Effect of Metformin and Rosiglitazone Combination Therapy in Patients With Type 2 Diabetes Mellitus:  A Randomized Controlled Trial FREE

Type 2 diabetes is characterized by decreased insulin secretion^1- 2 and insulin sensitivity in liver, adipose tissue, and skeletal muscle. Together these abnormalities confound efforts to treat diabetes because most antidiabetic agents target only 1 underlying cause of the disease. Approximately 50% of patients treated with monotherapy require additional therapy to achieve target glycosylated hemoglobin (HbA_1c) levels 3 years after diagnosis.³

Rosiglitazone maleate, a member of the thiazolidinedione class of antidiabetic agents that was recently approved by the US Food and Drug Administration, targets insulin resistance by binding to the transcription factor peroxisome proliferator-activated receptor-γ, promoting synthesis of glucose transporters and activating adipocyte differentiation.^4- 6 In contrast, metformin hydrochloride promotes glucose lowering by reducing hepatic glucose production and gluconeogenesis and by enhancing peripheral glucose uptake.^7- 10

Because metformin and rosiglitazone act through different mechanisms, their combined use may be indicated in patients whose disease is poorly controlled with a maintenance dose of metformin. This study evaluated the efficacy and safety of adding 4 mg/d and 8 mg/d of rosiglitazone maleate to maximal-dosage of metformin in patients with poorly controlled type 2 diabetes. Combined efficacy was assessed by comparing the level changes in HbA_1c, fasting plasma glucose (FPG), fructosamine, serum insulin, free fatty acids (FFA), lipids, lactate, and estimates of insulin sensitivity and β-cell function (BCF) between combined metformin-rosiglitazone treatment and metformin-placebo alone.¹¹

To detect a 0.75% absolute difference in HbA_1c between treatment groups, 65 evaluable patients per group would be required to achieve a power of 95%. Planned enrollment was 280 patients (approximately 93 per group). Persons between the ages of 40 and 80 years with type 2 diabetes as defined by the National Diabetes Data Group¹² with FPG concentrations of between 7.8 and 16.7 mmol/L (140 and 300 mg/dL) at screening and during the placebo-maintenance period while taking 2.5 g/d of metformin were eligible. All patients demonstrated insulin secretory capacity as determined by a fasting C-peptide concentration of 0.27 nmol/L (0.8 ng/mL) or more at screening. Subjects were required to have a body mass index, calculated as weight in kilograms divided by the square of height in meters, of 22 to 38 and a weight change of no more than 10% between screening and baseline.

Patients were excluded if they had clinically significant renal or hepatic disease, angina, New York Heart Association Classification class III or IV cardiac insufficiency, symptomatic diabetic neuropathy, significant clinical abnormality on electrocardiogram, abnormal laboratory test results (blood chemistry, hematology, or urinalysis), use of chronic insulin therapy, participated in any rosiglitazone-related study, or used any investigational drug (excluding metformin) within 30 days of study (or 5 half-lives of the investigational drug, if longer than 30 days). Anorectic agents were discontinued at least 30 days before screening. Patients with hyperlipemia, elevated cholesterol or triglyceride levels, or lipid metabolism disorders were eligible; lipid-lowering agents were maintained at the same dosage level throughout the study.

This multicenter, randomized, double-blind, placebo-controlled trial was conducted at 36 sites in the United States between April 1997 and March 1998. Before the study, patients discontinued all antihyperglycemic medications, with the exception of metformin. Metformin dose tolerability was determined during a 3-week period in which metformin was titrated to 2.5 g/d; afterward, patients entered a 4-week, single-blind metformin-placebo maintenance period with a weight-maintenance diet. During this maintenance period, only investigators were aware that patients were receiving the metformin-placebo treatment. Patients previously treated with metformin at 2.5 g/d proceeded directly to maintenance; thus, with the exception of metformin, patients refrained from medication for a minimum of 4 weeks and a maximum of 7 weeks.

