Introduction
Currently, there are only few weight-loss drugs with a favourable side-effect profile
available [1]. Drugs like rimonabant have been promising in the recent past. However, severe side
effects lead to their witdrawal from the market [2]. Presently, GLP-1 analogues such as Liraglutide are being promoted as a new strategy
to loose weight without major side effects however at high costs [3].
Metformin is a well established anti-diabetic drug, which has been used from the late
1950 s on and has since made an unrattled career as an antidiabetic drug which is
easy to handle and still first line therapy in type 2 diabetes [4]
[5]. In the last decade it underwent a second career as efficacious treatment of polycystic
ovary syndrome (PCOS). Metformin reduces insulin resistance which is the underlying
cause of both type 2 diabetes and PCOS [6]
[7]
[25]
[26]. Metformin has been observed to cause weight loss in type 2 diabetes patients [8]
[9]. Obesity in non-diabetic patients is also linked to insulin resistance [10]
[11]
[12]. Improving insulin sensitivity may therefore account for weight reduction under
metformin therapy, although the exact underlying pathomechanisms remain to be elucidated
[9]
[13]
[14].
Only one randomized controlled trial of high quality (Diabetes Prevention Program)
has been published that shows metformin to significantly reduce weight in non-diabetic
patients [15], this effect endured up to 8 years, if patients adhered to the medication [16].
However, results are still inconclusive to date. In 2 further placebo-controlled studies,
no significant weight-loss of metformin was found [17]
[18].
Even in very recently published reviews on anti-obesity drugs metformin is not mentioned
[19]
[19]
[20]. In another review metformin is listed as a weight reducing drug, but the scarcity
of these studies is underlined [21]. In adults metformin has merely been looked upon as an anti-diabetic drug and significant
weight loss under metformin has been described more as a welcome side effect in diabetic
and prediabetic patients [22]
[23]
[24]. Interestingly however, for special indications like obese psychiatric patients
[25]
[26] or obese children and adolescents [27]
[28] as well as for women suffering from PCOS [29]
[30]
[31] weight reduction by metformin has been well documented.
Moreover, while the efficacy of metformin to reduce weight in a standardized setting
has been tested in few randomized trials, data on effectiveness in the everyday endocrine
outpatient setting are still lacking. The aim of this study was to test the effectiveness
of metformin to reduce weight in obese, non-diabetic patients in a naturalistic outpatient
setting.
Subjects
In our Practice for Endocrinology in Regensburg, Germany, an outpatient setting for
endocrine patients, we screened overweight and obese patients desiring to loose weight
for insulin resistance during a recruitment period of 18 months. Patients were offered
metformin therapy when their body mass index was ≥ 27 kg/m2. Insulin sensitivity was evaluated by calculating HOMA- and Matsuda-index after conducting
an oral glucose tolerance test (OGTT). We recommended all patients to reduce carbohydrate
consumption in the evening, but none of the patients underwent a concomitant diet
program. We chose that study design because a diet program would confound the results
of the study as a consequence of varying compliance and adherence to the diet. Furthermore,
we did not include patients who were planning to change their lifestyle during the
observation period, especially regarding physical activity.
Patients suffering from overt diabetes or impaired glucose tolerance or already taking
anti-diabetic drugs (including metformin) were excluded. Patients under steroid or
antipsychotic medication were also excluded. We further excluded patients suffering
from depression or any form of drug addiction including alcoholism and severe smoking
(more than one package per day). All patients were euthyroid. Pregnant and nursing
women were excluded. At baseline, weight, height, blood pressure, HbA1c, liver enzymes,
cholesterol and triglyzerides were measured. Patients were then treated with metformin.
