OBJECTIVES: Patients with type 2 diabetes are known to make increased use of health-care resources, but the impact of specific macro and microvascular complications on costs is unclear. Here we use regression based methods to estimate the immediate and long-term impact of six diabetes-related complications on hospital costs, using data from the UKPDS, a large (n = 5102) and long-term (median duration 10.3 years) clinical trial. METHODS: Data on the occurrence and precise timing of pre-defined diabetes-related complications, and on all hospitalisations with associated specialties, lengths of stay, and procedures, were collected routinely for all patients during the trial. Panel data regression analysis was used to estimate the immediate impact (i.e. in the year event occurred) and long-term impact (i.e. in each subsequent year) of the following six diabetes-related complications on hospital costs: fatal and non fatal myocardial infarction (n = 828); fatal and non-fatal stroke (n = 271); heart failure (n = 166); angina (n = 319); blindness in one eye (n = 166); amputation (n = 67). Hospital costs were calculated using national average specialty-specific costs per inpatient day, expressed in 1999 £s UK. RESULTS: All six diabetes-related complications had a statistically significant impact on hospital costs. In the year in which the complication occurred, diabetes-related complications were associated with increased hospital costs ranging from £995 for loss of sight in one eye to £5478 for an amputation. In subsequent years the annual magnitude of this effect ranged from 13% to 33% of the costs in the initial year. CONCLUSIONS: Regression analysis on a large and well-validated, patient-specific data set yields plausible empirically based estimates of the hospital cost consequences of diabetes-related complications. These will be of use to other economists and health service researchers, particularly those interested in assessing the costs of diabetes and the cost-effectiveness of interventions within a modeling framework.