Most patients with type 2 diabetes are overweight or obese, overweight or obesity increases the risk of developing type 2 diabetes and obesity per se is strongly associated with multiple cardiometabolic risk factors. However, many antidiabetic treatments increase body weight. The oral antidiabetic agent, metformin, has been evaluated in hundreds of clinical studies in diverse patient populations during approximately five decades of clinical use. This review summarizes the effects of metformin on body weight, with special reference to studies of longer duration (6 months) as both diabetes and obesity are long-term conditions. Approximately half of studies in drug-naive type 2 diabetic patients demonstrated significant weight loss with metformin compared with baseline or comparator drugs, although pooled analyses have suggested no significant effect versus placebo. Similarly, metformin has been shown to induce weight loss in obese nondiabetic populations, although studies of long duration in this population are scarce. Metformin does appear to mitigate the adverse effects of insulin on body weight. The weight-neutral or weight-sparing effects of metformin constitute a therapeutic advantage in diabetes management where other first-line oral antidiabetic treatments often promote clinically significant weight gain.
The management of overweight or obesity is an important facet of overall diabetes management. Several classes of oral antidiabetic agents are associated with increased body weight, including insulin secretagogues (sulphonylureas and meglitinides), thiazolidinediones (TZDs) and insulin itself.1, 2 Newly developed agents acting through the incretin hormone system are weight neutral or promote weight loss,3, 4 but these agents are not approved for first-line use in some regions, including Europe. Two classes of oral antidiabetic agents have not been associated with weight gain: the biguanides (of which metformin is now effectively the sole representative) and α-glucosidase inhibitors. Current global guidelines for the management of type 2 diabetes recommended metformin as initial pharmacotherapy for all type 2 diabetes patients and even support such first-line use for non-overweight patients1 and concurrent use with lifestyle modification as an intervention partnership at diagnosis and throughout treatment.5 This article considers the role of overweight and obesity in the pathogenesis of dysglycemia and reviews the effects of metformin on body weight and adiposity.
Role of obesity in the pathogenesis and management of type 2 diabetes
Dysglycemia is associated strongly with a tendency to overweight or obesity.6 For example, the mean body mass index (BMI) in subjects with impaired glucose tolerance (IGT) was 34 kg/m2 in the Diabetes Prevention Program (DPP)7 and 31 kg/m2 in the Finnish Diabetes Prevention Study (FDPS),8 each consistent with a diagnosis of class I obesity according to criteria from the World Health Organization.9 Most type 2 diabetic patients are overweight or obese. Recent cross-sectional data revealed that 37% of 44 042 Swedish patients with type 2 diabetes had BMI 30 kg/m2).10 Other recent analyses have identified 75% of Brazilian type 2 diabetes patients as overweight (BMI>25 kg/m2) and 30% as obese,11 while in the United Kingdom12 and United States,13 85 and 86%, respectively, were overweight and 52 and 55%, respectively, were obese. The prevalence of both obesity and diabetes is increasing rapidly.6
The pathophysiology of obesity and diabetes are closely linked. Analyses from the Nurses’ Health Study and the Health Professionals Follow-up Study showed that the risk of diabetes increased in line with BMI and that obesity, especially abdominal obesity, increased the risk of subsequent diabetes by 10- to 11-fold.14, 15 A cross-sectional analysis in Japanese subjects examined the prevalence of type 2 diabetes according to the obesity phenotype.16 A similar prevalence of type 2 diabetes was found in subjects with either elevated BMI or waist circumference and multivariate adjustment for cardiometabolic risk factors had little effect on these associations. Adding adjustment for BMI to the model did not appreciably attenuate the effect of high waist circumference on the risk of having type 2 diabetes, but adjustment for waist circumference abolished the association between BMI and diabetes. Abdominal obesity is now recognized as a core component of the metabolic syndrome according to the International Diabetes Federation (IDF).17
A reduction in adiposity reduces the risk of developing type 2 diabetes and cardiometabolic risk factors. Intensive lifestyle interventions in the DPP18 and the FDPS8 reduced the risk of developing diabetes by 58% relative to standard lifestyle advice in each study. Changes in body weight in the subjects receiving the intensive lifestyle intervention in the DPP paralleled effects on the incidence of type 2 diabetes and associated cardiometabolic risk factors (the effects of metformin are described below).19, 20 In the FDPS, the risk of developing type 2 diabetes among subjects who adopted healthier lifestyles (low fat/high fiber diet) was about half of that in observed in a subgroup of subjects who did not.21 Weight loss by pharmacological means also reduces the risk of developing type 2 diabetes and improves the cardiovascular risk profile of obese patients.22, 23, 24, 25
Thus, patients who are overweight or obese are at a markedly higher risk of developing type 2 diabetes and patients who lose weight are more likely to avoid the burden of morbidity and early mortality associated with a diagnosis of diabetes and the metabolic syndrome. Indeed, the term ‘diabesity’ was coined as long ago as 1980 in the United States to describe the close pathogenetic links between obesity and type 2 diabetes.26 It is clearly important not to add to the burden of obesity in diabetes if possible, especially in patients who are already overweight.
