The Practical Evidence of Antidiabetic Combination Therapy in Korea (PEAK) study was a multicenter, double-blinded, randomized, placebo-controlled, three-arm parallel trial conducted across 13 centers in Korea. Between January 30, 2015 and October 4, 2018, we enrolled 279 patients with T2DM who were receiving metformin (≥1,000 mg). Of these, 214 eligible participants were randomized into one of three treatment groups: the metformin+alogliptin+pioglitazone placebo (Alo) group (n=75), the metformin+pioglitazone+alogliptin placebo (Pio) group (n=69), and the metformin+alogliptin+pioglit azone (Alo+Pio) group (n=70). Following a 4-week screening period, each patient was prescribed OHAs corresponding to their assigned group, maintaining the same metformin dosage from the screening period until the end of the trial (Supplementary Fig. 1).

The participants underwent assessments every 12 weeks for 24 weeks from the initiation of the randomized treatment. In addition to pharmaceutical intervention, all participants received education on diet, exercise, and lifestyle modification. They were also provided a diary for self-monitoring blood glucose. Demographic, anthropometric (height, weight, and body mass density), lifestyle (alcohol consumption and smoking habits), and medication data were collected. Physical examinations and laboratory assessments were performed at each visit.

Written informed consent was obtained from all study participants. The study was approved by the Institutional Review Board at each center and conducted following the ethical principles of the Declaration of Helsinki (Approval No: KC14MIM V0127) (Clinical Trial Registration Number: clinicaltrial.gov, NCT02231021).

This study’s inclusion criteria were as follows: age between 19 to 75 years, diagnosis of T2DM a minimum of 6 months prior to the initiation of the study, body mass index (BMI) within the range of 18.5 to 45 kg/m², baseline HbA1c levels from 7.0% to 10.0%, indicating suboptimal glycemic control, despite being on a stable metformin regimen (≥1,000 mg of the maximally tolerated dose), fasting C-peptide level >0.78 ng/mL (0.26 nmol/L), systolic/diastolic blood pressure ≤160/100 mm Hg, and a hemoglobin level of at least 12 g/dL for men and 10 g/dL for women.

Exclusion criteria included the use of medications that significantly influence blood glucose control, such as glucocorticoids and pregnancy or lactation. Those with clinically significant liver disease, manifested by aspartate aminotransferase and alanine aminotransferase levels ≥2.5 times the upper normal limit, or substantial renal disease, indicated by an estimated glomerular filtration rate of <50 mL/min, were also ineligible. Additional conditions for exclusion included New York Heart Association (NYHA) Class III–IV heart failure classification; diagnoses of hypopituitarism or adrenal insufficiency; a history of major surgical procedures, severe infections, or severe trauma within the past 6 months; a history of malignancy within the past 5 years; and the presence of active bladder cancer. Individuals with rare hereditary problems, such as galactose intolerance, Lapp lactase deficiency, or glucose-galactose malabsorption, were deemed unsuitable for this study. A documented history of hypersensitivity to alogliptin, pioglitazone, or any of their components also led to exclusion, as did heavy alcohol consumption patterns (≥30 g daily, ≥5 days a week).

The study protocol stipulated that if a participant exhibited an HbA1c level >9% and demonstrated symptoms of hyperglycemia at the 12-week visit, their involvement in the clinical trial would be discontinued, and rescue medications would be prescribed to maintain glucose control.

