Advertisement
Clinical Studies| Volume 103, ISSUE 6, P491-497, December 1997

Efficacy of Metformin in Type II Diabetes

Results of a Double-Blind, Placebo-controlled, Dose-Response Trialfn1

      Abstract

      PURPOSE: To study the efficacy and safety of various dosages of metformin as compared with placebo in patients with type II diabetes mellitus.
      PATIENTS AND METHODS: A 14-week, multicenter, double-blind, dose-response study was conducted. After a 3-week, single-blind, placebo-controlled washout, 451 patients with fasting plasma glucose levels of at least 180 mg/dL were randomized to receive an 11-week course of placebo or metformin given at 500, 1000, 1500, 2000, or 2500 mg daily.
      RESULTS: Metformin improved glucose variables as compared with placebo. The adjusted mean changes in fasting plasma glucose from baseline associated with each metformin group at week 7, 11, or at endpoint exceeded those associated with placebo by 19 to 84 mg/dL at dosages of 500 to 2000 mg daily, respectively. The corresponding between-group differences in glycated hemoglobin (HbA1c) ranged from 0.6% to 2.0% at dosages of 500 to 2000 mg daily, respectively. All between-group differences were significant (P < 0.05) for both fasting plasma glucose and HbA1c at week 7, week 11, and endpoint, except for the difference between placebo and metformin 500 mg in fasting plasma glucose at endpoint (P = 0.054). Treatment-related adverse events occurred in 15% of patients in the placebo group and in 28% in the metformin group (P = 0.02); these were primarily manifested as digestive disturbances, such as diarrhea.
      CONCLUSIONS: Metformin lowered fasting plasma glucose and HbA1c generally in a dose-related manner. Benefits were observed with as little as 500 mg of metformin; maximal benefits were observed at the upper limits of the recommended daily dosage. All dosages were well tolerated. Metformin appears to be a useful therapeutic option for physicians who wish to titrate drug therapy to achieve target glucose concentrations.
      The primary goal of treatment in patients with type II diabetes mellitus is to prevent chronic complications. The results of the Diabetes Control and Complications Trial[
      The Diabetes Control and Complications Trial Research Group
      The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus.
      ] indicate that intensive glycemic control prevents the development and delays the progression of chronic complications in patients with type I diabetes. The results of a similar, but smaller, study from Japan[
      • Ohkubo Y
      • Kishikawa H
      • Araki E
      • et al.
      Intensive insulin therapy prevents the progression of diabetic microvascular complications in Japanese patients with non-insulin-dependent diabetes mellitus: a randomized prospective 6-year study.
      ] indicate that virtually identical benefits of glycemic control result from intensive management in patients with type II diabetes. Until 1995, the only therapeutic agents for glycemic control were diet, sulfonylureas, and insulin.[
      • Harris MI
      • Hadden WC
      • Knowler WC
      • Bennett PH
      Prevalence of diabetes and impaired glucose tolerance and plasma glucose levels in US population aged 20–74 yr.
      ] Now three more therapeutic classes are available: a biguanide, an α-glucosidase inhibitor, and a thiazolidinedione; there is also a new sulfonylurea.
      Metformin is an oral biguanide that has been available in Europe for more than 30 years and in the United States since 1995. Metformin, which is classified as an antihyperglycemic agent, lowers glucose variables by increasing insulin sensitivity in peripheral tissues[
      • Johnson AB
      • Webster JM
      • Sum C-F
      • et al.
      The impact of metformin therapy on hepatic glucose production and skeletal muscle glycogen synthase activity in overweight type II diabetic patients.
      ,
      • Nosadini R
      • Avogaro A
      • Trevisan R
      • et al.
      Effect of metformin on insulin-stimulated glucose turnover and insulin binding to receptors in type II diabetes.
      ,
      • Hother-Nielsen O
      • Schmitz O
      • Andersen PH
      • et al.
      Metformin improves peripheral but not hepatic insulin action in obese patients with type II diabetes.
      ,
      • Klip A
      • Leiter LA
      Cellular mechanism of action of metformin.
      ,
      • Widén EIM
      • Eriksson JG
      • Groop LC
      Metformin normalizes nonoxidative glucose metabolism in insulin-resistant normoglycemic first-degree relatives of patients with NIDDM.
      ] and inhibiting hepatic glucose production.[
      • Johnson AB
      • Webster JM
      • Sum C-F
      • et al.
      The impact of metformin therapy on hepatic glucose production and skeletal muscle glycogen synthase activity in overweight type II diabetic patients.
      ,
      • Nosadini R
      • Avogaro A
      • Trevisan R
      • et al.
      Effect of metformin on insulin-stimulated glucose turnover and insulin binding to receptors in type II diabetes.
      ,
      • DeFronzo RA
      • Barzilai N
      • Simonson DC
      Mechanism of metformin action in obese and lean non-insulin-dependent diabetic subjects.
      ,
      • Perriello G
      • Misericordia P
      • Volpi E
      • et al.
      Acute antihyperglycemic mechanisms of metformin in NIDDM.
      ] Unlike sulfonylureas, metformin does not stimulate insulin secretion and, consequently, does not produce hypoglycemia.[
      • Bailey CJ
      Biguanides and NIDDM.
      ,
      • Campbell IW
      Metformin and the sulphonylureas: the comparative risk.
      ,
      • Reaven GM
      • Johnston P
      • Hollenbeck CB
      • et al.
      Combined metformin-sulfonylurea treatment of patients with non-insulin-dependent diabetes in fair to poor glycemic control.
      ] In fact, the plasma insulin response to glucose is unchanged or may be decreased in patients with hyperinsulinemia.[
      • Bailey CJ
      Biguanides and NIDDM.
      ,
      United Kingdom Prospective Diabetes Study Group
      United Kingdom Prospective Diabetes Study (UKPDS) 13: relative efficacy of randomly allocated diet, sulphonylurea, insulin, or metformin in patients with newly diagnosed non-insulin-dependent diabetes followed for three years.
      ] In view of these complementary mechanisms of action for biguanides as compared with sulfonylureas, it is understandable that metformin has synergistic action with sulfonylureas.[
      • Reaven GM
      • Johnston P
      • Hollenbeck CB
      • et al.
      Combined metformin-sulfonylurea treatment of patients with non-insulin-dependent diabetes in fair to poor glycemic control.
      ] Metformin alone also has beneficial effects on plasma lipid concentrations and in dyslipidemias, which are independent of glycemic control.[
      • DeFronzo RA
      • Barzilai N
      • Simonson DC
      Mechanism of metformin action in obese and lean non-insulin-dependent diabetic subjects.
      ,
      • Reaven GM
      • Johnston P
      • Hollenbeck CB
      • et al.
      Combined metformin-sulfonylurea treatment of patients with non-insulin-dependent diabetes in fair to poor glycemic control.
      ,
      • Nagi DK
      • Yudkin J
      Effects of metformin on insulin resistance, risk factors for cardiovascular disease, and plasminogen activator inhibitor in NIDDM subjects. A study of two ethnic groups.
      ] Finally, metformin promotes stabilization of weight and may even cause weight loss.[
      • Bailey CJ
      Biguanides and NIDDM.
      ]
      Despite this extensive clinical research and clinical experience worldwide, no formal dose-ranging study has been conducted because, when the drug was originally developed, such studies were not routinely performed in patients. The initial daily dosage selected for clinical studies,[
      • Azerad E
      • Lubetzki J
      The treatment of adult diabetes with N-N dimethyl-diguanide (LA 6023).
      ,
      • Azerad E
      The treatment of adult diabetes with N-N dimethyl-diguanide.
      ,

