Aims/hypothesis We compared the effect of biphasic, basal or prandial insulin regimens on glucose control, clinical outcomes and adverse events in people with type 2 diabetes. Methods We searched the Cochrane Library, MEDLINE, EMBASE and major American and European conference abstracts for randomised controlled trials up to October 2008. A systematic review and meta-analyses were performed. Results Twenty-two trials that randomised 4,379 patients were included. Seven trials reported both starting insulin dose and titration schedules. Hypoglycaemia definitions and glucose targets varied. Meta-analyses were performed pooling data from insulin-naive patients. Greater HbA 1c reductions were seen with biphasic and prandial insulin, compared with basal insulin, of 0.45% (95% CI 0.19-0.70, p=0.0006) and 0.45% (95% CI 0.16-0.73, p=0.002), respectively, but with lesser reductions of fasting glucose of 0.93 mmol/l (95% CI 0.21-1.65, p = 0.01) and 2.20 mmol/l (95% CI 1.70-2.70, p<0.00001), respectively. Larger insulin doses at study end were reported in biphasic and prandial arms compared with basal arms. No studies found differences in major hypoglycaemic events, but minor hypoglycaemic events for prandial and biphasic insulin were inconsistently reported as either higher than or equivalent to basal insulin. Greater weight gain was seen with prandial compared with basal insulin (1.86 kg, 95% CI 0.80-2.92, p=0.0006). Conclusions/interpretation Greater HbA 1c reduction may be obtained in type 2 diabetes when insulin is initiated using biphasic or prandial insulin rather than a basal regimen, but with an unquantified risk of hypoglycaemia. Studies with longer follow-up are required to determine the clinical relevance of this finding.