BACKGROUND: Seronegative coeliac disease is poorly defined. AIMS: To study clinical phenotypes and long-term outcomes of seronegative coeliac disease in a multicentre cohort over 20 years. METHODS: Seronegative coeliac disease was diagnosed in HLA-DQ2/DQ8-positive patients with villous atrophy (VA), negative IgA endomysial (EmA), tissue transglutaminase (tTG) and deamidated-gliadin antibodies (DGP), clinical and histological response to a gluten-free diet (GFD), and no alternative causes for VA. In patients with IgA deficiency, coeliac disease was diagnosed through VA, positive IgG EmA/tTG/DGP and clinical/histological response to a GFD (coeliac disease+IgAd). Patients with seropositive coeliac disease served as controls. RESULTS: Of 227 patients previously diagnosed with seronegative coeliac disease, true seronegative coeliac disease was confirmed in 84, coeliac disease+IgAd in 48, and excluded in 55. Lack of follow-up duodenal biopsy precluded diagnosing seronegative coeliac disease in 40 patients. 2084 patients with seropositive coeliac disease served as controls. True seronegative coeliac disease had more severe symptoms at diagnosis and a higher risk of complications (HR 10.87, 95% CI 6.11-19.33, P < 0.001) and mortality (HR 2.18, 95% CI 1.12-4.26, P < 0.01) than seropositive coeliac disease. There were no differences between true seronegative coeliac disease and coeliac disease+IgAd. On multivariate analysis, age at diagnosis, lack of clinical response to a GFD, true seronegative coeliac disease, coeliac disease+IgAd, and classical presentation predicted complications. Age at diagnosis, complications and absence of clinical response to a GFD predicted mortality. CONCLUSIONS: Seronegative coeliac disease has a more aggressive disease phenotype than seropositive coeliac disease. These data argue against over-reliance on serology for the diagnosis of coeliac disease and support a strict clinical and histologic follow-up in seronegative coeliac disease.