      Leblanc D. Contribution to the study of a synthetic hypoglycemic: NN dimethyl biguanide hydrochloride. Thesis Med. 1960;1–64.

      ] approximately 3000 mg, was based on animal data.[

      Duval D. Contribution to the study of the hypoglycemic action of biguanides. Thesis Pharm. 1960;77–109.

      ,

      Sterne J. Toxicology of the antidiabetic biguanides. Dimethylguanylguanide hydrochloride. 4th IDF Congress, Geneva, Switzerland. 1961;July 10–14:712–716.

      ,

      Sterne J, Duval D. Hypoglycemic effects of N-N-dimethylbiguanide. 3rd IDF Congress, Dûsseldorf, Germany. 1958:July 21–25:443–452.

      ] Because this dosage was often associated with digestive disturbances, the initial dose was lowered and the drug was titrated at weekly intervals to the final maintenance dosage. Consequently, the current dosing strategy of metformin was determined empirically, rather than by an understanding of the minimal effective dose or of its dose-response relationship in patients with type II diabetes.
      The purpose of the current trial was to study the pharmacodynamic effect of various dosages of metformin as compared with placebo in patients with type II diabetes as measured by changes in fasting plasma glucose. Additionally, glycated hemoglobin (HbA1c) was evaluated. A secondary objective was to determine the minimal effective dose of metformin.

      1. Patients and Methods

      1.1 Study Design

      This was a 14-week multicenter, double-blind, dose-response study of 451 patients in 6 parallel treatment groups. All patients were instructed to maintain their recommended dietary and exercise programs. After a 3-week, single-blind, placebo-controlled washout phase, patients were randomized to receive placebo or 1 of 5 dosages of metformin for 11 weeks. The final daily metformin dosages were 500, 1000, 1500, 2000, and 2500 mg. Metformin was supplied as 500-mg tablets. To maintain blinding, matching placebo tablets were used as needed, so that each patient received five tablets daily. The tablets were given 3 times daily (tid) with meals according to the following dose escalation schedule: 1 tablet tid, which was increased over 3 weeks to 2 tablets with breakfast, 1 with lunch, and 2 with dinner. The final regimens, which were given for a minimum of 8 weeks, were metformin 500 mg once daily; 500 mg twice daily (bid); 500 mg tid; 1000 mg bid; and 1000 mg with breakfast, 500 mg with lunch, and 1000 mg with dinner. Compliance was assessed by tablet count at each visit.

