OBJECTIVE: The goal was to compare the predictive values of the Prometheus Inflammatory Bowel Disease (IBD) Serology 7 (IBD7) panel (Prometheus Laboratories, San Diego, CA) with the predictive values of routine blood tests in a population of children referred for initial evaluation of suspected IBD.
METHODS: Medical records of pediatric patients referred for evaluation of IBD for whom IBD7 testing was performed at Prometheus Laboratories between January 2006 and November 2008 were reviewed. Patients underwent diagnosis by pediatric gastroenterologists on the basis of clinical, radiologic, endoscopic, and pathologic evaluations.
RESULTS: A total of 394 records were identified. We excluded 90 records on the basis of age of >21 years, previous diagnosis of IBD, or unclear diagnosis. The prevalence of IBD in this cohort was 38%. The sensitivity, specificity, positive predictive value, negative predictive value, and κ value for the serological panel were 67%, 76%, 63%, 79%, and 42%, respectively, compared with values for a combination of 3 abnormal routine laboratory test results of 72%, 94%, 85%, 79%, and 47%. The antiflagellin antibody assay, the newest assay added to the panel, had sensitivity of 50% and specificity of 53%. Repeat serological testing failed to produce consistent results for 4 of 10 patients.
CONCLUSION: Despite its recent inclusion of the antiflagellin assay, the IBD7 panel has lower predictive values than routine laboratory tests in pediatric screening for IBD.
- inflammatory bowel disease
- serological testing
- Crohn disease
- ulcerative colitis
- indeterminate colitis
- erythrocyte sedimentation rate
WHAT'S KNOWN ON THIS SUBJECT:
The sensitivity and specificity of previous serological panels in screening for IBD are poor. The more-sensitive anti-CBir1 assay has been included in the IBD7 panel available from Prometheus Laboratories since 2006.
WHAT THIS STUDY ADDS:
Despite inclusion of the anti-CBir1 assay, the IBD7 panel is of little value in screening for pediatric IBD. Repeat serological testing failed to produce consistent results. Routine laboratory tests had higher predictive values.
The diagnosis of inflammatory bowel disease (IBD) is established on the basis of clinical, laboratory, radiologic, endoscopic, and histologic findings. Laboratory testing is usually performed to identify patients at high risk, for whom definitive diagnostic procedures such as endoscopy and colonoscopy are warranted. Determinations of erythrocyte sedimentation rate (ESR), hemoglobin level, and platelet count are routine laboratory tests that assess inflammation; however, the sensitivity and specificity of these tools in predicting IBD vary between 15% and 91% and between 44% and 94%, respectively.1,–,7
A role for serological testing in screening for IBD was suggested by several studies.3,4,6,–,9 The first serological markers used for IBD testing were anti-Saccharomyces cerevisiae antibody (ASCA) and atypical perinuclear antineutrophil cytoplasmic antibody (pANCA). Antibodies to Escherichia coli outer membrane porin (OmpC), Pseudomonas fluorescens-associated sequence I2, and flagellin CBir1 are newer additions to serological evaluations. Values for the individual antibody assays vary considerably among different laboratories.7,8,10,–,12 Moreover, each of these assays has only a modest contribution, because the sensitivity of the individual markers for identifying patients with IBD is low, ranging from 11% to 83%.10,–,19 To maximize the sensitivity and specificity of these serological tests, the individual assays have been included in panels. The most widely used panel in the United States is the IBD Serology 7 (IBD7) panel (Prometheus Laboratories, San Diego, CA), which has been available since 2006. This panel's innovations are the inclusion of the anti-CBir1 assay and the use of a pattern-recognition algorithm. The objective of this study was to compare the predictive value of the Prometheus IBD7 panel with the predictive value of routine blood tests in a population of children referred for initial evaluation of suspected IBD.
The study was approved by the institutional review board at Massachusetts General Hospital (MGH). Medical charts of patients who were seen in the MGH pediatric gastrointestinal unit and for whom IBD7 testing was performed at Prometheus Laboratories between July 2006 and November 2008 were retrospectively reviewed for the following: demographic data, symptoms at presentation, family history of IBD, endoscopic, pathologic, and radiologic findings, complete blood count (CBC), ESR, IBD7 results, and clinical diagnosis. Exclusion criteria were age of >21 years and previous diagnosis of IBD (≥6 months before serological testing). Patients also were excluded if the clinical diagnosis was not known by the time of data analysis (February 1, 2009). Records lacking either CBC or ESR assessed within 1 month of serological testing were included in the statistical analyses for IBD7 but not in the analyses for the routine laboratory predictive values.
