Diagnostic accuracy of faecal calprotectin in a symptom-based algorithm for early diagnosis of inflammatory bowel disease adjusting for differential verification bias using a Bayesian approach

Abstract

Background Diagnostic delay in IBD is a major problem and diagnosis is frequently arrived when irreversible damage has already occurred. This study evaluated accuracy of faecal calprotectin (fCAL) integrated with diagnostic criteria for early diagnosis of IBD in a primary care setting. Methods General practitioners (GPs) were trained to recognize alarm symptoms for IBD classified as major and minor criteria. Fulfilment of one major or at least two minor criteria was followed by free fCAL testing and a visit by an IBD specialist and follow-up over 12 months. All patients with positive fCAL testing, i.e., ≥70 μg/g underwent colonoscopy. The diagnostic accuracy of fCAL was estimated after adjusting for differential-verification bias following a Bayesian approach. Results Thirty-four GPs participated in the study and 133 patients were tested for fCAL between July 2016 and August 2017. Positivity of fCAL was seen in 45/133 patients (34%) and a final IBD diagnosis was made in 10/45 (22%). According to the threshold of 70 μg/g, fCAL achieved a sensitivity of 74.8% (95%CI: 39.10–96.01%), a specificity of 70.4% (95%CI: 61.76–78.16%) and an overall diagnostic accuracy of 70.6% (95%CI: 61.04–78.37%). As for prognostic accuracy, despite positive predictive value being low, 21.9% (95%CI: 11.74–35.18%), the negative predictive value was definitely higher: 96.2% (95%CI: 84.96–99.51%). Conclusions fCAL with a threshold set at 70 μg/g seems to represent a potentially reliable negative test to be used in primary care settings for patients with symptoms suggestive of IBD.