- Email: [email protected]
Performance Characteristics of the Suspected Blood Indicator Feature in Capsule Endoscopy According to Indication for Study
CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2008;6:298 –301
Performance Characteristics of the Suspected Blood Indicator Feature in Capsule Endoscopy According to Indication for Study JONATHAN M. BUSCAGLIA, SAMUEL A. GIDAY, SERGEY V. KANTSEVOY, JOHN O. CLARKE, PRISCILLA MAGNO, ELAINE YONG, and GERARD E. MULLIN Division of Gastroenterology and Hepatology, Johns Hopkins University School of Medicine, Johns Hopkins Hospital, Baltimore, Maryland
Background & Aims: The suspected blood indicator (SBI) feature of wireless capsule endoscopy (WCE) was developed for rapid screening of intestinal lesions with bleeding potential. Our aim was to assess the accuracy and performance characteristics of the SBI according to the indications for study in a large cohort of patients. Methods: We reviewed collected data on all WCE studies performed at Johns Hopkins Hospital from January 2006 to June 2007. Study indications were as follows: anemia of unknown origin (n ⴝ 53), obscure gastrointestinal bleeding (n ⴝ 112), suspected Crohn’s disease (n ⴝ 122), and other (n ⴝ 4). Concordant and discordant findings between gastroenterologists’ readings and SBI were recorded for each patient. Results: A total of 221 lesions with bleeding potential was detected. The overall sensitivity, specificity, positive predictive value, and negative predictive value for the SBI were 56.4%, 33.5%, 24.0%, and 67.3%, respectively. For actively bleeding lesions, the SBI sensitivity and positive predictive value were only 58.3% and 70%, respectively. The sensitivity was highest (64%) in patients undergoing WCE for suspected Crohn’s disease, with a negative predictive value of 80.4%. The sensitivity was only 58.3% and 41.3%, respectively, in studies performed for obscure gastrointestinal bleeding and anemia. Conclusions: Performance characteristics of the currently available SBI feature in WCE are suboptimal and insufficient to screen for lesions with bleeding potential. Even in patients with active intestinal bleeding, the sensitivity of SBI was less than 60%, which is lower than previously reported. However, in patients with suspected Crohn’s disease, the high sensitivity and negative predictive value of SBI may make it a useful tool for the detection of large areas of abnormal mucosa.
W
ireless capsule endoscopy (WCE) was first developed for advanced imaging of the small intestine.1–3 Although its clinical efficacy is most proven in the diagnosis of obscure gastrointestinal bleeding,4 – 6 it also has been helpful as an adjunct to radiologic studies for patients with suspected Crohn’s disease, celiac disease, small-bowel tumors, anemia of unknown origin, chronic abdominal pain, and other indications.7–11 The WCE system (Given Imaging, Yoqneam, Israel) in combination with RAPID software (Duluth, Georgia) has a suspected blood indicator (SBI) feature that initially was developed to facilitate the screening of possible sites of active bleeding, thereby significantly reducing the reading time spent by a physician-reviewer.12 The SBI function works by identifying redcolored pixels on the viewing screen, and then automatically
annotating the study scroll bar with a red hash mark; thus indicating to the reader that a potential bleeding lesion or other site of pathology may be present. Previously reported studies on the accuracy of the SBI either have been dedicated solely to gastrointestinal bleeding,13 or were limited by a small number of enrolled patients.12,14 The aim of our study was to assess the accuracy and performance characteristics of the SBI according to the indications for the study in a large cohort of patients.
Methods Consecutive patients undergoing WCE at our institution from January 2006 through June 2007 were reviewed for the study. Permission to review patient records was granted by the Johns Hopkins University Institutional Review Board. All patients were asked to refrain from eating or drinking at least 8 hours before swallowing the Given M2A video capsule endoscope. Laxative bowel preparation was not used. Patients were allowed to eat and drink 4 hours after the start of their study. Each WCE study was interpreted by 1 of 5 board-certified/board-eligible gastroenterologists (J.M.B., S.A.G., P.M., J.O.C., and E.Y.). All 5 readers had experience reviewing greater than 50 cases each. Images were reviewed at a speed of 8 to 15 frames per second. Captured thumbnail images and summary reports were re-examined and verified by a separate, boardcertified gastroenterologist (G.E.M.) who had experience with at least 250 cases. There was greater than 95% concordance between the verifying reader (G.E.M.) and each of the 5 initial reviewers. After each study, the interpreting physician was asked to record all endoscopic findings within our WCE database. The following pathologic lesions were considered significant as it pertains to the SBI function of the Given software: ulcers, erosions, arteriovenous malformations, red spots, varices, venous ectasias, blood, and blood clots. Physicians were asked to document whether the SBI accurately predicted the lesions discovered within each patient’s study. Performance characteristics of the SBI were defined by concordance between what was positively sensed by the Given software system, and what was positively detected by the physician-reviewer. Cases in which a significant lesion was found by the reviewer, but not detected by the SBI, also were documented. Those cases in which SBI Abbreviations used in this paper: GI, gastrointestinal; NPV, negative predictive value; PPV, positive predictive value; SBI, suspected blood indicator; WCE, wireless capsule endoscopy. © 2008 by the AGA Institute 1542-3565/08/$34.00 doi:10.1016/j.cgh.2007.12.029
March 2008
PERFORMANCE OF THE SUSPECTED BLOOD INDICATOR
Table 1. Patient Demographics and Study Indications
Indication for WCE
Anemia of unknown origin (n ⫽ 53)
Obscure GI bleeding (n ⫽ 112)
Suspected Crohn’s disease (n ⫽ 122)
Total (%) (n ⫽ 287)
Males Females Mean age, y
25 28 53.7
64 48 58.9
36 86 46.3
125 (43.6) 162 (56.4) 53.0
correctly predicted the findings in one part of the study, but failed to detect the lesions in another part of the study, were recorded as well. The indications for each WCE study also were documented for all patients.
