Dig Dis Sci (2014) 59:2503–2507 DOI 10.1007/s10620-014-3327-8
ORIGINAL ARTICLE
A Rapid and Accurate Method to Detect Active Small Bowel Gastrointestinal Bleeding on Video Capsule Endoscopy Adam C. Stein • Anoop Appannagari Ibrahim Habib • Carol E. Semrad • David T. Rubin
•
Received: 22 January 2014 / Accepted: 7 August 2014 / Published online: 22 August 2014 Ó Springer Science+Business Media New York 2014
Abstract Background Video capsule endoscopy (VCE) is indicated to evaluate for suspected small bowel bleeding, but ‘‘standard view’’ (SV) evaluation is time-consuming. Rapid Reader 6.0 software (Given Imaging, Duluth GA) contains two computer algorithmic systems: (1) ‘‘Quickview’’ (QV) which automatically skips similar images and (2) a pixel analysis program that identifies suspected blood (SBI). Combining the two modalities (QV ? SBI) may provide a faster modality to assess for active small bowel bleeding. Aims This study was designed to assess the accuracy of QV ? SBI for small bowel bleeding compared to SV findings. Methods This is a retrospective, case–control study at a single tertiary care referral hospital including all patients with VCE performed for suspected small bowel bleeding from 4/2007 to 3/2011. All studies were previously read using SV by one of two experienced faculty (CS, DR). The primary outcome was diagnostic accuracy of QV ? SBI in assessing for active small bowel bleeding compared to SV.
Adam C. Stein and Anoop Appannagari are the co-first authors. A. C. Stein C. E. Semrad D. T. Rubin (&) Section of Gastroenterology, Hepatology, and Nutrition, Department of Medicine, University of Chicago Medicine, 5841 S. Maryland Ave., MC 4076, Chicago, IL 60637, USA e-mail: [email protected] A. Appannagari Division of Gastroenterology, Department of Medicine, Loyola University, Maywood, IL, USA I. Habib Department of Medicine, Advocate Lutheran General Hospital, Park Ridge, IL, USA
Results A total of 116 VCE were included, 28 with active small bowel bleeding identified by original SV. Using QV ? SBI, all 28 VCEs with active small bowel bleeding were identified. The sensitivity of QV ? SBI to detect active bleeding was 100 %, while the specificity was 93–94 %. The mean time to identify landmarks and read the entire study was 3 min 20 s. Conclusions The QV ? SBI reading format of VCE is an efficient, highly sensitive modality to assess for potential small bowel bleeding. Keywords Video capsule endoscopy Small bowel bleeding Quickview Suspected blood indicator Given
Introduction Video capsule endoscopy (VCE) is a well-established, noninvasive modality to identify obscure gastrointestinal bleeding (GIB) [1, 2]. VCE provides excellent diagnostic yield for the evaluation obscure GI bleeding. Despite this accuracy, a limitation to VCE has been the amount of time required to read each study. While the exam itself usually requires minimal effort by the physician, the time required for the exam itself, the amount of time required to read each study can be burdensome. The standard method of reading a VCE study involves watching streaming video of approximately 50,000 images and requires a significant time commitment, with prior studies reporting average or median reading times of 17–60 min [3–5]. For indications that may require urgent intervention based on results of the VCE, such as evaluation for GIB, prioritizing time to read a study in a timely manner may add a significant burden to a busy gastroenterologist. To improve the efficiency of reading VCE
123
2504
studies, several methods have been proposed including software that filters the number of images viewed and utilizing physician extenders (for example, nurses) to preliminarily review the study [4, 6–9]. The Rapid Reader 6.0 software program (Given imaging, Duluth, GA) contains two software algorithmic systems that allow for more rapid analysis of VCE studies. The ‘‘Quickview’’ (QV) system is a program that reduces the image quantity by analyzing specific patterns and colors to provide a shorter composite video. Theoretically, QV allows for a rapid and accurate overview of the most significant lesions detected by VCE. An additional algorithm contained in Rapid Reader 6.0, the suspected blood indicator (SBI), analyzes pixels to look for a red color that may be consistent with active or potential bleeding. There have been limited studies on the diagnostic efficacy of QV, and most results show that there may be an unacceptably high increase in diagnostic miss rate [5, 6, 8, 9]. The published data regarding SBI is also limited, with one study showing an improved sensitivity and negative predictive value (83, 97 %) for active bleeding but lower specificity and positive predictive value (66, 23 %) [10]. No studies have assessed combining QV and SBI for obscure gastrointestinal bleeding. We performed a study to assess the accuracy of QV ? SBI compared to standard view (SV) in patients with suspected small bowel bleeding. We then assessed the amount of time required to analyze each exam in efforts to develop an algorithm for suspected GIB.
