The present study was a multicenter, open-label, randomized, active-controlled study conducted at 16 centers in South Korea. The duration of the study was 17 weeks (five visits) and included a 1-week screening period (visits 1 to 2) and a 16-week open-label treatment period. Each patient attended an initial screening visit during which the inclusion and exclusion criteria were assessed. At visit 2, all eligible patients were randomly assigned to receive either vildagliptin (50 mg twice daily, n=117) or pioglitazone (15 mg once daily, n=111) as an add-on treatment to metformin using a 1:1 ratio. After the randomization procedure, the patients returned to the clinic for the following assessments: visit 3 (week 4), visit 4 (week 12), and visit 5 (week 16). The first patient visit took place on December 22, 2009 and the final patient visit took place on July 9, 2012. All participants provided written informed consent prior to participating. The protocol was approved by the institutional review board at each study site and the study was conducted in accordance with the Declaration of Helsinki (Clinical trial reg. no. NCT01882907).

Males and females between 18 and 80 years of age with type 2 diabetes were eligible to participate. Additional inclusion criteria consisted of an HbA1c level of 7.0% to 11.0% and an FPG level of <270 mg/dL at the initial screening (visit 1) and treatment with a stable dose of metformin (≥1,000 mg/day) for 4 weeks prior to the initiation of the study.

Patients were excluded from the present study based on the following criteria: pregnant or nursing females or females planning to become pregnant during the study period; reproductive females willing to prevent conception; patients with type 1 diabetes or diabetes due to a secondary cause; those with severe diabetes complications, active infections that could have influenced glycemic status in the 4 weeks prior to visit 1, Torsades de Pointes (clinically significant and persistent ventricular tachycardia and ventricular fibrillation), myocardial infarction, unstable angina, stroke, congestive heart failure (all New York Heart Association classes I to IV), liver disorders such as cirrhosis or chronic hepatitis, or hypersensitivity or allergic reactions to pioglitazone and/or rosiglitazone; those who had undergone percutaneous coronary intervention within the previous 3 months or coronary artery bypass surgery within the previous 6 months; patients receiving insulin treatment without a concomitant illness over the previous 4 weeks; those who had used glucocorticoids within the previous 8 weeks (over 7 days); and those with alanine aminotransferase or aspartate aminotransferase levels >2.5 times the upper limit of normal (ULN), direct bilirubin levels >1.3 times ULN, serum creatinine levels >1.5 mg/dL, and/or clinically abnormal thyroid-stimulating hormone levels.

The body weights and vital signs of each patient were measured at each visit. Fasting lipid profiles (triglyceride [TG], total cholesterol, low density lipoprotein cholesterol [LDL-C], high density lipoprotein cholesterol [HDL-C], and non-HDL-C levels), standard hematology and biochemistry laboratory assessments, and homeostasis model assessment as an index of insulin resistance (HOMA-IR) and the homeostatic model assessment β-cell function (HOMA-β) were obtained at the initial screening and at weeks 4, 12, and 16. HOMA-IR and HOMA-β represent insulin resistance and pancreatic β-cell function, respectively, and were calculated as follows: HOMA-IR=fasting insulin (FI, µU/mL)×FPG (mmol/L)/22.5, and HOMA-β=20×FI (µU/mL)/[FPG (mmol/L)–3.5]. In addition, electrocardiograms were performed at the screening and week 16.

At each post-baseline visit, all adverse events (AEs) were recorded and assessed for severity and a possible relationship to the study medications. Patients were provided with glucose monitoring devices to determine hypoglycemia, which was defined as symptoms suggestive of low glucose confirmed by a self-monitored glucose measurement of <56 mg/dL. Hypoglycemia was classified into three types: (1) asymptomatic hypoglycemia, which was a glucose level <56 mg/dL without typical symptoms; (2) hypoglycemia grade 1 (mild hypoglycemia), which was typical symptoms and a confirmed glucose level <56 mg/dL; and (3) hypoglycemia grade 2 (severe hypoglycemia), which was an event requiring the assistance of another person regardless of glucose levels.

The primary efficacy parameter was changes in HbA1c levels during the 16-week treatment period following the administration of either vildagliptin (50 mg twice daily) or pioglitazone (15 mg once daily) as an add-on treatment to metformin in the intention-to-treat (ITT) population. The secondary efficacy parameters included assessments of FPG and PPG levels, lipid profiles, and body weight. The mean changes in the HOMA-IR and the HOMA-β from baseline to 16 weeks were also investigated.

We compared the efficacies of vildagliptin (50 mg twice daily) to pioglitazone (15 mg once daily) in terms of reducing HbA1c levels from baseline to week 16 (non-inferiority margin: upper limit of the 97.5% confidence interval <0.3%). The primary and secondary endpoints were assessed in the ITT population and missing data were accounted for using the later observation carried forward method. Of the 287 patients who consented to the participation in this study, 59 were excluded during the screening and 228 were randomized. The safety population consisted of the 228 randomized patients that received at least one dose of study drug and had at least one post-baseline safety assessment. After excluding five patients who had no efficacy data available to assess their primary or secondary endpoints, the ITT population consisted of the 223 randomized patients that received at least one dose of a study drug and had at least one post-baseline assessment of the primary or secondary efficacy variables. The analyses of all efficacy endpoints were conducted with paired t-tests and a P<0.05 was considered to indicate statistical significance. The incidence rates of hypoglycemic events and other AEs through week 16 were assessed using descriptive statistics and used to determine the safety population.