Results A total of 291 patient WCE recordings were studied between January 2006 and June 2007. When the total number of studies was divided according to clinical indications, the single most common indication was suspicion of small-bowel Crohn’s disease in patients with diarrhea and abdominal pain (n ⫽ 122; 41.9%). Obscure gastrointestinal (GI) bleeding was the second most common indication (n ⫽ 112; 38.5%), followed by anemia of unknown origin (n ⫽ 53; 18.2%). Four additional studies were performed for other indications (1.4%). These 4 studies were not included in the analysis, making the final number of WCE studies equal to 287 (Table 1). There were 125 males (43.6%) and 162 females (56.4%), with a mean age of 53 years. A total of 221 pathologic lesions were identified in 287 studies by 5 interpreting physicians (Table 2). Mucosal red spots were the most common lesions recognized (69 of 221; 31.2%), followed by erosions (55 of 221; 24.9%), arteriovenous malformations (41 of 221; 18.6%), ulcers (41 of 221; 18.6%), venous ectasias (11 of 221; 5%), and intestinal varices (1 of 221; 0.5%). There were 15 cases in which blood was recognized, and 12 separate lesions were documented to account for the bleeding: 4 ulcers, 3 erosions, and 5 arteriovenous malformations. In 3 cases (1.4%), blood clots or active bleeding was documented in the small bowel by the interpreting physician, yet no discernable lesion was recognized (Table 2).
299
Table 3. Concordance Between Reader and the SBI Function Anemia of Suspected unknown Obscure GI Crohn’s origin bleeding disease Total Total number of analyzed recordings SBI present, significant lesion present (true positive), n SBI present, significant lesion absent (false positive), n SBI absent, significant lesion absent (true negative), n SBI absent, significant lesion present (false negative), n
53
112
122
287
7
21
16
44
24
55
60
139
12
21
37
70
10
15
9
34
There was a similar proportion of overall patients in the obscure GI bleeding group and the anemia group with positive findings detected by the interpreting physician: 32.1% (36 of 112) and 32.1% (17 of 53), respectively. In the suspected smallbowel Crohn’s disease group, 20.5% (25 of 122) of patients had positive findings. Table 3 highlights the concordance between the reader and the SBI function of the software. Performance characteristics of the SBI did not differ among the 5 different readers. Of a total of 287 studies, 44 patients had a significant lesion with an associated SBI (true positive). In 139 patients, an SBI was present without any discernable disease (false positive). The study was negative in 70 patients without an SBI present (true negative), and 34 patients had a lesion recognized by the phy-
Table 2. Pathologic Lesions With Bleeding Potential Found in 287 WCE Studies
Lesion Ulcers Erosions Arteriovenous malformations Red spots Varices Venous ectasias Blood clots and/or active bleeding Total
Without active bleeding (%)
Actively bleeding (%)
Total (%)
37 (16.7) 52 (23.5) 36 (16.3)
4 (1.8) 3 (1.4) 5 (2.3)
41 (18.6) 55 (24.9) 41 (18.6)
69 (31.2) 1 (0.5) 11 (5.0) —
0 0 0 3 (1.4)
69 (31.2) 1 (0.5) 11 (5.0) 3 (1.4)
206 (93.2)
15 (6.8)
221 (100)
Figure 1. Example of a false-negative SBI in a patient with obscureovert GI bleeding found to have lymphonodular hyperplasia on resection of the terminal ileum.17 PillCam (Given Imaging, Yogneam, Israel)
300
BUSCAGLIA ET AL
CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 6, No. 3
Table 4. Performance Characteristics of the SBI for the Prediction of Pathologic Lesions With Bleeding Potential Indication for WCE
Anemia of unknown origin
Obscure GI bleeding
Suspected Crohn’s disease
Total
Sensitivity Specificity PPV NPV
41.3 (19.4–66.6) 33.3 (19.1–51.1) 22.6 (10.3–41.5) 54.5 (32.7–74.9)
58.3 (40.1–74.0) 27.6 (18.3–39.3) 27.6 (18.3–39.3) 58.3 (40.1–74.0)
64.0 (42.6–81.3) 38.1 (28.6–48.6) 21.1 (12.9–32.2) 80.4 (65.6–90.1)
56.4 (44.7–67.4) 33.5 (27.2–40.4) 24.0 (18.2–31.0) 67.3 (57.3–76.0)
NOTE. Percentages (95% confidence intervals) shown.