Patients were separated into 4 groups, that is, Crohn disease (CD), ulcerative colitis (UC), indeterminate colitis (IC), and non-IBD groups. The diagnoses were made by board-certified, pediatric gastroenterologists on the basis of standard endoscopic, histologic, and radiologic criteria.20 All of the patients with IBD had documented endoscopy/colonoscopy and biopsy results consistent with the diagnosis. Clinicians were not blinded to the serological or routine laboratory test results.
The IBD7 panel was used for evaluation of serological markers. Cutoff values for positive serological marker findings were based on the reference ranges provided by Prometheus Laboratories, as follows: ASCA immunoglobulin A (IgA) level, >20 EU/mL; ASCA immunoglobulin G (IgG) level, >40 EU/mL; anti-OmpC level, >16.5 EU/mL; anti-CBir1 level, >21 EU/mL; pANCA level (enzyme-linked immunosorbent assay), >12 EU/mL. The presence of IBD-specific pANCA was assessed through immunofluorescent antibody and deoxyribonuclease sensitivity testing, as described previously.18
Routine Laboratory Tests
Anemia was assessed on the basis of age-adjusted hemoglobin level standards at the MGH laboratory. The lower limits were as follows: <2 years, 10.5 g/dL; 2 to 12 years, 11.5 g/dL; >12 years, 11.7 g/dL for girls and 13.5 g/dL for boys. Platelet counts of >450 × 109 cells per L were considered to indicate thrombocytosis, and ESR values of >20 mm/hour were considered elevated. A subsequent analysis was performed with reduced thresholds of >350 × 109 cells per L for thrombocytosis and >10 mm/hour for elevated ESR values. These thresholds were established as optimal thresholds for IBD detection in a previous study.5
Results were analyzed with SAS 9.1 (SAS Institute, Cary, NC). To compare the different test results and the clinical diagnosis, we examined sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). Exact binomial confidence intervals (CIs) were calculated. We also used Cohen's κ as a measure of agreement between the test results and the clinical diagnosis (1 equaled perfect agreement). Student's t test was used to determine the significance of the difference in mean antibody titers. P values of <.05 were considered significant.
A total of 404 tests were ordered for 394 patients (Fig 1). Ten tests were repeated. We excluded 19 records because of unknown diagnosis, 60 because of previous IBD diagnosis, and 11 because of age of >21 years. The prevalence of IBD was 37.8% (n = 115); 62 of the patients were diagnosed as having CD, 34 UC, and 19 IC. The distributions of age and gender among the diagnostic groups are noted in Table 1.
The most common presenting symptoms in this cohort were abdominal pain (86%), diarrhea (50%), weight loss (37%), and rectal bleeding (37%). Rectal bleeding was the symptom most associated with IBD, specifically with UC, with a prevalence of 68% in the IBD group, 94% in the UC group, 84% in the IC group, and 50% in the CD group. In the non-IBD group, only 26% of patients had rectal bleeding. The prevalence of diarrhea also was significantly higher in the IBD group (72%) than in the non-IBD group (37%). Weight loss was a prominent feature in the CD group, with a prevalence of 66%, compared with 35% in the UC group and 29% in the non-IBD group.
All of the patients diagnosed as having IBD had positive endoscopic evidence and biopsy pathologic findings consistent with the diagnosis, and 112 (97%) had both ESR and CBC findings obtained within 1 month of serological testing. Imaging of the small bowel was documented in 76 records (65%). Sixty-three percent of patients in the non-IBD group had undergone endoscopy with biopsy, 83% had both CBC and ESR findings available, and 61% had undergone imaging of the small bowel.
Non-IBD Clinical Diagnoses With False-Positive IBD7 Results
The most common diagnoses in the non-IBD group and the number of false-positive serological results per diagnosis are presented in Table 2. False CD results were more common than false UC results. Nine additional patients with false-positive CD serological results received clinical diagnoses of peptic ulcer, diverticulosis, duodenal polyp, dysphagia with autism, osteomyelitis, pilonidal cyst, pneumonia, transiently elevated liver enzyme levels, and vasculitis. One additional patient with a false-positive UC serological result received a clinical diagnosis of spondyloarthropathy.
Routine Laboratory Testing
The prevalence rates of anemia, thrombocytosis (>450 × 109 cells per L), and elevated ESR (>20 mm/hour) in the different diagnostic groups are presented in Fig 2. At least 1 abnormal result was found for 72% of patients with IBD, and ≥2 were found for 49%. In the non-IBD group, 25% of patients had ≥1 abnormal result and 6.3% had ≥2 abnormal values. When reduced thresholds (platelet counts of >350 × 109 cells per L and ESR values of >10 mm/hour) were examined, ≥1 abnormal result was found for 95% of patients in the IBD group and 69% of patients in the non-IBD group. At least 2 abnormal results were found for 63% of patients in the IBD group and 23% of patients in the non-IBD group.
