Aim Faecal calprotectin (FC) is increasingly used as a marker of intestinal inflammation, especially in diagnosis of inflammatory bowel disease (IBD). A meta-analysis (Rheenen et al. BMJ 2010) evaluated adult and paediatric studies to October 2009 and concluded that FC has good discriminative power to safely exclude IBD. However, the seven paediatric studies only included a total of 226 paediatric IBD (PIBD) patients, therefore small study effect may have existed. We aimed to determine the usefulness of FC at presentation in children with suspected IBD by evaluating all the available literature.

Methods A search was performed with keywords relating to IBD and calprotectin in several electronic resources from 1966 to November 2011. A hand search of articles was also performed, drawn from reference lists and meeting abstracts. Inclusion criteria were studies that reported FC levels prior to the endoscopic investigation of IBD in children less than 18 years old. There was no English language restriction. Each study was evaluated using the QUADAS tool and a meta-analysis of all included studies was performed using RevMan (v 5.1) and HSROC (R package). Pooled sensitivity and specificity was generated using a random effects model.

Results A total of 73 papers were identified during the initial search. All were reviewed but only 8 met the inclusion criteria (6 prospective and 2 retrospective case-control studies). Two studies in the original BMJ meta-analysis (Bunn 2001, Kolho 2006) were excluded as both the IBD and control groups did not represent children undergoing primary investigation for suspected IBD. In total the studies presented FC levels at presentation in 394 IBD patients and 353 non-IBD controls. Pooled specificity and sensitivity for the diagnostic utility of FC during the investigation of suspected PIBD were 0.998 (95% CI 0.960-1.000) and 0.665 (95% CI 0.537-0.820) respectively.

Conclusions FC has a high specificity but a modest specificity for the diagnosis of IBD in the paediatric population. The inclusion of two new larger studies and removal of two studies from the original meta-analysis led to an increase in the pooled sensitivity and reduction in pooled specificity. Furthermore, it is evident from other studies that use of a higher normal range (i.e. >200 ug/g) would significantly increase this specificity and therefore further work is required that report multiple cut off levels to truly appreciate the value of this test.