sician-reviewer without an associated SBI (false negative, Figure 1). Concordance values based on each of the 3 major indications for the study also are listed in Table 3. Table 4 lists the performance characteristics of the SBI in all 287 WCE studies, as well as performance characteristics according to each clinical indication. The overall sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the SBI were 56.4%, 33.5%, 24.0%, and 67.3%, respectively. In patients undergoing WCE for anemia of unknown origin, the sensitivity, specificity, PPV, and NPV of the SBI were 41.3%, 33.3%, 22.6%, and 54.5%, respectively (Table 4). In patients undergoing WCE for obscure GI bleeding, the sensitivity, specificity, PPV, and NPV of the SBI were 58.3%, 27.6%, 27.6%, and 58.3%, respectively. When performed for the suspicion of smallbowel Crohn’s disease, the sensitivity, specificity, PPV, and NPV were 64.0%, 38.1%, 21.1%, and 80.4%, respectively. Factoring in only those cases with active bleeding, the overall sensitivity of the SBI was 58.3%, with a PPV of 70.0% (Table 5). When accounting for all patients studied, the sensitivity of the SBI was highest (64%) in those undergoing WCE for suspicion of small-bowel Crohn’s disease. The specificity of the SBI was poor for all 3 indications, with the highest value being 38.1% in the suspected Crohn’s group (Table 4). The NPV also was highest in this same group, with a value of 80.4%. The PPV was poor in all groups, with a peak value of 27.6% in patients with obscure GI bleeding.
Discussion Since the invention of capsule endoscopy, it has been used most often and most effectively for the diagnosis of obscure GI bleeding.15,16 More recently, however, its indications have broadened with its usefulness shown in the diagnosis of inflammatory bowel disease and the evaluation of chronic abdominal pain.7–11 Early in its development, the WCE software system (Given Imaging) offered an SBI feature with each of its recordings. This small red line appearing at the top of the time tracing initially was designed as a rapid screening tool, allowing physicians to identify areas of active bleeding quickly.12 Although the SBI feature has been available for more than 5 years, reports of its clinical usefulness are relatively sparse, usually completed on a Table 5. Sensitivity and PPV of the SBI in 15 WCE Studies With Blood Clots or Active Bleeding Sensitivity PPV
58.3 (28.6–83.6) 70.0 (35.4–91.9)
NOTE. Percentages (95% confidence intervals) shown.
limited number of patients, and showing significant variations in performance characteristics.12–14 Furthermore, other studies have not reported the accuracy of SBI when capsule endoscopy is used for indications other than active bleeding (ie, suspected small-bowel Crohn’s disease, anemia of unknown origin, and so forth). In 2003, Liangpunsakul et al12 first reported the accuracy of SBI in 24 patients undergoing WCE. Eighteen studies were performed for occult GI bleeding or iron-deficiency anemia, and 6 for abdominal pain. Overall, the sensitivity of the SBI in the detection of small-bowel lesions with bleeding potential was poor at 25.7%. However, if only actively bleeding lesions in the intestine were considered, the sensitivity of the SBI was significantly better (81.2%). Signorelli et al13 also studied the performance characteristics of the SBI in 100 consecutive patients. The overall sensitivity was poor (40.9%), and it only increased to 60.9% in the group of patients with red blood or active bleeding in the small bowel. D’Halluin et al14 looked at a larger cohort of patients (n ⫽ 156) with obscure GI bleeding. The overall sensitivity of the SBI for detecting low-risk or high-risk lesions in the small bowel still was low (37%), and it increased to only 57% for the detection of active bleeding. The aim of our study was to verify the SBI accuracy in a large cohort of patients (n ⫽ 287), and to examine the SBI performance characteristics according to the indications for the study. Altogether, our findings suggest that when accounting for all possible lesions with bleeding potential in the small bowel, the sensitivity of the SBI may be higher than originally reported (56.4% vs 25%– 41%). More importantly, however, the sensitivity in our study was not improved significantly when the SBI was used only in patients with active bleeding (58.3% vs 60%– 81% in previous studies). This finding has significant clinical relevance, showing that the current version of the SBI cannot be a substitute for a complete and detailed review of the entire WCE study; therefore, it should not serve as the primary interpretation modality by those physicians who do not have time to read the entire report from mouth to cecum. The sensitivity of the SBI in our study was highest in patients undergoing WCE for the evaluation of diarrhea and abdominal pain with a clinical suspicion of small-bowel Crohn’s disease (64%). Previous publications showed significantly lower overall sensitivity values of 25% to 41% in studies performed for a similar indication.12–14 It appears from our study that with an NPV of 80.4%, the use of an SBI to aid in the detection of larger areas of abnormal mucosa in patients with suspected inflammatory bowel disease warrants further investigation. The main limitation of our study was its retrospective design. A large prospective study will be necessary to verify the SBI accuracy and performance characteristics shown by our retro-
