Trial Design and Oversight
We conducted this phase 3 trial (SEED) from July 18, 2017 to March 6, 2020, at 40 sites in China, in Chinese drug-naïve type 2 diabetes patients. The SEED trial had a 2-week screening period, a 4-week single-blind placebo run-in period, and then eligible patients were randomized at a 2:1 ratio into the dorzagliatin and placebo group in a 24-week double-blind, placebo-controlled treatment period, followed by an extended 28-week open-label treatment period with all patients receiving dorzagliatin orally, and a 1-week follow-up period without treatment (Fig. 1a).
The trial was conducted in accordance with the principles of Declaration of Helsinki, Good Clinical Practice guidelines, and laws and regulations in China. The study protocol was amended once during study enrollment, and the important changes to methods after trial commencement and reasons were summarized in the supplemental material. The trial protocol and amendments were approved by the local ethics committees of all study sites. Written informed consent was obtained from all patients before the initiation of any trial-related procedures. The trial was also conducted in accordance with the Chinese Diabetes Society guidelines, which requires physicians to educate and strictly enforce improved exercise and dietary control, as well as self-monitoring of blood glucose (at least 2 times per week), in treating type 2 diabetes patients.
Patients were recruited at the study site by investigators. Patients were required to sign the informed consent form before enrollment. After providing written informed consent, patients were screened for eligibility. The individuals (18 to 75 years of age) were eligible for inclusion in the trial if they met the following criteria: diagnosed with type 2 diabetes, on diet and exercise interventions for at least 3 months and had no antidiabetic therapy before screening, had HbA1c levels between 7.5 and 11.0%, and had a body-mass index between 18.5 and 35.0 kg per square meters at screening.
Individuals were excluded if they had severe hypoglycemia without cause within 3 months, or frequent hypoglycemia occurred no less than 3 times within 1 month before screening, fasting C-peptide level of less than 1.0 ng per milliliter at screening, a history of type 1 diabetes. Other exclusion criteria were the major cardio-cerebrovascular diseases within 6 months before screening; unstable or rapidly progressive kidney disease; active liver diseases; psychiatric disease; any type of malignancy; an alanine or aspartate transaminase level greater than 2.5 times of the upper limit of the normal range, or serum total bilirubin level greater than 1.5 times of the upper limit of the normal range. Complete lists of inclusion, exclusion, and randomization criteria are provided in the protocol as a supplemental material.
After screening, eligible patients entered a 4-week, single-blind placebo run-in period. At week 3 of run-in period, patients were reevaluated before randomization to confirm the eligibility. Eligibility criteria included HbA1c levels between 7.5 and 10.0%, and fasting plasma glucose levels between 7.0 and 13.3 mmol per liter at randomization. Diet and exercise counseling was provided during the entire experiment.
Eligible patients were randomly assigned, in a 2:1 ratio, to receive dorzagliatin or placebo for 24 weeks, on a background of a diet and exercise regimen. Randomization and drug dispense were performed with an interactive web response system (IWRS) (Balance, MedData Solution). The random allocation sequences were generated by the IWRS. A stratified randomization method with the permuted block randomization algorithm was used. The blocks were dynamically allocated to each site and stratum from the randomization list. A unique identification number (drug box serial number) was provided by the drug packing vendor and marked on the label of drug box. By central randomization, the randomization codes were assigned to subjects who meet randomization criteria by IWRS system based on randomization factors (baseline HbA1c level ≤ 8.5% or >8.5%) and the block size. Then, the randomization codes along with the corresponding drug box numbers were provided to subjects who meet the randomization criteria at each visit.
Investigators enrolled participants and used IWRS to assign participants to interventions. Double-blind method was used in the study. The random allocation sequences were concealed from patients, investigators and all other study members. The placebo tablets have the size, color, odor, and appearance same as the investigational drug. Blinding was maintained until the end of week 24. All doses of dorzagliatin 75 mg tablets and placebo tablets were administered orally twice daily, as single tablet.
All the patients who completed a 24-week double-blind treatment (dorzagliatin or placebo), received dorzagliatin treatment for a 28-week open-label period, during which patients assigned to dorzagliatin continued their treatment unchanged, whereas those assigned to placebo were switched from the placebo to dorzagliatin. After the 52-week treatment, all patients were observed for 1 week for safety evaluation.