Many clinical trials of varying design have evaluated the effects of metformin on body weight, glycemia and other cardiometabolic parameters. Studies on metformin reviewed here were selected from a PubMed search for ‘metformin and weight’, limited to ‘humans’ and ‘randomized controlled trial’ or ‘meta-analysis’. In some cases, further references were identified from reference lists. Studies in which patients had previously received metformin were excluded, except where indicated.
Preference was given to meta-analyses and studies of longer duration. Long-term clinical evaluations such as the DPP18 or the FDPS8 typically demonstrate an initial reduction in body weight during the early months of therapy, followed by a tendency for the effect to reverse during continued treatment. Furthermore, metformin has been shown to reduce food intake in studies of up to 15 days,27, 28 and the gastrointestinal side effects of metformin tend to appear early in therapy and may exert undue influence on the results of short-term studies.29
This is not a formal meta-analysis. However, the quality of evaluations of metformin quoted below have been graded in terms of their methodological quality, using the classification system devised by the American Diabetes Association for use in their diabetes management guidelines.30 Briefly, this provides four classifications, ranging from ‘clear evidence from well-conducted, generalizable, randomized controlled trials’ (level A) to ‘Expert consensus or clinical experience’ (E). Most trials quoted in this review are randomized, controlled trials corresponding to evidence level A, although some trials correspond most closely to evidence level C (‘supportive evidence from poorly controlled or uncontrolled studies’).
Effects of metformin on body weight in patients with type 2 diabetes
Patients suboptimally controlled on diet
The effect of metformin on body weight in randomized, controlled trials (evidence level A) in patients suboptimally controlled by diet was variable, with about half of studies demonstrating significant reductions in body weight with metformin, relative to baseline or comparator agents (Table 1 a).31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45 The longest treatment duration available was the 10-year follow up of metformin-treated overweight patients in the UK Prospective Diabetes Study (UKPDS).31 Patients receiving conventional, diet-based treatment in this study gained about 2 kg over the course of the trial, with a slightly smaller weight gain of about 1.5 kg in the metformin group and of about 4 kg in the glibenclamide group (Figure 1). The subsequent addition of metformin to sulphonylurea treatment in a subgroup of patients in this trial did not result in a significant weight change.
A Diabetes Progression Outcomes Trial randomized a population of 4360 patients uncontrolled by lifestyle intervention to monotherapy with metformin, glibenclamide or rosiglitazone for 4 years.32 Patients in the metformin group lost weight, on average, while weight gains occurred in the other treatment groups. The weight increase with rosiglitazone was significant compared with either of the other treatments. Other trials involved randomized treatment durations of up to 1 year. There was no indication of metformin-induced weight gain, but little indication of marked or significant weight loss in metformin groups relative to placebo, although significant differences were noted in comparison with TZD or sulphonylurea treatment. Further 6-month46 and 1-year47 randomized evaluations (evidence level A) of metformin at doses up to 3000 mg/day in type 2 diabetes patients reported no difference in body weight, but data were not presented.
