The aim of the study was to identify predictors of long-term glycaemic control in Type 1 diabetic patients after participation in an intensified insulin treatment and teaching programme. The study population consisted of 697 Type 1 diabetic patients (mean age 26 +/- 7 (SD) years, duration of diabetes 8 +/- 7 years) who participated in the same structured intensified insulin treatment and teaching programme in 10 hospitals and who were re-examined after 1, 2, and 3 years. Multiple and logistic regression analyses were performed including a set of demographic, disease-related, social, and psychosocial variables as potential predictors. As dependent variables the average HbA1 values during the 3-year follow-up period and a composite variable (average HbA1 values/frequency of severe hypoglycaemia)--dividing patients into three groups with good, moderate or poor metabolic control--were considered. Regression analysis of average HbA1 values revealed significance (p < 0.05) for seven independent predictors in descending order: smoking, age at onset of diabetes, frequency of home blood glucose monitoring, socioeconomic status, diabetes-related knowledge, perceived coping abilities, and sex (R2 (percentage of variation explained by the model) = 17%). In a second regression model, HbA1 values before the intervention programme were added to the model and achieved the highest standardized regression coefficient (0.38), increasing R2 to 29%. In the logistic regression models considering both HbA1 and severe hypoglycaemia as a composite dependent variable, diabetes-related knowledge, HbA1 values before the intervention, smoking, perceived coping abilities, age at onset of diabetes, and C-peptide levels were the strongest predictors of glycaemic control. In conclusion, the relationship between demographic, disease-related, psychosocial, and social variables and metabolic control is complex. Therefore, simplistic concepts of linear causality should be abandoned. In addition to HbA1 values before the intervention, smoking, diabetes-related knowledge, home blood glucose monitoring, age at onset of diabetes, perceived coping abilities and C-peptide levels were the most significant and consistent predictors of glycaemic control.