Materials and Methods We retrospectively identified all VCE’s performed to assess for obscure gastrointestinal bleeding, melena of unknown origin, and hematochezia of unknown origin at the University of Chicago Medicine from April 1, 2007 to March 31, 2011. The study was approved by the Investigational Review Board at the University of Chicago. Patients We identified all VCE’s ordered for the following indications: obscure GI bleed, melena of unknown origin, and hematochezia of unknown origin. Studies were excluded if the capsule did not reach the small bowel, if there were technical difficulties leading to inability to adequately visualize the small bowel mucosa, or if the study was unavailable at time of review. If multiple VCE’s were identified for the same patient, all studies were included. All identified VCE’s were previously read in SV by two experienced faculty members (CS, DR), whose reading served as the reference standard for this study. The experienced faculty members read each VCE using quad view
123
Dig Dis Sci (2014) 59:2503–2507
as the standard setting, toggling to dual view when needed. Each study was read at variable speed rates. Each VCE was read unblinded using SV to verify active bleeding and to assess whether areas of active bleeding would be identified using SBI alone as well as QV ? SBI. Patient charts of all studies included for analysis were reviewed to determine whether further investigation or intervention was recommended and performed based on SV VCE findings. Blinded Review Prior to the beginning of the study, two novice readers (AA, AS) were trained in VCE interpretation by reading 25 studies in SV under supervision by advanced readers. A recent study validated the learning curve for becoming adequately trained in VCE and found that with feedbackbased education, 20 studies would reduce the possibility of missing lesions [11]. All included VCEs were independently assessed for active small bowel bleeding using only the QV ? SBI mode by the two novice readers blinded to the original faculty member’s read. Reads were performed without restriction, including the ability to pause and rewind the VCE as desired. Readers were able to toggle between single view, dual view, and quad view, with the standard setting of dual view. Data recorded during the QV ? SBI read included assessment for active small bowel GIB, if the area of active small bowel GIB had a corresponding SBI, the presence of SBI without active small bowel GIB, and the total time required to complete the necessary data collection for each VCE. The readers did not attempt to identify lesions that may have represented a potential bleeding source. The two individual reads using QB ? SBI were compared independently to the original SV reads to assess for accuracy of identifying active small bowel GIB. Statistical Analysis Standard view findings were compared to each blinded reader under SBI alone as well as QV ? SBI for sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). Comparison of SBI alone with QV ? SBI for each blinded reader was done by chisquared analysis. Interobserver agreement was assessed between the two independent QV ? SBI novice reads.
Results A total of 131 individual VCEs were identified during the study period. Of these, 15 were excluded (eight were unavailable for review, six did not reach the small bowel,
Dig Dis Sci (2014) 59:2503–2507
2505
Table 1 Results of review with standard view and QV ? SBI
Active bleeding Standard view (gold standard)
a
All active bleeding seen on standard view was found on QV ? SBI by both reader 1 and reader 2
No bleeding
28
88
QV ? SBI, reader 1
33a
83
QV ? SBI, reader 2
34a
82
and one was of poor technical quality limiting adequate small bowel visualization). A total of 116 VCEs were included for analysis from 98 patients, with nine patients having two VCEs included. Of the nine patients with multiple VCE’s ,the median time between VCEs was 32 days with a range of 2–916 days. The indications for the VCEs which found active bleeding include hematochezia (5 of 28), melena (11 of 28), and obscure occult bleeding (12 of 28). Active Small Bowel Bleeding Twenty-eight out of 116 VCE’s included for analysis were found to have active small bowel GIB. All of these studies had at least one area of active GIB which would be identified using QV ? SBI, and 27 of 28 had at least one SBI corresponding to a frame with active GIB. Of note, the one VCE without SBI corresponding to active bleeding did have other frames with the SBI indicator. Using QV ? SBI to assess for active small bowel GIB, blinded reader 1 found 33 studies with active GIB and blinded reader 2 found 34 studies with active GIB (Table 1). Both blinded readers correctly identified all 28 studies found to have active small bowel GIB by SV. The indications for the additional VCEs identified by the readers were hematochezia (one) and