The dosage was slowly uptitrated starting with 500 mg per day during the first week,
dosage was then weekly increased by 500 mg daily up to the final dose. Patients with
a BMI <30 kg/m2 received 1 500 mg final dose per day, patients with a BMI ≥30 kg/m2 but <35 kg/m2 received 2 000 mg and patients with a BMI ≥35 kg/m2 2 500 mg. 5 patients did not tolerate the projected dose of metformin due to side
effects and were excluded from the study. These 5 patients are not part of the 154
metformin-treated patients analyzed for effectiveness. However, these 5 patients were
included in the analyses of adverse events. Therefore, 159 metformin-treated patients
were analyzed for adverse events.
Patients, who chose not to take a medication as a mean to reduce weight served as
a control group. All patients were followed-up for 6 months. All patients agreed to
refrain from diet programs as well from a change of their habits in physical excercise
for the time of the study and were regularly questioned on that issue.
Methods
All patients with a BMI of ≥27 kg/m2 were offered to participate. Prior to inclusion, a 75 g oral glucose tolerance test
was performed in all subjects. Blood glucose as well as plasma insulin levels were
measured at baseline, 1 h, and 2 h after glucose ingestion. Matsuda-Index and HOMA-Index
was calculated from measured insulin and blood glucose values [32]. Patients were included in the study when blood glucose levels in the performed
OGTT were within the normal range irrespective of the degree of insulin sensitivity.
Insulin resistance was established if HOMA index was >2.5 or if Matsuda index was
<6. Insulin was measured by a chemoluminescence assay (Roche, Mannheim, Germany).
Blood glucose was measured by Hemocue 201 RT (glucose-dehydrogenase method, HemoCue,
Großostheim, Germany).
We analyzed changes in weight after 6 months in metformin treated patients and untreated
controls. We additionally calculated the percentage of patients losing at least 5%
and 10% of their baseline weight, respectively. We then divided baseline BMI, age,
C-peptide levels, HOMA-index, and Matsuda-index in tertiles and analyzed weight loss
according to the respective tertiles. Additionally we analyzed sex-specific weight
loss.
Statistics
Significance of weight loss after 6 months of treatment as compared to initial body
weight was calculated by using the independent t-test. ANOVA was used to compare the
weight loss between subgroups. Mann-Whithney-U-Test was used to compare weight loss
between males and females, as well as to compare weight changes in treated and untreated
patients.
All calculations were performed with WINSTAT for Excel.
Results
A total of 199 consecutive patients were included in the study (179 women, 20 men).
The average age was 39.2 years, the medium BMI was 35.3 kg/m2 ±6.0. A total of 154 patients (138 women, 16 men) were included in the metformin
group and 45 patients (41 women, 4 men) were included in the control group. [Table 1] displays the clinical baseline characteristics of both groups. There were no differences
in age, BMI, blood pressure, lipids and HbA1c. Metformin treatment was in general
well tolerated by most of the patients.
Table 1 Baseline clinical characteristics, Means and standard deviation of metabolic parameters
are shown.
|
Metformin (n=154)
|
Control (n=45)
|
p-value
|
|
age [years]
|
37.8±12.9
|
40.3±11.4
|
0.09
|
|
BMI [kg/m2]
|
35.4±6.5
|
34.9±4.0
|
0.92
|
|
hypertension [%]
|
20.4%
|
22.2%
|
0.42
|
|
systol BP [mmHg]
|
122.8±15.6
|
123.5±12
|
0.89
|
|
diast BP [mmHg]
|
81.0±10.5
|
78.4±9.6
|
0.66
|
|
cholesterol [mg/dl]
|
208±35
|
217±39.5
|
0.11
|
|
HDL-C [mg/dl]
|
60±13.0
|
61±25.1
|
0.59
|
|
Triglycerides [mg/dl]
|
161±35.3
|
154±32.1
|
0.08
|
|
HbA1c [%]
|
5.6±0.6
|
5.7±0.4
|
0.9
|
The average dose of metformin was 2 230 mg daily in the metformin-treated group. 24
(15%) of patients under metformin complained about gastrointestinal side effects like
diarrhea, bloating and stomach pain. In these patients we increased the dosage less
rapidely, so that they reached the final dose later but all within 2 month.