March 2008
spective review. Improvements in the existing Given software system could increase the diagnostic yield of the SBI, and thus will make it a reliable screening tool for patients undergoing WCE. Until then, there is no substitution for an expert physician reviewing the study in full. In conclusion, our large retrospective study showed that the performance characteristics of the currently available SBI feature in WCE are suboptimal and insufficient for adequate detection of lesions with bleeding potential. Even in patients with active small intestinal bleeding, the sensitivity of the SBI still is less than 60%, which is lower than previously reported (60%– 81%). However, in patients with suspected small-bowel Crohn’s disease, our findings indicate that the high sensitivity and NPVs of SBI warrant further investigation of this tool as a clinical aid in the detection of large areas of abnormal mucosa. References 1. Iddan G, Meron G, Glukhovsky A, et al. Wireless capsule endoscopy. Nature 2000;405:417. 2. Meron GD. The development of the swallowable video capsule M2A. Gastrointest Endosc 2000;52:817– 819. 3. ASGE Technology Assessment Committee. Wireless capsule endoscopy. Gastrointest Endosc 2002;56:621– 624. 4. Saurin JC, Delvaux M, Vahedi K, et al. Clinical impact of capsule endoscopy compared to push enteroscopy: 1-year follow-up study. Endoscopy 2005;37:318 –323. 5. Hartman D, Schmidt H, Bolz G, et al. A prospective two-center study comparing wireless capsule endoscopy with intraoperative enteroscopy in patients with obscure GI bleeding. Gastrointest Endosc 2005;61:826 – 832. 6. Ell C, Remke S, May A, et al. The first prospective controlled trial comparing wireless capsule endoscopy with push enteroscopy in chronic gastrointestinal bleeding. Endoscopy 2002;34:685– 689. 7. Eliakim R, Fischer D, Suissa A, et al. Wireless capsule video endoscopy is a superior diagnostic tool in comparison to barium follow-through and computerized tomography in patients with suspected Crohn’s disease. Eur J Gastroenterol Hepatol 2003;15: 363–367. 8. Costamagna G, Shah SK, Riccioni ME, et al. A prospective trial comparing small bowel radiographs and video capsule endoscopy for suspected small bowel disease. Gastroenterology 2002; 123:999 –1005.
PERFORMANCE OF THE SUSPECTED BLOOD INDICATOR
301
9. Herreiras JM, Caunedo A, Rodriguez-Tellez M, et al. Capsule endoscopy in patients with suspected Crohn’s disease in negative endoscopy? Endoscopy 2003;35:1–5. 10. Bardan E, Nadler M, Chowers Y, et al. Capsule endoscopy for the evaluation of patients with chronic abdominal pain. Endoscopy 2003;35:688 – 689. 11. Vasquez-Iglesias J, Gonzalez-Conde B, Estevez-Prieto E, et al. A prospective study of COX-2 inhibitors versus nonspecific NSAIDs induced small bowel lesions using video capsule endoscopy. Endoscopy 2003;35:A183. 12. Liangpunsakul S, Mays L, Rex DK. Performance of Given suspected blood indicator. Am J Gastroenterol 2003;98:2676 – 2768. 13. 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. 14. Signorelli C, Villa F, Rondonotti E, et al. Sensitivity and specificity of the suspected blood identification system in video capsule enteroscopy. Endoscopy 2005;37:1170 –1173. 15. Soussan BE, Antonietti M, Herve S, et al. Diagnostic yield and therapeutic implications of capsule endoscopy in obscure gastrointestinal bleeding. Gastroenterol Clin Biol 2004;28:1068 – 1073. 16. Gupta R, Lakhtakia S, Tandan M, et al. Capsule endoscopy in obscure gastrointestinal bleeding—an Indian experience. Indian J Gastroenterol 2006;25:188 –190. 17. Buscaglia J, Carroll C, Daniels J, et al. Benign lymphoid hyperplasia: a rare cause of obscure overt bleeding in an adult. Gastrointest Endosc 2007;66:1248 –1250.
Address requests for reprints to: Jonathan M. Buscaglia, MD, Johns Hopkins Hospital, 1830 E. Monument Street, Room 7100-A, Baltimore, Maryland 21205. e-mail: [email protected]; fax: (410) 9552108. Jonathan Buscaglia, MD, initiated the study design, conducted the data analysis, and prepared the manuscript. Samuel Giday, MD, and Sergey Kantsevoy, MD, PhD, aided in the data analysis and the manuscript preparation. John Clarke, MD, Priscilla Magno, MD, and Elaine Yong, MD, collectively interpreted greater than 60% of the WCE studies, and each edited the manuscript draft. Gerard Mullin, MD, MHS, reviewed all WCE studies, contributed to the study design and data analysis, and edited the manuscript draft.