melena (five, one which blinded reader 1 did not identify as active GIB). The sensitivity to assess for active GIB using QV ? SBI was 100 % for both blinded readers, with a specificity of 94.3 % for blinded reader 1 and 93.2 % for blinded reader 2. This corresponded to a NPV of 100 % and a PPV of 82.4 %. There was excellent interobserver agreement (j = 0.9790) corresponding to a 99.14 % agreement between the two blinded readers. For SBI alone, the sensitivity was 93.9 % for blinded reader 1 and 94.1 % for blinded reader 2, while the specificities were 30.1 and 32.9 %, respectively (Table 2). This corresponded to a PPV of 34.8 % for blinded reviewer 1 and 36.8 % for blinded reviewer 2, with NPV of 92.6 and 93.1 %, respectively. Interobserver agreement for SBI was high (j = 0.8785) corresponding to a 92.2 % agreement. Both blinded readers correctly identified the one VCE which SBI did not correspond with active bleeding. The
Intrerobserver agreement (Kappa)
0.979 (99.14 %)
Sensitivity (%)
Specificity (%)
PPV (%)
NPV (%)
100
94.3
84.8
100
100
93.2
82.4
100
Table 2 Results of suspected blood indicator as it corresponds to active bleeding No red tag, no bleeding
Red tag, no bleeding
Red tag in frame corresponding active bleeding
Red tag not in frame corresponding to active bleeding
Blinded reader 1 No active bleeding on standard view
25
58
4
1
0
27
1
Active 0 bleeding on standard view Blinded reader 2 No active bleeding on standard view
27
55
5
1
Active bleeding on standard view
0
0
27
1
ability of QV ? SBI compared to SBI alone to accurately identify active GIB failed to reach statistical significance (blinded reader 1 p = 0.89; blinded reader 2 p = 0.78). Follow-Up Post-VCE Of the 28 studies read to have active bleeding, 23 studies (82.1 %) had further endoscopic evaluation (two standard esophagogastroduodenoscopies (EGDs), one push enteroscopy, 18 balloon-assisted enteroscopy, and two repeat VCEs). Out of these 23 studies, 17 identified a potential candidate lesion or lesions, with 14 requiring intervention (Fig. 1). All of the remaining 6 studies which did not show a candidate lesion were either negative or the lesions were thought to be not reached. Five of the 28 (17.9 %) studies showing active bleeding did not have any further
123
2506
Dig Dis Sci (2014) 59:2503–2507
Fig. 1 Evaluation of VCEs with active small bowel bleeding
investigation (one recommended to have double-balloon enteroscopy but not performed for unclear reasons, one patient died before recommended double-balloon enteroscopy could be performed, three with no endoscopy recommended). Eighty-eight VCE studies were negative for active bleeding. Of these, 30 (34.1 %) had a follow-up study recommended due to other small bowel pathology seen on SV (Fig. 2). Within the 30 follow-up studies, 11 found a significant lesion that necessitated intervention. The mean reading time between the two reviewers for the QV ? SBI read was 3 min and 20 s, with a range of 2 min and 6 min and 34 s.
Discussion This is the first study of its kind to demonstrate that active small bowel bleeding may be rapidly and accurately detected using the Quickview (QV) and suspected blood indicator (SBI) functions in GIVEN’s Rapid Reader 6.0 software. One of the limitations to rapid detection of active small bowel bleeding using VCE is the amount of time required to read each study, which may lead to capsule studies being read days after the study was performed. Currently no validated method exists to rapidly read VCE studies to assess for active small bowel bleeding. In our
123
Fig. 2 Evaluation of VCEs without active small bowel bleeding
retrospective study, we reviewed 116 VCE’s ordered to assess for possible small bowel bleeding, of which 28 studies contained active small bowel bleeding. To our knowledge, this is the largest case series of VCEs with active small bowel bleeding presented. Using this technique with QV ? SBI, we were able to accurately and efficiently identify every study with active small bowel bleeding, with a high degree of sensitivity and specificity. This technique took a mean diagnostic reading time of 3 min 20 s. This markedly reduced the diagnostic reading time required to accurately assess for active small bowel GIB compared to previously published data. While accurate to detect active bleeding in the small bowel, this method of QV ? SBI is not designed to replace a complete and thorough quality assessment of a VCE study. Instead, utilizing this technique, a clinician could use the QV ? SBI method as an initial VCE screen, quickly and accurately assessing for active bleeding. In cases of active bleeding, standard view of the capsule study is then immediately performed and intervention is scheduled. If no bleeding is observed, the clinician can read the complete VCE using standard methods at their convenience as per routine. Not surprisingly, our analysis of the negative studies showed that a significant amount (30/88, 34.1 %) of studies had findings concerning enough to have
