2.2.1. Possible Flaws in the Design of UKPDS

The results of the UKPDS data highlighting the benefits of metformin have been challenged for various reasons including;

• lack of blinding as the conventional group was not administered a placebo,
• change in significance threshold from initially chosen p < 0.01 to p < 0.05 during the study increasing the probability of results being due to chance alone and
• long period of follow-up leading to risk of attrition bias and difficulty in maintaining the comparability between the groups [22].

After UKPDS, an observational Danish study published in 2011 compared metformin to insulin secretagogues (SU and repaglinide) looking at all-cause mortality, cardiovascular mortality, and the composite of MI, stroke, and cardiovascular mortality in patients with or without previous MI [23]. Monotherapy with individual insulin secretagogues was associated with increased mortality and cardiovascular risk compared to metformin. A previous study published in 2005 to estimate the congestive heart failure (CHF) risk associated with specific therapies for diabetes showed lowest incidence of CHF in those taking metformin compared to other oral and injectable therapies [24]. Further consolidating the UKPDS findings were the results of the Diabetes Audit and Research in Tayside Scotland (DARTS) study published in 2006, which looked at prescribing database for population of Tayside, Scotland [25]. Of the 5730 study patients newly treated with one or more oral GLTs, those treated with metformin had lower risk of adverse cardiovascular outcomes than those treated with SU only, or combinations of SU and metformin. Although not all data has been very convincing, these observational studies support the notion that the findings of UKPDS were not merely by chance alone.

2.2.2. Participant Characteristics

One main criticism is that UKPDS only comprised of individuals with a new diagnosis of T2D not based on a tight criteria (fasting blood glucose range from 6.1 to 15.0 mmol/L) who were not at particularly high risk of CVD in the first place [26]. It is important to remember that the inclusion of obese individuals in UKPDS is consistent with general T2D population and is more representative of the current diabetes population. Later studies indicate that metformin has beneficial effects on CVD even in those with established risk factors or in those with longer duration of diabetes. In data looking at individuals with T2D and established acute coronary syndrome (ACS), metformin users were found to have a lower all-cause mortality rate (hazard ratio (HR) 0.50, 95% CI 0.26, 0.95; p = 0.0346) in the primary analysis compared to non-metformin users [27]. A large trial of 390 individuals treated with insulin investigated whether metformin addition has sustained beneficial metabolic and cardiovascular effects compared to placebo in individuals with T2D [28]. The primary outcome was a composite of microvascular and macrovascular endpoints and the secondary outcomes were microvascular and macrovascular morbidity and mortality separately. Although, metformin was not associated with an improvement in the combined primary outcome, it was associated with an improvement in the secondary macrovascular outcomes (HR 0.61, 95% CI 0.40, 0.94; p = 0.02).

2.2.3. Increased Mortality in Combination with SU

Another criticism is based around the increased mortality seen in those treated with combination of SU and metformin in UKPDS. In the UKPDS, early addition of metformin in SU-treated patients was associated with an increased risk of diabetes-related death (96%, 95% CI 2, 275, p = 0.039) compared with continued SU alone, although a sub-group analysis did not show such association with a combination of SU and metformin [2]. A meta-analysis looking at the safety of combination of SU and metformin showed an increased risk for a composite end-point of CVD hospitalisations and mortality but no significant effects of this combination therapy was seen on either CVD mortality or all-cause mortality alone [29]. An observational cohort study also showed a significantly increased mortality in individuals treated with combination of various SUs (glibenclamide and chlorpropamide) and biguanides compared to other combinations and monotherapies [30]. However, it was acknowledged that patients who were on combined treatment were more likely to be obese and showed more severe associated metabolic abnormalities. Most of the individuals on combination therapy were treated with older longer-acting SUs like glibenclamide. Newer 3rd generation SUs like glimepiride, which have lower affinity for myocardial ATP-dependent potassium channels compared to glibenclamide [31], could show a lower detrimental impact when used with biguanides [30] and animal models have confirmed that compared to conventional SUs, glimepiride has less harmful cardiovascular effects [32]. The newer generation of SUs also appear to confer less risk of hypoglycaemia compared to older generation of SUs used in UKPDS [33]. As a result, these shorter-acting SUs could be a safer option when combined with metformin compared to longer-acting SUs.