      1.2 Patients

      Eligible patients were men and women at least 30 years old who had type II diabetes, which was suboptimally controlled on diet alone or which was previously treated with an oral sulfonylurea. Three weeks after discontinuing previous drug therapy, they had to have a fasting plasma glucose of at least 180 mg/dL without symptoms.
      Patients were excluded if any of the following was present: significant disease or conditions likely to affect their diabetes or ability to complete the study, markedly symptomatic diabetes, biguanide hypersensitivity, previous insulin therapy, or concomitant treatment with nephrotoxic drugs or other investigational drugs. Women who were pregnant, nursing, or not using adequate methods of contraception were also excluded.
      The study protocol was approved by each institution’s review board or by the central review board. All patients gave written informed consent before enrollment.

      1.3 Evaluation of Safety and Efficacy

      Patients visited the local study site weekly during washout and during the first 3 weeks of double-blind treatment; thereafter, patients visited the center after 7 and 11 weeks of double-blind treatment. Physical examination was performed at enrollment and at study completion. Vital signs, adverse events, and concomitant medications were assessed at each visit. Fasting plasma glucose was measured at each visit. Other laboratory evaluations, including HbA1c, hematology profile, chemistry profile, and urinalysis, were performed at the beginning and end of washout, and after 7 and 11 weeks of double-blind treatment. Laboratory evaluations were performed by a central laboratory (MAYO Medical Laboratories, Rochester, Minnesota).
      The primary efficacy endpoint was defined as the last valid double-blind evaluation for each patient. To be eligible for the efficacy analysis, the patient had to have received at least 4 weeks of treatment and to have data from either week 7 or 11. The primary efficacy variable was fasting plasma glucose at endpoint; data from double-blind treatment weeks 7 and 11 were also analyzed. The other efficacy variable was HbA1c from corresponding evaluation periods.

      1.4 Statistical Analysis

      Baseline demographic characteristics were compared for homogeneity across treatment groups using a Cochran-Mantel-Haenszel test for categorical data and analysis of variance (ANOVA) for continuous variables.
      Efficacy analyses were based on intent-to-treat data, without regard to study withdrawal, compliance, or concomitant medications, and on change from baseline in fasting plasma glucose and HbA1c. ANOVA was performed using a two-way model with terms for treatment and center; treatment-by-center interactions were found to be insignificant. The dose response was evaluated by pairwise comparisons using a general linear model. William’s t-bar test[
      • Williams DA
      The comparison of several dose levels with a zero dose control.
      ] was used to determine the minimum effective dose.
      All patients were included in the safety analysis. Adverse experience data were analyzed by chi-square or Fisher’s exact test. Laboratory variables were categorized as low, normal, or high. The Stuart-Maxwell statistic, McNemar’s test, or the sign test was performed to assess the change in distribution of laboratory variables across the 3 categories from baseline and at double-blind treatment weeks 7 and 11. All tests were two-tailed with levels of significance of 0.05 for efficacy analyses, and 0.10 for treatment-by-center interaction and safety analyses.

      2. Results

      2.1 Patients

      Six hundred ninety-three patients were entered into the single-blind, placebo-controlled washout in 1995. Two-hundred forty-two patients (35%) were not randomized because of the presence of exclusion criteria (22%), patient withdrawal (6%), protocol violation (3%), loss to follow-up (2%), physician preference (1%), or an adverse event (1%) during the washout. The remaining 451 patients were evenly distributed across treatment groups based on demographic characteristics, except for age (Table 1). Mean glucose variables were similar across treatment groups at baseline.
      Table 1Baseline Demographic Characteristics, Baseline Glucose Variables, and Study Completion Status
      Metformin Dose (mg)
      VariablePlacebo (n = 79)500 (n = 73)1000 (n = 73)1500 (n = 76)2000 (n = 73)2500 (n = 77)
      Mean age ± SD (y)
      P = 0.03 across all treatment groups.
      55 ± 1157 ± 1055 ± 1059 ± 1060 ± 1159 ± 11
      Mean weight ± SD (kg)90.9 ± 20.190.0 ± 20.390.0 ± 18.989.6 ± 16.289.1 ± 20.394.5 ± 23.4
      Mean height ± SD (cm)168 ± 10168 ± 10168 ± 10168 ± 13170 ± 10168 ± 13
      Male (%)566255635365
      Caucasian (%)66747771
      Data missing for one patient.
      7079
      Black (%)22111212
      Data missing for one patient.
      1410
      Hispanic (%)11141015
      Data missing for one patient.
      1210
      Other (%)1112
      Data missing for one patient.
      40
      FPG
      FPG = fasting plasma glucose.
      ± SD (mg/dL)
      279 ± 59.1
      Data missing for one patient.
      282 ± 59.5281 ± 60.3262 ± 51.7288 ± 61.1287 ± 59.9
      HbA1c ± SD (%)9.9 ± 1.9
      Data missing for one patient.
      10.1 ± 1.7
      Data missing for one patient.
      10.0 ± 2.0
      Data missing for one patient.
      9.7 ± 1.5
      Data missing for one patient.
      10.1 ± 2.110.0 ± 1.8
      Study status (%)
      Completed707475828271
      Incomplete302625181829
      Adverse event related)6 (0)4 (0)5 (5)7 (3)7 (4)14 (10)
      Treatment failure9105316
      Lost to follow-up133301
      Personal preference647333
      Protocol violation343443
      Other511031
      a P = 0.03 across all treatment groups.
      b Data missing for one patient.
      c FPG = fasting plasma glucose.
      One hundred-ten patients did not complete the entire double-blind treatment but were included in the intent-to-treat analysis. Reasons for stopping treatment were evenly distributed across treatment groups (Table 1), except for adverse events, which are addressed in the safety analysis.