Dig Dis Sci (2014) 59:2503–2507
a follow-up study recommended for further evaluation. Ultimately, 11 of those studies had a significant lesion that was found and intervened on. On assessment of SBI alone, we validated previous studies showing strong sensitivity (93.9 and 94.1 % for readers 1 and 2, respectively) and NPV for the SBI (92.6 and 93.1 %). We also found it to have poor specificity and PPV. Using the SBI tool alone is not an acceptable way to review VCE studies as there are too many ‘‘red tags’’ that are clinically irrelevant and do not capture all bleeding lesions. Of note, one study with active bleeding on SV was missed by the SBI. There are limitations to this study. As a retrospective study, we did not have a gold standard endoscopic method of validating our findings for QV ? SBI and SV reads. The possibility exists that culprit bleeding lesions were missed by both QV reviewers as well as the clinician reading in SV. However, our primary aim was to assess the accuracy of QV ? SBI when compared to SV. As further prospective studies are warranted on this topic, studies can be designed to include endoscopic verification of capsule findings. Also, our study only looked at QV ? SBI for active bleeding. We did not attempt to locate any other potential causes of bleeding and thus missed other lesions that were clinically significant as noted above. Additionally, there were several studies with bleeding on QV ? SBI that were deemed to not have active bleeding on SV. While the false-positive rate is relatively low, it may add time to a busy gastroenterologist’s schedule with a more urgent review of the capsule in SV and is a limitation with the QV ? SBI technique. One explanation for the false-positive rate is lack of expertise and experience of the novice readers utilized to review each VCE using SBI alone, or QBV ? SBI may have contributed to the number of false-positive studies showing active small bowel bleeding compared to SV. However, it is important to note no studies with active bleeding were missed using QV ? SBI (false negative rate is zero). Quickview ? SBI is a rapid method of interpreting VCE performed for suspected small bowel bleeding and is extremely sensitive and specific in detecting active small bowel bleeding. While QV ? SBI is not a replacement for a standard view read, we advocate that the QV ? SBI
2507
method is useful as an initial evaluation to identify active small bowel bleeding that allows for early detection and urgent intervention. Conflict of interest Carol E. Semrad—has received consulting fees from Given Imaging. David T. Rubin—has received consulting fees from Given Imaging. For the remaining authors none were declared.
References 1. Appleyard M, Glukhovsky A, Swain P. Wireless-capsule diagnostic endoscopy for recurrent small-bowel bleeding. N Engl J Med. 2001;344:232–233. 2. Scapa E, Jacob H, Lewkowicz S, et al. Initial experience of wireless-capsule endoscopy for evaluating occult gastrointestinal bleeding and suspected small bowel pathology. Am J Gastroenterol. 2002;97:2776–2779. 3. Delvaux M, Gerard G. Capsule endoscopy in 2005: facts and perspectives. Best Pract Res Clin Gastroenterol. 2006;20:23–39. 4. Shiotani A, Honda K, Kawakami M, et al. Evaluation of RAPID((R)) 5 access software for examination of capsule endoscopies and reading of the capsule by an endoscopy nurse. J Gastroenterol. 2011;46:138–142. 5. Westerhof J, Koornstra JJ, Weersma RK. Can we reduce capsule endoscopy reading times? Gastrointest Endosc. 2009;69: 497–502. 6. Saurin JC, Lapalus MG, Cholet F, et al. Can we shorten the small-bowel capsule reading time with the ‘‘Quick-view’’ image detection system? Dig Liver Dis. 2012;44:477–481. 7. Kyriakos N, Karagiannis S, Galanis P, et al. Evaluation of four time-saving methods of reading capsule endoscopy videos. Eur J Gastroenterol Hepatol. 2012;24:1276–1280. 8. Koulaouzidis A, Smirnidis A, Douglas S, et al. QuickView in small-bowel capsule endoscopy is useful in certain clinical settings, but QuickView with Blue Mode is of no additional benefit. Eur J Gastroenterol Hepatol. 2012;24:1099–1104. 9. Shiotani A, Honda K, Kawakami M, et al. Analysis of smallbowel capsule endoscopy reading by using Quickview mode: training assistants for reading may produce a high diagnostic yield and save time for physicians. J Clin Gastroenterol. 2012;46: e92–e95. 10. D’Halluin PN, Delvaux M, Lapalus MG, et al. Does the ‘‘Suspected Blood Indicator’’ improve the detection of bleeding lesions by capsule endoscopy? Gastrointest Endosc. 2005;61: 243–249. 11. Hosoe N, Rey JF, Imaeda H, et al. Evaluations of capsule endoscopy software in reducing the reading time and the rate of false negatives by inexperienced endoscopists. Clin Res Hepatol Gastroenterol. 2012;36:66–71.