At the end of the maintenance period, patients with inadequate glycemic control (FPG concentration range, 7.7-16.7 mmol/L [140-300 mg/dL]) were randomly assigned (1:1:1 ratio) to receive double-blind metformin treatment in 1 of 3 combinations: placebo (control), 4 mg of rosiglitazone, or 8 mg of rosiglitazone once daily for 26 weeks. Randomization was computer generated with a fixed block size. No patient, investigator, or sponsor was aware of treatment allocation until study completion (Figure 1).

This study was conducted in accordance with the Declaration of Helsinki (as amended, 1989), Title 21 of the US Code of Federal Regulations, and Good Clinical Practice guidelines. The institutional review board at each center approved the protocol, and subjects provided informed consent before enrollment.

Laboratory measurements for efficacy and safety were performed by SmithKline Beecham Clinical Laboratories (Van Nuys, Calif) on blood collected in the fasting state. Fasting plasma glucose concentrations, total cholesterol, high-density lipoprotein cholesterol (HDL-C), and triglyceride levels were measured by an Olympus analyzer (Olympus Clinical Instruments Division, Lake Success, NY); levels of HbA_1c were measured by the high-performance liquid chromatography method (Variant, Bio-Rad, Hercules, Calif); C-peptide by radioimmunoassay (Diagnostic Products, Los Angeles, Calif); insulin by radioimmunoassay (Linco Research Inc, St Charles, Mo); fructosamine by colorimetric analysis (RoTAG fructosamine assay, Roche Diagnostic Systems, Indianapolis, Ind); and FFA by enzymatic/colorimetric analysis (Wako Diagnostic, Richmond, Va) using a COBAS analyzer (Roche Diagnostic Systems). Low-density lipoprotein cholesterol (LDL-C) concentrations were estimated from total cholesterol and HDL-C determinations using the Friedewald calculation.¹³ Lactate was measured by enzymatic spectrophotometric analysis using an Olympus analyzer (Olympus Clinical Instruments Division).

Estimates of insulin sensitivity determined by homeostasis model assessment (HOMA-S) and BCF (HOMA-B) were calculated using FPG and immunoreactive insulin values, or C-peptide levels. HOMA is a mathematical model based on glucose and insulin interaction in different organs, including the pancreas, liver, and peripheral tissues.¹¹ HOMA estimates of BCF and insulin sensitivity were calculated for each participant's FPG and insulin, or C-peptide levels, and expressed relative to values in a lean, nondiabetic reference population aged 18 to 25 years.^14- 16 HOMA-S determinations of insulin sensitivity or insulin resistance have been validated by comparison with results of glucose clamp studies,^11,14 intravenous glucose tolerance tests,^11,15 and continuous infusion of glucose with model assessment.¹⁵ The HOMA-B method has been validated by comparison with the intravenous glucose tolerance test and continuous infusion of glucose model assessment.¹⁷ Application of HOMA has also been used in epidemiological studies.^18- 19

Safety monitoring included physical examination, vital sign assessment, weight measurement, electrocardiogram, adverse experience query, and laboratory tests.

The primary population for efficacy analysis was the intention-to-treat population, those with at least 1 value while receiving therapy (last observation was carried forward in the case of missing data or early withdrawals). Efficacy and safety parameters were measured at baseline and after 26 weeks of treatment. Safety parameters were assessed based on week 26 data (without the last observation carried forward).

Treatment groups were compared using analysis of covariance with terms for baseline, treatment, and center. The assumptions of the statistical model were tested before application. The Levene test of heterogeneity across treatments was applied at a significance level of α = .01. If significant, the Shapiro-Wilk test of nonnormality (α = .01) was examined. Parametric analysis or nonparametric analysis was used, depending on results of test assumptions. If prospectively defined assumptions for parametric analysis were not met, the Wilcoxon rank sum test was used. Pairwise comparisons to placebo used Dunnett multiple comparison procedure to maintain a 2-sided .05 significance level within each parameter. The statistical significance of the within-group change from baseline was tested by a paired t test or a signed rank test. Safety parameters, including clinical laboratory tests, vital signs, and body weight, were examined using 1-way analysis of variance. Statistical analyses were performed using statistical software (SAS/STAT Software, Release 6.12, SAS Institute Inc, Cary, NC).