Performance Characteristics of the Suspected Blood Indicator Feature in Capsule Endoscopy According to Indication for Study JONATHAN M. BUSCAGLIA, SAMUEL A. GIDAY, SERGEY V. KANTSEVOY, JOHN O. CLARKE, PRISCILLA MAGNO, ELAINE YONG, and GERARD E. MULLIN Division of Gastroenterology and Hepatology, Johns Hopkins University School of Medicine, Johns Hopkins Hospital, Baltimore, Maryland
Background & Aims: The suspected blood indicator (SBI) feature of wireless capsule endoscopy (WCE) was developed for rapid screening of intestinal lesions with bleeding potential. Our aim was to assess the accuracy and performance characteristics of the SBI according to the indications for study in a large cohort of patients. Methods: We reviewed collected data on all WCE studies performed at Johns Hopkins Hospital from January 2006 to June 2007. Study indications were as follows: anemia of unknown origin (n ⴝ 53), obscure gastrointestinal bleeding (n ⴝ 112), suspected Crohn’s disease (n ⴝ 122), and other (n ⴝ 4). Concordant and discordant findings between gastroenterologists’ readings and SBI were recorded for each patient. Results: A total of 221 lesions with bleeding potential was detected. The overall sensitivity, specificity, positive predictive value, and negative predictive value for the SBI were 56.4%, 33.5%, 24.0%, and 67.3%, respectively. For actively bleeding lesions, the SBI sensitivity and positive predictive value were only 58.3% and 70%, respectively. The sensitivity was highest (64%) in patients undergoing WCE for suspected Crohn’s disease, with a negative predictive value of 80.4%. The sensitivity was only 58.3% and 41.3%, respectively, in studies performed for obscure gastrointestinal bleeding and anemia. Conclusions: Performance characteristics of the currently available SBI feature in WCE are suboptimal and insufficient to screen for lesions with bleeding potential. Even in patients with active intestinal bleeding, the sensitivity of SBI was less than 60%, which is lower than previously reported. However, in patients with suspected Crohn’s disease, the high sensitivity and negative predictive value of SBI may make it a useful tool for the detection of large areas of abnormal mucosa.
W
ireless capsule endoscopy (WCE) was first developed for advanced imaging of the small intestine.1–3 Although its clinical efficacy is most proven in the diagnosis of obscure gastrointestinal bleeding,4 – 6 it also has been helpful as an adjunct to radiologic studies for patients with suspected Crohn’s disease, celiac disease, small-bowel tumors, anemia of unknown origin, chronic abdominal pain, and other indications.7–11 The WCE system (Given Imaging, Yoqneam, Israel) in combination with RAPID software (Duluth, Georgia) has a suspected blood indicator (SBI) feature that initially was developed to facilitate the screening of possible sites of active bleeding, thereby significantly reducing the reading time spent by a physician-reviewer.12 The SBI function works by identifying redcolored pixels on the viewing screen, and then automatically
annotating the study scroll bar with a red hash mark; thus indicating to the reader that a potential bleeding lesion or other site of pathology may be present. Previously reported studies on the accuracy of the SBI either have been dedicated solely to gastrointestinal bleeding,13 or were limited by a small number of enrolled patients.12,14 The aim of our study was to assess the accuracy and performance characteristics of the SBI according to the indications for the study in a large cohort of patients.
Methods Consecutive patients undergoing WCE at our institution from January 2006 through June 2007 were reviewed for the study. Permission to review patient records was granted by the Johns Hopkins University Institutional Review Board. All patients were asked to refrain from eating or drinking at least 8 hours before swallowing the Given M2A video capsule endoscope. Laxative bowel preparation was not used. Patients were allowed to eat and drink 4 hours after the start of their study. Each WCE study was interpreted by 1 of 5 board-certified/board-eligible gastroenterologists (J.M.B., S.A.G., P.M., J.O.C., and E.Y.). All 5 readers had experience reviewing greater than 50 cases each. Images were reviewed at a speed of 8 to 15 frames per second. Captured thumbnail images and summary reports were re-examined and verified by a separate, boardcertified gastroenterologist (G.E.M.) who had experience with at least 250 cases. There was greater than 95% concordance between the verifying reader (G.E.M.) and each of the 5 initial reviewers. After each study, the interpreting physician was asked to record all endoscopic findings within our WCE database. The following pathologic lesions were considered significant as it pertains to the SBI function of the Given software: ulcers, erosions, arteriovenous malformations, red spots, varices, venous ectasias, blood, and blood clots. Physicians were asked to document whether the SBI accurately predicted the lesions discovered within each patient’s study. Performance characteristics of the SBI were defined by concordance between what was positively sensed by the Given software system, and what was positively detected by the physician-reviewer. Cases in which a significant lesion was found by the reviewer, but not detected by the SBI, also were documented. Those cases in which SBI Abbreviations used in this paper: GI, gastrointestinal; NPV, negative predictive value; PPV, positive predictive value; SBI, suspected blood indicator; WCE, wireless capsule endoscopy. © 2008 by the AGA Institute 1542-3565/08/$34.00 doi:10.1016/j.cgh.2007.12.029
March 2008
PERFORMANCE OF THE SUSPECTED BLOOD INDICATOR
Table 1. Patient Demographics and Study Indications
Indication for WCE
Anemia of unknown origin (n ⫽ 53)
Obscure GI bleeding (n ⫽ 112)
Suspected Crohn’s disease (n ⫽ 122)
Total (%) (n ⫽ 287)
Males Females Mean age, y
25 28 53.7
64 48 58.9
36 86 46.3
125 (43.6) 162 (56.4) 53.0
correctly predicted the findings in one part of the study, but failed to detect the lesions in another part of the study, were recorded as well. The indications for each WCE study also were documented for all patients.