      2.2 Efficacy Analysis

      In the placebo group, adjusted mean fasting plasma glucose increased by 0.4 mg/dL between baseline and 7 weeks, and decreased by 8 mg/dL at 11 weeks and at endpoint (Table 2). In contrast, metformin reduced adjusted mean fasting plasma glucose (FPG) by 24 to 88 mg/dL from baseline, depending on the dose and time of evaluation. The adjusted mean differences between the change from baseline in the placebo group versus the corresponding changes in each metformin group were significant at all evaluation times, except for the difference associated with metformin 500 mg at endpoint (P = 0.054). At endpoint, these between-group differences ranged from 19 mg/dL at the lowest dosage to 78 mg/dL at the 2000 mg dosage (Fig. 1).
      Table 2Adjusted Mean Changes from Baseline in Glucose Variables During Double-Blind Treatment
      Metformin Dose (mg)
      VariablePlacebo (n = 79)500 (n = 73)1000 (n = 73)1500 (n = 76)2000 (n = 73)2500 (n = 77)
      Fasting plasma glucose (mg/dL)
      Week 7+0.4−24
      P < 0.01;
      −41
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      −49
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      −84
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      −62
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      Week 11−8−29
      P < 0.05;
      −43
      P < 0.01;
      −52
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      −88
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      −73
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      Endpoint−8−27
      P = 0.054 for mean difference from placebo, adjusting for center effect in linear model.
      −39
      P < 0.01;
      −49
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      −86
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      −70
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      HbA1c (%)
      Week 7+1.1+0.4
      P < 0.01;
      −0.01
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      −0.3
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      −0.5
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      −0.1
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      Week 11+1.2+0.2
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      −0.1
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      −0.6
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      −0.9
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      −0.5
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      Endpoint+1.2+0.3
      P < 0.01;
      +0.01
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      −0.5
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      −0.8
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      −0.4
      P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      a P < 0.05;
      b P < 0.01;
      c P < 0.001 for mean difference from placebo, adjusting for center effect in linear model.
      d P = 0.054 for mean difference from placebo, adjusting for center effect in linear model.
      Figure thumbnail gr1
      Fig. 1Estimated difference from placebo (±SE) as measured by changes from respective baseline glucose variables at endpoint. ∗P = 0.054; ∗∗P < 0.01; ∗∗∗P < 0.001 for estimated difference from placebo.
      Analogous findings were observed for HbA1c. In the placebo group, adjusted mean HbA1c increased by 1.1%, 1.2%, and 1.2% between baseline and 7 weeks, 11 weeks, and endpoint, respectively (Table 2). In contrast, metformin reduced adjusted mean HbA1c by up to 0.9% from baseline. The adjusted mean between-group differences were significant at all evaluation times (P ≤ 0.01), ranging at endpoint from 0.9% at the lowest metformin dosage to 2.0% at the 2000 mg dosage (Fig. 1).
      The results of the William’s t-bar test for the minimum effective daily dosage (test statistic and two-sided critical values not shown) were consistent with the ANOVA results (P values shown in Table 2). At the 500 mg dosage, the change in FPG from baseline to endpoint was marginally insignificant in the intent-to-treat population (P = 0.054) and significant when 13 patients who received concomitant antidiabetic therapy were excluded (P = 0.03). At the 500 mg dosage, the corresponding change in HbA1c was significant in both patient populations (P < 0.001 and P < 0.001).
      Pairwise comparisons of adjusted mean changes in FPG from baseline between adjacent metformin dosages confirmed that the maximal response occurred at 2000 mg. All differences between 2000 and ≤1500 mg were significant (P < 0.001). The difference between 2000 and 2500 mg was not significant (P = 0.1), suggesting a plateau effect.

