Comparison of effectiveness of premixed insulin with long-acting insulin in diabetes: evidence from real-world cohort studies—systematic review and meta-analyses

The purpose of this research is to perform a systematic review and meta-analysis to compare effectiveness of premixed insulin (PM) and long-acting insulin (LA) based on cohort studies. Embase, Pubmed, Web of Science, Clinical.​gov, and Opengray were searched. We included all cohort studies comparing the effectiveness of PM with LA on intermediate- or long-term outcomes. The weighted mean difference (WMD) was recorded for intermediate outcomes, and the risk ratio (RR) estimates were combined and weighted to produce pooled RR for long-term outcomes. Subgroup analysis was conducted by different insulin type and population. Eight studies involving 95,415 patients were included, in which six studies were used for quantitative analysis. Follow-up duration ranged from 6 months to 5 years. The meta-analysis indicated a comparable effect of PM and LA on the mean reduction in hemoglobin A1c (HbA1c) (WMD = − 0.03% [95% CI − 0.12–0.07]). Less fasting plasma glucose (FPG) declines were noticed with PM than LA (0.11 mmol/L [95% CI 0.03–0.18]). There was not statistically significant difference for post-prandial glucose (WMD = − 0.61 mmol/L [95% CI [− 1.49–0.27]). PM was associated with further weight gain of 0.67 kg [95% CI 0.60–0.74] and more insulin consumption (WMD = 11.72 U/day [95% CI 8.02–15.43]. The rates of hypoglycemia events were low but varied considerably. There was no significant difference in the effectiveness of PM and LA in real-world setting on HbA1c reduction, although PM was related with less FPG reductions, slight weight gain, and more insulin consumptions. Future real-world studies should concern about balanced oral hypoglycemic agents regimens between insulin cohorts and other potential confounding factors and assess relevant complications and death rate.