Of 443 patients screened, 437 entered the titration and maintenance period and 348 were randomized to treatment (Figure 1). Most withdrawals were due to failing to meet inclusion criteria (69.7%). Baseline characteristics were similar among treatment groups (Table 1). Fifty-eight patients withdrew before completion of the double-blind phase: 22 from the placebo group and 18 from the 4-mg/d and 18 from the 8-mg/d rosiglitazone groups. Most participants withdrew because of adverse experiences or lack of efficacy (Figure 1).

The mean HbA_1clevels decreased significantly from baseline in a dose-dependent fashion in both rosiglitazone groups by 0.56% in the 4-mg/d and by 0.78% in the 8-mg/d rosiglitazone groups. But the control group experienced a significant increase in HbA_1c levels (0.45%) (Figure 2). Furthermore, both rosiglitazone groups had HbA_1c levels lower than those in the control group by 1.0% in the 4-mg/d and 1.2% in the 8-mg/d rosiglitazone groups. In contrast to results observed in the control group, the mean HbA_1c levels in the rosiglitazone groups decreased after week 4 and plateaued by week 18 (Figure 3). The percentage who achieved a 1.0% reduction in HbA_1c concentrations was 32.8% in the 4-mg/d and 37.3% in the 8-mg/d rosiglitazone groups and 7.1% in the control group.

Twenty-five (28.1%) of 89 patients taking 8 mg/d of rosiglitazone achieved the target HbA_1c control levels of 7.0%, and 51 patients (57.3%) in the same group achieved HbA_1c levels of 8.0%, or below the American Diabetes Association action point. Yet only 7.6% of the patients in the control group achieved HbA_1c levels of 7.0% and 35.9% achieved an HbA_1c level of 8.0%.

The mean baseline fructosamine levels of 341.73 µmol/L in the control group increased by 12.3 µmol/L. But in the rosiglitazone groups the levels decreased by 27.9 µmol/L from 340.9 µmol/L in the 4-mg/d group and by 36.8 µmol/L from 351.8 µmol/L in the 8-mg/d group (reference range, 200-278 µmol/L).

Although the mean FPG concentrations did not change significantly in the control group, they significantly decreased in a dose-dependent order from baseline in both rosiglitazone groups (1.8 mmol/L [–33.0 mg/dL], 4-mg/d rosiglitazone; –2.7 mmol/L [–48.4 mg/dL], 8-mg/d-rosiglitazone; P<.0001). The mean FPG concentrations in both rosiglitazone groups also had decreased compared with the control group (−2.2 mmol/L [−39.8 mg/dL], 4-mg/d rosiglitazone; –2.9 mmol/L [–52.9 mg/dL], 8-mg/d rosiglitazone; P<.0001) (Figure 4). Furthermore, FPG concentrations in both rosiglitazone groups decreased during the first 4 weeks, plateaued at 12 to 18 weeks, and remained stable thereafter (Figure 5). Nine patients (7.9%) in the control group, 25 (21.6%) in the 4-mg/d and 33 (30.0%) in the 8-mg/d rosiglitazone groups achieved FPG concentrations of less than7.8 mmol/L (140 mg/dL).

Adding rosiglitazone to maximum doses of metformin significantly increased HOMA-S values. The median baseline HOMA-S values ranged from 46.6 to 49.0 units. The HOMA-S values increased dose-dependently by 1.7 units in the 4-mg/d and by 3.8 units in the 8-mg/d rosiglitazone groups compared with the control group.

The metformin-rosiglitazone combination increased HOMA-B in a dose-dependent fashion. The median baseline HOMA-B values ranged from 32.5 to 35.8 units and were significantly increased by 10.3 to 13.7 units in the rosiglitazone groups compared with the control group.

In the control group, the insulin value decreased by 11.05 pmol/L from a baseline of 118.56 pmol/L after treatment (P = .03) and in the 4-mg/d and 8-mg/d rosiglitazone groups the insulin values respectively decreased by 12.98 pmol/L from 124.55 pmol/L (P = .01) and by 31.07 pmol/L from 136.73 pmol/L (P = .14). The C-peptide values respectively decreased by 0.10 nmol/L from 0.93 nmol/L (P<.001), by 0.07 nmol/L from 0.92 nmol/L (P = .01), and by 0.12 nmol/L from 0.93 nmol/L (P<.001).