      2.3 Safety Analysis

      The incidence of adverse events, which were considered by the investigator to be related to treatment, was higher in the collective metformin groups than in the placebo group (28% versus 15%; P = 0.02) (Table 3). This resulted primarily from digestive disturbances (24% versus 13%, P = 0.025), which were usually manifested as diarrhea (15% versus 5%; P = 0.02), nausea (9% versus 5%; P = 0.4), or dyspepsia (4% versus 1%; P = 0.5). The only other adverse events that occurred in at least 2% of patients within a treatment group were anorexia (2% versus 1%), abdominal pain (2% versus 0%), and hyperglycemia as defined by clinical judgment (1% versus 3%).
      Table 3Drug-Related Adverse Events Occurring in ≥2% of Patients Within a Treatment Group
      Percent of Patients with Adverse Event (Percent Who Stopped Treatment Due to Adverse Event)
      Metformin Dose (mg)
      Adverse EventPlacebo (n = 79)500 (n = 73)1000 (n = 73)1500 (n = 76)2000 (n = 73)2500 (n = 77)
      Any adverse event
      P < 0.05, metformin (all doses) versus placebo.
      Some patients experienced more than one adverse event.
      15 (0)25 90)30 (5)26 (3)27 (4)30 (10)
      Whole body disturbances3 (0)4 (0)1 (1)4 (0)0 (0)4 (0)
      Abdominal pain0 (0)3 (0)1 (1)4 (1)0 (0)3 (0)
      Digestive disturbances
      P < 0.05, metformin (all doses) versus placebo.
      13 (0)16 (0)29 (5)24 (3)23 (4)29 (10)
      Diarrhea
      P < 0.05, metformin (all doses) versus placebo.
      5 (0)8 (0)21 (4)12 (3)19 (3)14 (5)
      Nausea5 (0)7 (0)10 (3)8 (3)1 (1)12 (5)
      Dyspepsia1 (0)1 (0)1 (0)9 (0)3 (0)4 (3)
      Anorexia1 (0)0 (0)1 (0)3 (0)1 (0)5 (1)
      Metabolic disturbances3 (0)1 (0)0 (0)4 (0)3 (0)3 (0)
      Hyperglycemia3 (0)1 (0)0 (0)1 (0)0 (0)0 (0)
      a P < 0.05, metformin (all doses) versus placebo.
      b Some patients experienced more than one adverse event.
      When all adverse events were considered, including those considered to be unrelated to treatment, placebo was associated with a twofold higher incidence of hyperglycemia than metformin (16% versus 8%; P = 0.02). The only other between-group differences in all adverse events, digestive disturbances and diarrhea, were also reflected in the subset of treatment-related adverse events.
      Most treatment-related adverse events were mild or moderate. During treatment with metformin or placebo, only 11 patients experienced serious adverse events, including 1 who was hospitalized because of shortness of breath and chest pain and who was ultimately diagnosed with dyspepsia, which was of moderate severity and possibly related to receiving metformin 1500 mg daily. The other 10 cases were attributed to their underlying diseases.
      Metformin was stopped in 17 patients (5%) because of treatment-related adverse events, usually digestive disturbances. All adverse events that led to discontinuation of metformin were mild or moderate, except for one case of severe diarrhea and one case of severe abdominal pain, diarrhea, nausea, and vomiting. All of these adverse events resolved after stopping metformin.
      There were no clinically relevant changes in vital signs. Mean weight changes were generally evenly distributed across treatment groups; there was a nonsignificant tendency toward weight loss in all treatment groups. There were no clinically relevant shifts in laboratory variables as measured by the proportions of patients who experienced changes relative to normal values. There were only 3 episodes of hypoglycemia as defined by clinical judgment, 1 each at daily dosages of 1500, 2000, and 2500 mg; 2 were mild and 1 was moderate; none required discontinuation of treatment.