The primary efficacy endpoint was the change from baseline in the HbA1c level at week 24. Key secondary efficacy endpoints included the changes from baseline in 2-hour postprandial glucose, fasting plasma glucose levels at week 24, and HbA1c level at each visit (except week 24); the percentage of patients who reached a HbA1c level of less than 7.0% at week 24. Additional efficacy endpoints included homeostatic control rate (a HbA1c level of less than 7.0% and without hypoglycemia), as well as the change from baseline in the homeostasis model assessment 2-β (an index for β-cell function). A full list of other efficacy endpoints is provided in the statistical analysis plan as the supplemental material.
Adverse events, serious adverse events were assessed throughout the trial (definitions are provided in the protocol as a supplemental material). Hypoglycemic episodes were classified according to the American Diabetes Association definitions (see the protocol provided as a supplemental material). Vital signs and clinical laboratory test results were assessed, and physical examinations were performed.
We hypothesized that dorzagliatin would show superiority to placebo in the decrease of HbA1c level in patients after 24-week treatment. For the primary endpoint of the change from baseline in the HbA1c level at week 24, we calculated that a total sample size of 450 patients would provide the trial with 91.4% power to detect a difference of 0.4 percentage points between the dorzagliatin group and the placebo group in a 2:1 ratio of allocation at a 2-sided significance level of 0.05, assuming a standard deviation (SD) of 1.2 percentage points. The full analysis set included all randomized patients, who took at least one dose of study drugs and had at least one post-treatment measurement of the primary endpoint during double-blind treatment period. The safety set included all randomized patients, who took at least one dose of study drugs. Since this trial had only one confirmatory hypothesis test, i.e. the test of null hypothesis of no difference in the primary endpoint between dorzagliatin and placebo, there was no adjustment for multiplicity.
The primary analysis method for the primary efficacy endpoint was in the full analysis set using a mixed model for repeated measures without missing value imputation, which included treatment group, scheduled visit, interaction of treatment group with scheduled visit, pooled site and HbA1c baseline value as fixed effects. In the mixed model for repeated measures, data collected after initiation of prohibited antidiabetic medications (the amount and time of taking were reviewed in blind-data-review meeting and open-data-review meeting) were handled as missing data. The least-squares means of each treatment group, the estimated treatment difference between dorzagliatin and placebo and its 95% confidence interval (CI) were calculated. A sensitivity analysis was performed with the primary efficacy endpoint in the full analysis set using an analysis of covariance model with factors of treatment group, pooled site and interaction of treatment group with pooled site (interaction effect removed if not significant on a 2-sided alpha level of 0.1), and a covariate of HbA1c baseline value. In the analysis of covariance model, missing values were imputed using the method of last observation carried forward, which was applied after excluding data collected after initiation of prohibited antidiabetic medications. Another sensitivity analysis was performed with the primary efficacy endpoint using the same mixed model for repeated measures in the per-protocol set.
The secondary efficacy endpoints including changes from baseline in 2-hour postprandial glucose, and fasting plasma glucose levels at week 24 were assessed using the same mixed model for repeated measures for the primary endpoint. Another secondary efficacy endpoint of the percentage of patients who reached a HbA1c level of less than 7.0% and additional efficacy endpoint of the homeostatic control rate were estimated based on the data imputed by last observation carried forward approach in the full analysis set. Odds ratio and 95% confidence interval between two treatment groups were estimated using the logistic regression model with categorical independent variables of treatment group, pooled site and interaction of treatment group with pooled site (interaction effect removed from the model if not significant at a 2-sided alpha level of 0.1), and a continuous independent variable of baseline HbA1c level. For other efficacy endpoints, such as the change from baseline in the homeostasis model assessment 2-β were analyzed using the same analysis of covariance model as described above.
The post-hoc subgroup analyses were carried out with the same statistical methods as used for primary efficacy outcome. Patients were stratified into subgroups according to their HbA1c level at baseline (less than or equal to 8.0% vs greater than 8.0%). The analysis of covariance model was used to investigate the treatment effect on homeostasis model assessment 2-β. Kaplan-Meier method was employed to analyze the time to first reach target response (a HbA1c level of less than 7.0%) over 24 weeks.
The incidence of treatment-emergent adverse events that occurred between the first intake of double-blind study medications and the 7th day after the last dose of study medications was summarized by treatment group and compared between two treatment groups. The incidence of hypoglycemic events was also summarized by treatment group and compared between treatment groups, both patients that spontaneously reported low glucose events and low glucose events reported at on-site visit were counted as hypoglycemic events. Additional details regarding the statistical analysis are provided in the statistical analysis plan as a supplemental material.