Professor Evan Begg, Clinical Pharmacology, Department of Medicine, University of Otago–Christchurch, Private Bag 4345, Christchurch 8041, New Zealand. Tel:+ 64 3 364 1055 Fax:+ 64 3 364 1003 E-mail: firstname.lastname@example.org
WHAT IS ALREADY KNOWN ABOUT THIS SUBJECT
• Metformin, unlike the other major antihyperglycaemic drugs, is not associated with weight gain.
• Ghrelin is an appetite-stimulating hormone whose concentrations vary in relation to food, obesity and diabetes control.
• Reports are conflicting about how metformin affects ghrelin concentrations, and this study was aimed at resolving this issue in patients with Type 2 diabetes.
WHAT THIS STUDY ADDS
• In this study an increase in ghrelin concentrations was seen in response to metformin treatment in patients with Type 2 diabetes.
• This effect was opposite to what might be expected if the effect of metformin on weight control was mediated via suppression of ghrelin.
• It is likely that the ghrelin response was secondary to improved glycaemic control.
• Meal time changes in appetite and satiety did not correlate with changes in ghrelin, which suggests ghrelin may not be important in meal initiation.
AIMS Metformin treatment of Type 2 diabetes is not usually associated with weight gain, and may assist with weight reduction. Plasma ghrelin concentrations are inversely associated with obesity and food intake. Metformin might therefore affect ghrelin concentrations, although previous studies have shown variable results in this regard. The primary aim of this study was to determine the effect of metformin on plasma ghrelin, appetite and satiety in patients with Type 2 diabetes.
METHODS Eighteen patients with Type 2 diabetes were studied before and after 6 weeks of metformin treatment, which was titrated to 1 g b.d. On the study days patients were fed standard meals of 390 kcal at 08.00 and 12.30 h, plasma samples were collected at 15- and 30-min intervals, and appetite and satiety were measured on visual analogue scales. Changes in the area under the concentration–time curves (AUCs) of plasma ghrelin, insulin, glucose, appetite and satiety were assessed and examined for correlations with metformin AUCs. Changes in fasting adiponectin and leptin were also measured.
RESULTS Treatment with metformin increased the mean AUC (07.30–16.30 h) of plasma ghrelin by 24% (P= 0.003), while decreasing those of glucose by 19% (P < 0.001) and insulin by 19% (P= 0.001). No changes were detected in hunger and satiety, or in fasting adiponectin or leptin concentrations. There were no clear correlations between metformin plasma concentrations (AUC) and changes in plasma glucose, insulin or ghrelin.
CONCLUSIONS Treatment of Type 2 diabetes with metformin was associated with increased plasma ghrelin concentrations, without associated changes in hunger and satiety.
Metformin is widely used for the treatment of Type 2 diabetes. It has been shown to reduce mortality and cardiovascular events in obese patients with Type 2 diabetes [1, 2]. The principle action of metformin is to reduce hepatic gluconeogenesis by inhibiting hepatocyte mitochondrial respiratory chain oxidation . Metformin treatment in Type 2 diabetes is associated with modest weight loss or no weight change [4, 5], in contrast to sulphonylureas, thiazolidinediones and insulin, which are all associated with weight gain [6, 7].
Ghrelin is an appetite-stimulating hormone that increases growth hormone secretion and food intake in animals and humans [8, 9]. Ghrelin was first identified as an endogenous agonist of the growth hormone secretagogue receptor . Ghrelin receptors (GHS-R type 1a) are present in many organs. Plasma concentrations of ghrelin exhibit a diurnal rhythm and are inversely associated with food intake (acute effect) [11, 12] and obesity (chronic effect) . Fasting plasma ghrelin concentrations are lower and postprandial decreases in ghrelin are less in patients with Type 2 diabetes than in those without diabetes . A reciprocal relationship between ghrelin (increase) and insulin (decrease) is seen in response to treatment with growth hormone . The appetite-stimulating effects of ghrelin are thought to be mediated via the arcuate nucleus of the hypothalamus and the messenger peptides neuropeptide Y and Agouti-related protein .
Plasma concentrations of the adipokines adiponectin and leptin are associated with body composition and correlate directly and inversely, respectively, with plasma ghrelin concentrations [17, 18]. Metformin has been reported to have no effect on plasma adiponectin [19, 20], to have no effect on or modestly reduce plasma leptin with short-term use [21, 22], and to reduce plasma leptin with long-term use .