123
ORIGINAL ARTICLE
A Rapid and Accurate Method to Detect Active Small Bowel Gastrointestinal Bleeding on Video Capsule Endoscopy Adam C. Stein • Anoop Appannagari Ibrahim Habib • Carol E. Semrad • David T. Rubin
•
Received: 22 January 2014 / Accepted: 7 August 2014 / Published online: 22 August 2014 Ó Springer Science+Business Media New York 2014
Abstract Background Video capsule endoscopy (VCE) is indicated to evaluate for suspected small bowel bleeding, but ‘‘standard view’’ (SV) evaluation is time-consuming. Rapid Reader 6.0 software (Given Imaging, Duluth GA) contains two computer algorithmic systems: (1) ‘‘Quickview’’ (QV) which automatically skips similar images and (2) a pixel analysis program that identifies suspected blood (SBI). Combining the two modalities (QV ? SBI) may provide a faster modality to assess for active small bowel bleeding. Aims This study was designed to assess the accuracy of QV ? SBI for small bowel bleeding compared to SV findings. Methods This is a retrospective, case–control study at a single tertiary care referral hospital including all patients with VCE performed for suspected small bowel bleeding from 4/2007 to 3/2011. All studies were previously read using SV by one of two experienced faculty (CS, DR). The primary outcome was diagnostic accuracy of QV ? SBI in assessing for active small bowel bleeding compared to SV.
Adam C. Stein and Anoop Appannagari are the co-first authors. A. C. Stein C. E. Semrad D. T. Rubin (&) Section of Gastroenterology, Hepatology, and Nutrition, Department of Medicine, University of Chicago Medicine, 5841 S. Maryland Ave., MC 4076, Chicago, IL 60637, USA e-mail: [email protected] A. Appannagari Division of Gastroenterology, Department of Medicine, Loyola University, Maywood, IL, USA I. Habib Department of Medicine, Advocate Lutheran General Hospital, Park Ridge, IL, USA
Results A total of 116 VCE were included, 28 with active small bowel bleeding identified by original SV. Using QV ? SBI, all 28 VCEs with active small bowel bleeding were identified. The sensitivity of QV ? SBI to detect active bleeding was 100 %, while the specificity was 93–94 %. The mean time to identify landmarks and read the entire study was 3 min 20 s. Conclusions The QV ? SBI reading format of VCE is an efficient, highly sensitive modality to assess for potential small bowel bleeding. Keywords Video capsule endoscopy Small bowel bleeding Quickview Suspected blood indicator Given
Introduction Video capsule endoscopy (VCE) is a well-established, noninvasive modality to identify obscure gastrointestinal bleeding (GIB) [1, 2]. VCE provides excellent diagnostic yield for the evaluation obscure GI bleeding. Despite this accuracy, a limitation to VCE has been the amount of time required to read each study. While the exam itself usually requires minimal effort by the physician, the time required for the exam itself, the amount of time required to read each study can be burdensome. The standard method of reading a VCE study involves watching streaming video of approximately 50,000 images and requires a significant time commitment, with prior studies reporting average or median reading times of 17–60 min [3–5]. For indications that may require urgent intervention based on results of the VCE, such as evaluation for GIB, prioritizing time to read a study in a timely manner may add a significant burden to a busy gastroenterologist. To improve the efficiency of reading VCE
123
2504
studies, several methods have been proposed including software that filters the number of images viewed and utilizing physician extenders (for example, nurses) to preliminarily review the study [4, 6–9]. The Rapid Reader 6.0 software program (Given imaging, Duluth, GA) contains two software algorithmic systems that allow for more rapid analysis of VCE studies. The ‘‘Quickview’’ (QV) system is a program that reduces the image quantity by analyzing specific patterns and colors to provide a shorter composite video. Theoretically, QV allows for a rapid and accurate overview of the most significant lesions detected by VCE. An additional algorithm contained in Rapid Reader 6.0, the suspected blood indicator (SBI), analyzes pixels to look for a red color that may be consistent with active or potential bleeding. There have been limited studies on the diagnostic efficacy of QV, and most results show that there may be an unacceptably high increase in diagnostic miss rate [5, 6, 8, 9]. The published data regarding SBI is also limited, with one study showing an improved sensitivity and negative predictive value (83, 97 %) for active bleeding but lower specificity and positive predictive value (66, 23 %) [10]. No studies have assessed combining QV and SBI for obscure gastrointestinal bleeding. We performed a study to assess the accuracy of QV ? SBI compared to standard view (SV) in patients with suspected small bowel bleeding. We then assessed the amount of time required to analyze each exam in efforts to develop an algorithm for suspected GIB.