Results A total of 291 patient WCE recordings were studied between January 2006 and June 2007. When the total number of studies was divided according to clinical indications, the single most common indication was suspicion of small-bowel Crohn’s disease in patients with diarrhea and abdominal pain (n ⫽ 122; 41.9%). Obscure gastrointestinal (GI) bleeding was the second most common indication (n ⫽ 112; 38.5%), followed by anemia of unknown origin (n ⫽ 53; 18.2%). Four additional studies were performed for other indications (1.4%). These 4 studies were not included in the analysis, making the final number of WCE studies equal to 287 (Table 1). There were 125 males (43.6%) and 162 females (56.4%), with a mean age of 53 years. A total of 221 pathologic lesions were identified in 287 studies by 5 interpreting physicians (Table 2). Mucosal red spots were the most common lesions recognized (69 of 221; 31.2%), followed by erosions (55 of 221; 24.9%), arteriovenous malformations (41 of 221; 18.6%), ulcers (41 of 221; 18.6%), venous ectasias (11 of 221; 5%), and intestinal varices (1 of 221; 0.5%). There were 15 cases in which blood was recognized, and 12 separate lesions were documented to account for the bleeding: 4 ulcers, 3 erosions, and 5 arteriovenous malformations. In 3 cases (1.4%), blood clots or active bleeding was documented in the small bowel by the interpreting physician, yet no discernable lesion was recognized (Table 2).
299
Table 3. Concordance Between Reader and the SBI Function Anemia of Suspected unknown Obscure GI Crohn’s origin bleeding disease Total Total number of analyzed recordings SBI present, significant lesion present (true positive), n SBI present, significant lesion absent (false positive), n SBI absent, significant lesion absent (true negative), n SBI absent, significant lesion present (false negative), n
53
112
122
287
7
21
16
44
24
55
60
139
12
21
37
70
10
15
9
34
There was a similar proportion of overall patients in the obscure GI bleeding group and the anemia group with positive findings detected by the interpreting physician: 32.1% (36 of 112) and 32.1% (17 of 53), respectively. In the suspected smallbowel Crohn’s disease group, 20.5% (25 of 122) of patients had positive findings. Table 3 highlights the concordance between the reader and the SBI function of the software. Performance characteristics of the SBI did not differ among the 5 different readers. Of a total of 287 studies, 44 patients had a significant lesion with an associated SBI (true positive). In 139 patients, an SBI was present without any discernable disease (false positive). The study was negative in 70 patients without an SBI present (true negative), and 34 patients had a lesion recognized by the phy-
Table 2. Pathologic Lesions With Bleeding Potential Found in 287 WCE Studies
Lesion Ulcers Erosions Arteriovenous malformations Red spots Varices Venous ectasias Blood clots and/or active bleeding Total
Without active bleeding (%)
Actively bleeding (%)
Total (%)
37 (16.7) 52 (23.5) 36 (16.3)
4 (1.8) 3 (1.4) 5 (2.3)
41 (18.6) 55 (24.9) 41 (18.6)
69 (31.2) 1 (0.5) 11 (5.0) —
0 0 0 3 (1.4)
69 (31.2) 1 (0.5) 11 (5.0) 3 (1.4)
206 (93.2)
15 (6.8)
221 (100)
Figure 1. Example of a false-negative SBI in a patient with obscureovert GI bleeding found to have lymphonodular hyperplasia on resection of the terminal ileum.17 PillCam (Given Imaging, Yogneam, Israel)
300
BUSCAGLIA ET AL
CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 6, No. 3
Table 4. Performance Characteristics of the SBI for the Prediction of Pathologic Lesions With Bleeding Potential Indication for WCE
Anemia of unknown origin
Obscure GI bleeding
Suspected Crohn’s disease
Total
Sensitivity Specificity PPV NPV
41.3 (19.4–66.6) 33.3 (19.1–51.1) 22.6 (10.3–41.5) 54.5 (32.7–74.9)
58.3 (40.1–74.0) 27.6 (18.3–39.3) 27.6 (18.3–39.3) 58.3 (40.1–74.0)
64.0 (42.6–81.3) 38.1 (28.6–48.6) 21.1 (12.9–32.2) 80.4 (65.6–90.1)
56.4 (44.7–67.4) 33.5 (27.2–40.4) 24.0 (18.2–31.0) 67.3 (57.3–76.0)
NOTE. Percentages (95% confidence intervals) shown.