Immediate prescription of oral antidiabetic combination therapy is an option where severe hyperglycemia at presentation is unlikely to be controlled adequately by oral antidiabetic monotherapy.48 A 2-year randomized study (PRESERVE-β; evidence level A) evaluated initial treatment of drug-naive patients with co-administered combinations of glibenclamide or nateglinide in combination with metformin.41 Changes in weight were relatively small (−0.4 kg for metformin–nateglinide and +0.8 kg for metformin–glibenclamide), despite the relatively long duration of this study. This was presumably due to synergy between the components of the combination allowing the use of lower doses of the secretagogues than would be required as monotherapy, as seen elsewhere with similar combinations;49 mean final doses of metformin–nateglinide were 1105 and 357 mg, respectively, and mean final doses of metformin–glibenclamide were 1105 and 5.1 mg, respectively, and these were sufficient to produce 2-year reductions in HbA1C of 1.2 and 1.5%, respectively. Early use of metformin-based combinations in this way may provide another means of achieving marked antihyperglycemic efficacy while limiting weight gain. Changes in body weight with metformin-based combination therapy in patients who had previously received oral antidiabetic treatment are described in the following section.
Metformin may influence body fat distribution in patients with type 2 diabetes uncontrolled by diet. A post hoc analysis from a randomized study of 26 weeks’ duration (evidence level A) showed that metformin significantly reduced visceral fat mass versus placebo while rosiglitazone did not.37, 50 In addition, an uncontrolled study reported in abstract form (evidence level C), employed computed tomography scanning to measure abdominal fat distribution in African-American type 2 diabetes patients receiving metformin. The fat in the visceral and subcutaneous adipose tissue compartments was reduced by 16% (P<0.025) and by 6% (P=0.11), respectively, suggesting a redistribution of adiposity from visceral to subcutaneous depots.51
Patients suboptimally controlled on oral antidiabetic monotherapy
Two randomized trials, of 12 months’ and 7 months’ duration, provided data on body weight in patients previously treated with glibenclamide and randomized to receive metformin, glibenclamide or these two agents in combination (Table 1b).36, 43 In either study, mean body weight was significantly lower in the metformin group compared with sulphonylurea-containing groups at study end. The QUARTET study randomized 639 type 2 diabetes patients suboptimally controlled on a sulphonylurea to receive double-blind treatment with additional metformin or a TZD (pioglitazone) for 1 year (Table 1b).42 Patients receiving the resulting sulphonylurea–TZD combination gained 2.8 kg body weight, on average, compared with an average reduction of 1.0 kg in the sulphonylurea–metformin group (information of the significance of this effect was not provided). A shorter randomized study (3 months; evidence level A) showed that adding metformin to the regimens of patients suboptimally controlled on sulphonylurea resulted in smaller mean gains in weight (+0.4 kg) than adding insulin lispro (+3.4 kg) or NPH insulin (+2.3 kg).52
Where patients become uncontrolled on submaximal doses of metformin, increasing the metformin dose or adding a second agent are alternative management strategies for restoring glycemic control. Two randomized 6-month studies have evaluated these options in substantial patient populations (551 and 766 patients) using a TZD as the second agent (Table 1c).44, 45 Similar changes in weight occurred in the metformin–TZD arms of the studies (1.3 and 1.8 kg), and similar weight loss occurred in the intensified metformin therapy arms (−0.9 and −1.8 kg). The authors of one study speculated that the presence of metformin in the combination regimen may have mitigated weight gain with the TZD.44
Metformin as an adjunct to insulin
Metformin acts by improving the action of insulin, particularly in the liver,53 and metformin has been evaluated as an adjunct to insulin treatment. Most trials in this area are of <6 months’ duration. However, as weight gain is an early event after initiation of insulin, the short-term effects of metformin on body weight may be relevant in the setting of concomitant insulin use. A number of these trials, listed in Table 2, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66 demonstrated significant reductions in body weight between patients receiving insulin and metformin compared with insulin and placebo54, 56, 59, 60, 62 although other studies did not.55, 57, 58, 61, 63, 64 Two trials demonstrated amelioration of insulin-induced increase in body weight with metformin compared with intensified insulin.65, 66
Weight gain with insulin or insulin secretagogues arises, in part, through greater retention of energy due to reduced glycosuria, along with fluid retention and an anabolic response to insulin. One evaluation of metformin in combination with insulin calculated as the change in weight per 1% decrease in HbA1C, thus correcting for differences in glycemic control between studies.67 This trial suggested that combinations of metformin with insulin were associated with lower weight gain corrected for changes in glycemia than regimens containing a sulphonylurea or a TZD. The mechanism by which metformin may limit weight gain during insulin therapy is unclear. Reduced energy intake may be more important in this regard than changes in energy expenditure,65, 68, 69, 70, 71 and both metformin and insulin have been shown to modulate leptin secretion.72, 73
In summary, metformin may assist in ameliorating weight gain associated with insulin. It should be noted, however, that the principal benefits of addition of metformin to insulin probably arise from improved glycemia, lower insulin requirements secondary to improved insulin action and a consequently lower risk of hypoglycemic reactions.54, 56, 74, 75 Reductions in insulin requirements of 25% or more have been noted in patients receiving metformin as an adjunct to insulin therapy.67, 75
A meta-analysis of 11 trials published in 1995 which randomized patients to metformin or a sulphonylurea for 6–52 weeks included nine trials providing information on changes in body weight.76 Sulphonylurea treatment was associated with increased body weight in every trial. Decreases from baseline in body weight in the metformin group occurred in seven trials and body weight increased in two, with all but one of the trials demonstrating a significant weight reduction on metformin compared with increases in body weight on a sulphonylurea. Overall, there was an average treatment difference for metformin versus sulphonylurea of −4 kg.