We observed a weight loss of 10% or more in 16.2% and a weight loss of at least 5%
in 47.4% of the 154 metformin-treated patients.
In the group metformin-treated, the overall average weight loss during 6 month of
treatment was 5.8±7.0 kg (5.6±6.5%) (p<0.001). The range was from the loss of 35 kg
in one patient to a weight gain of 13 kg in another patient, the median was a weight
reduction of 5 kg. 32 patients out of the total of 154 treated patients (20.8%) did
not loose weight, 9 of them even gained weight under metformin. The weight loss in
the metformin treated group was significant as compared to the control group (p<0.0001).
In the control group an average weight gain of 0.8±3.5 kg (0.8±3.8%) was observed.
There was no significant difference in weight loss under metformin treatment between
men and women (6.6 kg, 5.6% and 5.7 kg, 5.6%, respectively, p=0.23).
Patients were divided in tertiles according to the measured parameters of insulin
resistance (Matsuda-Index, HOMA-Index and C-Peptide levels), age and BMI. [Table 2] and [Fig. 1] show the results. Both, absolute and relative weight loss increased with higher
degrees of insulin resistance as measured by Matsuda-index and HOMA-index. However,
weight loss was not associated with different levels of C-peptide. There was significantly
more absolute weight loss with increasing BMI, but no significant association of relative
weight loss with BMI. Moreover, age was unrelated to the degree of weight loss.
Fig. 1 Weight changes in body weight in relation to base line weight with regard to sex
and insulin sensitivity after 6 month of metformin treatment. Insulin resistant patients
(HOMA >2.5, n=111) lost significantly(*) more weight as compared to patients with
normal HOMA-index (n=43) (p=0.001).
Table 2 Weight loss in kg and % of baseline weight by 3 tertiles of Matsuda-Index, C-peptide
HOMA-index,age and BMI. Data are presented as mean±standard deviation.
|
Tertile I
|
Tertile II
|
Tertile III
|
p-value
|
|
Matsuda-Index
|
>3.5
|
2.1–3.4
|
<2.1
|
|
|
kg
|
5.0±8.2
|
6.1±7.4
|
6.4±4.8
|
p=0.005
|
|
%
|
5.0±7.8
|
6.0±6.9
|
6.0±4.2
|
p=0.0002
|
|
C-peptide
|
<3.0
|
3.0–4,4
|
>4.5
|
|
|
kg
|
6.0±6.6
|
6.2±7.8
|
5.9±6.5
|
p=0.12
|
|
%
|
5.9±5.8
|
5.5±7.3
|
5.6±6.3
|
p=0.57
|
|
HOMA
|
<3,0
|
3,0–4,5
|
>4.5
|
|
|
kg
|
5.1±8.5
|
6.0±6.7
|
6.4±5.4
|
p=0.004
|
|
%
|
5.1±8.1
|
5.9±6.0
|
5.9±5.0
|
p=0.001
|
|
Age
|
<31.5
|
31.6–42
|
43–72
|
|
|
kg
|
5.6±6.6
|
6.1±6.9
|
5.7±7.5
|
p=0.18
|
|
%
|
5.4±7.0
|
6.2±5.7
|
5.2±7.1
|
p=0.09
|
|
BMI
|
27–32.6
|
>32,6–37.4
|
37.5–50
|
|
|
kg
|
3.4±4.9
|
5.6±6.6
|
8.5±8.2
|
P=0.02
|
|
%
|
4.1±5.9
|
5.7±6.6
|
6.9±6.8
|
p=0.07
|
In the subgroup of patients with normal HOMA-index (< 2.5, n=43) the average weight
loss was 4.9±8.9 kg. In the subgroup of patients with normal HOMA- and normal Matsuda-index
(>6; n=14) the average weight loss was 5.3±8.9 kg.
Discussion
In our study we found an average weight loss of 5.8 kg (5.6%) under treatment with
metformin for 6 months in overweight and obese, mostly insulin-resistant patients.
