Introduction

Obesity is an energy metabolism disorder, progressing with excessive fat accumulation and leading to serious physical and psychological problems [1]. The number of overweight and obese people has increased substantially [2]. Many countries could face increasing pressure to take on more medical costs in obesity-related diseases [3]. Metformin was recommended as a first-line agent in the treatment of type 2 diabetes over 5 decades ago. Because of its multiple beneficial “side effects” [4], many researchers found metformin could be administered for the treatment of polycystic ovary syndrome [5, 6], nonalcoholic fatty liver disease [7], obesity [8], and some types of cancer in obese non-diabetic adolescents [9]. However, inconsistent results with regards to metformin in the treatment of overweight or obese adults has been reported. To evaluate the effectiveness of metformin in overweight and obese people without obesity related diseases, we carried out this meta-analysis to compare the outcomes of metformin with a placebo in people with simple obesity.

Methods

Search strategy

Our available relevant published reports came from pubmed, EMBASE, Cochrane, and Science Citation Index Expanded databases. We searched the references of all the papers found as supplementary research. Retrieval time is from the time each database was built to March 2018. We excluded papers not published in English. Searches for terms “metformin” or “biguanides” or “diguanide” and “obese” or “weight” or “body fat” or “diet” or “overweight” or “fitness” were performed.

Study selection

Two researchers reviewed published reports independently and evaluated relevant literature to decide whether the study should be included. If there were differences in opinion, a third investigator was consulted.

The main inclusion criteria are as follows: randomized controlled trials; patients diagnosed as being overweight or obese; patients were randomly assigned to receive metformin or control. Exclusion criteria were predefined as a diagnosis of obesity related diseases such as DM or PCOS.

Data extraction

Two researchers independently reviewed and extracted the patient and study characteristics (primary outcomes) such as body weight, BMI, waist circumference, and fasting blood glucose from the relevant literature using standardized forms. A third reviewer was consulted to settle discrepancy. We attempted to obtain missing data by contacting the authors of relevant literature. According to the quality standard of the Jadad scale, the Jadad score was used to evaluate the quality of the selected literature.

Statistical analysis

In each study, change in weight, change in waist circumference, and fasting blood glucose were extracted as mean ± standard deviation (SD). Some studies provided 95% confidence interval (CI) of means, and 95% CIs was converted to SD values. Our study used Review Manager 5.3 to analyze data. A p < 0.05 was considered statistically significant. Thirteen studies were used to estimate the heterogeneity. According to the heterogeneity standard of I² statistic, we used the I² statistic to assess heterogeneity between studies. For studies with I² ≥ 50%, heterogeneity was considered significant and we performed random effects estimates to calculate [10]. For studies with I² ≤ 50%, we used the fixed effects model. The weighted mean difference for continuous variables were used to explain outcomes with the 95% CI.

Results

Study selection

In total, 825 studies were identified using the above-mentioned search strategy, 13 of which were eligible for inclusion in our meta-analysis. 780 studies were ineligible for inclusion after screening on title and abstract, 32 were excluded because some contained research protocols only, some review articles failed our inclusion criteria, some were not randomized, some had inaccessible outcomes, and some were meta-analysis articles (Fig. 1). Table 1 lists the study characteristics.

Fig. 1
figure 1

Flow chart of study selection process

Full size image
Table 1 Characteristics of included studies
Full size table

Effect of metformin on body weight

Three studies with data available for change in weight were included in the analysis. A total of 327 patients were included, of whom 158 patients were in the metformin group and 169 patients were in the control group. The decline in body weight was more highly significant in the metformin group than in the control group (p = 0.02, I² = 88%, Fig.2); heterogeneity was considered significant and when compared with the placebo group, weighted mean difference was 2.33 kg (95% CI 0.31, 4.35) kg higher with metformin when we used the random-effects model.

Fig. 2
figure 2

Forest Plot showing the reduction of body weight (kg) comparing metformin and control, The decline in body weight was more highly significant in metformin group than in control group

Full size image

Effect of metformin on BMI

Six studies with data available for change in BMI were included in the analysis. A total of 487 patients were included, of whom 246 patients were in the metformin group and 241 patients were in the control group. The reduction of body mass index was more highly significant in the metformin group than in the control group (p < 0.00001, I² = 94%; Fig. 3); heterogeneity was considered significant and compared with the placebo group, weighted mean difference was 0.57 kg/m² (95% CI 0.35, 0.79) higher with the metformin group when we used the random-effects model.