A more recent meta-analysis pooled data from randomized evaluations of metformin in comparison with any other oral treatment, including placebo (12 trials), and sulphonylureas (13 trials).77 This analysis, which included evaluations of 12 weeks or greater duration, confirmed the results of the earlier meta-analysis in terms of a significant weight loss on metformin versus sulphonylurea (standardized percent mean difference in body weight or BMI of −0.45 (95% CI −0.80 to −0.10) in favor of metformin; P<0.0001). However, there was no difference in this parameter between metformin and placebo groups (P=0.92). Another meta-analysis reached a similar conclusion with regard to comparisons between placebo and sulphonylureas after analyzing 19 randomized, controlled trials.78
Effects of metformin on body weight in nondiabetic patients
Obese and/or abdominally obese subjects
A summary of prospective studies of at least 6 months’ duration in obese or abdominally obese subjects is shown in Table 3.79, 80, 81, 82 The first Biguanides and Prevention of the Risks in Obesity study 1 enrolled 324 patients with abdominal obesity (waist–hip ratio 0.95 in men and 0.80 in women) in order to recruit a trial population containing a high prevalence of insulin resistance.79 Subjects were also free of cardiovascular disease and diabetes at baseline. Randomization was to a low dose of metformin (850 mg) or placebo for 1 year; all patients also received advice on diet and exercise. There was a trend (P<0.06) toward a reduction in weight in the metformin group.
A randomized trial in 150 women with BMI >30 kg/m2 (of whom 15 had type 2 diabetes) evaluated sibutramine, metformin or orlistat for 6 months.80 Weight, waist and BMI were reduced in all groups (P<0.0001). Small 7-month studies in nondiabetic overweight (BMI >28 kg/m2)81 or morbidly obese subjects (mean BMI 43 kg/m2)82 demonstrated decreases in body weight with metformin, although only the latter study demonstrated a reduction in waist circumference. Two of the studies summarized in Table 3 suggest a greater weight reduction with metformin relative to placebo, but these differences did not achieve statistical significance. A systematic review of randomized, controlled evaluations of metformin (duration 15 days to 1 year) in populations without diabetes or polycystic ovary syndrome (PCOS) found insufficient evidence to consider metformin as a potential therapy for weight loss in overweight or obese patients.83
An uncontrolled, retrospective analysis (evidence level C) followed the progress of 26 nondiabetic women with hyperinsulinaemia and progressive weight gain refractory to diet and exercise who received a 1-year intervention of metformin together with a hypocaloric diet.84, 85 All but one patient (96%) lost 5% of initial body weight at 6 months, and 21 patients (81%) lost 10% of initial body weight at 12 months. Longer-term (2–4 years) follow-up of 21 of these women demonstrated weight maintenance in 19 women (90%), with the final body weight correlating strongly (P<0.001) with the initial weight loss during the original 1-year intervention. The design of this analysis requires caution in interpreting these data. However, the hyperinsulinaemia of these patients suggests the presence of insulin resistance and it is possible that this population, termed ‘syndrome W’,86 may be enriched with individuals sensitive to effects of metformin on body weight.