To our knowledge, this is the first study that assesses the effectiveness of metformin
for weight-loss in a natural endocrine outpatient setting with particular regard to
the degree of insulin resistance.
Although metformin is widely recognized by endocrinologists and diabetologists as
a weight reducing agent in clinical practice, there is little evidence in the literature
up to date supporting this empiric knowledge.
Even in recently published reviews on anti-obesity drugs metformin is not mentioned
[19]
[19]
[20]. One review analysing studies describing weight changes under metformin came to
the conclusion that there is not sufficient evidence to recommend metformin as a treatment
of overweight or grade 1 obesity [33]. Importantly however, in that review, the duration of treatment in the majority
of the examined studies was too short to see effects on weight (15 days to 3 months
in 6 out of 9 studies), most of the studies were not designed to investigate weight
reduction and the BMI measurements of study participants were limited to overweight
(BMI 25–29.9 kg/m2) or grade I obesity (BMI 30–34.9 kg/m2), therefore severe insulin resistance in these patients was unlikely.
One randomized controlled study of high quality and 2 randomized controlled studies
of fair quality investigated the efficacy of metformin as a weight reducing agent:
in subjects at risk for the development of diabetes (Diabetes Prevention Program,
base line weight 94.4 kg), in subjects with an elevated waist to hip ratio (base line
BMI 33.4) and in women suffering from PCOS (base line weight 97.0 kg) [15]
[16]
[17]
[18]. A limited or insignificant weight reducing effect of metformin therapy was reported
in these studies (Diabetes Prevention Program 2,7 kg, Fontbonne et al. and Gamberini
et al. no significant effect). The weight loss induced by metformin treatment in our
study of non-diabetic patients exceeds the amount of weight reduction reported by
the above mentioned studies.
One reason may be that in the study of Gambineri at al., who did not observe a significant
weight loss as compared to placebo, patients were already on a hypocaloric diet before
metformin treatment was started. It is possible that a hypocaloric diet diminishes
insulin resistance and thereby the effect of metformin [34]. Another reason maybe dosage of metformin. As we increased the dose of metformin
depending on the weight of the patient up to 2 500 mg per day, other authors used
lower doses. In the BigPro1 trial only 850 mg of metformin had been administered daily
to the study participants, possibly explaining the insignificant weight reduction
in this study [17].
Moreover, none of these studies was conducted under natural outpatient circumstances.
In a smaller open-label study in an outpatient-setting a weight loss of 6 kg was described
in severely obese patients with mostly elevated fasting insulin levels (base line
weight 117.3 kg), an effect that was similar to our observation [35].
The controversial results in these studies may be largely caused by different degrees
of insulin resistance in the examined study populations. As we could show in our study
that insulin resistant patients lost more weight under metformin as compared to insulin
sensitive patients, a turkish study described an average weight loss of 9 kg examining
patients with pronounced insulin resistance (baseline weight 96 kg; average HOMA-index
4,99) [36]. Unfortunately this study lacked an untreated control group, but compared metformin
to orlistat and sibutramine, describing metformin as equally effective as orlistat.
Another study observed in obese type 2 diabetes patients – insulin resistant by definition
– a weight loss of 8 kg after metformin treatment over 24 weeks [8].
Mogul et al. reported in 25 of 26 hyperinsulinemic, severly obese women (refractory
to diet) a weight loss of at least 5% within the first 6 month of metformin therapy
[37]. The hyperinsulinemia of these patients indicates the presence of insulin resistance
and it is possible that this population may be more sensitive to effects of metformin
on body weight.
Altogether, the discussed data suggest that the efficacy of metformin to reduce weight
depends – at least partially – on the degree of insulin resistance.
It has been shown that insulin-receptor binding, the tyrosine kinase activity of the
insulin receptor and glucose transport are altered in the same way in type 2 diabetes
patients and comparably obese individuals without diabetes, which supports the efficacy
of metformin in both conditions [38].
