Understanding legacy effects: epidemiology or metabolic memory?

Recently several trials with the new classes of antidiabetic drugs SGLT-2 inhibitors and GLP-1RA have demonstrated a beneficial effect of these drugs in reducing the risk for developing cardiovascular disease and nephropathy in type 2 diabetes.
One of the main issues in evaluating the results of these trials is the heterogeneity of the population included, which, in practice makes impossible to compare the studies between them. In some studies only people with a previous established cardiovascular event have been included, while in others there is also a consistent component of people with only several risk factors for cardiovascular disease.
Because the aim of thee studies is to show the efficacy/superiority of a drug compared to placebo at the enrolment the patients must have comparable glycated haemoglobin and duration of the disease.
The placebo, in reality, is just the daily practice, without the studied drug, aiming to keep during the study the same level of glycemic control. 
It is clear that, even in presence of a previous established cardiovascular event, we do not have enough information on the history of the patients until the time they are recruited.
Does this omission/ignorance introduce a possible bias in the final results of the study.
Certainly we are aware of the so called phenomenon of the “metabolic memory” or “legacy effect”.
Large randomized studies have established that early intensive glycemic control reduces the risk of diabetic complications, both micro and macrovascular. Moreover, epidemiological and prospective data support a long-term influence of early metabolic control on clinical outcomes. This phenomenon has recently been defined as "metabolic memory." Potential mechanisms for propagating this "memory" may be the production of reactive species unrelated to the presence of hyperglycemia, depending on the previous production of AGEs which can maintain RAGE over-expression, on the level of glycation of mitochondrial proteins and on the amount of mtDNA produced, all conditions able to induce an altered gene expression which may be persistent even when glycemia is normalized. 
According to this evidence, it is clear that we are unable to understand how much such phenomenon might have influenced the results of the CVOTs, because the population compared in the trials, at the end of the day, might be really not comparable.