This study, called the Practical Evidence of Antidiabetic Monotherapy Study (PEAM), consisted of a multicenter, randomized, double-blind trial performed in 15 centers in Korea. Between February 2007 and December 2008, 435 patients who had not received previous pharmacologic treatment for type 2 diabetes mellitus were screened and 349 patients were randomized to one of three treatment groups. We assigned 118 patients to the glimepiride group, 114 patients to the metformin group, and 117 patients to the rosiglitazone group (Fig. 1). After a 4-week life style intervention, the study drugs were prescribed to each patient. During the lifestyle intervention period, we provided individualized education to each study subject, according to current, recommended guidelines for medical nutritional treatment [10]. We also recommended that each subject perform at least 150 minutes per week of moderate-intensity aerobic physical activity, provided exercise was not contraindicated [10]. Trained dietitians or diabetic nurse specialists provided the education and made these recommendations; however, we did not later verify whether the subjects had followed these recommendations appropriately.

Participants were examined every 8 weeks for 48 weeks from the start of the randomization period. According to the results of HbA1c (<6.5% or ≥6.5%) and drug tolerability check that was performed at each visit, we performed scheduled up-titration of the study drugs (Table 1A). Fasting plasma glucose levels (FPG), HbA1c concentrations, medication compliance, and a physical examination, including taking vital signs and body weight (BW), were assessed or performed at every visit. To evaluate the study participants' medication compliance indirectly, we checked the number of remaining doses at each visit. The ratio of the remaining number of doses at the present visit to the number of doses prescribed at the previous visit was calculated. Subjects with a ratio greater than 30% were excluded in the outcome analysis, since their compliance rate was presumed to be less than 70%. At the outset of randomization and after 48 weeks, we performed a 75 g oral glucose tolerance test. Anthropometric characteristics, demographic characteristics, electrocardiogram, lipid profiles, complete blood counts (CBC), aspartate aminotransferase (AST), alanine aminotransferase (ALT), blood urea nitrogen (BUN), and creatinine were evaluated. All adverse events were recorded and judged for severity and possible relationship to study medications. The study protocol was approved by the Institutional Review Board at each center and all participants provided written informed consent.

Eligible study participants were between the ages of 30 and 65 years, with HbA1c levels ranging from 6.5% to 9.5%. None of the subjects had ever taken an oral hypoglycemic agent. Glucocorticoid users, pregnant women, patients who had clinically significant liver disease (AST, ALT>2.5 x upper normal limit), significant renal disease (serum creatinine>1.5 mg/dL in men,>1.4 mg/dL in women), a history of lactic acidosis, a history of unstable angina or severe angina pectoris, a history of or treatment for congestive heart failure, or contraindications to metformin or sulfonylurea treatment were excluded.

The primary goal of this study was to determine changes in HbA1c levels from baseline (randomization) to end point. The secondary goals were to determine changes in FPG, body weight, and the numbers of subjects achieving target HbA1c levels (<6.5%).

We evaluated differences in all adverse events for the study populations. Hypoglycemia was defined as the presence of typical adrenergic or neuroglycopenic symptoms and signs, regardless of the data for self-monitoring of blood glucose. Edema was defined when the subjects had any signs of fluid retention upon physical examination, or complained of systemic edema.

At baseline, 24 weeks, and 48 weeks, HbA1c levels were measured using high-performance liquid chromatography and FPG was determined using a hexokinase method in a central laboratory (Samkwang Medical Laboratories, Seoul, Korea). Levels of FPG and HbA1c in the other follow-up visits, and lipid levels, were determined using standard assays in each local laboratory.

Originally, we calculated a need to enroll 540 study participants to obtain 80% statistical power at a significance level of P=0.05, assuming a 20% attrition rate in each study group. At the end of the study, the overall attrition rate was 34.9%, with a total of 349 randomized subjects. We used an intention-to-treat analysis method in analyzing our data. The last observation carried forward (LOCF) method was used to fill in missing values at a later point in the study. Using per protocol analysis, we included, 82 subjects in the glimepiride group, 71 subjects in the metformin group, and 74 subjects in the rosiglitazone groups in our analysis (data not shown here).

