Does metformin remain as the first choice compared to all new hypoglycemic agents? An observational cohort study was recently published in JAMA InternMed with the aim of exploring the impact of the choice of starting oral hypoglycemic drug therapy on subsequent intensive treatment. The trial included 15,516 patients with untreated diabetes who were initially treated with metformin, sulfonylureas (SU), thiazolidinediones (TZD), or dipeptidyl peptidase 4-inhibitors (DPP4). The primary observation was the time from initiation of therapy to subsequent intensive therapy, with intensive therapy defined as requiring the addition of other types of oral hypoglycemic drug therapy. Secondary observations were time intervals between compound cardiovascular events (including coronary artery disease, congestive heart failure, unstable angina, ischemic stroke, acute myocardial infarction, or coronary artery revascularization), congestive heart failure alone, hypoglycemic emergencies or hospitalizations, and any other diabetes-related emergency management. After correcting for multiple covariates such as demographic characteristics, comorbidities, and differences in healthcare resource use, statistical analyses were performed using Cox proportional risk models. Metformin Remained the First Choice Of all subjects, 58% of patients initiated treatment with metformin, and an additional 23%, 6%, and 13% initiated treatment with SU, TZD, and DPP4 inhibitors, respectively. The follow-up period was slightly more than 1 year, during which the time interval between the initiation of treatment and subsequent intensification differed significantly between the groups. After multivariate analysis, the results showed that 25% of patients in the starting metformin group required the addition of a second oral agent, whereas 37%, 40%, and 36% of patients in the starting SU, TZD, and DPP4 inhibitor groups required the addition of a second oral agent, respectively. In addition, the risk of intensification was increased by 68%, 61%, and 62% in the groups using SU, TZD, and DPP4 inhibitors, respectively, compared with the metformin group. And, the risk of composite cardiovascular events, congestive heart failure, and hypoglycemia was also increased in the SU use group. Does metformin reduce mortality? And in a retrospective study published by Diabetes ObesMetab, researchers enrolled diabetic patients who initiated treatment with metformin or SU-based monotherapy between 2000 and 2012, with a duration of treatment of at least 180 days, using data from the U.K. Clinical Practice Research Database. Patients were also matched based on age, sex, cancer history, and smoking status, and non-diabetic individuals were included as controls. Individuals in both groups were followed for 5.5 years to compare subsequent mortality. Initiation of treatment with metformin was compared with SU treatment, while the diabetic group was compared with a matched non-diabetic control group. Survival times were compared by multivariate analysis, with covariates including age, sex, Charlson comorbidity index, smoking status, and prior use of antihypertensive and lipid-lowering medications and antiplatelet therapy. Survival time was longer in diabetic patients treated with metformin compared with the nondiabetic population The study included 78,241 diabetic patients starting treatment with metformin and 12,222 diabetic patients starting treatment with SU, respectively, and matched the same number of nondiabetic individuals for control. At baseline, the metformin-treated group was younger and heavier than the SU group (BMI: 32.4 vs 27.1 kg/m2), and baseline HbA1c levels were higher in both groups, but higher in the SU group than in the metformin group (9.2% vs 8.6%). Crude mortality was significantly lower in the metformin-using group than in the SU group (14.4 vs 50.9/1000/year). After correcting for multivariable, survival time was 38% lower in the SU group compared with the metformin group. The results of all subgroup analyses supported the superiority of metformin over SU, and the difference was statistically significant. The most surprising result was the 15% lower survival time in the matched no-diabetes control group compared with the metformin group. This result was also consistent with the results of the various other subgroup analyses, where almost all differences were statistically significant and were particularly pronounced in patients with high-risk comorbidities. In addition, the overall results were not affected after adjusting for cardiovascular prophylactic medication factors. ANALYSIS AND CONCLUSIONS Randomized controlled trials are an important method of scientific research, but they cannot explain all the issues associated with the superiority of one drug over another. And drug registration trials are too short to determine medium-term effects on clinical outcomes. Although glucose-lowering drugs currently require longer trials to demonstrate cardiovascular efficacy and safety, such studies may not provide a true picture of new drugs in the clinic because of overly stringent control of confounders. Therefore, there is a need for comparative drug efficacy studies that utilize the rich data resources available in electronic medical databases. However, these studies must be carefully designed to be informative, and the two articles mentioned above are good examples. Berkowitz and his colleagues compared four classes of hypoglycemic agents head-to-head, a comparison that would have been costly in a randomized trial. The time between initiation of therapy and the need for intensive therapy was consistent with the approach taken in the ADOPT trial, but not in the ADOPT trial for two classes of drugs. The follow-up period was too short, resulting in fewer secondary outcomes. Still, the results showed that metformin was significantly superior to SU, TZD, and DPP-4 inhibitors. The main shortcoming of the analysis was the inability to reference HbA1c values. Low baseline HbA1c levels are important predictors of glycemic attainment and maintenance. If baseline levels of HbA1c vary significantly between groups, this will affect the subsequent time interval at which intensive therapy is required. For example, patients starting a new drug, DPP4 inhibitor, may have poor glycemic control at baseline levels, which is a classic case of confounding due to medication indications. The differences between metformin and other drugs are also significant, and while correcting baseline HbA1c levels can reduce the impact on outcomes, it cannot completely eliminate it. The time interval between initiation of treatment and the need for intensive treatment is one of the commonly used methods for evaluating diabetes outcomes, with mortality serving as an observational endpoint for epidemiologic analyses in general. The innovative use of this method in this study (which may never be used as part of a clinical trial) was to match cohorts not on this drug as a control group. The protective effect of metformin relative to SU is unsurprising as, previously, it has been reported in epidemiologic analyses of the UKPDS.The results of the comparison between the SU group and its matched control group were also expected as previous findings have shown that the mortality rate in diabetic patients is approximately twice as high as that of non-diabetics. The most meaningful finding was that metformin demonstrated a significant protective effect compared with nondiabetics who did not use metformin. Metformin benefits diabetics, and it is hypothesized that it would also be beneficial in the non-diabetic population. However, the results of a recent randomized controlled trial in a nondiabetic population showed no significant difference between the progression of distal carotid intima-media thickness in the metformin-using group and the placebo group during 18 months of follow-up. However, this endpoint does not determine whether metformin has a benefit in terms of mortality in the nondiabetic population. Whether metformin reduces mortality is perhaps not definitively answered by a long-term randomized controlled trial, as well as by those head-to-head studies on which glucose-lowering drug to prioritize, given the increasing number of treatment options available. Nonetheless, there is much to be gained from these well-designed comparative studies of efficacy.