      3. Comments

      The results of this double-blind, placebo-controlled, multicenter trial in 451 patients with type II diabetes indicate that metformin lowered glucose variables in a generally dose-related manner. Metformin improved glucose variables by more than did placebo. Metformin, at dosages of 500 to 2000 mg daily, reduced adjusted mean FPG from baseline by 19 to 84 mg/dL and adjusted mean HbA1c by 0.6% to 2.0% more than did placebo. This disproportionate effect on glucose variables is not surprising; FPG rapidly reflects the full effect of treatment whereas the time required for HbA1c to reflect the full effect of treatment may have exceeded the 11-week study period. Importantly, these findings indicate that the minimal effective daily dosage of metformin is 500 mg, not 1500 mg as previously reported.[
      ,
      • Grant PJ
      The effects of high- and medium-dose metformin therapy on cardiovascular risk factors in patients with type II diabetes.
      ]
      Only one previous double-blind study[
      • Grant PJ
      The effects of high- and medium-dose metformin therapy on cardiovascular risk factors in patients with type II diabetes.
      ] provides direct evidence of a dose relationship. When 75 patients with type II diabetes were randomized to receive placebo or metformin 1500 or 3000 mg daily for 6 months, the between-group differences from their respective baseline FPG values were approximately 27 mg/dL for the lower dosage (NS) and 82 mg/dL for the higher dosage (P = 0.001). The authors[
      • Grant PJ
      The effects of high- and medium-dose metformin therapy on cardiovascular risk factors in patients with type II diabetes.
      ] attributed the lack of benefit at the lower dosage to good initial glycemic control; small sample size may also have been a confounding factor. In any event, the corresponding between-group differences for HbA1c were 1.5% (P < 0.001) and 1.8% (P < 0.001) for the lower and higher dosages, respectively.
      In the current study, the maximal efficacy of metformin was observed at 2000 mg, not at 2500 mg. The explanation for this finding is elusive. It is conceivable that the slightly higher discontinuation rate at the highest dosage may have diluted the effect on glucose variables at the highest dosage in this intent-to-treat analysis. Unfortunately, the current study was not designed to evaluate the clinical practice of titrating the dosage according to individual response because each patient was randomly assigned to a predetermined dosage. Although previous studies were not designed to evaluate dose relationships, their titration phases provide indirect evidence of a within-patient dose relationship up to daily dosages of 2500 mg. For example, the incremental reductions in mean FPG were 22, 23, and 15 mg/dL when the daily dosage was escalated from 850 mg to 1700 mg, then to 2550 mg, respectively.[
      • DeFronzo RA
      • Goodman AM
      The Multicenter Metformin Study Group. Efficacy of metformin in patients with non-insulin-dependent diabetes mellitus.
      ] Similarly, the incremental reductions in FPG were 14, 28, 13, 11, and 23 mg/dL when the daily dosage of metformin (in the presence of glyburide) was escalated from 500 mg to 2500 mg at 500-mg intervals.[
      • DeFronzo RA
      • Goodman AM
      The Multicenter Metformin Study Group. Efficacy of metformin in patients with non-insulin-dependent diabetes mellitus.
      ]
      The reductions in glucose variables observed in the current study are consistent with the findings of a previous study. A 29-week course of metformin 2550 mg reduced mean FPG by 58 mg/dL and mean HbA1c by 1.8% more than did placebo in a double-blind study of 289 moderately obese patients with type II diabetes.[
      • DeFronzo RA
      • Goodman AM
      The Multicenter Metformin Study Group. Efficacy of metformin in patients with non-insulin-dependent diabetes mellitus.
      ]
      The improvements in glycemic control have also been reported for other newly available antidiabetic agents in similar studies of patients with type II diabetes. Three daily dosages of the new sulfonylurea, glimepiride, were evaluated in a dose-response study.[
      • Goldberg RB
      • Holvey SM
      • Schneider J
      The Glimepiride Protocol #201 Study Group. A dose-response study of glimepiride in patients with NIDDM who have previously received sulfonylurea agents.
      ] Glimepiride reduced median FPG by 43 to 73 mg/dL and median HbA1c by 1.2% to 1.9% more than did placebo; however, the median within-treatment change from baseline was relatively slight for HbA1c, ranging from +0.2% to −0.3%.[
      • Goldberg RB
      • Holvey SM
      • Schneider J
      The Glimepiride Protocol #201 Study Group. A dose-response study of glimepiride in patients with NIDDM who have previously received sulfonylurea agents.
      ] Three daily dosages of the new α-glucosidase inhibitor, acarbose, significantly improved glucose variables compared with placebo, with between-group differences of 27 to 39 mg/dL for mean FPG and 0.8% to 1.0% for mean HbA1c.[
      • Coniff RF
      • Shapiro JA
      • Robbins D
      • et al.
      Reduction of glycosylated hemoglobin and postprandial hyperglycemia by acarbose in patients with NIDDM.
      ] The new thiazolidinedione, troglitazone, reduced mean FPG by 20 mg/dL and mean HbA1c by 0.6% more than did placebo.[
      • Iwamoto Y
      • Kosaka K
      • Kuzuya T
      • et al.
      Effects of troglitazone. A new hypoglycemic agent in patients with NIDDM poorly controlled by diet therapy.
      ] Of course, differences in study design and patient populations may have contributed to the between-study differences in glucose-lowering activity. For example, mean baseline FPG values ranged from 180 mg/dL in the troglitazone study[
      • Iwamoto Y
      • Kosaka K
      • Kuzuya T
      • et al.
      Effects of troglitazone. A new hypoglycemic agent in patients with NIDDM poorly controlled by diet therapy.
      ] up to approximately 230 mg/dL in the other two studies[
      • Goldberg RB
      • Holvey SM
      • Schneider J
      The Glimepiride Protocol #201 Study Group. A dose-response study of glimepiride in patients with NIDDM who have previously received sulfonylurea agents.
      ,
      • Coniff RF
      • Shapiro JA
      • Robbins D
      • et al.
      Reduction of glycosylated hemoglobin and postprandial hyperglycemia by acarbose in patients with NIDDM.
      ]; mean baseline HbA1c values were between 8% and 9% in the three studies.[
      • Goldberg RB
      • Holvey SM
      • Schneider J
      The Glimepiride Protocol #201 Study Group. A dose-response study of glimepiride in patients with NIDDM who have previously received sulfonylurea agents.
      ,
      • Coniff RF
      • Shapiro JA
      • Robbins D
      • et al.
      Reduction of glycosylated hemoglobin and postprandial hyperglycemia by acarbose in patients with NIDDM.
      ,
      • Iwamoto Y
      • Kosaka K
      • Kuzuya T
      • et al.
      Effects of troglitazone. A new hypoglycemic agent in patients with NIDDM poorly controlled by diet therapy.
      ] Nonetheless, these findings suggest that the slope of the dose-response curve is flatter for acarbose and troglitazone than it is for glimepiride.
      A flat dose-response curve has also been observed with sulfonylureas. In a double-blind, crossover study of 24 patients with type II diabetes,[
      • Stenman S
      • Melander A
      • Groop P-H
      • Groop L
      What is the benefit of increasing the sulfonylurea dose?.
      ] increasing the daily dose of glipizide from 10 mg to 20 mg or 40 mg was not associated with concomitant improvement in HbA1c. These findings underscore the importance of evaluating dose response in patients with type II diabetes and of defining the minimally and maximally effective doses.
      The dosages of metformin used in the current study were well tolerated. The most common adverse event, digestive disturbances, especially diarrhea, is consistent with previous experience.[
      • DeFronzo RA
      • Goodman AM
      The Multicenter Metformin Study Group. Efficacy of metformin in patients with non-insulin-dependent diabetes mellitus.
      ,
      • Hermann LS
      Metformin: a review of its pharmacological properties and therapeutic use.
      ,
      • Bailey CJ
      • Turner RC
      Metformin.
      ] Although the incidence of digestive disturbances has been reported to be dose-related,[
      • Hermann LS
      Metformin: a review of its pharmacological properties and therapeutic use.
      ,
      • Bailey CJ
      • Turner RC
      Metformin.
      ] there was no discernable association between the incidence of adverse events and metformin dosage in the current study. The higher incidence of digestive disturbances and diarrhea in the metformin groups versus placebo group was primarily due to dosages greater than 500 mg because the 500 mg dosage was not associated with any digestive disturbances. The incidences were essentially identical for dosages of 1000 through 2000 mg and somewhat higher for 2500 mg. This lack of a linear dose relationship for adverse events indicates that the use of gradual dose escalation at weekly intervals was at least partially successful.
      The results of the current study demonstrate that the antihyperglycemic activity of metformin is generally dose-dependent. Metformin is well tolerated with no significant differences in the incidence of adverse events compared with placebo, except for digestive disturbances. There were no cases of lactic acidosis or unexpected adverse events. When used in appropriately selected patients, metformin appears to be a safe and effective therapeutic alternative for physicians who wish to titrate drug therapy to achieve target glucose concentrations in individual patients with type II diabetes. The efficacy of metformin at dosages as low as 500 mg, along with the minimal risk of hypoglycemia, suggests that this agent will be useful in patients with mild to moderate hyperglycemia, and that higher dosages will be useful in those with moderate to severe disease. These data also suggest that most patients will achieve maximal efficacy at a daily dosage of 2000 mg (1000 mg bid), while some patients may achieve additional benefit if the dosage is increased to 2500 mg. We therefore conclude that twice-daily dosing may be sufficient for most patients with type II diabetes mellitus.