sician-reviewer without an associated SBI (false negative, Figure 1). Concordance values based on each of the 3 major indications for the study also are listed in Table 3. Table 4 lists the performance characteristics of the SBI in all 287 WCE studies, as well as performance characteristics according to each clinical indication. The overall sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the SBI were 56.4%, 33.5%, 24.0%, and 67.3%, respectively. In patients undergoing WCE for anemia of unknown origin, the sensitivity, specificity, PPV, and NPV of the SBI were 41.3%, 33.3%, 22.6%, and 54.5%, respectively (Table 4). In patients undergoing WCE for obscure GI bleeding, the sensitivity, specificity, PPV, and NPV of the SBI were 58.3%, 27.6%, 27.6%, and 58.3%, respectively. When performed for the suspicion of smallbowel Crohn’s disease, the sensitivity, specificity, PPV, and NPV were 64.0%, 38.1%, 21.1%, and 80.4%, respectively. Factoring in only those cases with active bleeding, the overall sensitivity of the SBI was 58.3%, with a PPV of 70.0% (Table 5). When accounting for all patients studied, the sensitivity of the SBI was highest (64%) in those undergoing WCE for suspicion of small-bowel Crohn’s disease. The specificity of the SBI was poor for all 3 indications, with the highest value being 38.1% in the suspected Crohn’s group (Table 4). The NPV also was highest in this same group, with a value of 80.4%. The PPV was poor in all groups, with a peak value of 27.6% in patients with obscure GI bleeding.
Discussion Since the invention of capsule endoscopy, it has been used most often and most effectively for the diagnosis of obscure GI bleeding.15,16 More recently, however, its indications have broadened with its usefulness shown in the diagnosis of inflammatory bowel disease and the evaluation of chronic abdominal pain.7–11 Early in its development, the WCE software system (Given Imaging) offered an SBI feature with each of its recordings. This small red line appearing at the top of the time tracing initially was designed as a rapid screening tool, allowing physicians to identify areas of active bleeding quickly.12 Although the SBI feature has been available for more than 5 years, reports of its clinical usefulness are relatively sparse, usually completed on a Table 5. Sensitivity and PPV of the SBI in 15 WCE Studies With Blood Clots or Active Bleeding Sensitivity PPV
58.3 (28.6–83.6) 70.0 (35.4–91.9)
NOTE. Percentages (95% confidence intervals) shown.
limited number of patients, and showing significant variations in performance characteristics.12–14 Furthermore, other studies have not reported the accuracy of SBI when capsule endoscopy is used for indications other than active bleeding (ie, suspected small-bowel Crohn’s disease, anemia of unknown origin, and so forth). In 2003, Liangpunsakul et al12 first reported the accuracy of SBI in 24 patients undergoing WCE. Eighteen studies were performed for occult GI bleeding or iron-deficiency anemia, and 6 for abdominal pain. Overall, the sensitivity of the SBI in the detection of small-bowel lesions with bleeding potential was poor at 25.7%. However, if only actively bleeding lesions in the intestine were considered, the sensitivity of the SBI was significantly better (81.2%). Signorelli et al13 also studied the performance characteristics of the SBI in 100 consecutive patients. The overall sensitivity was poor (40.9%), and it only increased to 60.9% in the group of patients with red blood or active bleeding in the small bowel. D’Halluin et al14 looked at a larger cohort of patients (n ⫽ 156) with obscure GI bleeding. The overall sensitivity of the SBI for detecting low-risk or high-risk lesions in the small bowel still was low (37%), and it increased to only 57% for the detection of active bleeding. The aim of our study was to verify the SBI accuracy in a large cohort of patients (n ⫽ 287), and to examine the SBI performance characteristics according to the indications for the study. Altogether, our findings suggest that when accounting for all possible lesions with bleeding potential in the small bowel, the sensitivity of the SBI may be higher than originally reported (56.4% vs 25%– 41%). More importantly, however, the sensitivity in our study was not improved significantly when the SBI was used only in patients with active bleeding (58.3% vs 60%– 81% in previous studies). This finding has significant clinical relevance, showing that the current version of the SBI cannot be a substitute for a complete and detailed review of the entire WCE study; therefore, it should not serve as the primary interpretation modality by those physicians who do not have time to read the entire report from mouth to cecum. The sensitivity of the SBI in our study was highest in patients undergoing WCE for the evaluation of diarrhea and abdominal pain with a clinical suspicion of small-bowel Crohn’s disease (64%). Previous publications showed significantly lower overall sensitivity values of 25% to 41% in studies performed for a similar indication.12–14 It appears from our study that with an NPV of 80.4%, the use of an SBI to aid in the detection of larger areas of abnormal mucosa in patients with suspected inflammatory bowel disease warrants further investigation. The main limitation of our study was its retrospective design. A large prospective study will be necessary to verify the SBI accuracy and performance characteristics shown by our retro-