Fig. 3
figure 3

Forest Plot showing the reduction of BMI (kg/m²) comparing metformin and control, The reduction of body mass index was higher significantly in metformin group than in control group

Full size image

Effect of metformin on waist circumference

Three studies with data available for the reduction of waist circumference were included in the analysis. A total of 417 patients were included, of whom 203 patients were in the metformin group and 214 patients were in the control group. There was no significant difference in waist circumference between the metformin group and the control group (p = 0.05, Fig. 4).

Fig. 4
figure 4

Forest Plot showing the change in waist circumference in the metformin and control group. There was no significant difference in waist circumference between metformin group and control group

Full size image

Effect of metformin on fasting blood glucose

Seven studies with data available for fasting blood glucose were included in our analysis. A total of 426 patients were included, of whom 206 patients were in the metformin group and 216 patients were in the control group. There was no significant difference in fasting blood glucose at baseline between the metformin group and the control group (p = 0.05) with no significant heterogeneity (Heterogeneity: I² = 25%; Fig. 5a). The fasting blood glucose was significantly lower in the metformin group than in the control group (p < 0.00001) after the treatment with no significant heterogeneity (Heterogeneity: I² = 0%; Fig. 5b). However, no hypoglycemia had been noted in the metformin group.

Fig. 5
figure 5

a Forest Plot showing the fasting blood glucose at baseline comparing the metformin and control group. There was no significant difference in fasting blood glucose at baseline between the metformin group and control group. b Forest Plot showing the fasting blood glucose after treatment comparing the metformin and control group. The fasting blood glucose was significantly lower in the metformin group than in the control group

Full size image

Discussion

Health issues associated with obesity have become serious public health problems [1]. Obesity treatment includes diet control, exercise and medication, and bariatric surgery. Metformin is mainly used to prevent and cure diabetes. Much evidence demonstrates that metformin may have other effects [4], including weight loss. We carried out a meta-analysis to compare metformin with a placebo on the outcomes of people with simple obesity.

Our results found that metformin significantly decreased BMI and weight. In other words, metformin could effectively improve clinical parameters such as BMI and weight related with obesity or obese adults who do not have obesity-related diseases. Our results were consistent with these two studies [11, 12] which showed that metformin could induce weight loss in obese nondiabetic populations. Metformin in therapy reduced BMI and body weight, and its mechanism may be as follows: Firstly, metformin may enhance leptin sensitivity and reduce leptin secretion [13]. Hyperleptinemia and leptin resistance are features of obesity [14]. Leptin is associated with fat mass, and regulates food intake and energy - consumption processes [15]. Hyperleptinemia is also harmful to the tissues [16]. Secondly, metformin can effectively control the level of blood glucose, improving insulin resistance [17]. The previous study showed that the development of obesity seems to be determined by insulin resistance [18]. Insulin resistance is also a feature of obesity [19, 20]. Research shows that there is a relationship between increasing insulin sensitivity and reducing liver volume in patients taking metformin [21]. Thirdly, metformin induces a significant decrease of total and LDL cholesterol levels [22]. It has been reported that the decrease of total and LDL cholesterol levels is related to weight loss [23].

In addition, this meta-analysis found no significant difference in waist circumference between the metformin and control group. We think the reason is that the amount of research we included and the number of participants involved in each study were small, and so our meta-analysis could not produce as accurate of results if we had a larger sample size. Although our results do not support that metformin reduced waist circumference, much research has shown that metformin could produce a significant reduction in waist circumference [24, 25]. Some studies have shown that metformin causes a significant redistribution of the visceral adipose tissue area [26, 27].

Our meta-analysis also showed that metformin could effectively control blood sugar in patients without DM. Antihyperglycemic activity of metformin is attributed to increased insulin sensitivity, glucose disposal, suppressed hepatic gluconeogenesis, and inhibited glucagon action [18]. However, as no hypoglycemia has been noted in the metformin group, it could be concluded that metformin does not increase insulin secretion. This could indicate that metformin did not induce hypoglycemia.

Although our meta-analysis produced promising results, there are some limitations. Firstly, because the number of studies and study participants in each trial was small, the tests for publication bias cannot be performed. Secondly, we did not conduct sensitivity analysis in our meta-analysis. The study participant and outcome of each article were different. Two of the studies [28, 29] have estimated metformin prevention of olanzapineinduced obesity weight gain. Nearly half the participants came from two of the studies [30, 31]. This weakens the proportional impact of the other articles. Finally, almost all studies were of short duration, so it could be that the benefits of metformin were higher in the beginning of the treatment [32]. Despite our limitations, it was still shown that metformin could produce weight loss. We suggest more well-designed, randomized, double-blind placebo-control clinical trials are needed to prove the results of this study.

In conclusion, metformin can reduce body weight and BMI, and metformin did not induce hypoglycemia in the treatment of patients with simple obesity,