Today, the strong link between weight gain and insulin resistance is undisputed. There
is compelling evidence that insulin resistance is a major contributor to abdominal
obesity [10]
[11]
[12]. Yet, it hasn’t been clarified if insulin resistance leads to or is rather a consequence
of obesity [39].
Improving insulin resistance mediates weight loss by the following mechanisms:
First of all, improved insulin sensitivity reduces the frequency of postprandial hypoglycaemia
which is caused by sluggish postprandial insulin secretion with a delayed insulin
peak [40]. Less postprandial hypoglycaemic events result in less carbohydrate cravings and
as a consequence in less compensatory food intake, especially of carbohydrates [41].
Moreover, metformin not only improves blood glucose control by insulin-stimulated
glucose disposal in the skeletal muscle, but also decreases hepatic glucose output,
inhibits gluconeogenesis and decreases intestinal glucose absorption from the gastrointestinal
tract. [9], reviewed by [42]. These mechanisms provide less glucose for energy storage in the adipose tissue.
Notably, it has been shown that, in contrast to weight loss induced by hypocaloric
diet, metformin diminishes adipose tissue but not lean body mass [9].
Furthermore, abdominal fat tissue tends to retain its sensitivity to insulin in the
face of hepatic and skeletal muscle resistance, which leads typically to a predominant
increment of abdominal fat tissue in insulin resistant obese patients [43]
[44]. Restoring peripheral insulin sensitivity by metformin reduces pancreatic hypersecretion
of insulin. As a consequence the insulin receptor of abdominal fat tissue is exposed
to lower levels of insulin, which may result in a lower energy uptake by abdominal
fat and thereby reduce abdominal fat mass.
Interestingly however, metformin did also cause significant weight loss in obese patients
without insulin resistance (normal HOMA- or Matsuda-index). This leads to the postulation
that other mechanisms than reducing insulin resistance by metformin contribute to
weight loss in metformin treated patients. In the treatment of PCOS patients metformin
has also be shown to be effective on oligomenorrhea and hyperandrogenemia in some
patients despite the lack of insulin resistance [45].
Metformin leads to a decrease in appetite [9]
[41] and certain authors discuss that metformin contains a primary anorectic factor [46]. One other reason may be a decrease of leptin levels, which has been observed under
metformin treatment in adipose tissue in vivo as well as in serum levels in vivo suggesting
an underlying improvement of leptin resistance [47]
[48]. Furthermore GLP-1 levels seem to rise significantly under metformin and may thus
promote weight loss [49]
[50].
In our study the amount of weight loss correlated with body weight at base line. It
is evident that patients with severe obesity have a greater capacity to loose weight
than patients with relatively mild obesity. Severely obese patients lost on average
more weight as compared to patients with slighter obesity. However there was no significant
difference in the percentage of weight reduction as compared to baseline under metformin
in the higher BMI tertiles as compared to the lowest tertile. We also observed that
the effectiveness of metformin as a weight reducing agent was independent of age or
sex.
It is a limitation of our study that the control group was not randomized. Possibly
the motivation to loose weight was different in this group. Also we only analyzed
a follow-up period of 6 months. We do not know with certainty if these effects will
endure. However the follow-up of the Diabetes Prevention Program points to a sustained
effect of metformin, which is in line with our experience (data not shown) [10].
A strength of our study is that it shows the effect of metformin in a real life situation,
and not in a clinical trial situation, including effects of compliance and potential
interactions with comedications, among others. Additionally, we carefully assessed
the association of insulin resistance with weight loss.
Taken together, our data suggest that if metformin is administered to obese patients
in high enough doses it is a beneficial and cost effective drug to reduce weight.
The effectiveness of metformin as a weight reducing agent is not restricted to insulin
resistant patients although a greater weight loss can be expected if patients exhibit
biochemical signs of insulin resistance prior to metformin treatment.