March 2008
spective review. Improvements in the existing Given software system could increase the diagnostic yield of the SBI, and thus will make it a reliable screening tool for patients undergoing WCE. Until then, there is no substitution for an expert physician reviewing the study in full. In conclusion, our large retrospective study showed that the performance characteristics of the currently available SBI feature in WCE are suboptimal and insufficient for adequate detection of lesions with bleeding potential. Even in patients with active small intestinal bleeding, the sensitivity of the SBI still is less than 60%, which is lower than previously reported (60%– 81%). However, in patients with suspected small-bowel Crohn’s disease, our findings indicate that the high sensitivity and NPVs of SBI warrant further investigation of this tool as a clinical aid in the detection of large areas of abnormal mucosa. References 1. Iddan G, Meron G, Glukhovsky A, et al. Wireless capsule endoscopy. Nature 2000;405:417. 2. Meron GD. The development of the swallowable video capsule M2A. Gastrointest Endosc 2000;52:817– 819. 3. ASGE Technology Assessment Committee. Wireless capsule endoscopy. Gastrointest Endosc 2002;56:621– 624. 4. Saurin JC, Delvaux M, Vahedi K, et al. Clinical impact of capsule endoscopy compared to push enteroscopy: 1-year follow-up study. Endoscopy 2005;37:318 –323. 5. Hartman D, Schmidt H, Bolz G, et al. A prospective two-center study comparing wireless capsule endoscopy with intraoperative enteroscopy in patients with obscure GI bleeding. Gastrointest Endosc 2005;61:826 – 832. 6. Ell C, Remke S, May A, et al. The first prospective controlled trial comparing wireless capsule endoscopy with push enteroscopy in chronic gastrointestinal bleeding. Endoscopy 2002;34:685– 689. 7. Eliakim R, Fischer D, Suissa A, et al. Wireless capsule video endoscopy is a superior diagnostic tool in comparison to barium follow-through and computerized tomography in patients with suspected Crohn’s disease. Eur J Gastroenterol Hepatol 2003;15: 363–367. 8. Costamagna G, Shah SK, Riccioni ME, et al. A prospective trial comparing small bowel radiographs and video capsule endoscopy for suspected small bowel disease. Gastroenterology 2002; 123:999 –1005.
PERFORMANCE OF THE SUSPECTED BLOOD INDICATOR
301
9. Herreiras JM, Caunedo A, Rodriguez-Tellez M, et al. Capsule endoscopy in patients with suspected Crohn’s disease in negative endoscopy? Endoscopy 2003;35:1–5. 10. Bardan E, Nadler M, Chowers Y, et al. Capsule endoscopy for the evaluation of patients with chronic abdominal pain. Endoscopy 2003;35:688 – 689. 11. Vasquez-Iglesias J, Gonzalez-Conde B, Estevez-Prieto E, et al. A prospective study of COX-2 inhibitors versus nonspecific NSAIDs induced small bowel lesions using video capsule endoscopy. Endoscopy 2003;35:A183. 12. Liangpunsakul S, Mays L, Rex DK. Performance of Given suspected blood indicator. Am J Gastroenterol 2003;98:2676 – 2768. 13. 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. 14. Signorelli C, Villa F, Rondonotti E, et al. Sensitivity and specificity of the suspected blood identification system in video capsule enteroscopy. Endoscopy 2005;37:1170 –1173. 15. Soussan BE, Antonietti M, Herve S, et al. Diagnostic yield and therapeutic implications of capsule endoscopy in obscure gastrointestinal bleeding. Gastroenterol Clin Biol 2004;28:1068 – 1073. 16. Gupta R, Lakhtakia S, Tandan M, et al. Capsule endoscopy in obscure gastrointestinal bleeding—an Indian experience. Indian J Gastroenterol 2006;25:188 –190. 17. Buscaglia J, Carroll C, Daniels J, et al. Benign lymphoid hyperplasia: a rare cause of obscure overt bleeding in an adult. Gastrointest Endosc 2007;66:1248 –1250.
Address requests for reprints to: Jonathan M. Buscaglia, MD, Johns Hopkins Hospital, 1830 E. Monument Street, Room 7100-A, Baltimore, Maryland 21205. e-mail: [email protected]; fax: (410) 9552108. Jonathan Buscaglia, MD, initiated the study design, conducted the data analysis, and prepared the manuscript. Samuel Giday, MD, and Sergey Kantsevoy, MD, PhD, aided in the data analysis and the manuscript preparation. John Clarke, MD, Priscilla Magno, MD, and Elaine Yong, MD, collectively interpreted greater than 60% of the WCE studies, and each edited the manuscript draft. Gerard Mullin, MD, MHS, reviewed all WCE studies, contributed to the study design and data analysis, and edited the manuscript draft.