Subjects with IGT
Principal diabetes prevention trials involving metformin are also summarized in Table 4.7, 87, 88, 89 The DPP enrolled a population of 3234 subjects with IGT and high–normal fasting plasma glucose and a mean BMI at baseline of 34 kg/m2 and randomized them to receive standard lifestyle advice plus metformin (n=1073) or placebo (n=1082) or an intensive lifestyle intervention program (n=1079).7, 18 Weight loss occurred in all groups, reaching a maximum after 6–12 months and reversing to some extent afterward. Mean weight loss in the placebo, metformin and intensive lifestyle groups was 0.1, 2.1 and 5.6 kg, respectively (P<0.001 for differences between groups). A corresponding effect of metformin on waist circumference was observed. Among subjects in the metformin group of the DPP who did not meet the NCEP/ATPIII waist circumference criterion for diagnosis of metabolic syndrome (waist 40 inches), 15% met this criterion at 3 years, compared with 33% of the placebo group (P<0.001).19 Conversely, of patients with waist >40 inches at baseline, 89% of the metformin group still had high waist circumference at study end, compared with 93% of the placebo group (P<0.001).19
The Indian DPP randomized 531 subjects with IGT to standard lifestyle advice, metformin, a DPP-like intensive lifestyle intervention, or intensive lifestyle intervention plus metformin for 3 years.87 Modest but significant increases in body weight occurred in the standard and intensive lifestyle intervention groups, but not in subjects receiving metformin-containing regimens. There was no apparent difference between the metformin and metformin plus intensive lifestyle intervention arms in terms of effects on body weight. No significant changes in waist circumference occurred in this study. A study in members of the Swedish Botnia cohort randomized 40 subjects with IGT to metformin or placebo for 6 months.89 A decrease in body weight was observed in the metformin group, but this did not achieve statistical significance relative to placebo (P=0.07). Finally, no effect on BMI was observed in a further 3-year study in 321 Chinese subjects with IGT, some of whom were treated with diet and exercise or metformin.88 The design of this study was weaker (level C) than those described above, as subjects were allocated to treatment by geographical area rather than in a standard randomization procedure.
Polycystic ovary syndrome
The observation that obesity and insulin resistance may play an important role in the pathogenesis of PCOS has prompted intense clinical interest in the potential of metformin for the management of this condition.90 Details of trials of at least 6 months’ duration are shown in Table 5.81, 91, 92, 93, 94 In general, these were well designed, randomized studies (evidence level A). However, one study92 randomized patients to one of two doses of metformin, without a non-metformin control group, and was thus categorized as evidence level C for the purposes of this review.
As in other populations, the results of individual studies were variable, with reports of significant weight loss with metformin compared with baseline but no evidence of significant changes relative to control groups in studies of at least 6 months’ duration. However, these studies were heterogeneous in design, and numbers of patients were usually small. Nevertheless, a systematic review and meta-analysis of 13 randomized, controlled trials published in 2003 found no overall effect of metformin on body weight.95
One of these trials evaluated the dose relationship of the effects of metformin on body weight in a population of women with PCOS stratified for the severity of obesity and treated for 8 months.92 Weight reductions were observed in obese patients (defined as BMI 30 and <37 kg/m2) and morbidly obese patients (defined as BMI 37 kg/m2). In the obese group, the weight loss with metformin appeared to be dose dependent; mean final body weight in the metformin 2550 mg group was 3.6 kg lower than the corresponding value at the start of the trial, compared with a difference of 1.1 kg for the metformin 1500 mg group (P=0.04 between treatments). There was no significant dose relationship for the overall population or for morbidly obese patients, however.
Four well-designed studies (evidence level A) have evaluated the effects of metformin on body weight and other parameters in pediatric populations. One randomized, double-blind trial in children and adolescents with type 2 diabetes demonstrated similar antihyperglycemic benefits to those observed in adults.96 These data underpinned an indication for the use of metformin in pediatric type 2 diabetes in Europe and in the United States. Mean weight gain did not differ between metformin and placebo groups. However, it should be noted that this nominally 4-month study was terminated after 2 months due to safety concerns over uncontrolled hyperglycemia in the placebo group. A further randomized, double-blind trial evaluated metformin versus glimepiride in 285 type 2 diabetes patients aged 8–17 years who were suboptimally controlled by lifestyle intervention or a single oral antidiabetic agent.97 Marked and significant differences occurred in the effects of metformin and glimepiride on body weight (mean changes 0.55 and 1.97 kg, respectively, P=0.005) and BMI (mean changes −0.33 and 0.26 kg/m2, respectively, P=0.003), despite there being no significant differences between treatments with respect to glycemic parameters.