The flow of the study patients from the initial screening to the study endpoint is summarized in Fig. 1. Of the initial 287 patients who were recruited, 49 were excluded during the screening process and 228 were randomly assigned to a drug group. Table 1 summarizes the demographic and baseline metabolic characteristics of the patients in the randomized population. The mean HbA1c levels of the vildagliptin (7.98%±0.77%) and pioglitazone (8.08%±0.98%) groups at baseline (visit 1) did not significantly differ (P=0.389). The FPG and PPG levels, lipid profiles, and body weights of the two groups were also similar at baseline and there were no significant differences in medication history between the two groups (Supplementary Table 1).

The time-courses of the changes in mean HbA1c levels in the ITT population during the 16-week treatment period following the administration of either vildagliptin (50 mg twice daily) or pioglitazone (15 mg once daily) as an add-on treatment to metformin in patients with type 2 diabetes are provided in Fig. 2B. The mean changes in HbA1c levels from baseline to week 16 were –0.94% in the vildagliptin group and –0.6% in the pioglitazone group (P=0.010) (Fig. 2A). The between-group difference (ITT population) in the changes in HbA1c levels from baseline to week 16 was 0.34%±0.96%. Based on a non-inferiority margin of 0.3%, this finding establishes the non-inferiority of vildagliptin relative to pioglitazone as an add-on therapy to metformin.

The time-courses of the mean changes in FPG and PPG levels during the 16-week treatment period following the administration of either vildagliptin (50 mg twice daily) or pioglitazone (15 mg once daily) as an add-on treatment to metformin in patients with type 2 diabetes are provided in Fig. 2C and D. The mean changes in FPG from baseline to week 16 were –20.4 mg/dL in the vildagliptin group and –15.0 mg/dL in the pioglitazone group (P=0.273). The mean changes in PPG levels from baseline to week 16 were –60.2 mg/dL in the vildagliptin group and –38.2 mg/dL in the pioglitazone group (P=0.040).

In the ITT population, the fasting lipid levels of the two treatment groups were very similar at baseline. The fasting TG, total cholesterol, LDL-C, HDL-C, and non-HDL-C levels averaged 142.8±69.7, 170.0±34.5, 98.0±28.7, 46.3±11.8, and 123.9±32.1 mg/dL, respectively, in the vildagliptin group and 144.7±68.8, 173.9±34.9, 100.2±32.0, 48.3±11.6, and 125.2±34.7 mg/dL, respectively, in the pioglitazone group. There were significant decreases in the levels of total cholesterol, LDL-C, HDL-C, and non-HDL-C in the vildagliptin group (mean changes from baseline to endpoint: –6.5±25.1, –6.0±21.2, –0.3±8.6, and –8.4±24.7 mg/dL, respectively) but increases in the pioglitazone group (mean changes from baseline to endpoint: 9.6±33.8, 5.4±30.4, 3.7±10.3, and 5.8±32.7 mg/dL, respectively). The fasting TG levels in the vildagliptin group exhibited a greater decrease (–9.2±56.0 mg/dL) than those of the pioglitazone group (–4.8±67.8 mg/dL) and the mean changes in total, LDL-C, HDL-C, and non-HDL-C from baseline to week 16 were significantly different between the two groups (all P values <0.05).

The time-courses of the changes in mean body weight during the 16-week treatment period following the administration of either vildagliptin (50 mg twice daily) or pioglitazone (15 mg once daily) as an add-on treatment to metformin in patients with type 2 diabetes are provided in Fig. 2E. The mean changes in body weight from baseline to week 16 were –0.07 kg in the vildagliptin group and 0.69 kg in the pioglitazone group, which were significantly different from each other (P=0.002).

The mean changes in HOMA-IR and HOMA-β from baseline to 16 weeks were analyzed (data not shown). The mean changes in HOMA-IR from baseline were –0.06±2.92 in the vildagliptin group and –3.33±25.16 in the pioglitazone group (P=0.183). And the mean changes in HOMA-β from baseline were 9.77±28.10 in the vildagliptin group and –24.26±294.88 in the pioglitazone group (P=0.183). The mean changes in insulin from baseline were 0.36±6.95 in the vildagliptin group and –8.21±70.06 in the pioglitazone group (P=0.209) and the mean changes in C-peptide were 0.07±0.84 in the vildagliptin group and –0.24±1.14 in the pioglitazone group (P=0.023).

The incidence rates of hypoglycemic events through week 16 are summarized in Table 2. Two patients (1.71%) in the vildagliptin group and three patients (2.70%) in the pioglitazone suffered a hypoglycemic event but these rates did not significantly differ. Likewise, the incidence rates of AEs did not differ between the two groups over the 16-week treatment period. Both study drugs were relatively safe and well-tolerated and there were no serious drug-related AEs or death. The most commonly occurring AEs in both groups were common colds, upper respiratory tract infections, nausea, diarrhea, dyspepsia, and dizziness (Supplementary Table 2).