It was proposed that Stevens-Johnson syndrome and toxic epidermal necrolysis differed from erythema multiforme majus by the pattern and localization of skin lesions. To evaluate the validity of this clinical separation. Case-control study. Active survey from 1989 to 1995 of 1800 hospital departments in Europe. A total of 552 patients and 1720 control subjects. Cases were sorted into 5 groups (erythema multiforme majus, Stevens-Johnson syndrome, Stevens-Johnson syndrome-toxic epidermal necrolysis overlap, toxic epidermal necrolysis, and unclassified erythema multiforme majus or Stevens-Johnson syndrome) by experts blinded as to exposure to drugs and other factors. Etiologic fractions for herpes and drugs obtained from case-control analyses were compared between these groups. Erythema multiforme majus significantly differed from Stevens-Johnson syndrome, overlap, and toxic epidermal necrolysis by occurrence in younger males, frequent recurrences, less fever, milder mucosal lesions, and lack of association with collagen vascular diseases, human immunodeficiency virus infection, or cancer. Recent or recurrent herpes was the principal risk factor for erythema multiforme majus (etiologic fractions of 29% and 17%, respectively) and had a role in Stevens-Johnson syndrome (etiologic fractions of 6% and 10%) but not in overlap cases or toxic epidermal necrolysis. Drugs had higher etiologic fractions for Stevens-Johnson syndrome, overlap, or toxic epidermal necrolysis (64%-66%) than for erythema multiforme majus (18%). Unclassified cases mostly behaved clinically like erythema multiforme. This large prospective study confirmed that erythema multiforme majus differs from Stevens-Johnson syndrome and toxic epidermal necrolysis not only in severity but also in several demographic characteristics and causes.