A previous study of metformin in polycystic ovarian syndrome showed an increase in plasma ghrelin in patients treated with metformin . In contrast, a recent case–control study showed that metformin prolongs the postprandial fall in plasma ghrelin concentrations in Type 2 diabetes . The effects of metformin on ghrelin therefore remain unresolved.
The primary aim of this study was to determine the effects of metformin on plasma ghrelin, appetite and satiety. The secondary aims were to determine the effect of metformin on adiponectin and leptin, and to define any relationships between metformin plasma concentrations and changes in glucose, insulin and ghrelin.
Twenty patients referred to the Christchurch Diabetes Centre with a diagnosis of Type 2 diabetes were recruited. All patients had confirmed diabetes (fasting plasma glucose ≥7 mmol l−1 on two occasions), had been treated with lifestyle advice (>3 months prior to entering the study) and were on no medications for diabetes. All were overweight or obese [body mass index (BMI) >25 kg m−2] with a calculated creatinine clearance >50 ml min−1. Metformin treatment was considered clinically indicated for all patients. No patient was concurrently treated with any other drug known to affect plasma glucose concentrations or appetite. The study was approved by Upper South B Regional Ethics Committee, New Zealand, and all study participants gave written informed consent.
Patients were studied on two separate days, about 6 weeks apart. Day 1 was prior to starting treatment with metformin and day 2 was 6–8 weeks after starting treatment with metformin. Following the first study day metformin was started at 500 mg once daily and titrated up by 500 mg day−1 weekly to the treatment dose of 1000 mg twice daily. Six weeks was chosen to allow patients to reach steady state in terms of metformin concentrations and, hopefully, glucose/insulin homeostasis, and to precede any significant change in body morphology. Additional study at later time points was not possible within the resources of the study.
Patients attended on the study days after an overnight fast and were fed standard solid 390-kcal meals at 08.00 and 12.30 h, made up of protein 20 g, carbohydrate 57 g and fat 9 g. A medical history and examination were conducted on both study days. Height, weight and waist and hip circumferences were measured, and body fat was estimated by bioelectrical impedance (TBF-310™; Tanita, Tokyo, Japan). Hunger and satiety were recorded using validated 100-mm visual analogue scales  at 30-min intervals on the study days. On the second study day the morning dose of metformin was administered under supervision. Compliance was evaluated by questioning the subjects and by comparing the predose trough concentration of plasma metformin with the calculated postdose trough concentration. A difference of >40% was considered to suggest noncompliance.
An intravenous cannula was inserted into a forearm vein for blood sampling. All blood samples were centrifuged immediately and the plasma separated and frozen at −32°C. Serial blood samples were drawn at 15-min intervals from half an hour before until 2.5 h after meals, and otherwise at 30-min intervals during the study days. Plasma glucose, insulin and ghrelin were measured in these samples (27 samples per patient study day). Plasma metformin was measured on the second study day in samples drawn predose (0 h) and 0.5, 1, 1.5, 2, 4, 6 and 8 h after the metformin dose. Plasma leptin and adiponectin were measured in the fasting specimens. All specimens were stored at −32°C and analysed as a batch on completion of the study. Fasting plasma was analysed for creatinine, sodium, potassium, liver function tests and lipids during routine laboratory runs.
Total ghrelin concentrations were measured by radioimmunoassay (RIA) as previously described , with limit of quantification 10 pmol l−1 and inter- and intraday coefficients of variation of 9.2 and 11.4%, respectively, in the concentration range 50–220 pmol l−1. Metformin was measured by high-performance liquid chromatography as previously described , with limit of quantification 20 µg l−1 and inter- and intraday coefficients of variation of <9%. RIA kits were used for determination of plasma, adiponectin and leptin concentrations (Linco Research Inc., St Charles, MO, USA). Concentrations of insulin were measured using a Roche Elecsys 2010™ analyser (Roche Diagnostics, Indianapolis, IN, USA) and of glucose using an Abbott Architect ci8200™ analyser (Abbott Laboratories, Abbott Park, IL, USA).
The possibility that metformin might interfere with the ghrelin assay was examined as follows. Two separate plasma samples (one male, one female) were split into three equal volumes. For each sample there were three categories: (i) no metformin added, (ii) metformin added at 0.33 mg l−1, and (iii) metformin added at 3.33 mg l−1. The ghrelin concentrations measured in these samples were 128, 138 and 132 pmol l−1 and 320, 320 and 307 pmol l−1, respectively, using the study assay. These results show that metformin is unlikely to influence the determined ghrelin concentrations in this study.