The remaining two trials were in nondiabetic pediatric populations. Metformin was compared with placebo in 29 obese adolescents (mean BMI 40 kg/m2) over a treatment period of 6 months.98 Mean BMI decreased by 0.5 kg/m2 in the metformin group and increased by 0.23 kg/m2 in the placebo group (P<0.02). These data are broadly similar to those from adults, described above, although further studies are required for a definitive assessment of the effect of metformin on body weight in this population. A recent randomized, double-blind 6-month crossover study in 28 pediatric subjects has confirmed these findings; treatment with metformin versus placebo was associated with significant improvements in body weight, BMI and waist circumference.99
Metformin and obesity-associated cardiometabolic risk factors
Amelioration of insulin resistance, considered by some to be a cardiovascular risk factor in its own right, is a clinically important mechanism of metformin,53 although this agent primarily controls glycemia in type 2 diabetes through a reduction in hepatic glucose output.100 Treatment with metformin often results in modest improvements in lipid parameters, notably triglycerides and LDL cholesterol.101 There is usually no change in HDL cholesterol101 or blood pressure102 with metformin treatment, although an improvement in endothelial function has been observed with this agent in patients with type 2 diabetes,103 and in nondiabetic populations with the metabolic syndrome,104 or PCOS.105
It should be noted that changes in classical cardiometabolic risk factors are insufficient to account for the improvement by metformin in clinical cardiovascular outcomes in type 2 diabetes observed in the UKPDS.31 A number of insulin-dependent and insulin-independent mechanisms have been proposed to account for the cardioprotective actions of metformin, including actions on the vasculature,106 mitochondria,107 hemostasis108 and the formation of advanced glycation end-products,109 among others.
Treatment with metformin was associated with reductions in body weight in about half of the studies in drug-naive type 2 diabetic patients reviewed above. Meta-analyses do not support a consistent weight-reducing effect of metformin relative to placebo in diabetic or nondiabetic patients. However, the weight-reducing effect of metformin appears greater when co-administered with insulin or a sulphonylurea, compared with either agent alone. These findings contrast with observations of increased body weight with insulin secretagogues and TZDs (α-glucosidase inhibitors are the only other commonly used oral antidiabetic agent not associated with weight gain110). Treatment-induced weight gain can be distressing and may hinder compliance with treatment.111 A neutral effect on body weight represents an advantage in other clinical settings, for example in elderly patients who are at risk of unintentional weight loss.112, 113 Certainly, as metformin did not reduce body weight in the UKPDS, it is unlikely that its effects on glycemia or cardiovascular outcomes resulted from a weight-dependent mechanism, in contrast to antiobesity drugs which rely on weight loss for a substantial proportion of their effects on cardiometabolic risk factors.22, 23, 25
While the favorable effects of metformin on body weight are modest, considerations relating to body weight have profoundly affected its therapeutic use. The antihyperglycemic efficacy of metformin does not vary with adiposity to a clinically significant extent.114 Nevertheless, because metformin was administered to overweight patients in the UKPDS, and induced significant improvements in cardiovascular outcomes compared with diet-based treatment,31 national and international guidelines (reviewed by Consoli et al.115) recommended that metformin should be used as first-line pharmacological therapy in overweight patients, but not preferentially in non-overweight patients. This position has shifted, with recent global diabetes management guidelines from the IDF,1 which regard metformin as the drug of choice for overweight patients and as a first-line pharmacological antidiabetic option for non-overweight patients, and a recent joint statement from European and US expert societies which regard metformin as the initial pharmacological management option for all patients.5
The effects of metformin on body weight are variable between patient populations. It is clear, however, that metformin does not increase body weight, in contrast to other oral antidiabetic agents, and may help to limit the weight gain associated with insulin- or sulphonylurea-based regimens. The improvements in glycemia and clinical cardiovascular outcomes demonstrated for metformin are independent of changes in body weight or adiposity. This observation supports current recommendations relating to the suitability of metformin for initiation of oral antidiabetic pharmacotherapy in patients with type 2 diabetes, irrespective of body weight. There is no compelling evidence for the use of metformin to control body weight in nondiabetic populations, although other benefits in terms of a reduced risk of diabetes in subjects with IGT18, 88 or reduced symptoms associated with insulin resistance in women with PCOS95 may be observed.