Mean total cholesterol-HDL-C, and LDL-C levels from baseline in both rosiglitazone groups achieved statistically significant increases in all treatment groups compared with the control group (Table 2). Total cholesterol–HDL-C ratios in the rosiglitazone groups were not significantly different from those in the control group.

Changes in LDL-C levels were evaluated based on those at baseline. In that analysis, we identified 2 subgroups: those with levels lower than 3.37 mmol/L (<130 mg/dL) and those at that level or higher. We did not provide P values for any of the subgroups because the values were not large enough for statistical analyses and because the subgroups were not randomized, so significance could not be established. In the lower subgroup, the median baseline LDL-C value increased by 0.13 mmol/L (5 mg/dL) from 2.59 mmol/L (100 mg/dL) in 51 patients in the control group. In both rosiglitazone groups, the LDL-C values increased by 0.54-mmol/L (21 mg/dL) from a median baseline value of 2.69 mmol/L (104-mg/dL) in 57 patients taking 4-mg/d and from 2.64-mmol/L (102 mg/dL) in 60 patients taking 8-mg/d, resulting in medians that remained below 3.37 (<130 mg/dL) for all 3 treatment groups.

In the higher subgroup, the median baseline LDL-C value increased by 0.07 mmol/L (3 mg/dL) from 3.78-mmol/L (146 mg/dL) in 30 patients in the control group. In the rosiglitazone groups, the median baseline LDL-C value increased by 0.31 mmol/L (12 mg/dL) from 3.72 mmol/L (144-mg mg/dL) in 27 patients taking 4-mg/d and by 0.34-mmol/L (13 mg/dL) from 4.20 mmol/L (162 mg/dL) in 20 patients taking 8-mg/d.

Changes in triglyceride levels also were evaluated based on baseline values, using 2 subgroups: those with levels lower than 2.26 mmol/L (<200 mg/dL) and those with that level or higher. In the lower subgroup, the median baseline triglyceride values increased by 0.15 mmol/L (13 mg/dL) from 1.44-mmol/L (128-mg/dL) in 52 patients in the control group. In the rosiglitazone groups, the median baseline triglyceride value increased by 0.16 mmol/L (15 mg/dL) from 1.67 mmol/L (148-mg/dL) in 56 patients taking 4-mg/d and by 0.07 mmol/L (6 mg/dL) from 1.34-mmol/L (119 mg/dL) in 55 patients taking 8-mg/d. The treatment values in all groups remained less than 2.25 mmol/L (200 mg/dL).

In the higher subgroup, the median baseline triglyceride values decreased by 0.12 mmol/L (11 mg/dL) from 3.24-mmol/L (287 mg/dL) in 41 patients in the control group. In the rosiglitazone groups, the baseline median triglyceride value increased by 0.15 mmol/L (13 mg/dL) from 3.50 mmol/L (310 mg/dL) in 43 patients taking 4-mg/d and decreased by 0.72 mmol/L (64 mg/dL) from 3.16-mmol/L (280 mg/dL) in the 8-mg/d rosiglitazone group.

Mean fasting lactate levels decreased significantly in patients taking both dose levels of rosiglitazone compared with those in the control group (4-mg/d rosiglitazone, P = .012; 8-mg/d rosiglitazone, P = .002). Free fatty acids concentrations decreased significantly from baseline in both rosiglitazone groups. (Table 3).

The percentage of patients with at least 1 adverse event were comparable among each group (75.2%, 4-mg/d rosiglitazone; 78.2%, 8-mg/d rosiglitazone; 76.7%, control). The most frequently reported adverse events were upper respiratory tract infection, diarrhea, and headache. One death due to acute myocardial infarction occurred in the 4-mg/d rosiglitazone group but was judged to be unrelated to study medication. Serious nonfatal adverse events occurred in 5 (4.3%) of 116 patients in the control group and in 5 (4.2%) of 119 patients in the 4-mg/d and 5 (4.4%) of 113 patients in the 8-mg/d rosiglitazone groups, none considered related to study medication.

