BACKGROUND: Estimates of the diagnostic performance of serologic testing and HLA-DQ typing for detecting celiac disease have mainly come from case-control studies. OBJECTIVE: To define the performance of serologic testing and HLA-DQ typing prospectively. DESIGN: Prospective cohort study. SETTING: University hospital. PATIENTS: Patients referred for small-bowel biopsy for the diagnosis of celiac disease. INTERVENTIONS: Celiac serologic testing (antigliadin antibodies [AGA], antitransglutaminase antibodies [TGA], and antiendomysium antibodies [EMA]) and HLA-DQ typing. MEASUREMENTS: Diagnostic performance of serologic testing and HLA-DQ typing compared with a reference standard of abnormal histologic findings and clinical resolution after a gluten-free diet. RESULTS: Sixteen of 463 participants had celiac disease (prevalence, 3.46% [95% CI, 1.99% to 5.55%]). A positive result on both TGA and EMA testing had a sensitivity of 81% (CI, 54% to 95.9%), specificity of 99.3% (CI, 98.0% to 99.9%), and negative predictive value of 99.3% (CI, 98.0% to 99.9%). Testing positive for either HLA-DQ type maximized sensitivity (100% [CI, 79% to 100%]) and negative predictive value (100% [CI, 98.6% to 100%]), whereas testing negative for both minimized the negative likelihood ratio (0.00 [CI, 0.00 to 0.40]) and posttest probability (0% [CI, 0% to 1.4%]). The addition of HLA-DQ typing to TGA and EMA testing, and the addition of serologic testing to HLA-DQ typing, did not change test performance compared with either testing strategy alone. LIMITATION: Few cases of celiac disease precluded meaningful comparisons of testing strategies. CONCLUSIONS: In a patient population referred for symptoms and signs of celiac disease with a prevalence of celiac disease of 3.46%, TGA and EMA testing were the most sensitive serum antibody tests and a negative HLA-DQ type excluded the diagnosis. However, the addition of HLA-DQ typing to TGA and EMA testing, and the addition of serologic testing to HLA-DQ typing, provided the same measures of test performance as either testing strategy alone.