Head-to-head Comparison of Two Fecal Biomarkers to Screen Children for IBD

Overview

RATIONALE: A substantial proportion of children and teenagers with suspected inflammatory bowel disease (IBD) referred for endoscopy do not have the disease. The investigators designed a clinical decision rule that included a calprotectin stool test to discern which patients require further investigations. The accuracy of this diagnostic strategy is 88.5% with a low risk of missing IBD cases. Although the number of negative endoscopies was reduced after introduction of this strategy, still 22% of the referred children and teenagers underwent an unnecessary invasive test. S100A12 (calgranulin C) is a cytoplasmic protein secreted exclusively by activated neutrophils and this stool marker may be more IBD-specific than calprotectin. OBJECTIVE: To determine whether the specificity of S100A12 is superior to the specificity of calprotectin without sacrificing sensitivity HYPOTHESIS: Inclusion of the calgranulin C stool test will improve the specificity of the screening-strategy.

DESIGN: A prospective diagnostic accuracy study in several outpatient clinics for general paediatrics and several tertiary care hospitals in the Netherlands and Belgium. STUDY POPULATION: Eligible for inclusion are consecutive children and teenagers between 6 and 18 years who consult their pediatrician and have gastro-intestinal symptoms suggestive of IBD. INTERVENTION: Patients will be managed according to a calprotectin-based-referral strategy. Those with an elevated calprotectin level without colon pathogens are considered to have a high probability of IBD and may require referral for endoscopy (the preferred reference standard). Patients with confirmed gastrointestinal infection are advised to have their stools retested. Patients with normal calprotectin levels are considered to have a low probability of IBD and will therefore have a low change to be subjected to endoscopy. In these patients with low probablity of IBD an alternative reference standard may be performed, being clinical follow-up for 6 months. The decision for endoscopy or clinical follow up is up to the clinician's discretion, based on the combination of all symptoms, physical examination, blood results, fecal markers and colon pathogens. Next to calprotectin, also S100A12 will be measured in all stool samples. We will perform a post hoc scenario analysis to compare the test characteristics of both fecal markers. OUTCOME MEASURES: The primary outcome measure is the difference in specificity between FC and S100A12 among the total number of non-IBD patients. We adjusted our previously formulated outcome measure, being the proportion of patients with non-inflammatory conditions among the total number of patient subjected to endoscopy, for two reasons. 1. During an interim analysis in August 2016 the proportion of patients subjected to endoscopy was lower as expected (34% instead 46%). To reach the required amount of patients with the reference standard (endoscopy) we needed to extend the study for several months. 2. Triggered by a recently published paper by Naaktgeboren et al in the BMJ, we realized that our initial design would lead to biased results. Secondary endpoints are the difference in sensitivity among the total number of patients with IBD and the diagnostic accuracy characteristics (sensitivity, specificity, positive predictive value, negative predictive value, area under the curve, best cut-off point) for both markers individually. All diagnostic accuracy characteristics will be calculated with 1) a pre-specified cut-off value based on literature, 2) the best cut-off point calculated with the data from this trial. POWER/DATA ANALYSIS: At the start of our trial we defined a sample size calculation, based on the previously described outcome measure including only patients with endoscopy. Based on a previous cohort study we expected that 46% of the recruited patients would undergo endoscopy. Using a sample size calculation based on independent samples calculated with a Fisher's exact test, we calculated that with 154 patients subjected to endoscopy the study would have 80% power to detect a 50% relative reduction of the primary outcome from 22% false positives with FC to 11% false positives with S100A12, at a one-sided alpha level of 0.05. The total number of patients to be recruited for this diagnostic accuracy study was therefore calculated at 335. In the slipstream of the adjustment of the primary outcome measure, we adjusted our sample size. We now use McNemar's test for paired samples to compare the proportion of concordant and discordant results between FC and S100A12 in all patients with the disease or without the disease. To calculate the new sample size we used a specificity of FC of 0.70, based on recent individual patient data meta-analysis of Degraeuwe and we expected S100A12 would lead to 50% relative improvement. A sample size of 130 subjects achieves 80% power to detect a difference of 0,15 between the two diagnostic tests whose specificities are 0,70 and 0,85. This procedure uses a two-sided McNemar test with a significance level of 0,05. The prevalence of non-IBD in the population is 0,64. The proportion of discordant pairs is 0,23. ETHICAL CONSIDERATIONS: The Medical Ethical Committee of the University Medical Center in Groningen has granted exemption from WMO-approval, as it involves the collection of data generated by routine medical care. After measurement of calprotectin levels and testing for microbial gut pathogens the residual material will be used for the measurement of calgranulin C levels. When patients and their parents give permission, residual feces will be stored for a maximum period of 15 years for future diagnostic research. TIME SCHEDULE: Total running time is 30 months, including 6 months to complete the follow up and 2 months for analysis and reporting.