Materials and Methods We retrospectively identified all VCE’s performed to assess for obscure gastrointestinal bleeding, melena of unknown origin, and hematochezia of unknown origin at the University of Chicago Medicine from April 1, 2007 to March 31, 2011. The study was approved by the Investigational Review Board at the University of Chicago. Patients We identified all VCE’s ordered for the following indications: obscure GI bleed, melena of unknown origin, and hematochezia of unknown origin. Studies were excluded if the capsule did not reach the small bowel, if there were technical difficulties leading to inability to adequately visualize the small bowel mucosa, or if the study was unavailable at time of review. If multiple VCE’s were identified for the same patient, all studies were included. All identified VCE’s were previously read in SV by two experienced faculty members (CS, DR), whose reading served as the reference standard for this study. The experienced faculty members read each VCE using quad view
123
Dig Dis Sci (2014) 59:2503–2507
as the standard setting, toggling to dual view when needed. Each study was read at variable speed rates. Each VCE was read unblinded using SV to verify active bleeding and to assess whether areas of active bleeding would be identified using SBI alone as well as QV ? SBI. Patient charts of all studies included for analysis were reviewed to determine whether further investigation or intervention was recommended and performed based on SV VCE findings. Blinded Review Prior to the beginning of the study, two novice readers (AA, AS) were trained in VCE interpretation by reading 25 studies in SV under supervision by advanced readers. A recent study validated the learning curve for becoming adequately trained in VCE and found that with feedbackbased education, 20 studies would reduce the possibility of missing lesions [11]. All included VCEs were independently assessed for active small bowel bleeding using only the QV ? SBI mode by the two novice readers blinded to the original faculty member’s read. Reads were performed without restriction, including the ability to pause and rewind the VCE as desired. Readers were able to toggle between single view, dual view, and quad view, with the standard setting of dual view. Data recorded during the QV ? SBI read included assessment for active small bowel GIB, if the area of active small bowel GIB had a corresponding SBI, the presence of SBI without active small bowel GIB, and the total time required to complete the necessary data collection for each VCE. The readers did not attempt to identify lesions that may have represented a potential bleeding source. The two individual reads using QB ? SBI were compared independently to the original SV reads to assess for accuracy of identifying active small bowel GIB. Statistical Analysis Standard view findings were compared to each blinded reader under SBI alone as well as QV ? SBI for sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). Comparison of SBI alone with QV ? SBI for each blinded reader was done by chisquared analysis. Interobserver agreement was assessed between the two independent QV ? SBI novice reads.
Results A total of 131 individual VCEs were identified during the study period. Of these, 15 were excluded (eight were unavailable for review, six did not reach the small bowel,
Dig Dis Sci (2014) 59:2503–2507
2505
Table 1 Results of review with standard view and QV ? SBI
Active bleeding Standard view (gold standard)
a
All active bleeding seen on standard view was found on QV ? SBI by both reader 1 and reader 2
No bleeding
28
88
QV ? SBI, reader 1
33a
83
QV ? SBI, reader 2
34a
82
and one was of poor technical quality limiting adequate small bowel visualization). A total of 116 VCEs were included for analysis from 98 patients, with nine patients having two VCEs included. Of the nine patients with multiple VCE’s ,the median time between VCEs was 32 days with a range of 2–916 days. The indications for the VCEs which found active bleeding include hematochezia (5 of 28), melena (11 of 28), and obscure occult bleeding (12 of 28). Active Small Bowel Bleeding Twenty-eight out of 116 VCE’s included for analysis were found to have active small bowel GIB. All of these studies had at least one area of active GIB which would be identified using QV ? SBI, and 27 of 28 had at least one SBI corresponding to a frame with active GIB. Of note, the one VCE without SBI corresponding to active bleeding did have other frames with the SBI indicator. Using QV ? SBI to assess for active small bowel GIB, blinded reader 1 found 33 studies with active GIB and blinded reader 2 found 34 studies with active GIB (Table 1). Both blinded readers correctly identified all 28 studies found to have active small bowel GIB by SV. The indications for the additional VCEs identified by the readers were hematochezia (one) and