IMPROVE (TM) is an open-label, multinational, non-randomised, 26-week observational study designed to evaluate the safety and effectiveness of biphasic insulin aspart 30 (BIAsp 30) in routine clinical practice. Here, we report data for patients switching to BIAsp 30 from human premixed insulin. Patients (n = 3856) with type 2 diabetes previously receiving human premixed insulin with or without oral antidiabetic drugs were eligible for inclusion. Demographic data, efficacy end-points (HbA(1c), fasting blood glucose and postprandial blood glucose) and safety end-points (serious adverse drug reactions, hypoglycaemia and adverse events) were collected at baseline and final visit. A subgroup analysis of mean dose change was also undertaken. Switching patients to BIAsp 30 resulted in significant improvements in glycaemic control combined with a reduced risk of hypoglycaemia. Patients who reached the HbA(1c) target (< 7%) had shorter diabetes duration, lower HbA(1c) at baseline and needed less insulin. Over 30% of patients were able to reach this target without experiencing hypoglycaemia over the 26-week period. Compared with asymmetric dose switching, unit-for-unit switching resulted in the highest proportion of patients reaching HbA(1c) target and incurred the least amount of dose titration. A unit-for-unit switch is the most effective as well as the simplest approach when transferring patients from biphasic human insulin 30 to BIAsp 30.