OBJECTIVE: To study the relationship between symptoms, metabolic control and insulin treatment in diabetes mellitus. DESIGN: A cross-sectional questionnaire study of diabetic patients, and a 1-year follow-up study of poorly regulated patients prescribed insulin. Regulation criteria were predefined and the patients acted as their own controls. SETTING: Five primary care practices in Nordland county, Norway. PATIENTS: 111 patients in the cross-sectional study, with 18 of them participating in the follow-up study. MAIN OUTCOME MEASURES: Symptom scores and sum scores, based on five general symptoms--dizziness, depression, fatigue, thirst and dry mouth--and on two urinary symptoms--urinary frequency and urination during the night. RESULTS: Poorly regulated diabetic patients had a higher sum score for general symptoms than better regulated patients (6.1 vs. 4.2, p = 0.078 with Wilcoxon two-sample rank sum test). With parametric analysis, the difference was significant, and remained so when adjusted for age and sex. Females reported more symptoms than males. Symptom relief with insulin was not statistically significant; however, there was a slight but consistent tendency towards less symptoms with better regulation. There was no correlation with HbA1 values at any time during the follow-up study. Only one patient wanted to stop taking insulin after 1 year. CONCLUSION: Better metabolic control and better general well-being seem to be achieved in many cases when poorly regulated patients with type 2 diabetes mellitus begin insulin treatment. The general practitioner should be cautious in promising relief of specific symptoms.