melena (five, one which blinded reader 1 did not identify as active GIB). The sensitivity to assess for active GIB using QV ? SBI was 100 % for both blinded readers, with a specificity of 94.3 % for blinded reader 1 and 93.2 % for blinded reader 2. This corresponded to a NPV of 100 % and a PPV of 82.4 %. There was excellent interobserver agreement (j = 0.9790) corresponding to a 99.14 % agreement between the two blinded readers. For SBI alone, the sensitivity was 93.9 % for blinded reader 1 and 94.1 % for blinded reader 2, while the specificities were 30.1 and 32.9 %, respectively (Table 2). This corresponded to a PPV of 34.8 % for blinded reviewer 1 and 36.8 % for blinded reviewer 2, with NPV of 92.6 and 93.1 %, respectively. Interobserver agreement for SBI was high (j = 0.8785) corresponding to a 92.2 % agreement. Both blinded readers correctly identified the one VCE which SBI did not correspond with active bleeding. The
Intrerobserver agreement (Kappa)
0.979 (99.14 %)
Sensitivity (%)
Specificity (%)
PPV (%)
NPV (%)
100
94.3
84.8
100
100
93.2
82.4
100
Table 2 Results of suspected blood indicator as it corresponds to active bleeding No red tag, no bleeding
Red tag, no bleeding
Red tag in frame corresponding active bleeding
Red tag not in frame corresponding to active bleeding
Blinded reader 1 No active bleeding on standard view
25
58
4
1
0
27
1
Active 0 bleeding on standard view Blinded reader 2 No active bleeding on standard view
27
55
5
1
Active bleeding on standard view
0
0
27
1
ability of QV ? SBI compared to SBI alone to accurately identify active GIB failed to reach statistical significance (blinded reader 1 p = 0.89; blinded reader 2 p = 0.78). Follow-Up Post-VCE Of the 28 studies read to have active bleeding, 23 studies (82.1 %) had further endoscopic evaluation (two standard esophagogastroduodenoscopies (EGDs), one push enteroscopy, 18 balloon-assisted enteroscopy, and two repeat VCEs). Out of these 23 studies, 17 identified a potential candidate lesion or lesions, with 14 requiring intervention (Fig. 1). All of the remaining 6 studies which did not show a candidate lesion were either negative or the lesions were thought to be not reached. Five of the 28 (17.9 %) studies showing active bleeding did not have any further
123
2506
Dig Dis Sci (2014) 59:2503–2507
Fig. 1 Evaluation of VCEs with active small bowel bleeding
investigation (one recommended to have double-balloon enteroscopy but not performed for unclear reasons, one patient died before recommended double-balloon enteroscopy could be performed, three with no endoscopy recommended). Eighty-eight VCE studies were negative for active bleeding. Of these, 30 (34.1 %) had a follow-up study recommended due to other small bowel pathology seen on SV (Fig. 2). Within the 30 follow-up studies, 11 found a significant lesion that necessitated intervention. The mean reading time between the two reviewers for the QV ? SBI read was 3 min and 20 s, with a range of 2 min and 6 min and 34 s.
Discussion This is the first study of its kind to demonstrate that active small bowel bleeding may be rapidly and accurately detected using the Quickview (QV) and suspected blood indicator (SBI) functions in GIVEN’s Rapid Reader 6.0 software. One of the limitations to rapid detection of active small bowel bleeding using VCE is the amount of time required to read each study, which may lead to capsule studies being read days after the study was performed. Currently no validated method exists to rapidly read VCE studies to assess for active small bowel bleeding. In our
123
Fig. 2 Evaluation of VCEs without active small bowel bleeding
retrospective study, we reviewed 116 VCE’s ordered to assess for possible small bowel bleeding, of which 28 studies contained active small bowel bleeding. To our knowledge, this is the largest case series of VCEs with active small bowel bleeding presented. Using this technique with QV ? SBI, we were able to accurately and efficiently identify every study with active small bowel bleeding, with a high degree of sensitivity and specificity. This technique took a mean diagnostic reading time of 3 min 20 s. This markedly reduced the diagnostic reading time required to accurately assess for active small bowel GIB compared to previously published data. While accurate to detect active bleeding in the small bowel, this method of QV ? SBI is not designed to replace a complete and thorough quality assessment of a VCE study. Instead, utilizing this technique, a clinician could use the QV ? SBI method as an initial VCE screen, quickly and accurately assessing for active bleeding. In cases of active bleeding, standard view of the capsule study is then immediately performed and intervention is scheduled. If no bleeding is observed, the clinician can read the complete VCE using standard methods at their convenience as per routine. Not surprisingly, our analysis of the negative studies showed that a significant amount (30/88, 34.1 %) of studies had findings concerning enough to have
