One hundred and fifty-two patients in whom a diagnosis of dermatitis herpetiformis was made at St John's Hospital for Diseases of the Skin, London, during 1950-85, were followed from the date of diagnosis to the end of 1989 for mortality, and from 1971, or the date of diagnosis if later, to 1986 for cancer incidence. Thirty-eight deaths occurred under the age of 85, slightly fewer than expected on the basis of national general population rates [standardized mortality ratio (SMR) = 87; 95% confidence interval (CI) 61-119]. All-cause mortality was somewhat lower in patients who had followed a gluten-free diet (SMR = 51; 17-120) than in those who had not (SMR = 97; 66-136), but the difference in SMRs was not significant (P = 0.3). Cancer mortality was non-significantly below expectations from national rates (SMR = 72; 31-142), but cancer incidence was significantly increased [standardized registration ratio (SRR) = 394; 180-749]. No particular cancer site accounted for the cancer incidence excess. One death occurred from cancer of the small intestine (SMR = 4953, P = 0.04), and one lymphoma was incident (SRR = 1555, P = 0.12). Increased risks of these malignancies have previously been found to be associated with coeliac disease (which is present in many patients with dermatitis herpetiformis), and with dermatitis herpetiformis, respectively. Mortality from ischaemic heart disease (IHD) was significantly below national rates (SMR = 37; 95% CI 12-86), and was similar in patients who had followed a gluten-free diet and those who had not.(ABSTRACT TRUNCATED AT 250 WORDS)