Unfortunately Syria, Lebanon and Libya did not participate to the network. Palestine was not able to comply with the requirement for the study. Israel could not obtain the ethical permission to use data already collected for a similar study.
Each referral center from Albania, Algeria, Bosnia Herzegovina, Croatia, Egypt, France, Greece, Italy - Messina and Naples - Malta, Montenegro, Morocco, Slovenia, Spain, Tunisia and Turkey provided a list of 50 unselected consecutive CD cases diagnosed in the center by the best locally available methodologies. Using a standardized form, the clinical and laboratory data leading to the diagnosis were collected for each case. The methods used in each center for the serum assay of tTG were recorded, together with their local normal values. We reviewed the pathology procedure used to evaluate small intestinal biopsy specimens and recorded the methods used to evaluate the HLA haplotype. EMA assay was available in 8/16 centers: due to the subjectivity of the method, we could not obtain a standardized comparable set of results.
According to the WGO diagnostic protocol, tTG are an acceptable proxy of the EMA antibodies, especially in countries with limited resources (“The EMA test requires expert observers, and ELISA tests for detecting tTG antibodies should therefore be recommended in settings with low expertise”). For these reasons, in the analysis, we did not considered the few available EMA values.
Based on the facilities and tests available in each center, for each case we estimated the probability of being diagnosed with CD by the new ESPGHAN diagnostic criteria alone, without small intestinal biopsy. For each case we summed up the number of diagnostic criteria met (clinical symptoms, tTG above 10 folds the upper limit of normal and specific HLA). The sum of the criteria determined the final diagnostic score, which thus ranged from 1 to 3. To reduce large tables, countries were often aggregated within their geographical region (Europe, Balkans, North Africa).
Data analysis was performed using the SPSS 16 software package (SPSS Inc., Chicago, IL, USA). Variables were log transformed when required because of their not normal distribution. Student’s t test and ANOVA were used to compare group means; non parametric tests were used when appropriate; p values <0.05 were considered significant.
Institutional review boards at each collection site approved the study in each country (Albania, Algeria, Bosnia Herzegovina, Croatia, Egypt, France, Greece, Italy, Malta, Montenegro, Morocco, Slovenia, Spain, Tunisia and Turkey) (Additional file 1).