2.2.4. Impact of Other Interventions

Critics also raised concerns because of the impact of other interventions including optimisation of blood pressure control and lipid profile during the long follow-up period in UKPDS, which could have affected the outcomes. The Steno-2 study demonstrated that early multifactorial risk reduction does lead to improvement in outcomes including CVD death [34] and treatment intensification is required over the course of T2D timeline.

2.2.5. UKPDS vs Newer Cardiovascular Outcome Trials (CVOTs)

It is important to remember that UKPDS was first reported more than two decades ago looking at effects of intensive GLT on long-term outcomes and was not designed to test the efficacy of individual glucose-lowering agents. Similarly, Action to Control Cardiovascular Risk in Diabetes (ACCORD) [35], Action in Diabetes and Vascular Disease: Preterax and Diamicron Modified Release Controlled Evaluation (ADVANCE) [36] and Veterans Affairs Diabetes Trial (VADT) [37] were trials comparing the effects of intensive vs standard glucose control on cardiovascular events in T2D populations with established or high risk of cardiovascular disease and were not drug efficacy trials per se. In all these three trials, individuals in both intensive and control arms received metformin therapy. In ACCORD, 94.7% individuals were treated with metformin in the intensive therapy arm compared to 86.9% in the control arm [35], in ADVANCE, 74% of participants received metformin in the intensive control group vs 67% in the standard control group [36] and finally in VADT [37], in both study groups, individuals either received metformin or glimepiride based on BMI in combination with rosiglitazone. As a result, it becomes difficult to draw any direct inferences regarding metformin efficacy based on described trial designs. Consequently, changes in trial standards over these years complicates drawing direct comparisons between metformin and newer agents. The history of CVOTs can broadly be categorised into pre-2008 and post-2008 periods after the US Food and Drug Administration (FDA) introduced the new guidelines for cardiovascular evaluation of GLTs largely focussed around the composite end-point of Major Adverse Cardiovascular Events (MACE) following the discovery of increased adverse cardiovascular events associated with rosiglitazone use [38,39]. Hence, considerable caution is required when comparing older and newer trials, which are more standardised in design and amenable to meta-analysis [26]. It is also worth noting that the modern CVOTs are relatively short trials recruiting mostly high-risk individuals who are unrepresentative of the general population. Hence, there is a need to widen the scope of these CVOTs by facilitating more pragmatic designs employing diverse recruitment strategies with flexibility in design, ensuring extended follow-up and exploring the additive effects of different combinations of GLTs [40].

2.2.6. The European Society of Cardiology (ESC) 2019 Guidelines on Diabetes, Pre-diabetes, and CVD

The MACE outcome trials of “newer” GLTs namely the glucagon-like peptide-1 receptor agonists (GLP-1RA) and sodium-glucose co-transporter 2 inhibitors (SGLT2i) have demonstrated cardiovascular benefits in large CVOTs. Based on these trials, the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD) recommend early addition of these agents to baseline metformin therapy in individuals with established cardiorenal disease or high risk for cardiorenal disease [41]. The ADA/EASD recommendations retain metformin as the fundamental starting therapy in individuals with new diagnosis of uncomplicated T2D.

However, In 2019, the ESC guidelines on diabetes, pre-diabetes, and CVD recommended use of either SGLT2i or GLP-1RA in individuals with established ASCVD or high/very high cardiovascular risk as first-line, whether they are treatment-naïve or already on metformin [42]. Similarly, others have also proposed an alternate algorithm for individuals with T2D and ASCVD in which SGLT2i and GLP-1RA are the first line irrespective of HbA1c targets [43]. It is important to remember that metformin was the baseline therapy in most participants in these CVOTs and the cardiovascular benefits of SGLT2i and GLP-1RA remain largely unknown in treatment-naïve individuals. The current advice from ESC may be a step too far to start these newer agents in treatment-naïve individuals before metformin.

It may appear as if there is an area of difference between ADA/EASD recommendations and ESC guidelines with regards to positioning of metformin. However, on closer inspection there are more similarities between the two as ADA/EASD emphasizes that patients at high risk for cardiorenal disease should be treated with SGLT2i or GLP-1RA independent of HbA1c. In addition, most patients with T2DM progress to requiring combination therapy soon after baseline metformin therapy. A recent meta-analysis examining comparative effectiveness of GLTs concluded that the use of metformin as first-line treatment in treatment-naïve individuals with low cardiovascular risk remains justified and given the lack of evidence, they could not reach a conclusion about the optimal initial treatment of treatment-naïve patients at increased cardiovascular risk [44]. Future studies will need to address whether SGLT2i and GLP-1RAs should be used for the prophylaxis of ASCVD and CKD in uncomplicated newly diagnosed treatment-naïve individuals with T2D to address this further.

