 92% (95% CI, 82%–97%), respectively, and the specificities were 82% (95% CI, 80%–83%) and 95% (95% CI, 94%–95%), respectively (Table 3).
Figure 2 summarizes the capsule accuracy for detecting patients with at least 1 polyp as well as 1 conventional adenoma 6 mm or larger and 10 mm or larger in size with matching lesions by segment analysis.
(Enlarge Image)
Figure 2.
Summary of capsule accuracy for detecting patients with at least 1 polyp (blue) and at least 1 conventional adenoma (red) 6 mm or larger and 10 mm or larger in size with matching lesions by segment. The error bars indicate 95% CIs. The specificity for adenomas 6 mm or larger is lower than the specificity for adenomas 10 mm or larger, probably because nonadenomatous polyps detected by the capsule were considered as capsule false positives for the adenoma analysis, and a higher proportion of 10 mm or larger polyps are adenomas.
Supplementary Table 2 shows the capsule accuracy in the right vs left colon for patients with polyps and conventional adenomas 6 mm or larger and 10 mm or larger in size. The sensitivity of the capsule in the right colon for patients with any polyp 6 mm or larger was 72% (95% CI, 65%–79%), which was less than the left colon sensitivity of 88% (95% CI, 82%–93%). There were no other significant differences in right vs left capsule performance.
There were no significant differences in the sensitivity or specificity between capsule readers except for one reader who had lower specificity for lesions 6 mm or larger (Supplementary Table 3).
Table 4 provides additional detail regarding per-patient sensitivity of the capsule by histologic class. The sensitivity for conventional adenomas first was calculated without reference to lesions in the serrated class (Table 4, top row). Next, the sensitivity of the capsule for detecting patients with a sessile serrated polyp 6 mm or larger and 10 mm or larger, but no conventional adenoma 6 mm or larger, was calculated without reference to patients with hyperplastic polyps (Table 4, row 2). Finally, patients with hyperplastic polyps 6 mm or larger but no conventional adenomas or sessile serrated polyps 6 mm or larger were considered (Table 4, row 3). Sessile serrated polyps accounted for 26% of false-negative capsule detections of patients with polyps 6 mm or larger, and hyperplastic polyps accounted for 37% of this group.
In a separate analysis we calculated the capsule performance for detecting patients with polyps and conventional adenomas 6 mm or larger and 10 mm or larger in size without requiring that polyps be matched by segment (Table 3). Sensitivity was numerically, but not significantly, higher with entire-colon compared with per-segment matching.
There were 52 cases in which a lesion 6 mm or larger was detected by capsule and not by conventional colonoscopy, including 24 cases in which a lesion 10 mm or larger was detected by capsule. Among the 52 cases, repeat colonoscopy identified a matching lesion in 22 cases, including 7 of 24 cases with a polyp 10 mm or larger detected by the capsule and not by conventional colonoscopy. These cases were designated true-positive capsule detections. Figure 3 shows examples of 6 mm or larger lesions shown by the capsule that were not verified by colonoscopy and were designated false-positive capsule detections.
(Enlarge Image)
Figure 3.
(A–D) Capsule photographs of polyps that were not verified by either the blinded or unblinded colonoscopy.
Conventional colonoscopy detected 4 cancers in 4 patients. The cancers ranged in size from 10 to 80 mm. The capsule identified 3 cancers for a per-lesion sensitivity of 75%. The cancer that was missed was a 10-mm sessile lesion in the sigmoid colon. Unblinded review of the capsule video determined that the cancer was photographed by the capsule in multiple frames. The same patient had 2 cecal lesions (an 11-mm tubular adenoma and a 7-mm hyperplastic polyp) identified by the capsule. Therefore, the per-patient sensitivity of capsule for the 4 cancers was 100%.
In the 77 patients excluded for both inadequate preparation and a colon transit time of fewer than 40 minutes, capsule sensitivities for detecting patients with any polyp 6 mm or larger and 10 mm or larger were 68% and 71%, respectively (polyps matched by segment). The capsule sensitivities for detecting patients with any conventional adenoma 6 mm or larger and 10 mm or larger were 70% and 73%, respectively.
There were no serious adverse events associated with the capsule. One patient was hospitalized for 1 day after colonoscopy with abdominal pain.
There were 142 nonserious adverse events in 101 patients (11% of the 884 enrolled patients), with 128 related to bowel preparation including vomiting, nausea, headache, dizziness, gagging, diarrhea, hemorrhoidal bleeding, bloating, rash, abdominal pain, syncope, tinnitus, chest burning, chills, and flu-like symptoms, which resolved in all cases within 1 month and in 92% within 1 day. Eleven (8%) of the 142 nonserious adverse events were related to colonoscopy and included fever, headache, abdominal pain, bloating, and rectal bleeding. All resolved within 8 days. Three nonserious adverse events were related to the capsule procedure included gagging, vomiting, and abdominal cramping, and each resolved on the same day.
