Background: Randomized controlled trials
(RCTs) of antidiabetic agents started in the 1960s. Updated meta-analyses of RCTs investigating glucose-lowering in patients with type 2 diabetes
mellitus are lacking. Also, no previous attempt was made to evaluate the role of blood pressure
(BP) reduction and LDL cholesterol (LDL-C) change on outcome incidence following glucose-lowering.
Three main clinical questions were investigated: the extent of different outcome reductions by glucose-lowering in patients with diabetes
, the proportionality of outcome reductions to glycated hemoglobin (HBA1c) reductions and whether ongoing BP and LDL-C difference in RCTs can change glucose-lowering outcome effects.
PubMed between 1960 and January 2019 (any language), Cochrane Collaboration Library and previous overviews were used as data sources to identify and select all RCTs comparing the glucose-lowering drugs with placebo or less intense treatment (intentional glucose-lowering RCTs); comparing glucose-lowering drugs with placebo without glucose-lowering intention, but HBA1c difference (nonintentional glucose-lowering RCTs); enrolling type 2 diabetes
mellitus patients; and reporting ongoing SBP and DBP difference. We excluded RCTs of acute care, glucose intolerance, type 1 diabetes
, multiple interventions applied and glucose-lowering by lifestyle or other interventions. Risk ratios and 95% confidence intervals, of seven fatal and nonfatal outcomes
and of treatment-related discontinuations were calculated (random-effects model) before and after adjustment for the ongoing BP difference, while LDL-C difference was also considered. The relationships of different outcome reductions to HBA1c reductions were investigated by meta-regressions.
A total of 25 RCTs (174 235 individuals, follow-up 3.5 years) were eligible, and the resulted ongoing SBP/DBP difference was −1.4/−0.4 mmHg. Both before and after adjustment for BP difference, glucose-lowering reduced CHD (coronary heart disease) and both composites of major cardiovascular events
were reduced by a mean of 8 and 5%, respectively, while before BP-adjustment the risk of treatment-related discontinuations was increased by 26% and the risk of stroke and all-cause death was reduced by 7 and 6%, respectively. Logarithmic risk ratios were related to HBA1c reductions for the composite of CHD and stroke and for treatment-related discontinuations. Glucose-lowering had no differential outcome effects, before and after estimate adjustment for the ongoing BP difference, at different HBA1c thresholds and targets, as well as when both baseline BP and achieved BP, overall cardiovascular risk and diabetes
mellitus duration were considered as dichotomous effect modifiers. Although heart failure
incidence was found increased by 15% in the early glucose-lowering RCTs, this effect faded away in contemporary RCTs. LDL-C change was overall trivial and did not change glucose-lowering outcome effects.
Meta-analyses of all glucose-lowering RCTs involving patients with diabetes
provide precise estimates of benefits for CHD and major cardiovascular events
after consideration of the resulting ongoing BP difference. No benefit or harm on mortality, heart failure
and stroke were noticed, while discontinuations related to adverse events because of treatment were increased following glucose-lowering. The extent of glucose-lowering is proportionally related to changes of CHD and stroke composite, and treatment-related discontinuations.