This study aimed to assess the efficacy and safety of alogliptin and pioglitazone as combination therapy compared to each drug individually in managing patients with T2DM who failed to achieve glycemic control with metformin monotherapy. The principal efficacy endpoint was the change in HbA1c level from baseline to week 24. Durability was assessed by comparing the number of participants who attained their target HbA1c levels (<7.0% or <6.5%) across each group at the end of 24 weeks. Secondary endpoints included changes in BMI, insulin sensitivity, β-cell function, glycoalbumin (GA), and the GA/HbA1c ratio to evaluate glucose variability as well as fasting insulin and glucose levels from baseline to week 24. The incidence rate of rescue treatments within the 24-week period was also recorded. Our exploratory data analysis evaluated the changes in lipid profile parameters, including total cholesterol, triglycerides, high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), free fatty acid (FFA), and apolipoprotein B, from baseline to week 24. Safety endpoints included the incidence rates of hypoglycemia, severe hypoglycemia (defined as hypoglycemia requiring the intervention of another person to elevate glucose levels and promote neurological recovery), and adverse events of special interest. These include heart failure, cardiovascular effects, edema, weight gain, bladder tumors, macular edema, bone fractures, pancreatitis, hypersensitivity to alogliptin or pioglitazone, and serious adverse events. All adverse events were meticulously monitored and documented by an investigator who adjudicated the severity and potential relationship with the medication. Efficacy at 12 and 24 weeks was evaluated using standard laboratory measurements conducted at a central laboratory (Meditree Central Lab Service, Seoul, Korea).

Baseline characteristics are expressed as mean±standard deviation for normally distributed variables or as median (interquartile range) for non-normally distributed variables. Changes from baseline in efficacy variables were analyzed using the analysis of covariance model, with treatment as a fixed factor and the baseline value serving as a continuous covariate. The primary analysis compared the least-squares mean change in HbA1c levels from baseline to week 24 between the treatment groups. The Tukey-Kramer adjustment was used for post hoc analyses, with the significance level set at α<0.05. Intra-group differences in measurements from baseline to weeks 12 and 24 were assessed using the paired t-test or Wilcoxon signed-rank test. The proportion of patients achieving HbA1c levels <7.0% and <6.5% at week 24 was analyzed using the chi-square test, depending on the data distribution. Missing values were extrapolated using the last observation carried forward in all the efficacy analyses. The study was designed to include a randomized sample size of 100 patients per treatment arm, providing 90% power to detect a 0.48% difference in HbA1c change (with an assumed standard deviation of 0.73%) between the Alo+Pio group and either the Alo or Pio group at a two-sided significance level of 0.05. This calculation accounts for an anticipated dropout rate of 20%. Randomization of the study groups was stratified according to participating sites and baseline HbA1c levels (<8.5% or ≥8.5%). All statistical analyses were performed using the SAS software version 9.4 (SAS Institute, Cary, NC, USA). Statistical significance was set at P<0.05.

Of the 216 eligible participants, 214 were randomized into three groups (Alo group, n=75; Pio group, n=69; Alo+Pio group, n=70) (Supplementary Figs. 1 and 2). The baseline characteristics of the study participants measured at the randomization point are described in Table 1. The average age of the study population was 58.9±9.2 years. The mean HbA1c was 8.0%±0.4%, and the average BMI was calculated to be 25.4±3.3 kg/m². The mean duration of T2DM was 9.1±5.7 years. There were no significant differences in mean age, sex distribution, BMI, HbA1c levels, blood pressure, fasting glucose, or lipid profiles among the three groups after randomization. There were no significant differences in alcohol consumption, smoking history, diabetes duration, physical examination, hypertension, dyslipidemia, other comorbidities, medication, and medical compliance (Table 1).

