Introduction

In 2010 it was estimated that the world prevalence of diabetes among adults (aged 20–79 years) was 6.4% (285 million adults).1 Perhaps even more concerning is that diabetes is projected to affect approximately 7.7% of the world’s population (439 million adults) by 2030.1 Type 1 diabetes mellitus (T1DM) is considered as an autoimmune disease resulting in absolute insulin deficiency. Consequently, patients who are diagnosed with T1DM require therapy with exogenous insulin. In contrast, type 2 diabetes mellitus is characterized by a decline in β-cell function and worsening insulin resistance.2 Therapy for type 2 diabetes mellitus focuses around improving glucose tolerance through diet, exercise and oral anti-diabetic medications. Of the individuals affected with diabetes, approximately 95% are diagnosed with type 2 diabetes mellitus.3

Despite the advances in modern medicine, DM continues to be the most common endocrine metabolic disorder and the disease is rapidly increasing worldwide affecting all parts of the world.4 Many of the available therapies (that is, biguanides, sulfonylureas, glinides and insulin) have serious adverse effects associated with their use making the search for a more effective and safer hypoglycemic agent an important area of investigation.4 It is currently estimated that up to one-third of patients with DM use some form of supplemental medicine.

Momordica charantia (bitter melon) is a popular fruit used as a supplementary agent to treat DM in the populations of Asia, South America, India and East Africa. The exact mechanism of M. charantia is unknown; however, the active components charatin, vicine and polypeptide p is thought to be structurally similar to human insulin.5 Thus, some of the proposed mechanism of M. charantia includes insulin-like effects, increased insulin secretion, tissue glucose uptake, liver muscle glycogen synthesis and decrease glucose absorption.

There is insufficient evidence in the literature to make a definitive conclusion about the effects of bitter melon on glucose control. Current trials evaluating bitter melon on glycemic control has led to conflicting results and hence a systematic review of the available evidence was conducted. The objective of this paper is to conduct a systematic review and meta-analysis to evaluate the use of bitter melon on glycemic outcomes in patients with DM.

Materials and methods

Study selection

We performed a systematic literature search of PubMed, EMBASE and Cochrane Central Register of Controlled Trials from database start through September 2014 without language restriction to identify randomized controlled trials which compared bitter melon to no treatment in patients with type 1 or type 2 DM. Search terms included combining MeSH and text keywords for bitter melon (M. charantia), type 1 and type 2 diabetes, glucose levels and hemoglobin A1c. Studies were included if they were randomized controlled trials that evaluated bitter melon versus no treatment in either type 1 or type 2 diabetes and report results on A1c or fasting plasma glucose (FPG). Study selection was conducted by two independent investigators with disagreements resolved by discussion or a third investigator.

Data extraction and quality assessment

The following data were extracted from the included studies: baseline characteristics, study duration, inclusion and exclusion criteria, bitter melon dose and specific product used, results on A1c and FPG. Data were extracted by two independent investigators with disagreements resolved by discussion or a third party. Validity assessment was performed by two investigators independently using the Cochrane Risk of Bias Tool.6 This checklist includes six validity questions covering the following domains: random sequence generation, allocation concealment, blinding, blinding of outcome assessment, incomplete data reporting and selective reporting. Each item was scored as a low, unclear or high risk of bias (Figure 2). Upon analyzing each domain in the Cochrane Risk of Bias tool, an overall risk of bias score of either low, unclear, or high per trial was derived. A low risk of bias score was assigned to a study if most of the information from that study was at low risk of bias. An unclear risk of bias score was assigned to a study if most of the information from that study was at low or an unclear risk of bias, and a high risk of bias score was assigned to a study if the proportion of information from that study at high risk of bias was sufficient to affect the interpretation of the results.6