      Acknowledgements

      We are indebted to the following principal investigators and their staffs: Stephen Aronoff, MD, Endocrine Associates of Dallas, Dallas, Texas; David Bell, MD, Division of Endocrinology, Kirklin Clinic, University of Alabama at Birmingham, Birmingham, Alabama; Lawrence Blonde, MD, Ochsner Clinic, New Orleans, Louisiana; Dana Clarke, MD, Diabetes Health Center, Salt Lake City, Utah; Jamie A. Davidson, MD, Endocrine & Diabetes Associates of Texas, Dallas, Texas; Daniel Einhorn, MD, Diabetes, Endocrinology, Metabolic Disorders, San Diego, California; Stephen Farrow, MD, Division of Endocrinology, VA Medical Center, Allen Park, Michigan; Ronald B. Goldberg, MD, Diabetes Research Institute, University of Miami School of Medicine, Miami, Florida; Fred Hofeldt, MD, Department of Medicine, Denver General Hospital, Denver, Colorado; Willa Hsueh, MD, University of Southern California School of Medicine, Los Angeles, California; Charles Kilo, MD, FACP, Kilo Clinical Research, Ltd., St. Louis, Missouri; John Paul Lock, MD, Diabetes & Endocrinology Group, Worcester, Massachusetts; Jennifer Marks, MD, University of Miami School of Medicine, Miami, Florida; Janet McGill, MD, Washington University School of Medicine, St. Louis, Missouri; Pasquale Palumbo, MD, Mayo Clinic, Scottsdale, Arizona; Anne Peters, MD, UCLA Medical Plaza, Los Angeles, California; Lawrence Phillips, MD, The Emory Clinic, Atlanta, Georgia; Sherwyn L. Schwartz, MD, Diabetes & Glandular Disease Clinic, PA, San Antonio, Texas; Stephen Swartz, MD, Endocrinology-Hypertension Division, Brigham & Women’s Hospital, Boston, Massachusetts; Kenneth Ward, MD, Portland Diabetes & Endocrinology Center, Portland, Oregon; David Smith, MD, Memorial Research Medical Clinic, Long Beach, California; Matthew Weir, MD, Division of Nephrology, University of Maryland Hospital, Baltimore, Maryland. We also thank Shein-Chung Chow, PhD, and William A. Fox, MS, Bristol-Myers Squibb Company, Princeton, New Jersey for performing the statistical analysis.