Over the study period, the average HbA1c level significantly decreased in all groups, with the Alo+Pio group showing the greatest reduction. From baseline to week 12, HbA1c decreased by –1.14%±0.2%, –0.92%±0.64%, and –0.63%±0.8% in the Alo+Pio, Alo, and Pio groups, respectively. This trend persisted through week 24, with the Alo+Pio group showing a further decrease to –1.38%±0.08%, while the Alo and Pio groups reached –1.03%±0.08% and –0.84%±0.08%, respectively (Fig. 1A and B). At week 24, 71.43% of the participants in the Alo+Pio group achieved HbA1c levels below 7.0%, significantly higher than that in the Alo (54.67%) and Pio (37.68%) groups (Fig. 1C). Similarly, the Alo+Pio group had a higher proportion of participants (44.29%) achieving an HbA1c level of less than 6.5% than the Alo (17.33%) and Pio groups (17.39%) (Fig. 1D). Furthermore, when comparing the rate of change (%) in HbA1c at the 12- and 24-week intervals across the groups, the Alo+Pio group demonstrated a more pronounced, continuous decrease in HbA1c. The number of participants achieving the target HbA1c level of less than 6.5% was significantly higher in the triple therapy group than in the other groups. None of the participants required hyperglycemic rescue treatment, a provision intended for individuals with HbA1c levels exceeding 9% at the 12-week checkpoint. In summary, the Alo+Pio combination was the most effective at reducing HbA1c levels, ensuring that a larger proportion of participants achieved the target glycemic range.

We further explored whether triple or dual combination add-on OHA regimens changed the metabolic parameters in patients with T2DM. The mean body weight and BMI in the Alo+Pio group (0.63±0.93 kg/m²) increased significantly at 24 weeks from baseline compared to the Alo group (–0.05±0.74 kg/m²), with no discernible difference when juxtaposed with the Pio group (0.50±0.78 kg/m²) (P<0.0001). The Alo+Pio cohort witnessed a substantial reduction in the mean change of fasting glucose (–39.00±3.21 mg/dL) when contrasted with the Alo (–25.57±3.10 mg/dL) and Pio (–25.43±3.22 mg/dL) groups, a difference that was statistically significant at 24 weeks (P<0.0001). As for fasting insulin, a notable decrease was identified in the Pio (–1.54±0.28 μU/mL) and Alo+Pio (–1.4±0.28 μU/mL) but not in Alo (0.02±0.27 μU/mL) groups (P<0.0001). Homeostatic model assessment for insulin resistance (HOMAIR) was significantly decreased in the Pio (–0.97±0.12) and Alo+Pio (–1.00±0.12) group, underscoring the benefits of pioglitazone’s efficacy in ameliorating insulin resistance. With regard to β-cell function, there was a significant increase of homeostatic model assessment for β-function (HOMA-β) in the Alo (9.00±1.66) and Alo+Pio (6.01±1.72) groups which reached a statistically significant difference from the Pio group (1.00±1.74, P=0.0042). The mean change in GA levels from baseline to 24 weeks demonstrated a significant drop in the Alo+ Pio group than that in the Pio group (P=0.0005). However, the change in GA/HbA1c ratio was not significantly different among the three groups (P=0.1610) (Table 2). Overall, the combination therapy of Alo and Pio concurrently enhanced insulin resistance and β-cell functionality in patients with T2DM.

Previous studies suggested a possible correlation between thiazolidinedione, DPP-4 inhibitor, and lipid metabolism [15-17]. Based on these previous reports, we further studied whether different alogliptin and/or pioglitazone regimens could change the lipid profiles in patients with T2DM. Upon evaluating the lipid profile changes at 24 weeks relative to the baseline, a trend towards a decrease in total cholesterol was observed exclusively in the Alo group; however, this trend was not statistically significant across the three groups. For triglycerides, the Pio (–37.09±11.22 mg/dL) and Alo+Pio (–36.33±11.15 mg/dL) groups showed a greater reduction than the Alo group (–4.19±10.73 mg/dL, P=0.0520). HDL-C levels were more increased in the Alo+Pio group (6.34±1.12 mg/dL) than in the Alo group (–2.06±1.08, P=0.0001) and were similar to those in the Pio group (6.23±1.13 mg/dL, P=0.9970). LDL-C levels tended to decrease in the Alo group, although no significant differences were detected between the groups. FFA changes were more substantial in the Pio (–176.54±30.06 μEq/L) and Alo+Pio (–147.02±29.99 μEq/L) groups compared to the Alo group (46.87±28.79 μEq/L) (P=0.0051). Apolipoprotein B levels showed a downward trend in all three groups, with the Alo+Pio group experiencing the greatest decrease, albeit without significant differences among the groups. In short, the triple therapy by the dual addition of alogliptin and pioglitazone exhibited benefits in lipid profile alterations comparable to the effects observed in dual therapy with the addition of pioglitazone alone.