Symptomatic mild or moderate hypoglycemia was reported by 2 patients in the control group and by 3 patients in the 4-mg/d and by 5 patients in the 8-mg/d rosiglitazone groups. No patient required third-party intervention or hospitalization, but the metformin dose was reduced from 2.5 g/d to 2.0 g/d in 2 patients. No one withdrew because of hypoglycemia, and there were no biochemically documented instances of FPG levels of less than 2.78-mmol (<50 mg/dL).

Both rosiglitazone groups experienced small but statistically significant decreases in hemoglobin and hematocrit levels, which occurred primarily during the first 12 to 18 weeks of treatment, after which values for both parameters increased slightly. The mean decreases in hemoglobin levels were −5.0 g/L in the 4-mg/d and –8.0 g/L in the 8-mg/d rosiglitazone groups (P<.0001 for both groups), and mean decreases in hematocrit were –1.8% in the 4-mg/d and –2.5% in the 8-mg/d rosiglitazone groups (P<.0001 for both groups). There were no significant changes in these parameters in the control group. One patient in each rosiglitazone group withdrew because of anemia, and 1 patient in the 4-mg/d rosiglitazone group with low hemoglobin and hematocrit levels was withdrawn from the study after week 8 because of evidence of gastrointestinal tract bleeding, considered by the investigator to be unrelated to the study medication.

There were no significant changes from baseline in vital signs or electrocardiogram parameters in the rosiglitazone groups compared with the control group. Although infrequent, edema was observed with greater frequency in the rosiglitazone groups (2.5%, 4-mg/d; 3.5%, 8-mg/d) than in the control group (0.9%). No one withdrew due to edema.

Those in the control group experienced a mean decrease in body mass of 1.2 kg from baseline, but those in the rosiglitazone groups experienced a mean body mass increase of 0.7 kg in the 4-mg/d and 1.9 kg in the 8-mg/d rosiglitazone groups (P = .0001 for both groups). There were no significant differences in waist-to-hip ratios among groups.

No one in the rosiglitazone groups experienced elevations of alanine aminotransferase (ALT) levels greater than 3 times the upper limit of the reference range. Mean changes in aspartate aminotransferase (AST), ALT, and total bilirubin levels were similar in all groups, with a slight decrease observed in mean ALT (–1.9 U/L, control; –1.9 U/L, 4-mg/d rosiglitazone; –3.4 U/L, 8-mg/d rosiglitazone). Mean alkaline phosphatase decreased in all groups (–3.5 U/L, control; –12.0 U/L, 4 mg/d rosiglitazone; –14.7 U/L, 8-mg/d rosiglitazone); the mean value for all groups was within the reference range. Two patients in the control group were noted to have liver function tests for potential clinical concern (>3 times the upper limit of the reference range) while in treatment. Both completed the study with elevated transaminase values.

This is the first large, multicenter, clinical trial demonstrating the efficacy and safety of combined rosiglitazone and metformin treatment in patients with type 2 diabetes. The combination treatment of metformin and rosiglitazone significantly reduced HbA_1c and FPG concentrations, in a dose-ordered fashion compared with baseline and with metformin alone. Conversely, treatment with metformin was associated with significant increases in HbA_1c concentrations, indicating that these agents complement each other to achieve optimal glycemic control and confirming the clinical utility of metformin in combination with a thiazolidinedione drug.²⁰

Consistent with the mechanisms of action of metformin and rosiglitazone, the reductions in FPG concentrations were proportionately smaller than those observed in HbA_1c concentrations. Maximum doses of metformin decrease hepatic gluconeogenesis, which principally affects FPG concentrations, whereas rosiglitazone enhances insulin sensitivity at the peripheral level and affects overall glucose disposal, including postprandial excursions. Because the relative contribution of postprandial glucose on glycemic control depends on the magnitude of FPG concentrations,²¹ rosiglitazone may have an effect on postprandial hyperglycemia, as demonstrated directly in a rosiglitazone trial that showed significant improvements in fasting and postprandial glucose concentrations and excursions.²²

