Twenty-four studies (n=5,516 participants) were included in the review: 13 diagnostic cohort studies, 10 nested case-control studies and one study that used both designs. The QUADAS items on which studies scored poorly were patient spectrum, appropriate reference standard and differential verification bias. Time between index test and reference standard and withdrawals were poorly described.
Symptoms (11 studies): None of the symptoms (diarrhoea, abdominal pain, blood in stools, weight loss) frequently associated with inflammatory bowel disease showed both good sensitivity and specificity; there was considerable heterogeneity across studies. Other symptoms were evaluated only in a small number of studies. Symptom-based criteria showed better accuracy, but estimates were variable. Kruis criteria (threshold <44) were assessed in three studies that showed high sensitivity (94% to 100%) but variable specificity (17% to 69%). Rome criteria showed lower sensitivity (68% to 89%) and better specificity (50% to 76%).
Blood and faecal tests (13 studies): Accuracy of C-reactive protein (four studies) was very variable. Sensitivity ranged from 55% to 100% and specificity ranged from 42% to 90%. Sensitivity of erythrocyte sedimentation rate (three studies) ranged from 56% to 78% and specificity ranged from 75% to 96%. Calprotectin (nine studies) showed consistently high sensitivity (range 84% to 100%) and specificity (range 71% to 100%), except for two studies that showed lower sensitivity (61% and 64%). Lactoferrin showed good accuracy (three studies). Sensitivity ranged from 78% to 100% and specificity ranged from 75% to 100%.
Abdominal ultrasonography (four studies): Summary sensitivity was 73% (95% CI 65% to 80%) and summary specificity was 95% (95% CI 91% to 97%).
Results of tests evaluated in single studies and subgroup analyses were reported.