In order to detect a mean change in ghrelin concentrations of 15 pmol l−1, with a standard error of 25 pmol l−1 using a two-tailed test, 20 patients were required (α= 0.05, β= 0.20). Statistical significance was set at P≤ 0.05.
Fasting glucose, insulin and ghrelin concentrations were taken as the mean of the three serial fasting results prior to breakfast. Descriptive data are presented as mean ± standard deviation unless otherwise stated. Changes from baseline to week 6 are reported as change in the mean ± standard error of the mean. Differences in paired data were evaluated using Student's paired t-test (two-tailed). Areas under the each of the concentration vs. time curves (AUCs) were calculated using the trapezoidal rule for metformin, ghrelin, glucose, insulin, hunger and satiety. Linear regression analyses were performed to compare metformin AUCs with changes in the other parameters and to identify variables independently associated with ghrelin AUC.
Analysis of data was carried out using SPSS 15.0 for Windows (SPSS Inc., Chicago, IL, USA), Graph Pad Prism™ (GraphPad Software, San Diego, CA, USA) and Excel™ (Microsoft Corporation). The homeostasis model assessment of steady-state β cell function (%B) and insulin sensitivity (%S) (HOMA2)  and the quantitative insulin-sensitivity check index (QUICKI)  were calculated from fasting glucose and insulin values. QUICKI = 1/[log(I0) + log(G0)], where I0 is fasting insulin (µU ml−1) and G0 is fasting glucose (mg dl−1).
Twenty patients completed day 1 of the study. One patient was subsequently excluded, as she was diagnosed with Type 1 diabetes. A second patient was also excluded after repeatedly failing to attend for the second study day and then found not to be taking metformin when he did attend. The baseline characteristics of the 18 subjects are shown in Table 1.
Table 1. The effects of 6 weeks of metformin on 18 patients with Type 2 diabetes
Data are means ± SD for absolute values and means (95% CI) for changes. P-values were determined by paired, two-tailed t-test. BMI, body mass index; QUICKI, quantitative insulin-sensitivity check index; HDL, high-density lipoprotein; LDL, low-density lipoprotein.
10 / 8
Creatinine clearance (ml min–1)
104.6 ± 11.5
103.0 ± 11.4
−1.6* (−2.7 to −0.5)
112.2 ± 9.9
113.5 ± 9.7
1.3* (−0.4 to 3.1)
123.9 ± 13.9
122.5 ± 12.6
−1.4 (−3.2 to 0.5)
BMI (kg m–2)
36.6 ± 5.3
36.3 ± 5.4
−0.2 (−0.6 to 0.2)
Ghrelin AUC (pmol l−1 h−1)
2160 ± 980
2670 ± 930
Glucose AUC (mmol l−1 h−1)
160 ± 42
129 ± 13
−31* (−48 to 15)
Insulin AUC (pmol l−1 h−1)
7030 ± 3350
5870 ± 3020
−1370* (−1950 to −790)
Fasting ghrelin (pmol l−1)
120 ± 53
147 ± 56
28* (10 to 45)
Fasting glucose (mmol l−1)
8.2 ± 2.1
6.9 ± 0.9
−1.3 (−2.1 to −0.1)
6.7 ± 1.0
6.4 ± 0.6
−0.4* (−0.8 to −0.05)
3.2 ± 1.7
3.0 ± 2.2
−0.23 (−0.68 to 0.22)
0.33 ± 0.03
0.34 ± 0.04
0.014* (0.006 to 0.023)
Fat mass by impedance (kg)
40.1 ± 12.2
40.2 ± 10.9
0.1 (−2.0 to 2.2)
% fat by impedance
38.1 ± 10.1
38.5 ± 8.8
0.5 (−1.7 to 2.6)
Adiponectin (mg l−1)
6.87 ± 4.01
7.09 ± 4.62
0.23 (−0.56 to 1.02)
Leptin (µg l−1)
31.4 ± 18.8
30.6 ± 17.7
−0.8 (−3.2 to 1.6))
Total cholesterol (mmol l−1)
5.0 ± 1.1
4.8 ± 1.1
−0.2* (−0.36 to −0.04)
Triglycerides (mmol l−1)
2.28 ± 1.35
2.13 ± 1.14
−0.16 (−0.50 to 0.19)
HDL (mmol l−1)
1.14 ± 0.36
1.15 ± 0.27
0.01 (−0.07 to 0.10)
LDL (mmol l−1)
2.7 ± 0.7
1.6 ± 0.8
−0.1 (−0.24 to 0.04)
All 18 subjects had measurable plasma metformin concentrations immediately prior to the study dose. The predose trough metformin concentration of two patients was <60% of the calculated postdose trough concentration. These two patients were therefore defined as having poor compliance, but their results were still included in the analysis. If these two patients are excluded from the analysis the results do not change and remain significant.