Results
Exclusions
Figure 1 diagrams the initial enrollment cohort to the final analysis cohort. Of 884 asymptomatic subjects enrolled, 695 were included in the analysis of all polyps. Six patients did not have a pathology report, therefore 689 were included in the adenoma analysis. The mean age of the 689 subjects was 57 years, and 56% were female.
(Enlarge Image)
Figure 1.
Flow chart diagramming the initial enrollment cohort to the final analysis cohort.
There were 77 subjects (9%) excluded for both inadequate cleansing and whole-colon transit time fewer than 40 minutes. Forty-five subjects dropped out before the capsule procedure, 15 subjects dropped out after the capsule procedure and before the colonoscopy, 15 subjects were dropped because a site was terminated early for repeated protocol noncompliance, in 12 subjects the capsule did not reach the colon by 12 hours after ingestion, 10 subjects had protocol violations during the colonoscopy, in 8 subjects the capsule did not leave the cecum, 5 subjects had moderate adverse events before the capsule (related to bowel preparation) and did not ingest the capsule, and 2 subjects could not ingest the capsule.
Procedures
Bowel preparation was adequate (excellent or good) for the entire colon in 80% (95% CI, 76%–83%) of capsule procedures. The cecal cleansing level was adequate in 67% (95% CI, 63%–70%), ascending colon cleansing level was adequate in 74% (95% CI, 71%–78%), transverse colon cleansing level was adequate in 77% (95% CI, 73%–80%), descending-sigmoid colon cleansing level was adequate in 82% (95% CI, 79%–85%), and rectum cleansing level was adequate in 65% (95% CI, 61%–69%). Among 689 subjects evaluated for conventional adenoma performance, 522 (76%; 95% CI, 72%–79%) excreted the capsule within 6 hours after ingestion, 607 (88%; 95% CI, 85%–90%) excreted the capsule within 8 hours, 626 (91%; 95% CI, 88%–923%) within 10 hours, and 636 (92%; 95% CI, 90%–94%) excreted the capsule within 12 hours. The median transit time of the capsule through the colon before and after all exclusions was 1 hour 16 minutes and 1 hour 39 minutes, respectively (Supplementary Table 1). The total transit time through the colon was shorter than transit times in previous trials of capsule colonoscopy (Supplementary Table 1).
Colonoscopy was complete to the cecum in 99.6% (95% CI, 99%–100%) of subjects. Bowel preparation was adequate (excellent or good) in 93% of subjects (95% CI, 91%–95%). The cecal cleansing level was adequate in 88% (95% CI, 85%–90%), ascending colon cleansing level was adequate in 91% (95% CI, 89%–93%), transverse colon cleansing level was adequate in 92% (95% CI, 90%–94%), descending-sigmoid colon cleansing level was adequate in 92% (95% CI, 90%–94%), and rectum cleansing level was adequate in 93% (95% CI, 91%–95%). The mean withdrawal time in subjects with no lesions detected was 10.6 minutes (SD, 4.2 min; range, 5–32 min). The collective adenoma detection rate (ADR) of colonoscopists for the study was 39%. Among 17 colonoscopists who performed 10 or more colonoscopies, the range of ADRs was 8%–63%, and 3 colonoscopists had ADRs below the recommended 20% threshold. The size of all polyps and the histology (when available) detected at colonoscopy is shown in Table 1. Of 340 lesions detected and retrieved for histology by colonoscopy that were 6 mm or larger in size (excluding the 4 cancers), 211 (62%) were conventional adenomas and 129 (38%) were in the serrated class.
Capsule Performance
Capsule identified 392 lesions 6 mm or larger and 111 lesions 10 mm or larger in size. Colonoscopy detected 368 lesions 6 mm or larger and 114 lesions 10 mm or larger (Table 2).
Capsule detection of patients with polyps and with conventional adenomas 6 mm or larger and 10 mm or larger is shown in Table 3. The capsule sensitivities for detecting subjects with any polyp 6 mm or larger and 10 mm or larger were 81% (95% CI, 77%–84%) and 80% (95% CI, 74%–86%), respectively (per-segment matching), and the specificities were 93% (95% CI, 91%–95%) and 97% (95% CI, 96%–98%), respectively.
For the analysis of capsule detection of patients with conventional adenomas, lesions in the serrated class were ignored, except that a lesion detected by capsule that proved to not be a conventional adenoma was considered a false-positive capsule detection. Capsule sensitivities for detecting subjects with conventional adenomas 6 mm or larger and 10 mm or larger were 88% (95% CI, 82%–93%) and 92% (95% CI, 82%–97%), respectively, and the specificities were 82% (95% CI, 80%–83%) and 95% (95% CI, 94%–95%), respectively (Table 3).