Results are described as mean±standard deviation or median±interquartile range. For comparisons of data among the three treatment groups, a repeated measured ANOVA test was used. For nonparametric statistical analysis, the Kruskal-Wallis test was used. Wilcoxon's signed rank test was used for comparison of pre- and post-treatment values. For analysis of differences in the frequency of adverse events, the chi-square test and Fisher's exact test were used. Statistical analyses were performed using SAS version 9.1.3 (SAS Institute, Cary, NC, USA). Differences among groups with a P<0.05 were considered statistically significant.

The subjects' baseline characteristics are described in Table 2. All variables in baseline characteristics were measured at the randomization point. There was no significant difference in HbA1c levels, body weight, blood pressure, and lipid profiles between the screening visit and the randomization visit (data from the screening period not shown here).

The mean age of the total study population was 50.9 years, the mean HbA1c level was 7.8%, and the mean body mass index (BMI) was 25.6 kg/m². No significant difference was observed in baseline demographics, anthropometrics and metabolic characteristics among the three groups after randomization. Medication compliance was not different in the three groups. The proportion of participants who completed the study was 69.4% in the glimepiride group, 62.3% in the metformin group, and 63.2% in the rosiglitazone group. The number of study participants at each drug dosage level is described in Table 1B.

HbA1c levels decreased from 7.8% to 6.9% in the glimepiride group (P<0.001), from 7.9% to 7.0% in the metformin group (P<0.001), and from 7.8% to 7.0% (P<0.001) in the rosiglitazone group (Fig. 2). Differences in HbA1c levels from randomization to the end point were -0.89±0.76% in the glimepiride group, -0.92±0.96% in the metformin group, and -0.82±0.79% in the rosiglitazone group. There was no significant difference in the HbA1c levels, or in the changes in HbA1c levels, between the three groups (P=0.62) (Fig. 3A).

Levels of FPG decreased from 145.0±32.0 mg/dL to 128.1±27.9 mg/dL in the glimepiride group (P<0.001), from 151.0±32.4 mg/dL to 130.8±25.8 mg/dL in the metformin group (P<0.001), and from 144.1±31.5 mg/dL to 128.6±35.2 mg/dL in the rosiglitazone group (P<0.001). Among the three study groups, levels of FPG at baseline and the end point were statistically different (P=0.52) (Fig. 3B).

Over the study period, significant weight gain was observed in the glimepiride group (ΔBW=1.4 kg, ΔBMI=0.54 kg/m²). In the rosiglitazone group, mean weight increased continuously (ΔBW=1.5 kg, ΔBMI=0.63 kg/m²); however there was significant weight reduction in the metformin group (ΔBW=-1.1 kg, ΔBMI=-0.44 kg/m²) (Fig. 3C).

Study participants achieving target HbA1c levels of less than 6.5% were 36.44% (43 subjects) of the 118 members of the glimepiride group, 24.56% (28 subjects) of the 114 members of the metformin group, and 32.48% (38 subjects) in the 117 members of the rosiglitazone group, without significant differences among the three groups (P=0.14). Those with HbA1c level less than 7.0% were 65.25% (77 subjects) of the glimepiride group, 58.77% (67 subjects) of the metformin group, and 58.97% (69 subjects) of the rosiglitazone group, with no significant differences among the three groups (P=0.51) (Fig. 4).

Symptomatic hypoglycemia was more frequent in the glimepiride group (19.49% in glimepiride group vs. 3.51% in metformin group and 6.84% in rosiglitazone group). Diarrhea was more frequent in the metformin group (3.39% of glimepiride group vs. 8.77% of metformin group and 1.71% of rosiglitazone group). Subjective edema as reported by study participants was more frequent in the glimepiride and rosiglitazone groups compared to the metformin group (6.78% of glimepiride group vs. 0.88% of metformin group and 7.69% of rosiglitazone group) (Table 3). Other adverse events, including abdominal discomfort, elevated liver enzymes and chest discomfort or dyspnea were not different in the three study populations.