The AUC (07.30–16.30 h) of plasma ghrelin concentrations increased significantly after metformin treatment (Figure 1a) by a mean of 24% (P= 0.003). Fasting ghrelin increased by a mean of 23% (P= 0.007). At the same time points, and in overall (AUC) effect, there were no measurable changes in hunger or satiety (Figure 1c,d). The 95% confidence intervals (CIs) for change in AUC with metformin treatment were 0.91, 1.03 for hunger and 0.96, 1.12 for satiety, confirming the absence of any significant signal.
As expected, plasma concentrations of glucose and insulin were significantly reduced during metformin treatment (Figure 2). The mean reduction in the AUCs of glucose, 19% (P < 0.001), and insulin (Figure 1b), 19% (P= 0.001), were similar to the magnitude of increase in the AUC of ghrelin (Figure 1a), 24% (P= 0.003). The reductions in insulin and glucose were greater postprandially than fasting. The subjects with lower pretreatment fasting glucose concentrations (<8 mmol l−1) had a greater reduction in plasma insulin (29%) than in glucose (12%), whereas the subjects with higher fasting glucose concentrations (>8 mmol l−1) had a lesser reduction in plasma insulin (6%) than in glucose (30%). Despite the overall increase in plasma ghrelin and the decreases in insulin and glucose, there were no significant correlations between the increase in plasma ghrelin and the decreases in insulin (r2= 0.02, P= 0.60) and glucose (r2 < 0.001, P= 0.99) within patients.
The 95% CIs for change in adiponectin and leptin with metformin treatment were −0.56, 1.02 mg l−1 and −3.2, 1.6 µg l−1, respectively, indicating no clear effect. There was no significant change in measures of body morphology between the two study days (Table 1). Waist and hip circumferences and body fat estimates remained steady and weights were within day-to-day variability for most subjects. Fasting lipids and liver function tests did not change significantly between the two study days.
The mean (range) AUC of plasma metformin (0–8 h) was 9.9 mg l−1 h−1 (range 5.8–16.8). The mean (range) maximum plasma concentration (Cmax) of metformin was 1.7 mg l−1 (1.0–2.8), the mean apparent oral clearance (CL/F) was 1490 ml min−1 (780–2460) and the mean plasma half-life (t1/2) was 3.9 h (2.9–7.8). Plasma metformin AUC (the primary measure of drug exposure) did not correlate with the changes in AUC of ghrelin (r2= 0.04, P= 0.40), insulin (r2= 0.08, P= 0.24) or glucose (r2= 0.04, P= 0.89). This was also the case when the analysis was extended to other aspects of the data (e.g. metformin Cmax, changes in fasting or peak concentrations of glucose or insulin, or responses to the standard meals).
Ghrelin AUC correlated weakly but significantly with male gender, fat mass, fat%, leptin and insulin AUC (r2= 0.11, 0.18, 0.15, 0.14 and 0.12; P= 0.02, 0.01, 0.02, 0.02 and 0.02, respectively) on univariate analysis, and did not correlate with other parameters including glucose AUC, HbA1c, BMI, waist circumference or adiponectin. On stepwise linear regression analysis, ghrelin AUC remained significantly negatively correlated with fat mass and insulin AUC, but relationships with gender, leptin and fat% were no longer significant. Fat mass and insulin AUC explained approximately 40% of the overall interindividual variability of ghrelin AUC observed (r2= 0.40, P < 0.001). Similar results were seen when the analysis was confined to fasting values.
There was a small but significant reduction in weight of 1.6 kg (P= 0.01) with metformin treatment, and weight is known to correlate inversely with plasma ghrelin concentration. However, the magnitude of change in ghrelin was much larger than one would expect to see with this degree of weight change . Of note, there was no trend towards decrease in either leptin, waist circumference or fat mass measured by bioelectric impedance, over the study interval.
Treatment with metformin for 6 weeks resulted in a significant increase in plasma ghrelin in the group of 18 subjects with Type 2 diabetes. In terms of individual response, a clear increase in ghrelin was seen in 14 of the 18 subjects, no change in two and a decrease in two. The increase in ghrelin suggests that the effect of metformin on appetite is unlikely to be mediated by ghrelin.