2.3.1. Meta-analysis Not Supporting UKPDS Findings

A meta-analysis by Griffin et al., examined randomised trials looking at the impact of metformin on CVD exclusively compared to diet, lifestyle or placebo [45]. All outcomes, with the exception of stroke, favoured metformin, but none achieved statistical significance and they concluded that there is uncertainty about whether metformin reduces risk of CVD among individuals with T2D. However, they acknowledged that this was mainly due to lack of evidence from a randomised-controlled CVOT comparing metformin against placebo. Similarly, Boussageon et al., did not find any evidence supporting metformin’s beneficial impact on all-cause mortality, cardiovascular mortality or any major macrovascular complications in their meta-analysis of randomised-controlled trials (RCTs) evaluating metformin cardiovascular efficacy compared to placebo, diet, no treatment as well as studies looking at metformin as an add-on therapy and studies of metformin withdrawal [46]. They suggested a need for further studies to clarify this as the existing number and qualities of studies were insufficient.

2.3.2. Meta-analysis/Systematic Reviews (Partially or Fully) Supporting UKPDS Findings

In a large meta-analysis by Han et al. looking at cardiovascular mortality between metformin and non-metformin groups, the adjusted HR was 0.81 (95% CI 0.79, 0.84; p < 0.00001) in favour of metformin [47]. Similarly, the adjusted HR between the metformin and non-metformin groups for all-cause mortality was 0.67 (95% CI 0.60, 0.75; p < 0.00001). Similarly, in a previous meta-analysis a significant cardiovascular benefit was observed in metformin versus placebo/no therapy trials with odds ratio (OR) of 0.79 (95% CI 0.64, 0.98; p < 0.031), but this superiority was not seen with metformin in active-comparator trials with OR of 1.03 (95% CI 0.72, 1.77; p = 0.89) [48]. They proposed that the reason behind this could be that metformin cardiovascular benefits are dependent on its glucose-lowering properties hence the observed superiority only compared to placebo or no therapy groups. However, the UKPDS data showed improvements in cardiovascular outcomes with metformin compared to insulin or SU despite similar glycaemic improvements. Campbell et al., performed an extensive meta-analysis comparing several cardiovascular outcomes of metformin against non-metformin therapy [49]. This supported UKPDS findings to an extent showing that all-cause mortality improved with metformin therapy compared to active comparators, but this was not seen in trials comparing metformin against diet-controlled diabetes. They thought that the difference could be due to diet therapy being a more likely option for individuals with early or less severe disease. A 2008 systematic review exploring cardiovascular outcomes of oral GLTs showed that metformin therapy was associated with a decreased risk of cardiovascular mortality compared with both placebo and any other oral GLT [50]. However, it failed to establish metformin’s superiority on all-cause mortality or cardiovascular morbidity suggesting that compared to other oral GLTs and placebo, metformin overall appeared only moderately protective. Finally, Crowley et al., looked at the cardiovascular outcomes of metformin in high-risk populations with CHF, chronic kidney disease (CKD) and chronic liver disease (CLD) [51]. They showed improvement in key clinical outcomes including all-cause mortality and MACE with metformin use even in individuals with moderate CKD, CHF, or CLD compared to non-metformin regimens, proving that metformin is not only safe but also potentially beneficial in such high-risk populations.

Table 1 below provides a summary of these systematic reviews and meta-analyses with main outcome results and important conclusions.

In summary, the results of these reviews are inconclusive in respect to the cardiovascular impact of metformin. It was acknowledged that a lot of this uncertainty was due to lack of a large randomised double-blind, placebo controlled trial with dedicated MACE end-points [45], lack of long-term evaluations [26] and the unexplained deleterious effect of the combination of metformin plus SU from the UKPDS data [25,46]. Additionally, differences in patient characteristics, study designs and duration could provide an explanation for the discrepancies observed.

In the majority of these meta-analyses and systematic reviews the balance does tilt in favour of metformin. Given its beneficial effects on HbA1c, weight and cardiovascular mortality and relative safety profile compared to most of the other GLTs, it is reasonable to support metformin as first-line therapy for individuals with new diagnosis of uncomplicated T2D [52].