This randomized, single- and double-blind, three-arm parallel-group, placebo-controlled trial was conducted in 26 study centers in the U.S. and Canada. Institutional review boards or independent ethics committees for each center approved the protocol. All patients provided written informed consent. The trial consisted of a 10- to 21-day qualification period, 12-week treatment phase, and 4-week follow-up phase. Starting at day −7, patients were instructed on a diet and exercise program, pursuant to American Diabetes Association or similar local guidelines, to be followed throughout the study. Until day 1, patients maintained their stable dose of OADs and insulin.
We used an adaptive trial design with two cohorts. The purpose of the first cohort was to identify a reduced insulin starting dose unlikely to cause hypoglycemia after addition of dapagliflozin. Four patients received single-blind 20 mg dapagliflozin after having their daily insulin dose decreased by 50%. If at least one patient recorded a glucose value ≤100 mg/dl in this cohort, lesser dose reductions (i.e., 30 or 10%) would not be tested, and the daily insulin dose reduction for patients in the larger second cohort would be set at 50%. This was the case, and in the second treatment cohort, patients were randomly assigned 1:1:1 on day 1 to double-blind placebo, 10 mg dapagliflozin, or 20 mg dapagliflozin once daily, in addition to open-label therapy with 50% of their usual daily insulin dose and their OAD(s).
Patients performed self-monitoring of blood glucose five times daily during the 3–5 days before clinic visits at weeks 1, 2, 4, 6, 8, 10, and 12. No dose modifications of blinded study medication or OAD(s) were allowed during the treatment phase. In patients with or at risk of hypoglycemia, insulin could be down-titrated for self-monitored blood glucose levels <54 mg/dl or mean daily glucose <100 mg/dl or when clinically necessary as determined by the investigator. Patients experiencing major hypoglycemia were discontinued from the study. For any fasting plasma glucose (FPG) level >240 mg/dl at weeks 4 and 6, >220 mg/dl at week 8, or >200 mg/dl at week 10, the insulin dose could be increased after a retest. Patients lacking glycemic control despite up-titration or whose modified insulin dose exceeded baseline were discontinued from the study.
Men and women with type 2 diabetes, aged 18–75 years, with BMI ≤45 kg/m2 and A1C 7.5–10%, were enrolled between October 2006 and November 2007. Patients were receiving stable-dose insulin sensitizer therapy (metformin ≥1,000 mg and/or pioglitazone ≥30 mg or rosiglitazone 4 mg) for ≥6 weeks and insulin therapy for ≥12 weeks before enrollment (insulin dose must have been ≥50 units of U100 daily and stable for ≥6 weeks). Laboratory criteria included fasting C-peptide ≥0.8 ng/ml, serum creatinine <1.5 mg/dl (men) or <1.4 mg/dl (women), and a urine microalbumin-to-creatinine ratio <300 mg/g or, if exceeded on spot check, a 24-h urine total protein <3 g/24 h. Major exclusion criteria were a history of type 1 diabetes, aspartate transaminase and/or alanine transaminase >2.5 times the upper limits of normal, creatine kinase ≥3 times the upper limits of normal, symptoms of severely uncontrolled diabetes, a history of severe hypoglycemia, and unstable condition or serious cardiovascular, renal, or hepatic disease.
Trial outcomes (dapagliflozin vs. placebo)
The primary efficacy measure was change from baseline in A1C at week 12 (last observation carried forward [LOCF]). Secondary efficacy measures at week 12 (LOCF) included changes from baseline in FPG and total daily dose of insulin (TDDI), the proportion of patients achieving a decrease in A1C ≥0.5% from baseline, and the proportion of patients achieving A1C <7%. Tertiary end points included changes from baseline in total body weight and in postprandial glucose (PPG) measured by an oral glucose tolerance test. Safety outcomes were assessed by treatment-emergent adverse events, vital signs, and laboratory measurements, including 24-h urine collections for volume and electrolytes.
For the treatment cohort, the sample size target of 22 patients per treatment group was chosen to allow for the calculation of 95% CI for the primary end point with a half-width of 0.42% for each treatment group, assuming a 1% SD; the half-width of a 95% CI for differences between mean treatment changes was estimated to be 0.59%. The primary efficacy dataset consisted of all randomly assigned patients who took ≥1 dose of double-blind study medication. Analyses of efficacy variables (except change from baseline in insulin dose) excluded data after insulin up-titration. Analyses for change from baseline in A1C, FPG, insulin dose, and total body weight at week 12 (LOCF) were performed using an ANCOVA model with treatment group as effect and baseline value as a covariate. No statistical hypothesis testing was planned for this study designed for exploratory analysis.