The complementary actions of combined metformin and rosiglitazone is further supported by the effects of rosiglitazone on insulin sensitivity despite maximum doses of metformin. Rosiglitazone may provide added therapeutic value by reducing peripheral insulin resistance. While HOMA-S is an indirect method for determining insulin sensitivity, these results are consistent with glucose-clamp studies using other thiazolidinedione drugs.^23- 24

The improvements in HOMA-B with metformin-rosiglitazone treatment (not observed with metformin alone) were unexpected and introduce an important potential therapeutic benefit of rosiglitazone. Although the exact mechanism underlying this improvement remains to be determined, rosiglitazone-mediated reductions in glucotoxicity²⁵ and lipotoxicity secondary to elevated concentrations of circulating FFA or both^26- 27 are candidate mechanisms by which rosiglitazone may improve BCF. The effects of rosiglitazone on BCF and insulin sensitivity are consistent with its effects on long-term glycemic control and suggest that it may possibly delay or prevent disease progression.

Despite significant increases in total cholesterol, HDL-C, and LDL-C with the metformin-rosiglitazone treatments, the total cholesterol–HDL-C ratio, which did not change significantly, may be a better predictor of cardiovascular outcome than either total cholesterol or HDL-C levels alone.^28- 30 Since this study was not designed to assess long-term lipid effects, the long-term significance of these changes is unknown; however, patients with baseline plasma LDL-C levels lower than 3.37 mmol/L (<130 mg/dL) remained less than that level after therapy. No significant changes in triglyceride levels were noted in any treatment group, and segregation of patients into subgroups revealed nonsignificant increases in patients with baseline triglyceride levels lower than 2.26 mmol/L (<200 mg/dL). Among patients in the 8-mg/d rosiglitazone group whose baseline was higher than 2.26 mmol/L (>200 mg/dL), there was a significant statistical decrease observed (64 mg/dL). The clinical significance of lipid level changes may be minimal, because lipid-lowering therapy may be often administered to patients with diabetes irrespective of prior heart disease history.^31- 32

Elevated FFA may play a role in the development of insulin resistance, because it is associated with increased hepatic glucose output^33- 34 and may contribute to β-cell dysfunction via a lipotoxic effect.^26- 27 Elevated FFA has also been linked to endothelial dysfunction and hypertension^35- 36 and enhanced platelet aggregation and coagulation,^37- 38 which may increase cardiovascular risk. Therefore metformin-rosiglitazone treatment was significantly more effective in lowering FFA than the metformin alone.

The weight gain observed in those receiving metformin-rosiglitazone treatment may be attributed to increased adipocyte differentiation,^39- 40 fluid retention,^39,41 or increased appetite.⁴² Despite weight increases, no significant differences in waist-to-hip ratio among groups were observed, suggesting that rosiglitazone treatment leads to increased energy storage in subcutaneous adipose sites that are not associated with increased cardiovascular risk.⁴³ The small decreases in hemoglobin and hematocrit levels associated with metformin-rosiglitazone therapy may relate to plasma volume expansion derived from fluid retention and hemodilution.⁴⁴

Metformin-rosiglitazone therapy may be a safe alternative therapy to attain optimal glycemic control where monotherapy has failed because the statistically significant decreases in lactate levels associated with metformin-rosiglitazone treatment indicate that rosiglitazone may correct metabolic abnormalities beyond reducing hyperglycemia, and further suggest differing and complementary actions of metformin and rosiglitazone; and ALT elevations greater than 3 times the upper limit of the reference range were not observed in either of the rosiglitazone groups.

In summary, combination metformin-rosiglitazone treatment is effective and safe in reducing hyperglycemia in patients with type 2 diabetes. In patients whose fundamental abnormality is insulin resistance, such a combination raises the exciting possibility of treating diabetes by targeting the underlying cause of the disease, rather than the traditional approach of stimulating insulin secretion. Nearly 30% of patients taking the combination therapy achieved HbA_1c levels of 7% or less. This level of glycemic control is 3-fold greater than what was achieved among those taking metformin alone. Additional investigation is needed to determine whether this combination will alter the long-term risk of cardiovascular disease or delay disease progression.