Dig Dis Sci (2014) 59:2503–2507
a follow-up study recommended for further evaluation. Ultimately, 11 of those studies had a significant lesion that was found and intervened on. On assessment of SBI alone, we validated previous studies showing strong sensitivity (93.9 and 94.1 % for readers 1 and 2, respectively) and NPV for the SBI (92.6 and 93.1 %). We also found it to have poor specificity and PPV. Using the SBI tool alone is not an acceptable way to review VCE studies as there are too many ‘‘red tags’’ that are clinically irrelevant and do not capture all bleeding lesions. Of note, one study with active bleeding on SV was missed by the SBI. There are limitations to this study. As a retrospective study, we did not have a gold standard endoscopic method of validating our findings for QV ? SBI and SV reads. The possibility exists that culprit bleeding lesions were missed by both QV reviewers as well as the clinician reading in SV. However, our primary aim was to assess the accuracy of QV ? SBI when compared to SV. As further prospective studies are warranted on this topic, studies can be designed to include endoscopic verification of capsule findings. Also, our study only looked at QV ? SBI for active bleeding. We did not attempt to locate any other potential causes of bleeding and thus missed other lesions that were clinically significant as noted above. Additionally, there were several studies with bleeding on QV ? SBI that were deemed to not have active bleeding on SV. While the false-positive rate is relatively low, it may add time to a busy gastroenterologist’s schedule with a more urgent review of the capsule in SV and is a limitation with the QV ? SBI technique. One explanation for the false-positive rate is lack of expertise and experience of the novice readers utilized to review each VCE using SBI alone, or QBV ? SBI may have contributed to the number of false-positive studies showing active small bowel bleeding compared to SV. However, it is important to note no studies with active bleeding were missed using QV ? SBI (false negative rate is zero). Quickview ? SBI is a rapid method of interpreting VCE performed for suspected small bowel bleeding and is extremely sensitive and specific in detecting active small bowel bleeding. While QV ? SBI is not a replacement for a standard view read, we advocate that the QV ? SBI
2507
method is useful as an initial evaluation to identify active small bowel bleeding that allows for early detection and urgent intervention. Conflict of interest Carol E. Semrad—has received consulting fees from Given Imaging. David T. Rubin—has received consulting fees from Given Imaging. For the remaining authors none were declared.
References 1. Appleyard M, Glukhovsky A, Swain P. Wireless-capsule diagnostic endoscopy for recurrent small-bowel bleeding. N Engl J Med. 2001;344:232–233. 2. Scapa E, Jacob H, Lewkowicz S, et al. Initial experience of wireless-capsule endoscopy for evaluating occult gastrointestinal bleeding and suspected small bowel pathology. Am J Gastroenterol. 2002;97:2776–2779. 3. Delvaux M, Gerard G. Capsule endoscopy in 2005: facts and perspectives. Best Pract Res Clin Gastroenterol. 2006;20:23–39. 4. Shiotani A, Honda K, Kawakami M, et al. Evaluation of RAPID((R)) 5 access software for examination of capsule endoscopies and reading of the capsule by an endoscopy nurse. J Gastroenterol. 2011;46:138–142. 5. Westerhof J, Koornstra JJ, Weersma RK. Can we reduce capsule endoscopy reading times? Gastrointest Endosc. 2009;69: 497–502. 6. Saurin JC, Lapalus MG, Cholet F, et al. Can we shorten the small-bowel capsule reading time with the ‘‘Quick-view’’ image detection system? Dig Liver Dis. 2012;44:477–481. 7. Kyriakos N, Karagiannis S, Galanis P, et al. Evaluation of four time-saving methods of reading capsule endoscopy videos. Eur J Gastroenterol Hepatol. 2012;24:1276–1280. 8. Koulaouzidis A, Smirnidis A, Douglas S, et al. QuickView in small-bowel capsule endoscopy is useful in certain clinical settings, but QuickView with Blue Mode is of no additional benefit. Eur J Gastroenterol Hepatol. 2012;24:1099–1104. 9. Shiotani A, Honda K, Kawakami M, et al. Analysis of smallbowel capsule endoscopy reading by using Quickview mode: training assistants for reading may produce a high diagnostic yield and save time for physicians. J Clin Gastroenterol. 2012;46: e92–e95. 10. D’Halluin PN, Delvaux M, Lapalus MG, et al. Does the ‘‘Suspected Blood Indicator’’ improve the detection of bleeding lesions by capsule endoscopy? Gastrointest Endosc. 2005;61: 243–249. 11. Hosoe N, Rey JF, Imaeda H, et al. Evaluations of capsule endoscopy software in reducing the reading time and the rate of false negatives by inexperienced endoscopists. Clin Res Hepatol Gastroenterol. 2012;36:66–71.
123