Figure 2 summarizes the capsule accuracy for detecting patients with at least 1 polyp as well as 1 conventional adenoma 6 mm or larger and 10 mm or larger in size with matching lesions by segment analysis.
(Enlarge Image)
Figure 2.
Summary of capsule accuracy for detecting patients with at least 1 polyp (blue) and at least 1 conventional adenoma (red) 6 mm or larger and 10 mm or larger in size with matching lesions by segment. The error bars indicate 95% CIs. The specificity for adenomas 6 mm or larger is lower than the specificity for adenomas 10 mm or larger, probably because nonadenomatous polyps detected by the capsule were considered as capsule false positives for the adenoma analysis, and a higher proportion of 10 mm or larger polyps are adenomas.
Supplementary Table 2 shows the capsule accuracy in the right vs left colon for patients with polyps and conventional adenomas 6 mm or larger and 10 mm or larger in size. The sensitivity of the capsule in the right colon for patients with any polyp 6 mm or larger was 72% (95% CI, 65%–79%), which was less than the left colon sensitivity of 88% (95% CI, 82%–93%). There were no other significant differences in right vs left capsule performance.
There were no significant differences in the sensitivity or specificity between capsule readers except for one reader who had lower specificity for lesions 6 mm or larger (Supplementary Table 3).
Detection of Lesions by Histologic Class
Table 4 provides additional detail regarding per-patient sensitivity of the capsule by histologic class. The sensitivity for conventional adenomas first was calculated without reference to lesions in the serrated class (Table 4, top row). Next, the sensitivity of the capsule for detecting patients with a sessile serrated polyp 6 mm or larger and 10 mm or larger, but no conventional adenoma 6 mm or larger, was calculated without reference to patients with hyperplastic polyps (Table 4, row 2). Finally, patients with hyperplastic polyps 6 mm or larger but no conventional adenomas or sessile serrated polyps 6 mm or larger were considered (Table 4, row 3). Sessile serrated polyps accounted for 26% of false-negative capsule detections of patients with polyps 6 mm or larger, and hyperplastic polyps accounted for 37% of this group.
Detection With Entire Colon Matching
In a separate analysis we calculated the capsule performance for detecting patients with polyps and conventional adenomas 6 mm or larger and 10 mm or larger in size without requiring that polyps be matched by segment (Table 3). Sensitivity was numerically, but not significantly, higher with entire-colon compared with per-segment matching.
Contribution of Unblinding
There were 52 cases in which a lesion 6 mm or larger was detected by capsule and not by conventional colonoscopy, including 24 cases in which a lesion 10 mm or larger was detected by capsule. Among the 52 cases, repeat colonoscopy identified a matching lesion in 22 cases, including 7 of 24 cases with a polyp 10 mm or larger detected by the capsule and not by conventional colonoscopy. These cases were designated true-positive capsule detections. Figure 3 shows examples of 6 mm or larger lesions shown by the capsule that were not verified by colonoscopy and were designated false-positive capsule detections.
(Enlarge Image)
Figure 3.
(A–D) Capsule photographs of polyps that were not verified by either the blinded or unblinded colonoscopy.
Cancers
Conventional colonoscopy detected 4 cancers in 4 patients. The cancers ranged in size from 10 to 80 mm. The capsule identified 3 cancers for a per-lesion sensitivity of 75%. The cancer that was missed was a 10-mm sessile lesion in the sigmoid colon. Unblinded review of the capsule video determined that the cancer was photographed by the capsule in multiple frames. The same patient had 2 cecal lesions (an 11-mm tubular adenoma and a 7-mm hyperplastic polyp) identified by the capsule. Therefore, the per-patient sensitivity of capsule for the 4 cancers was 100%.
Performance in Excluded Patients
In the 77 patients excluded for both inadequate preparation and a colon transit time of fewer than 40 minutes, capsule sensitivities for detecting patients with any polyp 6 mm or larger and 10 mm or larger were 68% and 71%, respectively (polyps matched by segment). The capsule sensitivities for detecting patients with any conventional adenoma 6 mm or larger and 10 mm or larger were 70% and 73%, respectively.
Adverse Events
There were no serious adverse events associated with the capsule. One patient was hospitalized for 1 day after colonoscopy with abdominal pain.
There were 142 nonserious adverse events in 101 patients (11% of the 884 enrolled patients), with 128 related to bowel preparation including vomiting, nausea, headache, dizziness, gagging, diarrhea, hemorrhoidal bleeding, bloating, rash, abdominal pain, syncope, tinnitus, chest burning, chills, and flu-like symptoms, which resolved in all cases within 1 month and in 92% within 1 day. Eleven (8%) of the 142 nonserious adverse events were related to colonoscopy and included fever, headache, abdominal pain, bloating, and rectal bleeding. All resolved within 8 days. Three nonserious adverse events were related to the capsule procedure included gagging, vomiting, and abdominal cramping, and each resolved on the same day.