What calprotectin cut-offs should apply for IBD in general practice?

To the Editor We write in response to the recent article by Dhaliwal et al 1 in Frontline Gastroenterology, which discussed the utility of faecal calprotectin (FC) levels in discriminating between inflammatory bowel disease (IBD) and irritable bowel syndrome (IBS). We applaud the authors' investigation of this important topic given the high prevalence of IBS in general practice, the understandable concern of practitioners in reliably excluding inflammatory conditions in patients with IBS and the imperative to avoid unnecessary diagnostic evaluations. The main reported results of the Dhaliwal study support previous findings in the published literature that, at a cutoff of 50 mg/g, FC determination demonstrates adequate sensitivity and specificity to distinguish between IBS and IBD. 2 The authors note, however, that raising the cutoff to 100 mg/g 'does further improve' sensitivity and specificity of the FC test, and improves the negative predictive value (NPV) to 97% from 87%. This statement highlights several practical concerns that we believe are of substantial importance for both general practitioners and medical policy decision-makers when interpreting the extensive literature on the role of FC in differentiating IBS and IBD. First, it is essential for readers to understand which condition (IBS or IBD) is selected as the 'target' when constructing receiver operating characteristic curves for sensitivity and specificity of FC. Although the authors state that '[ performance characteristics] for diagnosing IBD… were calculated at different FC cutoff values', it appears they actually use IBS as the target condition for IBS versus IBD in table 3, as evidenced by the increase in sensitivity and (slight) reduction in specificity produced by raising the cutoff. Using the authors' data and switching the target condition to IBD for an indication of IBS versus IBD at FC 50 μg/g yields a sensitivity of 77.7 and specificity of 87.5, the inverse of what is reported in table 3. Unfortunately, this inconsistency in selecting the target condition is not unusual in the published literature for FC and may contribute to the widespread perception among medical policy decision-makers in the USA that there is a 'lack of consensus for appropriate cutoff values for FC'. 3 Second, we would like to remind readers that, unlike sensitivity and specificity, which are fixed performance characteristics of a test, the NPV and positive predictive value (PPV) of a test are dependent on the prevalence of the target condition in the practitioner's population. 4 Although Dhaliwal et al do …