Data synthesis and analysis

The mean changes in FPG and A1c from baseline were treated as continuous variables, and the weighted mean differences and accompanying 95% confidence intervals were pooled using a DerSimonian and Laird random-effects model.7 Changes from baseline in outcomes were extracted from trials; in instances where changes were not reported directly, they were calculated from end-of-study and baseline results. Outcome data were extracted as analyzed in that specific trial without any additional adjustment for potential losses to follow-up. As suggested by Follmann et al.,8 we assumed a correlation coefficient of 0.5 between initial and final values. The statistical analysis was performed by using StatsDirect statistical software, version 2.7.8 (Altrincham, Chesire, UK). A P-value <0.05 was considered statistically significant for all analyses. Statistical heterogeneity was planned to be assessed using the I2 statistic, but too few studies were available to assess.9 Publication bias was planned to be assessed using visual inspection of funnel plots and Egger’s weighted regression statistics, but too few studies were available to assess.10

Results

The search strategy yielded 50 nonduplicate citations for screening. (Figure 1) Six full-text articles were screened and three trials (n=187) met all inclusion criteria. Only two trials reported usable data for meta-analysis of A1c11, 12 (Table 1) and two trials reported meta-analyzable data for FPG11, 13 (Table 1). Patients were randomized to be treated with either bitter melon or control (dosing range 1–6 g per day) in various dosage forms (tablets or capsules) for a period of 4–12 weeks (Table 1). Although our search strategy allowed for the inclusion of patients with type 1 diabetes, all studies that met inclusion criteria evaluated patients with type 2 diabetes. The risk of bias assessment for individual validity components is presented in Figure 2.

Figure 1
figure 1

Preferred reporting items in systematic reviews and meta-analyses flow diagram depicting study selection, inclusion and exclusion of trials evaluating bitter melon on glycemic parameters.

Full size image
Table 1 Baseline characteristics of trials evaluating bitter melon use on glycemic parameters
Full size table
Figure 2
figure 2

Risk of bias assessment of trials evaluating bitter melon on glycemic parameters. Note: +, low risk of bias, ?, unclear risk of bias, −, high risk of bias.

Full size image

Quantitative data synthesis

Upon meta-analysis, there was no statistically significant association between bitter melon usage and A1c (weighted mean difference −0.13%, 95% confidence interval −0.41 to 0.16) or FPG (weighted mean difference 2.23 mg dl−1, 95% confidence interval −14.91 to 19.37 mg dl−1) when compared with control (Figure 3a and b, respectively). Very few studies were available for each analysis to allow for the assessment of heterogeneity or publication bias.

Figure 3
figure 3

Forest plot depicting result from meta-analysis of trials evaluating the effects of bitter melon on A1c and FPG versus placebo based on the year of publication. Note: The squares represent the pooled results of that study in addition to all studies preceding it. Errors bars represent 95% Cls and the diamond represents the overall pooled results. The solid vertical line extending upward from 0 is the null value.

Full size image

Discussion

Previous data has offered mixed conclusions about the benefits of bitter melon in patients with diabetes. Although it has been suggested that bitter melon promotes post-meal insulin secretion, thereby improving glycemic response, this has only been showed in a single-dose experiment without a control group.14 This mechanism of activity may require the presence of viable pancreatic β cells to secrete insulin in order for hypoglycemic activity to take place. It is possible that the patients evaluated in randomized controlled trials have limited β-cell function, decreasing their ability to respond to bitter melon. In this meta-analysis, pooling together the available evidence found no statistically significant effect for the use of bitter melon versus control on A1c or FPG.

The small number of available trials on this topic is a limitation of this analysis. The small sample of trials may be insufficient to detect statistically significant effects; because of this, it is unclear whether bitter melon truly exerts no effect or if the analysis is underpowered. The variety of doses and dosage forms of bitter melon contribute to the potential heterogeneity of the evidence. The duration of study is also limited, with the longest trial at 12 weeks and the others only 4 weeks long; however, the outcome A1c generally requires at least 8–12 weeks of therapy to detect changes.15, 16, 17, 18 It is likely that this current evidence could not adequately detect effects on A1c if this effect were to exist. The effects of bitter melon on A1c may require a longer duration of therapy before any potential benefits on A1c are seen. Owing to the limited report of adverse effects, we could not meta-analyze safety parameters related to bitter melon. However, one trial reported that the most common complaints were about gastrointestinal discomfort.11

Overall, the evidence regarding the use of bitter melon on glycemic outcomes in patients with DM is inconclusive. Additional evidence in a larger sample of patients evaluated over a longer duration of time is needed to determine whether bitter melon is truly ineffective in patients with DM.