The purpose of the present retrospective study was to learn the long-term outcome of oral premalignant lesions, leukoplakia and erythroplakia, with or without surgical intervention and to relate the outcome to factors supposed to be significant for malignant development including clinical type, demarcation, size, site, presence of epithelial dysplasia, smoking and surgery. A total of 269 lesions in 236 patients were included. Ninety-four lesions were surgically removed, 39 lesions (41%) being homogenous and 46 (49%) non-homogenous leukoplakias whereas nine (5%) were erythroplakias. Seventy-three percent of the lesions were associated with tobacco habits. The mean size of the lesions was 486 mm(2), and 71% of the lesions showed a degree of epithelial dysplasia. After excision the defects were closed primarily by transposition of mucosal flaps or they were covered by free mucosal or skin grafts. A few defects were left for secondary healing. After surgical treatment the patients were followed (mean 6.8 yrs, range 1.5-18.6 yrs), and new biopsies taken in case of recurrences. One hundred and seventy five lesions had no surgical intervention, 149 lesions (85%) being homogenous and 20 (11%) non-homogenous leukoplakias, and 6 (3%) erythroplakias. Eighty-one percent of the lesions were associated with smoking. The mean size of the lesions was 503 mm(2) and 21 of the lesions (12%) exhibited epithelial dysplasia. Sixty-five lesions were not biopsied. These patients were also followed (mean 5.5 yrs, range 1.1-20.2 yrs), and biopsies taken in case of changes indicative of malignant development. All patients were encouraged to quit smoking and candidal infections were treated. The possible role of different variables for malignant development was estimated by means of logistic regression analysis. Following surgical treatment 11 lesions (12%) developed carcinoma after a mean follow-up period of 7.5 yrs. Non-homogenous leukoplakia accounted for the highest frequency of malignant development, i.e. 20%, whereas 3% of the homogenous leukoplakias developed carcinomas. Surgically treated lesions with slight, moderate, severe and no epithelial dysplasia developed carcinoma with similar frequencies, i.e. 9-11%. Without surgical intervention 16% of the 175 lesions disappeared whereas seven lesions (4%) developed carcinoma after a mean observation period of 6.6 yrs. The highest frequency of malignant development (15%) was seen for non-homogenous leukoplakias, this figure being 3% for homogenous leukoplakias. Fourteen percent of lesions with slight epithelial dysplasia developed malignancy and 2% of lesions with no dysplasia showed malignant transformation. Logistic regression analysis showed a seven times increased risk (OR = 7.0) of non-homogenous leukoplakia for malignant development as compared with homogenous leukoplakia and a 5.4 times increased risk for malignant development for lesions with a size exceeding 200 mm(2). No other examined variables including presence of any degree of epithelial dysplasia, site, demarcation, smoking and surgical intervention were statistically significant factors for malignant development.