The mean values for the individual markers in the different groups are presented in Table 3. Anti-CBir1 was the most-sensitive and least-specific marker. Of the 189 patients in the non-IBD group, 89 (47%) had positive anti-CBir1 findings. The mean anti-CBir1 titer in the non-IBD group was significantly higher than both the reference level and the mean titer in the UC group (P < .05).
All except 2 of the 48 patients with IBD with positive pANCA findings had colonic involvement; 27 of them had positive deoxyribonuclease sensitivity results for the IBD-specific pANCA. In the deoxyribonuclease-positive group, 16 of the patients had UC, 7 had CD, and 4 had IC. In the non-IBD group, 15 of the patients (8%) had positive deoxyribonuclease sensitivity findings. Their diagnoses were varied and did not necessarily involve the colon. The other serological markers did not correlate with the site of disease involvement.
Twenty-one patients in the IBD group who had positive ASCA IgA findings had CD. One additional patient with positive ASCA IgA findings was diagnosed as having UC. All 16 patients in the IBD group who had positive ASCA IgG findings had CD. In the non-IBD group, 6 patients had positive ASCA IgA findings and 4 had positive ASCA IgG findings.
The IBD7 panel results were positive for 77 (67%) of 115 subjects with IBD and 46 (24%) of 189 patients in the non-IBD group. The sensitivity, specificity, NPV, and PPV are presented in Table 4, in comparison with the predictive values for routine laboratory tests. Exclusion from analysis of 34 records with incomplete routine blood testing did not change the sensitivity and specificity of the serological panel (N = 270; sensitivity: 67%; specificity: 76%). Reduction of the ESR and thrombocytosis thresholds increased the sensitivity to 95% (95% CI: 91%–99%) and reduced the specificity to 77% (95% CI: 71%–84%). The NPV was 95% (95% CI: 92%–99%), and the PPV was 75% (95% CI: 68%–82%).
A reliable NPV could be beneficial in excluding IBD in cases with a low index of suspicion. To investigate whether the panel could be used reliably for this purpose, an additional analysis was conducted with a subset of patients determined to represent a low-risk group because of the absence of rectal bleeding, which was the clinical symptom most associated with IBD (Table 4). This subset consisted of 190 patients, with a 22% prevalence of IBD. A total of 166 of these patients also had routine laboratory test results available. In this group as well, the predictive values of routine laboratory test results were higher than those of serological test results, similar to the results for the total cohort (sensitivity: 70%; specificity: 93%; PPV: 67%; NPV: 89%). Reduced thrombocytosis and ESR thresholds yielded sensitivity of 93% (95% CI: 84%–100%), specificity of 65% (95% CI: 68%–83%), and NPV of 93% (95% CI: 85%–100%). Exclusion from analysis of 24 records with incomplete routine blood test results did not significantly change the sensitivity and specificity of the serological panel (N = 166; sensitivity: 63%; specificity: 79%).
Ten patients underwent repeat IBD7 testing, for reasons that were not clear. For those patients, only the results of the first test were included in the statistical analyses. Four of the 10 patients had changes in either overall IBD7 results or individual antibody titers (Table 5).
A screening test for the initial evaluation of pediatric IBD should be reliable, inexpensive, and minimally invasive. Standard screening evaluations often include CBCs and markers of inflammation such as ESRs or C-reactive protein levels. The sensitivity and specificity of routine laboratory evaluations in the current study were comparable to those reported in previous studies.1,–,7
Serological biomarkers have been associated with IBD disease phenotype, and assays are commonly used in clinical practice as screening tests. The previous IBD serological panel marketed by Prometheus was the IBD First Step and Confirmatory System, which was based on ASCA IgG, ASCA IgA, anti-OmpC, and pANCA assays. The sensitivity and specificity of that panel in pediatric IBD screening were found by Sabery and Bass4 to be 60% and 92%, respectively. These values were compared with those for anemia and/or increased ESR, which were more sensitive (83%) and specific (96%).4 The addition of anti-CBir1 is one of the innovations in the IBD7 panel. This antibody was reported to identify children with IBD who were not identified with other serological assays.21,–,23
Despite low reported predictive values, a simple categorical result produced through pattern recognition of multiple antibody assays is appealing to clinicians. However, the cost of a CBC and ESR test is less than $100, compared with the cost of the IBD7 panel, which is approximately $450.