      References

        • The Diabetes Control and Complications Trial Research Group
        The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus.
        NEJM. 1993; 329: 977-986
        • Ohkubo Y
        • Kishikawa H
        • Araki E
        • et al.
        Intensive insulin therapy prevents the progression of diabetic microvascular complications in Japanese patients with non-insulin-dependent diabetes mellitus: a randomized prospective 6-year study.
        Diabetes Res Clin Pract. 1995; 28: 103-117
        • Harris MI
        • Hadden WC
        • Knowler WC
        • Bennett PH
        Prevalence of diabetes and impaired glucose tolerance and plasma glucose levels in US population aged 20–74 yr.
        Diabetes. 1987; 36: 523-534
        • Johnson AB
        • Webster JM
        • Sum C-F
        • et al.
        The impact of metformin therapy on hepatic glucose production and skeletal muscle glycogen synthase activity in overweight type II diabetic patients.
        Metabolism. 1993; 42: 1217-1222
        • Nosadini R
        • Avogaro A
        • Trevisan R
        • et al.
        Effect of metformin on insulin-stimulated glucose turnover and insulin binding to receptors in type II diabetes.
        Diabetes Care. 1987; 10: 62-67
        • Hother-Nielsen O
        • Schmitz O
        • Andersen PH
        • et al.
        Metformin improves peripheral but not hepatic insulin action in obese patients with type II diabetes.
        Acta Endocrinol. 1989; 120: 257-265
        • Klip A
        • Leiter LA
        Cellular mechanism of action of metformin.
        Diabetes Care. 1990; 13: 696-704
        • Widén EIM
        • Eriksson JG
        • Groop LC
        Metformin normalizes nonoxidative glucose metabolism in insulin-resistant normoglycemic first-degree relatives of patients with NIDDM.
        Diabetes. 1992; 41: 354-358
        • DeFronzo RA
        • Barzilai N
        • Simonson DC
        Mechanism of metformin action in obese and lean non-insulin-dependent diabetic subjects.
        J Clin Endocrinol Metab. 1991; 73: 1294-1301
        • Perriello G
        • Misericordia P
        • Volpi E
        • et al.
        Acute antihyperglycemic mechanisms of metformin in NIDDM.
        Diabetes. 1994; 43: 920-928
        • Bailey CJ
        Biguanides and NIDDM.
        Diabetes Care. 1992; 15: 755-772
        • Campbell IW
        Metformin and the sulphonylureas: the comparative risk.
        Horm Metab Res Suppl. 1985; 15: 105-111
        • Reaven GM
        • Johnston P
        • Hollenbeck CB
        • et al.
        Combined metformin-sulfonylurea treatment of patients with non-insulin-dependent diabetes in fair to poor glycemic control.
        J Clin Endocrinol Metab. 1992; 74: 1020-1026
        • United Kingdom Prospective Diabetes Study Group
        United Kingdom Prospective Diabetes Study (UKPDS) 13: relative efficacy of randomly allocated diet, sulphonylurea, insulin, or metformin in patients with newly diagnosed non-insulin-dependent diabetes followed for three years.
        BMJ. 1995; 310: 83-88
        • Nagi DK
        • Yudkin J
        Effects of metformin on insulin resistance, risk factors for cardiovascular disease, and plasminogen activator inhibitor in NIDDM subjects. A study of two ethnic groups.
        Diabetes Care. 1993; 16: 621-629
        • Azerad E
        • Lubetzki J
        The treatment of adult diabetes with N-N dimethyl-diguanide (LA 6023).
        Bull Soc Med Hop (Paris). 1959; 3–4: 132-136
        • Azerad E
        The treatment of adult diabetes with N-N dimethyl-diguanide.
        Union Med Can. 1963; 92: 1251-1259
      1. Leblanc D. Contribution to the study of a synthetic hypoglycemic: NN dimethyl biguanide hydrochloride. Thesis Med. 1960;1–64.

      2. Duval D. Contribution to the study of the hypoglycemic action of biguanides. Thesis Pharm. 1960;77–109.

      3. Sterne J. Toxicology of the antidiabetic biguanides. Dimethylguanylguanide hydrochloride. 4th IDF Congress, Geneva, Switzerland. 1961;July 10–14:712–716.

      4. Sterne J, Duval D. Hypoglycemic effects of N-N-dimethylbiguanide. 3rd IDF Congress, Dûsseldorf, Germany. 1958:July 21–25:443–452.

        • Williams DA
        The comparison of several dose levels with a zero dose control.
        Biometrics. 1972; 28: 519-531
      5. Glucophage® (metformin hydrochloride tablets) 500 mg and 850 mg. Prescribing information. Bristol-Myers Squibb Company, Princeton, New JerseyFebruary 1995
        • Grant PJ
        The effects of high- and medium-dose metformin therapy on cardiovascular risk factors in patients with type II diabetes.
        Diabetes Care. 1996; 19: 64-66
        • DeFronzo RA
        • Goodman AM
        The Multicenter Metformin Study Group. Efficacy of metformin in patients with non-insulin-dependent diabetes mellitus.
        NEJM. 1995; 333: 541-549
        • Goldberg RB
        • Holvey SM
        • Schneider J
        The Glimepiride Protocol #201 Study Group. A dose-response study of glimepiride in patients with NIDDM who have previously received sulfonylurea agents.
        Diabetes Care. 1996; 19: 849-856
        • Coniff RF
        • Shapiro JA
        • Robbins D
        • et al.
        Reduction of glycosylated hemoglobin and postprandial hyperglycemia by acarbose in patients with NIDDM.
        Diabetes Care. 1995; 18: 817-824
        • Iwamoto Y
        • Kosaka K
        • Kuzuya T
        • et al.
        Effects of troglitazone. A new hypoglycemic agent in patients with NIDDM poorly controlled by diet therapy.
        Diabetes Care. 1996; 19: 151-156
        • Stenman S
        • Melander A
        • Groop P-H
        • Groop L
        What is the benefit of increasing the sulfonylurea dose?.
        Ann Intern Med. 1993; 118: 169-172
        • Hermann LS
        Metformin: a review of its pharmacological properties and therapeutic use.
        Diabete Metab (Paris). 1979; 5: 233-245
        • Bailey CJ
        • Turner RC
        Metformin.
        NEJM. 1996; 334: 574-579