OBJECTIVES: In a type 2 diabetes population, we calculated retrospectively the hospital costs according to different body mass index (BMI). METHODS: We studied a cohort (n = 2263) of type 2 diabetes patients included up to February 2002 and having undergone at least two visits. The cohort was stratified in 4 BMI groups calculated on the mean measured body weight: 18.5-24.9 kg/m2 (normal), 25-29.9 kg/m2 (overweight), 30-34.9 kg/m2 (obese) and over 35 kg/m2 (very obese). Annual hospital costs per patient for each BMI group were calculated by combining observed Health care resource utilization and official costs data of the Diagnosis Related Group (DRG) system. Health care resource considered and costs were related to management of diabetes (in-hospitalization, day care hospitalization) and complications (myocardial infarction, angina pectoris, transient ischaemic accident and stroke). RESULTS: Number of patients per group was 399, 914, 625, and 325 for normal, overweight, obese, and very obese respectively. Total annual hospital costs (in Euros) per patient were 2357, 2645, 2779 and 2972 for normal, overweight, obese and very obese respectively. In each group, the relative weight of in-hospitalization cost for management of diabetes were 51.2 %, 51 %, 58.1 % and 64.4 % for normal, overweight, obese and very obese respectively. In each group, the relative weight of cost of complications were 8 %, 10.8 %, 7.3 % and 8.9 % for normal, overweight, obese and very obese respectively.CONCLUSIONS:We confirm obese diabetic patients (BMI above 25 kg/m2) generate more hospital costs than non-obese diabetic patients. According to the increase of BMI, we note a progression of in-hospitalization costs for management of type 2 diabetes whereas the proportion of day care hospitalization costs is decreasing. Higher is the BMI, higher are costs dedicated to diabetes care. As opposite, cost of complications remains stable whatever the BMI. Prevention of complications in patient with high BMI may be an explanation to this pattern of care.