Previous studies used single-, double-, or multiple-antibody panels and demonstrated sensitivities ranging between 47% and 84%,3,4,6,–,8,12,14,–,19,21,23,24 comparable to the sensitivity of 67% found in the current study. One of the surprising findings in the current study was a low serological specificity of 76%, in comparison with specificities between 72% and 100% reported previously.3,4,6,–,8,12,14,–,19,21,23,24 The reduced specificity could be attributed to the inclusion of anti-CBir1 in the panel, because results for this antibody were positive for nearly one-half of the patients in the non-IBD group. A higher anti-CBir1 threshold for pediatric patients might improve specificity.
The reported correlation between involvement of the small bowel and anti-CBir1 and ASCA seroreactivity18,25,–,27 was not reproduced in the current study, possibly because of the lack of documented imaging of the small bowel in a large proportion of the records. However, a correlation was found between pANCA levels and disease involvement of the colon in both CD and UC. These findings are in accord with findings from previous studies.7,18,22,24,28
A coincidental observation was the change in antibody titers over short periods of time for 4 of 10 patients who underwent repeat IBD7 testing. Whereas routine blood tests serve as measures of disease activity and responses to treatment, ASCA and pANCA levels have been reported to remain stable in the presence of disease progression and in response to treatment.18,25,26,28,29 Although the aim of this study was not to evaluate the variation in serological titers over time, the changes in IBD7 results and antibody titers in 4 of 10 cases stress the limited reliability and reproducibility of this test.
A limitation of this study is its retrospective design. There were no predefined criteria for ordering of serological assessments. In cases in which the diagnosis was obvious, serological assessments might not have been ordered, which would exclude those patients from analysis. Furthermore, clinicians were not blinded to test results, which might have influenced the clinical diagnoses. Endoscopy and colonoscopy were not performed for all of the patients who were not ultimately diagnosed as having IBD. It could be argued that, if endoscopy or colonoscopy had been performed, then the clinical diagnoses might have been different.
The IBD7 panel is of little value as a screening tool for pediatric IBD, despite inclusion of the anti-CBir1 assay. The lower cost and higher sensitivity, specificity, PPV, and NPV of CBC and ESR testing support the use of such testing for the selection of patients to undergo additional diagnostic procedures.
This study was supported by an Advanced IBD Fellowship grant provided by Martin Schlaff.
We acknowledge Jeffrey Biller, Tim Buie, Robert Fusunyan, Stephan Hardy, Aubrey Katz, Gary Russell, Marc Salvatore, and Lauren Drake for their support or provision of access to their patients' serological reports.
- Accepted February 5, 2010.
- Address correspondence to Shira Benor, MD, Dana Hospital for Children, Tel Aviv Sourasky Medical Center, Department of Pediatrics, 6 Weizman St, Tel Aviv, Israel 64239. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
- IBD =
- inflammatory bowel disease •
- ESR =
- erythrocyte sedimentation rate •
- ASCA =
- anti-Saccharomyces cerevisiae antibody •
- pANCA =
- perinuclear antineutrophil cytoplasmic antibody •
- OmpC =
- Escherichia coli outer membrane porin •
- IBD7 =
- Inflammatory Bowel Disease Serology 7 •
- MGH =
- Massachusetts General Hospital •
- CBC =
- complete blood count •
- CD =
- Crohn disease •
- UC =
- ulcerative colitis •
- IC =
- indeterminate colitis •
- NPV =
- negative predictive value •
- PPV =
- positive predictive value •
- CI =
- confidence interval •
- IgG =
- immunoglobulin G •
- IgA =
- immunoglobulin A
- Mack DR,
- Langton C,
- Markowitz J,
- et al
- Beattie RM,
- Walker-Smith JA,
- Murch SH
- Sabery N,
- Bass D
- Cabrera-Abreu JC,
- Davies P,
- Matek Z,
- Murphy MS
- Jaskowski TD,
- Litwin CM,
- Hill HR
- Bossuyt X
- Anand V,
- Russell AS,
- Tsuyuki R,
- Fedorak R
- Zholudev A,
- Zurakowski D,
- Young W,
- Leichtner A,
- Bousvaros A
- Papp M,
- Altorjay I,
- Norman GL,
- et al
- Bousvaros A,
- Antonioli DA,
- Colletti RB,
- et al
- Romano C,
- Famiani A,
- Gallizzi R,
- Comito D,
- Ferrau' V,
- Rossi P
- Giaffer MH,
- Clark A,
- Holdsworth CD
- Copyright © 2010 by the American Academy of Pediatrics