The observed overall increase in ghrelin concentrations was accompanied by a decrease in insulin concentrations (Figure 1). Although the mechanism of this is unknown, it has been observed that insulin decreases and ghrelin increases in response to treatment with growth hormone . Similarly, insulin is known to inhibit growth hormone signalling via the growth hormone receptor, whereas ghrelin is the endogenous agonist of the growth hormone secretagogue receptor [10, 33]. There is thus considerable cross-talk between the insulin and growth hormone signalling pathways.
Whereas the results of this study are consistent with those of Schofl et al. that were used in the power calculations of this study  and another observational study , they contrast with the findings of two other recently published studies [25, 34]. English et al. conducted an intergroup comparison of 10 metformin and 11 nonmetformin-treated subjects, measuring plasma ghrelin after a standard meal. They showed no change in preprandial ghrelin concentration and prolonged postprandial suppression of ghrelin in the metformin treated group compared with the nonmetformin-treated group. Satiety (fullness) did not change in their study. Kusaka et al. conducted a within-group comparison of ghrelin response to a glucose tolerance test, and observed a decrease in ghrelin when treated with metformin. However, in their study the insulin concentrations in the metformin-treated subjects were increased, whereas one would normally expect insulin to decrease in metformin-treated patients. Strengths of our study include the measurement of plasma metformin, the within-group comparison, the within-patient consistency across two meals and the large number of sampling points per patient. Further studies are needed to resolve this question, and we suggest that these include careful measurement of IgF1 and growth hormone in addition to ghrelin, insulin, glucose and the adipokines.
Although ghrelin increased with metformin treatment, no change was observed in either hunger or satiety in response to two standard meals. Furthermore, the shape of the ghrelin curve was unlike those observed for hunger and satiety. These observations suggest ghrelin may not be as important in meal initiation as has been hypothesized . Alternatively, the lack of change in appetite and satiety may reflect lack of power or sensitivity of our measures of appetite and satiety. However, the consistency of the profiles between study days argues against this.
The lack of linearity in the concentration–effect relationships between metformin concentrations and changes in glucose, insulin or ghrelin suggests that individuals may differ in their determinants of response. A within-individual linear dose–effect relationship between metformin and glucose reduction has been demonstrated previously with a group maximum response at a dose of 2 g day−1. In our study there was no between-individual correlation between plasma metformin concentrations and effects, which is consistent with animal data . Differences in compliance are likely to occur no matter how robust the system, but the measured concentrations suggest this was a small effect in this study. Other possibilities are that individuals differ in their response downstream from the site of action, or in their delivery of metformin to hepatocyte mitochondria (the principle site of action). Variability in the hepatocellular uptake of metformin has been demonstrated in animal models , and genetic polymorphisms in the transporters responsible for this process have been demonstrated in humans .
The adipokines leptin and adiponectin have been the focus of intense research in recent years. The plasma concentration of leptin is positively correlated with fat cell mass, and adiponectin is inversely correlated with abdominal fat mass. Studies suggest that treatment with thiazolidinediones substantially increases plasma adiponectin [19, 40] and results in a small decrease or no change in plasma leptin [41, 42]. No change was seen in either leptin or adiponectin with short-term metformin treatment in this study, consistent with minimal direct effects of metformin on adipose tissues. This highlights differences in effect between metformin and the thiazolidinediones.
In summary, 6 weeks of metformin treatment of Type 2 diabetes was associated with a significant increase in plasma ghrelin concentrations along with decreases in insulin and glucose. There were no changes in hunger, satiety, or plasma concentrations of adiponectin or leptin. There was no correlation between metformin plasma concentrations and changes in glucose, insulin or ghrelin. More studies are needed to explain the effect of metformin on ghrelin and to determine why some individuals have a greater response to metformin than others.
This study was funded by a grant from the Diabetes and Heart Research Trust of Canterbury, New Zealand and supported by the Diabetes and Obstetric Medicine Research Trust, Christchurch, New Zealand. The authors thank the following: all patients; Marilyn Cullens, Caroline Adamson and the Christchurch Diabetes Centre; Kerry Watkins and the Endocrine Special Tests Centre; Sarah Raudsepp and the Cardioendocrine Research Group of the Christchurch School of Medicine; Jane Ellis, John Kinley and Endolab; Pete Elder and Canterbury Health Laboratories; and Chris Frampton for statistical advice.