To further elucidate who may benefit from early combination (triple) therapy, we analyzed the baseline characteristics of patients who failed to achieve the target glycemic level (HbA1c ≥6.5%) with alogliptin or pioglitazone add-on therapy (Table 3). In comparison to those who achieved the target HbA1c, patients who failed to respond to alogliptin mono add-on therapy exhibited higher HbA1c levels (7.61%±0.64% vs. 7.99%±0.68%, P=0.0222), lower HOMA-IR values (3.05±2.15 vs. 2.23±1.61, P=0.0678), and lower HOMA-β values (31.55±17.93 vs. 23.37±13.41, P=0.0678). Similarly, patients who did not achieve the target HbA1c with pioglitazone mono add-on therapy demonstrated higher HbA1c levels (7.48%±0.34% vs. 8.19%±0.79%, P=0.0055), fasting blood sugar (FBS; 138.33±15.67 mg/dL vs. 163.65±44.22 mg/dL, P=0.0366), lower fasting C-peptide levels (1.97±0.56 ng/dL vs. 1.75±0.94 ng/dL, P=0.0469), and lower HOMA-β values (37.28±12.95 vs. 24.99±15.48, P=0.0039). Additionally, patients who failed on pioglitazone monotherapy had a lower BMI (25.43±3.28 kg/m²) compared to those who achieved the target glycemic level (27.17±3.06 kg/m²), although this difference did not reach statistical significance (P=0.0595). No significant differences were found in age, sex, duration of T2DM, metformin dose, hypertension, dyslipidemia, smoking status, alcohol consumption, GA, and GA/HbA1c ratio between those who succeeded or failed on alogliptin or pioglitazone mono add-on therapy.

On the contrary, there was no difference in FBS (163.71±35.33 mg/dL vs. 166.90±38.53 mg/dL, P=0.6362), HOMA-IR (2.39±0.98 vs. 2.28±1.31, P=0.2820), or HOMA-β (24.38±15.82 vs. 21.87±14.46, P=0.3149) between those who succeeded or failed to achieve the target HbA1c with the alogliptin and pioglitazone triple combination therapy. To sum up, high FBS, low HOMA-IR level, and low HOMA-β levels are the characteristics of patients who fail on alogliptin or pioglitazone mono add-on therapy.

During the study period, the incidence of adverse events varied across the three groups: 34.67% (42 cases) in the Alo group, 23.19% (31 cases) in the Pio group, and 32.86% (33 cases) in the Alo+Pio group. However, the difference was not statistically significant, and no adverse reactions were attributed to the clinical trial treatment. Hypoglycemia incidence during the trial was low and similar across groups, as reported by one (1.33%) participant in the Alo group and one (1.43%) participant in the Alo+Pio group. Importantly, no severe hypoglycemic events were reported. Serious adverse events were reported in 7.14% (six cases) of the participants in the Alo group, 2.37% (two cases) in the Pio group, and 2.9% (two cases) in the Alo+Pio group. One patient in the Alo+Pio group experienced acute pancreatitis and acute cholecystitis due to biliary stones but was fully treated. Two cases of ureteral cancer and pruritus were potentially linked to drug usage, whereas the other cases were deemed unlikely or impossible to evaluate for drug association. Adverse events of special interest, such as cardiovascular events, were reported in 1.45% of the Pio group and 1.43% of the Alo+Pio group. Edema occurred in six participants: one (1.33%) in the Alo group, three (4.35%) in the Pio group, and two (2.86%) in the Alo+Pio group. However, these incidents did not differ significantly between the groups (Table 4). In summary, the combined use of Alo and Pio did not significantly increase the incidence of adverse events.