Study Design
At baseline we randomized the patients to receive either metformin or troglitazone for a 14-week period. The study was divided into 2 phases: a 2-week dose-titration period and a 12-week open-label comparison of metformin and troglitazone. If randomized to metformin, the patient took 500 mg with the evening meal for 2 days, then 500 mg twice daily with the morning and evening meals for 5 days. During the second week of the study, the patient took 500 mg with the morning meal and 1000 mg with the evening meal. After week 2, all patients randomized to metformin therapy were taking 1000 mg with the morning and evening meals. Patients randomized to troglitazone took 200 mg daily with the evening meal for 2 weeks and then 400 mg daily for the remaining 12 weeks of the study. If at any time during the study a patient experienced a FPG of less than 80 mg/dL, the oral sulfonylurea was decreased by one half of the original dose and then discontinued if further blood glucose readings were less than 80 mg/dL on more than one occasion.
Patients were required to make a screening visit at least 1 week before entry into the trial. We obtained a past medical history, body weight and height, and blood tests (serum creatinine, serum bicarbonate, liver enzymes, Hb A1c, FPG, and C-peptide levels). We recorded body weight and repeated blood tests 6 to 8 weeks after randomization and at the end of the study period (14 weeks). In addition, participants randomized to troglitazone had monthly liver enzyme tests. We also instructed patients to perform home blood glucose monitoring twice daily, in the morning before breakfast and at bedtime. We validated blood glucose monitors for accuracy by checking control solutions and performing check strip tests at study visits. Patients received standardized information about diabetes from a certified diabetes educator consisting of a general overview of diabetes mellitus, a medication review, instructions for blood glucose monitoring, a review of complications associated with diabetes, and nutritional advice. Each patient was given the same written information about diabetes and counseled on the signs and symptoms of high and low blood glucose. No patient enrolled in this trial reported problems reading or understanding written instructions. We assessed compliance using pill counts during each scheduled follow-up visit. We asked patients about adverse events at each visit after beginning drug therapy; any reported events were recorded.
Statistical Analysis
We performed an analysis of efficacy by intention to treat. We included all patients who received at least one dose of troglitazone or metformin, and selected a sample size adequate for detecting clinically meaningful differences in treatment effects. A sample of 16 patients in each group allowed detection of a mean absolute difference in Hb A1c level reduction between groups from a baseline of 1.2% (±0.2%) with a power greater than 0.80 (a = 0.05, 2-tailed test). We evaluated baseline differences between treatment groups using analysis of variance and chi-square procedures. Paired t tests were used to examine changes in variables over time. Analyses were performed using the Statistical Package for the Social Sciences for Personal Computers (Version 8.0, SPSS, Inc, Chicago, Ill).
Results
The baseline demographic and disease-related characteristics of the participants are outlined in Table 1. There were no significant differences at baseline between treatment groups with respect to age; body mass index (BMI); Hb A1c, FPG, or C-peptide levels; or the duration of diabetes. Ninety-seven percent of the patients took their assigned medication for the 14 weeks of the study. All patients were receiving an oral sulfonylurea, with 85% taking glipizide (Glucotrol XL) 10 mg to 20 mg per day, 9% taking glimepiride (Amaryl) 4 to 8 mg per day, and 6% taking glyburide (generic or DiaBeta) 10 to 20 mg per day.
Table 2 contains the changes in glycemic control parameters observed in each treatment group. At the end of a 3-month treatment period, Hb A1c values decreased significantly for each group when compared with the values obtained at baseline. The mean Hb A1c level among those receiving metformin fell from 9.9% ±1.6 to 7.8% ±1.3 (P <.001). Among patients in the troglitazone treatment group, the mean Hb A1c level fell from 10% ±1.6 to 7.4% ±1.7 (P <.001). The mean FPG level fell from 229 mg/dL ±75 to 138 mg/dL ±36 (P <.001) in the patients receiving metformin and from 210 mg/dL ±79 to 127 mg/dL ±33 (P <.001), in those receiving troglitazone. For each treatment group this amounts to a 60% reduction in FPG levels. For those patients receiving metformin, fasting C-peptide levels fell from 6.9 ng/mL ±2.3 to 4.7 ng/mL ±1.6 (P <.001), and in troglitazone-treated patients, it fell from 6.5 ng/mL ±3.9 to 4.5 ng/mL ±2.3 (P = .004). Although both troglitazone and metformin significantly improved glycemic control, there was no significant difference between the 2 groups in any treatment effect measured. In addition, BMI did not significantly change from baseline to the end of the study in either treatment group.