Twenty trials were included in the review (n=5,981 patients listed in individual trials but reported as totalling 6,178 in text and table). Eighteen trials were parallel design (n=5,601 patients) and two were crossover design (n=380 patients). Ten trials had adequate randomisation, 18 trials had adequate levels of drop-outs and 15 trials used intention-to-treat analyses.
Long-acting insulin analogues significantly reduced glycated haemoglobin (HbA1c) compared with neutral protamine Hagedorn human insulin (SMD -0.07, 95% CI -0.13 to -0.01%; 18 trials). Detemir and glargine insulin analogues did not significantly differ in their effect on HbA1C.
Detemir was associated with a significantly smaller increase in BMI trial endpoint compared with neutral protamine Hagedorn human insulin (SMD 0.26 kg/m2, 95% CI 0.06 to 0.47; nine trials). Analyses could not be carried out for the impact of glargine on BMI.
Long-acting insulin analogues significantly reduced the risk of severe hypoglycaemia (OR 0.73, 95% CI 0.60 to 0.89; 15 trials) and nocturnal hypoglycaemia (OR 0.69, 95% CI 0.55 to 0.86; 13 trials), but not overall hypoglycaemic events, compared with neutral protamine Hagedorn human insulin.